Merge commit 'origin/master' into gallium-0.2

Conflicts:

	src/mesa/glapi/descrip.mms
	src/mesa/shader/grammar/descrip.mms

58 files changed, 854 insertions, 1967 deletions

diff --git a/bin/mklib b/bin/mklib
index a25d5bfaeb..d7b740f8e3 100755
--- a/bin/mklib
+++ b/bin/mklib
@@ -260,9 +260,31 @@ case $ARCH in
                 OPTS=${ALTOPTS}
             fi
             rm -f ${LIBNAME}
+
+	    # expand any .a objects into constituent .o files.
+	    NEWOBJECTS=""
+	    DELETIA=""
+	    for OBJ in ${OBJECTS} ; do
+		if [ `expr match $OBJ '.*\.a'` -gt 0 ] ; then
+		    # extract the .o files from this .a archive
+		    FILES=`ar t $OBJ`
+		    ar x $OBJ
+		    NEWOBJECTS="$NEWOBJECTS $FILES"
+		    # keep track of temporary .o files and delete them below
+		    DELETIA="$DELETIA $FILES"
+		else
+		    # ordinary .o file
+		    NEWOBJECTS="$NEWOBJECTS $OBJ"
+		fi
+	    done
+
             # make lib
-            ${LINK} ${OPTS} ${LIBNAME} ${OBJECTS}
+            ${LINK} ${OPTS} ${LIBNAME} ${NEWOBJECTS}
             ranlib ${LIBNAME}
+
+	    # remove temporary extracted .o files
+	    rm -f ${DELETIA}
+
             # finish up
             FINAL_LIBS=${LIBNAME}
         else
diff --git a/configs/darwin b/configs/darwin
index aa9efb66f5..42897ac033 100644
--- a/configs/darwin
+++ b/configs/darwin
@@ -12,10 +12,10 @@ CXX = gcc
 PIC_FLAGS = -fPIC
 DEFINES =  -D_DARWIN_C_SOURCE -D_POSIX_SOURCE -D_POSIX_C_SOURCE=199309L \
 	   -D_BSD_SOURCE -D_SVID_SOURCE -D_GNU_SOURCE -DPTHREADS \
-	   -DGLX_INDIRECT_RENDERING \
-	   -DGLX_ALIAS_UNSUPPORTED
+	   -DGLX_ALIAS_UNSUPPORTED -DGLX_INDIRECT_RENDERING
 
 # -DGLX_DIRECT_RENDERING - pulls in libdrm stuff in glx/x11
+# -DGLX_USE_APPLEGL      - supposed to be used with GLX_DIRECT_RENDERING to use AGL rather than DRM, but doesn't compile
 # -DIN_DRI_DRIVER
 
 ARCH_FLAGS += $(RC_CFLAGS)
@@ -47,7 +47,7 @@ GLW_LIB_DEPS = -L$(TOP)/$(LIB_DIR) -l$(GL_LIB) -L$(INSTALL_DIR)/$(LIB_DIR) -lX11
 APP_LIB_DEPS = -L$(TOP)/lib -l$(GLUT_LIB) -l$(GLU_LIB) -l$(GL_LIB) -L$(INSTALL_DIR)/$(LIB_DIR) -lX11 -lXmu -lXt -lXi -lm
 
 # omit glw lib for now:
-SRC_DIRS = glx/x11 mesa glu glw glut/glx
+SRC_DIRS = glx/x11 mesa glu glut/glx
 GLU_DIRS = sgi
 DRIVER_DIRS = osmesa
 #DRIVER_DIRS = dri
diff --git a/src/mesa/drivers/common/descrip.mms b/src/mesa/drivers/common/descrip.mms
index 99a2ae6c37..d5bbc69dfd 100644
--- a/src/mesa/drivers/common/descrip.mms
+++ b/src/mesa/drivers/common/descrip.mms
@@ -1,6 +1,6 @@
 # Makefile for core library for VMS
 # contributed by Jouk Jansen  joukj@hrem.nano.tudelft.nl
-# Last revision : 3 October 2007
+# Last revision : 29 September 2008
 
 .first
 	define gl [----.include.gl]
@@ -19,7 +19,8 @@ VPATH = RCS
 
 INCDIR = [----.include],[--.main],[--.glapi],[--.shader]
 LIBDIR = [----.lib]
-CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm
+CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)\
+	/float=ieee/ieee=denorm/warn=disable=(PTRMISMATCH)
 
 SOURCES = driverfuncs.c
 
diff --git a/src/mesa/drivers/dri/i915/i830_vtbl.c b/src/mesa/drivers/dri/i915/i830_vtbl.c
index 0ab27704d5..773a8b4dd0 100644
--- a/src/mesa/drivers/dri/i915/i830_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i830_vtbl.c
@@ -566,6 +566,13 @@ i830_destroy_context(struct intel_context *intel)
    GLuint i;
    struct i830_context *i830 = i830_context(&intel->ctx);
 
+   intel_region_release(&i830->state.draw_region);
+   intel_region_release(&i830->state.depth_region);
+   intel_region_release(&i830->meta.draw_region);
+   intel_region_release(&i830->meta.depth_region);
+   intel_region_release(&i830->initial.draw_region);
+   intel_region_release(&i830->initial.depth_region);
+
    for (i = 0; i < I830_TEX_UNITS; i++) {
       if (i830->state.tex_buffer[i] != NULL) {
 	 dri_bo_unreference(i830->state.tex_buffer[i]);
diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c
index efcac911aa..e0ddc7fd61 100644
--- a/src/mesa/drivers/dri/i915/i915_context.c
+++ b/src/mesa/drivers/dri/i915/i915_context.c
@@ -56,8 +56,6 @@ static const struct dri_extension i915_extensions[] = {
    {"GL_ARB_shadow", NULL},
    {"GL_ARB_texture_non_power_of_two", NULL},
    {"GL_EXT_shadow_funcs", NULL},
-   /* ARB extn won't work if not enabled */
-   {"GL_SGIX_depth_texture", NULL},
    {NULL, NULL}
 };
 
diff --git a/src/mesa/drivers/dri/i915/i915_context.h b/src/mesa/drivers/dri/i915/i915_context.h
index c6958dd8d4..a2376e50e1 100644
--- a/src/mesa/drivers/dri/i915/i915_context.h
+++ b/src/mesa/drivers/dri/i915/i915_context.h
@@ -125,6 +125,9 @@ struct i915_fragment_program
    GLboolean on_hardware;
    GLboolean error;             /* If program is malformed for any reason. */
 
+   /** Record of which phases R registers were last written in. */
+   GLuint register_phases[16];
+   GLuint indirections;
    GLuint nr_tex_indirect;
    GLuint nr_tex_insn;
    GLuint nr_alu_insn;
diff --git a/src/mesa/drivers/dri/i915/i915_program.c b/src/mesa/drivers/dri/i915/i915_program.c
index 49193297a8..350da5e169 100644
--- a/src/mesa/drivers/dri/i915/i915_program.c
+++ b/src/mesa/drivers/dri/i915/i915_program.c
@@ -190,6 +190,9 @@ i915_emit_arith(struct i915_fragment_program * p,
    *(p->csr++) = (A1_SRC0(src0) | A1_SRC1(src1));
    *(p->csr++) = (A2_SRC1(src1) | A2_SRC2(src2));
 
+   if (GET_UREG_TYPE(dest) == REG_TYPE_R)
+      p->register_phases[GET_UREG_NR(dest)] = p->nr_tex_indirect;
+
    p->nr_alu_insn++;
    return dest;
 }
@@ -237,10 +240,22 @@ GLuint i915_emit_texld( struct i915_fragment_program *p,
    else {
       assert(GET_UREG_TYPE(dest) != REG_TYPE_CONST);
       assert(dest = UREG(GET_UREG_TYPE(dest), GET_UREG_NR(dest)));
+      /* Can't use unsaved temps for coords, as the phase boundary would result
+       * in the contents becoming undefined.
+       */
+      assert(GET_UREG_TYPE(coord) != REG_TYPE_U);
+
+      /* Output register being oC or oD defines a phase boundary */
+      if (GET_UREG_TYPE(dest) == REG_TYPE_OC ||
+	  GET_UREG_TYPE(dest) == REG_TYPE_OD)
+	 p->nr_tex_indirect++;
 
-      if (GET_UREG_TYPE(coord) != REG_TYPE_T) {
+      /* Reading from an r# register whose contents depend on output of the
+       * current phase defines a phase boundary.
+       */
+      if (GET_UREG_TYPE(coord) == REG_TYPE_R &&
+	  p->register_phases[GET_UREG_NR(coord)] == p->nr_tex_indirect)
 	 p->nr_tex_indirect++;
-      }
 
       *(p->csr++) = (op | 
 		     T0_DEST( dest ) |
@@ -249,6 +264,9 @@ GLuint i915_emit_texld( struct i915_fragment_program *p,
       *(p->csr++) = T1_ADDRESS_REG( coord );
       *(p->csr++) = T2_MBZ;
 
+      if (GET_UREG_TYPE(dest) == REG_TYPE_R)
+	 p->register_phases[GET_UREG_NR(dest)] = p->nr_tex_indirect;
+
       p->nr_tex_insn++;
       return dest;
    }
@@ -413,7 +431,8 @@ i915_init_program(struct i915_context *i915, struct i915_fragment_program *p)
    p->on_hardware = 0;
    p->error = 0;
 
-   p->nr_tex_indirect = 1;      /* correct? */
+   memset(&p->register_phases, 0, sizeof(p->register_phases));
+   p->nr_tex_indirect = 1;
    p->nr_tex_insn = 0;
    p->nr_alu_insn = 0;
    p->nr_decl_insn = 0;
diff --git a/src/mesa/drivers/dri/i915/i915_texstate.c b/src/mesa/drivers/dri/i915/i915_texstate.c
index ae42b102db..d1b0dcdf31 100644
--- a/src/mesa/drivers/dri/i915/i915_texstate.c
+++ b/src/mesa/drivers/dri/i915/i915_texstate.c
@@ -307,10 +307,21 @@ i915_update_tex_unit(struct intel_context *intel, GLuint unit, GLuint ss3)
    }
 
 
-   state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(tObj->_BorderChan[0],
-                                                tObj->_BorderChan[1],
-                                                tObj->_BorderChan[2],
-                                                tObj->_BorderChan[3]);
+   if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) {
+      /* GL specs that border color for depth textures is taken from the
+       * R channel, while the hardware uses A.  Spam R into all the channels
+       * for safety.
+       */
+      state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(tObj->_BorderChan[0],
+						   tObj->_BorderChan[0],
+						   tObj->_BorderChan[0],
+						   tObj->_BorderChan[0]);
+   } else {
+      state[I915_TEXREG_SS4] = INTEL_PACKCOLOR8888(tObj->_BorderChan[0],
+						   tObj->_BorderChan[1],
+						   tObj->_BorderChan[2],
+						   tObj->_BorderChan[3]);
+   }
 
 
    I915_ACTIVESTATE(i915, I915_UPLOAD_TEX(unit), GL_TRUE);
diff --git a/src/mesa/drivers/dri/i915/i915_vtbl.c b/src/mesa/drivers/dri/i915/i915_vtbl.c
index edbbe23e09..7431a9cf76 100644
--- a/src/mesa/drivers/dri/i915/i915_vtbl.c
+++ b/src/mesa/drivers/dri/i915/i915_vtbl.c
@@ -490,6 +490,13 @@ i915_destroy_context(struct intel_context *intel)
    GLuint i;
    struct i915_context *i915 = i915_context(&intel->ctx);
 
+   intel_region_release(&i915->state.draw_region);
+   intel_region_release(&i915->state.depth_region);
+   intel_region_release(&i915->meta.draw_region);
+   intel_region_release(&i915->meta.depth_region);
+   intel_region_release(&i915->initial.draw_region);
+   intel_region_release(&i915->initial.depth_region);
+
    for (i = 0; i < I915_TEX_UNITS; i++) {
       if (i915->state.tex_buffer[i] != NULL) {
 	 dri_bo_unreference(i915->state.tex_buffer[i]);
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index c2d555cd0c..005460f354 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -18,8 +18,9 @@ DRIVER_SOURCES = \
 	intel_screen.c \
 	intel_span.c \
 	intel_pixel.c \
-	intel_pixel_copy.c \
 	intel_pixel_bitmap.c \
+	intel_pixel_copy.c \
+	intel_pixel_draw.c \
 	intel_state.c \
 	intel_tex.c \
 	intel_tex_copy.c \
@@ -51,6 +52,7 @@ DRIVER_SOURCES = \
 	brw_metaops.c \
 	brw_misc_state.c \
 	brw_program.c \
+	brw_queryobj.c \
 	brw_sf.c \
 	brw_sf_emit.c \
 	brw_sf_state.c \
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index 92629016d9..474158b484 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -33,6 +33,7 @@
 #include "main/imports.h"
 #include "main/api_noop.h"
 #include "main/vtxfmt.h"
+#include "main/simple_list.h"
 #include "shader/shader_api.h"
 
 #include "brw_context.h"
@@ -68,6 +69,7 @@ static void brwInitDriverFunctions( struct dd_function_table *functions )
 
    brwInitFragProgFuncs( functions );
    brwInitProgFuncs( functions );
+   brw_init_queryobj_functions(functions);
 }
 
 
@@ -150,9 +152,9 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis,
    ctx->VertexProgram._MaintainTnlProgram = GL_TRUE;
    ctx->FragmentProgram._MaintainTexEnvProgram = GL_TRUE;
 
-   brw_draw_init( brw );
+   make_empty_list(&brw->query.active_head);
 
-   brw_ProgramCacheInit( ctx );
+   brw_draw_init( brw );
 
    return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index b04487ecee..1c6a0dede0 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -130,7 +130,6 @@ struct brw_context;
 #define BRW_NEW_CONTEXT                 0x80
 #define BRW_NEW_WM_INPUT_DIMENSIONS     0x100
 #define BRW_NEW_INPUT_VARYING           0x200
-#define BRW_NEW_TNL_PROGRAM             0x400
 #define BRW_NEW_PSP                     0x800
 #define BRW_NEW_METAOPS                 0x1000
 #define BRW_NEW_FENCE                   0x2000
@@ -411,7 +410,22 @@ struct brw_tnl_cache {
    GLuint size, n_items;
 };
 
+struct brw_query_object {
+   struct gl_query_object Base;
 
+   /** Doubly linked list of active query objects in the context. */
+   struct brw_query_object *prev, *next;
+
+   /** Last query BO associated with this query. */
+   dri_bo *bo;
+   /** First index in bo with query data for this object. */
+   int first_index;
+   /** Last index in bo with query data for this object. */
+   int last_index;
+
+   /* Total count of pixels from previous BOs */
+   unsigned int count;
+};
 
 struct brw_context 
 {
@@ -488,10 +502,6 @@ struct brw_context
       GLboolean active;
    } metaops;
 
-   /* Track fixed function t&l in a vertex program:
-    */
-   struct gl_vertex_program *tnl_program;
-   struct brw_tnl_cache tnl_program_cache;
 
    /* Active vertex program: 
     */
@@ -631,7 +641,12 @@ struct brw_context
       dri_bo *vp_bo;
    } cc;
 
-   
+   struct {
+      struct brw_query_object active_head;
+      dri_bo *bo;
+      int index;
+      GLboolean active;
+   } query;
    /* Used to give every program string a unique id
     */
    GLuint program_id;
@@ -656,7 +671,13 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis,
 			    __DRIcontextPrivate *driContextPriv,
 			    void *sharedContextPrivate);
 
-
+/*======================================================================
+ * brw_queryobj.c
+ */
+void brw_init_queryobj_functions(struct dd_function_table *functions);
+void brw_prepare_query_begin(struct brw_context *brw);
+void brw_emit_query_begin(struct brw_context *brw);
+void brw_emit_query_end(struct brw_context *brw);
 
 /*======================================================================
  * brw_state.c
diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index 9a353fc7b6..6c71b4abcf 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -382,7 +382,6 @@ void brw_draw_prims( GLcontext *ctx,
       return;
    }
 
-
    /* Make a first attempt at drawing:
     */
    retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
@@ -395,6 +394,7 @@ void brw_draw_prims( GLcontext *ctx,
        _swsetup_Wakeup(ctx);
       _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
    }
+
 }
 
 void brw_draw_init( struct brw_context *brw )
@@ -409,8 +409,18 @@ void brw_draw_init( struct brw_context *brw )
 
 void brw_draw_destroy( struct brw_context *brw )
 {
+   int i;
+
    if (brw->vb.upload.bo != NULL) {
       dri_bo_unreference(brw->vb.upload.bo);
       brw->vb.upload.bo = NULL;
    }
+
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      dri_bo_unreference(brw->vb.inputs[i].bo);
+      brw->vb.inputs[i].bo = NULL;
+   }
+
+   dri_bo_unreference(brw->ib.bo);
+   brw->ib.bo = NULL;
 }
diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c
index 303eaac5cf..7b88b5eaa1 100644
--- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c
@@ -365,8 +365,10 @@ static void brw_prepare_vertices(struct brw_context *brw)
 	 if (i == 0) {
 	    /* Position array not properly enabled:
 	     */
-	    if (input->glarray->StrideB == 0)
-	      return;
+            if (input->glarray->StrideB == 0) {
+               intel->Fallback = 1;
+               return;
+            }
 
 	    interleave = input->glarray->StrideB;
 	    ptr = input->glarray->Ptr;
@@ -413,6 +415,8 @@ static void brw_prepare_vertices(struct brw_context *brw)
           copy_array_to_vbo_array(brw, upload[i], upload[i]->element_size);
       }
    }
+
+   brw_prepare_query_begin(brw);
 }
 
 static void brw_emit_vertices(struct brw_context *brw)
@@ -433,6 +437,7 @@ static void brw_emit_vertices(struct brw_context *brw)
       enabled[nr_enabled++] = input;
    }
 
+   brw_emit_query_begin(brw);
 
    /* Now emit VB and VEP state packets.
     *
diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c
index 2f6b7febbd..4ea660a51a 100644
--- a/src/mesa/drivers/dri/i965/brw_fallback.c
+++ b/src/mesa/drivers/dri/i965/brw_fallback.c
@@ -74,10 +74,7 @@ static GLboolean do_check_fallback(struct brw_context *brw)
       if (texUnit->_ReallyEnabled) {
 	 struct intel_texture_object *intelObj = intel_texture_object(texUnit->_Current);
 	 struct gl_texture_image *texImage = intelObj->base.Image[0][intelObj->firstLevel];
-	 if (texImage->Border ||
-         ((texImage->_BaseFormat == GL_DEPTH_COMPONENT) &&
-          ((texImage->TexObject->WrapS == GL_CLAMP_TO_BORDER) || 
-           (texImage->TexObject->WrapT == GL_CLAMP_TO_BORDER)))) {
+	 if (texImage->Border) {
 	    DBG("FALLBACK: texture border\n");
 	    return GL_TRUE;
 	 }
diff --git a/src/mesa/drivers/dri/i965/brw_queryobj.c b/src/mesa/drivers/dri/i965/brw_queryobj.c
new file mode 100644
index 0000000000..a1a1353dee
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/brw_queryobj.c
@@ -0,0 +1,263 @@
+/*
+ * Copyright © 2008 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
+ * IN THE SOFTWARE.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+/** @file support for ARB_query_object
+ *
+ * ARB_query_object is implemented by using the PIPE_CONTROL command to stall
+ * execution on the completion of previous depth tests, and write the
+ * current PS_DEPTH_COUNT to a buffer object.
+ *
+ * We use before and after counts when drawing during a query so that
+ * we don't pick up other clients' query data in ours.  To reduce overhead,
+ * a single BO is used to record the query data for all active queries at
+ * once.  This also gives us a simple bound on how much batchbuffer space is
+ * required for handling queries, so that we can be sure that we won't
+ * have to emit a batchbuffer without getting the ending PS_DEPTH_COUNT.
+ */
+#include "main/simple_list.h"
+#include "main/imports.h"
+
+#include "brw_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_reg.h"
+
+/** Waits on the query object's BO and totals the results for this query */
+static void
+brw_queryobj_get_results(struct brw_query_object *query)
+{
+   int i;
+   uint64_t *results;
+
+   if (query->bo == NULL)
+      return;
+
+   /* Map and count the pixels from the current query BO */
+   dri_bo_map(query->bo, GL_FALSE);
+   results = query->bo->virtual;
+   for (i = query->first_index; i <= query->last_index; i++) {
+      query->Base.Result += results[i * 2 + 1] - results[i * 2];
+   }
+   dri_bo_unmap(query->bo);
+
+   dri_bo_unreference(query->bo);
+   query->bo = NULL;
+}
+
+static struct gl_query_object *
+brw_new_query_object(GLcontext *ctx, GLuint id)
+{
+   struct brw_query_object *query;
+
+   query = _mesa_calloc(sizeof(struct brw_query_object));
+
+   query->Base.Id = id;
+   query->Base.Result = 0;
+   query->Base.Active = GL_FALSE;
+   query->Base.Ready = GL_TRUE;
+
+   return &query->Base;
+}
+
+static void
+brw_delete_query(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct brw_query_object *query = (struct brw_query_object *)q;
+
+   dri_bo_unreference(query->bo);
+   _mesa_free(query);
+}
+
+static void
+brw_begin_query(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct brw_context *brw = brw_context(ctx);
+   struct intel_context *intel = intel_context(ctx);
+   struct brw_query_object *query = (struct brw_query_object *)q;
+
+   /* Reset our driver's tracking of query state. */
+   dri_bo_unreference(query->bo);
+   query->bo = NULL;
+   query->first_index = -1;
+   query->last_index = -1;
+
+   insert_at_head(&brw->query.active_head, query);
+   intel->stats_wm++;
+}
+
+/**
+ * Begin the ARB_occlusion_query query on a query object.
+ */
+static void
+brw_end_query(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct brw_context *brw = brw_context(ctx);
+   struct intel_context *intel = intel_context(ctx);
+   struct brw_query_object *query = (struct brw_query_object *)q;
+
+   /* Flush the batchbuffer in case it has writes to our query BO.
+    * Have later queries write to a new query BO so that further rendering
+    * doesn't delay the collection of our results.
+    */
+   if (query->bo) {
+      brw_emit_query_end(brw);
+      intel_batchbuffer_flush(intel->batch);
+
+      dri_bo_unreference(brw->query.bo);
+      brw->query.bo = NULL;
+   }
+
+   remove_from_list(query);
+
+   intel->stats_wm--;
+}
+
+static void brw_wait_query(GLcontext *ctx, struct gl_query_object *q)
+{
+   struct brw_query_object *query = (struct brw_query_object *)q;
+
+   brw_queryobj_get_results(query);
+   query->Base.Ready = GL_TRUE;
+}
+
+static void brw_check_query(GLcontext *ctx, struct gl_query_object *q)
+{
+   /* XXX: Need to expose dri_bo_is_idle from bufmgr. */
+#if 0
+   struct brw_query_object *query = (struct brw_query_object *)q;
+
+   if (dri_bo_is_idle(query->bo)) {
+      brw_queryobj_get_results(query);
+      query->Base.Ready = GL_TRUE;
+   }
+#else
+   brw_wait_query(ctx, q);
+#endif
+}
+
+/** Called to set up the query BO and account for its aperture space */
+void
+brw_prepare_query_begin(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+   dri_bo *aper_array[] = {
+      intel->batch->buf,
+      brw->query.bo,
+   };
+
+   /* Skip if we're not doing any queries. */
+   if (is_empty_list(&brw->query.active_head))
+      return;
+
+   /* Get a new query BO if we're going to need it. */
+   if (brw->query.bo == NULL ||
+       brw->query.index * 2 + 1 >= 4096 / sizeof(uint64_t)) {
+      dri_bo_unreference(brw->query.bo);
+      brw->query.bo = NULL;
+
+      brw->query.bo = dri_bo_alloc(intel->bufmgr, "query", 4096, 1);
+      brw->query.index = 0;
+   }
+
+   if (dri_bufmgr_check_aperture_space(aper_array, ARRAY_SIZE(aper_array)))
+      intel_batchbuffer_flush(intel->batch);
+}
+
+/** Called just before primitive drawing to get a beginning PS_DEPTH_COUNT. */
+void
+brw_emit_query_begin(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+   struct brw_query_object *query;
+
+   /* Skip if we're not doing any queries, or we've emitted the start. */
+   if (brw->query.active || is_empty_list(&brw->query.active_head))
+      return;
+
+   BEGIN_BATCH(4, IGNORE_CLIPRECTS);
+   OUT_BATCH(_3DSTATE_PIPE_CONTROL |
+	     PIPE_CONTROL_DEPTH_STALL |
+	     PIPE_CONTROL_WRITE_DEPTH_COUNT);
+   /* This object could be mapped cacheable, but we don't have an exposed
+    * mechanism to support that.  Since it's going uncached, tell GEM that
+    * we're writing to it.  The usual clflush should be all that's required
+    * to pick up the results.
+    */
+   OUT_RELOC(brw->query.bo,
+	     I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
+	     PIPE_CONTROL_GLOBAL_GTT_WRITE |
+	     ((brw->query.index * 2) * sizeof(uint64_t)));
+   OUT_BATCH(0);
+   OUT_BATCH(0);
+   ADVANCE_BATCH();
+
+   foreach(query, &brw->query.active_head) {
+      if (query->bo != brw->query.bo) {
+	 if (query->bo != NULL)
+	    brw_queryobj_get_results(query);
+	 dri_bo_reference(brw->query.bo);
+	 query->bo = brw->query.bo;
+	 query->first_index = brw->query.index;
+      }
+      query->last_index = brw->query.index;
+   }
+   brw->query.active = GL_TRUE;
+}
+
+/** Called at batchbuffer flush to get an ending PS_DEPTH_COUNT */
+void
+brw_emit_query_end(struct brw_context *brw)
+{
+   struct intel_context *intel = &brw->intel;
+
+   if (!brw->query.active)
+      return;
+
+   BEGIN_BATCH(4, IGNORE_CLIPRECTS);
+   OUT_BATCH(_3DSTATE_PIPE_CONTROL |
+	     PIPE_CONTROL_DEPTH_STALL |
+	     PIPE_CONTROL_WRITE_DEPTH_COUNT);
+   OUT_RELOC(brw->query.bo,
+	     I915_GEM_DOMAIN_INSTRUCTION, I915_GEM_DOMAIN_INSTRUCTION,
+	     PIPE_CONTROL_GLOBAL_GTT_WRITE |
+	     ((brw->query.index * 2 + 1) * sizeof(uint64_t)));
+   OUT_BATCH(0);
+   OUT_BATCH(0);
+   ADVANCE_BATCH();
+
+   brw->query.active = GL_FALSE;
+   brw->query.index++;
+}
+
+void brw_init_queryobj_functions(struct dd_function_table *functions)
+{
+   functions->NewQueryObject = brw_new_query_object;
+   functions->DeleteQuery = brw_delete_query;
+   functions->BeginQuery = brw_begin_query;
+   functions->EndQuery = brw_end_query;
+   functions->CheckQuery = brw_check_query;
+   functions->WaitQuery = brw_wait_query;
+}
diff --git a/src/mesa/drivers/dri/i965/brw_state.h b/src/mesa/drivers/dri/i965/brw_state.h
index 3ea6151ae9..4c04036ef0 100644
--- a/src/mesa/drivers/dri/i965/brw_state.h
+++ b/src/mesa/drivers/dri/i965/brw_state.h
@@ -74,7 +74,6 @@ const struct brw_tracked_state brw_wm_unit;
 const struct brw_tracked_state brw_psp_urb_cbs;
 
 const struct brw_tracked_state brw_active_vertprog;
-const struct brw_tracked_state brw_tnl_vertprog;
 const struct brw_tracked_state brw_pipe_control;
 
 const struct brw_tracked_state brw_clear_surface_cache;
diff --git a/src/mesa/drivers/dri/i965/brw_state_cache.c b/src/mesa/drivers/dri/i965/brw_state_cache.c
index 1318dea594..d5b5166406 100644
--- a/src/mesa/drivers/dri/i965/brw_state_cache.c
+++ b/src/mesa/drivers/dri/i965/brw_state_cache.c
@@ -497,9 +497,10 @@ void brw_destroy_cache( struct brw_context *brw )
    GLuint i;
 
    brw_clear_cache(brw);
-   for (i = 0; i < BRW_MAX_CACHE; i++)
+   for (i = 0; i < BRW_MAX_CACHE; i++) {
+      dri_bo_unreference(brw->cache.last_bo[i]);
       free(brw->cache.name[i]);
-
+   }
    free(brw->cache.items);
    brw->cache.items = NULL;
    brw->cache.size = 0;
diff --git a/src/mesa/drivers/dri/i965/brw_state_upload.c b/src/mesa/drivers/dri/i965/brw_state_upload.c
index 7d4fd467b1..b6a52843a8 100644
--- a/src/mesa/drivers/dri/i965/brw_state_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_state_upload.c
@@ -45,7 +45,6 @@ const struct brw_tracked_state *atoms[] =
 {
    &brw_check_fallback,
 
-   &brw_tnl_vertprog,
    &brw_active_vertprog,
    &brw_wm_input_sizes,
    &brw_vs_prog,
diff --git a/src/mesa/drivers/dri/i965/brw_vs.h b/src/mesa/drivers/dri/i965/brw_vs.h
index 41a33ffe38..22388ec99d 100644
--- a/src/mesa/drivers/dri/i965/brw_vs.h
+++ b/src/mesa/drivers/dri/i965/brw_vs.h
@@ -80,8 +80,4 @@ struct brw_vs_compile {
 
 void brw_vs_emit( struct brw_vs_compile *c );
 
-
-void brw_ProgramCacheDestroy( GLcontext *ctx );
-void brw_ProgramCacheInit( GLcontext *ctx );
-
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_vs_tnl.c b/src/mesa/drivers/dri/i965/brw_vs_tnl.c
index 9b04f19112..eacc289f1f 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_tnl.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_tnl.c
@@ -33,1635 +33,15 @@
 #include "main/glheader.h"
 #include "main/macros.h"
 #include "main/enums.h"
-#include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
-#include "shader/prog_statevars.h"
 #include "brw_vs.h"
 #include "brw_state.h"
 
 
-struct state_key {
-   unsigned light_global_enabled:1;
-   unsigned light_local_viewer:1;
-   unsigned light_twoside:1;
-   unsigned light_color_material:1;
-   unsigned light_color_material_mask:12;
-   unsigned light_material_mask:12;
-   unsigned normalize:1;
-   unsigned rescale_normals:1;
-   unsigned fog_source_is_depth:1;
-   unsigned tnl_do_vertex_fog:1;
-   unsigned separate_specular:1;
-   unsigned fog_option:2;
-   unsigned point_attenuated:1;
-   unsigned texture_enabled_global:1;
-   unsigned fragprog_inputs_read:12;
-
-   struct {
-      unsigned light_enabled:1;
-      unsigned light_eyepos3_is_zero:1;
-      unsigned light_spotcutoff_is_180:1;
-      unsigned light_attenuated:1;      
-      unsigned texunit_really_enabled:1;
-      unsigned texmat_enabled:1;
-      unsigned texgen_enabled:4;
-      unsigned texgen_mode0:4;
-      unsigned texgen_mode1:4;
-      unsigned texgen_mode2:4;
-      unsigned texgen_mode3:4;
-   } unit[8];
-};
-
-
-
-#define FOG_NONE   0
-#define FOG_LINEAR 1
-#define FOG_EXP    2
-#define FOG_EXP2   3
-
-static GLuint translate_fog_mode( GLenum mode )
-{
-   switch (mode) {
-   case GL_LINEAR: return FOG_LINEAR;
-   case GL_EXP: return FOG_EXP;
-   case GL_EXP2: return FOG_EXP2;
-   default: return FOG_NONE;
-   }
-}
-
-#define TXG_NONE           0
-#define TXG_OBJ_LINEAR     1
-#define TXG_EYE_LINEAR     2
-#define TXG_SPHERE_MAP     3
-#define TXG_REFLECTION_MAP 4
-#define TXG_NORMAL_MAP     5
-
-static GLuint translate_texgen( GLboolean enabled, GLenum mode )
-{
-   if (!enabled)
-      return TXG_NONE;
-
-   switch (mode) {
-   case GL_OBJECT_LINEAR: return TXG_OBJ_LINEAR;
-   case GL_EYE_LINEAR: return TXG_EYE_LINEAR;
-   case GL_SPHERE_MAP: return TXG_SPHERE_MAP;
-   case GL_REFLECTION_MAP_NV: return TXG_REFLECTION_MAP;
-   case GL_NORMAL_MAP_NV: return TXG_NORMAL_MAP;
-   default: return TXG_NONE;
-   }
-}
-
-static void make_state_key( GLcontext *ctx, struct state_key *key )
-{
-   struct brw_context *brw = brw_context(ctx);
-   const struct gl_fragment_program *fp = brw->fragment_program;
-   GLuint i;
-
-   /* This now relies on texenvprogram.c being active:
-    */
-   assert(fp);
-
-   memset(key, 0, sizeof(*key));
-
-   /* BRW_NEW_FRAGMENT_PROGRAM */
-   key->fragprog_inputs_read = fp->Base.InputsRead;
-
-   /* _NEW_LIGHT */
-   key->separate_specular = (brw->attribs.Light->Model.ColorControl ==
-			     GL_SEPARATE_SPECULAR_COLOR);
-
-   /* _NEW_LIGHT */
-   if (brw->attribs.Light->Enabled) {
-      key->light_global_enabled = 1;
-
-      if (brw->attribs.Light->Model.LocalViewer)
-	 key->light_local_viewer = 1;
-
-      if (brw->attribs.Light->Model.TwoSide)
-	 key->light_twoside = 1;
-
-      if (brw->attribs.Light->ColorMaterialEnabled) {
-	 key->light_color_material = 1;
-	 key->light_color_material_mask = brw->attribs.Light->ColorMaterialBitmask;
-      }
-
-      /* BRW_NEW_INPUT_VARYING */
-
-      /* For these programs, material values are stuffed into the
-       * generic slots:
-       */
-      for (i = 0 ; i < MAT_ATTRIB_MAX ; i++) 
-	 if (brw->vb.info.varying & (1<<(VERT_ATTRIB_GENERIC0 + i))) 
-	    key->light_material_mask |= 1<<i;
-
-      for (i = 0; i < MAX_LIGHTS; i++) {
-	 struct gl_light *light = &brw->attribs.Light->Light[i];
-
-	 if (light->Enabled) {
-	    key->unit[i].light_enabled = 1;
-
-	    if (light->EyePosition[3] == 0.0)
-	       key->unit[i].light_eyepos3_is_zero = 1;
-	    
-	    if (light->SpotCutoff == 180.0)
-	       key->unit[i].light_spotcutoff_is_180 = 1;
-
-	    if (light->ConstantAttenuation != 1.0 ||
-		light->LinearAttenuation != 0.0 ||
-		light->QuadraticAttenuation != 0.0)
-	       key->unit[i].light_attenuated = 1;
-	 }
-      }
-   }
-
-   /* _NEW_TRANSFORM */
-   if (brw->attribs.Transform->Normalize)
-      key->normalize = 1;
-
-   if (brw->attribs.Transform->RescaleNormals)
-      key->rescale_normals = 1;
-
-   /* BRW_NEW_FRAGMENT_PROGRAM */
-   key->fog_option = translate_fog_mode(fp->FogOption);
-   if (key->fog_option)
-      key->fragprog_inputs_read |= FRAG_BIT_FOGC;
-   
-   /* _NEW_FOG */
-   if (brw->attribs.Fog->FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT)
-      key->fog_source_is_depth = 1;
-   
-   /* _NEW_HINT, ??? */
-   if (1)
-      key->tnl_do_vertex_fog = 1;
-
-   /* _NEW_POINT */
-   if (brw->attribs.Point->_Attenuated)
-      key->point_attenuated = 1;
-
-   /* _NEW_TEXTURE */
-   if (brw->attribs.Texture->_TexGenEnabled ||
-       brw->attribs.Texture->_TexMatEnabled ||
-       brw->attribs.Texture->_EnabledUnits)
-      key->texture_enabled_global = 1;
-      
-   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
-      struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[i];
-
-      if (texUnit->_ReallyEnabled)
- 	 key->unit[i].texunit_really_enabled = 1;
-
-      if (brw->attribs.Texture->_TexMatEnabled & ENABLE_TEXMAT(i))      
-	 key->unit[i].texmat_enabled = 1;
-      
-      if (texUnit->TexGenEnabled) {
-	 key->unit[i].texgen_enabled = 1;
-      
-	 key->unit[i].texgen_mode0 = 
-	    translate_texgen( texUnit->TexGenEnabled & (1<<0),
-			      texUnit->GenModeS );
-	 key->unit[i].texgen_mode1 = 
-	    translate_texgen( texUnit->TexGenEnabled & (1<<1),
-			      texUnit->GenModeT );
-	 key->unit[i].texgen_mode2 = 
-	    translate_texgen( texUnit->TexGenEnabled & (1<<2),
-			      texUnit->GenModeR );
-	 key->unit[i].texgen_mode3 = 
-	    translate_texgen( texUnit->TexGenEnabled & (1<<3),
-			      texUnit->GenModeQ );
-      }
-   }
-}
-
-
-   
-/* Very useful debugging tool - produces annotated listing of
- * generated program with line/function references for each
- * instruction back into this file:
- */
-#define DISASSEM 0
-
-/* Should be tunable by the driver - do we want to do matrix
- * multiplications with DP4's or with MUL/MAD's?  SSE works better
- * with the latter, drivers may differ.
- */
-#define PREFER_DP4 1
-
-
-/* Use uregs to represent registers internally, translate to Mesa's
- * expected formats on emit.  
- *
- * NOTE: These are passed by value extensively in this file rather
- * than as usual by pointer reference.  If this disturbs you, try
- * remembering they are just 32bits in size.
- *
- * GCC is smart enough to deal with these dword-sized structures in
- * much the same way as if I had defined them as dwords and was using
- * macros to access and set the fields.  This is much nicer and easier
- * to evolve.
- */
-struct ureg {
-   GLuint file:4;
-   GLint idx:8;      /* relative addressing may be negative */
-   GLuint negate:1;
-   GLuint swz:12;
-   GLuint pad:7;
-};
-
-
-struct tnl_program {
-   const struct state_key *state;
-   struct gl_vertex_program *program;
-   
-   GLuint nr_instructions;
-   GLuint temp_in_use;
-   GLuint temp_reserved;
-   
-   struct ureg eye_position;
-   struct ureg eye_position_normalized;
-   struct ureg eye_normal;
-   struct ureg identity;
-
-   GLuint materials;
-   GLuint color_materials;
-};
-
-
-const static struct ureg undef = { 
-   PROGRAM_UNDEFINED,
-   ~0,
-   0,
-   0,
-   0
-};
-
-/* Local shorthand:
- */
-#define X    SWIZZLE_X
-#define Y    SWIZZLE_Y
-#define Z    SWIZZLE_Z
-#define W    SWIZZLE_W
-
-
-/* Construct a ureg:
- */
-static struct ureg make_ureg(GLuint file, GLint idx)
-{
-   struct ureg reg;
-   reg.file = file;
-   reg.idx = idx;
-   reg.negate = 0;
-   reg.swz = SWIZZLE_NOOP;
-   reg.pad = 0;
-   return reg;
-}
-
-
-
-static struct ureg ureg_negate( struct ureg reg )
-{
-   reg.negate ^= 1;
-   return reg;
-} 
-
-
-static struct ureg swizzle( struct ureg reg, int x, int y, int z, int w )
-{
-   reg.swz = MAKE_SWIZZLE4(GET_SWZ(reg.swz, x),
-			   GET_SWZ(reg.swz, y),
-			   GET_SWZ(reg.swz, z),
-			   GET_SWZ(reg.swz, w));
-
-   return reg;
-}
-
-static struct ureg swizzle1( struct ureg reg, int x )
-{
-   return swizzle(reg, x, x, x, x);
-}
-
-static struct ureg get_temp( struct tnl_program *p )
-{
-   int bit = ffs( ~p->temp_in_use );
-   if (!bit) {
-      fprintf(stderr, "%s: out of temporaries\n", __FILE__);
-      assert(0);
-   }
-
-   if (bit > p->program->Base.NumTemporaries)
-      p->program->Base.NumTemporaries = bit;
-
-   p->temp_in_use |= 1<<(bit-1);
-   return make_ureg(PROGRAM_TEMPORARY, bit-1);
-}
-
-static struct ureg reserve_temp( struct tnl_program *p )
-{
-   struct ureg temp = get_temp( p );
-   p->temp_reserved |= 1<<temp.idx;
-   return temp;
-}
-
-static void release_temp( struct tnl_program *p, struct ureg reg )
-{
-   if (reg.file == PROGRAM_TEMPORARY) {
-      p->temp_in_use &= ~(1<<reg.idx);
-      p->temp_in_use |= p->temp_reserved; /* can't release reserved temps */
-   }
-}
-
-static void release_temps( struct tnl_program *p )
-{
-   p->temp_in_use = p->temp_reserved;
-}
-
-
-
-static struct ureg register_input( struct tnl_program *p, GLuint input )
-{
-   assert(input < 32);
-
-   p->program->Base.InputsRead |= (1<<input);
-   return make_ureg(PROGRAM_INPUT, input);
-}
-
-static struct ureg register_output( struct tnl_program *p, GLuint output )
-{
-   p->program->Base.OutputsWritten |= (1<<output);
-   return make_ureg(PROGRAM_OUTPUT, output);
-}
-
-static struct ureg register_const4f( struct tnl_program *p, 
-			      GLfloat s0,
-			      GLfloat s1,
-			      GLfloat s2,
-			      GLfloat s3)
-{
-   GLfloat values[4];
-   GLint idx;
-   GLuint swizzle;
-   values[0] = s0;
-   values[1] = s1;
-   values[2] = s2;
-   values[3] = s3;
-   idx = _mesa_add_unnamed_constant( p->program->Base.Parameters, values, 4,
-                                     &swizzle);
-   assert(swizzle == SWIZZLE_NOOP); /* Need to handle swizzle in reg setup */
-   return make_ureg(PROGRAM_STATE_VAR, idx);
-}
-
-#define register_const1f(p, s0)         register_const4f(p, s0, 0, 0, 1)
-#define register_scalar_const(p, s0)    register_const4f(p, s0, s0, s0, s0)
-#define register_const2f(p, s0, s1)     register_const4f(p, s0, s1, 0, 1)
-#define register_const3f(p, s0, s1, s2) register_const4f(p, s0, s1, s2, 1)
-
-static GLboolean is_undef( struct ureg reg )
-{
-   return reg.file == PROGRAM_UNDEFINED;
-}
-
-static struct ureg get_identity_param( struct tnl_program *p )
-{
-   if (is_undef(p->identity)) 
-      p->identity = register_const4f(p, 0,0,0,1);
-
-   return p->identity;
-}
-
-static struct ureg register_param5( struct tnl_program *p, 
-                                    GLint s0,
-                                    GLint s1,
-                                    GLint s2,
-                                    GLint s3,
-                                    GLint s4)
-{
-   gl_state_index tokens[STATE_LENGTH];
-   GLint idx;
-   tokens[0] = s0;
-   tokens[1] = s1;
-   tokens[2] = s2;
-   tokens[3] = s3;
-   tokens[4] = s4;
-   idx = _mesa_add_state_reference( p->program->Base.Parameters, tokens );
-   return make_ureg(PROGRAM_STATE_VAR, idx);
-}
-
-
-#define register_param1(p,s0)          register_param5(p,s0,0,0,0,0)
-#define register_param2(p,s0,s1)       register_param5(p,s0,s1,0,0,0)
-#define register_param3(p,s0,s1,s2)    register_param5(p,s0,s1,s2,0,0)
-#define register_param4(p,s0,s1,s2,s3) register_param5(p,s0,s1,s2,s3,0)
-
-
-static void register_matrix_param5( struct tnl_program *p,
-				    GLint s0, /* matrix name */
-				    GLint s1, /* texture matrix number */
-				    GLint s2, /* first row */
-				    GLint s3, /* last row */
-				    GLint s4, /* modifier */
-				    struct ureg *matrix )
-{
-   GLint i;
-
-   /* This is a bit sad as the support is there to pull the whole
-    * matrix out in one go:
-    */
-   for (i = 0; i <= s3 - s2; i++) 
-      matrix[i] = register_param5( p, s0, s1, i, i, s4 );
-}
-
-
-static void emit_arg( struct prog_src_register *src,
-		      struct ureg reg )
-{
-   src->File = reg.file;
-   src->Index = reg.idx;
-   src->Swizzle = reg.swz;
-   src->RelAddr = 0;
-   src->NegateBase = reg.negate;
-   src->Abs = 0;
-   src->NegateAbs = 0;
-}
-
-static void emit_dst( struct prog_dst_register *dst,
-		      struct ureg reg, GLuint mask )
-{
-   dst->File = reg.file;
-   dst->Index = reg.idx;
-   /* allow zero as a shorthand for xyzw */
-   dst->WriteMask = mask ? mask : WRITEMASK_XYZW; 
-   dst->CondMask = 0;
-   dst->CondSwizzle = 0;
-   dst->CondSrc = 0;
-   dst->pad = 0;
-}
-
-static void debug_insn( struct prog_instruction *inst, const char *fn,
-			GLuint line )
-{
-   if (DISASSEM) {
-      static const char *last_fn;
-   
-      if (fn != last_fn) {
-	 last_fn = fn;
-	 _mesa_printf("%s:\n", fn);
-      }
-	 
-      _mesa_printf("%d:\t", line);
-      _mesa_print_instruction(inst);
-   }
-}
-
-
-static void emit_op3fn(struct tnl_program *p,
-		       GLuint op,
-		       struct ureg dest,
-		       GLuint mask,
-		       struct ureg src0,
-		       struct ureg src1,
-		       struct ureg src2,
-		       const char *fn,
-		       GLuint line)
-{
-   GLuint nr = p->program->Base.NumInstructions++;
-      
-   if (nr >= p->nr_instructions) {
-      int new_nr_instructions = p->nr_instructions * 2;
-
-      p->program->Base.Instructions = 
-	 _mesa_realloc(p->program->Base.Instructions,
-		       sizeof(struct prog_instruction) * p->nr_instructions,
-		       sizeof(struct prog_instruction) * new_nr_instructions);
-      p->nr_instructions = new_nr_instructions;
-   }
-
-   {      
-      struct prog_instruction *inst = &p->program->Base.Instructions[nr];
-      memset(inst, 0, sizeof(*inst));
-      inst->Opcode = op; 
-      inst->StringPos = 0;
-      inst->Data = 0;
-   
-      emit_arg( &inst->SrcReg[0], src0 );
-      emit_arg( &inst->SrcReg[1], src1 );
-      emit_arg( &inst->SrcReg[2], src2 );   
-
-      emit_dst( &inst->DstReg, dest, mask );
-
-      debug_insn(inst, fn, line);
-   }
-}
-
-   
-
-#define emit_op3(p, op, dst, mask, src0, src1, src2) \
-   emit_op3fn(p, op, dst, mask, src0, src1, src2, __FUNCTION__, __LINE__)
-
-#define emit_op2(p, op, dst, mask, src0, src1) \
-    emit_op3fn(p, op, dst, mask, src0, src1, undef, __FUNCTION__, __LINE__)
-
-#define emit_op1(p, op, dst, mask, src0) \
-    emit_op3fn(p, op, dst, mask, src0, undef, undef, __FUNCTION__, __LINE__)
-
-
-static struct ureg make_temp( struct tnl_program *p, struct ureg reg )
-{
-   if (reg.file == PROGRAM_TEMPORARY && 
-       !(p->temp_reserved & (1<<reg.idx)))
-      return reg;
-   else {
-      struct ureg temp = get_temp(p);
-      emit_op1(p, OPCODE_MOV, temp, 0, reg);
-      return temp;
-   }
-}
-
-
-/* Currently no tracking performed of input/output/register size or
- * active elements.  Could be used to reduce these operations, as
- * could the matrix type.
- */
-static void emit_matrix_transform_vec4( struct tnl_program *p,
-					struct ureg dest,
-					const struct ureg *mat,
-					struct ureg src)
-{
-   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_X, src, mat[0]);
-   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_Y, src, mat[1]);
-   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_Z, src, mat[2]);
-   emit_op2(p, OPCODE_DP4, dest, WRITEMASK_W, src, mat[3]);
-}
-
-/* This version is much easier to implement if writemasks are not
- * supported natively on the target or (like SSE), the target doesn't
- * have a clean/obvious dotproduct implementation.
- */
-static void emit_transpose_matrix_transform_vec4( struct tnl_program *p,
-						  struct ureg dest,
-						  const struct ureg *mat,
-						  struct ureg src)
-{
-   struct ureg tmp;
-
-   if (dest.file != PROGRAM_TEMPORARY)
-      tmp = get_temp(p);
-   else
-      tmp = dest;
-
-   emit_op2(p, OPCODE_MUL, tmp, 0, swizzle1(src,X), mat[0]);
-   emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Y), mat[1], tmp);
-   emit_op3(p, OPCODE_MAD, tmp, 0, swizzle1(src,Z), mat[2], tmp);
-   emit_op3(p, OPCODE_MAD, dest, 0, swizzle1(src,W), mat[3], tmp);
-
-   if (dest.file != PROGRAM_TEMPORARY)
-      release_temp(p, tmp);
-}
-
-static void emit_matrix_transform_vec3( struct tnl_program *p,
-					struct ureg dest,
-					const struct ureg *mat,
-					struct ureg src)
-{
-   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_X, src, mat[0]);
-   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_Y, src, mat[1]);
-   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_Z, src, mat[2]);
-}
-
-
-static void emit_normalize_vec3( struct tnl_program *p,
-				 struct ureg dest,
-				 struct ureg src )
-{
-   emit_op2(p, OPCODE_DP3, dest, WRITEMASK_W, src, src);
-   emit_op1(p, OPCODE_RSQ, dest, WRITEMASK_W, swizzle1(dest,W));
-   emit_op2(p, OPCODE_MUL, dest, WRITEMASK_XYZ, src, swizzle1(dest,W));
-}
-
-static void emit_passthrough( struct tnl_program *p, 
-			      GLuint input,
-			      GLuint output )
-{
-   struct ureg out = register_output(p, output);
-   emit_op1(p, OPCODE_MOV, out, 0, register_input(p, input)); 
-}
-
-static struct ureg get_eye_position( struct tnl_program *p )
-{
-   if (is_undef(p->eye_position)) {
-      struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
-      struct ureg modelview[4];
-
-      p->eye_position = reserve_temp(p);
-
-      if (PREFER_DP4) {
-	 register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3, 
-				 0, modelview );
-
-	 emit_matrix_transform_vec4(p, p->eye_position, modelview, pos);
-      }
-      else {
-	 register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 3, 
-				 STATE_MATRIX_TRANSPOSE, modelview );
-
-	 emit_transpose_matrix_transform_vec4(p, p->eye_position, modelview, pos);
-      }
-   }
-   
-   return p->eye_position;
-}
-
-
-#if 0
-static struct ureg get_eye_z( struct tnl_program *p )
-{
-   if (!is_undef(p->eye_position)) {
-      return swizzle1(p->eye_position, Z);
-   }
-   else if (!is_undef(p->eye_z)) {
-      struct ureg pos = register_input( p, BRW_ATTRIB_POS ); 
-      struct ureg modelview2;
-
-      p->eye_z = reserve_temp(p);
-
-      register_matrix_param6( p, STATE_MATRIX, STATE_MODELVIEW, 0, 2, 1, 
-			      STATE_MATRIX, &modelview2 );
-
-      emit_matrix_transform_vec4(p, p->eye_position, modelview, pos);
-      emit_op2(p, OPCODE_DP4, p->eye_z, WRITEMASK_Z, pos, modelview2);
-   }
-   
-   return swizzle1(p->eye_z, Z)
-}
-#endif
-
-
-
-static struct ureg get_eye_position_normalized( struct tnl_program *p )
-{
-   if (is_undef(p->eye_position_normalized)) {
-      struct ureg eye = get_eye_position(p);
-      p->eye_position_normalized = reserve_temp(p);
-      emit_normalize_vec3(p, p->eye_position_normalized, eye);
-   }
-   
-   return p->eye_position_normalized;
-}
-
-
-static struct ureg get_eye_normal( struct tnl_program *p )
-{
-   if (is_undef(p->eye_normal)) {
-      struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL );
-      struct ureg mvinv[3];
-
-      register_matrix_param5( p, STATE_MODELVIEW_MATRIX, 0, 0, 2,
-			      STATE_MATRIX_INVTRANS, mvinv );
-
-      p->eye_normal = reserve_temp(p);
-
-      /* Transform to eye space:
-       */
-      emit_matrix_transform_vec3( p, p->eye_normal, mvinv, normal );
-
-      /* Normalize/Rescale:
-       */
-      if (p->state->normalize) {
-	 emit_normalize_vec3( p, p->eye_normal, p->eye_normal );
-      }
-      else if (p->state->rescale_normals) {
-	 struct ureg rescale = register_param2(p, STATE_INTERNAL,
-					       STATE_NORMAL_SCALE);
-
-	 emit_op2( p, OPCODE_MUL, p->eye_normal, 0, p->eye_normal, 
-		   swizzle1(rescale, X));
-      }
-   }
-
-   return p->eye_normal;
-}
-
-
-
-static void build_hpos( struct tnl_program *p )
-{
-   struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
-   struct ureg hpos = register_output( p, VERT_RESULT_HPOS );
-   struct ureg mvp[4];
-
-   if (PREFER_DP4) {
-      register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, 
-			      0, mvp );
-      emit_matrix_transform_vec4( p, hpos, mvp, pos );
-   }
-   else {
-      register_matrix_param5( p, STATE_MVP_MATRIX, 0, 0, 3, 
-			      STATE_MATRIX_TRANSPOSE, mvp );
-      emit_transpose_matrix_transform_vec4( p, hpos, mvp, pos );
-   }
-}
-
-
-static GLuint material_attrib( GLuint side, GLuint property )
-{
-   return (property - STATE_AMBIENT) * 2 + side;
-}
-
-/* Get a bitmask of which material values vary on a per-vertex basis.
- */
-static void set_material_flags( struct tnl_program *p )
-{
-   p->color_materials = 0;
-   p->materials = 0;
-
-   if (p->state->light_color_material) {
-      p->materials = 
-	 p->color_materials = p->state->light_color_material_mask;
-   }
-
-   p->materials |= p->state->light_material_mask;
-}
-
-
-static struct ureg get_material( struct tnl_program *p, GLuint side, 
-				 GLuint property )
-{
-   GLuint attrib = material_attrib(side, property);
-
-   if (p->color_materials & (1<<attrib))
-      return register_input(p, VERT_ATTRIB_COLOR0);
-   else if (p->materials & (1<<attrib)) 
-      return register_input( p, attrib + _TNL_ATTRIB_MAT_FRONT_AMBIENT );
-   else
-      return register_param3( p, STATE_MATERIAL, side, property );
-}
-
-#define SCENE_COLOR_BITS(side) ((MAT_BIT_FRONT_EMISSION | \
-				 MAT_BIT_FRONT_AMBIENT | \
-				 MAT_BIT_FRONT_DIFFUSE) << (side))
-
-/* Either return a precalculated constant value or emit code to
- * calculate these values dynamically in the case where material calls
- * are present between begin/end pairs.
- *
- * Probably want to shift this to the program compilation phase - if
- * we always emitted the calculation here, a smart compiler could
- * detect that it was constant (given a certain set of inputs), and
- * lift it out of the main loop.  That way the programs created here
- * would be independent of the vertex_buffer details.
- */
-static struct ureg get_scenecolor( struct tnl_program *p, GLuint side )
-{
-   if (p->materials & SCENE_COLOR_BITS(side)) {
-      struct ureg lm_ambient = register_param1(p, STATE_LIGHTMODEL_AMBIENT);
-      struct ureg material_emission = get_material(p, side, STATE_EMISSION);
-      struct ureg material_ambient = get_material(p, side, STATE_AMBIENT);
-      struct ureg material_diffuse = get_material(p, side, STATE_DIFFUSE);
-      struct ureg tmp = make_temp(p, material_diffuse);
-      emit_op3(p, OPCODE_MAD, tmp,  WRITEMASK_XYZ, lm_ambient, 
-	       material_ambient, material_emission);
-      return tmp;
-   }
-   else
-      return register_param2( p, STATE_LIGHTMODEL_SCENECOLOR, side );
-}
-
-
-static struct ureg get_lightprod( struct tnl_program *p, GLuint light, 
-				  GLuint side, GLuint property )
-{
-   GLuint attrib = material_attrib(side, property);
-   if (p->materials & (1<<attrib)) {
-      struct ureg light_value = 
-	 register_param3(p, STATE_LIGHT, light, property);
-      struct ureg material_value = get_material(p, side, property);
-      struct ureg tmp = get_temp(p);
-      emit_op2(p, OPCODE_MUL, tmp,  0, light_value, material_value);
-      return tmp;
-   }
-   else
-      return register_param4(p, STATE_LIGHTPROD, light, side, property);
-}
-
-static struct ureg calculate_light_attenuation( struct tnl_program *p,
-						GLuint i, 
-						struct ureg VPpli,
-						struct ureg dist )
-{
-   struct ureg attenuation = register_param3(p, STATE_LIGHT, i,
-					     STATE_ATTENUATION);
-   struct ureg att = get_temp(p);
-
-   /* Calculate spot attenuation:
-    */
-   if (!p->state->unit[i].light_spotcutoff_is_180) {
-      struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL,
-						  STATE_LIGHT_SPOT_DIR_NORMALIZED, i);
-      struct ureg spot = get_temp(p);
-      struct ureg slt = get_temp(p);
-
-      emit_op2(p, OPCODE_DP3, spot, 0, ureg_negate(VPpli), spot_dir_norm);
-      emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir_norm,W), spot);
-      emit_op2(p, OPCODE_POW, spot, 0, spot, swizzle1(attenuation, W));
-      emit_op2(p, OPCODE_MUL, att, 0, slt, spot);
-
-      release_temp(p, spot);
-      release_temp(p, slt);
-   }
-
-   /* Calculate distance attenuation:
-    */
-   if (p->state->unit[i].light_attenuated) {
-
-      /* 1/d,d,d,1/d */
-      emit_op1(p, OPCODE_RCP, dist, WRITEMASK_YZ, dist); 
-      /* 1,d,d*d,1/d */
-      emit_op2(p, OPCODE_MUL, dist, WRITEMASK_XZ, dist, swizzle1(dist,Y)); 
-      /* 1/dist-atten */
-      emit_op2(p, OPCODE_DP3, dist, 0, attenuation, dist); 
-
-      if (!p->state->unit[i].light_spotcutoff_is_180) {
-	 /* dist-atten */
-	 emit_op1(p, OPCODE_RCP, dist, 0, dist); 
-	 /* spot-atten * dist-atten */
-	 emit_op2(p, OPCODE_MUL, att, 0, dist, att);	
-      } else {
-	 /* dist-atten */
-	 emit_op1(p, OPCODE_RCP, att, 0, dist); 
-      }
-   }
-
-   return att;
-}
-						
-
-
-
-
-/* Need to add some addtional parameters to allow lighting in object
- * space - STATE_SPOT_DIRECTION and STATE_HALF_VECTOR implicitly assume eye
- * space lighting.
- */
-static void build_lighting( struct tnl_program *p )
-{
-   const GLboolean twoside = p->state->light_twoside;
-   const GLboolean separate = p->state->separate_specular;
-   GLuint nr_lights = 0, count = 0;
-   struct ureg normal = get_eye_normal(p);
-   struct ureg lit = get_temp(p);
-   struct ureg dots = get_temp(p);
-   struct ureg _col0 = undef, _col1 = undef;
-   struct ureg _bfc0 = undef, _bfc1 = undef;
-   GLuint i;
-
-   for (i = 0; i < MAX_LIGHTS; i++) 
-      if (p->state->unit[i].light_enabled)
-	 nr_lights++;
-   
-   set_material_flags(p);
-
-   {
-      struct ureg shininess = get_material(p, 0, STATE_SHININESS);
-      emit_op1(p, OPCODE_MOV, dots,  WRITEMASK_W, swizzle1(shininess,X));
-      release_temp(p, shininess);
-
-      _col0 = make_temp(p, get_scenecolor(p, 0));
-      if (separate)
-	 _col1 = make_temp(p, get_identity_param(p));
-      else
-	 _col1 = _col0;
-
-   }
-
-   if (twoside) {
-      struct ureg shininess = get_material(p, 1, STATE_SHININESS);
-      emit_op1(p, OPCODE_MOV, dots, WRITEMASK_Z, 
-	       ureg_negate(swizzle1(shininess,X)));
-      release_temp(p, shininess);
-
-      _bfc0 = make_temp(p, get_scenecolor(p, 1));
-      if (separate)
-	 _bfc1 = make_temp(p, get_identity_param(p));
-      else
-	 _bfc1 = _bfc0;
-   }
-
-
-   /* If no lights, still need to emit the scenecolor.
-    */
-   /* KW: changed to do this always - v1.17 "Fix lighting alpha result"? 
-    */
-   if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
-   {
-      struct ureg res0 = register_output( p, VERT_RESULT_COL0 );
-      emit_op1(p, OPCODE_MOV, res0, 0, _col0);
-
-      if (twoside) {
-	 struct ureg res0 = register_output( p, VERT_RESULT_BFC0 );
-	 emit_op1(p, OPCODE_MOV, res0, 0, _bfc0);
-      }
-   }
-
-   if (separate && (p->state->fragprog_inputs_read & FRAG_BIT_COL1)) {
-
-      struct ureg res1 = register_output( p, VERT_RESULT_COL1 );
-      emit_op1(p, OPCODE_MOV, res1, 0, _col1);
-      
-      if (twoside) {
-	 struct ureg res1 = register_output( p, VERT_RESULT_BFC1 );
-	 emit_op1(p, OPCODE_MOV, res1, 0, _bfc1);
-      }
-   }
-      
-   if (nr_lights == 0) {
-      release_temps(p);
-      return;
-   }
-
-
-   for (i = 0; i < MAX_LIGHTS; i++) {
-      if (p->state->unit[i].light_enabled) {
-	 struct ureg half = undef;
-	 struct ureg att = undef, VPpli = undef;
-	  
-	 count++;
-
-	 if (p->state->unit[i].light_eyepos3_is_zero) {
-	    /* Can used precomputed constants in this case.
-	     * Attenuation never applies to infinite lights.
-	     */
-	    VPpli = register_param3(p, STATE_LIGHT, i, 
-				    STATE_LIGHT_POSITION_NORMALIZED);
-            if (p->state->light_local_viewer) {
-                struct ureg eye_hat = get_eye_position_normalized(p);
-                half = get_temp(p);
-                emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
-                emit_normalize_vec3(p, half, half);
-            } else {
-                half = register_param3(p, STATE_LIGHT, i, STATE_HALF_VECTOR);
-            }
-	 } 
-	 else {
-	    struct ureg Ppli = register_param3(p, STATE_LIGHT, i, 
-					       STATE_POSITION); 
-	    struct ureg V = get_eye_position(p);
-	    struct ureg dist = get_temp(p);
-	    struct ureg tmpPpli = get_temp(p);
-
-	    VPpli = get_temp(p); 
-	    half = get_temp(p);
-
-	    /* In homogeneous object coordinates
-	     */
-	    emit_op1(p, OPCODE_RCP, dist, 0, swizzle1(Ppli, W));
-	    emit_op2(p, OPCODE_MUL, tmpPpli, 0, Ppli, dist);
- 
-	    /* Calulate VPpli vector
-	     */
-	    emit_op2(p, OPCODE_SUB, VPpli, 0, tmpPpli, V); 
-
-	    /* Normalize VPpli.  The dist value also used in
-	     * attenuation below.
-	     */
-	    emit_op2(p, OPCODE_DP3, dist, 0, VPpli, VPpli);
-	    emit_op1(p, OPCODE_RSQ, dist, 0, dist);
-	    emit_op2(p, OPCODE_MUL, VPpli, 0, VPpli, dist);
-
-
-	    /* Calculate  attenuation:
-	     */ 
-	    if (!p->state->unit[i].light_spotcutoff_is_180 ||
-		p->state->unit[i].light_attenuated) {
-	       att = calculate_light_attenuation(p, i, VPpli, dist);
-	    }
-	 
-      
-	    /* Calculate viewer direction, or use infinite viewer:
-	     */
-	    if (p->state->light_local_viewer) {
-	       struct ureg eye_hat = get_eye_position_normalized(p);
-	       emit_op2(p, OPCODE_SUB, half, 0, VPpli, eye_hat);
-	    }
-	    else {
-	       struct ureg z_dir = swizzle(get_identity_param(p),X,Y,W,Z); 
-	       emit_op2(p, OPCODE_ADD, half, 0, VPpli, z_dir);
-	    }
-
-	    emit_normalize_vec3(p, half, half);
-
-	    release_temp(p, dist);
-	    release_temp(p, tmpPpli);
-	 }
-
-	 /* Calculate dot products:
-	  */
-	 emit_op2(p, OPCODE_DP3, dots, WRITEMASK_X, normal, VPpli);
-	 emit_op2(p, OPCODE_DP3, dots, WRITEMASK_Y, normal, half);
-
-	
-	 /* Front face lighting:
-	  */
-	 {
-	    struct ureg ambient = get_lightprod(p, i, 0, STATE_AMBIENT);
-	    struct ureg diffuse = get_lightprod(p, i, 0, STATE_DIFFUSE);
-	    struct ureg specular = get_lightprod(p, i, 0, STATE_SPECULAR);
-	    struct ureg res0, res1;
-	    GLuint mask0, mask1;
-
-	    emit_op1(p, OPCODE_LIT, lit, 0, dots);
-   
-	    if (!is_undef(att)) 
-	       emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
-
-
-	    mask0 = 0;
-	    mask1 = 0;
-	    res0 = _col0;
-	    res1 = _col1;
-	    
-	    if (count == nr_lights) {
-	       if (separate) {
-		  mask0 = WRITEMASK_XYZ;
-		  mask1 = WRITEMASK_XYZ;
-
-		  if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
-		     res0 = register_output( p, VERT_RESULT_COL0 );
-
-		  if (p->state->fragprog_inputs_read & FRAG_BIT_COL1)
-		     res1 = register_output( p, VERT_RESULT_COL1 );
-	       }
-	       else {
-		  mask1 = WRITEMASK_XYZ;
-
-		  if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
-		     res1 = register_output( p, VERT_RESULT_COL0 );
-	       }
-	    } 
-
-	    emit_op3(p, OPCODE_MAD, _col0, 0, swizzle1(lit,X), ambient, _col0);
-	    emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _col0);
-	    emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _col1);
-      
-	    release_temp(p, ambient);
-	    release_temp(p, diffuse);
-	    release_temp(p, specular);
-	 }
-
-	 /* Back face lighting:
-	  */
-	 if (twoside) {
-	    struct ureg ambient = get_lightprod(p, i, 1, STATE_AMBIENT);
-	    struct ureg diffuse = get_lightprod(p, i, 1, STATE_DIFFUSE);
-	    struct ureg specular = get_lightprod(p, i, 1, STATE_SPECULAR);
-	    struct ureg res0, res1;
-	    GLuint mask0, mask1;
-	       
-	    emit_op1(p, OPCODE_LIT, lit, 0, ureg_negate(swizzle(dots,X,Y,W,Z)));
-
-	    if (!is_undef(att)) 
-	       emit_op2(p, OPCODE_MUL, lit, 0, lit, att);
-
-	    mask0 = 0;
-	    mask1 = 0;
-	    res0 = _bfc0;
-	    res1 = _bfc1;
-
-	    if (count == nr_lights) {
-	       if (separate) {
-		  mask0 = WRITEMASK_XYZ;
-		  mask1 = WRITEMASK_XYZ;
-		  if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
-		     res0 = register_output( p, VERT_RESULT_BFC0 );
-
-		  if (p->state->fragprog_inputs_read & FRAG_BIT_COL1)
-		     res1 = register_output( p, VERT_RESULT_BFC1 );
-	       }
-	       else {
-		  mask1 = WRITEMASK_XYZ;
-
-		  if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
-		     res1 = register_output( p, VERT_RESULT_BFC0 );
-	       }
-	    }
-
-	    emit_op3(p, OPCODE_MAD, _bfc0, 0, swizzle1(lit,X), ambient, _bfc0);
-	    emit_op3(p, OPCODE_MAD, res0, mask0, swizzle1(lit,Y), diffuse, _bfc0);
-	    emit_op3(p, OPCODE_MAD, res1, mask1, swizzle1(lit,Z), specular, _bfc1);
-
-	    release_temp(p, ambient);
-	    release_temp(p, diffuse);
-	    release_temp(p, specular);
-	 }
-
-	 release_temp(p, half);
-	 release_temp(p, VPpli);
-	 release_temp(p, att);
-      }
-   }
-
-   release_temps( p );
-}
-
-
-static void build_fog( struct tnl_program *p )
-{
-   struct ureg fog = register_output(p, VERT_RESULT_FOGC);
-   struct ureg input;
-   GLuint useabs = p->state->fog_source_is_depth && p->state->fog_option &&
-		   (p->state->fog_option != FOG_EXP2);
-
-   if (p->state->fog_source_is_depth) {
-      input = swizzle1(get_eye_position(p), Z);
-   }
-   else {
-      input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
-      if (p->state->fog_option &&
-	  p->state->tnl_do_vertex_fog)
-	  input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
-      else
-	  input = register_input(p, VERT_ATTRIB_FOG);
-   }
-
-   if (p->state->fog_option &&
-       p->state->tnl_do_vertex_fog) {
-      struct ureg params = register_param2(p, STATE_INTERNAL,
-					   STATE_FOG_PARAMS_OPTIMIZED);
-      struct ureg tmp = get_temp(p);
-      struct ureg id = get_identity_param(p);
-
-      emit_op1(p, OPCODE_MOV, fog, 0, id);
-
-      if (useabs) {
-	 emit_op1(p, OPCODE_ABS, tmp, 0, input);
-      }
-
-      switch (p->state->fog_option) {
-      case FOG_LINEAR: {
-	 emit_op3(p, OPCODE_MAD, tmp, 0, useabs ? tmp : input,
-			swizzle1(params,X), swizzle1(params,Y));
-	 emit_op2(p, OPCODE_MAX, tmp, 0, tmp, swizzle1(id,X)); /* saturate */
-	 emit_op2(p, OPCODE_MIN, fog, WRITEMASK_X, tmp, swizzle1(id,W));
-	 break;
-      }
-      case FOG_EXP:
-	 emit_op2(p, OPCODE_MUL, tmp, 0, useabs ? tmp : input,
-			swizzle1(params,Z));
-	 emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, ureg_negate(tmp));
-	 break;
-      case FOG_EXP2:
-	 emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W));
-	 emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp);
-	 emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, ureg_negate(tmp));
-	 break;
-      }
-
-      release_temp(p, tmp);
-   }
-   else {
-      /* results = incoming fog coords (compute fog per-fragment later) 
-       *
-       * KW:  Is it really necessary to do anything in this case?
-       */
-      emit_op1(p, useabs ? OPCODE_ABS : OPCODE_MOV, fog, 0, input);
-   }
-}
- 
-static void build_reflect_texgen( struct tnl_program *p,
-				  struct ureg dest,
-				  GLuint writemask )
-{
-   struct ureg normal = get_eye_normal(p);
-   struct ureg eye_hat = get_eye_position_normalized(p);
-   struct ureg tmp = get_temp(p);
-
-   /* n.u */
-   emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); 
-   /* 2n.u */
-   emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); 
-   /* (-2n.u)n + u */
-   emit_op3(p, OPCODE_MAD, dest, writemask, ureg_negate(tmp), normal, eye_hat);
-
-   release_temp(p, tmp);
-}
-
-static void build_sphere_texgen( struct tnl_program *p,
-				 struct ureg dest,
-				 GLuint writemask )
-{
-   struct ureg normal = get_eye_normal(p);
-   struct ureg eye_hat = get_eye_position_normalized(p);
-   struct ureg tmp = get_temp(p);
-   struct ureg half = register_scalar_const(p, .5);
-   struct ureg r = get_temp(p);
-   struct ureg inv_m = get_temp(p);
-   struct ureg id = get_identity_param(p);
-
-   /* Could share the above calculations, but it would be
-    * a fairly odd state for someone to set (both sphere and
-    * reflection active for different texture coordinate
-    * components.  Of course - if two texture units enable
-    * reflect and/or sphere, things start to tilt in favour
-    * of seperating this out:
-    */
-
-   /* n.u */
-   emit_op2(p, OPCODE_DP3, tmp, 0, normal, eye_hat); 
-   /* 2n.u */
-   emit_op2(p, OPCODE_ADD, tmp, 0, tmp, tmp); 
-   /* (-2n.u)n + u */
-   emit_op3(p, OPCODE_MAD, r, 0, ureg_negate(tmp), normal, eye_hat); 
-   /* r + 0,0,1 */
-   emit_op2(p, OPCODE_ADD, tmp, 0, r, swizzle(id,X,Y,W,Z)); 
-   /* rx^2 + ry^2 + (rz+1)^2 */
-   emit_op2(p, OPCODE_DP3, tmp, 0, tmp, tmp); 
-   /* 2/m */
-   emit_op1(p, OPCODE_RSQ, tmp, 0, tmp); 
-   /* 1/m */
-   emit_op2(p, OPCODE_MUL, inv_m, 0, tmp, half); 
-   /* r/m + 1/2 */
-   emit_op3(p, OPCODE_MAD, dest, writemask, r, inv_m, half); 
-	       
-   release_temp(p, tmp);
-   release_temp(p, r);
-   release_temp(p, inv_m);
-}
-
-
-static void build_texture_transform( struct tnl_program *p )
-{
-   GLuint i, j;
-
-   for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
-
-      if (!(p->state->fragprog_inputs_read & (FRAG_BIT_TEX0<<i)))
-	 continue;
-							     
-      if (p->state->unit[i].texgen_enabled || 
-	  p->state->unit[i].texmat_enabled) {
-	 
-	 GLuint texmat_enabled = p->state->unit[i].texmat_enabled;
-	 struct ureg out = register_output(p, VERT_RESULT_TEX0 + i);
-	 struct ureg out_texgen = undef;
-
-	 if (p->state->unit[i].texgen_enabled) {
-	    GLuint copy_mask = 0;
-	    GLuint sphere_mask = 0;
-	    GLuint reflect_mask = 0;
-	    GLuint normal_mask = 0;
-	    GLuint modes[4];
-	 
-	    if (texmat_enabled) 
-	       out_texgen = get_temp(p);
-	    else
-	       out_texgen = out;
-
-	    modes[0] = p->state->unit[i].texgen_mode0;
-	    modes[1] = p->state->unit[i].texgen_mode1;
-	    modes[2] = p->state->unit[i].texgen_mode2;
-	    modes[3] = p->state->unit[i].texgen_mode3;
-
-	    for (j = 0; j < 4; j++) {
-	       switch (modes[j]) {
-	       case TXG_OBJ_LINEAR: {
-		  struct ureg obj = register_input(p, VERT_ATTRIB_POS);
-		  struct ureg plane = 
-		     register_param3(p, STATE_TEXGEN, i,
-				     STATE_TEXGEN_OBJECT_S + j);
-
-		  emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, 
-			   obj, plane );
-		  break;
-	       }
-	       case TXG_EYE_LINEAR: {
-		  struct ureg eye = get_eye_position(p);
-		  struct ureg plane = 
-		     register_param3(p, STATE_TEXGEN, i, 
-				     STATE_TEXGEN_EYE_S + j);
-
-		  emit_op2(p, OPCODE_DP4, out_texgen, WRITEMASK_X << j, 
-			   eye, plane );
-		  break;
-	       }
-	       case TXG_SPHERE_MAP: 
-		  sphere_mask |= WRITEMASK_X << j;
-		  break;
-	       case TXG_REFLECTION_MAP:
-		  reflect_mask |= WRITEMASK_X << j;
-		  break;
-	       case TXG_NORMAL_MAP: 
-		  normal_mask |= WRITEMASK_X << j;
-		  break;
-	       case TXG_NONE:
-		  copy_mask |= WRITEMASK_X << j;
-	       }
-
-	    }
-
-	 
-	    if (sphere_mask) {
-	       build_sphere_texgen(p, out_texgen, sphere_mask);
-	    }
-
-	    if (reflect_mask) {
-	       build_reflect_texgen(p, out_texgen, reflect_mask);
-	    }
-
-	    if (normal_mask) {
-	       struct ureg normal = get_eye_normal(p);
-	       emit_op1(p, OPCODE_MOV, out_texgen, normal_mask, normal );
-	    }
-
-	    if (copy_mask) {
-	       struct ureg in = register_input(p, VERT_ATTRIB_TEX0+i);
-	       emit_op1(p, OPCODE_MOV, out_texgen, copy_mask, in );
-	    }
-	 }
-
-	 if (texmat_enabled) {
-	    struct ureg texmat[4];
-	    struct ureg in = (!is_undef(out_texgen) ? 
-			      out_texgen : 
-			      register_input(p, VERT_ATTRIB_TEX0+i));
-	    if (PREFER_DP4) {
-	       register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3,
-				       0, texmat );
-	       emit_matrix_transform_vec4( p, out, texmat, in );
-	    }
-	    else {
-	       register_matrix_param5( p, STATE_TEXTURE_MATRIX, i, 0, 3,
-				       STATE_MATRIX_TRANSPOSE, texmat );
-	       emit_transpose_matrix_transform_vec4( p, out, texmat, in );
-	    }
-	 }
-
-	 release_temps(p);
-      } 
-      else {
-	 emit_passthrough(p, VERT_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i);
-      }
-   }
-}
-
-
-/* Seems like it could be tighter:
- */
-static void build_pointsize( struct tnl_program *p )
-{
-   struct ureg eye = get_eye_position(p);
-   struct ureg state_size = register_param1(p, STATE_POINT_SIZE);
-   struct ureg state_attenuation = register_param1(p, STATE_POINT_ATTENUATION);
-   struct ureg out = register_output(p, VERT_RESULT_PSIZ);
-   struct ureg ut = get_temp(p);
-
-   /* 1, Z, Z * Z, 1 */      
-   emit_op1(p, OPCODE_MOV, ut, WRITEMASK_XW, swizzle1(get_identity_param(p), W));
-   emit_op1(p, OPCODE_ABS, ut, WRITEMASK_YZ, swizzle1(eye, Z));
-   emit_op2(p, OPCODE_MUL, ut, WRITEMASK_Z, ut, ut);
-
-
-   /* p1 +  p2 * dist + p3 * dist * dist, 0 */
-   emit_op2(p, OPCODE_DP3, ut, WRITEMASK_X, ut, state_attenuation);
-
-   /* 1 / sqrt(factor) */
-   emit_op1(p, OPCODE_RSQ, ut, WRITEMASK_X, ut ); 
-
-   /* ut = pointSize / factor */
-   emit_op2(p, OPCODE_MUL, ut, WRITEMASK_X, ut, state_size); 
-
-   /* Clamp to min/max - state_size.[yz]
-    */
-   emit_op2(p, OPCODE_MAX, ut, WRITEMASK_X, ut, swizzle1(state_size, Y)); 
-   emit_op2(p, OPCODE_MIN, out, 0, swizzle1(ut, X), swizzle1(state_size, Z)); 
-   
-   release_temp(p, ut);
-}
-
-static void build_tnl_program( struct tnl_program *p )
-{  
-   /* Emit the program, starting with modelviewproject:
-    */
-   build_hpos(p);
-
-   /* Lighting calculations:
-    */
-   if (p->state->fragprog_inputs_read & (FRAG_BIT_COL0|FRAG_BIT_COL1)) {
-      if (p->state->light_global_enabled)
-	 build_lighting(p);
-      else {
-	 if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
-	    emit_passthrough(p, VERT_ATTRIB_COLOR0, VERT_RESULT_COL0);
-
-	 if (p->state->fragprog_inputs_read & FRAG_BIT_COL1)
-	    emit_passthrough(p, VERT_ATTRIB_COLOR1, VERT_RESULT_COL1);
-      }
-   }
-
-   if ((p->state->fragprog_inputs_read & FRAG_BIT_FOGC) ||
-       p->state->fog_option != FOG_NONE)
-      build_fog(p);
-
-   if (p->state->fragprog_inputs_read & FRAG_BITS_TEX_ANY)
-      build_texture_transform(p);
-
-   if (p->state->point_attenuated)
-      build_pointsize(p);
-
-   /* Finish up:
-    */
-   emit_op1(p, OPCODE_END, undef, 0, undef);
-
-   /* Disassemble:
-    */
-   if (DISASSEM) {
-      _mesa_printf ("\n");
-   }
-}
-
-
-static void build_new_tnl_program( const struct state_key *key,
-				   struct gl_vertex_program *program,
-				   GLuint max_temps)
-{
-   struct tnl_program p;
-
-   _mesa_memset(&p, 0, sizeof(p));
-   p.state = key;
-   p.program = program;
-   p.eye_position = undef;
-   p.eye_position_normalized = undef;
-   p.eye_normal = undef;
-   p.identity = undef;
-   p.temp_in_use = 0;
-   p.nr_instructions = 16;
-   
-   if (max_temps >= sizeof(int) * 8)
-      p.temp_reserved = 0;
-   else
-      p.temp_reserved = ~((1<<max_temps)-1);
-
-   p.program->Base.Instructions = 
-      _mesa_malloc(sizeof(struct prog_instruction) * p.nr_instructions);
-   p.program->Base.String = 0;
-   p.program->Base.NumInstructions =
-   p.program->Base.NumTemporaries =
-   p.program->Base.NumParameters =
-   p.program->Base.NumAttributes = p.program->Base.NumAddressRegs = 0;
-   p.program->Base.Parameters = _mesa_new_parameter_list();
-   p.program->Base.InputsRead = 0;
-   p.program->Base.OutputsWritten = 0;
-
-   build_tnl_program( &p );
-}
-
-static void *search_cache( struct brw_tnl_cache *cache,
-			   GLuint hash,
-			   const void *key,
-			   GLuint keysize)
-{
-   struct brw_tnl_cache_item *c;
-
-   for (c = cache->items[hash % cache->size]; c; c = c->next) {
-      if (c->hash == hash && memcmp(c->key, key, keysize) == 0)
-	 return c->data;
-   }
-
-   return NULL;
-}
-
-static void rehash( struct brw_tnl_cache *cache )
-{
-   struct brw_tnl_cache_item **items;
-   struct brw_tnl_cache_item *c, *next;
-   GLuint size, i;
-
-   size = cache->size * 3;
-   items = (struct brw_tnl_cache_item**) _mesa_malloc(size * sizeof(*items));
-   _mesa_memset(items, 0, size * sizeof(*items));
-
-   for (i = 0; i < cache->size; i++)
-      for (c = cache->items[i]; c; c = next) {
-	 next = c->next;
-	 c->next = items[c->hash % size];
-	 items[c->hash % size] = c;
-      }
-
-   FREE(cache->items);
-   cache->items = items;
-   cache->size = size;
-}
-
-static void cache_item( struct brw_tnl_cache *cache,
-			GLuint hash,
-			const struct state_key *key,
-			void *data )
-{
-   struct brw_tnl_cache_item *c = MALLOC(sizeof(*c));
-   c->hash = hash;
-
-   c->key = malloc(sizeof(*key));
-   memcpy(c->key, key, sizeof(*key));
-
-   c->data = data;
-
-   if (++cache->n_items > cache->size * 1.5)
-      rehash(cache);
-
-   c->next = cache->items[hash % cache->size];
-   cache->items[hash % cache->size] = c;
-}
-
-
-static GLuint hash_key( struct state_key *key )
-{
-   GLuint *ikey = (GLuint *)key;
-   GLuint hash = 0, i;
-
-   /* I'm sure this can be improved on, but speed is important:
-    */
-   for (i = 0; i < sizeof(*key)/sizeof(GLuint); i++)
-      hash += ikey[i];
-
-   return hash;
-}
-
-static void prepare_tnl_program( struct brw_context *brw )
-{
-   GLcontext *ctx = &brw->intel.ctx;
-   struct state_key key;
-   GLuint hash;
-   struct gl_vertex_program *old = brw->tnl_program;
-
-   /* _NEW_PROGRAM */
-   if (brw->attribs.VertexProgram->_Current) 
-      return;
-      
-   /* Grab all the relevent state and put it in a single structure:
-    */
-   make_state_key(ctx, &key);
-   hash = hash_key(&key);
-
-   /* Look for an already-prepared program for this state:
-    */
-   brw->tnl_program = (struct gl_vertex_program *)
-      search_cache( &brw->tnl_program_cache, hash, &key, sizeof(key) );
-   
-   /* OK, we'll have to build a new one:
-    */
-   if (!brw->tnl_program) {
-      brw->tnl_program = (struct gl_vertex_program *)
-	 ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0); 
-
-      build_new_tnl_program( &key, brw->tnl_program, 
-/* 			     ctx->Const.MaxVertexProgramTemps  */
-			     32
-	 );
-
-      if (ctx->Driver.ProgramStringNotify)
-	 ctx->Driver.ProgramStringNotify( ctx, GL_VERTEX_PROGRAM_ARB, 
-					  &brw->tnl_program->Base );
-
-      cache_item( &brw->tnl_program_cache, 
-		  hash, &key, brw->tnl_program );
-   }
-
-   if (old != brw->tnl_program)
-      brw->state.dirty.brw |= BRW_NEW_TNL_PROGRAM;
-   return;
-}
-
-/* Note: See brw_draw.c - the vertex program must not rely on
- * brw->primitive or brw->reduced_prim.
- */
-const struct brw_tracked_state brw_tnl_vertprog = {
-   .dirty = {
-      .mesa = (_NEW_PROGRAM | 
-	       _NEW_LIGHT | 
-	       _NEW_TRANSFORM | 
-	       _NEW_FOG | 
-	       _NEW_HINT | 
-	       _NEW_POINT | 
-	       _NEW_TEXTURE |
-          _NEW_TEXTURE_MATRIX),
-      .brw = (BRW_NEW_FRAGMENT_PROGRAM | 
-	      BRW_NEW_INPUT_VARYING),
-      .cache = 0
-   },
-   .prepare = prepare_tnl_program
-};
-
-
-
-
 static void prepare_active_vertprog( struct brw_context *brw )
 {
    const struct gl_vertex_program *prev = brw->vertex_program;
 
-   /* NEW_PROGRAM */
-   if (brw->attribs.VertexProgram->_Current) {
-      brw->vertex_program = brw->attribs.VertexProgram->_Current;
-   }
-   else {
-      /* BRW_NEW_TNL_PROGRAM */
-      brw->vertex_program = brw->tnl_program;
-   }
+   brw->vertex_program = brw->attribs.VertexProgram->_Current;
 
    if (brw->vertex_program != prev) 
       brw->state.dirty.brw |= BRW_NEW_VERTEX_PROGRAM;
@@ -1672,37 +52,8 @@ static void prepare_active_vertprog( struct brw_context *brw )
 const struct brw_tracked_state brw_active_vertprog = {
    .dirty = {
       .mesa = _NEW_PROGRAM,
-      .brw = BRW_NEW_TNL_PROGRAM,
+      .brw = 0,
       .cache = 0
    },
    .prepare = prepare_active_vertprog
 };
-
-
-void brw_ProgramCacheInit( GLcontext *ctx )
-{
-   struct brw_context *brw = brw_context(ctx);
-
-   brw->tnl_program_cache.size = 17;
-   brw->tnl_program_cache.n_items = 0;
-   brw->tnl_program_cache.items = (struct brw_tnl_cache_item **)
-      _mesa_calloc(brw->tnl_program_cache.size * 
-		   sizeof(struct brw_tnl_cache_item));
-}
-
-void brw_ProgramCacheDestroy( GLcontext *ctx )
-{
-   struct brw_context *brw = brw_context(ctx);
-   struct brw_tnl_cache_item *c, *next;
-   GLuint i;
-
-   for (i = 0; i < brw->tnl_program_cache.size; i++)
-      for (c = brw->tnl_program_cache.items[i]; c; c = next) {
-	 next = c->next;
-	 FREE(c->key);
-	 FREE(c->data);
-	 FREE(c);
-      }
-
-   FREE(brw->tnl_program_cache.items);
-}
diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c
index 2a03fc59f3..a64e437860 100644
--- a/src/mesa/drivers/dri/i965/brw_vtbl.c
+++ b/src/mesa/drivers/dri/i965/brw_vtbl.c
@@ -51,6 +51,12 @@
 #include "brw_vs.h"
 #include <stdarg.h>
 
+static void
+dri_bo_release(dri_bo **bo)
+{
+   dri_bo_unreference(*bo);
+   *bo = NULL;
+}
 
 /* called from intelDestroyContext()
  */
@@ -58,13 +64,40 @@ static void brw_destroy_context( struct intel_context *intel )
 {
    GLcontext *ctx = &intel->ctx;
    struct brw_context *brw = brw_context(&intel->ctx);
+   int i;
 
    brw_destroy_metaops(brw);
    brw_destroy_state(brw);
    brw_draw_destroy( brw );
 
-   brw_ProgramCacheDestroy( ctx );
    brw_FrameBufferTexDestroy( brw );
+
+   for (i = 0; i < brw->state.nr_draw_regions; i++)
+       intel_region_release(&brw->state.draw_regions[i]);
+   brw->state.nr_draw_regions = 0;
+   intel_region_release(&brw->state.depth_region);
+
+   dri_bo_release(&brw->curbe.curbe_bo);
+   dri_bo_release(&brw->vs.prog_bo);
+   dri_bo_release(&brw->vs.state_bo);
+   dri_bo_release(&brw->gs.prog_bo);
+   dri_bo_release(&brw->gs.state_bo);
+   dri_bo_release(&brw->clip.prog_bo);
+   dri_bo_release(&brw->clip.state_bo);
+   dri_bo_release(&brw->clip.vp_bo);
+   dri_bo_release(&brw->sf.prog_bo);
+   dri_bo_release(&brw->sf.state_bo);
+   dri_bo_release(&brw->sf.vp_bo);
+   for (i = 0; i < BRW_MAX_TEX_UNIT; i++)
+      dri_bo_release(&brw->wm.sdc_bo[i]);
+   dri_bo_release(&brw->wm.bind_bo);
+   for (i = 0; i < BRW_WM_MAX_SURF; i++)
+      dri_bo_release(&brw->wm.surf_bo[i]);
+   dri_bo_release(&brw->wm.prog_bo);
+   dri_bo_release(&brw->wm.state_bo);
+   dri_bo_release(&brw->cc.prog_bo);
+   dri_bo_release(&brw->cc.state_bo);
+   dri_bo_release(&brw->cc.vp_bo);
 }
 
 /* called from intelDrawBuffer()
@@ -87,6 +120,15 @@ static void brw_set_draw_region( struct intel_context *intel,
    brw->state.nr_draw_regions = num_regions;
 }
 
+/* called from intel_batchbuffer_flush and children before sending a
+ * batchbuffer off.
+ */
+static void brw_finish_batch(struct intel_context *intel)
+{
+   struct brw_context *brw = brw_context(&intel->ctx);
+
+   brw_emit_query_end(brw);
+}
 
 /* called from intelFlushBatchLocked
  */
@@ -185,6 +227,7 @@ void brwInitVtbl( struct brw_context *brw )
    brw->intel.vtbl.note_fence = brw_note_fence; 
    brw->intel.vtbl.note_unlock = brw_note_unlock; 
    brw->intel.vtbl.new_batch = brw_new_batch;
+   brw->intel.vtbl.finish_batch = brw_finish_batch;
    brw->intel.vtbl.destroy = brw_destroy_context;
    brw->intel.vtbl.set_draw_region = brw_set_draw_region;
    brw->intel.vtbl.flush_cmd = brw_flush_cmd;
diff --git a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
index e1db31ec08..f12ef47a7d 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_sampler_state.c
@@ -229,6 +229,9 @@ brw_wm_sampler_populate_key(struct brw_context *brw,
 	 struct wm_sampler_entry *entry = &key->sampler[unit];
 	 struct gl_texture_unit *texUnit = &brw->attribs.Texture->Unit[unit];
 	 struct gl_texture_object *texObj = texUnit->_Current;
+	 struct intel_texture_object *intelObj = intel_texture_object(texObj);
+	 struct gl_texture_image *firstImage =
+	    texObj->Image[0][intelObj->firstLevel];
 
 	 entry->wrap_r = texObj->WrapR;
 	 entry->wrap_s = texObj->WrapS;
@@ -244,8 +247,22 @@ brw_wm_sampler_populate_key(struct brw_context *brw,
     entry->comparefunc = texObj->CompareFunc;
 
 	 dri_bo_unreference(brw->wm.sdc_bo[unit]);
-	 brw->wm.sdc_bo[unit] = upload_default_color(brw, texObj->BorderColor);
-
+	 if (firstImage->_BaseFormat == GL_DEPTH_COMPONENT) {
+	    float bordercolor[4] = {
+	       texObj->BorderColor[0],
+	       texObj->BorderColor[0],
+	       texObj->BorderColor[0],
+	       texObj->BorderColor[0]
+	    };
+	    /* GL specs that border color for depth textures is taken from the
+	     * R channel, while the hardware uses A.  Spam R into all the
+	     * channels for safety.
+	     */
+	    brw->wm.sdc_bo[unit] = upload_default_color(brw, bordercolor);
+	 } else {
+	    brw->wm.sdc_bo[unit] = upload_default_color(brw,
+							texObj->BorderColor);
+	 }
 	 key->sampler_count = unit + 1;
       }
    }
@@ -304,7 +321,7 @@ static void upload_wm_samplers( struct brw_context *brw )
 	    continue;
 
 	 dri_bo_emit_reloc(brw->wm.sampler_bo,
-			   I915_GEM_DOMAIN_INSTRUCTION, 0,
+			   I915_GEM_DOMAIN_SAMPLER, 0,
 			   0,
 			   i * sizeof(struct brw_sampler_state) +
 			   offsetof(struct brw_sampler_state, ss2),
diff --git a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
index 47127c0421..3790b50c97 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_surface_state.c
@@ -274,7 +274,6 @@ brw_update_texture_surface( GLcontext *ctx, GLuint unit )
    key.width = firstImage->Width;
    key.height = firstImage->Height;
    key.cpp = intelObj->mt->cpp;
-   key.depth = firstImage->Depth;
    key.tiling = intelObj->mt->region->tiling;
 
    dri_bo_unreference(brw->wm.surf_bo[unit + MAX_DRAW_BUFFERS]);
diff --git a/src/mesa/drivers/dri/intel/intel_buffers.c b/src/mesa/drivers/dri/intel/intel_buffers.c
index defa5b173d..f5eaf765f3 100644
--- a/src/mesa/drivers/dri/intel/intel_buffers.c
+++ b/src/mesa/drivers/dri/intel/intel_buffers.c
@@ -135,6 +135,14 @@ intel_readbuf_region(struct intel_context *intel)
 static void
 intelSetRenderbufferClipRects(struct intel_context *intel)
 {
+   /* If the batch contents require looping over cliprects, flush them before
+    * we go changing which cliprects get referenced when that happens.
+    */
+   if (intel->batch->cliprect_mode == LOOP_CLIPRECTS &&
+       (intel->fboRect.x2 != intel->ctx.DrawBuffer->Width ||
+	intel->fboRect.x2 != intel->ctx.DrawBuffer->Height))
+      intel_batchbuffer_flush(intel->batch);
+
    assert(intel->ctx.DrawBuffer->Width > 0);
    assert(intel->ctx.DrawBuffer->Height > 0);
    intel->fboRect.x1 = 0;
@@ -160,6 +168,12 @@ intelSetFrontClipRects(struct intel_context *intel)
    if (!dPriv)
       return;
 
+   /* If the batch contents require looping over cliprects, flush them before
+    * we go changing which cliprects get referenced when that happens.
+    */
+   if (intel->batch->cliprect_mode == LOOP_CLIPRECTS &&
+       intel->pClipRects != dPriv->pClipRects)
+      intel_batchbuffer_flush(intel->batch);
    intel->numClipRects = dPriv->numClipRects;
    intel->pClipRects = dPriv->pClipRects;
    intel->drawX = dPriv->x;
@@ -183,6 +197,10 @@ intelSetBackClipRects(struct intel_context *intel)
 
    if (intel_fb->pf_active || dPriv->numBackClipRects == 0) {
       /* use the front clip rects */
+      if (intel->batch->cliprect_mode == LOOP_CLIPRECTS &&
+	  intel->pClipRects != dPriv->pClipRects)
+	 intel_batchbuffer_flush(intel->batch);
+
       intel->numClipRects = dPriv->numClipRects;
       intel->pClipRects = dPriv->pClipRects;
       intel->drawX = dPriv->x;
@@ -190,6 +208,10 @@ intelSetBackClipRects(struct intel_context *intel)
    }
    else {
       /* use the back clip rects */
+      if (intel->batch->cliprect_mode == LOOP_CLIPRECTS &&
+	  intel->pClipRects != dPriv->pBackClipRects)
+	 intel_batchbuffer_flush(intel->batch);
+
       intel->numClipRects = dPriv->numBackClipRects;
       intel->pClipRects = dPriv->pBackClipRects;
       intel->drawX = dPriv->backX;
@@ -900,12 +922,6 @@ intel_draw_buffer(GLcontext * ctx, struct gl_framebuffer *fb)
    if (fb->Name)
       intel_validate_paired_depth_stencil(ctx, fb);
 
-   /* If the batch contents require looping over cliprects, flush them before
-    * we go changing which cliprects get referenced when that happens.
-    */
-   if (intel->batch->cliprect_mode == LOOP_CLIPRECTS)
-      intel_batchbuffer_flush(intel->batch);
-
    /*
     * How many color buffers are we drawing into?
     */
diff --git a/src/mesa/drivers/dri/intel/intel_context.c b/src/mesa/drivers/dri/intel/intel_context.c
index 57e574447a..e6c0d3175e 100644
--- a/src/mesa/drivers/dri/intel/intel_context.c
+++ b/src/mesa/drivers/dri/intel/intel_context.c
@@ -68,14 +68,15 @@
 int INTEL_DEBUG = (0);
 #endif
 
-#define need_GL_NV_point_sprite
 #define need_GL_ARB_multisample
+#define need_GL_ARB_occlusion_query
 #define need_GL_ARB_point_parameters
+#define need_GL_ARB_shader_objects
 #define need_GL_ARB_texture_compression
 #define need_GL_ARB_vertex_buffer_object
 #define need_GL_ARB_vertex_program
+#define need_GL_ARB_vertex_shader
 #define need_GL_ARB_window_pos
-#define need_GL_ARB_occlusion_query
 #define need_GL_EXT_blend_color
 #define need_GL_EXT_blend_equation_separate
 #define need_GL_EXT_blend_func_separate
@@ -84,14 +85,13 @@ int INTEL_DEBUG = (0);
 #define need_GL_EXT_fog_coord
 #define need_GL_EXT_framebuffer_object
 #define need_GL_EXT_multi_draw_arrays
+#define need_GL_EXT_point_parameters
 #define need_GL_EXT_secondary_color
-#define need_GL_NV_vertex_program
 #define need_GL_ATI_separate_stencil
-#define need_GL_EXT_point_parameters
+#define need_GL_NV_point_sprite
+#define need_GL_NV_vertex_program
 #define need_GL_VERSION_2_0
 #define need_GL_VERSION_2_1
-#define need_GL_ARB_shader_objects
-#define need_GL_ARB_vertex_shader
 
 #include "extension_helper.h"
 
@@ -344,88 +344,78 @@ intel_viewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
  * i965_dri.
  */
 static const struct dri_extension card_extensions[] = {
-   {"GL_ARB_multisample", GL_ARB_multisample_functions},
-   {"GL_ARB_multitexture", NULL},
-   {"GL_ARB_point_parameters", GL_ARB_point_parameters_functions},
-   {"GL_NV_point_sprite", GL_NV_point_sprite_functions},
-   {"GL_ARB_texture_border_clamp", NULL},
-   {"GL_ARB_texture_compression", GL_ARB_texture_compression_functions},
-   {"GL_ARB_texture_cube_map", NULL},
-   {"GL_ARB_texture_env_add", NULL},
-   {"GL_ARB_texture_env_combine", NULL},
-   {"GL_ARB_texture_env_crossbar", NULL},
-   {"GL_ARB_texture_env_dot3", NULL},
-   {"GL_ARB_texture_mirrored_repeat", NULL},
-   {"GL_ARB_texture_non_power_of_two",   NULL },
-   {"GL_ARB_texture_rectangle", NULL},
-   {"GL_NV_texture_rectangle", NULL},
-   {"GL_EXT_texture_rectangle", NULL},
-   {"GL_ARB_point_parameters", NULL}, 
-   {"GL_ARB_vertex_buffer_object", GL_ARB_vertex_buffer_object_functions},
-   {"GL_ARB_vertex_program", GL_ARB_vertex_program_functions},
-   {"GL_ARB_window_pos", GL_ARB_window_pos_functions},
-   {"GL_EXT_blend_color", GL_EXT_blend_color_functions},
-   {"GL_EXT_blend_equation_separate",
-    GL_EXT_blend_equation_separate_functions},
-   {"GL_EXT_blend_func_separate", GL_EXT_blend_func_separate_functions},
-   {"GL_EXT_blend_minmax", GL_EXT_blend_minmax_functions},
-   {"GL_EXT_blend_logic_op", NULL},
-   {"GL_EXT_blend_subtract", NULL},
-   {"GL_EXT_cull_vertex", GL_EXT_cull_vertex_functions},
-   {"GL_EXT_fog_coord", GL_EXT_fog_coord_functions},
-   {"GL_EXT_multi_draw_arrays", GL_EXT_multi_draw_arrays_functions},
-   {"GL_ATI_separate_stencil", GL_ATI_separate_stencil_functions},
-#if 1                           /* XXX FBO temporary? */
-   {"GL_EXT_packed_depth_stencil", NULL},
-#endif
-   {"GL_EXT_secondary_color", GL_EXT_secondary_color_functions},
-   {"GL_EXT_stencil_wrap", NULL},
-   {"GL_EXT_texture_edge_clamp", NULL},
-   {"GL_EXT_texture_env_combine", NULL},
-   {"GL_EXT_texture_env_dot3", NULL},
-   {"GL_EXT_texture_filter_anisotropic", NULL},
-   {"GL_EXT_texture_lod_bias", NULL},
-   {"GL_3DFX_texture_compression_FXT1", NULL},
-   {"GL_APPLE_client_storage", NULL},
-   {"GL_MESA_pack_invert", NULL},
-   {"GL_MESA_ycbcr_texture", NULL},
-   {"GL_NV_blend_square", NULL},
-   {"GL_NV_vertex_program", GL_NV_vertex_program_functions},
-   {"GL_NV_vertex_program1_1", NULL},
-   { "GL_SGIS_generate_mipmap", NULL },
-   {NULL, NULL}
+   { "GL_ARB_multisample",                GL_ARB_multisample_functions },
+   { "GL_ARB_multitexture",               NULL },
+   { "GL_ARB_point_parameters",           GL_ARB_point_parameters_functions },
+   { "GL_ARB_texture_border_clamp",       NULL },
+   { "GL_ARB_texture_compression",        GL_ARB_texture_compression_functions },
+   { "GL_ARB_texture_cube_map",           NULL },
+   { "GL_ARB_texture_env_add",            NULL },
+   { "GL_ARB_texture_env_combine",        NULL },
+   { "GL_ARB_texture_env_crossbar",       NULL },
+   { "GL_ARB_texture_env_dot3",           NULL },
+   { "GL_ARB_texture_mirrored_repeat",    NULL },
+   { "GL_ARB_texture_rectangle",          NULL },
+   { "GL_ARB_vertex_buffer_object",       GL_ARB_vertex_buffer_object_functions },
+   { "GL_ARB_vertex_program",             GL_ARB_vertex_program_functions },
+   { "GL_ARB_window_pos",                 GL_ARB_window_pos_functions },
+   { "GL_EXT_blend_color",                GL_EXT_blend_color_functions },
+   { "GL_EXT_blend_equation_separate",    GL_EXT_blend_equation_separate_functions },
+   { "GL_EXT_blend_func_separate",        GL_EXT_blend_func_separate_functions },
+   { "GL_EXT_blend_minmax",               GL_EXT_blend_minmax_functions },
+   { "GL_EXT_blend_logic_op",             NULL },
+   { "GL_EXT_blend_subtract",             NULL },
+   { "GL_EXT_cull_vertex",                GL_EXT_cull_vertex_functions },
+   { "GL_EXT_fog_coord",                  GL_EXT_fog_coord_functions },
+   { "GL_EXT_multi_draw_arrays",          GL_EXT_multi_draw_arrays_functions },
+   { "GL_EXT_packed_depth_stencil",       NULL },
+   { "GL_EXT_secondary_color",            GL_EXT_secondary_color_functions },
+   { "GL_EXT_stencil_wrap",               NULL },
+   { "GL_EXT_texture_edge_clamp",         NULL },
+   { "GL_EXT_texture_env_combine",        NULL },
+   { "GL_EXT_texture_env_dot3",           NULL },
+   { "GL_EXT_texture_filter_anisotropic", NULL },
+   { "GL_EXT_texture_lod_bias",           NULL },
+   { "GL_3DFX_texture_compression_FXT1",  NULL },
+   { "GL_APPLE_client_storage",           NULL },
+   { "GL_MESA_pack_invert",               NULL },
+   { "GL_MESA_ycbcr_texture",             NULL },
+   { "GL_NV_blend_square",                NULL },
+   { "GL_NV_point_sprite",                GL_NV_point_sprite_functions },
+   { "GL_NV_vertex_program",              GL_NV_vertex_program_functions },
+   { "GL_NV_vertex_program1_1",           NULL },
+   { "GL_SGIS_generate_mipmap",           NULL },
+   { NULL, NULL }
 };
 
 static const struct dri_extension brw_extensions[] = {
-   { "GL_ARB_shading_language_100",       GL_VERSION_2_0_functions},
-   { "GL_ARB_shading_language_120",       GL_VERSION_2_1_functions},
-   { "GL_ARB_shader_objects",             GL_ARB_shader_objects_functions},
-   { "GL_ARB_vertex_shader",              GL_ARB_vertex_shader_functions},
-   { "GL_ARB_point_sprite", 		  NULL},
-   { "GL_ARB_fragment_shader",            NULL },
-   { "GL_ARB_draw_buffers",               NULL },
    { "GL_ARB_depth_texture",              NULL },
+   { "GL_ARB_draw_buffers",               NULL },
    { "GL_ARB_fragment_program",           NULL },
+   { "GL_ARB_fragment_program_shadow",    NULL },
+   { "GL_ARB_fragment_shader",            NULL },
+   { "GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions },
+   { "GL_ARB_point_sprite", 		  NULL },
+   { "GL_ARB_shader_objects",             GL_ARB_shader_objects_functions },
+   { "GL_ARB_shading_language_100",       GL_VERSION_2_0_functions },
+   { "GL_ARB_shading_language_120",       GL_VERSION_2_1_functions },
    { "GL_ARB_shadow",                     NULL },
+   { "GL_ARB_texture_non_power_of_two",   NULL },
+   { "GL_ARB_vertex_shader",              GL_ARB_vertex_shader_functions },
    { "GL_EXT_shadow_funcs",               NULL },
-   { "GL_ARB_fragment_program_shadow",    NULL },
-   /* ARB extn won't work if not enabled */
-   { "GL_SGIX_depth_texture",             NULL },
-   { "GL_EXT_texture_sRGB",		  NULL},
+   { "GL_EXT_texture_sRGB",		  NULL },
+   { "GL_ATI_separate_stencil",           GL_ATI_separate_stencil_functions },
    { NULL,                                NULL }
 };
 
-#ifdef I915_MMIO_READ
-static const struct dri_extension arb_oc_extensions[] = {
-   {"GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions},
-   {NULL, NULL}
+static const struct dri_extension arb_oq_extensions[] = {
+   { NULL, NULL }
 };
-#endif
 
 static const struct dri_extension ttm_extensions[] = {
-   {"GL_EXT_framebuffer_object", GL_EXT_framebuffer_object_functions},
-   {"GL_ARB_pixel_buffer_object", NULL},
-   {NULL, NULL}
+   { "GL_ARB_pixel_buffer_object",        NULL },
+   { "GL_EXT_framebuffer_object",         GL_EXT_framebuffer_object_functions },
+   { NULL, NULL }
 };
 
 /**
@@ -445,13 +435,6 @@ void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging)
    if (intel == NULL || intel->ttm)
       driInitExtensions(ctx, ttm_extensions, GL_FALSE);
 
-#ifdef I915_MMIO_READ
-   if (intel == NULL || 
-       (IS_965(intel->intelScreen->deviceID) && 
-	intel->intelScreen->drmMinor >= 8))
-      driInitExtensions(ctx, arb_oc_extensions, GL_FALSE);
-#endif
-
    if (intel == NULL || IS_965(intel->intelScreen->deviceID))
       driInitExtensions(ctx, brw_extensions, GL_FALSE);
 }
@@ -548,39 +531,6 @@ intelFinish(GLcontext * ctx)
    }
 }
 
-#ifdef I915_MMIO_READ
-static void
-intelBeginQuery(GLcontext *ctx, struct gl_query_object *q)
-{
-	struct intel_context *intel = intel_context( ctx );
-	struct drm_i915_mmio io = {
-		.read_write = I915_MMIO_READ,
-		.reg = MMIO_REGS_PS_DEPTH_COUNT,
-		.data = &q->Result 
-	};
-	intel->stats_wm++;
-	intelFinish(&intel->ctx);
-	drmCommandWrite(intel->driFd, DRM_I915_MMIO, &io, sizeof(io));
-}
-
-static void
-intelEndQuery(GLcontext *ctx, struct gl_query_object *q)
-{
-	struct intel_context *intel = intel_context( ctx );
-	GLuint64EXT tmp;	
-	struct drm_i915_mmio io = {
-		.read_write = I915_MMIO_READ,
-		.reg = MMIO_REGS_PS_DEPTH_COUNT,
-		.data = &tmp
-	};
-	intelFinish(&intel->ctx);
-	drmCommandWrite(intel->driFd, DRM_I915_MMIO, &io, sizeof(io));
-	q->Result = tmp - q->Result;
-	q->Ready = GL_TRUE;
-	intel->stats_wm--;
-}
-#endif
-
 void
 intelInitDriverFunctions(struct dd_function_table *functions)
 {
@@ -597,11 +547,6 @@ intelInitDriverFunctions(struct dd_function_table *functions)
    functions->CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D;
    functions->CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D;
 
-#ifdef I915_MMIO_READ
-   functions->BeginQuery = intelBeginQuery;
-   functions->EndQuery = intelEndQuery;
-#endif
-
    intelInitTextureFuncs(functions);
    intelInitStateFuncs(functions);
    intelInitBufferFuncs(functions);
@@ -810,7 +755,12 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
       intel->Fallback = 0;      /* don't call _swrast_Flush later */
 
       intel_batchbuffer_free(intel->batch);
+      intel->batch = NULL;
+
       free(intel->prim.vb);
+      intel->prim.vb = NULL;
+      dri_bo_unreference(intel->prim.vb_bo);
+      intel->prim.vb_bo = NULL;
 
       if (release_texture_heaps) {
          /* This share group is about to go away, free our private
@@ -820,6 +770,13 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
             fprintf(stderr, "do something to free texture heaps\n");
       }
 
+      intel_region_release(&intel->front_region);
+      intel_region_release(&intel->back_region);
+      intel_region_release(&intel->third_region);
+      intel_region_release(&intel->depth_region);
+
+      driDestroyOptionCache(&intel->optionCache);
+
       /* free the Mesa context */
       _mesa_free_context_data(&intel->ctx);
    }
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index 4af4cb9c96..554159ac44 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -226,7 +226,6 @@ struct intel_context
    GLenum reduced_primitive;
    GLuint vertex_size;
    GLubyte *verts;              /* points to tnl->clipspace.vertex_buf */
-   struct intel_region *draw_region;
 
    /* Fallback rasterization functions 
     */
diff --git a/src/mesa/drivers/dri/intel/intel_pixel.c b/src/mesa/drivers/dri/intel/intel_pixel.c
index f39fac13cf..b267ffd890 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel.c
@@ -181,9 +181,9 @@ intelInitPixelFuncs(struct dd_function_table *functions)
    if (!getenv("INTEL_NO_BLIT")) {
       functions->Bitmap = intelBitmap;
       functions->CopyPixels = intelCopyPixels;
+      functions->DrawPixels = intelDrawPixels;
 #ifdef I915
       functions->ReadPixels = intelReadPixels;
-      functions->DrawPixels = intelDrawPixels;
 #endif
    }
 }
diff --git a/src/mesa/drivers/dri/intel/intel_pixel_draw.c b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
index be213e7b96..b60dad7460 100644
--- a/src/mesa/drivers/dri/intel/intel_pixel_draw.c
+++ b/src/mesa/drivers/dri/intel/intel_pixel_draw.c
@@ -51,7 +51,6 @@
 #include "intel_regions.h"
 #include "intel_pixel.h"
 #include "intel_buffer_objects.h"
-#include "intel_tris.h"
 
 
 static GLboolean
diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h
index 96af7e1a03..c21f408093 100644
--- a/src/mesa/drivers/dri/intel/intel_reg.h
+++ b/src/mesa/drivers/dri/intel/intel_reg.h
@@ -45,6 +45,25 @@
 #define I1_LOAD_S(n)                      (1<<(4+n))
 
 /** @{
+ *
+ * PIPE_CONTROL operation, a combination MI_FLUSH and register write with
+ * additional flushing control.
+ */
+#define _3DSTATE_PIPE_CONTROL		(CMD_3D | (3 << 27) | (2 << 24) | 2)
+#define PIPE_CONTROL_NO_WRITE		(0 << 14)
+#define PIPE_CONTROL_WRITE_IMMEDIATE	(1 << 14)
+#define PIPE_CONTROL_WRITE_DEPTH_COUNT	(2 << 14)
+#define PIPE_CONTROL_WRITE_TIMESTAMP	(3 << 14)
+#define PIPE_CONTROL_DEPTH_STALL	(1 << 13)
+#define PIPE_CONTROL_WRITE_FLUSH	(1 << 12)
+#define PIPE_CONTROL_INSTRUCTION_FLUSH	(1 << 11)
+#define PIPE_CONTROL_INTERRUPT_ENABLE	(1 << 8)
+#define PIPE_CONTROL_PPGTT_WRITE	(0 << 2)
+#define PIPE_CONTROL_GLOBAL_GTT_WRITE	(1 << 2)
+
+/** @} */
+
+/** @{
  * 915 definitions
  */
 #define S0_VB_OFFSET_MASK		0xffffffc
diff --git a/src/mesa/drivers/dri/intel/intel_regions.c b/src/mesa/drivers/dri/intel/intel_regions.c
index cb0f4ba083..45faf64c71 100644
--- a/src/mesa/drivers/dri/intel/intel_regions.c
+++ b/src/mesa/drivers/dri/intel/intel_regions.c
@@ -478,6 +478,11 @@ intel_recreate_static(struct intel_context *intel,
    region->pitch = intelScreen->pitch;
    region->height = intelScreen->height;     /* needed? */
 
+   if (region->buffer != NULL) {
+      dri_bo_unreference(region->buffer);
+      region->buffer = NULL;
+   }
+
    if (intel->ttm) {
       assert(region_desc->bo_handle != -1);
       region->buffer = intel_bo_gem_create_from_name(intel->bufmgr,
@@ -486,6 +491,11 @@ intel_recreate_static(struct intel_context *intel,
 
       intel_set_region_tiling_gem(intel, region, region_desc->bo_handle);
    } else {
+      if (region->classic_map != NULL) {
+	 drmUnmap(region->classic_map,
+		  region->pitch * region->cpp * region->height);
+	 region->classic_map = NULL;
+      }
       ret = drmMap(intel->driFd, region_desc->handle,
 		   region->pitch * region->cpp * region->height,
 		   &region->classic_map);
diff --git a/src/mesa/drivers/dri/intel/intel_tex_copy.c b/src/mesa/drivers/dri/intel/intel_tex_copy.c
index a7b88b39c0..f4cb4a781c 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_copy.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_copy.c
@@ -60,7 +60,7 @@ get_teximage_source(struct intel_context *intel, GLenum internalFormat)
 
    switch (internalFormat) {
    case GL_DEPTH_COMPONENT:
-   case GL_DEPTH_COMPONENT16_ARB:
+   case GL_DEPTH_COMPONENT16:
       irb = intel_get_renderbuffer(intel->ctx.ReadBuffer, BUFFER_DEPTH);
       if (irb && irb->region && irb->region->cpp == 2)
          return irb->region;
diff --git a/src/mesa/drivers/dri/intel/intel_tex_validate.c b/src/mesa/drivers/dri/intel/intel_tex_validate.c
index 3dae738ac2..820683d42e 100644
--- a/src/mesa/drivers/dri/intel/intel_tex_validate.c
+++ b/src/mesa/drivers/dri/intel/intel_tex_validate.c
@@ -141,10 +141,7 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
 
    /* Fallback case:
     */
-   if (firstImage->base.Border ||
-       ((firstImage->base._BaseFormat == GL_DEPTH_COMPONENT) &&
-        ((tObj->WrapS == GL_CLAMP_TO_BORDER) ||
-         (tObj->WrapT == GL_CLAMP_TO_BORDER)))) {
+   if (firstImage->base.Border) {
       if (intelObj->mt) {
          intel_miptree_release(intel, &intelObj->mt);
       }
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index ee4a69dce3..37436275e3 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -138,7 +138,6 @@ const struct dri_extension card_extensions[] = {
   {"GL_NV_blend_square",		NULL},
   {"GL_NV_vertex_program",		GL_NV_vertex_program_functions},
   {"GL_SGIS_generate_mipmap",		NULL},
-  {"GL_SGIX_depth_texture",		NULL},
   {NULL,				NULL}
   /* *INDENT-ON* */
 };
diff --git a/src/mesa/glapi/glapi.c b/src/mesa/glapi/glapi.c
index 4b23b42223..9b5144a88b 100644
--- a/src/mesa/glapi/glapi.c
+++ b/src/mesa/glapi/glapi.c
@@ -294,7 +294,31 @@ _glapi_get_context(void)
 #endif
 }
 
+#ifdef USE_X86_ASM
 
+#if defined( GLX_USE_TLS )
+extern       GLubyte gl_dispatch_functions_start[];
+extern       GLubyte gl_dispatch_functions_end[];
+#else
+extern const GLubyte gl_dispatch_functions_start[];
+#endif
+
+#endif /* USE_X86_ASM */
+
+
+#if defined(USE_X64_64_ASM) && defined(GLX_USE_TLS)
+# define DISPATCH_FUNCTION_SIZE  16
+#elif defined(USE_X86_ASM)
+# if defined(THREADS) && !defined(GLX_USE_TLS)
+#  define DISPATCH_FUNCTION_SIZE  32
+# else
+#  define DISPATCH_FUNCTION_SIZE  16
+# endif
+#endif
+
+#if !defined(DISPATCH_FUNCTION_SIZE) && !defined(XFree86Server) && !defined(XGLServer)
+# define NEED_FUNCTION_POINTER
+#endif
 
 #if defined(PTHREADS) || defined(GLX_USE_TLS)
 /**
diff --git a/src/mesa/main/depthstencil.c b/src/mesa/main/depthstencil.c
index fb54d6184d..9d208e2997 100644
--- a/src/mesa/main/depthstencil.c
+++ b/src/mesa/main/depthstencil.c
@@ -282,8 +282,8 @@ _mesa_new_z24_renderbuffer_wrapper(GLcontext *ctx,
    z24rb->RefCount = 1;
    z24rb->Width = dsrb->Width;
    z24rb->Height = dsrb->Height;
-   z24rb->InternalFormat = GL_DEPTH_COMPONENT24_ARB;
-   z24rb->_ActualFormat = GL_DEPTH_COMPONENT24_ARB;
+   z24rb->InternalFormat = GL_DEPTH_COMPONENT24;
+   z24rb->_ActualFormat = GL_DEPTH_COMPONENT24;
    z24rb->_BaseFormat = GL_DEPTH_COMPONENT;
    z24rb->DataType = GL_UNSIGNED_INT;
    z24rb->DepthBits = 24;
diff --git a/src/mesa/main/descrip.mms b/src/mesa/main/descrip.mms
index 3ef215f47f..e49ec65d42 100644
--- a/src/mesa/main/descrip.mms
+++ b/src/mesa/main/descrip.mms
@@ -1,6 +1,6 @@
 # Makefile for core library for VMS
 # contributed by Jouk Jansen  joukj@hrem.nano.tudelft.nl
-# Last revision : 2 October 2007
+# Last revision : 29 September 2008
 
 .first
 	define gl [---.include.gl]
@@ -21,6 +21,7 @@ CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ie
 
 SOURCES =accum.c \
 	api_arrayelt.c \
+	api_exec.c \
 	api_loopback.c \
 	api_noop.c \
 	api_validate.c \
@@ -29,6 +30,7 @@ SOURCES =accum.c \
 	blend.c \
 	bufferobj.c \
 	buffers.c \
+	clear.c \
 	clip.c \
 	colortab.c \
 	context.c \
@@ -46,6 +48,7 @@ SOURCES =accum.c \
 	extensions.c \
 	fbobject.c \
 	feedback.c \
+	ffvertex_prog.c \
 	fog.c \
 	framebuffer.c \
 	get.c \
@@ -60,22 +63,29 @@ SOURCES =accum.c \
 	matrix.c \
 	mipmap.c \
 	mm.c \
+	multisample.c \
 	pixel.c \
+	pixelstore.c \
 	points.c \
 	polygon.c \
 	rastpos.c \
 	rbadaptors.c \
+	readpix.c \
 	renderbuffer.c \
+	scissor.c \
 	shaders.c \
 	state.c \
 	stencil.c \
 	texcompress.c \
 	texcompress_fxt1.c \
 	texcompress_s3tc.c \
+	texenv.c \
 	texenvprogram.c \
 	texformat.c \
+	texgen.c \
 	teximage.c \
 	texobj.c \
+	texparam.c \
 	texrender.c \
 	texstate.c \
 	texstore.c \
@@ -86,6 +96,7 @@ SOURCES =accum.c \
 
 OBJECTS=accum.obj,\
 api_arrayelt.obj,\
+api_exec.obj,\
 api_loopback.obj,\
 api_noop.obj,\
 api_validate.obj,\
@@ -94,6 +105,7 @@ attrib.obj,\
 blend.obj,\
 bufferobj.obj,\
 buffers.obj,\
+clear.obj,\
 clip.obj,\
 colortab.obj,\
 context.obj,\
@@ -111,6 +123,7 @@ execmem.obj,\
 extensions.obj,\
 fbobject.obj,\
 feedback.obj,\
+ffvertex_prog.obj,\
 fog.obj,\
 framebuffer.obj,\
 get.obj,\
@@ -125,21 +138,28 @@ lines.obj,\
 matrix.obj,\
 mipmap.obj,\
 mm.obj,\
+multisample.obj,\
 pixel.obj,\
+pixelstore.obj,\
 points.obj,\
 polygon.obj,\
 rastpos.obj,\
+readpix.obj,\
 renderbuffer.obj,\
+scissor.obj,\
 shaders.obj,\
 state.obj,\
 stencil.obj,\
 texcompress.obj,\
 texcompress_fxt1.obj,\
 texcompress_s3tc.obj,\
+texenv.obj,\
 texenvprogram.obj,\
 texformat.obj,\
+texgen.obj,\
 teximage.obj,\
 texobj.obj,\
+texparam.obj,\
 texrender.obj,\
 texstate.obj,\
 texstore.obj,\
@@ -226,3 +246,13 @@ vtxfmt.obj : vtxfmt.c
 shaders.obj : shaders.c
 queryobj.obj : queryobj.c
 rbadaptors.obj : rbadaptors.c
+clear.obj : clear.c
+multisample.obj : multisample.c
+scissor.obj : scissor.c
+texenv.obj : texenv.c
+texgen.obj : texgen.c
+texparam.obj : texparam.c
+readpix.obj : readpix.c
+ffvertex_prog.obj : ffvertex_prog.c
+api_exec.obj : api_exec.c
+pixelstore.obj : pixelstore.c
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index 8dfbfeb62e..95bf1165f4 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -165,7 +165,7 @@ static const struct {
    { OFF, "GL_SGIS_texture_border_clamp",      F(ARB_texture_border_clamp) },
    { ON,  "GL_SGIS_texture_edge_clamp",        F(SGIS_texture_edge_clamp) },
    { ON,  "GL_SGIS_texture_lod",               F(SGIS_texture_lod) },
-   { OFF, "GL_SGIX_depth_texture",             F(SGIX_depth_texture) },
+   { OFF, "GL_SGIX_depth_texture",             F(ARB_depth_texture) },
    { OFF, "GL_SGIX_shadow",                    F(SGIX_shadow) },
    { OFF, "GL_SGIX_shadow_ambient",            F(SGIX_shadow_ambient) },
    { OFF, "GL_SUN_multi_draw_arrays",          F(EXT_multi_draw_arrays) },
@@ -292,7 +292,6 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
    ctx->Extensions.SGI_texture_color_table = GL_TRUE;
    ctx->Extensions.SGIS_generate_mipmap = GL_TRUE;
    ctx->Extensions.SGIS_texture_edge_clamp = GL_TRUE;
-   ctx->Extensions.SGIX_depth_texture = GL_TRUE;
    ctx->Extensions.SGIX_shadow = GL_TRUE;
    ctx->Extensions.SGIX_shadow_ambient = GL_TRUE;
 #if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program
diff --git a/src/mesa/main/ffvertex_prog.c b/src/mesa/main/ffvertex_prog.c
index 787672be9f..308b4ef711 100644
--- a/src/mesa/main/ffvertex_prog.c
+++ b/src/mesa/main/ffvertex_prog.c
@@ -178,12 +178,12 @@ static GLboolean check_active_shininess( GLcontext *ctx,
 
 
 
-static struct state_key *make_state_key( GLcontext *ctx )
+static void make_state_key( GLcontext *ctx, struct state_key *key )
 {
    const struct gl_fragment_program *fp;
-   struct state_key *key = CALLOC_STRUCT(state_key);
    GLuint i;
 
+   memset(key, 0, sizeof(struct state_key));
    fp = ctx->FragmentProgram._Current;
 
    /* This now relies on texenvprogram.c being active:
@@ -301,8 +301,6 @@ static struct state_key *make_state_key( GLcontext *ctx )
 			      texUnit->GenModeQ );
       }
    }
-   
-   return key;
 }
 
 
@@ -1714,16 +1712,16 @@ struct gl_vertex_program *
 _mesa_get_fixed_func_vertex_program(GLcontext *ctx)
 {
    struct gl_vertex_program *prog;
-   struct state_key *key;
+   struct state_key key;
 
    /* Grab all the relevent state and put it in a single structure:
     */
-   key = make_state_key(ctx);
+   make_state_key(ctx, &key);
 
    /* Look for an already-prepared program for this state:
     */
    prog = (struct gl_vertex_program *)
-      _mesa_search_program_cache(ctx->VertexProgram.Cache, key, sizeof(*key));
+      _mesa_search_program_cache(ctx->VertexProgram.Cache, &key, sizeof(key));
    
    if (!prog) {
       /* OK, we'll have to build a new one */
@@ -1735,7 +1733,7 @@ _mesa_get_fixed_func_vertex_program(GLcontext *ctx)
       if (!prog)
          return NULL;
 
-      create_new_program( key, prog,
+      create_new_program( &key, prog,
                           ctx->Const.VertexProgram.MaxTemps );
 
 #if 0
@@ -1744,10 +1742,8 @@ _mesa_get_fixed_func_vertex_program(GLcontext *ctx)
                                           &prog->Base );
 #endif
       _mesa_program_cache_insert(ctx, ctx->VertexProgram.Cache,
-                                 key, sizeof(*key), &prog->Base);
+                                 &key, sizeof(key), &prog->Base);
    }
 
-   _mesa_free(key);
-
    return prog;
 }
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 8bf6858c1e..052da2c18e 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -2628,7 +2628,6 @@ struct gl_extensions
    GLboolean SGIS_generate_mipmap;
    GLboolean SGIS_texture_edge_clamp;
    GLboolean SGIS_texture_lod;
-   GLboolean SGIX_depth_texture;
    GLboolean SGIX_shadow;
    GLboolean SGIX_shadow_ambient; /* or GL_ARB_shadow_ambient */
    GLboolean TDFX_texture_compression_FXT1;
diff --git a/src/mesa/main/queryobj.c b/src/mesa/main/queryobj.c
index 2d06030030..554e0b0d18 100644
--- a/src/mesa/main/queryobj.c
+++ b/src/mesa/main/queryobj.c
@@ -90,9 +90,8 @@ _mesa_wait_query(GLcontext *ctx, struct gl_query_object *q)
 
 
 /**
- * Delete an occlusion query object.
+ * Delete a query object.  Called via ctx->Driver.DeleteQuery().
  * Not removed from hash table here.
- * XXX maybe add Delete() method to gl_query_object class and call that instead
  */
 void
 _mesa_delete_query(GLcontext *ctx, struct gl_query_object *q)
@@ -546,6 +545,6 @@ delete_queryobj_cb(GLuint id, void *data, void *userData)
 void
 _mesa_free_query_data(GLcontext *ctx)
 {
-   _mesa_HashDeleteAll(ctx->Query.QueryObjects, delete_queryobj_cb, NULL);
+   _mesa_HashDeleteAll(ctx->Query.QueryObjects, delete_queryobj_cb, ctx);
    _mesa_DeleteHashTable(ctx->Query.QueryObjects);
 }
diff --git a/src/mesa/main/shaders.c b/src/mesa/main/shaders.c
index f0db0d2a81..aeb5d4ca2a 100644
--- a/src/mesa/main/shaders.c
+++ b/src/mesa/main/shaders.c
@@ -233,10 +233,18 @@ _mesa_GetObjectParameterivARB(GLhandleARB object, GLenum pname, GLint *params)
    GET_CURRENT_CONTEXT(ctx);
    /* Implement in terms of GetProgramiv, GetShaderiv */
    if (ctx->Driver.IsProgram(ctx, object)) {
-      ctx->Driver.GetProgramiv(ctx, object, pname, params);
+      if (pname == GL_OBJECT_TYPE_ARB) {
+	 *params = GL_PROGRAM_OBJECT_ARB;
+      } else {
+	 ctx->Driver.GetProgramiv(ctx, object, pname, params);
+      }
    }
    else if (ctx->Driver.IsShader(ctx, object)) {
-      ctx->Driver.GetShaderiv(ctx, object, pname, params);
+      if (pname == GL_OBJECT_TYPE_ARB) {
+	 *params = GL_SHADER_OBJECT_ARB;
+      } else {
+	 ctx->Driver.GetShaderiv(ctx, object, pname, params);
+      }
    }
    else {
       _mesa_error(ctx, GL_INVALID_OPERATION, "glGetObjectParameterivARB");
diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c
index 7e44310d8d..2f0e7cc368 100644
--- a/src/mesa/main/state.c
+++ b/src/mesa/main/state.c
@@ -258,12 +258,6 @@ update_program(GLcontext *ctx)
       }
    }
 
-   if (ctx->VertexProgram._Current)
-      assert(ctx->VertexProgram._Current->Base.Parameters);
-   if (ctx->FragmentProgram._Current)
-      assert(ctx->FragmentProgram._Current->Base.Parameters);
-
-
    /* XXX: get rid of _Active flag.
     */
 #if 1
diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c
index 60f36c4a87..ce2772c299 100644
--- a/src/mesa/main/texformat.c
+++ b/src/mesa/main/texformat.c
@@ -1420,14 +1420,13 @@ _mesa_choose_tex_format( GLcontext *ctx, GLint internalFormat,
          ; /* fallthrough */
    }
 
-   if (ctx->Extensions.SGIX_depth_texture ||
-       ctx->Extensions.ARB_depth_texture) {
+   if (ctx->Extensions.ARB_depth_texture) {
       switch (internalFormat) {
          case GL_DEPTH_COMPONENT:
-         case GL_DEPTH_COMPONENT24_SGIX:
-         case GL_DEPTH_COMPONENT32_SGIX:
+         case GL_DEPTH_COMPONENT24:
+         case GL_DEPTH_COMPONENT32:
             return &_mesa_texformat_z32;
-         case GL_DEPTH_COMPONENT16_SGIX:
+         case GL_DEPTH_COMPONENT16:
             return &_mesa_texformat_z16;
          default:
             ; /* fallthrough */
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index 91c41f38c3..bf23a1f290 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -241,13 +241,12 @@ _mesa_base_tex_format( GLcontext *ctx, GLint internalFormat )
       }
    }
 
-   if (ctx->Extensions.SGIX_depth_texture ||
-       ctx->Extensions.ARB_depth_texture) {
+   if (ctx->Extensions.ARB_depth_texture) {
       switch (internalFormat) {
          case GL_DEPTH_COMPONENT:
-         case GL_DEPTH_COMPONENT16_SGIX:
-         case GL_DEPTH_COMPONENT24_SGIX:
-         case GL_DEPTH_COMPONENT32_SGIX:
+         case GL_DEPTH_COMPONENT16:
+         case GL_DEPTH_COMPONENT24:
+         case GL_DEPTH_COMPONENT32:
             return GL_DEPTH_COMPONENT;
          default:
             ; /* fallthrough */
@@ -526,9 +525,9 @@ static GLboolean
 is_depth_format(GLenum format)
 {
    switch (format) {
-      case GL_DEPTH_COMPONENT16_ARB:
-      case GL_DEPTH_COMPONENT24_ARB:
-      case GL_DEPTH_COMPONENT32_ARB:
+      case GL_DEPTH_COMPONENT16:
+      case GL_DEPTH_COMPONENT24:
+      case GL_DEPTH_COMPONENT32:
       case GL_DEPTH_COMPONENT:
          return GL_TRUE;
       default:
@@ -2308,8 +2307,7 @@ _mesa_GetTexImage( GLenum target, GLint level, GLenum format,
       return;
    }
 
-   if (!ctx->Extensions.SGIX_depth_texture &&
-       !ctx->Extensions.ARB_depth_texture && is_depth_format(format)) {
+   if (!ctx->Extensions.ARB_depth_texture && is_depth_format(format)) {
       _mesa_error(ctx, GL_INVALID_ENUM, "glGetTexImage(format)");
       return;
    }
diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c
index 664adadfb9..a9e752a637 100644
--- a/src/mesa/main/texparam.c
+++ b/src/mesa/main/texparam.c
@@ -642,8 +642,7 @@ _mesa_GetTexLevelParameteriv( GLenum target, GLint level,
             *params = 0;
          break;
       case GL_TEXTURE_DEPTH_SIZE_ARB:
-         if (ctx->Extensions.SGIX_depth_texture ||
-             ctx->Extensions.ARB_depth_texture)
+         if (ctx->Extensions.ARB_depth_texture)
             *params = img->TexFormat->DepthBits;
          else
             _mesa_error(ctx, GL_INVALID_ENUM,
@@ -903,9 +902,9 @@ _mesa_GetTexParameterfv( GLenum target, GLenum pname, GLfloat *params )
 #ifdef FEATURE_OES_draw_texture
       case GL_TEXTURE_CROP_RECT_OES:
          params[0] = obj->CropRect[0];
-         params[0] = obj->CropRect[1];
-         params[0] = obj->CropRect[2];
-         params[0] = obj->CropRect[3];
+         params[1] = obj->CropRect[1];
+         params[2] = obj->CropRect[2];
+         params[3] = obj->CropRect[3];
          break;
 #endif
       default:
@@ -1053,9 +1052,9 @@ _mesa_GetTexParameteriv( GLenum target, GLenum pname, GLint *params )
 #ifdef FEATURE_OES_draw_texture
       case GL_TEXTURE_CROP_RECT_OES:
          params[0] = obj->CropRect[0];
-         params[0] = obj->CropRect[1];
-         params[0] = obj->CropRect[2];
-         params[0] = obj->CropRect[3];
+         params[1] = obj->CropRect[1];
+         params[2] = obj->CropRect[2];
+         params[3] = obj->CropRect[3];
          break;
 #endif
       default:
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c
index 448251b223..536404bf97 100644
--- a/src/mesa/shader/arbprogparse.c
+++ b/src/mesa/shader/arbprogparse.c
@@ -3895,6 +3895,9 @@ _mesa_parse_arb_fragment_program(GLcontext* ctx, GLenum target,
    program->FogOption          = ap.FogOption;
    program->UsesKill          = ap.UsesKill;
 
+   if (program->FogOption)
+      program->Base.InputsRead |= FRAG_BIT_FOGC;
+      
    if (program->Base.Instructions)
       _mesa_free(program->Base.Instructions);
    program->Base.Instructions = ap.Base.Instructions;
diff --git a/src/mesa/shader/descrip.mms b/src/mesa/shader/descrip.mms
index bdac946efe..19bafd4830 100644
--- a/src/mesa/shader/descrip.mms
+++ b/src/mesa/shader/descrip.mms
@@ -1,6 +1,6 @@
 # Makefile for core library for VMS
 # contributed by Jouk Jansen  joukj@hrem.nano.tudelft.nl
-# Last revision : 27 May 2008
+# Last revision : 29 September 2008
 .first
 	define gl [---.include.gl]
 	define math [-.math]
@@ -34,6 +34,7 @@ SOURCES = \
 	prog_instruction.c \
 	prog_parameter.c \
 	prog_print.c \
+	prog_cache.c \
 	prog_statevars.c \
 	shader_api.c prog_uniform.c
 
@@ -52,7 +53,7 @@ OBJECTS = \
 	prog_parameter.obj,\
 	prog_print.obj,\
 	prog_statevars.obj,\
-	shader_api.obj,prog_uniform.obj
+	shader_api.obj,prog_uniform.obj,prog_cache.obj
 
 ##### RULES #####
 
@@ -91,3 +92,4 @@ prog_print.obj : prog_print.c
 prog_statevars.obj : prog_statevars.c
 shader_api.obj : shader_api.c
 prog_uniform.obj : prog_uniform.c
+prog_cache.obj : prog_cache.c
diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c
index 5d21cff672..37962f0e9b 100644
--- a/src/mesa/shader/program.c
+++ b/src/mesa/shader/program.c
@@ -372,7 +372,11 @@ _mesa_reference_program(GLcontext *ctx,
    assert(ptr);
    if (*ptr && prog) {
       /* sanity check */
-      ASSERT((*ptr)->Target == prog->Target);
+      if ((*ptr)->Target == GL_VERTEX_PROGRAM_ARB)
+         ASSERT(prog->Target == GL_VERTEX_PROGRAM_ARB);
+      else if ((*ptr)->Target == GL_FRAGMENT_PROGRAM_ARB)
+         ASSERT(prog->Target == GL_FRAGMENT_PROGRAM_ARB ||
+                prog->Target == GL_FRAGMENT_PROGRAM_NV);
    }
    if (*ptr == prog) {
       return;  /* no change */
diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index 854f8bfdaa..266ecc4ef2 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -47,7 +47,7 @@
 #include "shader/shader_api.h"
 #include "shader/slang/slang_compile.h"
 #include "shader/slang/slang_link.h"
-
+#include "glapi/dispatch.h"
 
 
 #ifndef GL_PROGRAM_BINARY_LENGTH_OES
@@ -455,7 +455,13 @@ _mesa_attach_shader(GLcontext *ctx, GLuint program, GLuint shader)
    n = shProg->NumShaders;
    for (i = 0; i < n; i++) {
       if (shProg->Shaders[i] == sh) {
-         /* already attached */
+         /* The shader is already attched to this program.  The
+          * GL_ARB_shader_objects spec says:
+          *
+          *     "The error INVALID_OPERATION is generated by AttachObjectARB
+          *     if <obj> is already attached to <containerObj>."
+          */
+         _mesa_error(ctx, GL_INVALID_OPERATION, "glAttachShader");
          return;
       }
    }
@@ -919,24 +925,15 @@ _mesa_get_attached_shaders(GLcontext *ctx, GLuint program, GLsizei maxCount,
 static GLuint
 _mesa_get_handle(GLcontext *ctx, GLenum pname)
 {
-#if 0
-   GET_CURRENT_CONTEXT(ctx);
-
-   switch (pname) {
-   case GL_PROGRAM_OBJECT_ARB:
-      {
-         struct gl2_program_intf **pro = ctx->Shader.CurrentProgram;
-
-         if (pro != NULL)
-            return (**pro)._container._generic.
-               GetName((struct gl2_generic_intf **) (pro));
-      }
-      break;
-   default:
+   GLint handle = 0;
+   
+   if (pname == GL_PROGRAM_OBJECT_ARB) {
+      CALL_GetIntegerv(ctx->Exec, (GL_CURRENT_PROGRAM, &handle));
+   } else {
       _mesa_error(ctx, GL_INVALID_ENUM, "glGetHandleARB");
    }
-#endif
-   return 0;
+
+   return handle;
 }
 
 
diff --git a/src/mesa/swrast/s_depth.c b/src/mesa/swrast/s_depth.c
index 293eb8628e..26e23f02d5 100644
--- a/src/mesa/swrast/s_depth.c
+++ b/src/mesa/swrast/s_depth.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.1
+ * Version:  7.2.1
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -534,15 +534,15 @@ depth_test_span( GLcontext *ctx, SWspan *span)
       if (rb->DataType == GL_UNSIGNED_SHORT) {
          GLushort zbuffer[MAX_WIDTH];
          rb->GetRow(ctx, rb, count, x, y, zbuffer);
-         passed = depth_test_span16(ctx, count, zbuffer, zValues, mask );
-         rb->PutRow(ctx, rb, count, x, y, zbuffer, NULL);
+         passed = depth_test_span16(ctx, count, zbuffer, zValues, mask);
+         rb->PutRow(ctx, rb, count, x, y, zbuffer, mask);
       }
       else {
          GLuint zbuffer[MAX_WIDTH];
          ASSERT(rb->DataType == GL_UNSIGNED_INT);
          rb->GetRow(ctx, rb, count, x, y, zbuffer);
-         passed = depth_test_span32(ctx, count, zbuffer, zValues, mask );
-         rb->PutRow(ctx, rb, count, x, y, zbuffer, NULL);
+         passed = depth_test_span32(ctx, count, zbuffer, zValues, mask);
+         rb->PutRow(ctx, rb, count, x, y, zbuffer, mask);
       }
    }
 
@@ -1080,15 +1080,15 @@ depth_test_pixels( GLcontext *ctx, SWspan *span )
       if (rb->DataType == GL_UNSIGNED_SHORT) {
          GLushort zbuffer[MAX_WIDTH];
          _swrast_get_values(ctx, rb, count, x, y, zbuffer, sizeof(GLushort));
-         depth_test_span16(ctx, count, zbuffer, z, mask );
-         rb->PutValues(ctx, rb, count, x, y, zbuffer, NULL);
+         depth_test_span16(ctx, count, zbuffer, z, mask);
+         rb->PutValues(ctx, rb, count, x, y, zbuffer, mask);
       }
       else {
          GLuint zbuffer[MAX_WIDTH];
          ASSERT(rb->DataType == GL_UNSIGNED_INT);
          _swrast_get_values(ctx, rb, count, x, y, zbuffer, sizeof(GLuint));
-         depth_test_span32(ctx, count, zbuffer, z, mask );
-         rb->PutValues(ctx, rb, count, x, y, zbuffer, NULL);
+         depth_test_span32(ctx, count, zbuffer, z, mask);
+         rb->PutValues(ctx, rb, count, x, y, zbuffer, mask);
       }
    }
 
diff --git a/src/mesa/swrast/s_points.c b/src/mesa/swrast/s_points.c
index 846c485f15..61ff4d0b84 100644
--- a/src/mesa/swrast/s_points.c
+++ b/src/mesa/swrast/s_points.c
@@ -129,13 +129,13 @@ sprite_point(GLcontext *ctx, const SWvertex *vert)
       s = 0.0;
       dsdx = 1.0 / size;
       if (ctx->Point.SpriteOrigin == GL_LOWER_LEFT) {
-         t0 = 0.0;
          dtdy = 1.0 / size;
+         t0 = 0.5 * dtdy;
       }
       else {
          /* GL_UPPER_LEFT */
-         t0 = 1.0;
          dtdy = -1.0 / size;
+         t0 = 1.0 + 0.5 * dtdy;
       }
 
       ATTRIB_LOOP_BEGIN
diff --git a/src/mesa/swrast/s_texstore.c b/src/mesa/swrast/s_texstore.c
index 15d52aa587..16b00b9fa1 100644
--- a/src/mesa/swrast/s_texstore.c
+++ b/src/mesa/swrast/s_texstore.c
@@ -216,9 +216,9 @@ is_depth_format(GLenum format)
 {
    switch (format) {
       case GL_DEPTH_COMPONENT:
-      case GL_DEPTH_COMPONENT16_SGIX:
-      case GL_DEPTH_COMPONENT24_SGIX:
-      case GL_DEPTH_COMPONENT32_SGIX:
+      case GL_DEPTH_COMPONENT16:
+      case GL_DEPTH_COMPONENT24:
+      case GL_DEPTH_COMPONENT32:
          return GL_TRUE;
       default:
          return GL_FALSE;
diff --git a/src/mesa/swrast_setup/ss_context.c b/src/mesa/swrast_setup/ss_context.c
index f4d90c514b..61172f9979 100644
--- a/src/mesa/swrast_setup/ss_context.c
+++ b/src/mesa/swrast_setup/ss_context.c
@@ -112,22 +112,25 @@ setup_vertex_format(GLcontext *ctx)
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    SScontext *swsetup = SWSETUP_CONTEXT(ctx);
+   GLboolean intColors = !ctx->FragmentProgram._Current
+                      && !ctx->ATIFragmentShader._Enabled
+                      && ctx->RenderMode == GL_RENDER
+                      && CHAN_TYPE == GL_UNSIGNED_BYTE;
 
-   if (!RENDERINPUTS_EQUAL(tnl->render_inputs_bitset,
+   if (intColors != swsetup->intColors ||
+       !RENDERINPUTS_EQUAL(tnl->render_inputs_bitset,
                            swsetup->last_index_bitset)) {
       DECLARE_RENDERINPUTS(index_bitset);
       struct tnl_attr_map map[_TNL_ATTRIB_MAX];
       int i, e = 0;
 
+      swsetup->intColors = intColors;
+
       RENDERINPUTS_COPY( index_bitset, tnl->render_inputs_bitset );
 
       EMIT_ATTR( _TNL_ATTRIB_POS, EMIT_4F_VIEWPORT, attrib[FRAG_ATTRIB_WPOS] );
 
       if (RENDERINPUTS_TEST( index_bitset, _TNL_ATTRIB_COLOR0 )) {
-         swsetup->intColors = !ctx->FragmentProgram._Current
-                           && !ctx->ATIFragmentShader._Enabled
-                           && ctx->RenderMode == GL_RENDER
-                           && CHAN_TYPE == GL_UNSIGNED_BYTE;
          if (swsetup->intColors)
             EMIT_ATTR( _TNL_ATTRIB_COLOR0, EMIT_4CHAN_4F_RGBA, color );
          else
diff --git a/src/mesa/tnl/descrip.mms b/src/mesa/tnl/descrip.mms
index f77f672dc8..25dd1aecb1 100644
--- a/src/mesa/tnl/descrip.mms
+++ b/src/mesa/tnl/descrip.mms
@@ -1,6 +1,6 @@
 # Makefile for core library for VMS
 # contributed by Jouk Jansen  joukj@hrem.nano.tudelft.nl
-# Last revision : 30 November 2007
+# Last revision : 39 September 2008
 
 .first
 	define gl [---.include.gl]
@@ -27,13 +27,13 @@ SOURCES = t_context.c t_draw.c \
 	t_pipeline.c t_vb_fog.c \
 	t_vb_light.c t_vb_normals.c t_vb_points.c t_vb_program.c \
 	t_vb_render.c t_vb_texgen.c t_vb_texmat.c t_vb_vertex.c \
-	t_vertex.c \
+	t_vertex.c t_rasterpos.c\
 	t_vertex_generic.c t_vp_build.c
 
 OBJECTS = t_context.obj,t_draw.obj,\
 	t_pipeline.obj,t_vb_fog.obj,t_vb_light.obj,t_vb_normals.obj,\
 	t_vb_points.obj,t_vb_program.obj,t_vb_render.obj,t_vb_texgen.obj,\
-	t_vb_texmat.obj,t_vb_vertex.obj,\
+	t_vb_texmat.obj,t_vb_vertex.obj,t_rasterpos.obj,\
 	t_vertex.obj,t_vertex_generic.obj,\
 	t_vp_build.obj
 
@@ -65,3 +65,4 @@ t_vb_vertex.obj : t_vb_vertex.c
 t_vertex.obj : t_vertex.c
 t_vertex_generic.obj : t_vertex_generic.c
 t_vp_build.obj : t_vp_build.c
+t_rasterpos.obj : t_rasterpos.c
diff --git a/src/mesa/tnl/t_vertex_generic.c b/src/mesa/tnl/t_vertex_generic.c
index db70ad4dad..f763522f91 100644
--- a/src/mesa/tnl/t_vertex_generic.c
+++ b/src/mesa/tnl/t_vertex_generic.c
@@ -34,6 +34,12 @@
 #include "t_vertex.h"
 
 
+#if 0
+#define DEBUG_INSERT printf("%s\n", __FUNCTION__)
+#else
+#define DEBUG_INSERT
+#endif
+
 
 /*
  * These functions take the NDC coordinates pointed to by 'in', apply the
@@ -45,7 +51,7 @@ static INLINE void insert_4f_viewport_4( const struct tnl_clipspace_attr *a, GLu
 {
    GLfloat *out = (GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[5] * in[1] + vp[13];
    out[2] = vp[10] * in[2] + vp[14];
@@ -57,7 +63,7 @@ static INLINE void insert_4f_viewport_3( const struct tnl_clipspace_attr *a, GLu
 {
    GLfloat *out = (GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[5] * in[1] + vp[13];
    out[2] = vp[10] * in[2] + vp[14];
@@ -69,7 +75,7 @@ static INLINE void insert_4f_viewport_2( const struct tnl_clipspace_attr *a, GLu
 {
    GLfloat *out = (GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[5] * in[1] + vp[13];
    out[2] = vp[14];
@@ -81,7 +87,7 @@ static INLINE void insert_4f_viewport_1( const struct tnl_clipspace_attr *a, GLu
 {
    GLfloat *out = (GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[13];
    out[2] = vp[14];
@@ -93,7 +99,7 @@ static INLINE void insert_3f_viewport_3( const struct tnl_clipspace_attr *a, GLu
 {
    GLfloat *out = (GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[5] * in[1] + vp[13];
    out[2] = vp[10] * in[2] + vp[14];
@@ -104,7 +110,7 @@ static INLINE void insert_3f_viewport_2( const struct tnl_clipspace_attr *a, GLu
 {
    GLfloat *out = (GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[5] * in[1] + vp[13];
    out[2] = vp[10] * in[2] + vp[14];
@@ -115,7 +121,7 @@ static INLINE void insert_3f_viewport_1( const struct tnl_clipspace_attr *a, GLu
 {
    GLfloat *out = (GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[13];
    out[2] = vp[14];
@@ -126,7 +132,7 @@ static INLINE void insert_2f_viewport_2( const struct tnl_clipspace_attr *a, GLu
 {
    GLfloat *out = (GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[5] * in[1] + vp[13];
 }
@@ -136,7 +142,7 @@ static INLINE void insert_2f_viewport_1( const struct tnl_clipspace_attr *a, GLu
 {
    GLfloat *out = (GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = vp[0] * in[0] + vp[12];
    out[1] = vp[13];
 }
@@ -150,7 +156,7 @@ static INLINE void insert_4f_4( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = in[1];
    out[2] = in[2];
@@ -161,7 +167,7 @@ static INLINE void insert_4f_3( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = in[1];
    out[2] = in[2];
@@ -172,7 +178,7 @@ static INLINE void insert_4f_2( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = in[1];
    out[2] = 0;
@@ -183,7 +189,7 @@ static INLINE void insert_4f_1( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = 0;
    out[2] = 0;
@@ -194,7 +200,7 @@ static INLINE void insert_3f_xyw_4( const struct tnl_clipspace_attr *a, GLubyte
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = in[1];
    out[2] = in[3];
@@ -203,6 +209,7 @@ static INLINE void insert_3f_xyw_4( const struct tnl_clipspace_attr *a, GLubyte
 static INLINE void insert_3f_xyw_err( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
 {
    (void) a; (void) v; (void) in;
+   DEBUG_INSERT;
    _mesa_exit(1);
 }
 
@@ -210,7 +217,7 @@ static INLINE void insert_3f_3( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = in[1];
    out[2] = in[2];
@@ -220,7 +227,7 @@ static INLINE void insert_3f_2( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = in[1];
    out[2] = 0;
@@ -230,7 +237,7 @@ static INLINE void insert_3f_1( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = 0;
    out[2] = 0;
@@ -241,7 +248,7 @@ static INLINE void insert_2f_2( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = in[1];
 }
@@ -250,7 +257,7 @@ static INLINE void insert_2f_1( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-   
+   DEBUG_INSERT;
    out[0] = in[0];
    out[1] = 0;
 }
@@ -259,12 +266,13 @@ static INLINE void insert_1f_1( const struct tnl_clipspace_attr *a, GLubyte *v,
 {
    GLfloat *out = (GLfloat *)(v);
    (void) a;
-
+   DEBUG_INSERT;
    out[0] = in[0];
 }
 
 static INLINE void insert_null( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a; (void) v; (void) in;
 }
 
@@ -272,6 +280,7 @@ static INLINE void insert_4chan_4f_rgba_4( const struct tnl_clipspace_attr *a, G
 				  const GLfloat *in )
 {
    GLchan *c = (GLchan *)v;
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
    UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
@@ -283,6 +292,7 @@ static INLINE void insert_4chan_4f_rgba_3( const struct tnl_clipspace_attr *a, G
 				  const GLfloat *in )
 {
    GLchan *c = (GLchan *)v;
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
    UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
@@ -294,6 +304,7 @@ static INLINE void insert_4chan_4f_rgba_2( const struct tnl_clipspace_attr *a, G
 				  const GLfloat *in )
 {
    GLchan *c = (GLchan *)v;
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
    UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
@@ -305,6 +316,7 @@ static INLINE void insert_4chan_4f_rgba_1( const struct tnl_clipspace_attr *a, G
 				  const GLfloat *in )
 {
    GLchan *c = (GLchan *)v;
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
    c[1] = 0;
@@ -315,6 +327,7 @@ static INLINE void insert_4chan_4f_rgba_1( const struct tnl_clipspace_attr *a, G
 static INLINE void insert_4ub_4f_rgba_4( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -325,6 +338,7 @@ static INLINE void insert_4ub_4f_rgba_4( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_rgba_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -335,6 +349,7 @@ static INLINE void insert_4ub_4f_rgba_3( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_rgba_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -345,6 +360,7 @@ static INLINE void insert_4ub_4f_rgba_2( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_rgba_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
    v[1] = 0;
@@ -355,6 +371,7 @@ static INLINE void insert_4ub_4f_rgba_1( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_bgra_4( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -365,6 +382,7 @@ static INLINE void insert_4ub_4f_bgra_4( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_bgra_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -375,6 +393,7 @@ static INLINE void insert_4ub_4f_bgra_3( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_bgra_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -385,6 +404,7 @@ static INLINE void insert_4ub_4f_bgra_2( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_bgra_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
    v[1] = 0;
@@ -395,6 +415,7 @@ static INLINE void insert_4ub_4f_bgra_1( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_argb_4( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
@@ -405,6 +426,7 @@ static INLINE void insert_4ub_4f_argb_4( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_argb_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
@@ -415,6 +437,7 @@ static INLINE void insert_4ub_4f_argb_3( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_argb_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
@@ -425,6 +448,7 @@ static INLINE void insert_4ub_4f_argb_2( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_argb_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
    v[2] = 0x00;
@@ -435,6 +459,7 @@ static INLINE void insert_4ub_4f_argb_1( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_abgr_4( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
@@ -445,6 +470,7 @@ static INLINE void insert_4ub_4f_abgr_4( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_abgr_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
@@ -455,6 +481,7 @@ static INLINE void insert_4ub_4f_abgr_3( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_abgr_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
@@ -465,6 +492,7 @@ static INLINE void insert_4ub_4f_abgr_2( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_4ub_4f_abgr_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
    v[2] = 0x00;
@@ -475,6 +503,7 @@ static INLINE void insert_4ub_4f_abgr_1( const struct tnl_clipspace_attr *a, GLu
 static INLINE void insert_3ub_3f_rgb_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
 			       const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -484,6 +513,7 @@ static INLINE void insert_3ub_3f_rgb_3( const struct tnl_clipspace_attr *a, GLub
 static INLINE void insert_3ub_3f_rgb_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
 			       const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -493,6 +523,7 @@ static INLINE void insert_3ub_3f_rgb_2( const struct tnl_clipspace_attr *a, GLub
 static INLINE void insert_3ub_3f_rgb_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
 			       const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
    v[1] = 0;
@@ -502,6 +533,7 @@ static INLINE void insert_3ub_3f_rgb_1( const struct tnl_clipspace_attr *a, GLub
 static INLINE void insert_3ub_3f_bgr_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				 const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -511,6 +543,7 @@ static INLINE void insert_3ub_3f_bgr_3( const struct tnl_clipspace_attr *a, GLub
 static INLINE void insert_3ub_3f_bgr_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				 const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
    UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
@@ -520,6 +553,7 @@ static INLINE void insert_3ub_3f_bgr_2( const struct tnl_clipspace_attr *a, GLub
 static INLINE void insert_3ub_3f_bgr_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
 				 const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
    v[1] = 0;
@@ -530,6 +564,7 @@ static INLINE void insert_3ub_3f_bgr_1( const struct tnl_clipspace_attr *a, GLub
 static INLINE void insert_1ub_1f_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
 			   const GLfloat *in )
 {
+   DEBUG_INSERT;
    (void) a;
    UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
 }
@@ -551,6 +586,7 @@ static void extract_4f_viewport( const struct tnl_clipspace_attr *a, GLfloat *ou
    /* Although included for completeness, the position coordinate is
     * usually handled differently during clipping.
     */
+   DEBUG_INSERT;
    out[0] = (in[0] - vp[12]) / vp[0];
    out[1] = (in[1] - vp[13]) / vp[5];
    out[2] = (in[2] - vp[14]) / vp[10];
@@ -562,7 +598,7 @@ static void extract_3f_viewport( const struct tnl_clipspace_attr *a, GLfloat *ou
 {
    const GLfloat *in = (const GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = (in[0] - vp[12]) / vp[0];
    out[1] = (in[1] - vp[13]) / vp[5];
    out[2] = (in[2] - vp[14]) / vp[10];
@@ -575,7 +611,7 @@ static void extract_2f_viewport( const struct tnl_clipspace_attr *a, GLfloat *ou
 {
    const GLfloat *in = (const GLfloat *)v;
    const GLfloat * const vp = a->vp;
-   
+   DEBUG_INSERT;
    out[0] = (in[0] - vp[12]) / vp[0];
    out[1] = (in[1] - vp[13]) / vp[5];
    out[2] = 0;
diff --git a/src/mesa/x86-64/x86-64.c b/src/mesa/x86-64/x86-64.c
index 9ec43c841d..96f8da87f0 100644
--- a/src/mesa/x86-64/x86-64.c
+++ b/src/mesa/x86-64/x86-64.c
@@ -41,7 +41,10 @@
 #include "math/m_debug.h"
 #endif
 
+extern void _mesa_x86_64_cpuid(unsigned int *regs);
+
 DECLARE_XFORM_GROUP( x86_64, 4 )
+DECLARE_XFORM_GROUP( 3dnow, 4 )
 
 #else
 /* just to silence warning below */
@@ -81,6 +84,7 @@ static void message( const char *msg )
 void _mesa_init_all_x86_64_transform_asm(void)
 {
 #ifdef USE_X86_64_ASM
+   unsigned int regs[4];
 
    if ( _mesa_getenv( "MESA_NO_ASM" ) ) {
      return;
@@ -88,24 +92,32 @@ void _mesa_init_all_x86_64_transform_asm(void)
 
    message("Initializing x86-64 optimizations\n");
 
-   ASSIGN_XFORM_GROUP( x86_64, 4 );
 
-   /*
    _mesa_transform_tab[4][MATRIX_GENERAL] =
       _mesa_x86_64_transform_points4_general;
    _mesa_transform_tab[4][MATRIX_IDENTITY] =
       _mesa_x86_64_transform_points4_identity;
    _mesa_transform_tab[4][MATRIX_3D] =
       _mesa_x86_64_transform_points4_3d;
-   _mesa_transform_tab[4][MATRIX_3D_NO_ROT] =
-      _mesa_x86_64_transform_points4_3d_no_rot;
-   _mesa_transform_tab[4][MATRIX_PERSPECTIVE] =
-      _mesa_x86_64_transform_points4_perspective;
-   _mesa_transform_tab[4][MATRIX_2D_NO_ROT] =
-      _mesa_x86_64_transform_points4_2d_no_rot;
-   _mesa_transform_tab[4][MATRIX_2D] =
-      _mesa_x86_64_transform_points4_2d;
-   */
+
+   regs[0] = 0x80000001;
+   regs[1] = 0x00000000;
+   regs[2] = 0x00000000;
+   regs[3] = 0x00000000;
+   _mesa_x86_64_cpuid(regs);
+   if (regs[3] & (1U << 31)) {
+      message("3Dnow! detected\n");
+      _mesa_transform_tab[4][MATRIX_3D_NO_ROT] =
+	  _mesa_3dnow_transform_points4_3d_no_rot;
+      _mesa_transform_tab[4][MATRIX_PERSPECTIVE] =
+	  _mesa_3dnow_transform_points4_perspective;
+      _mesa_transform_tab[4][MATRIX_2D_NO_ROT] =
+	  _mesa_3dnow_transform_points4_2d_no_rot;
+      _mesa_transform_tab[4][MATRIX_2D] =
+	  _mesa_3dnow_transform_points4_2d;
+
+   }
+
    
 #ifdef DEBUG_MATH
    _math_test_all_transform_functions("x86_64");
diff --git a/src/mesa/x86-64/xform4.S b/src/mesa/x86-64/xform4.S
index 3f9c9d56ab..805969127d 100644
--- a/src/mesa/x86-64/xform4.S
+++ b/src/mesa/x86-64/xform4.S
@@ -29,7 +29,22 @@
 .text
 
 .align 16
+.globl _mesa_x86_64_cpuid
+_mesa_x86_64_cpuid:
+	pushq	%rbx
+	movl	(%rdi), %eax
+	movl	8(%rdi), %ecx
+
+	cpuid
+
+	movl	%ebx, 4(%rdi)
+	movl	%eax, (%rdi)
+	movl	%ecx, 8(%rdi)
+	movl	%edx, 12(%rdi)
+	popq	%rbx
+	ret
 
+.align 16
 .globl _mesa_x86_64_transform_points4_general
 _mesa_x86_64_transform_points4_general:
 /*
@@ -204,8 +219,8 @@ p4_identity_done:
 
 	
 .align 16
-.globl _mesa_x86_64_transform_points4_3d_no_rot
-_mesa_x86_64_transform_points4_3d_no_rot:
+.globl _mesa_3dnow_transform_points4_3d_no_rot
+_mesa_3dnow_transform_points4_3d_no_rot:
 
 	movl V4F_COUNT(%rdx), %ecx	/* count */
 	movzx V4F_STRIDE(%rdx), %eax	/* stride */
@@ -268,8 +283,8 @@ p4_3d_no_rot_done:
 
 	
 .align 16
-.globl _mesa_x86_64_transform_points4_perspective
-_mesa_x86_64_transform_points4_perspective:
+.globl _mesa_3dnow_transform_points4_perspective
+_mesa_3dnow_transform_points4_perspective:
 
 	movl V4F_COUNT(%rdx), %ecx	/* count */
 	movzx V4F_STRIDE(%rdx), %eax	/* stride */
@@ -334,8 +349,8 @@ p4_perspective_done:
 	ret
 
 .align 16
-.globl _mesa_x86_64_transform_points4_2d_no_rot
-_mesa_x86_64_transform_points4_2d_no_rot:
+.globl _mesa_3dnow_transform_points4_2d_no_rot
+_mesa_3dnow_transform_points4_2d_no_rot:
 
 	movl V4F_COUNT(%rdx), %ecx	/* count */
 	movzx V4F_STRIDE(%rdx), %eax	/* stride */
@@ -389,8 +404,8 @@ p4_2d_no_rot_done:
 
 	
 .align 16
-.globl _mesa_x86_64_transform_points4_2d
-_mesa_x86_64_transform_points4_2d:
+.globl _mesa_3dnow_transform_points4_2d
+_mesa_3dnow_transform_points4_2d:
 
 	movl V4F_COUNT(%rdx), %ecx	/* count */
 	movzx V4F_STRIDE(%rdx), %eax	/* stride */