Merge branch 'origin' into glsl-compiler-1

Conflicts:

	src/mesa/main/state.c
	src/mesa/shader/program.c
	src/mesa/shader/program.h
	src/mesa/shader/programopt.c
	src/mesa/shader/slang/slang_execute.c
	src/mesa/sources
	src/mesa/swrast/s_arbshader.c
	src/mesa/swrast/s_context.c
	src/mesa/swrast/s_span.c
	src/mesa/swrast/s_zoom.c
	src/mesa/tnl/t_context.c
	src/mesa/tnl/t_save_api.c
	src/mesa/tnl/t_vb_arbprogram.c
	src/mesa/tnl/t_vp_build.c
	src/mesa/tnl/t_vtx_eval.c

338 files changed, 23795 insertions, 21901 deletions

diff --git a/src/mesa/Makefile b/src/mesa/Makefile
index 1cd9e5b2bb7..0ab4fc5149c 100644
--- a/src/mesa/Makefile
+++ b/src/mesa/Makefile
@@ -142,7 +142,7 @@ depend: $(ALL_SOURCES)
 	@ echo "running $(MKDEP)"
 	@ touch depend
 	@$(MKDEP) $(MKDEP_OPTIONS) $(DEFINES) $(INCLUDE_DIRS) $(ALL_SOURCES) \
-		> /dev/null 
+		> /dev/null 2>/dev/null
 
 
 subdirs:
diff --git a/src/mesa/array_cache/ac_context.c b/src/mesa/array_cache/ac_context.c
deleted file mode 100644
index 5d5ad5e363b..00000000000
--- a/src/mesa/array_cache/ac_context.c
+++ /dev/null
@@ -1,375 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  5.1
- *
- * Copyright (C) 1999-2002  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"),
- * 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 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
- * BRIAN PAUL 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:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "macros.h"
-#include "imports.h"
-#include "mtypes.h"
-
-#include "array_cache/ac_context.h"
-
-
-/*
- * Initialize the array fallbacks.  That is, by default the fallback arrays
- * point into the current vertex attribute values in ctx->Current.Attrib[]
- */
-static void _ac_fallbacks_init( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   struct gl_client_array *cl;
-   GLuint i;
-
-   cl = &ac->Fallback.Normal;
-   cl->Size = 3;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) ctx->Current.Attrib[VERT_ATTRIB_NORMAL];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Fallback.Color;
-   cl->Size = 4;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Fallback.SecondaryColor;
-   cl->Size = 3;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) ctx->Current.Attrib[VERT_ATTRIB_COLOR1];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Fallback.FogCoord;
-   cl->Size = 1;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) &ctx->Current.Attrib[VERT_ATTRIB_FOG];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Fallback.Index;
-   cl->Size = 1;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) &ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX];
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   for (i = 0 ; i < MAX_TEXTURE_COORD_UNITS ; i++) {
-      cl = &ac->Fallback.TexCoord[i];
-      cl->Size = 4;
-      cl->Type = GL_FLOAT;
-      cl->Stride = 0;
-      cl->StrideB = 0;
-      cl->Ptr = (GLubyte *) ctx->Current.Attrib[VERT_ATTRIB_TEX0 + i];
-      cl->Enabled = 1;
-      cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-      cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-   }
-
-   cl = &ac->Fallback.EdgeFlag;
-   cl->Size = 1;
-   cl->Type = GL_UNSIGNED_BYTE;
-   cl->Stride = 0;
-   cl->StrideB = 0;
-   cl->Ptr = (GLubyte *) &ctx->Current.EdgeFlag;
-   cl->Enabled = 1;
-   cl->Flags = CA_CLIENT_DATA;	/* hack */
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-      cl = &ac->Fallback.Attrib[i];
-      cl->Size = 4;
-      cl->Type = GL_FLOAT;
-      cl->Stride = 0;
-      cl->StrideB = 0;
-      cl->Ptr = (GLubyte *) ctx->Current.Attrib[i];
-      cl->Enabled = 1;
-      cl->Flags = CA_CLIENT_DATA; /* hack */
-#if FEATURE_ARB_vertex_buffer_object
-      cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-   }
-}
-
-
-/*
- * Initialize the array cache pointers, types, strides, etc.
- */
-static void _ac_cache_init( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   struct gl_client_array *cl;
-   GLuint size = ctx->Const.MaxArrayLockSize + MAX_CLIPPED_VERTICES;
-   GLuint i;
-
-   cl = &ac->Cache.Vertex;
-   cl->Size = 4;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 4 * sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.Normal;
-   cl->Size = 3;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 3 * sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.Color;
-   cl->Size = 4;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 4 * sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.SecondaryColor;
-   cl->Size = 3;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = 4 * sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.FogCoord;
-   cl->Size = 1;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   cl = &ac->Cache.Index;
-   cl->Size = 1;
-   cl->Type = GL_FLOAT;
-   cl->Stride = 0;
-   cl->StrideB = sizeof(GLfloat);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
-      cl = &ac->Cache.TexCoord[i];
-      cl->Size = 4;
-      cl->Type = GL_FLOAT;
-      cl->Stride = 0;
-      cl->StrideB = 4 * sizeof(GLfloat);
-      cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-      cl->Enabled = 1;
-      cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-      cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-   }
-
-   cl = &ac->Cache.EdgeFlag;
-   cl->Size = 1;
-   cl->Type = GL_UNSIGNED_BYTE;
-   cl->Stride = 0;
-   cl->StrideB = sizeof(GLubyte);
-   cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-   cl->Enabled = 1;
-   cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-   cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-
-   for (i = 0 ; i < VERT_ATTRIB_MAX; i++) {
-      cl = &ac->Cache.Attrib[i];
-      cl->Size = 4;
-      cl->Type = GL_FLOAT;
-      cl->Stride = 0;
-      cl->StrideB = 4 * sizeof(GLfloat);
-      cl->Ptr = (GLubyte *) MALLOC( cl->StrideB * size );
-      cl->Enabled = 1;
-      cl->Flags = 0;
-#if FEATURE_ARB_vertex_buffer_object
-      cl->BufferObj = ctx->Array.NullBufferObj;
-#endif
-   }
-}
-
-
-/* This storage used to hold translated client data if type or stride
- * need to be fixed.
- */
-static void _ac_elts_init( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   GLuint size = 1000;
-
-   ac->Elts = (GLuint *)MALLOC( sizeof(GLuint) * size );
-   ac->elt_size = size;
-}
-
-static void _ac_raw_init( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   GLuint i;
-
-   ac->Raw.Color = ac->Fallback.Color;
-   ac->Raw.EdgeFlag = ac->Fallback.EdgeFlag;
-   ac->Raw.FogCoord = ac->Fallback.FogCoord;
-   ac->Raw.Index = ac->Fallback.Index;
-   ac->Raw.Normal = ac->Fallback.Normal;
-   ac->Raw.SecondaryColor = ac->Fallback.SecondaryColor;
-   ac->Raw.Vertex = ctx->Array.ArrayObj->Vertex;
-
-   ac->IsCached.Color = GL_FALSE;
-   ac->IsCached.EdgeFlag = GL_FALSE;
-   ac->IsCached.FogCoord = GL_FALSE;
-   ac->IsCached.Index = GL_FALSE;
-   ac->IsCached.Normal = GL_FALSE;
-   ac->IsCached.SecondaryColor = GL_FALSE;
-   ac->IsCached.Vertex = GL_FALSE;
-
-   for (i = 0 ; i < MAX_TEXTURE_COORD_UNITS ; i++) {
-      ac->Raw.TexCoord[i] = ac->Fallback.TexCoord[i];
-      ac->IsCached.TexCoord[i] = GL_FALSE;
-   }
-
-   for (i = 0 ; i < VERT_ATTRIB_MAX ; i++) {
-      ac->Raw.Attrib[i] = ac->Fallback.Attrib[i];
-      ac->IsCached.Attrib[i] = GL_FALSE;
-   }
-}
-
-GLboolean _ac_CreateContext( GLcontext *ctx )
-{
-   ctx->acache_context = CALLOC(sizeof(ACcontext));
-   if (ctx->acache_context) {
-      _ac_cache_init( ctx );
-      _ac_fallbacks_init( ctx );
-      _ac_raw_init( ctx );
-      _ac_elts_init( ctx );
-      return GL_TRUE;
-   }
-   return GL_FALSE;
-}
-
-void _ac_DestroyContext( GLcontext *ctx )
-{
-   struct gl_buffer_object *nullObj = ctx->Array.NullBufferObj;
-   ACcontext *ac = AC_CONTEXT(ctx);
-   GLint i;
-
-   /* only free vertex data if it's really a pointer to vertex data and
-    * not an offset into a buffer object.
-    */
-   if (ac->Cache.Vertex.Ptr && ac->Cache.Vertex.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.Vertex.Ptr );
-   if (ac->Cache.Normal.Ptr && ac->Cache.Normal.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.Normal.Ptr );
-   if (ac->Cache.Color.Ptr && ac->Cache.Color.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.Color.Ptr );
-   if (ac->Cache.SecondaryColor.Ptr && ac->Cache.SecondaryColor.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.SecondaryColor.Ptr );
-   if (ac->Cache.EdgeFlag.Ptr && ac->Cache.EdgeFlag.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.EdgeFlag.Ptr );
-   if (ac->Cache.Index.Ptr && ac->Cache.Index.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.Index.Ptr );
-   if (ac->Cache.FogCoord.Ptr && ac->Cache.FogCoord.BufferObj == nullObj)
-      FREE( (void *) ac->Cache.FogCoord.Ptr );
-
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++) {
-      if (ac->Cache.TexCoord[i].Ptr && ac->Cache.TexCoord[i].BufferObj == nullObj)
-	 FREE( (void *) ac->Cache.TexCoord[i].Ptr );
-   }
-
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-      if (ac->Cache.Attrib[i].Ptr && ac->Cache.Attrib[i].BufferObj == nullObj)
-	 FREE( (void *) ac->Cache.Attrib[i].Ptr );
-   }
-
-   if (ac->Elts)
-      FREE( ac->Elts );
-
-   /* Free the context structure itself */
-   FREE(ac);
-   ctx->acache_context = NULL;
-}
-
-void _ac_InvalidateState( GLcontext *ctx, GLuint new_state )
-{
-   AC_CONTEXT(ctx)->NewState |= new_state;
-   AC_CONTEXT(ctx)->NewArrayState |= ctx->Array.NewState;
-}
diff --git a/src/mesa/array_cache/ac_context.h b/src/mesa/array_cache/ac_context.h
deleted file mode 100644
index 36aded3152c..00000000000
--- a/src/mesa/array_cache/ac_context.h
+++ /dev/null
@@ -1,99 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  5.1
- *
- * Copyright (C) 1999-2002  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"),
- * 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 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
- * BRIAN PAUL 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:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef _AC_CONTEXT_H
-#define _AC_CONTEXT_H
-
-#include "glheader.h"
-#include "mtypes.h"
-
-#include "array_cache/acache.h"
-
-/* These are used to make the ctx->Current values look like
- * arrays (with zero StrideB).
- */
-struct ac_arrays {
-   struct gl_client_array Vertex;
-   struct gl_client_array Normal;
-   struct gl_client_array Color;
-   struct gl_client_array SecondaryColor;
-   struct gl_client_array FogCoord;
-   struct gl_client_array Index;
-   struct gl_client_array TexCoord[MAX_TEXTURE_COORD_UNITS];
-   struct gl_client_array EdgeFlag;
-   struct gl_client_array Attrib[VERT_ATTRIB_MAX];  /* GL_NV_vertex_program */
-};
-
-struct ac_array_pointers {
-   struct gl_client_array *Vertex;
-   struct gl_client_array *Normal;
-   struct gl_client_array *Color;
-   struct gl_client_array *SecondaryColor;
-   struct gl_client_array *FogCoord;
-   struct gl_client_array *Index;
-   struct gl_client_array *TexCoord[MAX_TEXTURE_COORD_UNITS];
-   struct gl_client_array *EdgeFlag;
-   struct gl_client_array *Attrib[VERT_ATTRIB_MAX];  /* GL_NV_vertex_program */
-};
-
-struct ac_array_flags {
-   GLboolean Vertex;
-   GLboolean Normal;
-   GLboolean Color;
-   GLboolean SecondaryColor;
-   GLboolean FogCoord;
-   GLboolean Index;
-   GLboolean TexCoord[MAX_TEXTURE_COORD_UNITS];
-   GLboolean EdgeFlag;
-   GLboolean Attrib[VERT_ATTRIB_MAX];  /* GL_NV_vertex_program */
-};
-
-
-typedef struct {
-   GLuint NewState;		/* not needed? */
-   GLuint NewArrayState;
-
-   /* Facility for importing and caching array data:
-    */
-   struct ac_arrays Fallback;
-   struct ac_arrays Cache;
-   struct ac_arrays Raw;
-   struct ac_array_flags IsCached;
-   GLuint start;
-   GLuint count;
-
-   /* Facility for importing element lists:
-    */
-   GLuint *Elts;
-   GLuint elt_size;
-
-} ACcontext;
-
-#define AC_CONTEXT(ctx) ((ACcontext *)ctx->acache_context)
-
-#endif
diff --git a/src/mesa/array_cache/ac_import.c b/src/mesa/array_cache/ac_import.c
deleted file mode 100644
index 119e58bf249..00000000000
--- a/src/mesa/array_cache/ac_import.c
+++ /dev/null
@@ -1,922 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5
- *
- * Copyright (C) 1999-2006  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"),
- * 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 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
- * BRIAN PAUL 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:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "macros.h"
-#include "imports.h"
-#include "mtypes.h"
-
-#include "math/m_translate.h"
-#include "array_cache/ac_context.h"
-#include "math/m_translate.h"
-
-#define STRIDE_ARRAY( array, offset ) 					\
-do {									\
-   GLubyte *tmp = ADD_POINTERS( (array).BufferObj->Data, (array).Ptr )	\
-                + (offset) * (array).StrideB;				\
-   (array).Ptr = tmp;							\
-} while (0)
-
-
-/* Set the array pointer back to its source when the cached data is
- * invalidated:
- */
-static void
-reset_texcoord( GLcontext *ctx, GLuint unit )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->TexCoord[unit].Enabled) {
-      ac->Raw.TexCoord[unit] = ctx->Array.ArrayObj->TexCoord[unit];
-      STRIDE_ARRAY(ac->Raw.TexCoord[unit], ac->start);
-   }
-   else {
-      ac->Raw.TexCoord[unit] = ac->Fallback.TexCoord[unit];
-
-      if (ctx->Current.Attrib[VERT_ATTRIB_TEX0 + unit][3] != 1.0)
-	 ac->Raw.TexCoord[unit].Size = 4;
-      else if (ctx->Current.Attrib[VERT_ATTRIB_TEX0 + unit][2] != 0.0)
-	 ac->Raw.TexCoord[unit].Size = 3;
-      else
-	 ac->Raw.TexCoord[unit].Size = 2;
-   }
-
-   ac->IsCached.TexCoord[unit] = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_TEXCOORD(unit);
-}
-
-static void
-reset_vertex( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   ASSERT(ctx->Array.ArrayObj->Vertex.Enabled
-          || (ctx->VertexProgram._Enabled && ctx->Array.ArrayObj->VertexAttrib[0].Enabled));
-   ac->Raw.Vertex = ctx->Array.ArrayObj->Vertex;
-   STRIDE_ARRAY(ac->Raw.Vertex, ac->start);
-   ac->IsCached.Vertex = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_VERTEX;
-}
-
-
-static void
-reset_normal( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->Normal.Enabled) {
-      ac->Raw.Normal = ctx->Array.ArrayObj->Normal;
-      STRIDE_ARRAY(ac->Raw.Normal, ac->start);
-   }
-   else {
-      ac->Raw.Normal = ac->Fallback.Normal;
-   }
-
-   ac->IsCached.Normal = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_NORMAL;
-}
-
-
-static void
-reset_color( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->Color.Enabled) {
-      ac->Raw.Color = ctx->Array.ArrayObj->Color;
-      STRIDE_ARRAY(ac->Raw.Color, ac->start);
-   }
-   else
-      ac->Raw.Color = ac->Fallback.Color;
-
-   ac->IsCached.Color = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_COLOR0;
-}
-
-
-static void
-reset_secondarycolor( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->SecondaryColor.Enabled) {
-      ac->Raw.SecondaryColor = ctx->Array.ArrayObj->SecondaryColor;
-      STRIDE_ARRAY(ac->Raw.SecondaryColor, ac->start);
-   }
-   else
-      ac->Raw.SecondaryColor = ac->Fallback.SecondaryColor;
-
-   ac->IsCached.SecondaryColor = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_COLOR1;
-}
-
-
-static void
-reset_index( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->Index.Enabled) {
-      ac->Raw.Index = ctx->Array.ArrayObj->Index;
-      STRIDE_ARRAY(ac->Raw.Index, ac->start);
-   }
-   else
-      ac->Raw.Index = ac->Fallback.Index;
-
-   ac->IsCached.Index = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_INDEX;
-}
-
-
-static void
-reset_fogcoord( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->FogCoord.Enabled) {
-      ac->Raw.FogCoord = ctx->Array.ArrayObj->FogCoord;
-      STRIDE_ARRAY(ac->Raw.FogCoord, ac->start);
-   }
-   else
-      ac->Raw.FogCoord = ac->Fallback.FogCoord;
-
-   ac->IsCached.FogCoord = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_FOGCOORD;
-}
-
-
-static void
-reset_edgeflag( GLcontext *ctx )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->EdgeFlag.Enabled) {
-      ac->Raw.EdgeFlag = ctx->Array.ArrayObj->EdgeFlag;
-      STRIDE_ARRAY(ac->Raw.EdgeFlag, ac->start);
-   }
-   else
-      ac->Raw.EdgeFlag = ac->Fallback.EdgeFlag;
-
-   ac->IsCached.EdgeFlag = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_EDGEFLAG;
-}
-
-
-/**
- * \param index  the generic vertex array number.
- */
-static void
-reset_attrib( GLcontext *ctx, GLuint index )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (ctx->Array.ArrayObj->VertexAttrib[index].Enabled) {
-      ac->Raw.Attrib[index] = ctx->Array.ArrayObj->VertexAttrib[index];
-      STRIDE_ARRAY(ac->Raw.Attrib[index], ac->start);
-   }
-   else
-      ac->Raw.Attrib[index] = ac->Fallback.Attrib[index];
-
-   ac->IsCached.Attrib[index] = GL_FALSE;
-   ac->NewArrayState &= ~_NEW_ARRAY_ATTRIB(index);
-}
-
-
-/**
- * Generic import function for color data
- */
-static void
-import( const GLcontext *ctx,
-        GLenum destType,
-        struct gl_client_array *to,
-        const struct gl_client_array *from )
-{
-   const ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (destType == 0) 
-      destType = from->Type;
-
-   switch (destType) {
-   case GL_FLOAT:
-      _math_trans_4fn( (GLfloat (*)[4]) to->Ptr,
-                       from->Ptr,
-		       from->StrideB,
-		       from->Type,
-		       from->Size,
-		       0,
-		       ac->count - ac->start);
-
-      to->StrideB = 4 * sizeof(GLfloat);
-      to->Type = GL_FLOAT;
-      break;
-      
-   case GL_UNSIGNED_BYTE:
-      _math_trans_4ub( (GLubyte (*)[4]) to->Ptr,
-                       from->Ptr,
-		       from->StrideB,
-		       from->Type,
-		       from->Size,
-		       0,
-		       ac->count - ac->start);
-
-      to->StrideB = 4 * sizeof(GLubyte);
-      to->Type = GL_UNSIGNED_BYTE;
-      break;
-
-   case GL_UNSIGNED_SHORT:
-      _math_trans_4us( (GLushort (*)[4]) to->Ptr,
-                       from->Ptr,
-		       from->StrideB,
-		       from->Type,
-		       from->Size,
-		       0,
-		       ac->count - ac->start);
-
-      to->StrideB = 4 * sizeof(GLushort);
-      to->Type = GL_UNSIGNED_SHORT;
-      break;
-      
-   default:
-      _mesa_problem(ctx, "Unexpected dest format in import()");
-      break;
-   }
-}
-
-
-
-/*
- * Functions to import array ranges with specified types and strides.
- * For example, if the vertex data is GLshort[2] and we want GLfloat[3]
- * we'll use an import function to do the data conversion.
- */
-
-static void
-import_texcoord( GLcontext *ctx, GLuint unit, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.TexCoord[unit];
-   struct gl_client_array *to = &ac->Cache.TexCoord[unit];
-   (void) type; (void) stride;
-
-   ASSERT(unit < ctx->Const.MaxTextureCoordUnits);
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == 4*sizeof(GLfloat) || stride == 0);
-   ASSERT(ac->count - ac->start < ctx->Const.MaxArrayLockSize);
-
-   _math_trans_4f( (GLfloat (*)[4]) to->Ptr,
-		   from->Ptr,
-		   from->StrideB,
-		   from->Type,
-		   from->Size,
-		   0,
-		   ac->count - ac->start);
-
-   to->Size = from->Size;
-   to->StrideB = 4 * sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.TexCoord[unit] = GL_TRUE;
-}
-
-static void
-import_vertex( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Vertex;
-   struct gl_client_array *to = &ac->Cache.Vertex;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == 4*sizeof(GLfloat) || stride == 0);
-
-   _math_trans_4f( (GLfloat (*)[4]) to->Ptr,
-		   from->Ptr,
-		   from->StrideB,
-		   from->Type,
-		   from->Size,
-		   0,
-		   ac->count - ac->start);
-
-   to->Size = from->Size;
-   to->StrideB = 4 * sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.Vertex = GL_TRUE;
-}
-
-static void
-import_normal( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Normal;
-   struct gl_client_array *to = &ac->Cache.Normal;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == 3*sizeof(GLfloat) || stride == 0);
-
-   _math_trans_3fn((GLfloat (*)[3]) to->Ptr,
-		   from->Ptr,
-		   from->StrideB,
-		   from->Type,
-		   0,
-		   ac->count - ac->start);
-
-   to->StrideB = 3 * sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.Normal = GL_TRUE;
-}
-
-static void
-import_color( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Color;
-   struct gl_client_array *to = &ac->Cache.Color;
-   (void) stride;
-
-   import( ctx, type, to, from );
-   
-   ac->IsCached.Color = GL_TRUE;
-}
-
-static void
-import_index( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Index;
-   struct gl_client_array *to = &ac->Cache.Index;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_UNSIGNED_INT);
-   ASSERT(stride == sizeof(GLuint) || stride == 0);
-
-   _math_trans_1ui( (GLuint *) to->Ptr,
-		    from->Ptr,
-		    from->StrideB,
-		    from->Type,
-		    0,
-		    ac->count - ac->start);
-
-   to->StrideB = sizeof(GLuint);
-   to->Type = GL_UNSIGNED_INT;
-   ac->IsCached.Index = GL_TRUE;
-}
-
-static void
-import_secondarycolor( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.SecondaryColor;
-   struct gl_client_array *to = &ac->Cache.SecondaryColor;
-   (void) stride;
-
-   import( ctx, type, to, from );
-
-   ac->IsCached.SecondaryColor = GL_TRUE;
-}
-
-static void
-import_fogcoord( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.FogCoord;
-   struct gl_client_array *to = &ac->Cache.FogCoord;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == sizeof(GLfloat) || stride == 0);
-
-   _math_trans_1f( (GLfloat *) to->Ptr,
-		   from->Ptr,
-		   from->StrideB,
-		   from->Type,
-		   0,
-		   ac->count - ac->start);
-
-   to->StrideB = sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.FogCoord = GL_TRUE;
-}
-
-static void
-import_edgeflag( GLcontext *ctx, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.EdgeFlag;
-   struct gl_client_array *to = &ac->Cache.EdgeFlag;
-   (void) type; (void) stride;
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_UNSIGNED_BYTE);
-   ASSERT(stride == sizeof(GLubyte) || stride == 0);
-
-   _math_trans_1ub( (GLubyte *) to->Ptr,
-		    from->Ptr,
-		    from->StrideB,
-		    from->Type,
-		    0,
-		    ac->count - ac->start);
-
-   to->StrideB = sizeof(GLubyte);
-   to->Type = GL_UNSIGNED_BYTE;
-   ac->IsCached.EdgeFlag = GL_TRUE;
-}
-
-/**
- * \param index  the generic vertex array number
- */
-static void
-import_attrib( GLcontext *ctx, GLuint index, GLenum type, GLuint stride )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-   const struct gl_client_array *from = &ac->Raw.Attrib[index];
-   struct gl_client_array *to = &ac->Cache.Attrib[index];
-   (void) type; (void) stride;
-
-   ASSERT(index < MAX_VERTEX_PROGRAM_ATTRIBS);
-
-   /* Limited choices at this stage:
-    */
-   ASSERT(type == GL_FLOAT);
-   ASSERT(stride == 4*sizeof(GLfloat) || stride == 0);
-   ASSERT(ac->count - ac->start < ctx->Const.MaxArrayLockSize);
-
-   if (from->Normalized) {
-      _math_trans_4fn( (GLfloat (*)[4]) to->Ptr,
-                      from->Ptr,
-                      from->StrideB,
-                      from->Type,
-                      from->Size,
-                      0,
-                      ac->count - ac->start);
-   }
-   else {
-      _math_trans_4f( (GLfloat (*)[4]) to->Ptr,
-                      from->Ptr,
-                      from->StrideB,
-                      from->Type,
-                      from->Size,
-                      0,
-                      ac->count - ac->start);
-   }
-
-   to->Size = from->Size;
-   to->StrideB = 4 * sizeof(GLfloat);
-   to->Type = GL_FLOAT;
-   ac->IsCached.Attrib[index] = GL_TRUE;
-}
-
-
-
-/*
- * Externals to request arrays with specific properties:
- */
-
-
-struct gl_client_array *
-_ac_import_texcoord( GLcontext *ctx,
-                     GLuint unit,
-                     GLenum type,
-                     GLuint reqstride,
-                     GLuint reqsize,
-                     GLboolean reqwriteable,
-                     GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   ASSERT(unit < MAX_TEXTURE_COORD_UNITS);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_TEXCOORD(unit))
-      reset_texcoord( ctx, unit );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.TexCoord[unit].Size > (GLint) reqsize)
-      return NULL;
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.TexCoord[unit].Type != type ||
-       (reqstride != 0 && ac->Raw.TexCoord[unit].StrideB != (GLint)reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.TexCoord[unit])
-	 import_texcoord(ctx, unit, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.TexCoord[unit];
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.TexCoord[unit];
-   }
-}
-
-struct gl_client_array *
-_ac_import_vertex( GLcontext *ctx,
-                   GLenum type,
-                   GLuint reqstride,
-                   GLuint reqsize,
-                   GLboolean reqwriteable,
-                   GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_VERTEX)
-      reset_vertex( ctx );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.Vertex.Size > (GLint) reqsize)
-      return NULL;
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.Vertex.Type != type ||
-       (reqstride != 0 && ac->Raw.Vertex.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Vertex)
-	 import_vertex(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.Vertex;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Vertex;
-   }
-}
-
-struct gl_client_array *
-_ac_import_normal( GLcontext *ctx,
-                   GLenum type,
-                   GLuint reqstride,
-                   GLboolean reqwriteable,
-                   GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_NORMAL)
-      reset_normal( ctx );
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.Normal.Type != type ||
-       (reqstride != 0 && ac->Raw.Normal.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Normal)
-	 import_normal(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.Normal;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Normal;
-   }
-}
-
-struct gl_client_array *
-_ac_import_color( GLcontext *ctx,
-                  GLenum type,
-                  GLuint reqstride,
-                  GLuint reqsize,
-                  GLboolean reqwriteable,
-                  GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_COLOR0)
-      reset_color( ctx );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.Color.Size > (GLint) reqsize) {
-      return NULL;
-   }
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if ((type != 0 && ac->Raw.Color.Type != type) ||
-       (reqstride != 0 && ac->Raw.Color.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Color) {
-      	 import_color(ctx, type, reqstride );
-      }
-      *writeable = GL_TRUE;
-      return &ac->Cache.Color;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Color;
-   }
-}
-
-struct gl_client_array *
-_ac_import_index( GLcontext *ctx,
-                  GLenum type,
-                  GLuint reqstride,
-                  GLboolean reqwriteable,
-                  GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_INDEX)
-      reset_index( ctx );
-
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.Index.Type != type ||
-       (reqstride != 0 && ac->Raw.Index.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Index)
-	 import_index(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.Index;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Index;
-   }
-}
-
-struct gl_client_array *
-_ac_import_secondarycolor( GLcontext *ctx,
-                           GLenum type,
-                           GLuint reqstride,
-                           GLuint reqsize,
-                           GLboolean reqwriteable,
-                           GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_COLOR1)
-      reset_secondarycolor( ctx );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.SecondaryColor.Size > (GLint) reqsize)
-      return NULL;
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if ((type != 0 && ac->Raw.SecondaryColor.Type != type) ||
-       (reqstride != 0 && ac->Raw.SecondaryColor.StrideB != (GLint)reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.SecondaryColor)
-	 import_secondarycolor(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.SecondaryColor;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.SecondaryColor;
-   }
-}
-
-struct gl_client_array *
-_ac_import_fogcoord( GLcontext *ctx,
-                     GLenum type,
-                     GLuint reqstride,
-                     GLboolean reqwriteable,
-                     GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_FOGCOORD)
-      reset_fogcoord( ctx );
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.FogCoord.Type != type ||
-       (reqstride != 0 && ac->Raw.FogCoord.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.FogCoord)
-	 import_fogcoord(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.FogCoord;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.FogCoord;
-   }
-}
-
-struct gl_client_array *
-_ac_import_edgeflag( GLcontext *ctx,
-                     GLenum type,
-                     GLuint reqstride,
-                     GLboolean reqwriteable,
-                     GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_EDGEFLAG)
-      reset_edgeflag( ctx );
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.EdgeFlag.Type != type ||
-       (reqstride != 0 && ac->Raw.EdgeFlag.StrideB != (GLint) reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.EdgeFlag)
-	 import_edgeflag(ctx, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.EdgeFlag;
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.EdgeFlag;
-   }
-}
-
-/**
- * For GL_ARB/NV_vertex_program
- * \param index  index of the vertex array, starting at zero.
- */
-struct gl_client_array *
-_ac_import_attrib( GLcontext *ctx,
-                   GLuint index,
-                   GLenum type,
-                   GLuint reqstride,
-                   GLuint reqsize,
-                   GLboolean reqwriteable,
-                   GLboolean *writeable )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   ASSERT(index < VERT_ATTRIB_MAX);
-
-   /* Can we keep the existing version?
-    */
-   if (ac->NewArrayState & _NEW_ARRAY_ATTRIB(index))
-      reset_attrib( ctx, index );
-
-   /* Is the request impossible?
-    */
-   if (reqsize != 0 && ac->Raw.Attrib[index].Size > (GLint) reqsize)
-      return NULL;
-
-   /* Do we need to pull in a copy of the client data:
-    */
-   if (ac->Raw.Attrib[index].Type != type ||
-       (reqstride != 0 && ac->Raw.Attrib[index].StrideB != (GLint)reqstride) ||
-       reqwriteable)
-   {
-      if (!ac->IsCached.Attrib[index])
-	 import_attrib(ctx, index, type, reqstride );
-      *writeable = GL_TRUE;
-      return &ac->Cache.Attrib[index];
-   }
-   else {
-      *writeable = GL_FALSE;
-      return &ac->Raw.Attrib[index];
-   }
-}
-
-
-/* Clients must call this function to validate state and set bounds
- * before importing any data:
- */
-void
-_ac_import_range( GLcontext *ctx, GLuint start, GLuint count )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (!ctx->Array.LockCount) {
-      /* Not locked, discard cached data.  Changes to lock
-       * status are caught via. _ac_invalidate_state().
-       */
-      ac->NewArrayState = _NEW_ARRAY_ALL;
-      ac->start = start;
-      ac->count = count;
-   }
-   else {
-      /* Locked, discard data for any disabled arrays.  Require that
-       * the whole locked range always be dealt with, otherwise hard to
-       * maintain cached data in the face of clipping.
-       */
-      ac->NewArrayState |= ~ctx->Array.ArrayObj->_Enabled;
-      ac->start = ctx->Array.LockFirst;
-      ac->count = ctx->Array.LockCount;
-      ASSERT(ac->start == start); /* hmm? */
-      ASSERT(ac->count == count);
-   }
-}
-
-
-
-/* Additional convienence function for importing the element list
- * for glDrawElements() and glDrawRangeElements().
- */
-CONST void *
-_ac_import_elements( GLcontext *ctx,
-		     GLenum new_type,
-		     GLuint count,
-		     GLenum old_type,
-		     CONST void *indices )
-{
-   ACcontext *ac = AC_CONTEXT(ctx);
-
-   if (old_type == new_type)
-      return indices;
-
-   if (ac->elt_size < count * sizeof(GLuint)) {
-      if (ac->Elts) FREE(ac->Elts);
-      while (ac->elt_size < count * sizeof(GLuint))
-	 ac->elt_size *= 2;
-      ac->Elts = (GLuint *) MALLOC(ac->elt_size);
-   }
-
-   switch (new_type) {
-   case GL_UNSIGNED_BYTE:
-      ASSERT(0);
-      return NULL;
-   case GL_UNSIGNED_SHORT:
-      ASSERT(0);
-      return NULL;
-   case GL_UNSIGNED_INT: {
-      GLuint *out = (GLuint *)ac->Elts;
-      GLuint i;
-
-      switch (old_type) {
-      case GL_UNSIGNED_BYTE: {
-	 CONST GLubyte *in = (CONST GLubyte *)indices;
-	 for (i = 0 ; i < count ; i++)
-	    out[i] = in[i];
-	 break;
-      }
-      case GL_UNSIGNED_SHORT: {
-	 CONST GLushort *in = (CONST GLushort *)indices;
-	 for (i = 0 ; i < count ; i++)
-	       out[i] = in[i];
-	 break;
-      }
-      default:
-	 ASSERT(0);
-      }
-
-      return (CONST void *)out;
-   }
-   default:
-      ASSERT(0);
-      break;
-   }
-
-   return NULL;
-}
diff --git a/src/mesa/array_cache/acache.h b/src/mesa/array_cache/acache.h
deleted file mode 100644
index 2bb64f3656c..00000000000
--- a/src/mesa/array_cache/acache.h
+++ /dev/null
@@ -1,132 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  4.1
- *
- * Copyright (C) 1999-2002  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"),
- * 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 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
- * BRIAN PAUL 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:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef _ARRAYCACHE_H
-#define _ARRAYCACHE_H
-
-#include "mtypes.h"
-
-
-extern GLboolean
-_ac_CreateContext( GLcontext *ctx );
-
-extern void
-_ac_DestroyContext( GLcontext *ctx );
-
-extern void
-_ac_InvalidateState( GLcontext *ctx, GLuint new_state );
-
-extern struct gl_client_array *
-_ac_import_vertex( GLcontext *ctx,
-		   GLenum type,
-		   GLuint reqstride,
-		   GLuint reqsize,
-		   GLboolean reqwritable,
-		   GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_normal( GLcontext *ctx,
-		   GLenum type,
-		   GLuint reqstride,
-		   GLboolean reqwritable,
-		   GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_color( GLcontext *ctx,
-		  GLenum type,
-		  GLuint reqstride,
-		  GLuint reqsize,
-		  GLboolean reqwritable,
-		  GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_index( GLcontext *ctx,
-		  GLenum type,
-		  GLuint reqstride,
-		  GLboolean reqwritable,
-		  GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_secondarycolor( GLcontext *ctx,
-			   GLenum type,
-			   GLuint reqstride,
-			   GLuint reqsize,
-			   GLboolean reqwritable,
-			   GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_fogcoord( GLcontext *ctx,
-		     GLenum type,
-		     GLuint reqstride,
-		     GLboolean reqwritable,
-		     GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_edgeflag( GLcontext *ctx,
-		     GLenum type,
-		     GLuint reqstride,
-		     GLboolean reqwritable,
-		     GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_texcoord( GLcontext *ctx,
-		     GLuint unit,
-		     GLenum type,
-		     GLuint reqstride,
-		     GLuint reqsize,
-		     GLboolean reqwritable,
-		     GLboolean *writable );
-
-extern struct gl_client_array *
-_ac_import_attrib( GLcontext *ctx,
-                   GLuint index,
-                   GLenum type,
-                   GLuint reqstride,
-                   GLuint reqsize,
-                   GLboolean reqwritable,
-                   GLboolean *writable );
-
-
-/* Clients must call this function to validate state and set bounds
- * before importing any data:
- */
-extern void
-_ac_import_range( GLcontext *ctx, GLuint start, GLuint count );
-
-
-/* Additional convenience function:
- */
-extern CONST void *
-_ac_import_elements( GLcontext *ctx,
-		     GLenum new_type,
-		     GLuint count,
-		     GLenum old_type,
-		     CONST void *indices );
-
-
-#endif
diff --git a/src/mesa/array_cache/descrip.mms b/src/mesa/array_cache/descrip.mms
deleted file mode 100644
index 0607a01d1f9..00000000000
--- a/src/mesa/array_cache/descrip.mms
+++ /dev/null
@@ -1,37 +0,0 @@
-# Makefile for core library for VMS
-# contributed by Jouk Jansen  joukj@hrem.stm.tudelft.nl
-# Last revision : 16 June 2003
-
-.first
-	define gl [---.include.gl]
-	define math [-.math]
-	define array_cache [-.array_cache]
-
-.include [---]mms-config.
-
-##### MACROS #####
-
-VPATH = RCS
-
-INCDIR = [---.include],[-.main],[-.glapi]
-LIBDIR = [---.lib]
-CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm
-
-SOURCES = ac_context.c ac_import.c
-
-OBJECTS = ac_context.obj,ac_import.obj
-##### RULES #####
-
-VERSION=Mesa V3.4
-
-##### TARGETS #####
-# Make the library
-$(LIBDIR)$(GL_LIB) : $(OBJECTS)
-  @ library $(LIBDIR)$(GL_LIB) $(OBJECTS)
-
-clean :
-	purge
-	delete *.obj;*
-
-ac_context.obj : ac_context.c
-ac_import.obj : ac_import.c
diff --git a/src/mesa/array_cache/sources b/src/mesa/array_cache/sources
deleted file mode 100644
index fb3328d10e6..00000000000
--- a/src/mesa/array_cache/sources
+++ /dev/null
@@ -1,7 +0,0 @@
-MESA_ARRAY_CACHE_SOURCES = \
-ac_context.c \
-ac_import.c
-
-MESA_ARRAY_CACHE_HEADERS = \
-ac_context.h \
-acache.h
diff --git a/src/mesa/drivers/beos/GLView.cpp b/src/mesa/drivers/beos/GLView.cpp
index 45473a8ef27..9e4a7ebe56e 100644
--- a/src/mesa/drivers/beos/GLView.cpp
+++ b/src/mesa/drivers/beos/GLView.cpp
@@ -43,7 +43,7 @@ extern "C" {
 #include "texobj.h"
 #include "teximage.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast/s_context.h"
@@ -344,7 +344,7 @@ BGLView::BGLView(BRect rect, char *name,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext(ctx);
-   _ac_CreateContext(ctx);
+   _vbo_CreateContext(ctx);
    _tnl_CreateContext(ctx);
    _swsetup_CreateContext(ctx);
    _swsetup_Wakeup(ctx);
@@ -830,7 +830,7 @@ void MesaDriver::UpdateState( GLcontext *ctx, GLuint new_state )
 
 	_swrast_InvalidateState( ctx, new_state );
 	_swsetup_InvalidateState( ctx, new_state );
-	_ac_InvalidateState( ctx, new_state );
+	_vbo_InvalidateState( ctx, new_state );
 	_tnl_InvalidateState( ctx, new_state );
 
 	if (ctx->Color.DrawBuffer[0] == GL_FRONT) {
diff --git a/src/mesa/drivers/directfb/idirectfbgl_mesa.c b/src/mesa/drivers/directfb/idirectfbgl_mesa.c
index 3c8c6c256e6..88ac4bb31d5 100644
--- a/src/mesa/drivers/directfb/idirectfbgl_mesa.c
+++ b/src/mesa/drivers/directfb/idirectfbgl_mesa.c
@@ -1,18 +1,19 @@
 /*
- * Copyright (C) 2004-2006 Claudio Ciccani <klan@users.sf.net>
+ * Copyright (C) 2004-2007 Claudio Ciccani <klan@directfb.org>
  *
- * This program is free software; you can redistribute it and/or modify
- * it under the terms of the GNU General Public License version 2 as
- * published by the Free Software Foundation.
- *
- * This program is distributed in the hope that it will be useful,
+ * This library is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU Lesser General Public
+ * License as published by the Free Software Foundation; either
+ * version 2 of the License, or (at your option) any later version.
+ * 
+ * This library is distributed in the hope that it will be useful,
  * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
- * GNU General Public License for more details.
- *
- * You should have received a copy of the GNU General Public License
- * along with this program; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * Lesser General Public License for more details.
+ * 
+ * You should have received a copy of the GNU Lesser General Public
+ * License along with this library; if not, write to the Free Software
+ * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
  *
  * 
  * Based on glfbdev.c, written by Brian Paul.
@@ -45,7 +46,7 @@
 #include "texformat.h"
 #include "teximage.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
@@ -373,7 +374,7 @@ dfbUpdateState( GLcontext *ctx, GLuint new_state )
 {
      _swrast_InvalidateState( ctx, new_state );
      _swsetup_InvalidateState( ctx, new_state );
-     _ac_InvalidateState( ctx, new_state );
+     _vbo_InvalidateState( ctx, new_state );
      _tnl_InvalidateState( ctx, new_state );
 }
 
@@ -786,7 +787,7 @@ directfbgl_create_context( GLcontext             *context,
      }
 
      _swrast_CreateContext( context );
-     _ac_CreateContext( context );
+     _vbo_CreateContext( context );
      _tnl_CreateContext( context );
      _swsetup_CreateContext( context );
      _swsetup_Wakeup( context );
diff --git a/src/mesa/drivers/dri/common/dri_util.c b/src/mesa/drivers/dri/common/dri_util.c
index ba251a81430..cc3dcf9d8db 100644
--- a/src/mesa/drivers/dri/common/dri_util.c
+++ b/src/mesa/drivers/dri/common/dri_util.c
@@ -482,8 +482,27 @@ __driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp)
 static void driSwapBuffers( __DRInativeDisplay *dpy, void *drawablePrivate )
 {
     __DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate;
+    drm_clip_rect_t rect;
+
     dPriv->swapBuffers(dPriv);
-    (void) dpy;
+
+    /* Check that we actually have the new damage report method */
+    if (api_ver < 20070105 || dri_interface->reportDamage == NULL)
+	return;
+
+    /* Assume it's affecting the whole drawable for now */
+    rect.x1 = 0;
+    rect.y1 = 0;
+    rect.x2 = rect.x1 + dPriv->w;
+    rect.y2 = rect.y1 + dPriv->h;
+
+    /* Report the damage.  Currently, all our drivers draw directly to the
+     * front buffer, so we report the damage there rather than to the backing
+     * store (if any).
+     */
+    (*dri_interface->reportDamage)(dpy, dPriv->screen, dPriv->draw,
+				   dPriv->x, dPriv->y,
+				   &rect, 1, GL_TRUE);
 }
 
 /**
diff --git a/src/mesa/drivers/dri/common/extension_helper.h b/src/mesa/drivers/dri/common/extension_helper.h
index d6d51cdd165..c7984964255 100644
--- a/src/mesa/drivers/dri/common/extension_helper.h
+++ b/src/mesa/drivers/dri/common/extension_helper.h
@@ -2466,10 +2466,10 @@ static const char MultiTexCoord4ivARB_names[] =
     "";
 #endif
 
-#if defined(need_GL_VERSION_2_0)
-static const char GetVertexAttribPointervARB_names[] = 
-    "iip\0" /* Parameter signature */
-    "glGetVertexAttribPointerv\0"
+#if defined(need_GL_EXT_gpu_program_parameters)
+static const char ProgramLocalParameters4fvEXT_names[] = 
+    "iiip\0" /* Parameter signature */
+    "glProgramLocalParameters4fvEXT\0"
     "";
 #endif
 
@@ -3152,13 +3152,6 @@ static const char VertexAttribPointerNV_names[] =
     "";
 #endif
 
-#if defined(need_GL_EXT_gpu_program_parameters)
-static const char ProgramLocalParameters4fvEXT_names[] = 
-    "iiip\0" /* Parameter signature */
-    "glProgramLocalParameters4fvEXT\0"
-    "";
-#endif
-
 #if defined(need_GL_EXT_framebuffer_object)
 static const char GetFramebufferAttachmentParameterivEXT_names[] = 
     "iiip\0" /* Parameter signature */
@@ -3781,9 +3774,10 @@ static const char ReplacementCodeuiColor3fVertex3fSUN_names[] =
     "";
 #endif
 
-#if defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
+#if defined(need_GL_VERSION_2_0) || defined(need_GL_ARB_vertex_program) || defined(need_GL_NV_vertex_program)
 static const char GetVertexAttribPointervNV_names[] = 
     "iip\0" /* Parameter signature */
+    "glGetVertexAttribPointerv\0"
     "glGetVertexAttribPointervARB\0"
     "glGetVertexAttribPointervNV\0"
     "";
@@ -5343,8 +5337,8 @@ static const struct dri_extension_function GL_EXT_framebuffer_object_functions[]
 
 #if defined(need_GL_EXT_gpu_program_parameters)
 static const struct dri_extension_function GL_EXT_gpu_program_parameters_functions[] = {
-    { ProgramEnvParameters4fvEXT_names, ProgramEnvParameters4fvEXT_remap_index, -1 },
     { ProgramLocalParameters4fvEXT_names, ProgramLocalParameters4fvEXT_remap_index, -1 },
+    { ProgramEnvParameters4fvEXT_names, ProgramEnvParameters4fvEXT_remap_index, -1 },
     { NULL, 0, 0 }
 };
 #endif
@@ -6263,7 +6257,6 @@ static const struct dri_extension_function GL_VERSION_2_0_functions[] = {
     { GetVertexAttribfvARB_names, GetVertexAttribfvARB_remap_index, -1 },
     { GetAttribLocationARB_names, GetAttribLocationARB_remap_index, -1 },
     { Uniform3ivARB_names, Uniform3ivARB_remap_index, -1 },
-    { GetVertexAttribPointervARB_names, GetVertexAttribPointervARB_remap_index, -1 },
     { VertexAttrib4sARB_names, VertexAttrib4sARB_remap_index, -1 },
     { VertexAttrib2dvARB_names, VertexAttrib2dvARB_remap_index, -1 },
     { VertexAttrib2fvARB_names, VertexAttrib2fvARB_remap_index, -1 },
@@ -6295,6 +6288,7 @@ static const struct dri_extension_function GL_VERSION_2_0_functions[] = {
     { Uniform4iARB_names, Uniform4iARB_remap_index, -1 },
     { UseProgramObjectARB_names, UseProgramObjectARB_remap_index, -1 },
     { DeleteProgram_names, DeleteProgram_remap_index, -1 },
+    { GetVertexAttribPointervNV_names, GetVertexAttribPointervNV_remap_index, -1 },
     { Uniform2iARB_names, Uniform2iARB_remap_index, -1 },
     { VertexAttrib4dARB_names, VertexAttrib4dARB_remap_index, -1 },
     { GetUniformLocationARB_names, GetUniformLocationARB_remap_index, -1 },
diff --git a/src/mesa/drivers/dri/fb/fb_dri.c b/src/mesa/drivers/dri/fb/fb_dri.c
index d6ba23bc6e6..08b52b4d8f7 100644
--- a/src/mesa/drivers/dri/fb/fb_dri.c
+++ b/src/mesa/drivers/dri/fb/fb_dri.c
@@ -50,7 +50,7 @@
 #include "extensions.h"
 #include "framebuffer.h"
 #include "renderbuffer.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
@@ -93,7 +93,7 @@ update_state( GLcontext *ctx, GLuint new_state )
    /* not much to do here - pass it on */
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
 }
 
@@ -365,7 +365,7 @@ fbCreateContext( const __GLcontextModes *glVisual,
 
    /* Create module contexts */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
    _swsetup_Wakeup( ctx );
@@ -399,7 +399,7 @@ fbDestroyContext( __DRIcontextPrivate *driContextPriv )
    if ( fbmesa ) {
       _swsetup_DestroyContext( fbmesa->glCtx );
       _tnl_DestroyContext( fbmesa->glCtx );
-      _ac_DestroyContext( fbmesa->glCtx );
+      _vbo_DestroyContext( fbmesa->glCtx );
       _swrast_DestroyContext( fbmesa->glCtx );
 
       /* free the Mesa context */
diff --git a/src/mesa/drivers/dri/fb/fb_egl.c b/src/mesa/drivers/dri/fb/fb_egl.c
index 5c74b0b0f47..517e71f8880 100644
--- a/src/mesa/drivers/dri/fb/fb_egl.c
+++ b/src/mesa/drivers/dri/fb/fb_egl.c
@@ -17,7 +17,7 @@
 #include "extensions.h"
 #include "framebuffer.h"
 #include "renderbuffer.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
@@ -388,7 +388,7 @@ update_state( GLcontext *ctx, GLuint new_state )
    /* not much to do here - pass it on */
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
 }
 
@@ -491,7 +491,7 @@ fbCreateContext(_EGLDriver *drv, EGLDisplay dpy, EGLConfig config, EGLContext sh
 
    /* Create module contexts */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
    _swsetup_Wakeup( ctx );
diff --git a/src/mesa/drivers/dri/ffb/ffb_state.c b/src/mesa/drivers/dri/ffb/ffb_state.c
index b81d94de252..eb13478166a 100644
--- a/src/mesa/drivers/dri/ffb/ffb_state.c
+++ b/src/mesa/drivers/dri/ffb/ffb_state.c
@@ -40,7 +40,7 @@
 #include "enums.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1034,7 +1034,7 @@ static void ffbDDUpdateState(GLcontext *ctx, GLuint newstate)
 
 	_swrast_InvalidateState( ctx, newstate );
 	_swsetup_InvalidateState( ctx, newstate );
-	_ac_InvalidateState( ctx, newstate );
+	_vbo_InvalidateState( ctx, newstate );
 	_tnl_InvalidateState( ctx, newstate );
 
 	if (newstate & _NEW_TEXTURE)
diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.c b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
index 99256d76b6e..215aaf8ffba 100644
--- a/src/mesa/drivers/dri/ffb/ffb_xmesa.c
+++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
@@ -38,7 +38,7 @@
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "drivers/common/driverfuncs.h"
 
 #include "ffb_context.h"
@@ -277,7 +277,7 @@ ffbCreateContext(const __GLcontextModes *mesaVis,
 
 	/* Initialize the software rasterizer and helper modules. */
 	_swrast_CreateContext( ctx );
-	_ac_CreateContext( ctx );
+	_vbo_CreateContext( ctx );
 	_tnl_CreateContext( ctx );
 	_swsetup_CreateContext( ctx );
 
@@ -313,7 +313,7 @@ ffbDestroyContext(__DRIcontextPrivate *driContextPriv)
 
 		_swsetup_DestroyContext( fmesa->glCtx );
 		_tnl_DestroyContext( fmesa->glCtx );
-		_ac_DestroyContext( fmesa->glCtx );
+		_vbo_DestroyContext( fmesa->glCtx );
 		_swrast_DestroyContext( fmesa->glCtx );
 
                 /* free the Mesa context */
diff --git a/src/mesa/drivers/dri/gamma/gamma_context.c b/src/mesa/drivers/dri/gamma/gamma_context.c
index ffaf45459bc..b1dcbfcdcfa 100644
--- a/src/mesa/drivers/dri/gamma/gamma_context.c
+++ b/src/mesa/drivers/dri/gamma/gamma_context.c
@@ -28,7 +28,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -147,7 +147,7 @@ GLboolean gammaCreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
diff --git a/src/mesa/drivers/dri/gamma/gamma_state.c b/src/mesa/drivers/dri/gamma/gamma_state.c
index 08749bf5d39..8dbe0a97caa 100644
--- a/src/mesa/drivers/dri/gamma/gamma_state.c
+++ b/src/mesa/drivers/dri/gamma/gamma_state.c
@@ -33,7 +33,7 @@
 #include "colormac.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 
 #define ENABLELIGHTING 0
@@ -1663,7 +1663,7 @@ static void gammaDDUpdateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    GAMMA_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/gamma/gamma_xmesa.c b/src/mesa/drivers/dri/gamma/gamma_xmesa.c
index 00f6aa32aeb..e8922b15036 100644
--- a/src/mesa/drivers/dri/gamma/gamma_xmesa.c
+++ b/src/mesa/drivers/dri/gamma/gamma_xmesa.c
@@ -34,7 +34,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 static GLboolean 
 gammaInitDriver(__DRIscreenPrivate *sPriv)
@@ -57,7 +57,7 @@ gammaDestroyContext(__DRIcontextPrivate *driContextPriv)
     if (gmesa) {
       _swsetup_DestroyContext( gmesa->glCtx );
       _tnl_DestroyContext( gmesa->glCtx );
-      _ac_DestroyContext( gmesa->glCtx );
+      _vbo_DestroyContext( gmesa->glCtx );
       _swrast_DestroyContext( gmesa->glCtx );
 
       gammaFreeVB( gmesa->glCtx );
diff --git a/src/mesa/drivers/dri/i810/i810context.c b/src/mesa/drivers/dri/i810/i810context.c
index c0f4efd7a31..db8f7a19a23 100644
--- a/src/mesa/drivers/dri/i810/i810context.c
+++ b/src/mesa/drivers/dri/i810/i810context.c
@@ -44,7 +44,7 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 
@@ -287,7 +287,7 @@ i810CreateContext( const __GLcontextModes *mesaVis,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -350,7 +350,7 @@ i810DestroyContext(__DRIcontextPrivate *driContextPriv)
       release_texture_heaps = (imesa->glCtx->Shared->RefCount == 1);
       _swsetup_DestroyContext( imesa->glCtx );
       _tnl_DestroyContext( imesa->glCtx );
-      _ac_DestroyContext( imesa->glCtx );
+      _vbo_DestroyContext( imesa->glCtx );
       _swrast_DestroyContext( imesa->glCtx );
 
       i810FreeVB( imesa->glCtx );
diff --git a/src/mesa/drivers/dri/i810/i810state.c b/src/mesa/drivers/dri/i810/i810state.c
index 6f9d4b5bd47..3ad25282d90 100644
--- a/src/mesa/drivers/dri/i810/i810state.c
+++ b/src/mesa/drivers/dri/i810/i810state.c
@@ -21,8 +21,8 @@
 #include "i810ioctl.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
 #include "tnl/tnl.h"
+#include "vbo/vbo.h"
 #include "swrast_setup/swrast_setup.h"
 
 #include "tnl/t_pipeline.h"
@@ -953,7 +953,7 @@ static void i810InvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    I810_CONTEXT(ctx)->new_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/i915/i915_context.c b/src/mesa/drivers/dri/i915/i915_context.c
index b2e9ba31874..e58c9498ce0 100644
--- a/src/mesa/drivers/dri/i915/i915_context.c
+++ b/src/mesa/drivers/dri/i915/i915_context.c
@@ -36,7 +36,8 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
+
 
 #include "utils.h"
 #include "i915_reg.h"
@@ -63,7 +64,7 @@ static void i915InvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _tnl_invalidate_vertex_state( ctx, new_state );
    INTEL_CONTEXT(ctx)->NewGLState |= new_state;
diff --git a/src/mesa/drivers/dri/i915/i915_state.c b/src/mesa/drivers/dri/i915/i915_state.c
index 3cec6a2ddfc..fd11e10652e 100644
--- a/src/mesa/drivers/dri/i915/i915_state.c
+++ b/src/mesa/drivers/dri/i915/i915_state.c
@@ -611,10 +611,12 @@ void i915_update_fog( GLcontext *ctx )
 	 i915->state.Ctx[I915_CTXREG_LIS5] &= ~S5_FOG_ENABLE;
    }
 
-   if (enabled) {
-      _tnl_allow_vertex_fog( ctx, (i915->vertex_fog == I915_FOG_VERTEX) ); 
-      _tnl_allow_pixel_fog( ctx, (i915->vertex_fog != I915_FOG_VERTEX) ); 
-   }
+   /* always enbale pixel fog
+    * vertex fog use precaculted fog coord will conflict with appended
+    * fog program
+    */
+    _tnl_allow_vertex_fog( ctx, 0 );
+    _tnl_allow_pixel_fog( ctx, 1 );
 }
 
 static void i915Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *param)
diff --git a/src/mesa/drivers/dri/i915/intel_context.c b/src/mesa/drivers/dri/i915/intel_context.c
index 0a2e33ffd53..7bb00d98339 100644
--- a/src/mesa/drivers/dri/i915/intel_context.c
+++ b/src/mesa/drivers/dri/i915/intel_context.c
@@ -37,7 +37,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 #include "tnl/t_vertex.h"
@@ -228,7 +228,7 @@ static void intelInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _tnl_invalidate_vertex_state( ctx, new_state );
    INTEL_CONTEXT(ctx)->NewGLState |= new_state;
@@ -304,7 +304,7 @@ GLboolean intelInitContext( intelContextPtr intel,
 
    /* Initialize the software rasterizer and helper modules. */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -423,7 +423,7 @@ void intelDestroyContext(__DRIcontextPrivate *driContextPriv)
       release_texture_heaps = (intel->ctx.Shared->RefCount == 1);
       _swsetup_DestroyContext (&intel->ctx);
       _tnl_DestroyContext (&intel->ctx);
-      _ac_DestroyContext (&intel->ctx);
+      _vbo_DestroyContext (&intel->ctx);
 
       _swrast_DestroyContext (&intel->ctx);
       intel->Fallback = 0;	/* don't call _swrast_Flush later */
diff --git a/src/mesa/drivers/dri/i915tex/Makefile b/src/mesa/drivers/dri/i915tex/Makefile
index 94879d209f7..3b3f3f5a3f7 100644
--- a/src/mesa/drivers/dri/i915tex/Makefile
+++ b/src/mesa/drivers/dri/i915tex/Makefile
@@ -20,6 +20,7 @@ DRIVER_SOURCES = \
 	intel_batchbuffer.c \
 	intel_mipmap_tree.c \
 	i915_tex_layout.c \
+	intel_tex_layout.c \
 	intel_tex_image.c \
 	intel_tex_subimage.c \
 	intel_tex_copy.c \
@@ -59,8 +60,10 @@ C_SOURCES = \
 
 ASM_SOURCES = 
 
-
+DRIVER_DEFINES = -I../intel
 
 include ../Makefile.template
 
+intel_tex_layout.o: ../intel/intel_tex_layout.c
+
 symlinks:
diff --git a/src/mesa/drivers/dri/i915tex/i915_context.c b/src/mesa/drivers/dri/i915tex/i915_context.c
index 4cbe29d79d6..9b4d72eab3e 100644
--- a/src/mesa/drivers/dri/i915tex/i915_context.c
+++ b/src/mesa/drivers/dri/i915tex/i915_context.c
@@ -36,7 +36,6 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
 
 #include "utils.h"
 #include "i915_reg.h"
@@ -67,7 +66,7 @@ i915InvalidateState(GLcontext * ctx, GLuint new_state)
 {
    _swrast_InvalidateState(ctx, new_state);
    _swsetup_InvalidateState(ctx, new_state);
-   _ac_InvalidateState(ctx, new_state);
+   _vbo_InvalidateState(ctx, new_state);
    _tnl_InvalidateState(ctx, new_state);
    _tnl_invalidate_vertex_state(ctx, new_state);
    intel_context(ctx)->NewGLState |= new_state;
diff --git a/src/mesa/drivers/dri/i915tex/i915_context.h b/src/mesa/drivers/dri/i915tex/i915_context.h
index 5ae76fcd185..d2713e88f96 100644
--- a/src/mesa/drivers/dri/i915tex/i915_context.h
+++ b/src/mesa/drivers/dri/i915tex/i915_context.h
@@ -243,7 +243,6 @@ struct i915_context
    GLuint lodbias_ss2[MAX_TEXTURE_UNITS];
 
 
-   struct i915_fragment_program tex_program;
    struct i915_fragment_program *current_program;
 
    struct i915_hw_state meta, initial, state, *current;
diff --git a/src/mesa/drivers/dri/i915tex/i915_state.c b/src/mesa/drivers/dri/i915tex/i915_state.c
index 2f5a30787ee..78ae4bdb5fc 100644
--- a/src/mesa/drivers/dri/i915tex/i915_state.c
+++ b/src/mesa/drivers/dri/i915tex/i915_state.c
@@ -520,7 +520,6 @@ update_specular(GLcontext * ctx)
    /* A hack to trigger the rebuild of the fragment program.
     */
    intel_context(ctx)->NewGLState |= _NEW_TEXTURE;
-   I915_CONTEXT(ctx)->tex_program.translated = 0;
 }
 
 static void
@@ -646,10 +645,12 @@ i915_update_fog(GLcontext * ctx)
          i915->state.Ctx[I915_CTXREG_LIS5] &= ~S5_FOG_ENABLE;
    }
 
-   if (enabled) {
-      _tnl_allow_vertex_fog(ctx, (i915->vertex_fog == I915_FOG_VERTEX));
-      _tnl_allow_pixel_fog(ctx, (i915->vertex_fog != I915_FOG_VERTEX));
-   }
+    /* always enbale pixel fog
+    * vertex fog use precaculted fog coord will conflict with appended
+    * fog program
+    */
+    _tnl_allow_vertex_fog( ctx, 0 );
+    _tnl_allow_pixel_fog( ctx, 1 );
 }
 
 static void
diff --git a/src/mesa/drivers/dri/i915tex/i915_tex.c b/src/mesa/drivers/dri/i915tex/i915_tex.c
index a53abe9a92f..59e148ca04e 100644
--- a/src/mesa/drivers/dri/i915tex/i915_tex.c
+++ b/src/mesa/drivers/dri/i915tex/i915_tex.c
@@ -52,27 +52,6 @@ i915TexEnv(GLcontext * ctx, GLenum target,
    struct i915_context *i915 = I915_CONTEXT(ctx);
 
    switch (pname) {
-   case GL_TEXTURE_ENV_COLOR:  /* Should be a tracked param */
-   case GL_TEXTURE_ENV_MODE:
-   case GL_COMBINE_RGB:
-   case GL_COMBINE_ALPHA:
-   case GL_SOURCE0_RGB:
-   case GL_SOURCE1_RGB:
-   case GL_SOURCE2_RGB:
-   case GL_SOURCE0_ALPHA:
-   case GL_SOURCE1_ALPHA:
-   case GL_SOURCE2_ALPHA:
-   case GL_OPERAND0_RGB:
-   case GL_OPERAND1_RGB:
-   case GL_OPERAND2_RGB:
-   case GL_OPERAND0_ALPHA:
-   case GL_OPERAND1_ALPHA:
-   case GL_OPERAND2_ALPHA:
-   case GL_RGB_SCALE:
-   case GL_ALPHA_SCALE:
-      i915->tex_program.translated = 0;
-      break;
-
    case GL_TEXTURE_LOD_BIAS:{
          GLuint unit = ctx->Texture.CurrentUnit;
          GLint b = (int) ((*param) * 16.0);
@@ -92,22 +71,8 @@ i915TexEnv(GLcontext * ctx, GLenum target,
 }
 
 
-static void
-i915BindTexture(GLcontext * ctx, GLenum target,
-                struct gl_texture_object *texobj)
-{
-   /* Need this if image format changes between bound textures.
-    * Could try and shortcircuit by checking for differences in
-    * state between incoming and outgoing textures:
-    */
-   I915_CONTEXT(ctx)->tex_program.translated = 0;
-}
-
-
-
 void
 i915InitTextureFuncs(struct dd_function_table *functions)
 {
-   functions->BindTexture = i915BindTexture;
    functions->TexEnv = i915TexEnv;
 }
diff --git a/src/mesa/drivers/dri/i915tex/i915_tex_layout.c b/src/mesa/drivers/dri/i915tex/i915_tex_layout.c
index e9360ecea83..333fefef859 100644
--- a/src/mesa/drivers/dri/i915tex/i915_tex_layout.c
+++ b/src/mesa/drivers/dri/i915tex/i915_tex_layout.c
@@ -30,6 +30,7 @@
  */
 
 #include "intel_mipmap_tree.h"
+#include "intel_tex_layout.h"
 #include "macros.h"
 #include "intel_context.h"
 
@@ -52,12 +53,6 @@ static GLint step_offsets[6][2] = { {0, 2},
 {-1, 1}
 };
 
-static GLuint
-minify(GLuint d)
-{
-   return MAX2(1, d >> 1);
-}
-
 GLboolean
 i915_miptree_layout(struct intel_mipmap_tree * mt)
 {
@@ -217,7 +212,7 @@ i945_miptree_layout(struct intel_mipmap_tree * mt)
                y = mt->total_height - 4;
                x = (face - 4) * 8;
             }
-            else if (dim < 4) {
+            else if (dim < 4 && (face > 0 || mt->first_level > 0)) {
                y = mt->total_height - 4;
                x = face * 8;
             }
@@ -322,52 +317,9 @@ i945_miptree_layout(struct intel_mipmap_tree * mt)
 
    case GL_TEXTURE_1D:
    case GL_TEXTURE_2D:
-   case GL_TEXTURE_RECTANGLE_ARB:{
-         GLuint x = 0;
-         GLuint y = 0;
-         GLuint width = mt->width0;
-         GLuint height = mt->height0;
-	 GLint align_h = 2;
-
-         mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp;
-         mt->total_height = 0;
-
-         for (level = mt->first_level; level <= mt->last_level; level++) {
-	    GLuint img_height;
-
-            intel_miptree_set_level_info(mt, level, 1,
-                                         x, y, 
-					 width, 
-					 mt->compressed ? height/4 : height, 1);
-
-
-	    if (mt->compressed)
-               img_height = MAX2(1, height / 4);
-            else
-               img_height = MAX2(align_h, height);
-
-            /* LPT change: step right after second mipmap.
-             */
-            if (level == mt->first_level + 1) {
-               x += mt->pitch / 2;
-	       x = (x + 3) & ~3;
-	    }
-            else {
-	       y += img_height;
-	       y += align_h - 1;
-	       y &= ~(align_h - 1);
-	    }
-
-            /* Because the images are packed better, the final offset
-             * might not be the maximal one:
-             */
-            mt->total_height = MAX2(mt->total_height, y);
-
-            width = minify(width);
-            height = minify(height);
-         }
+   case GL_TEXTURE_RECTANGLE_ARB:
+         i945_miptree_layout_2d(mt);
          break;
-      }
    default:
       _mesa_problem(NULL, "Unexpected tex target in i945_miptree_layout()");
    }
diff --git a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c
index b4e0b74f161..c92b83bcb3a 100644
--- a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c
+++ b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.c
@@ -252,6 +252,7 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch)
 {
    struct intel_context *intel = batch->intel;
    GLuint used = batch->ptr - batch->map;
+   GLboolean was_locked = intel->locked;
 
    if (used == 0)
       return batch->last_fence;
@@ -278,17 +279,14 @@ intel_batchbuffer_flush(struct intel_batchbuffer *batch)
    /* TODO: Just pass the relocation list and dma buffer up to the
     * kernel.
     */
-   if (!intel->locked) {
-      assert(!(batch->flags & INTEL_BATCH_NO_CLIPRECTS));
-
+   if (!was_locked)
       LOCK_HARDWARE(intel);
-      do_flush_locked(batch, used, GL_FALSE, GL_TRUE);
+
+   do_flush_locked(batch, used, !(batch->flags & INTEL_BATCH_CLIPRECTS),
+		   GL_FALSE);
+
+   if (!was_locked)
       UNLOCK_HARDWARE(intel);
-   }
-   else {
-      GLboolean ignore_cliprects = !(batch->flags & INTEL_BATCH_CLIPRECTS);
-      do_flush_locked(batch, used, ignore_cliprects, GL_FALSE);
-   }
 
    /* Reset the buffer:
     */
@@ -313,7 +311,7 @@ intel_batchbuffer_emit_reloc(struct intel_batchbuffer *batch,
                              struct _DriBufferObject *buffer,
                              GLuint flags, GLuint mask, GLuint delta)
 {
-   assert(batch->nr_relocs <= MAX_RELOCS);
+   assert(batch->nr_relocs < MAX_RELOCS);
 
    driBOAddListItem(&batch->list, buffer, flags, mask);
 
diff --git a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h
index a83dbf423df..59261f72741 100644
--- a/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h
+++ b/src/mesa/drivers/dri/i915tex/intel_batchbuffer.h
@@ -9,7 +9,7 @@ struct intel_context;
 #define BATCH_SZ 16384
 #define BATCH_RESERVED 16
 
-#define MAX_RELOCS 100
+#define MAX_RELOCS 400
 
 #define INTEL_BATCH_NO_CLIPRECTS 0x1
 #define INTEL_BATCH_CLIPRECTS    0x2
diff --git a/src/mesa/drivers/dri/i915tex/intel_buffer_objects.c b/src/mesa/drivers/dri/i915tex/intel_buffer_objects.c
index 31c41d86857..ba3c7f0c1f7 100644
--- a/src/mesa/drivers/dri/i915tex/intel_buffer_objects.c
+++ b/src/mesa/drivers/dri/i915tex/intel_buffer_objects.c
@@ -50,7 +50,10 @@ intel_bufferobj_alloc(GLcontext * ctx, GLuint name, GLenum target)
    _mesa_initialize_buffer_object(&obj->Base, name, target);
 
    driGenBuffers(intel->intelScreen->regionPool,
-                 "bufferobj", 1, &obj->buffer, 64, 0, 0);
+                 "bufferobj", 1, &obj->buffer, 64,
+		 DRM_BO_FLAG_MEM_LOCAL |
+		 DRM_BO_FLAG_READ | DRM_BO_FLAG_WRITE,
+		 0);
 
    return &obj->Base;
 }
diff --git a/src/mesa/drivers/dri/i915tex/intel_context.c b/src/mesa/drivers/dri/i915tex/intel_context.c
index c77d365360d..a5ce08b1701 100644
--- a/src/mesa/drivers/dri/i915tex/intel_context.c
+++ b/src/mesa/drivers/dri/i915tex/intel_context.c
@@ -37,7 +37,6 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
 
 #include "tnl/t_pipeline.h"
 #include "tnl/t_vertex.h"
@@ -241,7 +240,7 @@ intelInvalidateState(GLcontext * ctx, GLuint new_state)
 {
    _swrast_InvalidateState(ctx, new_state);
    _swsetup_InvalidateState(ctx, new_state);
-   _ac_InvalidateState(ctx, new_state);
+   _vbo_InvalidateState(ctx, new_state);
    _tnl_InvalidateState(ctx, new_state);
    _tnl_invalidate_vertex_state(ctx, new_state);
    intel_context(ctx)->NewGLState |= new_state;
@@ -357,6 +356,10 @@ intelInitContext(struct intel_context *intel,
    intel->driScreen = sPriv;
    intel->sarea = saPriv;
 
+   intel->width = intelScreen->width;
+   intel->height = intelScreen->height;
+   intel->current_rotation = intelScreen->current_rotation;
+
    if (!lockMutexInit) {
       lockMutexInit = GL_TRUE;
       _glthread_INIT_MUTEX(lockMutex);
@@ -390,7 +393,7 @@ intelInitContext(struct intel_context *intel,
 
    /* Initialize the software rasterizer and helper modules. */
    _swrast_CreateContext(ctx);
-   _ac_CreateContext(ctx);
+   _vbo_CreateContext(ctx);
    _tnl_CreateContext(ctx);
    _swsetup_CreateContext(ctx);
 
@@ -500,7 +503,7 @@ intelDestroyContext(__DRIcontextPrivate * driContextPriv)
       release_texture_heaps = (intel->ctx.Shared->RefCount == 1);
       _swsetup_DestroyContext(&intel->ctx);
       _tnl_DestroyContext(&intel->ctx);
-      _ac_DestroyContext(&intel->ctx);
+      _vbo_DestroyContext(&intel->ctx);
 
       _swrast_DestroyContext(&intel->ctx);
       intel->Fallback = 0;      /* don't call _swrast_Flush later */
@@ -635,14 +638,35 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
        sarea->rotation != intelScreen->current_rotation) {
 
       intelUpdateScreenRotation(sPriv, sarea);
+   }
 
-      /* 
+   if (sarea->width != intel->width ||
+       sarea->height != intel->height ||
+       sarea->rotation != intel->current_rotation) {
+      
+      void *batchMap = intel->batch->map;
+      
+      /*
+       * FIXME: Really only need to do this when drawing to a
+       * common back- or front buffer.
+       */
+
+      /*
        * This will drop the outstanding batchbuffer on the floor
-       * FIXME: This should be done for all contexts?
        */
 
+      if (batchMap != NULL) {
+	 driBOUnmap(intel->batch->buffer);
+	 intel->batch->map = NULL;
+      }
+
       intel_batchbuffer_reset(intel->batch);
 
+      if (batchMap == NULL) {
+	 driBOUnmap(intel->batch->buffer);
+	 intel->batch->map = NULL;
+      }
+
       /* lose all primitives */
       intel->prim.primitive = ~0;
       intel->prim.start_ptr = 0;
@@ -653,6 +677,10 @@ intelContendedLock(struct intel_context *intel, GLuint flags)
 
       /* force window update */
       intel->lastStamp = 0;
+
+      intel->width = sarea->width;
+      intel->height = sarea->height;
+      intel->current_rotation = sarea->rotation;
    }
 
 
diff --git a/src/mesa/drivers/dri/i915tex/intel_context.h b/src/mesa/drivers/dri/i915tex/intel_context.h
index 7654e4ecd51..96b911501f3 100644
--- a/src/mesa/drivers/dri/i915tex/intel_context.h
+++ b/src/mesa/drivers/dri/i915tex/intel_context.h
@@ -286,6 +286,15 @@ struct intel_context
    GLuint swap_missed_count;
 
    GLuint swap_scheduled;
+
+  /* Rotation. Need to match that of the
+   * current screen.
+   */
+
+  int width;
+  int height;
+  int current_rotation;
+
 };
 
 /* These are functions now:
diff --git a/src/mesa/drivers/dri/i915tex/intel_screen.c b/src/mesa/drivers/dri/i915tex/intel_screen.c
index 9bbfabbb8c3..efa1b014a61 100644
--- a/src/mesa/drivers/dri/i915tex/intel_screen.c
+++ b/src/mesa/drivers/dri/i915tex/intel_screen.c
@@ -737,6 +737,9 @@ intelFillInModes(unsigned pixel_bits, unsigned depth_bits,
     */
    stencil_bits_array[0] = 0;
    stencil_bits_array[1] = 0;
+   if (depth_bits == 24)
+      stencil_bits_array[1] = (stencil_bits == 0) ? 8 : stencil_bits;
+
    stencil_bits_array[2] = (stencil_bits == 0) ? 8 : stencil_bits;
 
    depth_buffer_factor = ((depth_bits != 0) || (stencil_bits != 0)) ? 3 : 1;
diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_layout.c b/src/mesa/drivers/dri/i915tex/intel_tex_layout.c
new file mode 120000
index 00000000000..fe61b441945
--- /dev/null
+++ b/src/mesa/drivers/dri/i915tex/intel_tex_layout.c
@@ -0,0 +1 @@
+../intel/intel_tex_layout.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_subimage.c b/src/mesa/drivers/dri/i915tex/intel_tex_subimage.c
index 25a2dca6857..3935787806b 100644
--- a/src/mesa/drivers/dri/i915tex/intel_tex_subimage.c
+++ b/src/mesa/drivers/dri/i915tex/intel_tex_subimage.c
@@ -50,7 +50,6 @@ intelTexSubimage(GLcontext * ctx,
 {
    struct intel_context *intel = intel_context(ctx);
    struct intel_texture_image *intelImage = intel_texture_image(texImage);
-   GLuint dstImageStride;
    GLuint dstRowStride;
 
    DBG("%s target %s level %d offset %d,%d %dx%d\n", __FUNCTION__,
@@ -79,7 +78,7 @@ intelTexSubimage(GLcontext * ctx,
                                                intelImage->face,
                                                intelImage->level,
                                                &dstRowStride,
-                                               &dstImageStride);
+                                               texImage->ImageOffsets);
 
    assert(dstRowStride);
 
diff --git a/src/mesa/drivers/dri/i915tex/intel_tex_validate.c b/src/mesa/drivers/dri/i915tex/intel_tex_validate.c
index 5f82dfa19e6..79d587a1744 100644
--- a/src/mesa/drivers/dri/i915tex/intel_tex_validate.c
+++ b/src/mesa/drivers/dri/i915tex/intel_tex_validate.c
@@ -2,6 +2,7 @@
 #include "macros.h"
 
 #include "intel_context.h"
+#include "intel_batchbuffer.h"
 #include "intel_mipmap_tree.h"
 #include "intel_tex.h"
 
@@ -109,6 +110,8 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
    GLuint nr_faces = 0;
    struct intel_texture_image *firstImage;
 
+   GLboolean need_flush = GL_FALSE;
+
    /* We know/require this is true by now: 
     */
    assert(intelObj->base.Complete);
@@ -155,9 +158,15 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
     * leaving the tree alone.
     */
    if (intelObj->mt &&
-       ((intelObj->mt->first_level > intelObj->firstLevel) ||
-        (intelObj->mt->last_level < intelObj->lastLevel) ||
-        (intelObj->mt->internal_format != firstImage->base.InternalFormat))) {
+       (intelObj->mt->target != intelObj->base.Target ||
+	intelObj->mt->internal_format != firstImage->base.InternalFormat ||
+	intelObj->mt->first_level != intelObj->firstLevel ||
+	intelObj->mt->last_level != intelObj->lastLevel ||
+	intelObj->mt->width0 != firstImage->base.Width ||
+	intelObj->mt->height0 != firstImage->base.Height ||
+	intelObj->mt->depth0 != firstImage->base.Depth ||
+	intelObj->mt->cpp != firstImage->base.TexFormat->TexelBytes ||
+	intelObj->mt->compressed != firstImage->base.IsCompressed)) {
       intel_miptree_release(intel, &intelObj->mt);
    }
 
@@ -194,10 +203,14 @@ intel_finalize_mipmap_tree(struct intel_context *intel, GLuint unit)
           */
          if (intelObj->mt != intelImage->mt) {
             copy_image_data_to_tree(intel, intelObj, intelImage);
+	    need_flush = GL_TRUE;
          }
       }
    }
 
+   if (need_flush)
+      intel_batchbuffer_flush(intel->batch);
+
    return GL_TRUE;
 }
 
diff --git a/src/mesa/drivers/dri/i965/Makefile b/src/mesa/drivers/dri/i965/Makefile
index 213eac895cb..9e4ff112dc3 100644
--- a/src/mesa/drivers/dri/i965/Makefile
+++ b/src/mesa/drivers/dri/i965/Makefile
@@ -20,6 +20,7 @@ DRIVER_SOURCES = \
 	intel_pixel_bitmap.c \
 	intel_state.c \
 	intel_tex.c \
+	intel_tex_layout.c \
 	intel_tex_validate.c \
 	brw_aub.c \
 	brw_aub_playback.c \
@@ -34,17 +35,11 @@ DRIVER_SOURCES = \
 	brw_context.c \
 	brw_curbe.c \
 	brw_draw.c \
-	brw_draw_current.c \
 	brw_draw_upload.c \
 	brw_eu.c \
 	brw_eu_debug.c \
 	brw_eu_emit.c \
 	brw_eu_util.c \
-	brw_exec.c \
-	brw_exec_api.c \
-	brw_exec_array.c \
-	brw_exec_draw.c \
-	brw_exec_eval.c \
 	brw_fallback.c \
 	brw_gs.c \
 	brw_gs_emit.c \
@@ -53,9 +48,6 @@ DRIVER_SOURCES = \
 	brw_metaops.c \
 	brw_misc_state.c \
 	brw_program.c \
-	brw_save.c \
-	brw_save_api.c \
-	brw_save_draw.c \
 	brw_sf.c \
 	brw_sf_emit.c \
 	brw_sf_state.c \
@@ -92,8 +84,10 @@ C_SOURCES = \
 
 ASM_SOURCES = 
 
-
+DRIVER_DEFINES = -I../intel
 
 include ../Makefile.template
 
+intel_tex_layout.o: ../intel/intel_tex_layout.c
+
 symlinks:
diff --git a/src/mesa/drivers/dri/i965/brw_clip.c b/src/mesa/drivers/dri/i965/brw_clip.c
index 0e8591aaa8d..3bec153075a 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.c
+++ b/src/mesa/drivers/dri/i965/brw_clip.c
@@ -62,6 +62,8 @@ static void compile_clip_prog( struct brw_context *brw,
     */
    brw_init_compile(&c.func);
 
+   c.func.single_program_flow = 1;
+
    c.key = *key;
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_clip.h b/src/mesa/drivers/dri/i965/brw_clip.h
index c3967c8c614..49b2770a514 100644
--- a/src/mesa/drivers/dri/i965/brw_clip.h
+++ b/src/mesa/drivers/dri/i965/brw_clip.h
@@ -116,7 +116,7 @@ struct brw_clip_compile {
    GLuint last_mrf;
 
    GLuint header_position_offset;
-   GLuint offset[BRW_ATTRIB_MAX];
+   GLuint offset[VERT_ATTRIB_MAX];
 };
 
 #define ATTR_SIZE  (4*4)
diff --git a/src/mesa/drivers/dri/i965/brw_context.c b/src/mesa/drivers/dri/i965/brw_context.c
index bc422c1a50b..6faee65542b 100644
--- a/src/mesa/drivers/dri/i965/brw_context.c
+++ b/src/mesa/drivers/dri/i965/brw_context.c
@@ -34,8 +34,6 @@
 #include "brw_aub.h"
 #include "brw_defines.h"
 #include "brw_draw.h"
-#include "brw_exec.h"
-#include "brw_save.h"
 #include "brw_vs.h"
 #include "imports.h"
 #include "intel_tex.h"
@@ -158,12 +156,6 @@ GLboolean brwCreateContext( const __GLcontextModes *mesaVis,
 
    brw_FrameBufferTexInit( brw );
 
-   /* Hook our functions into exec and compile dispatch tables.  Only
-    * fallback on out-of-memory situations.
-    */
-   brw_exec_init( ctx );
-   brw_save_init( ctx );
-
    {
       const char *filename = getenv("INTEL_REPLAY");
       if (filename) {
diff --git a/src/mesa/drivers/dri/i965/brw_context.h b/src/mesa/drivers/dri/i965/brw_context.h
index cac4d86d3ab..08fdc545205 100644
--- a/src/mesa/drivers/dri/i965/brw_context.h
+++ b/src/mesa/drivers/dri/i965/brw_context.h
@@ -36,7 +36,6 @@
 #include "intel_context.h"
 #include "brw_structs.h"
 #include "imports.h"
-#include "brw_attrib.h"
 
 
 /* Glossary:
@@ -215,7 +214,7 @@ struct brw_vs_prog_data {
    GLuint total_grf;
    GLuint outputs_written;
 
-   GLuint64EXT inputs_read;
+   GLuint inputs_read;
 
    /* Used for calculating urb partitions:
     */
@@ -382,10 +381,10 @@ struct brw_cached_batch_item {
    
 
 
-/* Protect against a future where BRW_ATTRIB_MAX > 32.  Wouldn't life
+/* Protect against a future where VERT_ATTRIB_MAX > 32.  Wouldn't life
  * be easier if C allowed arrays of packed elements?
  */
-#define ATTRIB_BIT_DWORDS  ((BRW_ATTRIB_MAX+31)/32)
+#define ATTRIB_BIT_DWORDS  ((VERT_ATTRIB_MAX+31)/32)
 
 struct brw_vertex_element {
    const struct gl_client_array *glarray;
@@ -401,8 +400,8 @@ struct brw_vertex_element {
 
 
 struct brw_vertex_info {
-   GLuint64EXT varying;  /* varying:1[BRW_ATTRIB_MAX] */
-   GLuint sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[BRW_ATTRIB_MAX] */
+   GLuint varying;  /* varying:1[VERT_ATTRIB_MAX] */
+   GLuint sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[VERT_ATTRIB_MAX] */
 };
 
 
@@ -449,14 +448,13 @@ struct brw_context
    struct brw_cached_batch_item *cached_batch_items;
 
    struct {
-      /* Fallback values for inputs not supplied: 
-       */
-      struct gl_client_array current_values[BRW_ATTRIB_MAX];
 
       /* Arrays with buffer objects to copy non-bufferobj arrays into
        * for upload:
        */
-      struct gl_client_array vbo_array[BRW_ATTRIB_MAX];
+      struct gl_client_array vbo_array[VERT_ATTRIB_MAX];
+
+      struct brw_vertex_element inputs[VERT_ATTRIB_MAX];
 
 #define BRW_NR_UPLOAD_BUFS 17
 #define BRW_UPLOAD_INIT_SIZE (128*1024)
@@ -469,11 +467,6 @@ struct brw_context
 	 GLuint wrap;
       } upload;
 
-      /* Currenly bound arrays, including fallbacks to current_values
-       * above:
-       */
-      struct brw_vertex_element inputs[BRW_ATTRIB_MAX];
-
       /* Summary of size and varying of active arrays, so we can check
        * for changes to this state:
        */
@@ -505,8 +498,8 @@ struct brw_context
 
    /* Active vertex program: 
     */
-   struct gl_vertex_program *vertex_program;
-   struct gl_fragment_program *fragment_program;
+   const struct gl_vertex_program *vertex_program;
+   const struct gl_fragment_program *fragment_program;
 
 
    /* For populating the gtt:
diff --git a/src/mesa/drivers/dri/i965/brw_draw.c b/src/mesa/drivers/dri/i965/brw_draw.c
index 471fda9f7ef..092c6bafc2d 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.c
+++ b/src/mesa/drivers/dri/i965/brw_draw.c
@@ -35,7 +35,6 @@
 
 #include "brw_draw.h"
 #include "brw_defines.h"
-#include "brw_attrib.h"
 #include "brw_context.h"
 #include "brw_aub.h"
 #include "brw_state.h"
@@ -45,7 +44,8 @@
 #include "intel_batchbuffer.h"
 #include "intel_buffer_objects.h"
 
-
+#include "tnl/tnl.h"
+#include "vbo/vbo_context.h"
 
 
 
@@ -143,7 +143,7 @@ static void brw_emit_cliprect( struct brw_context *brw,
 
 
 static void brw_emit_prim( struct brw_context *brw, 
-			   const struct brw_draw_prim *prim )
+			   const struct _mesa_prim *prim )
 
 {
    struct brw_3d_primitive prim_packet;
@@ -170,34 +170,9 @@ static void brw_emit_prim( struct brw_context *brw,
    }
 }
 
-
-
-static void update_current_size( struct gl_client_array *array)
-{
-   const GLfloat *ptr = (const GLfloat *)array->Ptr;
-
-   assert(array->StrideB == 0);
-   assert(array->Type == GL_FLOAT || array->Type == GL_UNSIGNED_BYTE);
-
-   if (ptr[3] != 1.0) 
-      array->Size = 4;
-   else if (ptr[2] != 0.0) 
-      array->Size = 3;
-   else if (ptr[1] != 0.0) 
-      array->Size = 2;
-   else
-      array->Size = 1;
-}
-
-
-
-/* Fill in any gaps in passed arrays with pointers to current
- * attributes:
- */
 static void brw_merge_inputs( struct brw_context *brw,
 		       const struct gl_client_array *arrays[])
 {
-   struct gl_client_array *current_values = brw->vb.current_values;
    struct brw_vertex_element *inputs = brw->vb.inputs;
    struct brw_vertex_info old = brw->vb.info;
    GLuint i;
@@ -205,19 +180,16 @@ static void brw_merge_inputs( struct brw_context *brw,
    memset(inputs, 0, sizeof(*inputs));
    memset(&brw->vb.info, 0, sizeof(brw->vb.info));
 
-   for (i = 0; i < BRW_ATTRIB_MAX; i++) {
-      if (arrays[i] && arrays[i]->Enabled)
-      {
-	 brw->vb.inputs[i].glarray = arrays[i];
-	 brw->vb.info.varying |= (GLuint64EXT) 1 << i;
-      }
-      else 
-      {
-	 brw->vb.inputs[i].glarray = &current_values[i];
-	 update_current_size(&current_values[i]);
-      }
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      brw->vb.inputs[i].glarray = arrays[i];
+
+      /* XXX: metaops passes null arrays */
+      if (arrays[i]) {
+	 if (arrays[i]->StrideB != 0)
+	    brw->vb.info.varying |= 1 << i;
 
-      brw->vb.info.sizes[i/16] |= (inputs[i].glarray->Size - 1) << ((i%16) * 2);
+	 brw->vb.info.sizes[i/16] |= (inputs[i].glarray->Size - 1) << ((i%16) * 2);
+      }
    }
 
    /* Raise statechanges if input sizes and varying have changed: 
@@ -229,8 +201,11 @@ static void brw_merge_inputs( struct brw_context *brw,
       brw->state.dirty.brw |= BRW_NEW_INPUT_VARYING;
 }
 
+/* XXX: could split the primitive list to fallback only on the
+ * non-conformant primitives.
+ */
 static GLboolean check_fallbacks( struct brw_context *brw,
-				  const struct brw_draw_prim *prim,
+				  const struct _mesa_prim *prim,
 				  GLuint nr_prims )
 {
    GLuint i;
@@ -281,15 +256,16 @@ static GLboolean check_fallbacks( struct brw_context *brw,
    return GL_FALSE;
 }
 
-
+/* May fail if out of video memory for texture or vbo upload, or on
+ * fallback conditions.
+ */
 static GLboolean brw_try_draw_prims( GLcontext *ctx,
 				     const struct gl_client_array *arrays[],
-				     const struct brw_draw_prim *prim,
+				     const struct _mesa_prim *prim,
 				     GLuint nr_prims,
-				     const struct brw_draw_index_buffer *ib,
+				     const struct _mesa_index_buffer *ib,
 				     GLuint min_index,
-				     GLuint max_index,
-				     GLuint flags )
+				     GLuint max_index )
 {
    struct intel_context *intel = intel_context(ctx);
    struct brw_context *brw = brw_context(ctx);
@@ -298,11 +274,11 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
 
    if (ctx->NewState)
       _mesa_update_state( ctx );
-      
+
    /* Bind all inputs, derive varying and size information:
     */
    brw_merge_inputs( brw, arrays );
-
+      
    /* Have to validate state quite late.  Will rebuild tnl_program,
     * which depends on varying information.  
     * 
@@ -319,10 +295,6 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
    }
 
    {
-      assert(intel->locked);
-      
-
-
       /* Set the first primitive early, ahead of validate_state:
        */
       brw_set_prim(brw, prim[0].mode);
@@ -411,44 +383,88 @@ static GLboolean brw_try_draw_prims( GLcontext *ctx,
    return retval;
 }
 
+static GLboolean brw_need_rebase( GLcontext *ctx,
+				  const struct gl_client_array *arrays[],
+				  const struct _mesa_index_buffer *ib,
+				  GLuint min_index )
+{
+   if (min_index == 0) 
+      return GL_FALSE;
 
-GLboolean brw_draw_prims( GLcontext *ctx,
-			  const struct gl_client_array *arrays[],
-			  const struct brw_draw_prim *prim,
-			  GLuint nr_prims,
-			  const struct brw_draw_index_buffer *ib,
-			  GLuint min_index,
-			  GLuint max_index,
-			  GLuint flags )
+   if (ib) {
+      if (!vbo_all_varyings_in_vbos(arrays))
+	 return GL_TRUE;
+      else
+	 return GL_FALSE;
+   }
+   else {
+      /* Hmm.  This isn't quite what I wanted.  BRW can actually
+       * handle the mixed case well enough that we shouldn't need to
+       * rebase.  However, it's probably not very common, nor hugely
+       * expensive to do it this way:
+       */
+      if (!vbo_all_varyings_in_vbos(arrays))
+	 return GL_TRUE;
+      else
+	 return GL_FALSE;
+   }
+}
+				  
+
+void brw_draw_prims( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     GLuint min_index,
+		     GLuint max_index )
 {
    struct intel_context *intel = intel_context(ctx);
    GLboolean retval;
 
-   retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index, flags);
+   /* Decide if we want to rebase.  If so we end up recursing once
+    * only into this function.
+    */
+   if (brw_need_rebase( ctx, arrays, ib, min_index )) {
+      vbo_rebase_prims( ctx, arrays, 
+			prim, nr_prims, 
+			ib, min_index, max_index, 
+			brw_draw_prims );
+      
+      return;
+   }
 
-   
-   if (!retval && bmError(intel)) {
 
-      DBG("retrying\n");
-      /* This looks like out-of-memory but potentially we have
-       * situation where there is enough memory but it has become
-       * fragmented.  Clear out all heaps and start from scratch by
-       * faking a contended lock event:  (done elsewhere)
-       */
+   /* Make a first attempt at drawing:
+    */
+   retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
 
+   
+   /* This looks like out-of-memory but potentially we have
+    * situation where there is enough memory but it has become
+    * fragmented.  Clear out all heaps and start from scratch by
+    * faking a contended lock event:  (done elsewhere)
+    */
+   if (!retval && !intel->Fallback && bmError(intel)) {
+      DBG("retrying\n");
       /* Then try a second time only to upload textures and draw the
        * primitives:
        */
-      retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index, flags);
+      retval = brw_try_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+   }
+
+   /* Otherwise, we really are out of memory.  Pass the drawing
+    * command to the software tnl module and which will in turn call
+    * swrast to do the drawing.
+    */
+   if (!retval) {
+      _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
    }
 
    if (intel->aub_file && (INTEL_DEBUG & DEBUG_SYNC)) {
       intelFinish( &intel->ctx );
       intel->aub_wrap = 1;
    }
-
-
-   return retval;
 }
 
 
@@ -461,21 +477,25 @@ static void brw_invalidate_vbo_cb( struct intel_context *intel, void *ptr )
 void brw_draw_init( struct brw_context *brw )
 {
    GLcontext *ctx = &brw->intel.ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLuint i;
    
+   /* Register our drawing function: 
+    */
+   vbo->draw_prims = brw_draw_prims;
+
    brw->vb.upload.size = BRW_UPLOAD_INIT_SIZE;
 
    for (i = 0; i < BRW_NR_UPLOAD_BUFS; i++) {
       brw->vb.upload.vbo[i] = ctx->Driver.NewBufferObject(ctx, 1, GL_ARRAY_BUFFER_ARB);
       
-      /* XXX: Set these to no-backing-store
+      /* NOTE:  These are set to no-backing-store.
        */
       bmBufferSetInvalidateCB(&brw->intel,
 			      intel_bufferobj_buffer(intel_buffer_object(brw->vb.upload.vbo[i])),
 			      brw_invalidate_vbo_cb,
 			      &brw->intel,
 			      GL_TRUE);
-
    }
 
    ctx->Driver.BufferData( ctx, 
@@ -484,9 +504,6 @@ void brw_draw_init( struct brw_context *brw )
 			   NULL,
 			   GL_DYNAMIC_DRAW_ARB,
 			   brw->vb.upload.vbo[0] );
-						     
-   
-   brw_init_current_values(ctx, brw->vb.current_values);
 }
 
 void brw_draw_destroy( struct brw_context *brw )
diff --git a/src/mesa/drivers/dri/i965/brw_draw.h b/src/mesa/drivers/dri/i965/brw_draw.h
index 92640bf725b..0f7b7383102 100644
--- a/src/mesa/drivers/dri/i965/brw_draw.h
+++ b/src/mesa/drivers/dri/i965/brw_draw.h
@@ -29,44 +29,18 @@
 #define BRW_DRAW_H
 
 #include "mtypes.h"		/* for GLcontext... */
-#include "brw_attrib.h"
+#include "vbo/vbo.h"
 
 struct brw_context;
 
-struct brw_draw_prim {
-   GLuint mode:8;
-   GLuint indexed:1;
-   GLuint begin:1;
-   GLuint end:1;
-   GLuint weak:1;
-   GLuint pad:20;
 
-   GLuint start;
-   GLuint count;
-};
-
-struct brw_draw_index_buffer {
-   GLuint count;
-   GLenum type;
-   struct gl_buffer_object *obj;
-   const void *ptr;
-   GLuint rebase;
-};
-
-
-#define BRW_DRAW_SORTED           0x1
-#define BRW_DRAW_ALL_INTERLEAVED  0x2
-#define BRW_DRAW_NON_INTERLEAVED  0x4
-#define BRW_DRAW_LOCKED           0x8
-
-GLboolean brw_draw_prims( GLcontext *ctx,
-			  const struct gl_client_array *arrays[],
-			  const struct brw_draw_prim *prims,
-			  GLuint nr_prims,
-			  const struct brw_draw_index_buffer *ib,
-			  GLuint min_index,
-			  GLuint max_index,
-			  GLuint flags );
+void brw_draw_prims( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prims,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     GLuint min_index,
+		     GLuint max_index );
 
 void brw_draw_init( struct brw_context *brw );
 void brw_draw_destroy( struct brw_context *brw );
@@ -80,25 +54,12 @@ void brw_init_current_values(GLcontext *ctx,
 /* brw_draw_upload.c
  */
 void brw_upload_indices( struct brw_context *brw,
-			 const struct brw_draw_index_buffer *index_buffer);
+			 const struct _mesa_index_buffer *index_buffer);
 
 GLboolean brw_upload_vertices( struct brw_context *brw,
 			       GLuint min_index,
 			       GLuint max_index );
 
 
-/* Helpers for save, exec.  Should probably have their own file:
- */
-struct brw_exec_context;
-struct brw_save_context;
-
-struct brw_exec_save {
-   struct brw_exec_context *exec;
-   struct brw_save_context *save;
-};
-
-/* Doesn't really belong here:
- */
-#define IMM_CONTEXT(ctx) ((struct brw_exec_save *)((ctx)->swtnl_im))
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_draw_current.c b/src/mesa/drivers/dri/i965/brw_draw_current.c
deleted file mode 100644
index 98d930738e5..00000000000
--- a/src/mesa/drivers/dri/i965/brw_draw_current.c
+++ /dev/null
@@ -1,103 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * 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"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, 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 NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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.
- * 
- **************************************************************************/
-
-#include <stdlib.h>
-
-#include "glheader.h"
-#include "context.h"
-#include "state.h"
-#include "api_validate.h"
-#include "enums.h"
-
-#include "brw_context.h"
-#include "brw_draw.h"
-
-#include "bufmgr.h"
-#include "intel_buffer_objects.h"
-
-
-void brw_init_current_values(GLcontext *ctx,
-			     struct gl_client_array *arrays)
-{
-   GLuint i;
-
-   memset(arrays, 0, sizeof(*arrays) * BRW_ATTRIB_MAX);
-
-   /* Set up a constant (StrideB == 0) array for each current
-    * attribute:
-    */
-   for (i = 0; i < BRW_ATTRIB_MAX; i++) {
-      struct gl_client_array *cl = &arrays[i];
-
-      switch (i) {
-      case BRW_ATTRIB_MAT_FRONT_SHININESS:
-      case BRW_ATTRIB_MAT_BACK_SHININESS:
-      case BRW_ATTRIB_INDEX:
-      case BRW_ATTRIB_EDGEFLAG:
-	 cl->Size = 1;
-	 break;
-      case BRW_ATTRIB_MAT_FRONT_INDEXES:
-      case BRW_ATTRIB_MAT_BACK_INDEXES:
-	 cl->Size = 3;
-	 break;
-      default:
-	 /* This is fixed for the material attributes, for others will
-	  * be determined at runtime:
-	  */
-	 if (i >= BRW_ATTRIB_MAT_FRONT_AMBIENT)
-	    cl->Size = 4;
-	 else
-	    cl->Size = 1;
-	 break;
-      }
-
-      switch (i) {
-      case BRW_ATTRIB_EDGEFLAG:
-	 cl->Type = GL_UNSIGNED_BYTE;
-	 cl->Ptr = (const void *)&ctx->Current.EdgeFlag;
-	 break;
-      case BRW_ATTRIB_INDEX:
-	 cl->Type = GL_FLOAT;
-	 cl->Ptr = (const void *)&ctx->Current.Index;
-	 break;
-      default:
-	 cl->Type = GL_FLOAT;
-	 if (i < BRW_ATTRIB_FIRST_MATERIAL)
-	    cl->Ptr = (const void *)ctx->Current.Attrib[i];
-	 else 
-	    cl->Ptr = (const void *)ctx->Light.Material.Attrib[i - BRW_ATTRIB_FIRST_MATERIAL];
-	 break;
-      }
-
-      cl->Stride = 0;
-      cl->StrideB = 0;
-      cl->Enabled = 1;
-      cl->Flags = 0;
-      cl->BufferObj = ctx->Array.NullBufferObj;
-   }
-}
-
diff --git a/src/mesa/drivers/dri/i965/brw_draw_upload.c b/src/mesa/drivers/dri/i965/brw_draw_upload.c
index 57ee294f0cf..6150cac4aa3 100644
--- a/src/mesa/drivers/dri/i965/brw_draw_upload.c
+++ b/src/mesa/drivers/dri/i965/brw_draw_upload.c
@@ -35,7 +35,6 @@
 
 #include "brw_draw.h"
 #include "brw_defines.h"
-#include "brw_attrib.h"
 #include "brw_context.h"
 #include "brw_aub.h"
 #include "brw_state.h"
@@ -310,7 +309,6 @@ copy_array_to_vbo_array( struct brw_context *brw,
 			 GLuint i,
 			 const struct gl_client_array *array,
 			 GLuint element_size,
-			 GLuint min_index,
 			 GLuint count)
 {
    GLcontext *ctx = &brw->intel.ctx;
@@ -337,7 +335,6 @@ copy_array_to_vbo_array( struct brw_context *brw,
    vbo_array->Enabled = 1;
    vbo_array->Normalized = array->Normalized;
    vbo_array->_MaxElement = array->_MaxElement;	/* ? */
-   vbo_array->Flags = array->Flags; /* ? */
    vbo_array->BufferObj = vbo;
 
    {
@@ -349,7 +346,7 @@ copy_array_to_vbo_array( struct brw_context *brw,
       map += offset;
 
       copy_strided_array( map, 
-			  array->Ptr + min_index * array->StrideB,
+			  array->Ptr,
 			  element_size,
 			  array->StrideB,
 			  count);
@@ -380,7 +377,6 @@ interleaved_vbo_array( struct brw_context *brw,
    vbo_array->Enabled = 1;
    vbo_array->Normalized = array->Normalized;
    vbo_array->_MaxElement = array->_MaxElement;	
-   vbo_array->Flags = array->Flags; /* ? */
    vbo_array->BufferObj = uploaded_array->BufferObj;
 
    return vbo_array;
@@ -393,17 +389,17 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
 {
    GLcontext *ctx = &brw->intel.ctx;
    struct intel_context *intel = intel_context(ctx);
-   GLuint64EXT tmp = brw->vs.prog_data->inputs_read; 
+   GLuint tmp = brw->vs.prog_data->inputs_read; 
    struct brw_vertex_element_packet vep;
    struct brw_array_state vbp;
    GLuint i;
    const void *ptr = NULL;
    GLuint interleave = 0;
 
-   struct brw_vertex_element *enabled[BRW_ATTRIB_MAX];
+   struct brw_vertex_element *enabled[VERT_ATTRIB_MAX];
    GLuint nr_enabled = 0;
 
-   struct brw_vertex_element *upload[BRW_ATTRIB_MAX];
+   struct brw_vertex_element *upload[VERT_ATTRIB_MAX];
    GLuint nr_uploads = 0;
    
 
@@ -412,17 +408,19 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
 
    /* First build an array of pointers to ve's in vb.inputs_read
     */
+   if (0)
+      _mesa_printf("%s %d..%d\n", __FUNCTION__, min_index, max_index);
    
    while (tmp) {
-      GLuint i = ffsll(tmp)-1;
+      GLuint i = _mesa_ffsll(tmp)-1;
       struct brw_vertex_element *input = &brw->vb.inputs[i];
 
-      tmp &= ~((GLuint64EXT)1<<i);
+      tmp &= ~(1<<i);
       enabled[nr_enabled++] = input;
 
       input->index = i;
       input->element_size = get_size(input->glarray->Type) * input->glarray->Size;
-      input->count = input->glarray->StrideB ? max_index - min_index : 1;
+      input->count = input->glarray->StrideB ? max_index + 1 - min_index : 1;
 
       if (!input->glarray->BufferObj->Name) {
 	 if (i == 0) {
@@ -441,10 +439,16 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
 	 }
 
 	 upload[nr_uploads++] = input;
-	 input->vbo_rebase_offset = 0;
+	 
+	 /* We rebase drawing to start at element zero only when
+	  * varyings are not in vbos, which means we can end up
+	  * uploading non-varying arrays (stride != 0) when min_index
+	  * is zero.  This doesn't matter as the amount to upload is
+	  * the same for these arrays whether the draw call is rebased
+	  * or not - we just have to upload the one element.
+	  */
+	 assert(min_index == 0 || input->glarray->StrideB == 0);
       }
-      else 
-	 input->vbo_rebase_offset = min_index * input->glarray->StrideB;
    }
 
    /* Upload interleaved arrays if all uploads are interleaved
@@ -457,7 +461,6 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
       input0->glarray = copy_array_to_vbo_array(brw, 0,
 						input0->glarray, 
 						interleave,
-						min_index,
 						input0->count);
 
       for (i = 1; i < nr_uploads; i++) {
@@ -475,7 +478,6 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
 	 input->glarray = copy_array_to_vbo_array(brw, i, 
 						  input->glarray,
 						  input->element_size,
-						  min_index,
 						  input->count);
 
       }
@@ -523,9 +525,9 @@ GLboolean brw_upload_vertices( struct brw_context *brw,
       vbp.vb[i].vb0.bits.pad = 0;
       vbp.vb[i].vb0.bits.access_type = BRW_VERTEXBUFFER_ACCESS_VERTEXDATA;
       vbp.vb[i].vb0.bits.vb_index = i;
-      vbp.vb[i].offset = (GLuint)input->glarray->Ptr + input->vbo_rebase_offset;
+      vbp.vb[i].offset = (GLuint)input->glarray->Ptr;
       vbp.vb[i].buffer = array_buffer(input->glarray);
-      vbp.vb[i].max_index = max_index - min_index;
+      vbp.vb[i].max_index = max_index;
    }
 
 
@@ -566,94 +568,32 @@ static GLuint element_size( GLenum type )
 
 
 
-
-static void rebase_indices_to_vbo_indices( struct brw_context *brw, 
-					   const struct brw_draw_index_buffer *index_buffer,
-					   struct gl_buffer_object **vbo_return,
-					   GLuint *offset_return )
+void brw_upload_indices( struct brw_context *brw,
+			 const struct _mesa_index_buffer *index_buffer )
 {
    GLcontext *ctx = &brw->intel.ctx;
-   GLuint min_index = index_buffer->rebase;
-   const void *indices = index_buffer->ptr;
-   GLsizei count = index_buffer->count;
-   GLenum type = index_buffer->type;
-   GLuint size = element_size(type) * count;
-   struct gl_buffer_object *bufferobj;
-   GLuint offset;
-   GLuint i;
-
-   get_space(brw, size, &bufferobj, &offset);
+   struct intel_context *intel = &brw->intel;
+   GLuint ib_size = get_size(index_buffer->type) * index_buffer->count;
+   struct gl_buffer_object *bufferobj = index_buffer->obj;
+   GLuint offset = (GLuint)index_buffer->ptr;
 
-   *vbo_return = bufferobj;
-   *offset_return = offset;
+   /* Turn into a proper VBO:
+    */
+   if (!bufferobj->Name) {
+     
+      /* Get new bufferobj, offset:
+       */
+      get_space(brw, ib_size, &bufferobj, &offset);
 
-   if (min_index == 0) {
       /* Straight upload
        */
       ctx->Driver.BufferSubData( ctx,
 				 GL_ELEMENT_ARRAY_BUFFER_ARB,
 				 offset, 
-				 size,
-				 indices,
+				 ib_size,
+				 index_buffer->ptr,
 				 bufferobj);
    }
-   else {
-      void *map = ctx->Driver.MapBuffer(ctx,
-					GL_ELEMENT_ARRAY_BUFFER_ARB,
-					GL_DYNAMIC_DRAW_ARB,
-					bufferobj);
-
-      map += offset;
-
-      switch (type) {
-      case GL_UNSIGNED_INT: {
-	 GLuint *ui_map = (GLuint *)map;
-	 const GLuint *ui_indices = (const GLuint *)indices;
-
-	 for (i = 0; i < count; i++)
-	    ui_map[i] = ui_indices[i] - min_index;
-	 break;
-      }
-      case GL_UNSIGNED_SHORT:  {
-	 GLushort *us_map = (GLushort *)map;
-	 const GLushort *us_indices = (const GLushort *)indices;
-
-	 for (i = 0; i < count; i++)
-	    us_map[i] = us_indices[i] - min_index;
-	 break;
-      }
-      case GL_UNSIGNED_BYTE:  {
-	 GLubyte *ub_map = (GLubyte *)map;
-	 const GLubyte *ub_indices = (const GLubyte *)indices;
-
-	 for (i = 0; i < count; i++)
-	    ub_map[i] = ub_indices[i] - min_index;
-	 break;
-      }
-      }
-
-      ctx->Driver.UnmapBuffer(ctx, 
-			      GL_ELEMENT_ARRAY_BUFFER_ARB, 
-			      bufferobj);
-
-   }
-}
-
-
-
-void brw_upload_indices( struct brw_context *brw,
-			 const struct brw_draw_index_buffer *index_buffer)
-{
-   struct intel_context *intel = &brw->intel;
-   GLuint ib_size = get_size(index_buffer->type) * index_buffer->count;
-   struct gl_buffer_object *bufferobj = index_buffer->obj;
-   GLuint offset = (GLuint)index_buffer->ptr;
-
-   /* Already turned into a proper VBO:
-    */
-   if (!index_buffer->obj->Name) {
-      rebase_indices_to_vbo_indices(brw, index_buffer, &bufferobj, &offset );
-   }
 
    /* Emit the indexbuffer packet:
     */
diff --git a/src/mesa/drivers/dri/i965/brw_eu.h b/src/mesa/drivers/dri/i965/brw_eu.h
index 1afa0f816b8..d4dbcf38a7a 100644
--- a/src/mesa/drivers/dri/i965/brw_eu.h
+++ b/src/mesa/drivers/dri/i965/brw_eu.h
@@ -104,6 +104,7 @@ struct brw_compile {
    struct brw_instruction *current;
 
    GLuint flag_value;
+   GLboolean single_program_flow;
 };
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_eu_emit.c b/src/mesa/drivers/dri/i965/brw_eu_emit.c
index 6425c91450c..9992b47d8ae 100644
--- a/src/mesa/drivers/dri/i965/brw_eu_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_eu_emit.c
@@ -464,7 +464,6 @@ struct brw_instruction *brw_JMPI(struct brw_compile *p,
    return insn;
 }
 
-
 /* EU takes the value from the flag register and pushes it onto some
  * sort of a stack (presumably merging with any flag value already on
  * the stack).  Within an if block, the flags at the top of the stack
@@ -482,7 +481,16 @@ struct brw_instruction *brw_JMPI(struct brw_compile *p,
  */
 struct brw_instruction *brw_IF(struct brw_compile *p, GLuint execute_size)
 {
-   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_IF);   
+   struct brw_instruction *insn;
+
+   if (p->single_program_flow) {
+      assert(execute_size == BRW_EXECUTE_1);
+
+      insn = next_insn(p, BRW_OPCODE_ADD);
+      insn->header.predicate_inverse = 1;
+   } else {
+      insn = next_insn(p, BRW_OPCODE_IF);
+   }
 
    /* Override the defaults for this instruction:
     */
@@ -504,7 +512,13 @@ struct brw_instruction *brw_IF(struct brw_compile *p, GLuint execute_size)
 struct brw_instruction *brw_ELSE(struct brw_compile *p, 
 				 struct brw_instruction *if_insn)
 {
-   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_ELSE);   
+   struct brw_instruction *insn;
+
+   if (p->single_program_flow) {
+      insn = next_insn(p, BRW_OPCODE_ADD);
+   } else {
+      insn = next_insn(p, BRW_OPCODE_ELSE);
+   }
 
    brw_set_dest(insn, brw_ip_reg());
    brw_set_src0(insn, brw_ip_reg());
@@ -516,11 +530,17 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p,
 
    /* Patch the if instruction to point at this instruction.
     */
-   assert(if_insn->header.opcode == BRW_OPCODE_IF);
+   if (p->single_program_flow) {
+      assert(if_insn->header.opcode == BRW_OPCODE_ADD);
 
-   if_insn->bits3.if_else.jump_count = insn - if_insn; 
-   if_insn->bits3.if_else.pop_count = 1;
-   if_insn->bits3.if_else.pad0 = 0;
+      if_insn->bits3.ud = (insn - if_insn + 1) * 16;
+   } else {
+      assert(if_insn->header.opcode == BRW_OPCODE_IF);
+
+      if_insn->bits3.if_else.jump_count = insn - if_insn;
+      if_insn->bits3.if_else.pop_count = 1;
+      if_insn->bits3.if_else.pad0 = 0;
+   }
 
    return insn;
 }
@@ -528,63 +548,76 @@ struct brw_instruction *brw_ELSE(struct brw_compile *p,
 void brw_ENDIF(struct brw_compile *p, 
 	       struct brw_instruction *patch_insn)
 {
-   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_ENDIF);   
+   if (p->single_program_flow) {
+      /* In single program flow mode, there's no need to execute an ENDIF,
+       * since we don't need to do any stack operations, and if we're executing
+       * currently, we want to just continue executing.
+       */
+      struct brw_instruction *next = &p->store[p->nr_insn];
 
-   brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
-   brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
-   brw_set_src1(insn, brw_imm_d(0x0));
+      assert(patch_insn->header.opcode == BRW_OPCODE_ADD);
 
-   insn->header.compression_control = BRW_COMPRESSION_NONE;
-   insn->header.execution_size = patch_insn->header.execution_size;
-   insn->header.mask_control = BRW_MASK_ENABLE;
+      patch_insn->bits3.ud = (next - patch_insn) * 16;
+   } else {
+      struct brw_instruction *insn = next_insn(p, BRW_OPCODE_ENDIF);
 
-   assert(patch_insn->bits3.if_else.jump_count == 0);
-      
-   /* Patch the if or else instructions to point at this or the next
-    * instruction respectively.
-    */
-   if (patch_insn->header.opcode == BRW_OPCODE_IF) {
-      /* Automagically turn it into an IFF:
+      brw_set_dest(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
+      brw_set_src0(insn, retype(brw_vec4_grf(0,0), BRW_REGISTER_TYPE_UD));
+      brw_set_src1(insn, brw_imm_d(0x0));
+
+      insn->header.compression_control = BRW_COMPRESSION_NONE;
+      insn->header.execution_size = patch_insn->header.execution_size;
+      insn->header.mask_control = BRW_MASK_ENABLE;
+
+      assert(patch_insn->bits3.if_else.jump_count == 0);
+
+      /* Patch the if or else instructions to point at this or the next
+       * instruction respectively.
        */
-      patch_insn->header.opcode = BRW_OPCODE_IFF;
-      patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
-      patch_insn->bits3.if_else.pop_count = 0;
-      patch_insn->bits3.if_else.pad0 = 0;
+      if (patch_insn->header.opcode == BRW_OPCODE_IF) {
+	 /* Automagically turn it into an IFF:
+	  */
+	 patch_insn->header.opcode = BRW_OPCODE_IFF;
+	 patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
+	 patch_insn->bits3.if_else.pop_count = 0;
+	 patch_insn->bits3.if_else.pad0 = 0;
+      } else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) {
+	 patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
+	 patch_insn->bits3.if_else.pop_count = 1;
+	 patch_insn->bits3.if_else.pad0 = 0;
+      } else {
+	 assert(0);
+      }
 
+      /* Also pop item off the stack in the endif instruction:
+       */
+      insn->bits3.if_else.jump_count = 0;
+      insn->bits3.if_else.pop_count = 1;
+      insn->bits3.if_else.pad0 = 0;
    }
-   else if (patch_insn->header.opcode == BRW_OPCODE_ELSE) {
-      patch_insn->bits3.if_else.jump_count = insn - patch_insn + 1;
-      patch_insn->bits3.if_else.pop_count = 1;
-      patch_insn->bits3.if_else.pad0 = 0;
-   }
-   else {
-      assert(0);
-   }
-
-   /* Also pop item off the stack in the endif instruction:
-    */
-   insn->bits3.if_else.jump_count = 0;
-   insn->bits3.if_else.pop_count = 1; 
-   insn->bits3.if_else.pad0 = 0;
 }
 
 /* DO/WHILE loop:
  */
 struct brw_instruction *brw_DO(struct brw_compile *p, GLuint execute_size)
 {
-   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_DO);   
+   if (p->single_program_flow) {
+      return &p->store[p->nr_insn];
+   } else {
+      struct brw_instruction *insn = next_insn(p, BRW_OPCODE_DO);
 
-   /* Override the defaults for this instruction:
-    */
-   brw_set_dest(insn, retype(brw_vec1_grf(0,0), BRW_REGISTER_TYPE_UD));
-   brw_set_src0(insn, retype(brw_vec1_grf(0,0), BRW_REGISTER_TYPE_UD));
-   brw_set_src1(insn, retype(brw_vec1_grf(0,0), BRW_REGISTER_TYPE_UD));
+      /* Override the defaults for this instruction:
+       */
+      brw_set_dest(insn, retype(brw_vec1_grf(0,0), BRW_REGISTER_TYPE_UD));
+      brw_set_src0(insn, retype(brw_vec1_grf(0,0), BRW_REGISTER_TYPE_UD));
+      brw_set_src1(insn, retype(brw_vec1_grf(0,0), BRW_REGISTER_TYPE_UD));
 
-   insn->header.compression_control = BRW_COMPRESSION_NONE;
-   insn->header.execution_size = execute_size;
-/*    insn->header.mask_control = BRW_MASK_ENABLE; */
+      insn->header.compression_control = BRW_COMPRESSION_NONE;
+      insn->header.execution_size = execute_size;
+      /* insn->header.mask_control = BRW_MASK_ENABLE; */
 
-   return insn;
+      return insn;
+   }
 }
 
 
@@ -592,19 +625,31 @@ struct brw_instruction *brw_DO(struct brw_compile *p, GLuint execute_size)
 void brw_WHILE(struct brw_compile *p, 
 	       struct brw_instruction *do_insn)
 {
-   struct brw_instruction *insn = next_insn(p, BRW_OPCODE_WHILE);
+   struct brw_instruction *insn;
+
+   if (p->single_program_flow)
+      insn = next_insn(p, BRW_OPCODE_ADD);
+   else
+      insn = next_insn(p, BRW_OPCODE_WHILE);
 
    brw_set_dest(insn, brw_ip_reg());
    brw_set_src0(insn, brw_ip_reg());
    brw_set_src1(insn, brw_imm_d(0x0));
 
    insn->header.compression_control = BRW_COMPRESSION_NONE;
-   insn->header.execution_size = do_insn->header.execution_size;
 
-   assert(do_insn->header.opcode == BRW_OPCODE_DO);
-   insn->bits3.if_else.jump_count = do_insn - insn;
-   insn->bits3.if_else.pop_count = 0;
-   insn->bits3.if_else.pad0 = 0;
+   if (p->single_program_flow) {
+      insn->header.execution_size = BRW_EXECUTE_1;
+
+      insn->bits3.d = (do_insn - insn) * 16;
+   } else {
+      insn->header.execution_size = do_insn->header.execution_size;
+
+      assert(do_insn->header.opcode == BRW_OPCODE_DO);
+      insn->bits3.if_else.jump_count = do_insn - insn;
+      insn->bits3.if_else.pop_count = 0;
+      insn->bits3.if_else.pad0 = 0;
+   }
 
 /*    insn->header.mask_control = BRW_MASK_ENABLE; */
 
diff --git a/src/mesa/drivers/dri/i965/brw_exec.h b/src/mesa/drivers/dri/i965/brw_exec.h
deleted file mode 100644
index f07b4485874..00000000000
--- a/src/mesa/drivers/dri/i965/brw_exec.h
+++ /dev/null
@@ -1,150 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
-TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#ifndef __BRW_EXEC_H__
-#define __BRW_EXEC_H__
-
-#include "mtypes.h"
-#include "brw_attrib.h"
-#include "brw_draw.h"
-
-
-#define BRW_MAX_PRIM 64
-
-/* Wierd implementation stuff:
- */
-#define BRW_VERT_BUFFER_SIZE (1024*16)	/* dwords == 64k */
-#define BRW_MAX_ATTR_CODEGEN 16 
-#define ERROR_ATTRIB 16
-
-
-
-
-struct brw_exec_eval1_map {
-   struct gl_1d_map *map;
-   GLuint sz;
-};
-
-struct brw_exec_eval2_map {
-   struct gl_2d_map *map;
-   GLuint sz;
-};
-
-
-
-struct brw_exec_copied_vtx {
-   GLfloat buffer[BRW_ATTRIB_MAX * 4 * BRW_MAX_COPIED_VERTS];
-   GLuint nr;
-};
-
-
-typedef void (*brw_attrfv_func)( const GLfloat * );
-
-
-struct brw_exec_context
-{
-   GLcontext *ctx;   
-   GLvertexformat vtxfmt;
-
-   struct {
-      struct gl_buffer_object *bufferobj;
-      GLubyte *buffer_map;
-
-      GLuint vertex_size;
-
-      struct brw_draw_prim prim[BRW_MAX_PRIM];
-      GLuint prim_count;
-
-      GLfloat *vbptr;		     /* cursor, points into buffer */
-      GLfloat vertex[BRW_ATTRIB_MAX*4]; /* current vertex */
-
-      GLfloat *current[BRW_ATTRIB_MAX]; /* points into ctx->Current, ctx->Light.Material */
-      GLfloat CurrentFloatEdgeFlag;
-
-      GLuint vert_count;
-      GLuint max_vert;
-      struct brw_exec_copied_vtx copied;
-
-      GLubyte attrsz[BRW_ATTRIB_MAX];
-      GLubyte active_sz[BRW_ATTRIB_MAX];
-
-      GLfloat *attrptr[BRW_ATTRIB_MAX]; 
-      struct gl_client_array arrays[BRW_ATTRIB_MAX];
-      const struct gl_client_array *inputs[BRW_ATTRIB_MAX];
-   } vtx;
-
-   
-   struct {
-      GLboolean recalculate_maps;
-      struct brw_exec_eval1_map map1[BRW_ATTRIB_MAX];
-      struct brw_exec_eval2_map map2[BRW_ATTRIB_MAX];
-   } eval;
-
-   struct {
-      const struct gl_client_array *inputs[BRW_ATTRIB_MAX];
-
-      struct gl_buffer_object *index_obj;
-   } array;
-};
-
-
-
-/* External API:
- */
-void brw_exec_init( GLcontext *ctx );
-void brw_exec_destroy( GLcontext *ctx );
-void brw_exec_invalidate_state( GLcontext *ctx, GLuint new_state );
-void brw_exec_FlushVertices( GLcontext *ctx, GLuint flags );
-void brw_exec_wakeup( GLcontext *ctx );
-
-
-/* Internal functions:
- */
-void brw_exec_array_init( struct brw_exec_context *exec );
-void brw_exec_array_destroy( struct brw_exec_context *exec );
-
-
-void brw_exec_vtx_init( struct brw_exec_context *exec );
-void brw_exec_vtx_destroy( struct brw_exec_context *exec );
-void brw_exec_vtx_flush( struct brw_exec_context *exec );
-void brw_exec_vtx_wrap( struct brw_exec_context *exec );
-
-void brw_exec_eval_update( struct brw_exec_context *exec );
-
-void brw_exec_do_EvalCoord2f( struct brw_exec_context *exec, 
-				     GLfloat u, GLfloat v );
-
-void brw_exec_do_EvalCoord1f( struct brw_exec_context *exec,
-				     GLfloat u);
-
-#endif
diff --git a/src/mesa/drivers/dri/i965/brw_exec_array.c b/src/mesa/drivers/dri/i965/brw_exec_array.c
deleted file mode 100644
index ca19a198374..00000000000
--- a/src/mesa/drivers/dri/i965/brw_exec_array.c
+++ /dev/null
@@ -1,283 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
- * 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"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, 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 NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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.
- * 
- **************************************************************************/
-
-#include "glheader.h"
-#include "context.h"
-#include "state.h"
-#include "api_validate.h"
-#include "api_noop.h"
-#include "dispatch.h"
-
-#include "brw_attrib.h"
-#include "brw_draw.h"
-#include "brw_exec.h"
-#include "brw_fallback.h"
-
-static GLuint get_max_index( GLuint count, GLuint type, 
-			     const GLvoid *indices )
-{
-   GLint i;
-
-   /* Compute max element.  This is only needed for upload of non-VBO,
-    * non-constant data elements.
-    *
-    * XXX: Postpone this calculation until it is known that it is
-    * needed.  Otherwise could scan this pointlessly in the all-vbo
-    * case.
-    */
-   switch(type) {
-   case GL_UNSIGNED_INT: {
-      const GLuint *ui_indices = (const GLuint *)indices;
-      GLuint max_ui = 0;
-      for (i = 0; i < count; i++)
-	 if (ui_indices[i] > max_ui)
-	    max_ui = ui_indices[i];
-      return max_ui;
-   }
-   case GL_UNSIGNED_SHORT: {
-      const GLushort *us_indices = (const GLushort *)indices;
-      GLuint max_us = 0;
-      for (i = 0; i < count; i++)
-	 if (us_indices[i] > max_us)
-	    max_us = us_indices[i];
-      return max_us;
-   }
-   case GL_UNSIGNED_BYTE: {
-      const GLubyte *ub_indices = (const GLubyte *)indices;
-      GLuint max_ub = 0;
-      for (i = 0; i < count; i++)
-	 if (ub_indices[i] > max_ub)
-	    max_ub = ub_indices[i];
-      return max_ub;
-   }
-   default:
-      return 0;
-   }
-}
-
-
-
-
-/***********************************************************************
- * API functions.
- */
-
-static void GLAPIENTRY
-brw_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
-   struct brw_draw_prim prim[1];
-   GLboolean ok;
-
-   if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
-      return;
-
-   FLUSH_CURRENT( ctx, 0 );
-
-   if (ctx->NewState)
-      _mesa_update_state( ctx );
-      
-   prim[0].begin = 1;
-   prim[0].end = 1;
-   prim[0].weak = 0;
-   prim[0].pad = 0;
-
-   if (exec->array.inputs[0]->BufferObj->Name) {
-      /* Use vertex attribute as a hint to tell us if we expect all
-       * arrays to be in VBO's and if so, don't worry about avoiding
-       * the upload of elements < start.
-       */
-      prim[0].mode = mode;
-      prim[0].start = start;
-      prim[0].count = count;
-      prim[0].indexed = 0;
-
-      ok = brw_draw_prims( ctx, exec->array.inputs, prim, 1, NULL, 0, start + count, 0 );
-   }
-   else {
-      /* If not using VBO's, we don't want to upload any more elements
-       * than necessary from the arrays as they will not be valid next
-       * time the application tries to draw with them.
-       */
-      prim[0].mode = mode;
-      prim[0].start = 0;
-      prim[0].count = count;
-      prim[0].indexed = 0;
-
-      ok = brw_draw_prims( ctx, exec->array.inputs, prim, 1, NULL, start, start + count, 0 );
-   }
-   
-   if (!ok) {
-      brw_fallback(ctx);
-      CALL_DrawArrays(ctx->Exec, ( mode, start, count ));
-      brw_unfallback(ctx);
-   }
-}
-
-
-
-static void GLAPIENTRY
-brw_exec_DrawRangeElements(GLenum mode,
-			   GLuint start, GLuint end,
-			   GLsizei count, GLenum type, const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
-   struct brw_draw_index_buffer ib;
-   struct brw_draw_prim prim[1];
-
-   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count, type, indices ))
-      return;
-
-   FLUSH_CURRENT( ctx, 0 );
-
-   if (ctx->NewState)
-      _mesa_update_state( ctx );
-      
-   ib.count = count;
-   ib.type = type; 
-   ib.obj = ctx->Array.ElementArrayBufferObj;
-   ib.ptr = indices;
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      /* Use the fact that indices are in a VBO as a hint that the
-       * program has put all the arrays in VBO's and we don't have to
-       * worry about performance implications of start > 0.
-       *
-       * XXX: consider passing start as min_index to draw_prims instead.
-       */
-      ib.rebase = 0;
-   }
-   else {
-      ib.rebase = start;
-   }
-
-   prim[0].begin = 1;
-   prim[0].end = 1;
-   prim[0].weak = 0;
-   prim[0].pad = 0;
-   prim[0].mode = mode;
-   prim[0].start = 0;
-   prim[0].count = count;
-   prim[0].indexed = 1;
-
-   if (!brw_draw_prims( ctx, exec->array.inputs, prim, 1, &ib, ib.rebase, end+1, 0 )) {
-      brw_fallback(ctx);
-      CALL_DrawRangeElements(ctx->Exec, (mode, start, end, count, type, indices));
-      brw_unfallback(ctx);
-   }
-}
-
-
-static void GLAPIENTRY
-brw_exec_DrawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLuint max_index;
-
-   if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
-      return;
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      const GLvoid *map = ctx->Driver.MapBuffer(ctx,
-						 GL_ELEMENT_ARRAY_BUFFER_ARB,
-						 GL_DYNAMIC_READ_ARB,
-						 ctx->Array.ElementArrayBufferObj);
-
-      max_index = get_max_index(count, type, ADD_POINTERS(map, indices));
-
-      ctx->Driver.UnmapBuffer(ctx,
-			      GL_ELEMENT_ARRAY_BUFFER_ARB,
-			      ctx->Array.ElementArrayBufferObj);
-   }
-   else {
-      max_index = get_max_index(count, type, indices);
-   }
-
-   brw_exec_DrawRangeElements(mode, 0, max_index, count, type, indices);
-}
-
-
-/***********************************************************************
- * Initialization
- */
-
-
-static void init_arrays( GLcontext *ctx, 
-			 const struct gl_client_array *arrays[] )
-{
-   struct gl_array_object *obj = ctx->Array.ArrayObj;
-   GLuint i;
-
-   memset(arrays, 0, sizeof(*arrays) * BRW_ATTRIB_MAX);
-
-   arrays[BRW_ATTRIB_POS] = &obj->Vertex;
-   arrays[BRW_ATTRIB_NORMAL] = &obj->Normal;
-   arrays[BRW_ATTRIB_COLOR0] = &obj->Color;
-   arrays[BRW_ATTRIB_COLOR1] = &obj->SecondaryColor;
-   arrays[BRW_ATTRIB_FOG] = &obj->FogCoord;
-
-   for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
-      arrays[BRW_ATTRIB_TEX0 + i] = &obj->TexCoord[i];
-
-   arrays[BRW_ATTRIB_INDEX] = &obj->Index;
-   arrays[BRW_ATTRIB_EDGEFLAG] = &obj->EdgeFlag;
-
-   for (i = BRW_ATTRIB_GENERIC0; i <= BRW_ATTRIB_GENERIC15; i++)
-      arrays[i] = &obj->VertexAttrib[i - BRW_ATTRIB_GENERIC0];
-}
-
-
-
-
-void brw_exec_array_init( struct brw_exec_context *exec )
-{
-   GLcontext *ctx = exec->ctx;
-
-   init_arrays(ctx, exec->array.inputs);
-
-#if 1
-   exec->vtxfmt.DrawArrays = brw_exec_DrawArrays;
-   exec->vtxfmt.DrawElements = brw_exec_DrawElements;
-   exec->vtxfmt.DrawRangeElements = brw_exec_DrawRangeElements;
-#else
-   exec->vtxfmt.DrawArrays = _mesa_noop_DrawArrays;
-   exec->vtxfmt.DrawElements = _mesa_noop_DrawElements;
-   exec->vtxfmt.DrawRangeElements = _mesa_noop_DrawRangeElements;
-#endif
-
-   exec->array.index_obj = ctx->Driver.NewBufferObject(ctx, 1, GL_ARRAY_BUFFER_ARB);
-}
-
-
-void brw_exec_array_destroy( struct brw_exec_context *exec )
-{
-   GLcontext *ctx = exec->ctx;
-
-   ctx->Driver.DeleteBuffer(ctx, exec->array.index_obj);
-}
diff --git a/src/mesa/drivers/dri/i965/brw_fallback.c b/src/mesa/drivers/dri/i965/brw_fallback.c
index 56e21e339ef..86464b2ec5f 100644
--- a/src/mesa/drivers/dri/i965/brw_fallback.c
+++ b/src/mesa/drivers/dri/i965/brw_fallback.c
@@ -30,8 +30,6 @@
 #include "tnl/tnl.h"
 #include "context.h"
 #include "brw_context.h"
-#include "brw_exec.h"
-#include "brw_save.h"
 #include "brw_fallback.h"
 
 #include "glheader.h"
@@ -40,297 +38,6 @@
 #include "imports.h"
 #include "macros.h"
 #include "mtypes.h"
-#include "dispatch.h"
-
-
-typedef void (*attr_func)( GLcontext *ctx, GLint target, const GLfloat * );
-
-
-/* Wrapper functions in case glVertexAttrib*fvNV doesn't exist */
-static void VertexAttrib1fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvNV(ctx->Exec, (target, v));
-}
-
-static attr_func vert_attrfunc[4] = {
-   VertexAttrib1fvNV,
-   VertexAttrib2fvNV,
-   VertexAttrib3fvNV,
-   VertexAttrib4fvNV
-};
-
-#if 0
-static void VertexAttrib1fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvARB(ctx->Exec, (target, v));
-}
-
-
-static attr_func vert_attrfunc_arb[4] = {
-   VertexAttrib1fvARB,
-   VertexAttrib2fvARB,
-   VertexAttrib3fvARB,
-   VertexAttrib4fvARB
-};
-#endif
-
-
-
-
-
-
-static void mat_attr1fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case BRW_ATTRIB_MAT_FRONT_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SHININESS, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SHININESS, v ));
-      break;
-   }
-}
-
-
-static void mat_attr3fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case BRW_ATTRIB_MAT_FRONT_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_COLOR_INDEXES, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_COLOR_INDEXES, v ));
-      break;
-   }
-}
-
-
-static void mat_attr4fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case BRW_ATTRIB_MAT_FRONT_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_EMISSION, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_EMISSION, v ));
-      break;
-   case BRW_ATTRIB_MAT_FRONT_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_AMBIENT, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_AMBIENT, v ));
-      break;
-   case BRW_ATTRIB_MAT_FRONT_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_DIFFUSE, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_DIFFUSE, v ));
-      break;
-   case BRW_ATTRIB_MAT_FRONT_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SPECULAR, v ));
-      break;
-   case BRW_ATTRIB_MAT_BACK_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SPECULAR, v ));
-      break;
-   }
-}
-
-
-static attr_func mat_attrfunc[4] = {
-   mat_attr1fv,
-   NULL,
-   mat_attr3fv,
-   mat_attr4fv
-};
-
-
-static void index_attr1fv(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   (void) target;
-   CALL_Indexf(ctx->Exec, (v[0]));
-}
-
-static void edgeflag_attr1fv(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   (void) target;
-   CALL_EdgeFlag(ctx->Exec, ((GLboolean)(v[0] == 1.0)));
-}
-
-struct loopback_attr {
-   GLint target;
-   GLint sz;
-   attr_func func;
-};
-
-/* Don't emit ends and begins on wrapped primitives.  Don't replay
- * wrapped vertices.  If we get here, it's probably because the the
- * precalculated wrapping is wrong.
- */
-static void loopback_prim( GLcontext *ctx,
-			   const GLfloat *buffer,
-			   const struct brw_draw_prim *prim,
-			   GLuint wrap_count,
-			   GLuint vertex_size,
-			   const struct loopback_attr *la, GLuint nr )
-{
-   GLint start = prim->start;
-   GLint end = start + prim->count;
-   const GLfloat *data;
-   GLint j;
-   GLuint k;
-
-   if (0)
-      _mesa_printf("loopback prim %s(%s,%s) verts %d..%d\n",
-		   _mesa_lookup_enum_by_nr(prim->mode),
-		   prim->begin ? "begin" : "..",
-		   prim->end ? "end" : "..",
-		   start, 
-		   end);
-
-   if (prim->begin) {
-      CALL_Begin(GET_DISPATCH(), ( prim->mode ));
-   }
-   else {
-      assert(start == 0);
-      start += wrap_count;
-   }
-
-   data = buffer + start * vertex_size;
-
-   for (j = start ; j < end ; j++) {
-      const GLfloat *tmp = data + la[0].sz;
-
-      for (k = 1 ; k < nr ; k++) {
-	 la[k].func( ctx, la[k].target, tmp );
-	 tmp += la[k].sz;
-      }
-	 
-      /* Fire the vertex
-       */
-      la[0].func( ctx, BRW_ATTRIB_POS, data );
-      data = tmp;
-   }
-
-   if (prim->end) {
-      CALL_End(GET_DISPATCH(), ());
-   }
-}
-
-/* Primitives generated by DrawArrays/DrawElements/Rectf may be
- * caught here.  If there is no primitive in progress, execute them
- * normally, otherwise need to track and discard the generated
- * primitives.
- */
-static void loopback_weak_prim( GLcontext *ctx,
-				const struct brw_draw_prim *prim )
-{
-   /* Use the prim_weak flag to ensure that if this primitive
-    * wraps, we don't mistake future vertex_lists for part of the
-    * surrounding primitive.
-    *
-    * While this flag is set, we are simply disposing of data
-    * generated by an operation now known to be a noop.
-    */
-   if (prim->begin)
-      ctx->Driver.CurrentExecPrimitive |= BRW_SAVE_PRIM_WEAK;
-   if (prim->end)
-      ctx->Driver.CurrentExecPrimitive &= ~BRW_SAVE_PRIM_WEAK;
-}
-
-
-void brw_loopback_vertex_list( GLcontext *ctx,
-			       const GLfloat *buffer,
-			       const GLubyte *attrsz,
-			       const struct brw_draw_prim *prim,
-			       GLuint prim_count,
-			       GLuint wrap_count,
-			       GLuint vertex_size)
-{
-   struct loopback_attr la[BRW_ATTRIB_MAX];
-   GLuint i, nr = 0;
-
-   for (i = 0 ; i <= BRW_ATTRIB_TEX7 ; i++) {
-      if (i == BRW_ATTRIB_INDEX || i == BRW_ATTRIB_EDGEFLAG)
-                   continue;
-
-      if (attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = attrsz[i];
-	 la[nr].func = vert_attrfunc[attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   for (i = BRW_ATTRIB_MAT_FRONT_AMBIENT ; 
-	i <= BRW_ATTRIB_MAT_BACK_INDEXES ; 
-	i++) {
-      if (attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = attrsz[i];
-	 la[nr].func = mat_attrfunc[attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   if (attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      la[nr].target = BRW_ATTRIB_EDGEFLAG;
-      la[nr].sz = attrsz[BRW_ATTRIB_EDGEFLAG];
-      la[nr].func = edgeflag_attr1fv;
-      nr++;
-   }
-
-   if (attrsz[BRW_ATTRIB_INDEX]) {
-      la[nr].target = BRW_ATTRIB_INDEX;
-      la[nr].sz = attrsz[BRW_ATTRIB_INDEX];
-      la[nr].func = index_attr1fv;
-      nr++;
-   }
-
-   /* XXX ARB vertex attribs */
-
-   for (i = 0 ; i < prim_count ; i++) {
-      if ((prim[i].mode & BRW_SAVE_PRIM_WEAK) &&
-	  (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END))
-      {
-	 loopback_weak_prim( ctx, &prim[i] );
-      }
-      else
-      {
-	 loopback_prim( ctx, buffer, &prim[i], wrap_count, vertex_size, la, nr );
-      }
-   }
-}
-
 
 
 
@@ -405,62 +112,6 @@ const struct brw_tracked_state brw_check_fallback = {
 
 
 
-/* If there is a fallback, fallback to software rasterization and
- * transformation together.  There is never a requirement to have
- * software t&l but hardware rasterization.
- * 
- * Further, all fallbacks are based on GL state, not on eg. primitive
- * or vertex data.
- */
-
-static void do_fallback( struct brw_context *brw,
-			 GLboolean fallback )
-{
-   GLcontext *ctx = &brw->intel.ctx;
-
-   /* flush:
-    */
-   ctx->Driver.Flush( ctx );
-
-   if (fallback) {
-      _swsetup_Wakeup( ctx );
-      _tnl_wakeup_exec( ctx );	
-
-      /* Need this because tnl_wakeup_exec does too much: 
-       */
-      brw_save_wakeup(ctx);
-      brw_save_fallback(ctx, GL_TRUE);
-   }
-   else {
-      /* Flush vertices and copy-to-current:
-       */
-      FLUSH_CURRENT(ctx, 0); 
-
-      _swrast_flush( ctx );
-
-      brw_exec_wakeup(ctx);
-
-      /* Need this because tnl_wakeup_exec does too much: 
-       */
-      brw_save_wakeup(ctx);
-      brw_save_fallback(ctx, GL_FALSE);	 
-   }
-}
-
-
-void brw_fallback( GLcontext *ctx )
-{
-   struct brw_context *brw = brw_context(ctx);
-   do_fallback(brw, 1);
-}
-
-
-void brw_unfallback( GLcontext *ctx )
-{
-   struct brw_context *brw = brw_context(ctx);
-   do_fallback(brw, 0);
-}
-
 /* Not used:
  */
 void intelFallback( struct intel_context *intel, GLuint bit, GLboolean mode )
diff --git a/src/mesa/drivers/dri/i965/brw_fallback.h b/src/mesa/drivers/dri/i965/brw_fallback.h
index 81a2d344b81..684a46cd170 100644
--- a/src/mesa/drivers/dri/i965/brw_fallback.h
+++ b/src/mesa/drivers/dri/i965/brw_fallback.h
@@ -31,7 +31,7 @@
 #include "mtypes.h"		/* for GLcontext... */
 
 struct brw_context;
-struct brw_draw_prim;
+struct vbo_prim;
 
 void brw_fallback( GLcontext *ctx );
 void brw_unfallback( GLcontext *ctx );
@@ -39,7 +39,7 @@ void brw_unfallback( GLcontext *ctx );
 void brw_loopback_vertex_list( GLcontext *ctx,
 			       const GLfloat *buffer,
 			       const GLubyte *attrsz,
-			       const struct brw_draw_prim *prim,
+			       const struct vbo_prim *prim,
 			       GLuint prim_count,
 			       GLuint wrap_count,
 			       GLuint vertex_size);
diff --git a/src/mesa/drivers/dri/i965/brw_gs.c b/src/mesa/drivers/dri/i965/brw_gs.c
index 7d3f9dd5e3b..73263a5fff4 100644
--- a/src/mesa/drivers/dri/i965/brw_gs.c
+++ b/src/mesa/drivers/dri/i965/brw_gs.c
@@ -66,7 +66,9 @@ static void compile_gs_prog( struct brw_context *brw,
    /* Begin the compilation:
     */
    brw_init_compile(&c.func);
-	
+
+   c.func.single_program_flow = 1;
+
    /* For some reason the thread is spawned with only 4 channels
     * unmasked.  
     */
@@ -80,6 +82,9 @@ static void compile_gs_prog( struct brw_context *brw,
    case GL_QUADS:
       brw_gs_quads( &c ); 
       break;
+   case GL_QUAD_STRIP:
+      brw_gs_quad_strip( &c );
+      break;
    case GL_LINE_LOOP:
       brw_gs_lines( &c );
       break;
@@ -143,7 +148,7 @@ static const GLenum gs_prim[GL_POLYGON+1] = {
    GL_TRIANGLES,
    GL_TRIANGLES,
    GL_QUADS,
-   GL_QUADS,
+   GL_QUAD_STRIP,
    GL_TRIANGLES
 };
 
diff --git a/src/mesa/drivers/dri/i965/brw_gs.h b/src/mesa/drivers/dri/i965/brw_gs.h
index f9aa71d9199..29a4e80ce1b 100644
--- a/src/mesa/drivers/dri/i965/brw_gs.h
+++ b/src/mesa/drivers/dri/i965/brw_gs.h
@@ -67,6 +67,7 @@ struct brw_gs_compile {
 #define ATTR_SIZE  (4*4)
 
 void brw_gs_quads( struct brw_gs_compile *c );
+void brw_gs_quad_strip( struct brw_gs_compile *c );
 void brw_gs_tris( struct brw_gs_compile *c );
 void brw_gs_lines( struct brw_gs_compile *c );
 void brw_gs_points( struct brw_gs_compile *c );
diff --git a/src/mesa/drivers/dri/i965/brw_gs_emit.c b/src/mesa/drivers/dri/i965/brw_gs_emit.c
index e4eed36a464..9abb94d82ed 100644
--- a/src/mesa/drivers/dri/i965/brw_gs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_gs_emit.c
@@ -116,6 +116,16 @@ void brw_gs_quads( struct brw_gs_compile *c )
    brw_gs_emit_vue(c, c->reg.vertex[2], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END));
 }
 
+void brw_gs_quad_strip( struct brw_gs_compile *c )
+{
+   brw_gs_alloc_regs(c, 4);
+   
+   brw_gs_emit_vue(c, c->reg.vertex[2], 0, ((_3DPRIM_POLYGON << 2) | R02_PRIM_START));
+   brw_gs_emit_vue(c, c->reg.vertex[3], 0, (_3DPRIM_POLYGON << 2));
+   brw_gs_emit_vue(c, c->reg.vertex[0], 0, (_3DPRIM_POLYGON << 2)); 
+   brw_gs_emit_vue(c, c->reg.vertex[1], 1, ((_3DPRIM_POLYGON << 2) | R02_PRIM_END));
+}
+
 void brw_gs_tris( struct brw_gs_compile *c )
 {
    brw_gs_alloc_regs(c, 3);
diff --git a/src/mesa/drivers/dri/i965/brw_metaops.c b/src/mesa/drivers/dri/i965/brw_metaops.c
index 2deec5eae3d..1579762b6d1 100644
--- a/src/mesa/drivers/dri/i965/brw_metaops.c
+++ b/src/mesa/drivers/dri/i965/brw_metaops.c
@@ -47,7 +47,6 @@
 #include "brw_context.h"
 #include "brw_defines.h"
 #include "brw_draw.h"
-#include "brw_attrib.h"
 #include "brw_fallback.h"
 
 #define INIT(brw, STRUCT, ATTRIB) 		\
@@ -388,8 +387,8 @@ static void meta_draw_quad(struct intel_context *intel,
    struct brw_context *brw = brw_context(&intel->ctx);
    struct gl_client_array pos_array;
    struct gl_client_array color_array;
-   struct gl_client_array *attribs[BRW_ATTRIB_MAX];
-   struct brw_draw_prim prim[1];
+   struct gl_client_array *attribs[VERT_ATTRIB_MAX];
+   struct _mesa_prim prim[1];
    GLfloat pos[4][3];
    GLubyte color[4];
 
@@ -439,29 +438,29 @@ static void meta_draw_quad(struct intel_context *intel,
    /* Ignoring texture coords. 
     */
 
-   memset(attribs, 0, BRW_ATTRIB_MAX * sizeof(*attribs));
-
-   attribs[BRW_ATTRIB_POS] = &pos_array;
-   attribs[BRW_ATTRIB_POS]->Ptr = 0;
-   attribs[BRW_ATTRIB_POS]->Type = GL_FLOAT;
-   attribs[BRW_ATTRIB_POS]->Enabled = 1;
-   attribs[BRW_ATTRIB_POS]->Size = 3;
-   attribs[BRW_ATTRIB_POS]->StrideB = 3 * sizeof(GLfloat);
-   attribs[BRW_ATTRIB_POS]->Stride = 3 * sizeof(GLfloat);
-   attribs[BRW_ATTRIB_POS]->_MaxElement = 4;
-   attribs[BRW_ATTRIB_POS]->Normalized = 0;
-   attribs[BRW_ATTRIB_POS]->BufferObj = brw->metaops.vbo;
-
-   attribs[BRW_ATTRIB_COLOR0] = &color_array;
-   attribs[BRW_ATTRIB_COLOR0]->Ptr = (const GLubyte *)sizeof(pos);
-   attribs[BRW_ATTRIB_COLOR0]->Type = GL_UNSIGNED_BYTE;
-   attribs[BRW_ATTRIB_COLOR0]->Enabled = 1;
-   attribs[BRW_ATTRIB_COLOR0]->Size = 4;
-   attribs[BRW_ATTRIB_COLOR0]->StrideB = 0;
-   attribs[BRW_ATTRIB_COLOR0]->Stride = 0;
-   attribs[BRW_ATTRIB_COLOR0]->_MaxElement = 1;
-   attribs[BRW_ATTRIB_COLOR0]->Normalized = 1;
-   attribs[BRW_ATTRIB_COLOR0]->BufferObj = brw->metaops.vbo;
+   memset(attribs, 0, VERT_ATTRIB_MAX * sizeof(*attribs));
+
+   attribs[VERT_ATTRIB_POS] = &pos_array;
+   attribs[VERT_ATTRIB_POS]->Ptr = 0;
+   attribs[VERT_ATTRIB_POS]->Type = GL_FLOAT;
+   attribs[VERT_ATTRIB_POS]->Enabled = 1;
+   attribs[VERT_ATTRIB_POS]->Size = 3;
+   attribs[VERT_ATTRIB_POS]->StrideB = 3 * sizeof(GLfloat);
+   attribs[VERT_ATTRIB_POS]->Stride = 3 * sizeof(GLfloat);
+   attribs[VERT_ATTRIB_POS]->_MaxElement = 4;
+   attribs[VERT_ATTRIB_POS]->Normalized = 0;
+   attribs[VERT_ATTRIB_POS]->BufferObj = brw->metaops.vbo;
+
+   attribs[VERT_ATTRIB_COLOR0] = &color_array;
+   attribs[VERT_ATTRIB_COLOR0]->Ptr = (const GLubyte *)sizeof(pos);
+   attribs[VERT_ATTRIB_COLOR0]->Type = GL_UNSIGNED_BYTE;
+   attribs[VERT_ATTRIB_COLOR0]->Enabled = 1;
+   attribs[VERT_ATTRIB_COLOR0]->Size = 4;
+   attribs[VERT_ATTRIB_COLOR0]->StrideB = 0;
+   attribs[VERT_ATTRIB_COLOR0]->Stride = 0;
+   attribs[VERT_ATTRIB_COLOR0]->_MaxElement = 1;
+   attribs[VERT_ATTRIB_COLOR0]->Normalized = 1;
+   attribs[VERT_ATTRIB_COLOR0]->BufferObj = brw->metaops.vbo;
    
    /* Just ignoring texture coordinates for now. 
     */
@@ -476,24 +475,16 @@ static void meta_draw_quad(struct intel_context *intel,
    prim[0].start = 0;
    prim[0].count = 4;
 
-   if (!brw_draw_prims(&brw->intel.ctx, 
-		       (const struct gl_client_array **)attribs,
-		       prim, 1,
-		       NULL,
-		       0,
-		       4,
-		       BRW_DRAW_LOCKED ))
-   {
-      /* This should not be possible:
-       */
-      _mesa_printf("brw_draw_prims failed in metaops!\n");
-      assert(0);
-   }
+   brw_draw_prims(&brw->intel.ctx, 
+		  (const struct gl_client_array **)attribs,
+		  prim, 1,
+		  NULL,
+		  0,
+		  3 );
 }
 
 
-static void install_meta_state( struct intel_context *intel,
-				GLenum state )
+static void install_meta_state( struct intel_context *intel )
 {
    GLcontext *ctx = &intel->ctx;
    struct brw_context *brw = brw_context(ctx);
diff --git a/src/mesa/drivers/dri/i965/brw_save.c b/src/mesa/drivers/dri/i965/brw_save.c
deleted file mode 100644
index 1af7791c4d4..00000000000
--- a/src/mesa/drivers/dri/i965/brw_save.c
+++ /dev/null
@@ -1,126 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.3
- *
- * Copyright (C) 1999-2005  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"),
- * 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 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
- * BRIAN PAUL 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:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-
-#include "mtypes.h"
-#include "api_arrayelt.h"
-#include "dlist.h"
-#include "vtxfmt.h"
-#include "imports.h"
-
-#include "brw_save.h"
-
-
-
-void brw_save_init( GLcontext *ctx )
-{
-   struct brw_save_context *save = CALLOC_STRUCT(brw_save_context);
-
-   if (ctx->swtnl_im == NULL) {
-      ctx->swtnl_im = CALLOC_STRUCT(brw_exec_save);
-   }
-
-   save->ctx = ctx;
-   IMM_CONTEXT(ctx)->save = save;
-
-   /* Initialize the arrayelt helper
-    */
-   if (!ctx->aelt_context &&
-       !_ae_create_context( ctx )) 
-      return;
-
-   brw_save_api_init( save );
-   brw_save_wakeup(ctx);
-
-   ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
-}
-
-
-void brw_save_destroy( GLcontext *ctx )
-{
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
-   if (save) {
-      FREE(save);
-      IMM_CONTEXT(ctx)->save = NULL;
-   }
-
-   if (ctx->aelt_context) {
-      _ae_destroy_context( ctx );
-      ctx->aelt_context = NULL;
-   }
-
-   if (IMM_CONTEXT(ctx)->exec == NULL &&
-       IMM_CONTEXT(ctx)->save == NULL) {
-      FREE(IMM_CONTEXT(ctx));
-      ctx->swtnl_im = NULL;
-   }
-}
-
-
-void brw_save_invalidate_state( GLcontext *ctx, GLuint new_state )
-{
-   _ae_invalidate_state(ctx, new_state);
-}
-
-
-/* Note that this can occur during the playback of a display list:
- */
-void brw_save_fallback( GLcontext *ctx, GLboolean fallback )
-{
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
-
-   if (fallback)
-      save->replay_flags |= BRW_SAVE_FALLBACK;
-   else
-      save->replay_flags &= ~BRW_SAVE_FALLBACK;
-}
-
-
-/* I don't see any reason to swap this code out on fallbacks.  It
- * wouldn't really mean anything to do so anyway as the old lists are
- * still around from pre-fallback.  Instead, the above code ensures
- * that vertices are routed back through immediate mode dispatch on
- * fallback.
- *
- * The below can be moved into init or removed:
- */
-void brw_save_wakeup( GLcontext *ctx )
-{
-   ctx->Driver.NewList = brw_save_NewList;
-   ctx->Driver.EndList = brw_save_EndList;
-   ctx->Driver.SaveFlushVertices = brw_save_SaveFlushVertices;
-   ctx->Driver.BeginCallList = brw_save_BeginCallList;
-   ctx->Driver.EndCallList = brw_save_EndCallList;
-   ctx->Driver.NotifySaveBegin = brw_save_NotifyBegin;
-
-   /* Assume we haven't been getting state updates either:
-    */
-   brw_save_invalidate_state( ctx, ~0 );
-}
-
-
-
diff --git a/src/mesa/drivers/dri/i965/brw_structs.h b/src/mesa/drivers/dri/i965/brw_structs.h
index 25acdcfe947..10fee944e8d 100644
--- a/src/mesa/drivers/dri/i965/brw_structs.h
+++ b/src/mesa/drivers/dri/i965/brw_structs.h
@@ -519,7 +519,22 @@ struct thread3
 struct brw_clip_unit_state
 {
    struct thread0 thread0;
-   struct thread1 thread1;
+   struct
+   {
+      GLuint pad0:7;
+      GLuint sw_exception_enable:1;
+      GLuint pad1:3;
+      GLuint mask_stack_exception_enable:1;
+      GLuint pad2:1;
+      GLuint illegal_op_exception_enable:1;
+      GLuint pad3:2;
+      GLuint floating_point_mode:1;
+      GLuint thread_priority:1;
+      GLuint binding_table_entry_count:8;
+      GLuint pad4:5;
+      GLuint single_program_flow:1;
+   } thread1;
+
    struct thread2 thread2;
    struct thread3 thread3;
 
@@ -532,8 +547,8 @@ struct brw_clip_unit_state
       GLuint pad1:1;
       GLuint urb_entry_allocation_size:5; 
       GLuint pad2:1;
-      GLuint max_threads:6; 	/* may be less */
-      GLuint pad3:1;
+      GLuint max_threads:1; 	/* may be less */
+      GLuint pad3:6;
    } thread4;   
       
    struct
@@ -1322,6 +1337,7 @@ struct brw_instruction
 	 GLuint end_of_thread:1;
       } generic;
 
+      GLint d;
       GLuint ud;
    } bits3;
 };
diff --git a/src/mesa/drivers/dri/i965/brw_tex_layout.c b/src/mesa/drivers/dri/i965/brw_tex_layout.c
index 1353325afff..af1ad0f1ef1 100644
--- a/src/mesa/drivers/dri/i965/brw_tex_layout.c
+++ b/src/mesa/drivers/dri/i965/brw_tex_layout.c
@@ -34,21 +34,15 @@
  */
 
 #include "intel_mipmap_tree.h"
+#include "intel_tex_layout.h"
 #include "macros.h"
 
-static GLuint minify( GLuint d )
-{
-   return MAX2(1, d>>1);
-}
-
 
 GLboolean brw_miptree_layout( struct intel_mipmap_tree *mt )
 {
    /* XXX: these vary depending on image format: 
     */
 /*    GLint align_w = 4; */
-   GLint align_h = 2;
-
 
    switch (mt->target) {
    case GL_TEXTURE_CUBE_MAP: 
@@ -107,53 +101,10 @@ GLboolean brw_miptree_layout( struct intel_mipmap_tree *mt )
       break;
    }
 
-   default: {
-      GLuint level;
-      GLuint x = 0;
-      GLuint y = 0;
-      GLuint width = mt->width0;
-      GLuint height = mt->height0;
-
-      mt->pitch = ((mt->width0 * mt->cpp + 3) & ~3) / mt->cpp;
-      mt->total_height = 0;
-
-      for ( level = mt->first_level ; level <= mt->last_level ; level++ ) {
-	 GLuint img_height;
-
-	 intel_miptree_set_level_info(mt, level, 1,
-				      x, y,
-				      width, 
-				      mt->compressed ? height/4 : height, 1);
-
-	 if (mt->compressed)
-	    img_height = MAX2(1, height/4);
-	 else
-	    img_height = MAX2(align_h, height);
-
-
-	 /* Because the images are packed better, the final offset
-	  * might not be the maximal one:
-	  */
-	 mt->total_height = MAX2(mt->total_height, y + img_height);
-	 
-	 /* Layout_below: step right after second mipmap.
-	  */
-	 if (level == mt->first_level + 1) {
-	    x += mt->pitch / 2;
-	    x = (x + 3) & ~ 3;
-	 }
-	 else {
-	    y += img_height;
-	    y += align_h - 1;
-	    y &= ~(align_h - 1);
-	 }
-
-	 width  = minify(width);
-	 height = minify(height);
-      }
+   default:
+      i945_miptree_layout_2d(mt);
       break;
    }
-   }
    DBG("%s: %dx%dx%d - sz 0x%x\n", __FUNCTION__, 
 		mt->pitch, 
 		mt->total_height,
diff --git a/src/mesa/drivers/dri/i965/brw_vs.c b/src/mesa/drivers/dri/i965/brw_vs.c
index e5a28b96e32..74c9d88e464 100644
--- a/src/mesa/drivers/dri/i965/brw_vs.c
+++ b/src/mesa/drivers/dri/i965/brw_vs.c
@@ -54,12 +54,11 @@ static void do_vs_prog( struct brw_context *brw,
    c.vp = vp;
 
    c.prog_data.outputs_written = vp->program.Base.OutputsWritten;
-   c.prog_data.inputs_read = brw_translate_inputs(brw->intel.ctx.VertexProgram._Enabled,
-						  vp->program.Base.InputsRead);
+   c.prog_data.inputs_read = vp->program.Base.InputsRead;
 
    if (c.key.copy_edgeflag) {
       c.prog_data.outputs_written |= 1<<VERT_RESULT_EDGE;
-      c.prog_data.inputs_read |= 1<<BRW_ATTRIB_EDGEFLAG;
+      c.prog_data.inputs_read |= 1<<VERT_ATTRIB_EDGEFLAG;
    }
 
    if (0)
diff --git a/src/mesa/drivers/dri/i965/brw_vs_constval.c b/src/mesa/drivers/dri/i965/brw_vs_constval.c
index 502d8283970..528e164db8c 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_constval.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_constval.c
@@ -162,6 +162,7 @@ static GLuint get_input_size(struct brw_context *brw,
    GLuint sizes_dword = brw->vb.info.sizes[attr/16];
    GLuint sizes_bits = (sizes_dword>>((attr%16)*2)) & 0x3;
    return sizes_bits + 1;
+/*    return brw->vb.inputs[attr].glarray->Size; */
 }
 
 /* Calculate sizes of vertex program outputs.  Size is the largest
@@ -176,8 +177,6 @@ static void calc_wm_input_sizes( struct brw_context *brw )
    struct tracker t;
    GLuint insn;
    GLuint i;
-   GLuint64EXT inputs = brw_translate_inputs(brw->intel.ctx.VertexProgram._Enabled,
-					     vp->program.Base.InputsRead);
 
    memset(&t, 0, sizeof(t));
 
@@ -185,8 +184,8 @@ static void calc_wm_input_sizes( struct brw_context *brw )
    if (brw->attribs.Light->Model.TwoSide)
       t.twoside = 1;
 
-   for (i = 0; i < BRW_ATTRIB_MAX; i++) 
-      if (inputs & (1<<i))
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) 
+      if (vp->program.Base.InputsRead & (1<<i))
 	 set_active_component(&t, PROGRAM_INPUT, i, 
 			      szflag[get_input_size(brw, i)]);
       
diff --git a/src/mesa/drivers/dri/i965/brw_vs_emit.c b/src/mesa/drivers/dri/i965/brw_vs_emit.c
index 8403e1bd7b6..ffdb843e85f 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_emit.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_emit.c
@@ -77,8 +77,8 @@ static void brw_vs_alloc_regs( struct brw_vs_compile *c )
    /* Allocate input regs:  
     */
    c->nr_inputs = 0;
-   for (i = 0; i < BRW_ATTRIB_MAX; i++) {
-      if (c->prog_data.inputs_read & ((GLuint64EXT)1<<i)) {
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      if (c->prog_data.inputs_read & (1<<i)) {
 	 c->nr_inputs++;
 	 c->regs[PROGRAM_INPUT][i] = brw_vec8_grf(reg, 0);
 	 reg++;
@@ -791,7 +791,7 @@ static void emit_vertex_write( struct brw_vs_compile *c)
    if (c->key.copy_edgeflag) {
       brw_MOV(p, 
 	      get_reg(c, PROGRAM_OUTPUT, VERT_RESULT_EDGE),
-	      get_reg(c, PROGRAM_INPUT, BRW_ATTRIB_EDGEFLAG));
+	      get_reg(c, PROGRAM_INPUT, VERT_ATTRIB_EDGEFLAG));
    }
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_vs_tnl.c b/src/mesa/drivers/dri/i965/brw_vs_tnl.c
index b7893ca3e55..0d61092247a 100644
--- a/src/mesa/drivers/dri/i965/brw_vs_tnl.c
+++ b/src/mesa/drivers/dri/i965/brw_vs_tnl.c
@@ -114,7 +114,7 @@ static GLuint translate_texgen( GLboolean enabled, GLenum mode )
 static void make_state_key( GLcontext *ctx, struct state_key *key )
 {
    struct brw_context *brw = brw_context(ctx);
-   struct gl_fragment_program *fp = brw->fragment_program;
+   const struct gl_fragment_program *fp = brw->fragment_program;
    GLuint i;
 
    /* This now relies on texenvprogram.c being active:
@@ -146,9 +146,13 @@ static void make_state_key( GLcontext *ctx, struct state_key *key )
       }
 
       /* BRW_NEW_INPUT_VARYING */
-      for (i = BRW_ATTRIB_MAT_FRONT_AMBIENT ; i < BRW_ATTRIB_MAX ; i++) 
-	 if (brw->vb.info.varying & ((GLuint64EXT)1<<i)) 
-	    key->light_material_mask |= 1<<(i-BRW_ATTRIB_MAT_FRONT_AMBIENT);
+
+      /* 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];
@@ -374,17 +378,10 @@ static void release_temps( struct tnl_program *p )
 
 static struct ureg register_input( struct tnl_program *p, GLuint input )
 {
-   GLuint orig_input = input;
-   /* Cram the material flags into the generic range.  We'll translate
-    * them back later.
-    */
-   if (input >= BRW_ATTRIB_MAT_FRONT_AMBIENT)
-      input -= BRW_ATTRIB_MAT_FRONT_AMBIENT - BRW_ATTRIB_GENERIC0;
-
    assert(input < 32);
 
    p->program->Base.InputsRead |= (1<<input);
-   return make_ureg(PROGRAM_INPUT, orig_input);
+   return make_ureg(PROGRAM_INPUT, input);
 }
 
 static struct ureg register_output( struct tnl_program *p, GLuint output )
@@ -647,7 +644,7 @@ static void emit_passthrough( struct tnl_program *p,
 static struct ureg get_eye_position( struct tnl_program *p )
 {
    if (is_undef(p->eye_position)) {
-      struct ureg pos = register_input( p, BRW_ATTRIB_POS ); 
+      struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
       struct ureg modelview[4];
 
       p->eye_position = reserve_temp(p);
@@ -710,7 +707,7 @@ static struct ureg get_eye_position_normalized( struct tnl_program *p )
 static struct ureg get_eye_normal( struct tnl_program *p )
 {
    if (is_undef(p->eye_normal)) {
-      struct ureg normal = register_input(p, BRW_ATTRIB_NORMAL );
+      struct ureg normal = register_input(p, VERT_ATTRIB_NORMAL );
       struct ureg mvinv[3];
 
       register_matrix_param6( p, STATE_MATRIX, STATE_MODELVIEW, 0, 0, 2,
@@ -743,7 +740,7 @@ static struct ureg get_eye_normal( struct tnl_program *p )
 
 static void build_hpos( struct tnl_program *p )
 {
-   struct ureg pos = register_input( p, BRW_ATTRIB_POS ); 
+   struct ureg pos = register_input( p, VERT_ATTRIB_POS ); 
    struct ureg hpos = register_output( p, VERT_RESULT_HPOS );
    struct ureg mvp[4];
 
@@ -787,9 +784,9 @@ static struct ureg get_material( struct tnl_program *p, GLuint side,
    GLuint attrib = material_attrib(side, property);
 
    if (p->color_materials & (1<<attrib))
-      return register_input(p, BRW_ATTRIB_COLOR0);
+      return register_input(p, VERT_ATTRIB_COLOR0);
    else if (p->materials & (1<<attrib)) 
-      return register_input( p, attrib + BRW_ATTRIB_MAT_FRONT_AMBIENT );
+      return register_input( p, attrib + _TNL_ATTRIB_MAT_FRONT_AMBIENT );
    else
       return register_param3( p, STATE_MATERIAL, side, property );
 }
@@ -1157,7 +1154,7 @@ static void build_fog( struct tnl_program *p )
       input = swizzle1(get_eye_position(p), Z);
    }
    else {
-      input = swizzle1(register_input(p, BRW_ATTRIB_FOG), X);
+      input = swizzle1(register_input(p, VERT_ATTRIB_FOG), X);
    }
 
    if (p->state->fog_option &&
@@ -1299,7 +1296,7 @@ static void build_texture_transform( struct tnl_program *p )
 	    for (j = 0; j < 4; j++) {
 	       switch (modes[j]) {
 	       case TXG_OBJ_LINEAR: {
-		  struct ureg obj = register_input(p, BRW_ATTRIB_POS);
+		  struct ureg obj = register_input(p, VERT_ATTRIB_POS);
 		  struct ureg plane = 
 		     register_param3(p, STATE_TEXGEN, i,
 				     STATE_TEXGEN_OBJECT_S + j);
@@ -1348,7 +1345,7 @@ static void build_texture_transform( struct tnl_program *p )
 	    }
 
 	    if (copy_mask) {
-	       struct ureg in = register_input(p, BRW_ATTRIB_TEX0+i);
+	       struct ureg in = register_input(p, VERT_ATTRIB_TEX0+i);
 	       emit_op1(p, OPCODE_MOV, out_texgen, copy_mask, in );
 	    }
 	 }
@@ -1357,7 +1354,7 @@ static void build_texture_transform( struct tnl_program *p )
 	    struct ureg texmat[4];
 	    struct ureg in = (!is_undef(out_texgen) ? 
 			      out_texgen : 
-			      register_input(p, BRW_ATTRIB_TEX0+i));
+			      register_input(p, VERT_ATTRIB_TEX0+i));
 	    if (PREFER_DP4) {
 	       register_matrix_param6( p, STATE_MATRIX, STATE_TEXTURE, i, 
 				       0, 3, STATE_MATRIX, texmat );
@@ -1373,7 +1370,7 @@ static void build_texture_transform( struct tnl_program *p )
 	 release_temps(p);
       } 
       else {
-	 emit_passthrough(p, BRW_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i);
+	 emit_passthrough(p, VERT_ATTRIB_TEX0+i, VERT_RESULT_TEX0+i);
       }
    }
 }
@@ -1425,10 +1422,10 @@ static void build_tnl_program( struct tnl_program *p )
 	 build_lighting(p);
       else {
 	 if (p->state->fragprog_inputs_read & FRAG_BIT_COL0)
-	    emit_passthrough(p, BRW_ATTRIB_COLOR0, VERT_RESULT_COL0);
+	    emit_passthrough(p, VERT_ATTRIB_COLOR0, VERT_RESULT_COL0);
 
 	 if (p->state->fragprog_inputs_read & FRAG_BIT_COL1)
-	    emit_passthrough(p, BRW_ATTRIB_COLOR1, VERT_RESULT_COL1);
+	    emit_passthrough(p, VERT_ATTRIB_COLOR1, VERT_RESULT_COL1);
       }
    }
 
@@ -1628,7 +1625,7 @@ const struct brw_tracked_state brw_tnl_vertprog = {
 
 static void update_active_vertprog( struct brw_context *brw )
 {
-   struct gl_vertex_program *prev = brw->vertex_program;
+   const struct gl_vertex_program *prev = brw->vertex_program;
 
    /* NEW_PROGRAM */
    if (brw->attribs.VertexProgram->_Enabled) {
diff --git a/src/mesa/drivers/dri/i965/brw_vtbl.c b/src/mesa/drivers/dri/i965/brw_vtbl.c
index ac09754e3ae..786f30e641e 100644
--- a/src/mesa/drivers/dri/i965/brw_vtbl.c
+++ b/src/mesa/drivers/dri/i965/brw_vtbl.c
@@ -46,8 +46,6 @@
 #include "brw_state.h"
 
 #include "brw_draw.h"
-#include "brw_exec.h"
-#include "brw_save.h"
 #include "brw_state.h"
 #include "brw_aub.h"
 #include "brw_fallback.h"
@@ -68,9 +66,6 @@ static void brw_destroy_context( struct intel_context *intel )
    brw_destroy_state(brw);
    brw_draw_destroy( brw );
 
-   brw_exec_destroy( ctx );
-   brw_save_destroy( ctx );
-
    brw_ProgramCacheDestroy( ctx );
    brw_FrameBufferTexDestroy( brw );
 }
@@ -166,10 +161,7 @@ static GLuint brw_flush_cmd( void )
 
 static void brw_invalidate_state( struct intel_context *intel, GLuint new_state )
 {
-   GLcontext *ctx = &intel->ctx;
-
-   brw_exec_invalidate_state(ctx, new_state);
-   brw_save_invalidate_state(ctx, new_state);
+   /* nothing */
 }
 
 
diff --git a/src/mesa/drivers/dri/i965/brw_wm_state.c b/src/mesa/drivers/dri/i965/brw_wm_state.c
index 4707a709e77..ff5cb31bdd1 100644
--- a/src/mesa/drivers/dri/i965/brw_wm_state.c
+++ b/src/mesa/drivers/dri/i965/brw_wm_state.c
@@ -122,7 +122,7 @@ static void upload_wm_unit(struct brw_context *brw )
 
    /* BRW_NEW_FRAGMENT_PROGRAM */
    {
-      struct gl_fragment_program *fp = brw->fragment_program; 
+      const struct gl_fragment_program *fp = brw->fragment_program; 
 
       if (fp->Base.InputsRead & (1<<FRAG_ATTRIB_WPOS)) 
 	 wm.wm5.program_uses_depth = 1; /* as far as we can tell */
@@ -168,7 +168,7 @@ static void upload_wm_unit(struct brw_context *brw )
       wm.wm5.line_stipple = 1;
    }
 
-   if (INTEL_DEBUG & DEBUG_STATS)
+   if (INTEL_DEBUG & DEBUG_STATS || intel->stats_wm)
       wm.wm4.stats_enable = 1;
 
    brw->wm.state_gs_offset = brw_cache_data( &brw->cache[BRW_WM_UNIT], &wm );
diff --git a/src/mesa/drivers/dri/i965/bufmgr.h b/src/mesa/drivers/dri/i965/bufmgr.h
index 6932522d3d3..e748c0d6d0b 100644
--- a/src/mesa/drivers/dri/i965/bufmgr.h
+++ b/src/mesa/drivers/dri/i965/bufmgr.h
@@ -199,9 +199,11 @@ void *bmFindVirtual( struct intel_context *intel,
  * For now they can stay, but will likely change/move before final:
  */
 unsigned bmSetFence( struct intel_context * );
+unsigned bmSetFenceLock( struct intel_context * );
 unsigned bmLockAndFence( struct intel_context *intel );
 int bmTestFence( struct intel_context *, unsigned fence );
 void bmFinishFence( struct intel_context *, unsigned fence );
+void bmFinishFenceLock( struct intel_context *, unsigned fence );
 
 void bm_fake_NotifyContendedLockTake( struct intel_context * );
 
diff --git a/src/mesa/drivers/dri/i965/bufmgr_fake.c b/src/mesa/drivers/dri/i965/bufmgr_fake.c
index ed88ab3797a..24ee11edd8c 100644
--- a/src/mesa/drivers/dri/i965/bufmgr_fake.c
+++ b/src/mesa/drivers/dri/i965/bufmgr_fake.c
@@ -338,7 +338,6 @@ static int evict_mru( struct intel_context *intel, GLuint *pool )
 }
 
 
-
 static int check_fenced( struct intel_context *intel )
 {
    struct bufmgr *bm = intel->bm;
@@ -1328,11 +1327,21 @@ unsigned bmSetFence( struct intel_context *intel )
    return intel->bm->last_fence;
 }
 
+unsigned bmSetFenceLock( struct intel_context *intel )
+{
+  unsigned last;
+  LOCK(intel->bm);
+  last = bmSetFence(intel);
+  UNLOCK(intel->bm);
+  return last;
+}
 unsigned bmLockAndFence( struct intel_context *intel )
 {
    if (intel->bm->need_fence) {
       LOCK_HARDWARE(intel);
+      LOCK(intel->bm);
       bmSetFence(intel);
+      UNLOCK(intel->bm);
       UNLOCK_HARDWARE(intel);
    }
 
@@ -1350,7 +1359,12 @@ void bmFinishFence( struct intel_context *intel, unsigned fence )
    check_fenced(intel);
 }
 
-
+void bmFinishFenceLock( struct intel_context *intel, unsigned fence )
+{
+   LOCK(intel->bm);
+   bmFinishFence(intel, fence);
+   UNLOCK(intel->bm);
+}
 
 
 /* Specifically ignore texture memory sharing.
diff --git a/src/mesa/drivers/dri/i965/intel_blit.c b/src/mesa/drivers/dri/i965/intel_blit.c
index c8c5bf93c95..173d1d5b6c2 100644
--- a/src/mesa/drivers/dri/i965/intel_blit.c
+++ b/src/mesa/drivers/dri/i965/intel_blit.c
@@ -66,7 +66,7 @@ void intelCopyBuffer( const __DRIdrawablePrivate *dPriv,
    intelFlush( &intel->ctx );
 
 
-   bmFinishFence(intel, intel->last_swap_fence);
+   bmFinishFenceLock(intel, intel->last_swap_fence);
 
    /* The LOCK_HARDWARE is required for the cliprects.  Buffer offsets
     * should work regardless.
@@ -155,7 +155,7 @@ void intelCopyBuffer( const __DRIdrawablePrivate *dPriv,
 
    intel_batchbuffer_flush( intel->batch );
    intel->second_last_swap_fence = intel->last_swap_fence;
-   intel->last_swap_fence = bmSetFence( intel );
+   intel->last_swap_fence = bmSetFenceLock( intel );
    UNLOCK_HARDWARE( intel );
 
    if (!rect)
diff --git a/src/mesa/drivers/dri/i965/intel_context.c b/src/mesa/drivers/dri/i965/intel_context.c
index d13e287fa76..34560e4353e 100644
--- a/src/mesa/drivers/dri/i965/intel_context.c
+++ b/src/mesa/drivers/dri/i965/intel_context.c
@@ -33,11 +33,12 @@
 #include "extensions.h"
 #include "framebuffer.h"
 #include "imports.h"
+#include "points.h"
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 #include "tnl/t_vertex.h"
@@ -60,6 +61,7 @@
 #include "bufmgr.h"
 
 #include "utils.h"
+#include "vblank.h"
 #ifndef INTEL_DEBUG
 int INTEL_DEBUG = (0);
 #endif
@@ -70,6 +72,7 @@ int INTEL_DEBUG = (0);
 #define need_GL_ARB_vertex_buffer_object
 #define need_GL_ARB_vertex_program
 #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
@@ -84,11 +87,6 @@ int INTEL_DEBUG = (0);
 int VERBOSE = 0;
 #endif
 
-#if DEBUG_LOCKING
-char *prevLockFile;
-int prevLockLine;
-#endif
-
 /***************************************
  * Mesa's Driver Functions
  ***************************************/
@@ -182,7 +180,16 @@ const struct dri_extension card_extensions[] =
     { NULL,                                NULL }
 };
 
+const struct dri_extension arb_oc_extension = 
+    { "GL_ARB_occlusion_query",            GL_ARB_occlusion_query_functions};
 
+void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging)
+{	     
+	struct intel_context *intel = ctx?intel_context(ctx):NULL;
+	driInitExtensions(ctx, card_extensions, enable_imaging);
+	if (!ctx || intel->intelScreen->drmMinor >= 8)
+		driInitSingleExtension (ctx, &arb_oc_extension);
+}
 
 static const struct dri_debug_control debug_control[] =
 {
@@ -216,7 +223,7 @@ static void intelInvalidateState( GLcontext *ctx, GLuint new_state )
 
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _tnl_invalidate_vertex_state( ctx, new_state );
    
@@ -241,6 +248,37 @@ void intelFinish( GLcontext *ctx )
    bmFinishFence(intel, bmLockAndFence(intel));
 }
 
+static void
+intelBeginQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
+{
+	struct intel_context *intel = intel_context( ctx );
+	drmI830MMIO io = {
+		.read_write = MMIO_READ,
+		.reg = MMIO_REGS_PS_DEPTH_COUNT,
+		.data = &q->Result 
+	};
+	intel->stats_wm++;
+	intelFinish(&intel->ctx);
+	drmCommandRead(intel->driFd, DRM_I830_MMIO, &io, sizeof(io));
+}
+
+static void
+intelEndQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
+{
+	struct intel_context *intel = intel_context( ctx );
+	GLuint64EXT tmp;	
+	drmI830MMIO io = {
+		.read_write = MMIO_READ,
+		.reg = MMIO_REGS_PS_DEPTH_COUNT,
+		.data = &tmp
+	};
+	intelFinish(&intel->ctx);
+	drmCommandRead(intel->driFd, DRM_I830_MMIO, &io, sizeof(io));
+	q->Result = tmp - q->Result;
+	q->Ready = GL_TRUE;
+	intel->stats_wm--;
+}
+
 
 void intelInitDriverFunctions( struct dd_function_table *functions )
 {
@@ -250,6 +288,8 @@ void intelInitDriverFunctions( struct dd_function_table *functions )
    functions->Finish = intelFinish;
    functions->GetString = intelGetString;
    functions->UpdateState = intelInvalidateState;
+   functions->BeginQuery = intelBeginQuery;
+   functions->EndQuery = intelEndQuery;
 
    /* CopyPixels can be accelerated even with the current memory
     * manager:
@@ -292,6 +332,11 @@ GLboolean intelInitContext( struct intel_context *intel,
    intel->driScreen = sPriv;
    intel->sarea = saPriv;
 
+   driParseConfigFiles (&intel->optionCache, &intelScreen->optionCache,
+		   intel->driScreen->myNum, "i965");
+
+   intel->vblank_flags = (intel->intelScreen->irq_active != 0)
+	   ? driGetDefaultVBlankFlags(&intel->optionCache) : VBLANK_FLAG_NO_IRQ;
 
    ctx->Const.MaxTextureMaxAnisotropy = 2.0;
 
@@ -327,7 +372,7 @@ GLboolean intelInitContext( struct intel_context *intel,
 
    /* Initialize the software rasterizer and helper modules. */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -374,9 +419,7 @@ GLboolean intelInitContext( struct intel_context *intel,
       _mesa_printf("IRQs not active.  Exiting\n");
       exit(1);
    }
- 
-   driInitExtensions( ctx, card_extensions, 
-		      GL_TRUE );
+   intelInitExtensions(ctx, GL_TRUE); 
 
    INTEL_DEBUG  = driParseDebugString( getenv( "INTEL_DEBUG" ),
 				       debug_control );
@@ -403,8 +446,8 @@ GLboolean intelInitContext( struct intel_context *intel,
 				 intelScreen->cpp,
 				 intelScreen->front.pitch / intelScreen->cpp,
 				 intelScreen->height,
-				 GL_FALSE);
-
+				 intelScreen->front.size,
+				 intelScreen->front.tiled != 0);
 
    intel->back_region = 
       intel_region_create_static(intel,
@@ -414,7 +457,8 @@ GLboolean intelInitContext( struct intel_context *intel,
 				 intelScreen->cpp,
 				 intelScreen->back.pitch / intelScreen->cpp,
 				 intelScreen->height,
-				 (INTEL_DEBUG & DEBUG_TILE) ? 0 : 1);
+				 intelScreen->back.size,
+                                 intelScreen->back.tiled != 0);
 
    /* Still assuming front.cpp == depth.cpp
     *
@@ -430,7 +474,8 @@ GLboolean intelInitContext( struct intel_context *intel,
 				 intelScreen->cpp,
 				 intelScreen->depth.pitch / intelScreen->cpp,
 				 intelScreen->height,
-				 (INTEL_DEBUG & DEBUG_TILE) ? 0 : 1);
+				 intelScreen->depth.size,
+                                 intelScreen->depth.tiled != 0);
    
    intel_bufferobj_init( intel );
    intel->batch = intel_batchbuffer_alloc( intel );
@@ -472,7 +517,7 @@ void intelDestroyContext(__DRIcontextPrivate *driContextPriv)
       release_texture_heaps = (intel->ctx.Shared->RefCount == 1);
       _swsetup_DestroyContext (&intel->ctx);
       _tnl_DestroyContext (&intel->ctx);
-      _ac_DestroyContext (&intel->ctx);
+      _vbo_DestroyContext (&intel->ctx);
 
       _swrast_DestroyContext (&intel->ctx);
       intel->Fallback = 0;	/* don't call _swrast_Flush later */
@@ -521,6 +566,9 @@ GLboolean intelMakeCurrent(__DRIcontextPrivate *driContextPriv,
 
       if ( intel->driDrawable != driDrawPriv ) {
 	 /* Shouldn't the readbuffer be stored also? */
+	 driDrawableInitVBlank( driDrawPriv, intel->vblank_flags,
+		      &intel->vbl_seq );
+
 	 intel->driDrawable = driDrawPriv;
 	 intelWindowMoved( intel );
       }
@@ -655,3 +703,4 @@ void UNLOCK_HARDWARE( struct intel_context *intel )
    _glthread_UNLOCK_MUTEX(lockMutex); 
 }
 
+
diff --git a/src/mesa/drivers/dri/i965/intel_context.h b/src/mesa/drivers/dri/i965/intel_context.h
index 8367a957109..a0f392f9353 100644
--- a/src/mesa/drivers/dri/i965/intel_context.h
+++ b/src/mesa/drivers/dri/i965/intel_context.h
@@ -177,6 +177,7 @@ struct intel_context
    GLuint second_last_swap_fence;
    
    GLboolean aub_wrap;
+   GLuint stats_wm;
 
    struct intel_batchbuffer *batch;
 
@@ -499,6 +500,7 @@ void intelBitmap(GLcontext * ctx,
 		 const struct gl_pixelstore_attrib *unpack,
 		 const GLubyte * pixels);
 
+void intelInitExtensions(GLcontext *ctx, GLboolean enable_imaging);
 #define _NEW_WINDOW_POS 0x40000000
 
 
@@ -521,6 +523,5 @@ static inline struct intel_texture_image *intel_texture_image( struct gl_texture
    return (struct intel_texture_image *)img;
 }
 
-
 #endif
 
diff --git a/src/mesa/drivers/dri/i965/intel_ioctl.c b/src/mesa/drivers/dri/i965/intel_ioctl.c
index f3b76db2153..0a8e976f706 100644
--- a/src/mesa/drivers/dri/i965/intel_ioctl.c
+++ b/src/mesa/drivers/dri/i965/intel_ioctl.c
@@ -43,6 +43,26 @@
 #include "drm.h"
 #include "bufmgr.h"
 
+static int intelWaitIdleLocked( struct intel_context *intel )
+{
+   static int in_wait_idle = 0;
+   unsigned int fence;
+
+   if (!in_wait_idle) {
+      if (INTEL_DEBUG & DEBUG_SYNC) {
+	 fprintf(stderr, "waiting for idle\n");
+      }
+
+      in_wait_idle = 1;
+      fence = bmSetFence(intel);
+      intelWaitIrq(intel, fence);
+      in_wait_idle = 0;
+
+      return bmTestFence(intel, fence);
+   } else {
+      return 1;
+   }
+}
 
 int intelEmitIrqLocked( struct intel_context *intel )
 {
@@ -88,7 +108,8 @@ void intelWaitIrq( struct intel_context *intel, int seq )
 
 	 /* This seems quite often to return before it should!?! 
 	  */
-      } while (ret == -EAGAIN || ret == -EINTR || (ret == -EBUSY && lastdispatch != intel->sarea->last_dispatch) || (ret == 0 && seq > intel->sarea->last_dispatch));
+      } while (ret == -EAGAIN || ret == -EINTR || (ret == -EBUSY && lastdispatch != intel->sarea->last_dispatch) || (ret == 0 && seq > intel->sarea->last_dispatch)
+	       || (ret == 0 && intel->sarea->last_dispatch - seq >= (1 << 24)));
       
 
       if ( ret ) {
@@ -140,7 +161,11 @@ void intel_batch_ioctl( struct intel_context *intel,
 	 UNLOCK_HARDWARE(intel);
 	 exit(1);
       }
-   }      
+
+      if (INTEL_DEBUG & DEBUG_SYNC) {
+	intelWaitIdleLocked(intel);
+      }
+   }
 }
 
 void intel_cmd_ioctl( struct intel_context *intel, 
@@ -172,5 +197,9 @@ void intel_cmd_ioctl( struct intel_context *intel,
 	 UNLOCK_HARDWARE(intel);
 	 exit(1);
       }
-   }      
+
+      if (INTEL_DEBUG & DEBUG_SYNC) {
+	intelWaitIdleLocked(intel);
+      }
+   }
 }
diff --git a/src/mesa/drivers/dri/i965/intel_regions.c b/src/mesa/drivers/dri/i965/intel_regions.c
index 398b0a0a3b5..835ecdd7257 100644
--- a/src/mesa/drivers/dri/i965/intel_regions.c
+++ b/src/mesa/drivers/dri/i965/intel_regions.c
@@ -122,10 +122,10 @@ struct intel_region *intel_region_create_static( struct intel_context *intel,
 						 GLuint cpp,
 						 GLuint pitch, 
 						 GLuint height,
+						 GLuint size,
 						 GLboolean tiled )
 {
    struct intel_region *region = calloc(sizeof(*region), 1);
-   GLuint size = cpp * pitch * height;
    GLint pool;
 
    DBG("%s\n", __FUNCTION__);
diff --git a/src/mesa/drivers/dri/i965/intel_regions.h b/src/mesa/drivers/dri/i965/intel_regions.h
index 2413f0de33c..d2235f1275b 100644
--- a/src/mesa/drivers/dri/i965/intel_regions.h
+++ b/src/mesa/drivers/dri/i965/intel_regions.h
@@ -78,6 +78,7 @@ struct intel_region *intel_region_create_static( struct intel_context *intel,
 						 GLuint cpp,
 						 GLuint pitch,
 						 GLuint height,
+						 GLuint size,
 						 GLboolean tiled );
 
 /* Map/unmap regions.  This is refcounted also: 
diff --git a/src/mesa/drivers/dri/i965/intel_screen.c b/src/mesa/drivers/dri/i965/intel_screen.c
index 56e6a792fa2..08f0bb340f7 100644
--- a/src/mesa/drivers/dri/i965/intel_screen.c
+++ b/src/mesa/drivers/dri/i965/intel_screen.c
@@ -38,6 +38,7 @@
 
 #include "intel_screen.h"
 
+#include "intel_context.h"
 #include "intel_tex.h"
 #include "intel_span.h"
 #include "intel_ioctl.h"
@@ -61,8 +62,6 @@ const GLuint __driNConfigOptions = 4;
 static PFNGLXCREATECONTEXTMODES create_context_modes = NULL;
 #endif /*USE_NEW_INTERFACE*/
 
-extern const struct dri_extension card_extensions[];
-
 /**
  * Map all the memory regions described by the screen.
  * \return GL_TRUE if success, GL_FALSE if error.
@@ -230,16 +229,19 @@ intelUpdateScreenFromSAREA(intelScreenPrivate *intelScreen,
    intelScreen->front.pitch = sarea->pitch * intelScreen->cpp;
    intelScreen->front.handle = sarea->front_handle;
    intelScreen->front.size = sarea->front_size;
+   intelScreen->front.tiled = sarea->front_tiled;
 
    intelScreen->back.offset = sarea->back_offset;
    intelScreen->back.pitch = sarea->pitch * intelScreen->cpp;
    intelScreen->back.handle = sarea->back_handle;
    intelScreen->back.size = sarea->back_size;
-			 
+   intelScreen->back.tiled = sarea->back_tiled;
+
    intelScreen->depth.offset = sarea->depth_offset;
    intelScreen->depth.pitch = sarea->pitch * intelScreen->cpp;
    intelScreen->depth.handle = sarea->depth_handle;
    intelScreen->depth.size = sarea->depth_size;
+   intelScreen->depth.tiled = sarea->depth_tiled;
 
    intelScreen->tex.offset = sarea->tex_offset;
    intelScreen->logTextureGranularity = sarea->log_tex_granularity;
@@ -249,6 +251,7 @@ intelUpdateScreenFromSAREA(intelScreenPrivate *intelScreen,
    intelScreen->rotated.offset = sarea->rotated_offset;
    intelScreen->rotated.pitch = sarea->rotated_pitch * intelScreen->cpp;
    intelScreen->rotated.size = sarea->rotated_size;
+   intelScreen->rotated.tiled = sarea->rotated_tiled;
    intelScreen->current_rotation = sarea->rotation;
 #if 0
    matrix23Rotate(&intelScreen->rotMatrix,
@@ -683,7 +686,6 @@ void * __driCreateNewScreen_20050727( __DRInativeDisplay *dpy, int scrn, __DRIsc
 					(dri_priv->cpp == 2) ? 16 : 24,
 					(dri_priv->cpp == 2) ? 0  : 8,
 					GL_TRUE );
-
       /* Calling driInitExtensions here, with a NULL context pointer, does not actually
        * enable the extensions.  It just makes sure that all the dispatch offsets for all
        * the extensions that *might* be enables are known.  This is needed because the
@@ -692,7 +694,7 @@ void * __driCreateNewScreen_20050727( __DRInativeDisplay *dpy, int scrn, __DRIsc
        *
        * Hello chicken.  Hello egg.  How are you two today?
        */
-      driInitExtensions( NULL, card_extensions, GL_FALSE );
+      intelInitExtensions(NULL, GL_FALSE);
    }
 
    return (void *) psp;
diff --git a/src/mesa/drivers/dri/i965/intel_screen.h b/src/mesa/drivers/dri/i965/intel_screen.h
index 094158afd85..bf9a716082d 100644
--- a/src/mesa/drivers/dri/i965/intel_screen.h
+++ b/src/mesa/drivers/dri/i965/intel_screen.h
@@ -42,6 +42,7 @@ typedef struct {
    char *map;           /* memory map */
    int offset;          /* from start of video mem, in bytes */
    int pitch;           /* row stride, in pixels */
+   unsigned int tiled; 
 } intelRegion;
 
 typedef struct 
diff --git a/src/mesa/drivers/dri/i965/intel_tex_layout.c b/src/mesa/drivers/dri/i965/intel_tex_layout.c
new file mode 120000
index 00000000000..fe61b441945
--- /dev/null
+++ b/src/mesa/drivers/dri/i965/intel_tex_layout.c
@@ -0,0 +1 @@
+../intel/intel_tex_layout.c
\ No newline at end of file
diff --git a/src/mesa/drivers/dri/i965/server/i830_common.h b/src/mesa/drivers/dri/i965/server/i830_common.h
index e3bbdc79070..f320378c2a5 100644
--- a/src/mesa/drivers/dri/i965/server/i830_common.h
+++ b/src/mesa/drivers/dri/i965/server/i830_common.h
@@ -52,6 +52,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define DRM_I830_INIT_HEAP                0x0a
 #define DRM_I830_CMDBUFFER                0x0b
 #define DRM_I830_DESTROY_HEAP             0x0c
+#define DRM_I830_MMIO		       	  0x10
 
 typedef struct {
    enum {
@@ -199,5 +200,23 @@ typedef struct {
 	int region;
 } drmI830MemDestroyHeap;
 
+#define MMIO_READ  0
+#define MMIO_WRITE 1
+
+#define MMIO_REGS_IA_PRIMATIVES_COUNT           0
+#define MMIO_REGS_IA_VERTICES_COUNT             1
+#define MMIO_REGS_VS_INVOCATION_COUNT           2
+#define MMIO_REGS_GS_PRIMITIVES_COUNT           3
+#define MMIO_REGS_GS_INVOCATION_COUNT           4
+#define MMIO_REGS_CL_PRIMITIVES_COUNT           5
+#define MMIO_REGS_CL_INVOCATION_COUNT           6
+#define MMIO_REGS_PS_INVOCATION_COUNT           7
+#define MMIO_REGS_PS_DEPTH_COUNT                8
+
+typedef struct {
+        unsigned int read_write:1;
+        unsigned int reg:31;
+        void __user *data;
+} drmI830MMIO;
 
 #endif /* _I830_DRM_H_ */
diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.c b/src/mesa/drivers/dri/intel/intel_tex_layout.c
new file mode 100644
index 00000000000..f3564802171
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_tex_layout.c
@@ -0,0 +1,102 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * 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"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, 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 NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Michel Dänzer <michel@tungstengraphics.com>
+  */
+
+#include "intel_mipmap_tree.h"
+#include "intel_tex_layout.h"
+#include "macros.h"
+
+
+static int align(int value, int alignment)
+{
+   return (value + alignment - 1) & ~(alignment - 1);
+}
+
+void i945_miptree_layout_2d( struct intel_mipmap_tree *mt )
+{
+   GLint align_h = 2, align_w = 4;
+   GLuint level;
+   GLuint x = 0;
+   GLuint y = 0;
+   GLuint width = mt->width0;
+   GLuint height = mt->height0;
+
+   mt->pitch = mt->width0;
+
+   /* May need to adjust pitch to accomodate the placement of
+    * the 2nd mipmap.  This occurs when the alignment
+    * constraints of mipmap placement push the right edge of the
+    * 2nd mipmap out past the width of its parent.
+    */
+   if (mt->first_level != mt->last_level) {
+      GLuint mip1_width = align(minify(mt->width0), align_w)
+			+ minify(minify(mt->width0));
+
+      if (mip1_width > mt->width0)
+	 mt->pitch = mip1_width;
+   }
+
+   /* Pitch must be a whole number of dwords, even though we
+    * express it in texels.
+    */
+   mt->pitch = align(mt->pitch * mt->cpp, 4) / mt->cpp;
+   mt->total_height = 0;
+
+   for ( level = mt->first_level ; level <= mt->last_level ; level++ ) {
+      GLuint img_height;
+
+      intel_miptree_set_level_info(mt, level, 1, x, y, width, 
+				   mt->compressed ? height/4 : height, 1);
+
+      if (mt->compressed)
+	 img_height = MAX2(1, height/4);
+      else
+	 img_height = align(height, align_h);
+
+
+      /* Because the images are packed better, the final offset
+       * might not be the maximal one:
+       */
+      mt->total_height = MAX2(mt->total_height, y + img_height);
+
+      /* Layout_below: step right after second mipmap.
+       */
+      if (level == mt->first_level + 1) {
+	 x += align(width, align_w);
+      }
+      else {
+	 y += img_height;
+      }
+
+      width  = minify(width);
+      height = minify(height);
+   }
+}
diff --git a/src/mesa/drivers/dri/intel/intel_tex_layout.h b/src/mesa/drivers/dri/intel/intel_tex_layout.h
new file mode 100644
index 00000000000..1e37f8f525f
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_tex_layout.h
@@ -0,0 +1,41 @@
+/**************************************************************************
+ * 
+ * Copyright 2006 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * 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"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, 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 NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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:
+  *   Keith Whitwell <keith@tungstengraphics.com>
+  *   Michel Dänzer <michel@tungstengraphics.com>
+  */
+
+#include "macros.h"
+
+
+static GLuint minify( GLuint d )
+{
+   return MAX2(1, d>>1);
+}
+
+extern void i945_miptree_layout_2d( struct intel_mipmap_tree *mt );
diff --git a/src/mesa/drivers/dri/mach64/mach64_context.c b/src/mesa/drivers/dri/mach64/mach64_context.c
index 2ab1cf2617e..5a6c301da2d 100644
--- a/src/mesa/drivers/dri/mach64/mach64_context.c
+++ b/src/mesa/drivers/dri/mach64/mach64_context.c
@@ -38,7 +38,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -225,7 +225,7 @@ GLboolean mach64CreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -278,7 +278,7 @@ void mach64DestroyContext( __DRIcontextPrivate *driContextPriv  )
 
       _swsetup_DestroyContext( mmesa->glCtx );
       _tnl_DestroyContext( mmesa->glCtx );
-      _ac_DestroyContext( mmesa->glCtx );
+      _vbo_DestroyContext( mmesa->glCtx );
       _swrast_DestroyContext( mmesa->glCtx );
 
       if (release_texture_heaps) {
diff --git a/src/mesa/drivers/dri/mach64/mach64_state.c b/src/mesa/drivers/dri/mach64/mach64_state.c
index d4804a2c550..667a3945206 100644
--- a/src/mesa/drivers/dri/mach64/mach64_state.c
+++ b/src/mesa/drivers/dri/mach64/mach64_state.c
@@ -40,7 +40,7 @@
 #include "enums.h"
 #include "colormac.h"
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1023,7 +1023,7 @@ static void mach64DDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    MACH64_CONTEXT(ctx)->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/mga/mga_xmesa.c b/src/mesa/drivers/dri/mga/mga_xmesa.c
index f02913c6af2..f024f73eb6a 100644
--- a/src/mesa/drivers/dri/mga/mga_xmesa.c
+++ b/src/mesa/drivers/dri/mga/mga_xmesa.c
@@ -45,7 +45,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 
@@ -593,7 +593,7 @@ mgaCreateContext( const __GLcontextModes *mesaVis,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    
    _swsetup_CreateContext( ctx );
@@ -676,7 +676,7 @@ mgaDestroyContext(__DRIcontextPrivate *driContextPriv)
       release_texture_heaps = (mmesa->glCtx->Shared->RefCount == 1);
       _swsetup_DestroyContext( mmesa->glCtx );
       _tnl_DestroyContext( mmesa->glCtx );
-      _ac_DestroyContext( mmesa->glCtx );
+      _vbo_DestroyContext( mmesa->glCtx );
       _swrast_DestroyContext( mmesa->glCtx );
 
       mgaFreeVB( mmesa->glCtx );
diff --git a/src/mesa/drivers/dri/mga/mgastate.c b/src/mesa/drivers/dri/mga/mgastate.c
index 075fd348cf4..c20a76f29ef 100644
--- a/src/mesa/drivers/dri/mga/mgastate.c
+++ b/src/mesa/drivers/dri/mga/mgastate.c
@@ -41,7 +41,7 @@
 #include "mgaregs.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
@@ -1042,7 +1042,7 @@ static void mgaDDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    MGA_CONTEXT(ctx)->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/nouveau/Makefile b/src/mesa/drivers/dri/nouveau/Makefile
new file mode 100644
index 00000000000..20d2de5eefb
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/Makefile
@@ -0,0 +1,53 @@
+# src/mesa/drivers/dri/nouveau/Makefile
+
+TOP = ../../../../..
+include $(TOP)/configs/current
+
+LIBNAME = nouveau_dri.so
+
+MINIGLX_SOURCES = 
+
+DRIVER_SOURCES = \
+	nouveau_bufferobj.c      \
+	nouveau_buffers.c        \
+	nouveau_card.c           \
+	nouveau_context.c        \
+	nouveau_driver.c         \
+	nouveau_fifo.c           \
+	nouveau_lock.c           \
+	nouveau_object.c         \
+	nouveau_screen.c         \
+	nouveau_span.c           \
+	nouveau_state.c          \
+	nouveau_state_cache.c    \
+	nouveau_shader.c         \
+	nouveau_shader_0.c       \
+	nouveau_shader_1.c       \
+	nouveau_shader_2.c       \
+	nouveau_tex.c            \
+	nouveau_swtcl.c          \
+	nouveau_sync.c           \
+	nouveau_query.c          \
+	nv04_state.c             \
+	nv04_swtcl.c             \
+	nv10_state.c             \
+	nv10_swtcl.c             \
+	nv20_state.c             \
+	nv20_vertprog.c          \
+	nv30_state.c             \
+	nv30_fragprog.c          \
+	nv30_vertprog.c          \
+	nv40_fragprog.c          \
+	nv40_vertprog.c          \
+	nv50_state.c
+
+C_SOURCES = \
+	$(COMMON_SOURCES) \
+	$(DRIVER_SOURCES) 
+
+ASM_SOURCES = 
+
+
+include ../Makefile.template
+
+symlinks:
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c
new file mode 100644
index 00000000000..684ed7b017d
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.c
@@ -0,0 +1,616 @@
+#include "bufferobj.h"
+#include "enums.h"
+
+#include "nouveau_bufferobj.h"
+#include "nouveau_buffers.h"
+#include "nouveau_context.h"
+#include "nouveau_drm.h"
+#include "nouveau_object.h"
+#include "nouveau_msg.h"
+
+#define NOUVEAU_MEM_FREE(mem) do {      \
+	nouveau_mem_free(ctx, (mem));   \
+	(mem) = NULL;                   \
+} while(0)
+
+#define DEBUG(fmt,args...) do {                \
+	if (NOUVEAU_DEBUG & DEBUG_BUFFEROBJ) { \
+		fprintf(stderr, "%s: "fmt, __func__, ##args);  \
+	}                                      \
+} while(0)
+
+static GLboolean
+nouveau_bo_download_from_screen(GLcontext *ctx,	GLuint offset, GLuint size,
+						struct gl_buffer_object *bo)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+	nouveau_mem *in_mem;
+
+	DEBUG("bo=%p, offset=%d, size=%d\n", bo, offset, size);
+
+	/* If there's a permanent backing store, blit directly into it */
+	if (nbo->cpu_mem) {
+		if (nbo->cpu_mem != nbo->gpu_mem) {
+			DEBUG("..cpu_mem\n");
+			nouveau_memformat_flat_emit(ctx, nbo->cpu_mem,
+						    nbo->gpu_mem,
+						    offset, offset, size);
+		}
+	} else {
+		DEBUG("..sys_mem\n");
+		in_mem = nouveau_mem_alloc(ctx, NOUVEAU_MEM_AGP, size, 0);
+		if (in_mem) {
+			DEBUG("....via AGP\n");
+			/* otherwise, try blitting to faster memory and
+			 * copying from there
+			 */
+			nouveau_memformat_flat_emit(ctx, in_mem, nbo->gpu_mem,
+							 0, offset, size);
+			nouveau_notifier_wait_nop(ctx, nmesa->syncNotifier,
+						       NvSubMemFormat);
+			_mesa_memcpy(nbo->cpu_mem_sys + offset,
+					in_mem->map, size);
+			NOUVEAU_MEM_FREE(in_mem);
+		} else {
+			DEBUG("....direct VRAM copy\n");
+			/* worst case, copy directly from vram */
+			_mesa_memcpy(nbo->cpu_mem_sys + offset,
+				     nbo->gpu_mem + offset,
+				     size);
+		}
+	}
+
+	return GL_TRUE;
+}
+
+static GLboolean
+nouveau_bo_upload_to_screen(GLcontext *ctx, GLuint offset, GLuint size,
+					    struct gl_buffer_object *bo)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+	nouveau_mem *out_mem;
+
+	DEBUG("bo=%p, offset=%d, size=%d\n", bo, offset, size);
+
+	if (nbo->cpu_mem) {
+		if (nbo->cpu_mem != nbo->gpu_mem) {
+			DEBUG("..cpu_mem\n");
+			nouveau_memformat_flat_emit(ctx, nbo->gpu_mem,
+						    nbo->cpu_mem,
+						    offset, offset, size);
+		}
+	} else {
+		out_mem = nouveau_mem_alloc(ctx, NOUVEAU_MEM_AGP |
+						 NOUVEAU_MEM_MAPPED,
+						 size, 0);
+		if (out_mem) {
+			DEBUG("....via AGP\n");
+			_mesa_memcpy(out_mem->map,
+					nbo->cpu_mem_sys + offset, size);
+			nouveau_memformat_flat_emit(ctx, nbo->gpu_mem, out_mem,
+						    offset, 0, size);
+			nouveau_notifier_wait_nop(ctx, nmesa->syncNotifier,
+						       NvSubMemFormat);
+			NOUVEAU_MEM_FREE(out_mem);
+		} else {
+			DEBUG("....direct VRAM copy\n");
+			_mesa_memcpy(nbo->gpu_mem->map + offset,
+				     nbo->cpu_mem_sys + offset,
+				     size);
+		}
+	}
+
+	return GL_TRUE;
+}
+
+GLboolean
+nouveau_bo_move_in(GLcontext *ctx, struct gl_buffer_object *bo)
+{
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+
+	DEBUG("bo=%p\n", bo);
+
+	if (bo->OnCard)
+		return GL_TRUE;
+	assert(nbo->gpu_mem_flags);
+
+	nbo->gpu_mem = nouveau_mem_alloc(ctx, nbo->gpu_mem_flags |
+					      NOUVEAU_MEM_MAPPED,
+					      bo->Size, 0);
+	assert(nbo->gpu_mem);
+
+	if (nbo->cpu_mem_flags) {
+		if ((nbo->cpu_mem_flags|NOUVEAU_MEM_MAPPED) != nbo->gpu_mem->type) {
+			DEBUG("..need cpu_mem buffer\n");
+
+			nbo->cpu_mem = nouveau_mem_alloc(ctx,
+							 nbo->cpu_mem_flags |
+							 NOUVEAU_MEM_MAPPED,
+							 bo->Size, 0);
+
+			if (nbo->cpu_mem) {
+				DEBUG("....alloc ok, kill sys_mem buffer\n");
+				_mesa_memcpy(nbo->cpu_mem->map,
+					     nbo->cpu_mem_sys, bo->Size);
+				FREE(nbo->cpu_mem_sys);
+			}
+		} else {
+			DEBUG("..cpu direct access to GPU buffer\n");
+			nbo->cpu_mem = nbo->gpu_mem;
+		}
+	}
+	nouveau_bo_upload_to_screen(ctx, 0, bo->Size, bo);
+
+	bo->OnCard = GL_TRUE;
+	return GL_TRUE;
+}
+
+GLboolean
+nouveau_bo_move_out(GLcontext *ctx, struct gl_buffer_object *bo)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+	GLuint nr_dirty;
+
+	DEBUG("bo=%p\n", bo);
+	if (!bo->OnCard)
+		return GL_TRUE;
+
+	nr_dirty = nouveau_bo_download_dirty(ctx, bo);
+	if (nbo->cpu_mem) {
+		if (nr_dirty && nbo->cpu_mem != nbo->gpu_mem)
+			nouveau_notifier_wait_nop(ctx, nmesa->syncNotifier,
+						       NvSubMemFormat);
+		DEBUG("..destroy cpu_mem buffer\n");
+		nbo->cpu_mem_sys = malloc(bo->Size);
+		assert(nbo->cpu_mem_sys);
+		_mesa_memcpy(nbo->cpu_mem_sys, nbo->cpu_mem->map, bo->Size);
+		if (nbo->cpu_mem == nbo->gpu_mem)
+			nbo->cpu_mem = NULL;
+		else
+			NOUVEAU_MEM_FREE(nbo->cpu_mem);
+	}
+	NOUVEAU_MEM_FREE(nbo->gpu_mem);
+
+	bo->OnCard = GL_FALSE;
+	return GL_TRUE;
+}
+
+static void
+nouveau_bo_choose_storage_method(GLcontext *ctx, GLenum usage,
+						 struct gl_buffer_object *bo)
+{
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+	GLuint gpu_type = 0;
+	GLuint cpu_type = 0;
+
+	switch (usage) {
+	/* Client source, changes often, used by GL many times */
+	case GL_DYNAMIC_DRAW_ARB:
+		gpu_type = NOUVEAU_MEM_AGP | NOUVEAU_MEM_FB_ACCEPTABLE;
+		cpu_type = NOUVEAU_MEM_AGP;
+		break;
+	/* GL source, changes often, client reads many times */
+	case GL_DYNAMIC_READ_ARB:
+	/* Client source, specified once, used by GL many times */
+	case GL_STATIC_DRAW_ARB:
+	/* GL source, specified once, client reads many times */
+	case GL_STATIC_READ_ARB:
+	/* Client source, specified once, used by GL a few times */
+	case GL_STREAM_DRAW_ARB:
+	/* GL source, specified once, client reads a few times */
+	case GL_STREAM_READ_ARB:
+	/* GL source, changes often, used by GL many times*/
+	case GL_DYNAMIC_COPY_ARB:
+	/* GL source, specified once, used by GL many times */
+	case GL_STATIC_COPY_ARB:
+	/* GL source, specified once, used by GL a few times */
+	case GL_STREAM_COPY_ARB:
+		gpu_type = NOUVEAU_MEM_FB;
+		break;
+	default: 
+		assert(0);
+	}
+
+	nbo->gpu_mem_flags = gpu_type;
+	nbo->cpu_mem_flags = cpu_type;
+	nbo->usage	   = usage;
+}
+
+void
+nouveau_bo_init_storage(GLcontext *ctx,	GLuint valid_gpu_access,
+					GLsizeiptrARB size,
+					const GLvoid *data,
+					GLenum usage,
+					struct gl_buffer_object *bo)
+{
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+
+	DEBUG("bo=%p\n", bo);
+
+	/* Free up previous buffers if we can't reuse them */
+	if (nbo->usage != usage ||
+			(nbo->gpu_mem && (nbo->gpu_mem->size != size))) {
+		if (nbo->cpu_mem_sys)
+			FREE(nbo->cpu_mem_sys);
+		if (nbo->cpu_mem) {
+			if (nbo->cpu_mem != nbo->gpu_mem)
+				NOUVEAU_MEM_FREE(nbo->cpu_mem);
+			else
+				nbo->cpu_mem = NULL;
+		}
+		if (nbo->gpu_mem)
+			NOUVEAU_MEM_FREE(nbo->gpu_mem);
+
+		bo->OnCard = GL_FALSE;
+		nbo->cpu_mem_sys = calloc(1, size);
+	}
+
+	nouveau_bo_choose_storage_method(ctx, usage, bo);
+	/* Force off flags that may not be ok for a given buffer */
+	nbo->gpu_mem_flags &= valid_gpu_access;
+
+	bo->Usage  = usage;
+	bo->Size   = size;
+
+	if (data) {
+		GLvoid *map = nouveau_bo_map(ctx, GL_WRITE_ONLY_ARB, bo);
+		_mesa_memcpy(map, data, size);
+		nouveau_bo_dirty_all(ctx, GL_FALSE, bo);
+		nouveau_bo_unmap(ctx, bo);
+	}
+}
+
+void *
+nouveau_bo_map(GLcontext *ctx, GLenum access, struct gl_buffer_object *bo)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+
+	DEBUG("bo=%p, access=%s\n", bo, _mesa_lookup_enum_by_nr(access));
+
+	if (bo->OnCard && 
+		(access == GL_READ_ONLY_ARB || access == GL_READ_WRITE_ARB)) {
+		GLuint nr_dirty;
+
+		DEBUG("..on card\n");
+		nr_dirty = nouveau_bo_download_dirty(ctx, bo);
+
+		/* nouveau_bo_download_dirty won't wait unless it needs to
+		 * free a temp buffer, which isn't the case if cpu_mem is
+		 * present.
+		 */
+		if (nr_dirty && nbo->cpu_mem && nbo->cpu_mem != nbo->gpu_mem)
+			nouveau_notifier_wait_nop(ctx, nmesa->syncNotifier,
+						       NvSubMemFormat);
+	}
+
+	if (nbo->cpu_mem) {
+		DEBUG("..access via cpu_mem\n");
+		return nbo->cpu_mem->map;
+	} else {
+		DEBUG("..access via cpu_mem_sys\n");
+		return nbo->cpu_mem_sys;
+	}
+}
+
+void
+nouveau_bo_unmap(GLcontext *ctx, struct gl_buffer_object *bo)
+{
+	DEBUG("unmap bo=%p\n", bo);
+}
+
+uint32_t
+nouveau_bo_gpu_ref(GLcontext *ctx, struct gl_buffer_object *bo)
+{
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+
+	assert(nbo->mapped == GL_FALSE);
+
+	DEBUG("gpu_ref\n");
+	
+	if (!bo->OnCard) {
+		nouveau_bo_move_in(ctx, bo);
+		bo->OnCard = GL_TRUE;
+	}
+	nouveau_bo_upload_dirty(ctx, bo);
+
+	return nouveau_mem_gpu_offset_get(ctx, nbo->gpu_mem);
+}
+
+void
+nouveau_bo_dirty_linear(GLcontext *ctx, GLboolean on_card,
+			uint32_t offset, uint32_t size,
+			struct gl_buffer_object *bo)
+{
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+	nouveau_bufferobj_dirty *dirty;
+	uint32_t start = offset;
+	uint32_t end = offset + size;
+	int i;
+
+	if (nbo->cpu_mem == nbo->gpu_mem)
+		return;
+
+	dirty = on_card ? &nbo->gpu_dirty : &nbo->cpu_dirty;
+
+	DEBUG("on_card=%d, offset=%d, size=%d, bo=%p\n",
+			on_card, offset, size, bo);
+
+	for (i=0; i<dirty->nr_dirty; i++) {
+		nouveau_bufferobj_region *r = &dirty->dirty[i];
+
+		/* already dirty */
+		if (start >= r->start && end <= r->end) {
+			DEBUG("..already dirty\n");
+			return;
+		}
+
+		/* add to the end of a region */
+		if (start >= r->start && start <= r->end) {
+			if (end > r->end) {
+				DEBUG("..extend end of region\n");
+				r->end = end;
+				return;
+			}
+		}
+
+		/* add to the start of a region */
+		if (start < r->start && end >= r->end) {
+			DEBUG("..extend start of region\n");
+			r->start = start;
+			/* .. and to the end */
+			if (end > r->end) {
+				DEBUG("....and end\n");
+				r->end = end;
+			}
+			return;
+		}
+	}
+
+	/* new region */
+	DEBUG("..new dirty\n");
+	dirty->nr_dirty++;
+	dirty->dirty = realloc(dirty->dirty,
+			       sizeof(nouveau_bufferobj_region) *
+			       dirty->nr_dirty);
+	dirty->dirty[dirty->nr_dirty - 1].start = start;
+	dirty->dirty[dirty->nr_dirty - 1].end   = end;
+}
+
+void
+nouveau_bo_dirty_all(GLcontext *ctx, GLboolean on_card,
+		     struct gl_buffer_object *bo)
+{
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+	nouveau_bufferobj_dirty *dirty;
+
+	dirty = on_card ? &nbo->gpu_dirty : &nbo->cpu_dirty;
+	
+	DEBUG("dirty all\n");
+	if (dirty->nr_dirty) {
+		FREE(dirty->dirty);
+		dirty->dirty    = NULL;
+		dirty->nr_dirty = 0;
+	}
+
+	nouveau_bo_dirty_linear(ctx, on_card, 0, bo->Size, bo);
+}
+
+GLuint
+nouveau_bo_upload_dirty(GLcontext *ctx, struct gl_buffer_object *bo)
+{
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+	nouveau_bufferobj_dirty *dirty = &nbo->cpu_dirty;
+	GLuint nr_dirty;
+	int i;
+
+	nr_dirty = dirty->nr_dirty;
+	if (!nr_dirty) {
+		DEBUG("clean\n");
+		return nr_dirty;
+	}
+
+	for (i=0; i<nr_dirty; i++) {
+		nouveau_bufferobj_region *r = &dirty->dirty[i];
+
+		DEBUG("dirty %d: o=0x%08x, s=0x%08x\n",
+				i, r->start, r->end - r->start);
+		nouveau_bo_upload_to_screen(ctx,
+					    r->start, r->end - r->start, bo);
+	}
+
+	FREE(dirty->dirty);
+	dirty->dirty    = NULL;
+	dirty->nr_dirty = 0;
+
+	return nr_dirty;
+}
+
+GLuint
+nouveau_bo_download_dirty(GLcontext *ctx, struct gl_buffer_object *bo)
+{
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)bo;
+	nouveau_bufferobj_dirty *dirty = &nbo->gpu_dirty;
+	GLuint nr_dirty;
+	int i;
+
+	nr_dirty = dirty->nr_dirty;
+	if (nr_dirty) {
+		DEBUG("clean\n");
+		return nr_dirty;
+	}
+	
+	for (i=0; i<nr_dirty; i++) {
+		nouveau_bufferobj_region *r = &dirty->dirty[i];
+
+		DEBUG("dirty %d: o=0x%08x, s=0x%08x\n",
+				i, r->start, r->end - r->start);
+		nouveau_bo_download_from_screen(ctx,
+						r->start,
+						r->end - r->start, bo);
+	}
+
+	FREE(dirty->dirty);
+	dirty->dirty    = NULL;
+	dirty->nr_dirty = 0;
+
+	return nr_dirty;
+}
+
+static void
+nouveauBindBuffer(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj)
+{
+}
+
+static struct gl_buffer_object *
+nouveauNewBufferObject(GLcontext *ctx, GLuint buffer, GLenum target)
+{
+	nouveau_buffer_object *nbo;
+
+	nbo = CALLOC_STRUCT(nouveau_buffer_object_t);
+	if (nbo)
+		_mesa_initialize_buffer_object(&nbo->mesa, buffer, target);
+	DEBUG("bo=%p\n", nbo);
+
+	return nbo ? &nbo->mesa : NULL;
+}
+
+static void
+nouveauDeleteBuffer(GLcontext *ctx, struct gl_buffer_object *obj)
+{
+	nouveau_buffer_object *nbo = (nouveau_buffer_object *)obj;
+
+	if (nbo->gpu_dirty.nr_dirty)
+		FREE(nbo->gpu_dirty.dirty);
+	if (nbo->cpu_dirty.nr_dirty)
+		FREE(nbo->cpu_dirty.dirty);
+	if (nbo->cpu_mem) nouveau_mem_free(ctx, nbo->cpu_mem);
+	if (nbo->gpu_mem) nouveau_mem_free(ctx, nbo->gpu_mem);
+
+	_mesa_delete_buffer_object(ctx, obj);
+}
+
+static void
+nouveauBufferData(GLcontext *ctx, GLenum target, GLsizeiptrARB size,
+		  const GLvoid *data, GLenum usage,
+		  struct gl_buffer_object *obj)
+{
+	GLuint gpu_flags;
+
+	DEBUG("target=%s, size=%d, data=%p, usage=%s, obj=%p\n",
+			_mesa_lookup_enum_by_nr(target),
+			(GLuint)size, data,
+			_mesa_lookup_enum_by_nr(usage),
+			obj);
+
+	switch (target) {
+	case GL_ELEMENT_ARRAY_BUFFER_ARB:
+		gpu_flags = 0;
+		break;
+	default:
+		gpu_flags = NOUVEAU_BO_VRAM_OK | NOUVEAU_BO_AGP_OK;
+		break;
+	}
+	nouveau_bo_init_storage(ctx, gpu_flags, size, data, usage, obj);
+}
+
+static void
+nouveauBufferSubData(GLcontext *ctx, GLenum target, GLintptrARB offset,
+		     GLsizeiptrARB size, const GLvoid *data,
+		     struct gl_buffer_object *obj)
+{
+	GLvoid *out;
+
+	DEBUG("target=%s, offset=0x%x, size=%d, data=%p, obj=%p\n",
+			_mesa_lookup_enum_by_nr(target),
+			(GLuint)offset, (GLuint)size, data, obj);
+
+	out = nouveau_bo_map(ctx, GL_WRITE_ONLY_ARB, obj);
+	_mesa_memcpy(out + offset, data, size);
+	nouveau_bo_dirty_linear(ctx, GL_FALSE, offset, size, obj);
+	nouveau_bo_unmap(ctx, obj);
+}
+
+static void
+nouveauGetBufferSubData(GLcontext *ctx, GLenum target, GLintptrARB offset,
+		     GLsizeiptrARB size, GLvoid *data,
+		     struct gl_buffer_object *obj)
+{
+	const GLvoid *in;
+
+	DEBUG("target=%s, offset=0x%x, size=%d, data=%p, obj=%p\n",
+			_mesa_lookup_enum_by_nr(target),
+			(GLuint)offset, (GLuint)size, data, obj);
+
+	in = nouveau_bo_map(ctx, GL_READ_ONLY_ARB, obj);
+	_mesa_memcpy(data, in + offset, size);
+	nouveau_bo_unmap(ctx, obj);
+}
+
+static void *
+nouveauMapBuffer(GLcontext *ctx, GLenum target, GLenum access,
+		 struct gl_buffer_object *obj)
+{
+	DEBUG("target=%s, access=%s, obj=%p\n",
+			_mesa_lookup_enum_by_nr(target),
+			_mesa_lookup_enum_by_nr(access),
+			obj
+			);
+
+	/* Already mapped.. */
+	if (obj->Pointer)
+		return NULL;
+
+	/* Have to pass READ_WRITE here, nouveau_bo_map will only ensure that
+	 * the cpu_mem buffer is up-to-date if we ask for read access.
+	 *
+	 * However, even if the client only asks for write access, we're still
+	 * forced to reupload the entire buffer.  So, we need the cpu_mem buffer
+	 * to have the correct data all the time.
+	 */
+	obj->Pointer = nouveau_bo_map(ctx, GL_READ_WRITE_ARB, obj);
+
+	/* The GL spec says that a client attempting to write to a bufferobj
+	 * mapped READ_ONLY object may have unpredictable results, possibly
+	 * even program termination.
+	 *
+	 * We're going to use this, and only mark the buffer as dirtied if
+	 * the client asks for write access.
+	 */
+	if (target != GL_READ_ONLY_ARB) {
+		/* We have no way of knowing what was modified by the client,
+		 * so the entire buffer gets dirtied. */
+		nouveau_bo_dirty_all(ctx, GL_FALSE, obj);
+	}
+
+	return obj->Pointer;
+}
+
+static GLboolean
+nouveauUnmapBuffer(GLcontext *ctx, GLenum target, struct gl_buffer_object *obj)
+{
+	DEBUG("target=%s, obj=%p\n", _mesa_lookup_enum_by_nr(target), obj);
+
+	assert(obj->Pointer);
+
+	nouveau_bo_unmap(ctx, obj);
+	obj->Pointer = NULL;
+	return GL_TRUE;
+}
+	  
+void
+nouveauInitBufferObjects(GLcontext *ctx)
+{
+	ctx->Driver.BindBuffer		= nouveauBindBuffer;
+	ctx->Driver.NewBufferObject	= nouveauNewBufferObject;
+	ctx->Driver.DeleteBuffer	= nouveauDeleteBuffer;
+	ctx->Driver.BufferData		= nouveauBufferData;
+	ctx->Driver.BufferSubData	= nouveauBufferSubData;
+	ctx->Driver.GetBufferSubData	= nouveauGetBufferSubData;
+	ctx->Driver.MapBuffer		= nouveauMapBuffer;
+	ctx->Driver.UnmapBuffer		= nouveauUnmapBuffer;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h
new file mode 100644
index 00000000000..932450fd877
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_bufferobj.h
@@ -0,0 +1,77 @@
+#ifndef __NOUVEAU_BUFFEROBJ_H__
+#define __NOUVEAU_BUFFEROBJ_H__
+
+#include "mtypes.h"
+#include "nouveau_buffers.h"
+
+#define NOUVEAU_BO_VRAM_OK (NOUVEAU_MEM_FB | NOUVEAU_MEM_FB_ACCEPTABLE)
+#define NOUVEAU_BO_AGP_OK  (NOUVEAU_MEM_AGP | NOUVEAU_MEM_AGP_ACCEPTABLE)
+
+typedef struct nouveau_bufferobj_region_t {
+	uint32_t start;
+	uint32_t end;
+} nouveau_bufferobj_region;
+
+typedef struct nouveau_bufferobj_dirty_t {
+	nouveau_bufferobj_region *dirty;
+	int nr_dirty;
+} nouveau_bufferobj_dirty;
+
+typedef struct nouveau_buffer_object_t {
+	/* Base class, must be first */
+	struct gl_buffer_object mesa;
+
+	GLboolean		mapped;
+	GLenum			usage;
+
+	/* Memory used for GPU access to the buffer*/
+	GLuint			gpu_mem_flags;
+	nouveau_mem *		gpu_mem;
+	nouveau_bufferobj_dirty	gpu_dirty;
+
+	/* Memory used for CPU access to the buffer */
+	GLuint			cpu_mem_flags;
+	nouveau_mem *		cpu_mem;
+	GLvoid *		cpu_mem_sys;
+	nouveau_bufferobj_dirty	cpu_dirty;
+} nouveau_buffer_object;
+
+extern void
+nouveau_bo_init_storage(GLcontext *ctx, GLuint valid_gpu_access,
+			GLsizeiptrARB size, const GLvoid *data, GLenum usage,
+			struct gl_buffer_object *bo);
+
+extern GLboolean
+nouveau_bo_move_in(GLcontext *ctx, struct gl_buffer_object *bo);
+
+extern GLboolean
+nouveau_bo_move_out(GLcontext *ctx, struct gl_buffer_object *bo);
+
+extern void *
+nouveau_bo_map(GLcontext *ctx, GLenum usage, struct gl_buffer_object *bo);
+
+extern void
+nouveau_bo_unmap(GLcontext *ctx, struct gl_buffer_object *bo);
+
+extern uint32_t
+nouveau_bo_gpu_ref(GLcontext *ctx, struct gl_buffer_object *bo);
+
+extern void
+nouveau_bo_dirty_linear(GLcontext *ctx, GLboolean on_card,
+			uint32_t offset, uint32_t size,
+			struct gl_buffer_object *bo);
+
+extern void
+nouveau_bo_dirty_all(GLcontext *ctx, GLboolean on_card,
+		     struct gl_buffer_object *bo);
+
+extern GLuint
+nouveau_bo_upload_dirty(GLcontext *ctx, struct gl_buffer_object *bo);
+
+extern GLuint
+nouveau_bo_download_dirty(GLcontext *ctx, struct gl_buffer_object *bo);
+
+extern void
+nouveauInitBufferObjects(GLcontext *ctx);
+
+#endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_buffers.c b/src/mesa/drivers/dri/nouveau/nouveau_buffers.c
new file mode 100644
index 00000000000..b54f68f4023
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_buffers.c
@@ -0,0 +1,434 @@
+#include "utils.h"
+#include "framebuffer.h"
+#include "renderbuffer.h"
+#include "fbobject.h"
+
+#include "nouveau_context.h"
+#include "nouveau_buffers.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+#include "nouveau_msg.h"
+
+#define MAX_MEMFMT_LENGTH 32768
+
+/* Unstrided blit using NV_MEMORY_TO_MEMORY_FORMAT */
+GLboolean
+nouveau_memformat_flat_emit(GLcontext *ctx,
+      			    nouveau_mem *dst, nouveau_mem *src,
+			    GLuint dst_offset, GLuint src_offset,
+			    GLuint size)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   uint32_t src_handle, dst_handle;
+   GLuint count;
+
+   if (src_offset + size > src->size) {
+      MESSAGE("src out of nouveau_mem bounds\n");
+      return GL_FALSE;
+   }
+   if (dst_offset + size > dst->size) {
+      MESSAGE("dst out of nouveau_mem bounds\n");
+      return GL_FALSE;
+   }
+
+   src_handle = (src->type & NOUVEAU_MEM_FB) ? NvDmaFB : NvDmaAGP;
+   dst_handle = (dst->type & NOUVEAU_MEM_FB) ? NvDmaFB : NvDmaAGP;
+   src_offset += nouveau_mem_gpu_offset_get(ctx, src);
+   dst_offset += nouveau_mem_gpu_offset_get(ctx, dst);
+
+   BEGIN_RING_SIZE(NvSubMemFormat, NV_MEMORY_TO_MEMORY_FORMAT_OBJECT_IN, 2);
+   OUT_RING       (src_handle);
+   OUT_RING       (dst_handle);
+
+   count = (size / MAX_MEMFMT_LENGTH) + ((size % MAX_MEMFMT_LENGTH) ? 1 : 0);
+
+   while (count--) {
+      GLuint length = (size > MAX_MEMFMT_LENGTH) ? MAX_MEMFMT_LENGTH : size;
+
+      BEGIN_RING_SIZE(NvSubMemFormat, NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN, 8);
+      OUT_RING       (src_offset);
+      OUT_RING       (dst_offset);
+      OUT_RING       (0); /* pitch in */
+      OUT_RING       (0); /* pitch out */
+      OUT_RING       (length); /* line length */
+      OUT_RING       (1); /* number of lines */
+      OUT_RING       ((1 << 8) /* dst_inc */ | (1 << 0) /* src_inc */);
+      OUT_RING       (0); /* buffer notify? */
+      FIRE_RING();
+
+      src_offset += length;
+      dst_offset += length;
+      size       -= length;
+   }
+
+   return GL_TRUE;
+}
+
+void
+nouveau_mem_free(GLcontext *ctx, nouveau_mem *mem)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   drm_nouveau_mem_free_t memf;
+
+   if (NOUVEAU_DEBUG & DEBUG_MEM)  {
+      fprintf(stderr, "%s: type=0x%x, offset=0x%x, size=0x%x\n",
+	    __func__, mem->type, (GLuint)mem->offset, (GLuint)mem->size);
+   }
+
+   if (mem->map)
+      drmUnmap(mem->map, mem->size);
+   memf.flags         = mem->type;
+   memf.region_offset = mem->offset;
+   drmCommandWrite(nmesa->driFd, DRM_NOUVEAU_MEM_FREE, &memf, sizeof(memf));
+   FREE(mem);
+}
+
+nouveau_mem *
+nouveau_mem_alloc(GLcontext *ctx, int type, GLuint size, GLuint align)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   drm_nouveau_mem_alloc_t mema;
+   nouveau_mem *mem;
+   int ret;
+
+   if (NOUVEAU_DEBUG & DEBUG_MEM)  {
+      fprintf(stderr, "%s: requested: type=0x%x, size=0x%x, align=0x%x\n",
+	    __func__, type, (GLuint)size, align);
+   }
+
+   mem = CALLOC(sizeof(nouveau_mem));
+   if (!mem)
+      return NULL;
+
+   mema.flags     = type;
+   mema.size      = mem->size = size;
+   mema.alignment = align;
+   mem->map       = NULL;
+   ret = drmCommandWriteRead(nmesa->driFd, DRM_NOUVEAU_MEM_ALLOC,
+	 		     &mema, sizeof(mema));
+   if (ret) {
+      FREE(mem);
+      return NULL;
+   }
+   mem->offset = mema.region_offset;
+   mem->type   = mema.flags;
+
+   if (NOUVEAU_DEBUG & DEBUG_MEM)  {
+      fprintf(stderr, "%s: actual: type=0x%x, offset=0x%x, size=0x%x\n",
+	    __func__, mem->type, (GLuint)mem->offset, (GLuint)mem->size);
+   }
+
+   if (type & NOUVEAU_MEM_MAPPED)
+      ret = drmMap(nmesa->driFd, mem->offset, mem->size, &mem->map);
+   if (ret) {
+      mem->map = NULL;
+      nouveau_mem_free(ctx, mem);
+      mem = NULL;
+   }
+
+   return mem;
+}
+
+uint32_t
+nouveau_mem_gpu_offset_get(GLcontext *ctx, nouveau_mem *mem)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+   if (mem->type & NOUVEAU_MEM_FB)
+      return (uint32_t)mem->offset - nmesa->vram_phys;
+   else if (mem->type & NOUVEAU_MEM_AGP)
+      return (uint32_t)mem->offset - nmesa->agp_phys;
+   else
+      return 0xDEADF00D;
+}
+
+static GLboolean
+nouveau_renderbuffer_pixelformat(nouveau_renderbuffer *nrb,
+      				 GLenum internalFormat)
+{
+      nrb->mesa.InternalFormat	= internalFormat;
+
+   /*TODO: We probably want to extend this a bit, and maybe make
+    *      card-specific? 
+    */
+      switch (internalFormat) {
+      case GL_RGBA:
+      case GL_RGBA8:
+	 nrb->mesa._BaseFormat	= GL_RGBA;
+	 nrb->mesa._ActualFormat= GL_RGBA8;
+	 nrb->mesa.DataType	= GL_UNSIGNED_BYTE;
+	 nrb->mesa.RedBits	= 8;
+	 nrb->mesa.GreenBits	= 8;
+	 nrb->mesa.BlueBits	= 8;
+	 nrb->mesa.AlphaBits	= 8;
+	 nrb->cpp		= 4;
+	 break;
+      case GL_RGB:
+      case GL_RGB5:
+	 nrb->mesa._BaseFormat	= GL_RGB;
+	 nrb->mesa._ActualFormat= GL_RGB5;
+	 nrb->mesa.DataType	= GL_UNSIGNED_BYTE;
+	 nrb->mesa.RedBits	= 5;
+	 nrb->mesa.GreenBits	= 6;
+	 nrb->mesa.BlueBits	= 5;
+	 nrb->mesa.AlphaBits	= 0;
+	 nrb->cpp		= 2;
+	 break;
+      case GL_DEPTH_COMPONENT16:
+	 nrb->mesa._BaseFormat	= GL_DEPTH_COMPONENT;
+	 nrb->mesa._ActualFormat= GL_DEPTH_COMPONENT16;
+	 nrb->mesa.DataType	= GL_UNSIGNED_SHORT;
+	 nrb->mesa.DepthBits	= 16;
+	 nrb->cpp		= 2;
+	 break;
+      case GL_DEPTH_COMPONENT24:
+	 nrb->mesa._BaseFormat	= GL_DEPTH_COMPONENT;
+	 nrb->mesa._ActualFormat= GL_DEPTH24_STENCIL8_EXT;
+	 nrb->mesa.DataType	= GL_UNSIGNED_INT_24_8_EXT;
+	 nrb->mesa.DepthBits	= 24;
+	 nrb->cpp		= 4;
+	 break;
+      case GL_STENCIL_INDEX8_EXT:
+	 nrb->mesa._BaseFormat	= GL_STENCIL_INDEX;
+	 nrb->mesa._ActualFormat= GL_DEPTH24_STENCIL8_EXT;
+	 nrb->mesa.DataType	= GL_UNSIGNED_INT_24_8_EXT;
+	 nrb->mesa.StencilBits	= 8;
+	 nrb->cpp		= 4;
+	 break;
+      case GL_DEPTH24_STENCIL8_EXT:
+	 nrb->mesa._BaseFormat	= GL_DEPTH_STENCIL_EXT;
+	 nrb->mesa._ActualFormat= GL_DEPTH24_STENCIL8_EXT;
+	 nrb->mesa.DataType	= GL_UNSIGNED_INT_24_8_EXT;
+	 nrb->mesa.DepthBits	= 24;
+	 nrb->mesa.StencilBits  = 8;
+	 nrb->cpp		= 4;
+	 break;
+      default:
+	 return GL_FALSE;
+	 break;
+      }
+
+      return GL_TRUE;
+}
+
+static GLboolean
+nouveau_renderbuffer_storage(GLcontext *ctx, struct gl_renderbuffer *rb,
+      			     GLenum internalFormat,
+			     GLuint width,
+			     GLuint height)
+{
+   nouveau_renderbuffer *nrb = (nouveau_renderbuffer*)rb;
+
+   if (!nouveau_renderbuffer_pixelformat(nrb, internalFormat)) {
+      fprintf(stderr, "%s: unknown internalFormat\n", __func__);
+      return GL_FALSE;
+   }
+
+   /* If this buffer isn't statically alloc'd, we may need to ask the
+    * drm for more memory */
+   if (!nrb->dPriv && (rb->Width != width || rb->Height != height)) {
+      GLuint pitch;
+
+      /* align pitches to 64 bytes */
+      pitch = ((width * nrb->cpp) + 63) & ~63;
+
+      if (nrb->mem)
+	 nouveau_mem_free(ctx, nrb->mem);
+      nrb->mem = nouveau_mem_alloc(ctx,
+	    			   NOUVEAU_MEM_FB | NOUVEAU_MEM_MAPPED,
+				   pitch*height,
+				   0);
+      if (!nrb->mem)
+	 return GL_FALSE;
+
+      /* update nouveau_renderbuffer info */
+      nrb->offset = nouveau_mem_gpu_offset_get(ctx, nrb->mem);
+      nrb->pitch  = pitch;
+   }
+
+   rb->Width = width;
+   rb->Height = height;
+   rb->InternalFormat = internalFormat;
+   return GL_TRUE;
+}
+
+static void
+nouveau_renderbuffer_delete(struct gl_renderbuffer *rb)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   nouveau_renderbuffer *nrb = (nouveau_renderbuffer*)rb;
+
+   if (nrb->mem)
+      nouveau_mem_free(ctx, nrb->mem);
+   FREE(nrb);
+}
+
+nouveau_renderbuffer *
+nouveau_renderbuffer_new(GLenum internalFormat, GLvoid *map,
+      			 GLuint offset,  GLuint pitch,
+			 __DRIdrawablePrivate *dPriv)
+{
+   nouveau_renderbuffer *nrb;
+
+   nrb = CALLOC_STRUCT(nouveau_renderbuffer_t);
+   if (nrb) {
+      _mesa_init_renderbuffer(&nrb->mesa, 0);
+
+      nouveau_renderbuffer_pixelformat(nrb, internalFormat);
+
+      nrb->mesa.AllocStorage	= nouveau_renderbuffer_storage;
+      nrb->mesa.Delete		= nouveau_renderbuffer_delete;
+
+      nrb->dPriv  = dPriv;
+      nrb->offset = offset;
+      nrb->pitch  = pitch;
+      nrb->map    = map;
+   }
+
+   return nrb;
+}
+
+static void
+nouveau_cliprects_drawable_set(nouveauContextPtr nmesa,
+      			       nouveau_renderbuffer *nrb)
+{
+   __DRIdrawablePrivate *dPriv = nrb->dPriv;
+
+   nmesa->numClipRects	= dPriv->numClipRects;
+   nmesa->pClipRects	= dPriv->pClipRects;
+   nmesa->drawX		= dPriv->x;
+   nmesa->drawY		= dPriv->y;
+}
+
+static void
+nouveau_cliprects_renderbuffer_set(nouveauContextPtr nmesa,
+      				   nouveau_renderbuffer *nrb)
+{
+   nmesa->numClipRects	= 1;
+   nmesa->pClipRects	= &nmesa->osClipRect;
+   nmesa->osClipRect.x1	= 0;
+   nmesa->osClipRect.y1	= 0;
+   nmesa->osClipRect.x2	= nrb->mesa.Width;
+   nmesa->osClipRect.y2	= nrb->mesa.Height;
+   nmesa->drawX		= 0;
+   nmesa->drawY		= 0;
+}
+
+void
+nouveau_window_moved(GLcontext *ctx)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   nouveau_renderbuffer *nrb;
+
+   nrb = (nouveau_renderbuffer *)ctx->DrawBuffer->_ColorDrawBuffers[0][0];
+   if (!nrb)
+      return;
+
+   if (!nrb->dPriv)
+      nouveau_cliprects_renderbuffer_set(nmesa, nrb);
+   else
+      nouveau_cliprects_drawable_set(nmesa, nrb);
+
+   /* Viewport depends on window size/position, nouveauCalcViewport
+    * will take care of calling the hw-specific WindowMoved
+    */
+   ctx->Driver.Viewport(ctx, ctx->Viewport.X, ctx->Viewport.Y,
+	 		     ctx->Viewport.Width, ctx->Viewport.Height);
+   /* Scissor depends on window position */
+   ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
+	 		    ctx->Scissor.Width, ctx->Scissor.Height);
+}
+
+GLboolean
+nouveau_build_framebuffer(GLcontext *ctx, struct gl_framebuffer *fb)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   nouveau_renderbuffer *color[MAX_DRAW_BUFFERS];
+   nouveau_renderbuffer *depth;
+
+   _mesa_update_framebuffer(ctx);
+   _mesa_update_draw_buffer_bounds(ctx);
+
+   color[0] = (nouveau_renderbuffer *)fb->_ColorDrawBuffers[0][0];
+   if (fb->_DepthBuffer && fb->_DepthBuffer->Wrapped)
+      depth = (nouveau_renderbuffer *)fb->_DepthBuffer->Wrapped;
+   else
+      depth = (nouveau_renderbuffer *)fb->_DepthBuffer;
+
+   if (!nmesa->hw_func.BindBuffers(nmesa, 1, color, depth))
+      return GL_FALSE;
+   nouveau_window_moved(ctx);
+
+   return GL_TRUE;
+}
+
+static void
+nouveauDrawBuffer(GLcontext *ctx, GLenum buffer)
+{
+   nouveau_build_framebuffer(ctx, ctx->DrawBuffer);
+}
+
+static struct gl_framebuffer *
+nouveauNewFramebuffer(GLcontext *ctx, GLuint name)
+{
+   return _mesa_new_framebuffer(ctx, name);
+}
+
+static struct gl_renderbuffer *
+nouveauNewRenderbuffer(GLcontext *ctx, GLuint name)
+{
+   nouveau_renderbuffer *nrb;
+
+   nrb = CALLOC_STRUCT(nouveau_renderbuffer_t);
+   if (nrb) {
+      _mesa_init_renderbuffer(&nrb->mesa, name);
+
+      nrb->mesa.AllocStorage	= nouveau_renderbuffer_storage;
+      nrb->mesa.Delete		= nouveau_renderbuffer_delete;
+   }
+   return &nrb->mesa;
+}
+
+static void
+nouveauBindFramebuffer(GLcontext *ctx, GLenum target, struct gl_framebuffer *fb)
+{
+   nouveau_build_framebuffer(ctx, fb);
+}
+
+static void
+nouveauFramebufferRenderbuffer(GLcontext *ctx,
+      			       struct gl_framebuffer *fb,
+      			       GLenum attachment,
+			       struct gl_renderbuffer *rb)
+{
+   _mesa_framebuffer_renderbuffer(ctx, fb, attachment, rb);
+   nouveau_build_framebuffer(ctx, fb);
+}
+
+static void
+nouveauRenderTexture(GLcontext *ctx,
+      		     struct gl_framebuffer *fb,
+      		     struct gl_renderbuffer_attachment *att)
+{
+}
+
+static void
+nouveauFinishRenderTexture(GLcontext *ctx,
+      			   struct gl_renderbuffer_attachment *att)
+{
+}
+
+void
+nouveauInitBufferFuncs(struct dd_function_table *func)
+{
+   func->DrawBuffer		 = nouveauDrawBuffer;
+
+   func->NewFramebuffer		 = nouveauNewFramebuffer;
+   func->NewRenderbuffer	 = nouveauNewRenderbuffer;
+   func->BindFramebuffer	 = nouveauBindFramebuffer;
+   func->FramebufferRenderbuffer = nouveauFramebufferRenderbuffer;
+   func->RenderTexture		 = nouveauRenderTexture;
+   func->FinishRenderTexture	 = nouveauFinishRenderTexture;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_buffers.h b/src/mesa/drivers/dri/nouveau/nouveau_buffers.h
new file mode 100644
index 00000000000..d86455184c2
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_buffers.h
@@ -0,0 +1,48 @@
+#ifndef __NOUVEAU_BUFFERS_H__
+#define __NOUVEAU_BUFFERS_H__
+
+#include <stdint.h>
+#include "mtypes.h"
+#include "utils.h"
+#include "renderbuffer.h"
+
+typedef struct nouveau_mem_t {
+   int type;
+   uint64_t offset;
+   uint64_t size;
+   void*    map;
+} nouveau_mem;
+
+extern nouveau_mem *nouveau_mem_alloc(GLcontext *ctx, int type,
+      				      GLuint size, GLuint align);
+extern void nouveau_mem_free(GLcontext *ctx, nouveau_mem *mem);
+extern uint32_t nouveau_mem_gpu_offset_get(GLcontext *ctx, nouveau_mem *mem);
+
+extern GLboolean nouveau_memformat_flat_emit(GLcontext *ctx,
+      					     nouveau_mem *dst,
+					     nouveau_mem *src,
+					     GLuint dst_offset,
+					     GLuint src_offset,
+					     GLuint size);
+
+typedef struct nouveau_renderbuffer_t {
+   struct gl_renderbuffer mesa; /* must be first! */
+   __DRIdrawablePrivate  *dPriv;
+
+   nouveau_mem *mem;
+   void *	map;
+
+   int		cpp;
+   uint32_t	offset;
+   uint32_t	pitch;
+} nouveau_renderbuffer;
+
+extern nouveau_renderbuffer *nouveau_renderbuffer_new(GLenum internalFormat,
+      GLvoid *map, GLuint offset, GLuint pitch, __DRIdrawablePrivate *dPriv);
+extern void nouveau_window_moved(GLcontext *ctx);
+extern GLboolean nouveau_build_framebuffer(GLcontext *, struct gl_framebuffer *);
+extern nouveau_renderbuffer *nouveau_current_draw_buffer(GLcontext *ctx);
+
+extern void nouveauInitBufferFuncs(struct dd_function_table *func);
+
+#endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_card.c b/src/mesa/drivers/dri/nouveau/nouveau_card.c
new file mode 100644
index 00000000000..91f12f0d704
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_card.c
@@ -0,0 +1,17 @@
+
+#include "nouveau_card.h"
+#include "nouveau_reg.h"
+#include "nouveau_drm.h"
+#include "nouveau_card_list.h"
+
+
+nouveau_card* nouveau_card_lookup(uint32_t device_id)
+{
+	int i;
+	for(i=0;i<sizeof(nouveau_card_list)/sizeof(nouveau_card)-1;i++)
+		if (nouveau_card_list[i].id==(device_id&0xffff))
+			return &(nouveau_card_list[i]);
+	return NULL;
+}
+
+
diff --git a/src/mesa/tnl/t_save_api.h b/src/mesa/drivers/dri/nouveau/nouveau_card.h
index c30e0e8ecdb..8a4c5f2244e 100644
--- a/src/mesa/tnl/t_save_api.h
+++ b/src/mesa/drivers/dri/nouveau/nouveau_card.h
@@ -1,7 +1,6 @@
 /**************************************************************************
 
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
+Copyright 2006 Stephane Marchesin
 All Rights Reserved.
 
 Permission is hereby granted, free of charge, to any person obtaining a
@@ -18,40 +17,33 @@ 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 NON-INFRINGEMENT. IN NO EVENT SHALL
-TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS BE LIABLE FOR ANY CLAIM,
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#ifndef __T_SAVE_API_H__
-#define __T_SAVE_API_H__
-
-#include "t_context.h"
 
-extern GLboolean _tnl_weak_begin( GLcontext *ctx, GLenum mode );
+#ifndef __NOUVEAU_CARD_H__
+#define __NOUVEAU_CARD_H__
 
-extern void _tnl_EndList( GLcontext *ctx );
-extern void _tnl_NewList( GLcontext *ctx, GLuint list, GLenum mode );
+#include "dri_util.h"
+#include "drm.h"
+#include "nouveau_drm.h"
 
-extern void _tnl_EndCallList( GLcontext *ctx );
-extern void _tnl_BeginCallList( GLcontext *ctx, struct mesa_display_list *list );
+typedef struct nouveau_card_t {
+	uint16_t id; /* last 4 digits of pci id, last digit is always 0 */
+	char* name; /* the user-friendly card name */
+	uint32_t class_3d; /* the object class this card uses for 3D */
+	uint32_t type; /* the major card family */
+	uint32_t flags;
+}
+nouveau_card;
 
-extern void _tnl_SaveFlushVertices( GLcontext *ctx );
+#define NV_HAS_LMA 0x00000001
 
-extern void _tnl_save_init( GLcontext *ctx );
-extern void _tnl_save_destroy( GLcontext *ctx );
-
-extern void _tnl_loopback_vertex_list( GLcontext *ctx,
-				       const struct tnl_vertex_list *list );
-
-extern void _tnl_playback_vertex_list( GLcontext *ctx, void *data );
+extern nouveau_card* nouveau_card_lookup(uint32_t device_id);
 
 #endif
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_card_list.h b/src/mesa/drivers/dri/nouveau/nouveau_card_list.h
new file mode 100644
index 00000000000..8ec5c4a188a
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_card_list.h
@@ -0,0 +1,232 @@
+static nouveau_card nouveau_card_list[]={
+{0x0008,  "EDGE 3D",                               0,                             NV_03, 0},
+{0x0009,  "EDGE 3D",                               0,                             NV_03, 0},
+{0x0010,  "Mutara V08",                            0,                             NV_03, 0},
+{0x0020,  "RIVA TNT",                              NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x0028,  "RIVA TNT2/TNT2 Pro",                    NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x0029,  "RIVA TNT2 Ultra",                       NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x002A,  "Riva TnT2",                             NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x002B,  "Riva TnT2",                             NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x002C,  "Vanta/Vanta LT",                        NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x002D,  "RIVA TNT2 Model 64/Model 64 Pro",       NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x002E,  "Vanta",                                 NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x002F,  "Vanta",                                 NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x0040,  "GeForce 6800 Ultra",                    NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0041,  "GeForce 6800",                          NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0042,  "GeForce 6800 LE",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0043,  "NV40.3",                                NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0044,  "GeForce 6800 XT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0045,  "GeForce 6800 GT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0046,  "GeForce 6800 GT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0047,  "GeForce 6800 GS",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0048,  "GeForce 6800 XT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0049,  "NV40GL",                                NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x004D,  "Quadro FX 4000",                        NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x004E,  "Quadro FX 4000",                        NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0090,  "GeForce 7800 GTX",                      NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0091,  "GeForce 7800 GTX",                      NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0092,  "GeForce 7800 GT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0093,  "GeForce 7800 GS",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0098,  "GeForce Go 7800",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0099,  "GE Force Go 7800 GTX",                  NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x009D,  "Quadro FX4500",                         NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00A0,  "Aladdin TNT2",                          NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x00C0,  "GeForce 6800 GS",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00C1,  "GeForce 6800",                          NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00C2,  "GeForce 6800 LE",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00C3,  "Geforce 6800 XT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00C8,  "GeForce Go 6800",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00C9,  "GeForce Go 6800 Ultra",                 NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00CC,  "Quadro FX Go1400",                      NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00CD,  "Quadro FX 3450/4000 SDI",               NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00CE,  "Quadro FX 1400",                        NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00F0,  "GeForce 6800/GeForce 6800 Ultra",       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00F1,  "GeForce 6600/GeForce 6600 GT",          NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00F2,  "GeForce 6600/GeForce 6600 GT",          NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00F3,  "GeForce 6200",                          NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00F4,  "GeForce 6600 LE",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00F5,  "GeForce 7800 GS",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00F6,  "GeForce 6600 GS",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00F8,  "Quadro FX 3400/4400",                   NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00F9,  "GeForce 6800 Ultra/GeForce 6800 GT",    NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x00FA,  "GeForce PCX 5750",                      NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x00FB,  "GeForce PCX 5900",                      NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x00FC,  "Quadro FX 330/GeForce PCX 5300",        NV30_TCL_PRIMITIVE_3D|0x0600,  NV_30, 0},
+{0x00FD,  "Quadro FX 330/Quadro NVS280",           NV30_TCL_PRIMITIVE_3D|0x0600,  NV_30, 0},
+{0x00FE,  "Quadro FX 1300",                        NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x00FF,  "GeForce PCX 4300",                      NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0100,  "GeForce 256 SDR",                       NV10_TCL_PRIMITIVE_3D,         NV_10, 0},
+{0x0101,  "GeForce 256 DDR",                       NV10_TCL_PRIMITIVE_3D,         NV_10, 0},
+{0x0103,  "Quadro",                                NV10_TCL_PRIMITIVE_3D,         NV_10, 0},
+{0x0110,  "GeForce2 MX/MX 400",                    NV11_TCL_PRIMITIVE_3D,         NV_11, 0},
+{0x0111,  "GeForce2 MX 100 DDR/200 DDR",           NV11_TCL_PRIMITIVE_3D,         NV_11, 0},
+{0x0112,  "GeForce2 Go",                           NV11_TCL_PRIMITIVE_3D,         NV_11, 0},
+{0x0113,  "Quadro2 MXR/EX/Go",                     NV11_TCL_PRIMITIVE_3D,         NV_11, 0},
+{0x0140,  "GeForce 6600 GT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0141,  "GeForce 6600",                          NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0142,  "GeForce 6600 PCIe",                     NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0144,  "GeForce Go 6600",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0145,  "GeForce 6610 XL",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0146,  "Geforce Go 6600TE/6200TE",              NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0148,  "GeForce Go 6600",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0149,  "GeForce Go 6600 GT",                    NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x014A,  "Quadro NVS 440",                        NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x014C,  "Quadro FX 550",                         NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x014D,  "Quadro FX 550",                         NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x014E,  "Quadro FX 540",                         NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x014F,  "GeForce 6200",                          NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0150,  "GeForce2 GTS/Pro",                      NV11_TCL_PRIMITIVE_3D,         NV_15, 0},
+{0x0151,  "GeForce2 Ti",                           NV11_TCL_PRIMITIVE_3D,         NV_15, 0},
+{0x0152,  "GeForce2 Ultra, Bladerunner",           NV11_TCL_PRIMITIVE_3D,         NV_15, 0},
+{0x0153,  "Quadro2 Pro",                           NV11_TCL_PRIMITIVE_3D,         NV_15, 0},
+{0x0161,  "GeForce 6200 TurboCache(TM)",           NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0162,  "GeForce 6200 SE TurboCache (TM)",       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0163,  "GeForce 6200 LE",                       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0164,  "GeForce Go 6200",                       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0165,  "Quadro NVS 285",                        NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0166,  "GeForce Go 6400",                       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0167,  "GeForce Go 6200 TurboCache",            NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0168,  "GeForce Go 6200 TurboCache",            NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0170,  "GeForce4 MX 460",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0171,  "GeForce4 MX 440",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0172,  "GeForce4 MX 420",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0173,  "GeForce4 MX 440-SE",                    NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0174,  "GeForce4 440 Go",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0175,  "GeForce4 420 Go",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0176,  "GeForce4 420 Go 32M",                   NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0177,  "GeForce4 460 Go",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0178,  "Quadro4 550 XGL",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0179,  "GeForce4 420 Go 32M",                   NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x017A,  "Quadro4 200/400 NVS",                   NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x017B,  "Quadro4 550 XGL",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x017C,  "Quadro4 500 GoGL",                      NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x017D,  "GeForce4 410 Go 16M",                   NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0181,  "GeForce4 MX 440 AGP 8x",                NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0182,  "GeForce4 MX 440SE AGP 8x",              NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0183,  "GeForce4 MX 420 AGP 8x",                NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0185,  "GeForce4 MX 4000 AGP 8x",               NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0186,  "GeForce4 448 Go",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0187,  "GeForce4 488 Go",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0188,  "Quadro4 580 XGL",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x018A,  "Quadro4 NVS AGP 8x",                    NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x018B,  "Quadro4 380 XGL",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x018C,  "Quadro NVS 50 PCI",                     NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x018D,  "GeForce4 448 Go",                       NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0191,  "GeForce 8800 GTX",                      NV30_TCL_PRIMITIVE_3D|0x5000,  NV_50, 0},
+{0x0193,  "GeForce 8800 GTS",                      NV30_TCL_PRIMITIVE_3D|0x5000,  NV_50, 0},
+{0x01A0,  "GeForce2 MX Integrated Graphics",       NV11_TCL_PRIMITIVE_3D,         NV_11, 0},
+{0x01D1,  "GeForce 7300 LE",                       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x01D6,  "GeForce Go 7200",                       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x01D7,  "Quadro NVS 110M / GeForce Go 7300",     NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x01D8,  "GeForce Go 7400",                       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x01DA,  "Quadro NVS 110M",                       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x01DF,  "GeForce 7300 GS",                       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x01F0,  "GeForce4 MX - nForce GPU",              NV17_TCL_PRIMITIVE_3D,         NV_17, 0},
+{0x0200,  "GeForce3",                              NV20_TCL_PRIMITIVE_3D,         NV_20, 0},
+{0x0201,  "GeForce3 Ti 200",                       NV20_TCL_PRIMITIVE_3D,         NV_20, 0},
+{0x0202,  "GeForce3 Ti 500",                       NV20_TCL_PRIMITIVE_3D,         NV_20, 0},
+{0x0203,  "Quadro DCC",                            NV20_TCL_PRIMITIVE_3D,         NV_20, 0},
+{0x0211,  "GeForce 6800",                          NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0212,  "GeForce 6800 LE",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0215,  "GeForce 6800 GT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0218,  "GeForce 6800 XT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0221,  "GeForce 6200",                          NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0240,  "GeForce 6150",                          NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0242,  "GeForce 6100",                          NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0244,  "Geforce 6150 Go",                       NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0250,  "GeForce4 Ti 4600",                      NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0251,  "GeForce4 Ti 4400",                      NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0252,  "GeForce4 Ti",                           NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0253,  "GeForce4 Ti 4200",                      NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0258,  "Quadro4 900 XGL",                       NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0259,  "Quadro4 750 XGL",                       NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x025B,  "Quadro4 700 XGL",                       NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0280,  "GeForce4 Ti 4800",                      NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0281,  "GeForce4 Ti 4200 AGP 8x",               NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0282,  "GeForce4 Ti 4800 SE",                   NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0286,  "GeForce4 Ti 4200 Go AGP 8x",            NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0288,  "Quadro4 980 XGL",                       NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0289,  "Quadro4 780 XGL",                       NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x028C,  "Quadro4 700 GoGL",                      NV20_TCL_PRIMITIVE_3D|0x0500,  NV_25, 0},
+{0x0290,  "GeForce 7900 GTX",                      NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0291,  "GeForce 7900 GT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0292,  "GeForce 7900 GS",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0298,  "GeForce Go 7900 GS",                    NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0299,  "GeForce Go 7900 GTX",                   NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x029A,  "Quadro FX 2500M",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x029B,  "Quadro FX 1500M",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x029C,  "Quadro FX 5500",                        NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x029D,  "Quadro FX 3500",                        NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x029E,  "Quadro FX 1500",                        NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x029F,  "Quadro FX 4500 X2",                     NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x02A0,  "XGPU",                                  NV20_TCL_PRIMITIVE_3D,         NV_20, 0},
+{0x02E1,  "GeForce 7600 GS",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0300,  "GeForce FX",                            NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0301,  "GeForce FX 5800 Ultra",                 NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0302,  "GeForce FX 5800",                       NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0308,  "Quadro FX 2000",                        NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0309,  "Quadro FX 1000",                        NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0311,  "GeForce FX 5600 Ultra",                 NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0312,  "GeForce FX 5600",                       NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0313,  "NV31",                                  NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0314,  "GeForce FX 5600XT",                     NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0316,  "NV31M",                                 NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0317,  "NV31M Pro",                             NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x031A,  "GeForce FX Go5600",                     NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x031B,  "GeForce FX Go5650",                     NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x031C,  "NVIDIA Quadro FX Go700",                NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x031D,  "NV31GLM",                               NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x031E,  "NV31GLM Pro",                           NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x031F,  "NV31GLM Pro",                           NV30_TCL_PRIMITIVE_3D|0x0300,  NV_30, 0},
+{0x0320,  "GeForce FX 5200",                       NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0321,  "GeForce FX 5200 Ultra",                 NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0322,  "GeForce FX 5200",                       NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0323,  "GeForce FX 5200LE",                     NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0324,  "GeForce FX Go5200",                     NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0325,  "GeForce FX Go5250",                     NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0326,  "GeForce FX 5500",                       NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0327,  "GeForce FX 5100",                       NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0328,  "GeForce FX Go5200 32M/64M",             NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0329,  "GeForce FX Go5200",                     NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x032A,  "Quadro NVS 280 PCI",                    NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x032B,  "Quadro FX 500/600 PCI",                 NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x032C,  "GeForce FX Go 5300",                    NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x032D,  "GeForce FX Go5100",                     NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x032F,  "NV34GL",                                NV30_TCL_PRIMITIVE_3D|0x0600,  NV_34, 0},
+{0x0330,  "GeForce FX 5900 Ultra",                 NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0331,  "GeForce FX 5900",                       NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0332,  "GeForce FX 5900XT",                     NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0333,  "GeForce FX 5950 Ultra",                 NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0334,  "GeForce FX 5900ZT",                     NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0338,  "Quadro FX 3000",                        NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x033F,  "Quadro FX 700",                         NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0341,  "GeForce FX 5700 Ultra",                 NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0342,  "GeForce FX 5700",                       NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0343,  "GeForce FX 5700LE",                     NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0344,  "GeForce FX 5700VE",                     NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0345,  "NV36.5",                                NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0347,  "GeForce FX Go5700",                     NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0348,  "GeForce FX Go5700",                     NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0349,  "NV36M Pro",                             NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x034B,  "NV36MAP",                               NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x034C,  "Quadro FX Go1000",                      NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x034E,  "Quadro FX 1100",                        NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x034F,  "NV36GL",                                NV30_TCL_PRIMITIVE_3D|0x0400,  NV_30, 0},
+{0x0391,  "GeForce 7600 GT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0392,  "GeForce 7600 GS",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0393,  "GeForce 7300 GT",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x0398,  "GeForce Go 7600",                       NV30_TCL_PRIMITIVE_3D|0x4000,  NV_40, 0},
+{0x03D0,  "GeForce 6100 nForce 430",               NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x03D1,  "GeForce 6100 nForce 405",               NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x03D2,  "GeForce 6100 nForce 400",               NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x03D5,  "GeForce 6100 nForce 420",               NV30_TCL_PRIMITIVE_3D|0x4400,  NV_44, 0},
+{0x0008,  "NV1",                                   0,                             NV_03, 0},
+{0x0009,  "DAC64",                                 0,                             NV_03, 0},
+{0x0018,  "Riva128",                               0,                             NV_03, 0},
+{0x0019,  "Riva128ZX",                             0,                             NV_03, 0},
+{0x0020,  "TNT",                                   NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x0028,  "TNT2",                                  NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x0029,  "UTNT2",                                 NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x002C,  "VTNT2",                                 NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+{0x00A0,  "ITNT2",                                 NV04_DX5_TEXTURED_TRIANGLE,    NV_04, 0},
+};
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.c b/src/mesa/drivers/dri/nouveau/nouveau_context.c
new file mode 100644
index 00000000000..f811dc1b72e
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_context.c
@@ -0,0 +1,371 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#include "glheader.h"
+#include "context.h"
+#include "simple_list.h"
+#include "imports.h"
+#include "matrix.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "framebuffer.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+#include "tnl/t_vp_build.h"
+
+#include "drivers/common/driverfuncs.h"
+
+#include "nouveau_context.h"
+#include "nouveau_driver.h"
+//#include "nouveau_state.h"
+#include "nouveau_span.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_tex.h"
+#include "nouveau_msg.h"
+#include "nouveau_reg.h"
+#include "nouveau_lock.h"
+#include "nouveau_query.h"
+#include "nv04_swtcl.h"
+#include "nv10_swtcl.h"
+
+#include "vblank.h"
+#include "utils.h"
+#include "texmem.h"
+#include "xmlpool.h" /* for symbolic values of enum-type options */
+
+#ifndef NOUVEAU_DEBUG
+int NOUVEAU_DEBUG = 0;
+#endif
+
+static const struct dri_debug_control debug_control[] =
+{
+	{ "shaders"   , DEBUG_SHADERS    },
+	{ "mem"       , DEBUG_MEM        },
+	{ "bufferobj" , DEBUG_BUFFEROBJ  },
+	{ NULL        , 0                }
+};
+
+#define need_GL_ARB_vertex_program
+#define need_GL_ARB_occlusion_query
+#include "extension_helper.h"
+
+const struct dri_extension common_extensions[] =
+{
+	{ NULL,    0 }
+};
+
+const struct dri_extension nv10_extensions[] =
+{
+	{ NULL,    0 }
+};
+
+const struct dri_extension nv20_extensions[] =
+{
+	{ NULL,    0 }
+};
+
+const struct dri_extension nv30_extensions[] =
+{
+	{ "GL_ARB_fragment_program",	NULL                            },
+	{ NULL,    0 }
+};
+
+const struct dri_extension nv40_extensions[] =
+{
+   /* ARB_vp can be moved to nv20/30 once the shader backend has been
+    * written for those cards.
+    */
+	{ "GL_ARB_vertex_program",	GL_ARB_vertex_program_functions },
+	{ "GL_ARB_occlusion_query",	GL_ARB_occlusion_query_functions},
+	{ NULL, 0 }
+};
+
+const struct dri_extension nv50_extensions[] =
+{
+	{ NULL,    0 }
+};
+
+/* Create the device specific context.
+ */
+GLboolean nouveauCreateContext( const __GLcontextModes *glVisual,
+		__DRIcontextPrivate *driContextPriv,
+		void *sharedContextPrivate )
+{
+	GLcontext *ctx, *shareCtx;
+	__DRIscreenPrivate *sPriv = driContextPriv->driScreenPriv;
+	struct dd_function_table functions;
+	nouveauContextPtr nmesa;
+	nouveauScreenPtr screen;
+
+	/* Allocate the context */
+	nmesa = (nouveauContextPtr) CALLOC( sizeof(*nmesa) );
+	if ( !nmesa )
+		return GL_FALSE;
+
+	nmesa->driContext = driContextPriv;
+	nmesa->driScreen = sPriv;
+	nmesa->driDrawable = NULL;
+	nmesa->hHWContext = driContextPriv->hHWContext;
+	nmesa->driHwLock = &sPriv->pSAREA->lock;
+	nmesa->driFd = sPriv->fd;
+
+	nmesa->screen = (nouveauScreenPtr)(sPriv->private);
+	screen=nmesa->screen;
+
+	/* Create the hardware context */
+	if (!nouveauDRMGetParam(nmesa, NOUVEAU_GETPARAM_FB_PHYSICAL,
+		 		&nmesa->vram_phys))
+	   return GL_FALSE;
+	if (!nouveauDRMGetParam(nmesa, NOUVEAU_GETPARAM_AGP_PHYSICAL,
+		 		&nmesa->agp_phys))
+	   return GL_FALSE;
+	if (!nouveauFifoInit(nmesa))
+	   return GL_FALSE;
+	nouveauObjectInit(nmesa);
+
+
+	/* Init default driver functions then plug in our nouveau-specific functions
+	 * (the texture functions are especially important)
+	 */
+	_mesa_init_driver_functions( &functions );
+	nouveauDriverInitFunctions( &functions );
+	nouveauTexInitFunctions( &functions );
+
+	/* Allocate the Mesa context */
+	if (sharedContextPrivate)
+		shareCtx = ((nouveauContextPtr) sharedContextPrivate)->glCtx;
+	else 
+		shareCtx = NULL;
+	nmesa->glCtx = _mesa_create_context(glVisual, shareCtx,
+			&functions, (void *) nmesa);
+	if (!nmesa->glCtx) {
+		FREE(nmesa);
+		return GL_FALSE;
+	}
+	driContextPriv->driverPrivate = nmesa;
+	ctx = nmesa->glCtx;
+
+	/* Parse configuration files */
+	driParseConfigFiles (&nmesa->optionCache, &screen->optionCache,
+			screen->driScreen->myNum, "nouveau");
+
+	nmesa->sarea = (drm_nouveau_sarea_t *)((char *)sPriv->pSAREA +
+			screen->sarea_priv_offset);
+
+	/* Enable any supported extensions */
+	driInitExtensions(ctx, common_extensions, GL_TRUE);
+	if (nmesa->screen->card->type >= NV_10)
+		driInitExtensions(ctx, nv10_extensions, GL_FALSE);
+	if (nmesa->screen->card->type >= NV_20)
+		driInitExtensions(ctx, nv20_extensions, GL_FALSE);
+	if (nmesa->screen->card->type >= NV_30)
+		driInitExtensions(ctx, nv30_extensions, GL_FALSE);
+	if (nmesa->screen->card->type >= NV_40)
+		driInitExtensions(ctx, nv40_extensions, GL_FALSE);
+	if (nmesa->screen->card->type >= NV_50)
+		driInitExtensions(ctx, nv50_extensions, GL_FALSE);
+
+	nmesa->current_primitive = -1;
+
+	nouveauShaderInitFuncs(ctx);
+	/* Install Mesa's fixed-function texenv shader support */
+	if (nmesa->screen->card->type >= NV_40)
+		ctx->_MaintainTexEnvProgram = GL_TRUE;
+
+	/* Initialize the swrast */
+	_swrast_CreateContext( ctx );
+	_vbo_CreateContext( ctx );
+	_tnl_CreateContext( ctx );
+	_swsetup_CreateContext( ctx );
+
+	_math_matrix_ctr(&nmesa->viewport);
+
+	nouveauDDInitStateFuncs( ctx );
+	nouveauSpanInitFunctions( ctx );
+	nouveauDDInitState( nmesa );
+	switch(nmesa->screen->card->type)
+	{
+		case NV_03:
+			//nv03TriInitFunctions( ctx );
+			break;
+		case NV_04:
+		case NV_05:
+			nv04TriInitFunctions( ctx );
+			break;
+		case NV_10:
+		case NV_20:
+		case NV_30:
+		case NV_40:
+		case NV_44:
+		case NV_50:
+		default:
+			nv10TriInitFunctions( ctx );
+			break;
+	}
+
+	nouveauInitBufferObjects(ctx);
+	if (!nouveauSyncInitFuncs(ctx))
+	   return GL_FALSE;
+	nouveauQueryInitFuncs(ctx);
+	nmesa->hw_func.InitCard(nmesa);
+        nouveauInitState(ctx);
+
+	driContextPriv->driverPrivate = (void *)nmesa;
+
+	NOUVEAU_DEBUG = driParseDebugString( getenv( "NOUVEAU_DEBUG" ),
+			debug_control );
+
+	if (driQueryOptionb(&nmesa->optionCache, "no_rast")) {
+		fprintf(stderr, "disabling 3D acceleration\n");
+		FALLBACK(nmesa, NOUVEAU_FALLBACK_DISABLE, 1);
+	}
+
+	return GL_TRUE;
+}
+
+/* Destroy the device specific context. */
+void nouveauDestroyContext( __DRIcontextPrivate *driContextPriv  )
+{
+	nouveauContextPtr nmesa = (nouveauContextPtr) driContextPriv->driverPrivate;
+
+	assert(nmesa);
+	if ( nmesa ) {
+		/* free the option cache */
+		driDestroyOptionCache (&nmesa->optionCache);
+
+		FREE( nmesa );
+	}
+
+}
+
+
+/* Force the context `c' to be the current context and associate with it
+ * buffer `b'.
+ */
+GLboolean nouveauMakeCurrent( __DRIcontextPrivate *driContextPriv,
+		__DRIdrawablePrivate *driDrawPriv,
+		__DRIdrawablePrivate *driReadPriv )
+{
+	if ( driContextPriv ) {
+		nouveauContextPtr nmesa = (nouveauContextPtr) driContextPriv->driverPrivate;
+		struct gl_framebuffer *draw_fb =
+			(struct gl_framebuffer*)driDrawPriv->driverPrivate;
+		struct gl_framebuffer *read_fb =
+			(struct gl_framebuffer*)driReadPriv->driverPrivate;
+
+		driDrawableInitVBlank(driDrawPriv, nmesa->vblank_flags, &nmesa->vblank_seq );
+		nmesa->driDrawable = driDrawPriv;
+
+		_mesa_resize_framebuffer(nmesa->glCtx, draw_fb,
+					 driDrawPriv->w, driDrawPriv->h);
+		if (draw_fb != read_fb) {
+			_mesa_resize_framebuffer(nmesa->glCtx, draw_fb,
+						 driReadPriv->w,
+						 driReadPriv->h);
+		}
+		_mesa_make_current(nmesa->glCtx, draw_fb, read_fb);
+
+		nouveau_build_framebuffer(nmesa->glCtx,
+		      			  driDrawPriv->driverPrivate);
+	} else {
+		_mesa_make_current( NULL, NULL, NULL );
+	}
+
+	return GL_TRUE;
+}
+
+
+/* Force the context `c' to be unbound from its buffer.
+ */
+GLboolean nouveauUnbindContext( __DRIcontextPrivate *driContextPriv )
+{
+	return GL_TRUE;
+}
+
+static void nouveauDoSwapBuffers(nouveauContextPtr nmesa,
+				 __DRIdrawablePrivate *dPriv)
+{
+	struct gl_framebuffer *fb;
+	nouveau_renderbuffer *src, *dst;
+	drm_clip_rect_t *box;
+	int nbox, i;
+
+	fb = (struct gl_framebuffer *)dPriv->driverPrivate;
+	dst = (nouveau_renderbuffer*)
+		fb->Attachment[BUFFER_FRONT_LEFT].Renderbuffer;
+	src = (nouveau_renderbuffer*)
+		fb->Attachment[BUFFER_BACK_LEFT].Renderbuffer;
+
+#ifdef ALLOW_MULTI_SUBCHANNEL
+	LOCK_HARDWARE(nmesa);
+	nbox = dPriv->numClipRects;
+	box  = dPriv->pClipRects;
+
+	if (nbox) {
+		BEGIN_RING_SIZE(NvSubCtxSurf2D,
+				NV10_CONTEXT_SURFACES_2D_FORMAT, 4);
+		if (src->mesa._ActualFormat == GL_RGBA8)
+			OUT_RING       (6); /* X8R8G8B8 */
+		else
+			OUT_RING       (4); /* R5G6B5 */
+		OUT_RING       ((dst->pitch << 16) | src->pitch);
+		OUT_RING       (src->offset);
+		OUT_RING       (dst->offset);
+	}
+
+	for (i=0; i<nbox; i++, box++) {
+		BEGIN_RING_SIZE(NvSubImageBlit, NV10_IMAGE_BLIT_SET_POINT, 3);
+		OUT_RING       (((box->y1 - dPriv->y) << 16) |
+				(box->x1 - dPriv->x));
+		OUT_RING       ((box->y1 << 16) | box->x1);
+		OUT_RING       (((box->y2 - box->y1) << 16) |
+				(box->x2 - box->x1));
+	}
+
+	UNLOCK_HARDWARE(nmesa);
+#endif
+}
+
+void nouveauSwapBuffers(__DRIdrawablePrivate *dPriv)
+{
+	if (dPriv->driContextPriv && dPriv->driContextPriv->driverPrivate) {
+		nouveauContextPtr nmesa = dPriv->driContextPriv->driverPrivate;
+
+		if (nmesa->glCtx->Visual.doubleBufferMode) {
+			_mesa_notifySwapBuffers(nmesa->glCtx);
+			nouveauDoSwapBuffers(nmesa, dPriv);
+		}
+
+	}
+}
+
+void nouveauCopySubBuffer(__DRIdrawablePrivate *dPriv,
+			  int x, int y, int w, int h)
+{
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_context.h b/src/mesa/drivers/dri/nouveau/nouveau_context.h
new file mode 100644
index 00000000000..f61fcbb48c0
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_context.h
@@ -0,0 +1,233 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+
+#ifndef __NOUVEAU_CONTEXT_H__
+#define __NOUVEAU_CONTEXT_H__
+
+#include "dri_util.h"
+#include "drm.h"
+#include "nouveau_drm.h"
+
+#include "mtypes.h"
+#include "tnl/t_vertex.h"
+
+#include "nouveau_screen.h"
+#include "nouveau_state_cache.h"
+#include "nouveau_buffers.h"
+#include "nouveau_shader.h"
+#include "nouveau_sync.h"
+
+#include "xmlconfig.h"
+
+typedef struct nouveau_fifo_t{
+	u_int32_t* buffer;
+	u_int32_t* mmio;
+	u_int32_t put_base;
+	u_int32_t current;
+	u_int32_t put;
+	u_int32_t free;
+	u_int32_t max;
+}
+nouveau_fifo;
+
+#define TAG(x) nouveau##x
+#include "tnl_dd/t_dd_vertex.h"
+#undef TAG
+
+/* Subpixel offsets for window coordinates (triangles): */
+#define SUBPIXEL_X  (0.0F)
+#define SUBPIXEL_Y  (0.125F)
+
+struct nouveau_context;
+
+typedef void (*nouveau_tri_func)( struct nouveau_context*, 
+		nouveauVertex *,
+		nouveauVertex *,
+		nouveauVertex * );
+
+typedef void (*nouveau_line_func)( struct nouveau_context*, 
+		nouveauVertex *,
+		nouveauVertex * );
+
+typedef void (*nouveau_point_func)( struct nouveau_context*,
+		nouveauVertex * );
+
+typedef struct nouveau_hw_func_t {
+	/* Initialise any card-specific non-GL related state */
+	GLboolean (*InitCard)(struct nouveau_context *);
+	/* Update buffer offset/pitch/format */
+	GLboolean (*BindBuffers)(struct nouveau_context *, int num_color,
+				 nouveau_renderbuffer **color,
+				 nouveau_renderbuffer *depth);
+	/* Update anything that depends on the window position/size */
+	void      (*WindowMoved)(struct nouveau_context *);
+} nouveau_hw_func;
+
+typedef struct nouveau_context {
+	/* Mesa context */
+	GLcontext *glCtx;
+
+	/* The per-context fifo */
+	nouveau_fifo fifo;
+
+	/* The read-only regs */
+	volatile unsigned char* mmio;
+
+	/* Physical addresses of AGP/VRAM apertures */
+	uint64_t vram_phys;
+	uint64_t agp_phys;
+
+	/* Channel synchronisation */
+	nouveau_notifier *syncNotifier;
+
+	/* ARB_occlusion_query / EXT_timer_query */
+	GLuint		  query_object_max;
+	GLboolean *	  query_alloc;
+	nouveau_notifier *queryNotifier;
+
+	/* Additional hw-specific functions */
+	nouveau_hw_func hw_func;
+
+	/* FIXME : do we want to put all state into a separate struct ? */
+	/* State for tris */
+	GLuint color_offset;
+	GLuint specular_offset;
+
+	/* Vertex state */
+	GLuint vertex_size;
+	GLubyte *verts;
+	struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
+	GLuint vertex_attr_count;
+
+	/* Color buffer clear value */
+	uint32_t clear_color_value;
+
+	/* Depth/stencil clear value */
+	uint32_t clear_value;
+
+	/* Light state */
+	GLboolean lighting_enabled;
+	uint32_t enabled_lights;
+
+	/* Cached state */
+	nouveau_state_cache state_cache;
+
+	/* The drawing fallbacks */
+	GLuint Fallback;
+	nouveau_tri_func draw_tri;
+	nouveau_line_func draw_line;
+	nouveau_point_func draw_point;
+
+	/* Cliprects information */
+	GLuint numClipRects;
+	drm_clip_rect_t *pClipRects;
+	drm_clip_rect_t osClipRect;
+	GLuint drawX, drawY;
+
+	/* The rendering context information */
+	GLenum current_primitive; /* the current primitive enum */
+	DECLARE_RENDERINPUTS(render_inputs_bitset); /* the current render inputs */
+
+	/* Shader state */
+	nvsFunc VPfunc;
+	nvsFunc FPfunc;
+	nouveauShader *current_fragprog;
+	nouveauShader *current_vertprog;
+	nouveauShader *passthrough_vp;
+	nouveauShader *passthrough_fp;
+
+	nouveauScreenRec *screen;
+	drm_nouveau_sarea_t *sarea;
+
+	__DRIcontextPrivate  *driContext;    /* DRI context */
+	__DRIscreenPrivate   *driScreen;     /* DRI screen */
+	__DRIdrawablePrivate *driDrawable;   /* DRI drawable bound to this ctx */
+	GLint lastStamp;
+
+	drm_context_t hHWContext;
+	drm_hw_lock_t *driHwLock;
+	int driFd;
+
+	/* Configuration cache */
+	driOptionCache optionCache;
+
+	/* vblank stuff */
+	uint32_t vblank_flags;
+	uint32_t vblank_seq;
+
+	GLuint new_state;
+	GLuint new_render_state;
+	GLuint render_index;
+	GLmatrix viewport;
+	GLfloat depth_scale;
+
+}nouveauContextRec, *nouveauContextPtr;
+
+
+#define NOUVEAU_CONTEXT(ctx)		((nouveauContextPtr)(ctx->DriverCtx))
+
+/* Flags for software fallback cases: */
+#define NOUVEAU_FALLBACK_TEXTURE		0x0001
+#define NOUVEAU_FALLBACK_DRAW_BUFFER		0x0002
+#define NOUVEAU_FALLBACK_READ_BUFFER		0x0004
+#define NOUVEAU_FALLBACK_STENCIL		0x0008
+#define NOUVEAU_FALLBACK_RENDER_MODE		0x0010
+#define NOUVEAU_FALLBACK_LOGICOP		0x0020
+#define NOUVEAU_FALLBACK_SEP_SPECULAR		0x0040
+#define NOUVEAU_FALLBACK_BLEND_EQ		0x0080
+#define NOUVEAU_FALLBACK_BLEND_FUNC		0x0100
+#define NOUVEAU_FALLBACK_PROJTEX		0x0200
+#define NOUVEAU_FALLBACK_DISABLE		0x0400
+
+
+extern GLboolean nouveauCreateContext( const __GLcontextModes *glVisual,
+		__DRIcontextPrivate *driContextPriv,
+		void *sharedContextPrivate );
+
+extern void nouveauDestroyContext( __DRIcontextPrivate * );
+
+extern GLboolean nouveauMakeCurrent( __DRIcontextPrivate *driContextPriv,
+		__DRIdrawablePrivate *driDrawPriv,
+		__DRIdrawablePrivate *driReadPriv );
+
+extern GLboolean nouveauUnbindContext( __DRIcontextPrivate *driContextPriv );
+
+extern void nouveauSwapBuffers(__DRIdrawablePrivate *dPriv);
+
+extern void nouveauCopySubBuffer(__DRIdrawablePrivate *dPriv,
+				 int x, int y, int w, int h);
+
+/* Debugging utils: */
+extern int NOUVEAU_DEBUG;
+
+#define DEBUG_SHADERS	0x00000001
+#define DEBUG_MEM	0x00000002
+#define DEBUG_BUFFEROBJ 0x00000004
+
+#endif /* __NOUVEAU_CONTEXT_H__ */
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_ctrlreg.h b/src/mesa/drivers/dri/nouveau/nouveau_ctrlreg.h
new file mode 100644
index 00000000000..c9b2d590077
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_ctrlreg.h
@@ -0,0 +1,44 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin, Sylvain Munaut
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+
+
+#define NV03_STATUS                                        0x004006b0
+#define NV04_STATUS                                        0x00400700
+
+#define NV03_FIFO_REGS_SIZE                                0x10000
+#    define NV03_FIFO_REGS_DMAPUT                          0x00000040
+#    define NV03_FIFO_REGS_DMAGET                          0x00000044
+
+/* Fifo commands. These are not regs, neither masks */
+#define NV03_FIFO_CMD_JUMP                                 0x20000000
+#define NV03_FIFO_CMD_JUMP_OFFSET_MASK                     0x1ffffffc
+#define NV03_FIFO_CMD_REWIND                               (NV03_FIFO_CMD_JUMP | (0 & NV03_FIFO_CMD_JUMP_OFFSET_MASK))
+
+
+#define NONINC_METHOD                                      0x40000000
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_dri.h b/src/mesa/drivers/dri/nouveau/nouveau_dri.h
new file mode 100644
index 00000000000..ce3c3fb9cc4
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_dri.h
@@ -0,0 +1,28 @@
+#ifndef _NOUVEAU_DRI_
+#define _NOUVEAU_DRI_
+
+#include "xf86drm.h"
+#include "drm.h"
+#include "nouveau_drm.h"
+
+typedef struct {
+	uint32_t device_id;	/**< \brief PCI device ID */
+	uint32_t width;		/**< \brief width in pixels of display */
+	uint32_t height;	/**< \brief height in scanlines of display */
+	uint32_t depth;		/**< \brief depth of display (8, 15, 16, 24) */
+	uint32_t bpp;		/**< \brief bit depth of display (8, 16, 24, 32) */
+
+	uint32_t bus_type;	/**< \brief ths bus type */
+	uint32_t bus_mode;	/**< \brief bus mode (used for AGP, maybe also for PCI-E ?) */
+
+	uint32_t front_offset;	/**< \brief front buffer offset */
+	uint32_t front_pitch;	/**< \brief front buffer pitch */
+	uint32_t back_offset;	/**< \brief private back buffer offset */
+	uint32_t back_pitch;	/**< \brief private back buffer pitch */
+	uint32_t depth_offset;	/**< \brief private depth buffer offset */
+	uint32_t depth_pitch;	/**< \brief private depth buffer pitch */
+
+} NOUVEAUDRIRec, *NOUVEAUDRIPtr;
+
+#endif
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_driver.c b/src/mesa/drivers/dri/nouveau/nouveau_driver.c
new file mode 100644
index 00000000000..00956aa8f8b
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_driver.c
@@ -0,0 +1,146 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#include "nouveau_context.h"
+//#include "nouveau_state.h"
+#include "nouveau_lock.h"
+#include "nouveau_fifo.h"
+#include "nouveau_driver.h"
+#include "swrast/swrast.h"
+
+#include "context.h"
+#include "framebuffer.h"
+
+#include "utils.h"
+
+/* Wrapper for DRM_NOUVEAU_GETPARAM ioctl */
+GLboolean nouveauDRMGetParam(nouveauContextPtr nmesa,
+			     unsigned int      param,
+			     uint64_t*         value)
+{
+	drm_nouveau_getparam_t getp;
+
+	getp.param = param;
+	if (!value || drmCommandWriteRead(nmesa->driFd, DRM_NOUVEAU_GETPARAM,
+					  &getp, sizeof(getp)))
+		return GL_FALSE;
+	*value = getp.value;
+	return GL_TRUE;
+}
+
+/* Wrapper for DRM_NOUVEAU_GETPARAM ioctl */
+GLboolean nouveauDRMSetParam(nouveauContextPtr nmesa,
+			     unsigned int      param,
+			     uint64_t          value)
+{
+	drm_nouveau_setparam_t setp;
+
+	setp.param = param;
+	setp.value = value;
+	if (drmCommandWrite(nmesa->driFd, DRM_NOUVEAU_SETPARAM, &setp,
+				sizeof(setp)))
+		return GL_FALSE;
+	return GL_TRUE;
+}
+
+/* Return the width and height of the current color buffer */
+static void nouveauGetBufferSize( GLframebuffer *buffer,
+		GLuint *width, GLuint *height )
+{
+	GET_CURRENT_CONTEXT(ctx);
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	LOCK_HARDWARE( nmesa );
+	*width  = nmesa->driDrawable->w;
+	*height = nmesa->driDrawable->h;
+	UNLOCK_HARDWARE( nmesa );
+}
+
+/* glGetString */
+static const GLubyte *nouveauGetString( GLcontext *ctx, GLenum name )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	static char buffer[128];
+	const char * card_name = "Unknown";
+	GLuint agp_mode = 0;
+
+	switch ( name ) {
+		case GL_VENDOR:
+			return (GLubyte *)DRIVER_AUTHOR;
+
+		case GL_RENDERER:
+			card_name=nmesa->screen->card->name;
+
+			switch(nmesa->screen->bus_type)
+			{
+				case NV_PCI:
+				case NV_PCIE:
+				default:
+					agp_mode=0;
+					break;
+				case NV_AGP:
+					agp_mode=nmesa->screen->agp_mode;
+					break;
+			}
+			driGetRendererString( buffer, card_name, DRIVER_DATE,
+					agp_mode );
+			return (GLubyte *)buffer;
+		default:
+			return NULL;
+	}
+}
+
+/* glFlush */
+static void nouveauFlush( GLcontext *ctx )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	FIRE_RING();
+}
+
+/* glFinish */
+static void nouveauFinish( GLcontext *ctx )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveauFlush( ctx );
+	nouveauWaitForIdle( nmesa );
+}
+
+/* glClear */
+static void nouveauClear( GLcontext *ctx, GLbitfield mask )
+{
+	// XXX we really should do something here...
+}
+
+void nouveauDriverInitFunctions( struct dd_function_table *functions )
+{
+	functions->GetBufferSize	= nouveauGetBufferSize;
+	functions->ResizeBuffers	= _mesa_resize_framebuffer;
+	functions->GetString		= nouveauGetString;
+	functions->Flush		= nouveauFlush;
+	functions->Finish		= nouveauFinish;
+	functions->Clear		= nouveauClear;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_driver.h b/src/mesa/drivers/dri/nouveau/nouveau_driver.h
new file mode 100644
index 00000000000..6164012b5b7
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_driver.h
@@ -0,0 +1,42 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+
+#ifndef __NOUVEAU_DRIVER_H__
+#define __NOUVEAU_DRIVER_H__
+
+#define DRIVER_DATE	"20060219"
+#define DRIVER_AUTHOR	"Stephane Marchesin"
+
+extern void nouveauDriverInitFunctions( struct dd_function_table *functions );
+extern GLboolean nouveauDRMGetParam(nouveauContextPtr nmesa, unsigned int param,
+      				    uint64_t *value);
+extern GLboolean nouveauDRMSetParam(nouveauContextPtr nmesa, unsigned int param,
+      				    uint64_t value);
+
+#endif /* __NOUVEAU_DRIVER_H__ */
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_fifo.c b/src/mesa/drivers/dri/nouveau/nouveau_fifo.c
new file mode 100644
index 00000000000..7af9f1e3c2f
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_fifo.c
@@ -0,0 +1,137 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+#include "vblank.h"
+#include <errno.h>
+#include "mtypes.h"
+#include "macros.h"
+#include "dd.h"
+#include "swrast/swrast.h"
+#include "nouveau_context.h"
+#include "nouveau_msg.h"
+#include "nouveau_fifo.h"
+#include "nouveau_lock.h"
+#include "nouveau_object.h"
+#include "nouveau_sync.h"
+
+
+#define RING_SKIPS 8
+
+void WAIT_RING(nouveauContextPtr nmesa,u_int32_t size)
+{
+#ifdef NOUVEAU_RING_DEBUG
+	return;
+#endif
+	u_int32_t fifo_get;
+	while(nmesa->fifo.free < size+1) {
+		fifo_get = NV_FIFO_READ_GET();
+
+		if(nmesa->fifo.put >= fifo_get) {
+			nmesa->fifo.free = nmesa->fifo.max - nmesa->fifo.current;
+			if(nmesa->fifo.free < size+1) {
+				OUT_RING(NV03_FIFO_CMD_JUMP | nmesa->fifo.put_base);
+				if(fifo_get <= RING_SKIPS) {
+					if(nmesa->fifo.put <= RING_SKIPS) /* corner case - will be idle */
+						NV_FIFO_WRITE_PUT(RING_SKIPS + 1);
+					do { fifo_get = NV_FIFO_READ_GET(); }
+					while(fifo_get <= RING_SKIPS);
+				}
+				NV_FIFO_WRITE_PUT(RING_SKIPS);
+				nmesa->fifo.current = nmesa->fifo.put = RING_SKIPS;
+				nmesa->fifo.free = fifo_get - (RING_SKIPS + 1);
+			}
+		} else 
+			nmesa->fifo.free = fifo_get - nmesa->fifo.current - 1;
+	}
+}
+
+/* 
+ * Wait for the channel to be idle 
+ */
+void nouveauWaitForIdleLocked(nouveauContextPtr nmesa)
+{
+	/* Wait for FIFO idle */
+	FIRE_RING();
+	while(RING_AHEAD()>0);
+
+	/* Wait on notifier to indicate all commands in the channel have
+	 * been completed.
+	 */
+	nouveau_notifier_wait_nop(nmesa->glCtx, nmesa->syncNotifier, NvSub3D);
+}
+
+void nouveauWaitForIdle(nouveauContextPtr nmesa)
+{
+	LOCK_HARDWARE(nmesa);
+	nouveauWaitForIdleLocked(nmesa);
+	UNLOCK_HARDWARE(nmesa);
+}
+
+// here we call the fifo initialization ioctl and fill in stuff accordingly
+GLboolean nouveauFifoInit(nouveauContextPtr nmesa)
+{
+	drm_nouveau_fifo_alloc_t fifo_init;
+	int i;
+
+#ifdef NOUVEAU_RING_DEBUG
+	return GL_TRUE;
+#endif
+
+	int ret;
+	ret=drmCommandWriteRead(nmesa->driFd, DRM_NOUVEAU_FIFO_ALLOC, &fifo_init, sizeof(fifo_init));
+	if (ret) {
+		FATAL("Fifo initialization ioctl failed (returned %d)\n",ret);
+		return GL_FALSE;
+	}
+
+	ret = drmMap(nmesa->driFd, fifo_init.cmdbuf, fifo_init.cmdbuf_size, &nmesa->fifo.buffer);
+	if (ret) {
+		FATAL("Unable to map the fifo (returned %d)\n",ret);
+		return GL_FALSE;
+	}
+	ret = drmMap(nmesa->driFd, fifo_init.ctrl, fifo_init.ctrl_size, &nmesa->fifo.mmio);
+	if (ret) {
+		FATAL("Unable to map the control regs (returned %d)\n",ret);
+		return GL_FALSE;
+	}
+
+	/* Setup our initial FIFO tracking params */
+	nmesa->fifo.put_base = fifo_init.put_base;
+	nmesa->fifo.current  = 0;
+	nmesa->fifo.put      = 0;
+	nmesa->fifo.max      = (fifo_init.cmdbuf_size >> 2) - 1;
+	nmesa->fifo.free     = nmesa->fifo.max - nmesa->fifo.current;
+
+	for (i=0; i<RING_SKIPS; i++)
+	   OUT_RING(0);
+	nmesa->fifo.free -= RING_SKIPS;
+
+	MESSAGE("Fifo init ok. Using context %d\n", fifo_init.channel);
+	return GL_TRUE;
+}
+
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_fifo.h b/src/mesa/drivers/dri/nouveau/nouveau_fifo.h
new file mode 100644
index 00000000000..490089f71a4
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_fifo.h
@@ -0,0 +1,179 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+
+#ifndef __NOUVEAU_FIFO_H__
+#define __NOUVEAU_FIFO_H__
+
+#include "nouveau_context.h"
+#include "nouveau_ctrlreg.h"
+#include "nouveau_state_cache.h"
+
+//#define NOUVEAU_RING_TRACE
+//#define NOUVEAU_RING_DEBUG
+//#define NOUVEAU_STATE_CACHE_DISABLE
+
+#ifndef NOUVEAU_RING_TRACE
+#define NOUVEAU_RING_TRACE 0
+#else
+#undef NOUVEAU_RING_TRACE
+#define NOUVEAU_RING_TRACE 1
+#endif
+
+#define NV_READ(reg) *(volatile u_int32_t *)(nmesa->mmio + (reg))
+
+#define NV_FIFO_READ(reg) *(volatile u_int32_t *)(nmesa->fifo.mmio + (reg/4))
+#define NV_FIFO_WRITE(reg,value) *(volatile u_int32_t *)(nmesa->fifo.mmio + (reg/4)) = value;
+#define NV_FIFO_READ_GET() ((NV_FIFO_READ(NV03_FIFO_REGS_DMAGET) - nmesa->fifo.put_base) >> 2)
+#define NV_FIFO_WRITE_PUT(val) do { \
+	if (NOUVEAU_RING_TRACE) {\
+		printf("FIRE_RING : 0x%08x\n", nmesa->fifo.current << 2); \
+		fflush(stdout); \
+		sleep(1); \
+	} \
+	NV_FIFO_WRITE(NV03_FIFO_REGS_DMAPUT, ((val)<<2) + nmesa->fifo.put_base); \
+} while(0)
+
+/* 
+ * Ring/fifo interface
+ *
+ * - Begin a ring section with BEGIN_RING_SIZE (if you know the full size in advance)
+ * - Output stuff to the ring with either OUT_RINGp (outputs a raw mem chunk), OUT_RING (1 uint32_t) or OUT_RINGf (1 float)
+ * - RING_AVAILABLE returns the available fifo (in uint32_ts)
+ * - RING_AHEAD returns how much ahead of the last submission point we are
+ * - FIRE_RING fires whatever we have that wasn't fired before
+ * - WAIT_RING waits for size (in uint32_ts) to be available in the fifo
+ */
+
+/* Enable for ring debugging.  Prints out writes to the ring buffer
+ * but does not actually write to it.
+ */
+#ifdef NOUVEAU_RING_DEBUG
+
+#define OUT_RINGp(ptr,sz) do {                                                  \
+uint32_t* p=(uint32_t*)(ptr);							\
+int i; printf("OUT_RINGp: (size 0x%x dwords)\n",sz); for(i=0;i<sz;i++) printf(" 0x%08x   %f\n", *(p+i), *((float*)(p+i))); 	\
+}while(0)
+
+#define OUT_RING(n) do {                                                        \
+    printf("OUT_RINGn: 0x%08x (%s)\n", n, __func__);                            \
+}while(0)
+
+#define OUT_RINGf(n) do {                                                       \
+    printf("OUT_RINGf: %.04f (%s)\n", n, __func__);                             \
+}while(0)
+
+#else
+
+#define OUT_RINGp(ptr,sz) do{							\
+	if (NOUVEAU_RING_TRACE) { \
+		uint32_t* p=(uint32_t*)(ptr);							\
+		int i; printf("OUT_RINGp: (size 0x%x dwords) (%s)\n",sz, __func__); for(i=0;i<sz;i++) printf(" [0x%08x] 0x%08x   %f\n", (nmesa->fifo.current+i) << 2, *(p+i), *((float*)(p+i))); 	\
+	} \
+	memcpy(nmesa->fifo.buffer+nmesa->fifo.current,ptr,(sz)*4);		\
+	nmesa->fifo.current+=(sz);						\
+}while(0)
+
+#define OUT_RING(n) do {							\
+if (NOUVEAU_RING_TRACE) \
+    printf("OUT_RINGn: [0x%08x] 0x%08x (%s)\n", nmesa->fifo.current << 2, n, __func__);        \
+nmesa->fifo.buffer[nmesa->fifo.current++]=(n);					\
+}while(0)
+
+#define OUT_RINGf(n) do {							\
+if (NOUVEAU_RING_TRACE) \
+    printf("OUT_RINGf: [0x%08x] %.04f (%s)\n", nmesa->fifo.current << 2, n, __func__);        \
+*((float*)(nmesa->fifo.buffer+nmesa->fifo.current++))=(n);			\
+}while(0)
+
+#endif
+
+#define BEGIN_RING_SIZE(subchannel,tag,size) do {					\
+	nouveau_state_cache_flush(nmesa);						\
+	if (nmesa->fifo.free <= (size))							\
+		WAIT_RING(nmesa,(size));						\
+	OUT_RING( ((size)<<18) | ((subchannel) << 13) | (tag));				\
+	nmesa->fifo.free -= ((size) + 1);                                               \
+}while(0)
+
+extern void WAIT_RING(nouveauContextPtr nmesa,u_int32_t size);
+extern void nouveau_state_cache_flush(nouveauContextPtr nmesa);
+extern void nouveau_state_cache_init(nouveauContextPtr nmesa);
+
+#ifdef NOUVEAU_STATE_CACHE_DISABLE
+#define BEGIN_RING_CACHE(subc,tag,size) BEGIN_RING_SIZE((subc), (tag), (size))
+#define OUT_RING_CACHE(n) OUT_RING((n))
+#define OUT_RING_CACHEf(n) OUT_RINGf((n))
+#define OUT_RING_CACHEp(ptr, sz) OUT_RINGp((ptr), (sz))
+#else
+#define BEGIN_RING_CACHE(subchannel,tag,size) do {					\
+	nmesa->state_cache.dirty=1;	 						\
+	nmesa->state_cache.current_pos=((tag)/4);					\
+}while(0)
+
+#define OUT_RING_CACHE(n) do {									\
+	if (nmesa->state_cache.atoms[nmesa->state_cache.current_pos].value!=(n))	{	\
+		nmesa->state_cache.atoms[nmesa->state_cache.current_pos].dirty=1; 		\
+		nmesa->state_cache.hdirty[nmesa->state_cache.current_pos/NOUVEAU_STATE_CACHE_HIER_SIZE]=1; 		\
+		nmesa->state_cache.atoms[nmesa->state_cache.current_pos].value=(n);		\
+	}											\
+	nmesa->state_cache.current_pos++;							\
+}while(0)
+
+#define OUT_RING_CACHEf(n) do {									\
+	if ((*(float*)(&nmesa->state_cache.atoms[nmesa->state_cache.current_pos].value))!=(n)){	\
+		nmesa->state_cache.atoms[nmesa->state_cache.current_pos].dirty=1;	 	\
+		nmesa->state_cache.hdirty[nmesa->state_cache.current_pos/NOUVEAU_STATE_CACHE_HIER_SIZE]=1; 		\
+		(*(float*)(&nmesa->state_cache.atoms[nmesa->state_cache.current_pos].value))=(n);\
+	}											\
+	nmesa->state_cache.current_pos++;							\
+}while(0)
+
+#define OUT_RING_CACHEp(ptr,sz) do {							\
+uint32_t* p=(uint32_t*)(ptr);								\
+int i; for(i=0;i<sz;i++) OUT_RING_CACHE(*(p+i)); 					\
+}while(0)
+#endif
+
+#define RING_AVAILABLE() (nmesa->fifo.free-1)
+
+#define RING_AHEAD() ((nmesa->fifo.put<=nmesa->fifo.current)?(nmesa->fifo.current-nmesa->fifo.put):nmesa->fifo.max-nmesa->fifo.put+nmesa->fifo.current)
+
+#define FIRE_RING() do {                             \
+	if (nmesa->fifo.current!=nmesa->fifo.put) {  \
+		nmesa->fifo.put=nmesa->fifo.current; \
+		NV_FIFO_WRITE_PUT(nmesa->fifo.put);  \
+	}                                            \
+}while(0)
+
+extern void nouveauWaitForIdle(nouveauContextPtr nmesa);
+extern void nouveauWaitForIdleLocked(nouveauContextPtr nmesa);
+extern GLboolean nouveauFifoInit(nouveauContextPtr nmesa);
+
+#endif /* __NOUVEAU_FIFO_H__ */
+
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_lock.c b/src/mesa/drivers/dri/nouveau/nouveau_lock.c
new file mode 100644
index 00000000000..c119d14dd75
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_lock.c
@@ -0,0 +1,81 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+#include "nouveau_context.h"
+#include "nouveau_lock.h"
+
+#include "drirenderbuffer.h"
+#include "framebuffer.h"
+
+
+/* Update the hardware state.  This is called if another context has
+ * grabbed the hardware lock, which includes the X server.  This
+ * function also updates the driver's window state after the X server
+ * moves, resizes or restacks a window -- the change will be reflected
+ * in the drawable position and clip rects.  Since the X server grabs
+ * the hardware lock when it changes the window state, this routine will
+ * automatically be called after such a change.
+ */
+void nouveauGetLock( nouveauContextPtr nmesa, GLuint flags )
+{
+   __DRIdrawablePrivate *dPriv = nmesa->driDrawable;
+   __DRIscreenPrivate *sPriv = nmesa->driScreen;
+   drm_nouveau_sarea_t *sarea = nmesa->sarea;
+
+   drmGetLock( nmesa->driFd, nmesa->hHWContext, flags );
+
+   /* The window might have moved, so we might need to get new clip
+    * rects.
+    *
+    * NOTE: This releases and regrabs the hw lock to allow the X server
+    * to respond to the DRI protocol request for new drawable info.
+    * Since the hardware state depends on having the latest drawable
+    * clip rects, all state checking must be done _after_ this call.
+    */
+   DRI_VALIDATE_DRAWABLE_INFO( sPriv, dPriv );
+
+   /* If timestamps don't match, the window has been changed */
+   if (nmesa->lastStamp != dPriv->lastStamp) {
+      struct gl_framebuffer *fb = (struct gl_framebuffer *)dPriv->driverPrivate;
+
+      /* _mesa_resize_framebuffer will take care of calling the renderbuffer's
+       * AllocStorage function if we need more memory to hold it */
+      if (fb->Width != dPriv->w || fb->Height != dPriv->h) {
+	 _mesa_resize_framebuffer(nmesa->glCtx, fb, dPriv->w, dPriv->h);
+	 /* resize buffers, will call nouveau_window_moved */
+	 nouveau_build_framebuffer(nmesa->glCtx, fb);
+      } else {
+	 nouveau_window_moved(nmesa->glCtx);
+      }
+
+      nmesa->lastStamp = dPriv->lastStamp;
+   }
+
+   nmesa->numClipRects = dPriv->numClipRects;
+   nmesa->pClipRects = dPriv->pClipRects;
+
+}
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_lock.h b/src/mesa/drivers/dri/nouveau/nouveau_lock.h
new file mode 100644
index 00000000000..38bb0014258
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_lock.h
@@ -0,0 +1,69 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+#ifndef __NOUVEAU_LOCK_H__
+#define __NOUVEAU_LOCK_H__
+
+#include "nouveau_context.h"
+
+extern void nouveauGetLock( nouveauContextPtr nmesa, GLuint flags );
+
+/*
+ * !!! We may want to separate locks from locks with validation.  This
+ * could be used to improve performance for those things commands that
+ * do not do any drawing !!!
+ */
+
+/* Lock the hardware and validate our state.
+ */
+#define LOCK_HARDWARE( nmesa )						\
+   do {									\
+      char __ret = 0;							\
+      DEBUG_CHECK_LOCK();						\
+      DRM_CAS( nmesa->driHwLock, nmesa->hHWContext,			\
+	       (DRM_LOCK_HELD | nmesa->hHWContext), __ret );		\
+      if ( __ret )							\
+	 nouveauGetLock( nmesa, 0 );					\
+      DEBUG_LOCK();							\
+   } while (0)
+
+/* Unlock the hardware.
+ */
+#define UNLOCK_HARDWARE( nmesa )					\
+   do {									\
+      DRM_UNLOCK( nmesa->driFd,						\
+		  nmesa->driHwLock,					\
+		  nmesa->hHWContext );					\
+      DEBUG_RESET();							\
+   } while (0)
+
+#define DEBUG_LOCK()
+#define DEBUG_RESET()
+#define DEBUG_CHECK_LOCK()
+
+
+#endif /* __NOUVEAU_LOCK_H__ */
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_msg.h b/src/mesa/drivers/dri/nouveau/nouveau_msg.h
new file mode 100644
index 00000000000..5dea2189c72
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_msg.h
@@ -0,0 +1,67 @@
+/*
+Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
+Copyright 2006 Stephane Marchesin. All Rights Reserved
+
+The Weather Channel (TM) funded Tungsten Graphics to develop the
+initial release of the Radeon 8500 driver under the XFree86 license.
+This notice must be preserved.
+
+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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
+ *   Keith Whitwell <keith@tungstengraphics.com>
+ *   Nicolai Haehnle <prefect_@gmx.net>
+ */
+
+
+#ifndef __NOUVEAU_MSG_H__
+#define __NOUVEAU_MSG_H__
+
+#define WARN_ONCE(a, ...) do {\
+	        static int warn##__LINE__=1;\
+	        if(warn##__LINE__){\
+			                fprintf(stderr, "*********************************WARN_ONCE*********************************\n");\
+			                fprintf(stderr, "File %s function %s line %d\n", __FILE__, __FUNCTION__, __LINE__);\
+			                fprintf(stderr,  a, ## __VA_ARGS__);\
+			                fprintf(stderr, "***************************************************************************\n");\
+			                warn##__LINE__=0;\
+			                } \
+	        }while(0)
+
+#define MESSAGE(a, ...) do{\
+	                fprintf(stderr, "************************************INFO***********************************\n");\
+	                fprintf(stderr, "File %s function %s line %d\n", __FILE__, __FUNCTION__, __LINE__); \
+	                fprintf(stderr,  a, ## __VA_ARGS__);\
+	                fprintf(stderr, "***************************************************************************\n");\
+	        }while(0)
+
+#define FATAL(a, ...) do{\
+	                fprintf(stderr, "***********************************FATAL***********************************\n");\
+	                fprintf(stderr, "File %s function %s line %d\n", __FILE__, __FUNCTION__, __LINE__); \
+	                fprintf(stderr,  a, ## __VA_ARGS__);\
+	                fprintf(stderr, "***************************************************************************\n");\
+	        }while(0)
+
+#endif /* __NOUVEAU_MSG_H__ */
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_object.c b/src/mesa/drivers/dri/nouveau/nouveau_object.c
new file mode 100644
index 00000000000..302009c8b17
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_object.c
@@ -0,0 +1,101 @@
+
+#include "nouveau_fifo.h"
+#include "nouveau_object.h"
+#include "nouveau_reg.h"
+
+
+GLboolean nouveauCreateContextObject(nouveauContextPtr nmesa, int handle, int class, uint32_t flags, uint32_t dma_in, uint32_t dma_out, uint32_t dma_notifier)
+{
+	drm_nouveau_object_init_t cto;
+	int ret;
+
+	cto.handle = handle;
+	cto.class  = class;
+	cto.flags  = flags;
+	cto.dma0= dma_in;
+	cto.dma1= dma_out;
+	cto.dma_notifier = dma_notifier;
+	ret = drmCommandWrite(nmesa->driFd, DRM_NOUVEAU_OBJECT_INIT, &cto, sizeof(cto));
+
+	return ret == 0;
+}
+
+GLboolean nouveauCreateDmaObject(nouveauContextPtr nmesa,
+      				 uint32_t handle,
+				 uint32_t offset,
+				 uint32_t size,
+				 int	  target,
+				 int	  access)
+{
+	drm_nouveau_dma_object_init_t dma;
+	int ret;
+
+	dma.handle = handle;
+	dma.target = target;
+	dma.access = access;
+	dma.offset = offset;
+	dma.size   = size;
+	ret = drmCommandWriteRead(nmesa->driFd, DRM_NOUVEAU_DMA_OBJECT_INIT,
+				  &dma, sizeof(dma));
+	return ret == 0;
+}
+
+void nouveauObjectOnSubchannel(nouveauContextPtr nmesa, int subchannel, int handle)
+{
+	BEGIN_RING_SIZE(subchannel, 0, 1);
+	OUT_RING(handle);
+}
+
+void nouveauObjectInit(nouveauContextPtr nmesa)
+{
+#ifdef NOUVEAU_RING_DEBUG
+	return;
+#endif
+
+/* We need to know vram size.. and AGP size (and even if the card is AGP..) */
+	nouveauCreateDmaObject( nmesa, NvDmaFB,
+				0, (256*1024*1024),
+				0 /*NV_DMA_TARGET_FB*/, 0 /*NV_DMA_ACCESS_RW*/);
+	nouveauCreateDmaObject( nmesa, NvDmaAGP,
+	      			nmesa->agp_phys, (128*1024*1024),
+				3 /* AGP */, 0 /* RW */);
+
+	nouveauCreateContextObject(nmesa, Nv3D, nmesa->screen->card->class_3d,
+	      			   0, 0, 0, 0);
+	if (nmesa->screen->card->type>=NV_10) {
+		nouveauCreateContextObject(nmesa, NvCtxSurf2D, NV10_CONTEXT_SURFACES_2D,
+	      			   0, 0, 0, 0);
+		nouveauCreateContextObject(nmesa, NvImageBlit, NV10_IMAGE_BLIT,
+	      			   NV_DMA_CONTEXT_FLAGS_PATCH_SRCCOPY, 0, 0, 0);
+	} else {
+		nouveauCreateContextObject(nmesa, NvCtxSurf2D, NV04_CONTEXT_SURFACES_2D,
+	      			   0, 0, 0, 0);
+		nouveauCreateContextObject(nmesa, NvCtxSurf3D, NV04_CONTEXT_SURFACES_3D,
+	      			   0, 0, 0, 0);
+		nouveauCreateContextObject(nmesa, NvImageBlit, NV_IMAGE_BLIT,
+	      			   NV_DMA_CONTEXT_FLAGS_PATCH_SRCCOPY, 0, 0, 0);
+	}
+	nouveauCreateContextObject(nmesa, NvMemFormat,
+	      			   NV_MEMORY_TO_MEMORY_FORMAT,
+	      			   0, 0, 0, 0);
+
+#ifdef ALLOW_MULTI_SUBCHANNEL
+	nouveauObjectOnSubchannel(nmesa, NvSubCtxSurf2D, NvCtxSurf2D);
+	BEGIN_RING_SIZE(NvSubCtxSurf2D, NV10_CONTEXT_SURFACES_2D_SET_DMA_IN_MEMORY0, 2);
+	OUT_RING(NvDmaFB);
+	OUT_RING(NvDmaFB);
+
+	nouveauObjectOnSubchannel(nmesa, NvSubImageBlit, NvImageBlit);
+	BEGIN_RING_SIZE(NvSubImageBlit, NV10_IMAGE_BLIT_SET_CONTEXT_SURFACES_2D, 1);
+	OUT_RING(NvCtxSurf2D);
+	BEGIN_RING_SIZE(NvSubImageBlit, NV10_IMAGE_BLIT_SET_OPERATION, 1);
+	OUT_RING(3); /* SRCCOPY */
+
+	nouveauObjectOnSubchannel(nmesa, NvSubMemFormat, NvMemFormat);
+#endif
+
+	nouveauObjectOnSubchannel(nmesa, NvSub3D, Nv3D);
+}
+
+
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_object.h b/src/mesa/drivers/dri/nouveau/nouveau_object.h
new file mode 100644
index 00000000000..daad281029f
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_object.h
@@ -0,0 +1,44 @@
+#ifndef __NOUVEAU_OBJECT_H__
+#define __NOUVEAU_OBJECT_H__
+
+#include "nouveau_context.h"
+
+#define ALLOW_MULTI_SUBCHANNEL
+
+void nouveauObjectInit(nouveauContextPtr nmesa);
+
+enum DMAObjects {
+	Nv3D                    = 0x80000019,
+	NvCtxSurf2D		= 0x80000020,
+	NvImageBlit		= 0x80000021,
+	NvMemFormat		= 0x80000022,
+	NvCtxSurf3D		= 0x80000023,
+	NvDmaFB			= 0xD0FB0001,
+	NvDmaAGP		= 0xD0AA0001,
+	NvSyncNotify		= 0xD0000001,
+	NvQueryNotify		= 0xD0000002
+};
+
+enum DMASubchannel {
+	NvSubCtxSurf2D	= 0,
+	NvSubImageBlit	= 1,
+	NvSubMemFormat	= 2,
+	NvSubCtxSurf3D	= 3,
+	NvSub3D		= 7,
+};
+
+extern void nouveauObjectOnSubchannel(nouveauContextPtr nmesa, int subchannel, int handle);
+
+extern GLboolean nouveauCreateContextObject(nouveauContextPtr nmesa,
+      					    int handle, int class,
+					    uint32_t flags,
+					    uint32_t dma_in,
+					    uint32_t dma_out,
+					    uint32_t dma_notifier);
+extern GLboolean nouveauCreateDmaObject(nouveauContextPtr nmesa,
+      					uint32_t handle,
+					uint32_t offset,
+					uint32_t size,
+					int      target,
+					int      access);
+#endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_query.c b/src/mesa/drivers/dri/nouveau/nouveau_query.c
new file mode 100644
index 00000000000..de3f5b0378b
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_query.c
@@ -0,0 +1,198 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ *
+ * 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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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.
+ *
+ */
+
+/* GL_ARB_occlusion_query support for NV20/30/40 */
+
+#include "mtypes.h"
+
+#include "nouveau_fifo.h"
+#include "nouveau_msg.h"
+#include "nouveau_object.h"
+#include "nouveau_reg.h"
+#include "nouveau_sync.h"
+#include "nouveau_query.h"
+
+static struct gl_query_object *
+nouveauNewQueryObject(GLcontext *ctx, GLuint id)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveau_query_object *nq;
+	int i;
+
+	for (i=0; i<nmesa->query_object_max; i++)
+		if (nmesa->query_alloc[i] == GL_FALSE)
+			break;
+	if (i==nmesa->query_object_max)
+		return NULL;
+
+	nq = CALLOC_STRUCT(nouveau_query_object_t);
+	if (nq) {
+		nq->notifier_id	= i;
+
+		nq->mesa.Id     = id;
+		nq->mesa.Result = 0;
+		nq->mesa.Active = GL_FALSE;
+		nq->mesa.Ready  = GL_TRUE;
+	}
+
+	return (struct gl_query_object *)nq;
+}
+
+static void
+nouveauBeginQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveau_query_object *nq = (nouveau_query_object *)q;
+
+	nouveau_notifier_reset(nmesa->queryNotifier, nq->notifier_id);
+
+	switch (nmesa->screen->card->type) {
+	case NV_20:
+		BEGIN_RING_CACHE(NvSub3D, 0x17c8, 1);
+		OUT_RING_CACHE  (1);
+		BEGIN_RING_CACHE(NvSub3D, 0x17cc, 1);
+		OUT_RING_CACHE  (1);
+		break;
+	case NV_30:
+	case NV_40:
+	case NV_44:
+		/* I don't think this is OCC_QUERY enable, but it *is* needed to make
+		 * the SET_OBJECT7 notifier block work with STORE_RESULT.
+		 *
+		 * Also, this appears to reset the pixel pass counter */
+		BEGIN_RING_SIZE(NvSub3D,
+				 NV30_TCL_PRIMITIVE_3D_OCC_QUERY_OR_COLOR_BUFF_ENABLE,
+				 1);
+		OUT_RING        (1);
+		/* Probably OCC_QUERY_ENABLE */
+		BEGIN_RING_CACHE(NvSub3D, 0x17cc, 1);
+		OUT_RING_CACHE  (1);
+		break;
+	default:
+		WARN_ONCE("no support for this card\n");
+		break;
+	}
+}
+
+static void
+nouveauUpdateQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveau_query_object *nq = (nouveau_query_object *)q;
+	int status;
+
+	status = nouveau_notifier_status(nmesa->queryNotifier,
+					 nq->notifier_id);
+
+	q->Ready = (status == NV_NOTIFY_STATE_STATUS_COMPLETED);
+	if (q->Ready)
+		q->Result = nouveau_notifier_return_val(nmesa->queryNotifier,
+							nq->notifier_id);
+}
+
+static void
+nouveauWaitQueryResult(GLcontext *ctx, GLenum target, struct gl_query_object *q)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveau_query_object *nq = (nouveau_query_object *)q;
+
+	nouveau_notifier_wait_status(nmesa->queryNotifier, nq->notifier_id,
+				     NV_NOTIFY_STATE_STATUS_COMPLETED, 0);
+	nouveauUpdateQuery(ctx, target, q);
+}
+
+static void
+nouveauEndQuery(GLcontext *ctx, GLenum target, struct gl_query_object *q)
+{
+	nouveau_query_object *nq = (nouveau_query_object *)q;
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	switch (nmesa->screen->card->type) {
+	case NV_20:
+		BEGIN_RING_SIZE(NvSub3D, 0x17d0, 1);
+		OUT_RING       (0x01000000 | nq->notifier_id*32);
+		break;
+	case NV_30:
+	case NV_40:
+	case NV_44:
+		BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_STORE_RESULT, 1);
+		OUT_RING       (0x01000000 | nq->notifier_id*32);
+		break;
+	default:
+		WARN_ONCE("no support for this card\n");
+		break;
+	}
+	FIRE_RING();
+
+	/*XXX: wait for query to complete, mesa doesn't give the driver
+	 *     an interface to query the status of a query object so 
+	 *     this has to stall the channel.
+	 */
+	nouveauWaitQueryResult(ctx, target, q);
+
+	BEGIN_RING_CACHE(NvSub3D, 0x17cc, 1);
+	OUT_RING_CACHE  (0);
+}
+
+void
+nouveauQueryInitFuncs(GLcontext *ctx)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (nmesa->screen->card->type < NV_20)
+		return;
+
+	nmesa->query_object_max = (0x4000 / 32);
+	nmesa->queryNotifier =
+		nouveau_notifier_new(ctx, NvQueryNotify,
+					  nmesa->query_object_max);
+	nmesa->query_alloc = calloc(nmesa->query_object_max, sizeof(GLboolean));
+
+	switch (nmesa->screen->card->type) {
+	case NV_20:
+		BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SET_OBJECT8, 1);
+		OUT_RING_CACHE  (NvQueryNotify);
+		break;
+	case NV_30:
+	case NV_40:
+	case NV_44:
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_OBJECT7, 1);
+		OUT_RING_CACHE  (NvQueryNotify);
+		break;
+	default:
+		break;
+	};
+
+	ctx->Driver.NewQueryObject	= nouveauNewQueryObject;
+	ctx->Driver.BeginQuery		= nouveauBeginQuery;
+	ctx->Driver.EndQuery		= nouveauEndQuery;
+#if 0
+	ctx->Driver.UpdateQuery		= nouveauUpdateQuery;
+	ctx->Driver.WaitQueryResult	= nouveauWaitQueryResult;
+#endif
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_query.h b/src/mesa/drivers/dri/nouveau/nouveau_query.h
new file mode 100644
index 00000000000..3ded41417e1
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_query.h
@@ -0,0 +1,38 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ *
+ * 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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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.
+ *
+ */
+
+#ifndef __NOUVEAU_QUERY_H__
+#define __NOUVEAU_QUERY_H__
+
+typedef struct nouveau_query_object_t {
+	struct gl_query_object mesa;
+
+	int notifier_id;
+} nouveau_query_object;
+
+extern void nouveauQueryInitFuncs(GLcontext *ctx);
+#endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_reg.h b/src/mesa/drivers/dri/nouveau/nouveau_reg.h
new file mode 100644
index 00000000000..8758b538c85
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_reg.h
@@ -0,0 +1,1505 @@
+/*
+   Autogenerated file, do not edit !
+
+**************************************************************************
+
+   Copyright (C) 2006 :
+   Dmitry Baryshkov,
+   Laurent Carlier,
+   Matthieu Castet,
+   Dawid Gajownik,
+   Jeremy Kolb,
+   Stephane Loeuillet,
+   Patrice Mandin,
+   Stephane Marchesin,
+   Serge Martin,
+   Sylvain Munaut,
+   Ben Skeggs,
+   Erik Waling,
+   koala_br,
+   sturmflut.
+
+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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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.
+
+**************************************************************************
+
+   Created from objects.c rev. 1.398
+*/
+
+#ifndef _NOUVEAU_REG_H
+#define _NOUVEAU_REG_H
+
+/****************************************** 
+Object NV01_CONTEXT_CLIP_RECTANGLE used on: NV03 NV04 NV10 NV15 NV20 NV40 G70
+*/
+#define			NV01_CONTEXT_CLIP_RECTANGLE				0x00000019
+#	define		NV01_CONTEXT_CLIP_RECTANGLE_SET_POINT			0x00000300	/* Parameters: x y */
+#	define		NV01_CONTEXT_CLIP_RECTANGLE_SET_SIZE			0x00000304	/* Parameters: width height */
+
+/****************************************** 
+Object NV_MEMORY_TO_MEMORY_FORMAT used on: NV04 NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV_MEMORY_TO_MEMORY_FORMAT				0x00000039
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_NOP				0x00000100
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_NOTIFY			0x00000104
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY			0x00000180
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_OBJECT_IN			0x00000184
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_OBJECT_OUT			0x00000188
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_IN			0x0000030c
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_OFFSET_OUT			0x00000310
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_PITCH_IN			0x00000314
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_PITCH_OUT			0x00000318
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_LINE_LENGTH_IN		0x0000031c
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_LINE_COUNT			0x00000320
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_FORMAT			0x00000324	/* Parameters: src_inc dst_inc */
+#	define		NV_MEMORY_TO_MEMORY_FORMAT_BUF_NOTIFY			0x00000328
+
+/****************************************** 
+Object NV03_PRIMITIVE_RASTER_OP used on: NV03 NV04 NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV03_PRIMITIVE_RASTER_OP				0x00000043
+#	define		NV03_PRIMITIVE_RASTER_OP_NOTIFY				0x00000100
+#	define		NV03_PRIMITIVE_RASTER_OP_DMA_NOTIFY			0x00000180
+#	define		NV03_PRIMITIVE_RASTER_OP_LOGIC_OP			0x00000300	/* Parameters: logic_op */
+
+/****************************************** 
+Object NV04_GDI_RECTANGLE_TEXT used on: NV04 NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV04_GDI_RECTANGLE_TEXT					0x0000004a
+#	define		NV04_GDI_RECTANGLE_TEXT_SET_DMA_NOTIFY			0x00000180
+#	define		NV04_GDI_RECTANGLE_TEXT_PATTERN				0x00000188
+#	define		NV04_GDI_RECTANGLE_TEXT_ROP5				0x0000018c
+#	define		NV04_GDI_RECTANGLE_TEXT_SURFACE				0x00000198
+#	define		NV04_GDI_RECTANGLE_TEXT_OPERATION			0x000002fc
+#	define		NV04_GDI_RECTANGLE_TEXT_FORMAT				0x00000300
+#	define		NV04_GDI_RECTANGLE_TEXT_BLOCK_LEVEL1_TL			0x000005f4	/* Parameters: left top */
+#	define		NV04_GDI_RECTANGLE_TEXT_BLOCK_LEVEL1_BR			0x000005f8	/* Parameters: right bottom */
+#	define		NV04_GDI_RECTANGLE_TEXT_FILL_VALUE			0x000005fc
+#	define		NV04_GDI_RECTANGLE_TEXT_BLOCK_LEVEL2_TL			0x00000600	/* Parameters: left top */
+#	define		NV04_GDI_RECTANGLE_TEXT_BLOCK_LEVEL2_BR			0x00000604	/* Parameters: right bottom */
+
+/****************************************** 
+Object NV04_SWIZZLED_SURFACE used on: NV04 NV10 NV15 
+*/
+#define			NV04_SWIZZLED_SURFACE					0x00000052
+#	define		NV04_SWIZZLED_SURFACE_DMA_NOTIFY			0x00000180
+#	define		NV04_SWIZZLED_SURFACE_DMA_IMAGE				0x00000184
+#	define		NV04_SWIZZLED_SURFACE_FORMAT				0x00000300	/* Parameters: log2(height) log2(width) color */
+#	define		NV04_SWIZZLED_SURFACE_OFFSET				0x00000304
+
+/****************************************** 
+Object NV04_CONTEXT_SURFACES_3D used on: NV04 
+*/
+#define			NV04_CONTEXT_SURFACES_3D				0x00000053
+#	define		NV04_CONTEXT_SURFACES_3D_DMA_NOTIFY			0x00000180
+#	define		NV04_CONTEXT_SURFACES_3D_DMA_COLOR			0x00000184
+#	define		NV04_CONTEXT_SURFACES_3D_DMA_ZETA			0x00000188
+#	define		NV04_CONTEXT_SURFACES_3D_CLIP_HORIZONTAL		0x000002f8	/* Parameters: x width */
+#	define		NV04_CONTEXT_SURFACES_3D_CLIP_VERTICAL			0x000002fc	/* Parameters: y height */
+#	define		NV04_CONTEXT_SURFACES_3D_FORMAT				0x00000300	/* Parameters: color type width height */
+#	define		NV04_CONTEXT_SURFACES_3D_CLIP_SIZE			0x00000304	/* Parameters: width height */
+#	define		NV04_CONTEXT_SURFACES_3D_PITCH				0x00000308	/* Parameters: color zeta */
+#	define		NV04_CONTEXT_SURFACES_3D_OFFSET_COLOR			0x0000030c
+#	define		NV04_CONTEXT_SURFACES_3D_OFFSET_ZETA			0x00000310
+
+/****************************************** 
+Object NV04_DX5_TEXTURED_TRIANGLE used on: NV04 
+*/
+#define			NV04_DX5_TEXTURED_TRIANGLE				0x00000054
+#	define		NV04_DX5_TEXTURED_TRIANGLE_NOP				0x00000100
+#	define		NV04_DX5_TEXTURED_TRIANGLE_NOTIFY			0x00000104
+#	define		NV04_DX5_TEXTURED_TRIANGLE_DMA_NOTIFY			0x00000180
+#	define		NV04_DX5_TEXTURED_TRIANGLE_DMA_1			0x00000184
+#	define		NV04_DX5_TEXTURED_TRIANGLE_DMA_2			0x00000188
+#	define		NV04_DX5_TEXTURED_TRIANGLE_SURFACE			0x0000018c
+#	define		NV04_DX5_TEXTURED_TRIANGLE_COLOR_KEY			0x00000300
+#	define		NV04_DX5_TEXTURED_TRIANGLE_TEXTURE_OFFSET		0x00000304
+#	define		NV04_DX5_TEXTURED_TRIANGLE_TEXTURE_FORMAT		0x00000308	/* Parameters: color mipmaps log(u) log(v) wrap_s wrap_t */
+#	define		NV04_DX5_TEXTURED_TRIANGLE_TEXTURE_FILTER		0x0000030c	/* Parameters: magfilter minfilter lodbias */
+#	define		NV04_DX5_TEXTURED_TRIANGLE_BLEND			0x00000310	/* Parameters: texture benable dst src */
+#	define		NV04_DX5_TEXTURED_TRIANGLE_CONTROL			0x00000314	/* Parameters: alpharef alphafunc alphaenable zenable zwrite zfunc cullmode */
+#	define		NV04_DX5_TEXTURED_TRIANGLE_FOG_COLOR			0x00000318
+#	define		NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX( d)		(0x00000400 + d * 0x0020)
+#	define		NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SY( d)		(0x00000404 + d * 0x0020)
+#	define		NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SZ( d)		(0x00000408 + d * 0x0020)
+#	define		NV04_DX5_TEXTURED_TRIANGLE_INV_W( d)			(0x0000040c + d * 0x0020)
+#	define		NV04_DX5_TEXTURED_TRIANGLE_COLOR( d)			(0x00000410 + d * 0x0020)
+#	define		NV04_DX5_TEXTURED_TRIANGLE_SPECULAR( d)			(0x00000414 + d * 0x0020)
+#	define		NV04_DX5_TEXTURED_TRIANGLE_TEXTURE_S( d)		(0x00000418 + d * 0x0020)
+#	define		NV04_DX5_TEXTURED_TRIANGLE_TEXTURE_T( d)		(0x0000041c + d * 0x0020)
+#	define		NV04_DX5_TEXTURED_TRIANGLE_DRAW				0x00000600	/* Parameters: v0 v1 v2 v3 v4 v5 */
+
+/****************************************** 
+Object NV04_DX6_MULTITEX_TRIANGLE used on: NV04 NV10 NV15 
+*/
+#define			NV04_DX6_MULTITEX_TRIANGLE				0x00000055
+#	define		NV04_DX6_MULTITEX_TRIANGLE_NOP				0x00000100
+#	define		NV04_DX6_MULTITEX_TRIANGLE_NOTIFY			0x00000104
+#	define		NV04_DX6_MULTITEX_TRIANGLE_DMA_NOTIFY			0x00000180
+#	define		NV04_DX6_MULTITEX_TRIANGLE_DMA_1			0x00000184
+#	define		NV04_DX6_MULTITEX_TRIANGLE_DMA_2			0x00000188
+#	define		NV04_DX6_MULTITEX_TRIANGLE_SURFACE			0x0000018c
+#	define		NV04_DX6_MULTITEX_TRIANGLE_OFFSET0			0x00000308
+#	define		NV04_DX6_MULTITEX_TRIANGLE_OFFSET1			0x0000030c
+#	define		NV04_DX6_MULTITEX_TRIANGLE_FORMAT0			0x00000310	/* Parameters: color mipmaps log(u) log(v) wrap_s wrap_t */
+#	define		NV04_DX6_MULTITEX_TRIANGLE_FORMAT1			0x00000314	/* Parameters: color mipmaps log(u) log(v) wrap_s wrap_t */
+#	define		NV04_DX6_MULTITEX_TRIANGLE_FILTER0			0x00000318	/* Parameters: magfilter minfilter lodbias */
+#	define		NV04_DX6_MULTITEX_TRIANGLE_FILTER1			0x0000031c	/* Parameters: magfilter minfilter lodbias */
+#	define		NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_ALPHA		0x00000320
+#	define		NV04_DX6_MULTITEX_TRIANGLE_COMBINE_0_COLOR		0x00000324
+#	define		NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_ALPHA		0x0000032c
+#	define		NV04_DX6_MULTITEX_TRIANGLE_COMBINE_1_COLOR		0x00000330
+#	define		NV04_DX6_MULTITEX_TRIANGLE_COMBINE_FACTOR		0x00000334
+#	define		NV04_DX6_MULTITEX_TRIANGLE_BLEND			0x00000338	/* Parameters: benable dst src */
+#	define		NV04_DX6_MULTITEX_TRIANGLE_CONTROL0			0x0000033c	/* Parameters: red_write green_write blue_write alpha_write alpha_write stencil_write alpharef alphafunc alphaenable zenable zwrite zfunc cullmode */
+#	define		NV04_DX6_MULTITEX_TRIANGLE_CONTROL1			0x00000340	/* Parameters: stencil_enable stencil_mask_write stencil_mask_read stencilref stencilfunc */
+#	define		NV04_DX6_MULTITEX_TRIANGLE_CONTROL2			0x00000344	/* Parameters: stencil_fail stencil_zfail stencil_zpass */
+#	define		NV04_DX6_MULTITEX_TRIANGLE_FOG_COLOR			0x00000348
+#	define		NV04_DX6_MULTITEX_TRIANGLE_TLVERTEX_SX( d)		(0x00000400 + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_TLVERTEX_SY( d)		(0x00000404 + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_TLVERTEX_SZ( d)		(0x00000408 + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_INV_W( d)			(0x0000040c + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_COLOR( d)			(0x00000410 + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_SPECULAR( d)			(0x00000414 + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_TEXTURE0_S( d)		(0x00000418 + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_TEXTURE0_T( d)		(0x0000041c + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_TEXTURE1_S( d)		(0x00000420 + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_TEXTURE1_T( d)		(0x00000424 + d * 0x0028)
+#	define		NV04_DX6_MULTITEX_TRIANGLE_DRAW				0x00000540	/* Parameters: v0 v1 v2 v3 v4 v5 */
+
+/****************************************** 
+Object NV04_COLOR_KEY used on: NV04 NV10 NV15 NV20 NV40 
+*/
+#define			NV04_COLOR_KEY						0x00000057
+#	define		NV04_COLOR_KEY_SET_DMA_NOTIFY				0x00000180
+#	define		NV04_COLOR_KEY_FORMAT					0x00000300
+#	define		NV04_COLOR_KEY_VALUE					0x00000304
+
+/****************************************** 
+Object NV04_SOLID_LINE used on: NV04 
+*/
+#define			NV04_SOLID_LINE						0x0000005c
+#	define		NV04_SOLID_LINE_CLIP_RECTANGLE				0x00000184
+#	define		NV04_SOLID_LINE_PATTERN					0x00000188
+#	define		NV04_SOLID_LINE_ROP					0x0000018c
+#	define		NV04_SOLID_LINE_SURFACE					0x00000198
+#	define		NV04_SOLID_LINE_OPERATION				0x000002fc
+#	define		NV04_SOLID_LINE_COLOR_FORMAT				0x00000300
+#	define		NV04_SOLID_LINE_COLOR_VALUE				0x00000304
+#	define		NV04_SOLID_LINE_START					0x00000400	/* Parameters: x y */
+#	define		NV04_SOLID_LINE_END					0x00000400	/* Parameters: x y */
+
+/****************************************** 
+Object NV04_UNK005E used on: NV04 
+*/
+#define			NV04_UNK005E						0x0000005e
+#	define		NV04_UNK005E_SET_SURFACE				0x00000198
+#	define		NV04_UNK005E_UNK02fc					0x000002fc
+#	define		NV04_UNK005E_UNK0300					0x00000300
+#	define		NV04_UNK005E_COUNTER					0x00000304
+
+/****************************************** 
+Object NV05_SCALED_IMAGE_FROM_MEMORY used on: NV04 
+*/
+#define			NV05_SCALED_IMAGE_FROM_MEMORY				0x00000063
+#	define		NV05_SCALED_IMAGE_FROM_MEMORY_SURFACE			0x00000198
+#	define		NV05_SCALED_IMAGE_FROM_MEMORY_COLOR_CONVERSION		0x000002fc
+#	define		NV05_SCALED_IMAGE_FROM_MEMORY_OPERATION			0x00000304
+
+/****************************************** 
+Object NV04_SCALED_IMAGE_FROM_MEMORY used on: NV04 
+*/
+#define			NV04_SCALED_IMAGE_FROM_MEMORY				0x00000077
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_DMA_NOTIFY		0x00000180
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_DMA_IMAGE			0x00000184
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_SURFACE			0x00000198
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_COLOR_FORMAT		0x00000300
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_OPERATION			0x00000304
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_POS			0x00000308	/* Parameters: x y */
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE			0x0000030c	/* Parameters: width height */
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_OUT_POS			0x00000310	/* Parameters: x y */
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_OUT_SIZE			0x00000314	/* Parameters: width height */
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_DU_DX			0x00000318	/* Parameters: int frac*0x100000 */
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_DV_DY			0x0000031c	/* Parameters: int frac*0x100000 */
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_SIZE			0x00000400	/* Parameters: width height */
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_FORMAT			0x00000404	/* Parameters: pitch origin filter */
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_OFFSET			0x00000408
+#	define		NV04_SCALED_IMAGE_FROM_MEMORY_POINT			0x0000040c	/* Parameters: u_int u_frac*0x10 v_int v_frac*0x10 */
+
+/****************************************** 
+Object NV_IMAGE_FROM_CPU used on: NV04 
+*/
+#define			NV_IMAGE_FROM_CPU					0x00000061
+#	define		NV_IMAGE_FROM_CPU_DMA_NOTIFY				0x00000180
+#	define		NV_IMAGE_FROM_CPU_CLIP_RECTANGLE			0x00000188
+#	define		NV_IMAGE_FROM_CPU_PATTERN				0x0000018c
+#	define		NV_IMAGE_FROM_CPU_ROP					0x00000190
+#	define		NV_IMAGE_FROM_CPU_SURFACE				0x0000019c
+#	define		NV_IMAGE_FROM_CPU_OPERATION				0x000002fc
+#	define		NV_IMAGE_FROM_CPU_FORMAT				0x00000300
+
+/****************************************** 
+Object NV05_IMAGE_FROM_CPU used on: NV04 
+*/
+#define			NV05_IMAGE_FROM_CPU					0x00000065
+#	define		NV05_IMAGE_FROM_CPU_DMA_NOTIFY				0x00000180
+#	define		NV05_IMAGE_FROM_CPU_CLIP_RECTANGLE			0x00000188
+#	define		NV05_IMAGE_FROM_CPU_PATTERN				0x0000018c
+#	define		NV05_IMAGE_FROM_CPU_ROP					0x00000190
+#	define		NV05_IMAGE_FROM_CPU_SURFACE				0x0000019c
+#	define		NV05_IMAGE_FROM_CPU_OPERATION				0x000002fc
+#	define		NV05_IMAGE_FROM_CPU_FORMAT				0x00000300
+#	define		NV05_IMAGE_FROM_CPU_POINT				0x00000304	/* Parameters: x y */
+#	define		NV05_IMAGE_FROM_CPU_SIZE_OUT				0x00000308	/* Parameters: x y */
+#	define		NV05_IMAGE_FROM_CPU_SIZE_IN				0x0000030c	/* Parameters: x y */
+#	define		NV05_IMAGE_FROM_CPU_COLOR( d)				(0x00000400 + d * 0x0004)
+
+/****************************************** 
+Object NV_IMAGE_BLIT used on: NV04 NV10 NV15 NV20 NV40 
+*/
+#define			NV_IMAGE_BLIT						0x0000005f
+#	define		NV_IMAGE_BLIT_DMA_NOTIFY				0x00000180
+#	define		NV_IMAGE_BLIT_COLOR_KEY					0x00000184
+#	define		NV_IMAGE_BLIT_CLIP_RECTANGLE				0x00000188
+#	define		NV_IMAGE_BLIT_PATTERN					0x0000018c
+#	define		NV_IMAGE_BLIT_ROP5					0x00000190
+#	define		NV_IMAGE_BLIT_SURFACE					0x0000019c
+#	define		NV_IMAGE_BLIT_OPERATION					0x000002fc
+#	define		NV_IMAGE_BLIT_POINT_IN					0x00000300	/* Parameters: x y */
+#	define		NV_IMAGE_BLIT_POINT_OUT					0x00000304	/* Parameters: x y */
+#	define		NV_IMAGE_BLIT_SIZE					0x00000308	/* Parameters: width height */
+
+/****************************************** 
+Object NV10_TCL_PRIMITIVE_3D used on: NV10 
+*/
+#define			NV10_TCL_PRIMITIVE_3D					0x00000056
+
+/****************************************** 
+Object NV17_TCL_PRIMITIVE_3D used on: NV15 
+*/
+#define			NV17_TCL_PRIMITIVE_3D					0x00000099
+
+/****************************************** 
+Object NV11_TCL_PRIMITIVE_3D used on: NV15 
+*/
+#define			NV11_TCL_PRIMITIVE_3D					0x00000096
+#	define		NV10_TCL_PRIMITIVE_3D_NOP				0x00000100
+#	define		NV10_TCL_PRIMITIVE_3D_NOTIFY				0x00000104
+#	define		NV10_TCL_PRIMITIVE_3D_SET_DMA_NOTIFY			0x00000180
+#	define		NV10_TCL_PRIMITIVE_3D_SET_DMA_IN_MEMORY0		0x00000184
+#	define		NV10_TCL_PRIMITIVE_3D_SET_DMA_IN_MEMORY1		0x00000188
+#	define		NV10_TCL_PRIMITIVE_3D_SET_DISPLAY_LIST			0x0000018c
+#	define		NV10_TCL_PRIMITIVE_3D_SET_DMA_IN_MEMORY2		0x00000194
+#	define		NV10_TCL_PRIMITIVE_3D_SET_DMA_IN_MEMORY3		0x00000198
+#	define		NV17_TCL_PRIMITIVE_3D_SET_DMA_IN_MEMORY4		0x000001ac
+#	define		NV17_TCL_PRIMITIVE_3D_SET_DMA_IN_MEMORY5		0x000001b0
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_HORIZ			0x00000200	/* Parameters: width x */
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_VERT			0x00000204	/* Parameters: height y */
+#	define		NV10_TCL_PRIMITIVE_3D_BUFFER_FORMAT			0x00000208	/* Parameters: type color */
+#	define		NV10_TCL_PRIMITIVE_3D_BUFFER_PITCH			0x0000020c	/* Parameters: depth/stencil buffer pitch color buffer pitch */
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_OFFSET			0x00000210
+#	define		NV10_TCL_PRIMITIVE_3D_DEPTH_OFFSET			0x00000214
+#	define		NV10_TCL_PRIMITIVE_3D_TX_OFFSET(d)			(0x00000218 + d * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_TX_FORMAT(d)			(0x00000220 + d * 0x0004)	/* Parameters: wrap_t wrap_s log2(height) log2(width) lod npot format cube_map */
+#	define		NV10_TCL_PRIMITIVE_3D_TX_ENABLE(d)			(0x00000228 + d * 0x0004)	/* Parameters: enable anisotropy */
+#	define		NV10_TCL_PRIMITIVE_3D_TX_NPOT_PITCH(d)			(0x00000230 + d * 0x0004)	/* Parameters: pitch */
+#	define		NV10_TCL_PRIMITIVE_3D_TX_NPOT_SIZE(d)			(0x00000240 + d * 0x0004)	/* Parameters: width height */
+#	define		NV10_TCL_PRIMITIVE_3D_TX_FILTER(d)			(0x00000248 + d * 0x0004)	/* Parameters: mag_filter min_filter */
+#	define		NV10_TCL_PRIMITIVE_3D_TX_PALETTE_OFFSET(d)		(0x00000250 + d * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_TX_MATRIX_ENABLE(d)		(0x000003e0 + d * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_TX_MATRIX(x,y)			(0x00000540 + y * 0x0010 + x * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_RC_IN_ALPHA(d)			(0x00000260 + d * 0x0004)	/* Parameters: vara_mapping vara_component_usage vara_input varb_mapping varb_component_usage varb_input varc_mapping varc_component_usage varc_input vard_mapping vard_component_usage vard_input */
+#	define		NV10_TCL_PRIMITIVE_3D_RC_IN_RGB(d)			(0x00000268 + d * 0x0004)	/* Parameters: vara_mapping vara_component_usage vara_input varb_mapping varb_component_usage varb_input varc_mapping varc_component_usage varc_input vard_mapping vard_component_usage vard_input */
+#	define		NV10_TCL_PRIMITIVE_3D_RC_OUT_ALPHA(d)			(0x00000278 + d * 0x0004)	/* Parameters: scale bias mux_sum ab_dot_product cd_dot_product sum_output ab_output cd_output */
+#	define		NV10_TCL_PRIMITIVE_3D_RC_OUT_RGB(d)			(0x00000280 + d * 0x0004)	/* Parameters: rc1_tx_units_enabled rc1_rc_enabled scale bias mux_sum ab_dot_product cd_dot_product sum_output ab_output cd_output */
+#	define		NV10_TCL_PRIMITIVE_3D_RC_COLOR0				0x00000270	/* Parameters: a r g b */
+#	define		NV10_TCL_PRIMITIVE_3D_RC_COLOR1				0x00000274	/* Parameters: a r g b */
+#	define		NV10_TCL_PRIMITIVE_3D_RC_FINAL0				0x00000288	/* Parameters: vara_mapping vara_component_usage vara_input varb_mapping varb_component_usage varb_input varc_mapping varc_component_usage varc_input vard_mapping vard_component_usage vard_input */
+#	define		NV10_TCL_PRIMITIVE_3D_RC_FINAL1				0x0000028c	/* Parameters: vare_mapping vare_component_usage vare_input varf_mapping varf_component_usage varf_input varg_mapping varg_component_usage varg_input color_sum_clamp */
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_MODEL			0x00000294	/* Parameters: local_viewer color_control */
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_MATERIAL_ENABLE		0x00000298	/* Parameters: specular diffuse ambient emission */
+#	define		NV10_TCL_PRIMITIVE_3D_FOG_MODE				0x0000029c
+#	define		NV10_TCL_PRIMITIVE_3D_FOG_COORD_DIST			0x000002a0
+#	define		NV10_TCL_PRIMITIVE_3D_FOG_ENABLE			0x000002a4
+#	define		NV10_TCL_PRIMITIVE_3D_FOG_COLOR				0x000002a8	/* Parameters: a b g r */
+#	define		NV17_TCL_PRIMITIVE_3D_COLOR_MASK_ENABLE			0x000002bc
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_HORIZ(d)		(0x000002c0 + d * 0x0004)	/* Parameters: x2 x1 */
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_VERT(d)		(0x000002e0 + d * 0x0004)	/* Parameters: y2 y1 */
+#	define		NV10_TCL_PRIMITIVE_3D_ALPHA_FUNC_ENABLE			0x00000300
+#	define		NV10_TCL_PRIMITIVE_3D_BLEND_FUNC_ENABLE			0x00000304
+#	define		NV10_TCL_PRIMITIVE_3D_CULL_FACE_ENABLE			0x00000308
+#	define		NV10_TCL_PRIMITIVE_3D_DEPTH_TEST_ENABLE			0x0000030c
+#	define		NV10_TCL_PRIMITIVE_3D_DITHER_ENABLE			0x00000310
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHTING_ENABLE			0x00000314
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_PARAMETERS_ENABLE		0x00000318
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_SMOOTH_ENABLE		0x0000031c
+#	define		NV10_TCL_PRIMITIVE_3D_LINE_SMOOTH_ENABLE		0x00000320
+#	define		NV10_TCL_PRIMITIVE_3D_POLYGON_SMOOTH_ENABLE		0x00000324
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_WEIGHT_ENABLE		0x00000328
+#	define		NV10_TCL_PRIMITIVE_3D_STENCIL_ENABLE			0x0000032c
+#	define		NV10_TCL_PRIMITIVE_3D_POLYGON_OFFSET_POINT_ENABLE	0x00000330
+#	define		NV10_TCL_PRIMITIVE_3D_POLYGON_OFFSET_LINE_ENABLE	0x00000334
+#	define		NV10_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FILL_ENABLE	0x00000338
+#	define		NV10_TCL_PRIMITIVE_3D_ALPHA_FUNC_FUNC			0x0000033c
+#	define		NV10_TCL_PRIMITIVE_3D_ALPHA_FUNC_REF			0x00000340
+#	define		NV10_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC			0x00000344
+#	define		NV10_TCL_PRIMITIVE_3D_BLEND_FUNC_DST			0x00000348
+#	define		NV10_TCL_PRIMITIVE_3D_BLEND_COLOR			0x0000034c	/* Parameters: a r g b */
+#	define		NV10_TCL_PRIMITIVE_3D_BLEND_EQUATION			0x00000350
+#	define		NV10_TCL_PRIMITIVE_3D_DEPTH_FUNC			0x00000354
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_MASK			0x00000358	/* Parameters: a r g b */
+#	define		NV10_TCL_PRIMITIVE_3D_DEPTH_WRITE_ENABLE		0x0000035c
+#	define		NV10_TCL_PRIMITIVE_3D_STENCIL_MASK			0x00000360
+#	define		NV10_TCL_PRIMITIVE_3D_STENCIL_FUNC_FUNC			0x00000364
+#	define		NV10_TCL_PRIMITIVE_3D_STENCIL_FUNC_REF			0x00000368
+#	define		NV10_TCL_PRIMITIVE_3D_STENCIL_FUNC_MASK			0x0000036c
+#	define		NV10_TCL_PRIMITIVE_3D_STENCIL_OP_FAIL			0x00000370
+#	define		NV10_TCL_PRIMITIVE_3D_STENCIL_OP_ZFAIL			0x00000374
+#	define		NV10_TCL_PRIMITIVE_3D_STENCIL_OP_ZPASS			0x00000378
+#	define		NV10_TCL_PRIMITIVE_3D_SHADE_MODEL			0x0000037c
+#	define		NV10_TCL_PRIMITIVE_3D_LINE_WIDTH			0x00000380
+#	define		NV10_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FACTOR		0x00000384
+#	define		NV10_TCL_PRIMITIVE_3D_POLYGON_OFFSET_UNITS		0x00000388
+#	define		NV10_TCL_PRIMITIVE_3D_POLYGON_MODE_FRONT		0x0000038c
+#	define		NV10_TCL_PRIMITIVE_3D_POLYGON_MODE_BACK			0x00000390
+#	define		NV10_TCL_PRIMITIVE_3D_DEPTH_RANGE_NEAR			0x00000394
+#	define		NV10_TCL_PRIMITIVE_3D_DEPTH_RANGE_FAR			0x00000398
+#	define		NV10_TCL_PRIMITIVE_3D_CULL_FACE				0x0000039c
+#	define		NV10_TCL_PRIMITIVE_3D_FRONT_FACE			0x000003a0
+#	define		NV10_TCL_PRIMITIVE_3D_NORMALIZE_ENABLE			0x000003a4
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_MATERIAL_R			0x000003a8
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_MATERIAL_G			0x000003ac
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_MATERIAL_B			0x000003b0
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_MATERIAL_A			0x000003b4
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_CONTROL			0x000003b8	/* Parameters: color_control */
+#	define		NV10_TCL_PRIMITIVE_3D_ENABLED_LIGHTS			0x000003bc	/* Parameters: light 7 light 6 light 5 light 4 light 3 light 2 light 1 light 0 */
+#	define		NV10_TCL_PRIMITIVE_3D_CLIP_PLANE_ENABLE( d)		(0x000003c0 + d * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_VIEW_MATRIX_ENABLE		0x000003e8	/* Parameters: projection modelview0 modelview1 */
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_SIZE			0x000003ec
+#	define		NV10_TCL_PRIMITIVE_3D_MODELVIEW0_MATRIX( d)		(0x00000400 + d * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_MODELVIEW1_MATRIX( d)		(0x00000440 + d * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_INVERSE_MODELVIEW0_MATRIX( d)	(0x00000480 + d * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_INVERSE_MODELVIEW1_MATRIX( d)	(0x000004c0 + d * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_PROJECTION_MATRIX( d)		(0x00000500 + d * 0x0004)
+#	define		NV10_TCL_PRIMITIVE_3D_CLIP_PLANE_A(d)			(0x00000600 + d * 0x0010)
+#	define		NV10_TCL_PRIMITIVE_3D_CLIP_PLANE_B(d)			(0x00000604 + d * 0x0010)
+#	define		NV10_TCL_PRIMITIVE_3D_CLIP_PLANE_C(d)			(0x00000608 + d * 0x0010)
+#	define		NV10_TCL_PRIMITIVE_3D_CLIP_PLANE_D(d)			(0x0000060c + d * 0x0010)
+#	define		NV10_TCL_PRIMITIVE_3D_FOG_EQUATION_CONSTANT		0x00000680
+#	define		NV10_TCL_PRIMITIVE_3D_FOG_EQUATION_LINEAR		0x00000684
+#	define		NV10_TCL_PRIMITIVE_3D_FOG_EQUATION_QUADRATIC		0x00000688
+#	define		NV10_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_A	0x000006a0
+#	define		NV10_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_B	0x000006a4
+#	define		NV10_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_C	0x000006a8
+#	define		NV10_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_D	0x000006ac
+#	define		NV10_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_E	0x000006b0
+#	define		NV10_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_F	0x000006b4
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R	0x000006c4
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G	0x000006c8
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B	0x000006cc
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_SCALE_X			0x000006e8
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_SCALE_Y			0x000006ec
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_SCALE_Z			0x000006f0
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_SCALE_W			0x000006f4
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_PARAMETER_A			0x000006f8
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_PARAMETER_B			0x000006fc
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_PARAMETER_C			0x00000700
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_PARAMETER_D			0x00000704
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_PARAMETER_E			0x00000708
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_PARAMETER_F			0x0000070c
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_PARAMETER_G			0x00000710
+#	define		NV10_TCL_PRIMITIVE_3D_POINT_PARAMETER_H			0x00000714
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R(d)	(0x00000800 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_G(d)	(0x00000804 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_B(d)	(0x00000808 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R(d)	(0x0000080c + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_G(d)	(0x00000810 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_B(d)	(0x00000814 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R(d)	(0x00000818 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_G(d)	(0x0000081c + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_B(d)	(0x00000820 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_HALF_VECTOR_X(d)		(0x00000828 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_HALF_VECTOR_Y(d)		(0x0000082c + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_HALF_VECTOR_Z(d)		(0x00000830 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_DIRECTION_X(d)		(0x00000834 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_DIRECTION_Y(d)		(0x00000838 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_DIRECTION_Z(d)		(0x0000083c + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_A(d)		(0x00000840 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_EXPONENT(d)		(0x00000844 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_B(d)		(0x00000848 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_X(d)		(0x0000084c + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_Y(d)		(0x00000850 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_Z(d)		(0x00000854 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_C(d)		(0x00000858 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_POSITION_X(d)		(0x0000085c + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_POSITION_Y(d)		(0x00000860 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_POSITION_Z(d)		(0x00000864 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_CONSTANT_ATTENUATION(d)	(0x00000868 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_LINEAR_ATTENUATION(d)	(0x0000086c + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_LIGHT_QUADRATIC_ATTENUATION(d)	(0x00000870 + d * 0x0080)
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_POS_3F_X			0x00000c00
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_POS_3F_Y			0x00000c04
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_POS_3F_Z			0x00000c08
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_POS_4F_X			0x00000c18
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_POS_4F_Y			0x00000c1c
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_POS_4F_Z			0x00000c20
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_POS_4F_W			0x00000c24
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_NOR_3F_X			0x00000c30
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_NOR_3F_Y			0x00000c34
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_NOR_3F_Z			0x00000c38
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_NOR_3I_XY			0x00000c40	/* Parameters: y x */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_NOR_3I_Z			0x00000c44	/* Parameters: z */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL_4F_R			0x00000c50
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL_4F_G			0x00000c54
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL_4F_B			0x00000c58
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL_4F_A			0x00000c5c
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL_3F_R			0x00000c60
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL_3F_G			0x00000c64
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL_3F_B			0x00000c68
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL_4I			0x00000c6c	/* Parameters: a b g r */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL2_3F_R			0x00000c80
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL2_3F_G			0x00000c84
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL2_3F_B			0x00000c88
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_COL2_3I			0x00000c8c	/* Parameters: a b g r */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX0_2F_S			0x00000c90
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX0_2F_T			0x00000c94
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX0_2I			0x00000c98	/* Parameters: t s */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_S			0x00000ca0
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_T			0x00000ca4
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_R			0x00000ca8
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_Q			0x00000cac
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX0_4I_ST			0x00000cb0	/* Parameters: t s */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX0_4I_RQ			0x00000cb4	/* Parameters: q r */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX1_2F_S			0x00000cb8
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX1_2F_T			0x00000cbc
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX1_2I			0x00000cc0	/* Parameters: t s */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_S			0x00000cc8
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_T			0x00000ccc
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_R			0x00000cd0
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_Q			0x00000cd4
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX1_4I_ST			0x00000cd8	/* Parameters: t s */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_TX1_4I_RQ			0x00000cdc	/* Parameters: q r */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_FOG_1F			0x00000ce0
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_WGH_1F			0x00000ce4
+#	define		NV10_TCL_PRIMITIVE_3D_EDGEFLAG_ENABLE			0x00000cec
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ATTR( d)			(0x00000d04 + d * 0x0008)	/* Parameters: stride fields type */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_VALIDATE		0x00000cf0
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_OFFSET_POS		0x00000d00
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_FORMAT_POS		0x00000d04	/* Parameters: stride fields type */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_OFFSET_COL		0x00000d08
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_FORMAT_COL		0x00000d0c	/* Parameters: stride fields type */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_OFFSET_COL2		0x00000d10
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_FORMAT_COL2		0x00000d14	/* Parameters: stride fields type */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_OFFSET_TX0		0x00000d18
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_FORMAT_TX0		0x00000d1c	/* Parameters: stride fields type */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_OFFSET_TX1		0x00000d20
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_FORMAT_TX1		0x00000d24	/* Parameters: stride fields type */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_OFFSET_NOR		0x00000d28
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_FORMAT_NOR		0x00000d2c	/* Parameters: stride fields type */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_OFFSET_WGH		0x00000d30
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_FORMAT_WGH		0x00000d34	/* Parameters: stride fields type */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_OFFSET_FOG		0x00000d38
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_FORMAT_FOG		0x00000d3c	/* Parameters: stride fields type */
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_ENABLE		0x00000d40
+#	define		NV10_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_OP			0x00000d44
+#	define		NV17_TCL_PRIMITIVE_3D_LMA_DEPTH_BUFFER_PITCH		0x00000d5c	/* Parameters: pitch */
+#	define		NV17_TCL_PRIMITIVE_3D_LMA_DEPTH_BUFFER_OFFSET		0x00000d60
+#	define		NV17_TCL_PRIMITIVE_3D_LMA_DEPTH_FILL_VALUE		0x00000d68
+#	define		NV17_TCL_PRIMITIVE_3D_LMA_DEPTH_CLEAR_ENABLE		0x00000d6c
+#	define		NV10_TCL_PRIMITIVE_3D_BEGIN_END				0x00000dfc
+#	define		NV10_TCL_PRIMITIVE_3D_INDEX_DATA			0x00000e00	/* Parameters: index1 index0 */
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_BUFFER_BEGIN_END		0x000013fc
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_BUFFER_DRAW_ARRAYS		0x00001400	/* Parameters: count-1 first */
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_X			0x00001638
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_Y			0x0000163c
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_Z			0x00001640
+#	define		NV10_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_W			0x00001644
+#	define		NV17_TCL_PRIMITIVE_3D_LMA_DEPTH_ENABLE			0x00001658
+#	define		NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_DATA			0x00001800
+
+/****************************************** 
+Object NV10_IMAGE_FROM_CPU used on: NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV10_IMAGE_FROM_CPU					0x0000008a
+#	define		NV10_IMAGE_FROM_CPU_SET_DMA_NOTIFY			0x00000180
+#	define		NV10_IMAGE_FROM_CPU_SET_CONTEXT_CLIP_RECTANGLE		0x00000188
+#	define		NV10_IMAGE_FROM_CPU_SET_IMAGE_PATTERN			0x0000018c
+#	define		NV10_IMAGE_FROM_CPU_SET_RASTER_OP			0x00000190
+#	define		NV10_IMAGE_FROM_CPU_SET_CONTEXT_SURFACES_2D		0x0000019c
+#	define		NV10_IMAGE_FROM_CPU_OPERATION				0x000002fc
+#	define		NV10_IMAGE_FROM_CPU_FORMAT				0x00000300
+#	define		NV10_IMAGE_FROM_CPU_POINT				0x00000304	/* Parameters: x y */
+#	define		NV10_IMAGE_FROM_CPU_SIZE_OUT				0x00000308	/* Parameters: width height */
+#	define		NV10_IMAGE_FROM_CPU_SIZE_IN				0x0000030c	/* Parameters: width height */
+#	define		NV10_IMAGE_FROM_CPU_HLINE				0x00000400
+
+/****************************************** 
+Object NV10_PRIMITIVE_2D used on: NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV10_PRIMITIVE_2D					0x0000007b
+#	define		NV10_PRIMITIVE_2D_SET_DMA_NOTIFY			0x00000180
+#	define		NV10_PRIMITIVE_2D_SET_SURFACE				0x00000184
+#	define		NV10_PRIMITIVE_2D_SET_FORMAT				0x00000300
+#	define		NV10_PRIMITIVE_2D_SET_POINT				0x00000304	/* Parameters: x y */
+#	define		NV10_PRIMITIVE_2D_SET_SIZE				0x00000308	/* Parameters: width height */
+#	define		NV10_PRIMITIVE_2D_SET_CLIP_HORIZ			0x0000030c	/* Parameters: width x */
+#	define		NV10_PRIMITIVE_2D_SET_CLIP_VERT				0x00000310	/* Parameters: height y */
+#	define		NV10_PRIMITIVE_2D_SET_DATA( d)				(0x00000400 + d * 0x0004)
+
+/****************************************** 
+Object NV10_IMAGE_BLIT used on: NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV10_IMAGE_BLIT						0x0000009f
+#	define		NV10_IMAGE_BLIT_NOP					0x00000100
+#	define		NV10_IMAGE_BLIT_NOTIFY					0x00000104
+#	define		NV10_IMAGE_BLIT_SET_DMA_NOTIFY				0x00000180
+#	define		NV10_IMAGE_BLIT_SET_CONTEXT_CLIP_RECTANGLE		0x00000188
+#	define		NV10_IMAGE_BLIT_SET_IMAGE_PATTERN			0x0000018c
+#	define		NV10_IMAGE_BLIT_SET_RASTER_OP				0x00000190
+#	define		NV10_IMAGE_BLIT_SET_CONTEXT_SURFACES_2D			0x0000019c
+#	define		NV10_IMAGE_BLIT_SET_OPERATION				0x000002fc
+#	define		NV10_IMAGE_BLIT_SET_POINT				0x00000300	/* Parameters: x y */
+#	define		NV10_IMAGE_BLIT_SET_PITCH				0x00000304	/* Parameters: skip */
+#	define		NV10_IMAGE_BLIT_SET_SIZE				0x00000308	/* Parameters: width height */
+
+/****************************************** 
+Object NV10_VIDEO_DISPLAY used on: NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV10_VIDEO_DISPLAY					0x0000007c
+#	define		NV10_VIDEO_DISPLAY_COUNTER				0x00000050
+#	define		NV10_VIDEO_DISPLAY_SET_DMA_FROM_MEMORY			0x00000180
+#	define		NV10_VIDEO_DISPLAY_SET_DMA_IN_MEMORY0			0x00000184
+#	define		NV10_VIDEO_DISPLAY_SET_DMA_IN_MEMORY1			0x00000188
+#	define		NV10_VIDEO_DISPLAY_SET_OBJECT3				0x0000019c
+#	define		NV10_VIDEO_DISPLAY_SIZE					0x000002f8	/* Parameters: height width */
+#	define		NV10_VIDEO_DISPLAY_OFFSET				0x00000300
+
+/****************************************** 
+Object NV10_UNK0072 used on: NV10 NV15 NV20 NV40 G70
+*/
+#define			NV10_UNK0072						0x00000072
+#	define		NV10_UNK0072_COUNTER					0x00000050
+#	define		NV40_UNK0072_SET_OBJECT					0x00000060
+#	define		NV10_UNK0072_SET_DMA_NOTIFY				0x00000180
+
+/****************************************** 
+Object NV10_SCALED_IMAGE_FROM_MEMORY used on: NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV10_SCALED_IMAGE_FROM_MEMORY				0x00000089
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_COUNTER			0x00000050
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_SET_DMA_IN_MEMORY		0x00000184
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_SET_RASTER_OP		0x0000018c
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_SET_IMAGE_PATTERN		0x00000188
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_SET_SURFACE		0x00000198
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_OPERATION			0x00000304
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_CLIP_POS			0x00000308	/* Parameters: x y */
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_CLIP_SIZE			0x0000030c	/* Parameters: width height */
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_OUT_POS			0x00000310	/* Parameters: x y */
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_OUT_SIZE			0x00000314	/* Parameters: width height */
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_SIZE			0x00000400	/* Parameters: width height */
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_FORMAT			0x00000404	/* Parameters: pitch */
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_OFFSET			0x00000408
+#	define		NV10_SCALED_IMAGE_FROM_MEMORY_POINT			0x0000040c	/* Parameters: u_int u_frac*0x10 v_int v_frac*0x10 */
+
+/****************************************** 
+Object NV10_CONTEXT_SURFACES_2D used on: NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV10_CONTEXT_SURFACES_2D				0x00000062
+#	define		NV10_CONTEXT_SURFACES_2D_SET_DMA_NOTIFY			0x00000180
+#	define		NV10_CONTEXT_SURFACES_2D_SET_DMA_IN_MEMORY0		0x00000184
+#	define		NV10_CONTEXT_SURFACES_2D_SET_DMA_IN_MEMORY1		0x00000188
+#	define		NV10_CONTEXT_SURFACES_2D_FORMAT				0x00000300	/* Parameters: color type width height */
+#	define		NV10_CONTEXT_SURFACES_2D_PITCH				0x00000304	/* Parameters: src dst */
+#	define		NV10_CONTEXT_SURFACES_2D_OFFSET_SRC			0x00000308
+#	define		NV10_CONTEXT_SURFACES_2D_OFFSET_DST			0x0000030c
+
+/****************************************** 
+Object NV04_CONTEXT_SURFACES_2D used on: NV04 NV10 NV15 
+*/
+#define			NV04_CONTEXT_SURFACES_2D				0x00000042
+#	define		NV04_CONTEXT_SURFACES_2D_NOTIFY				0x00000104
+#	define		NV04_CONTEXT_SURFACES_2D_SET_DMA_NOTIFY			0x00000180
+#	define		NV04_CONTEXT_SURFACES_2D_SET_DMA_IMAGE_SRC		0x00000184
+#	define		NV04_CONTEXT_SURFACES_2D_SET_DMA_IMAGE_DST		0x00000188
+#	define		NV04_CONTEXT_SURFACES_2D_FORMAT				0x00000300
+#	define		NV04_CONTEXT_SURFACES_2D_PITCH				0x00000304	/* Parameters: src dst */
+#	define		NV04_CONTEXT_SURFACES_2D_OFFSET_SRC			0x00000308
+#	define		NV04_CONTEXT_SURFACES_2D_OFFSET_DST			0x0000030c
+
+/****************************************** 
+Object NV04_IMAGE_PATTERN used on: NV04 NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV04_IMAGE_PATTERN					0x00000044
+#	define		NV04_IMAGE_PATTERN_COLOR_FORMAT				0x00000300
+#	define		NV04_IMAGE_PATTERN_MONO_FORMAT				0x00000304
+#	define		NV04_IMAGE_PATTERN_SELECT				0x0000030c
+#	define		NV04_IMAGE_PATTERN_MONOCHROME_SHAPE			0x00000308
+#	define		NV04_IMAGE_PATTERN_MONOCHROME_COLOR0			0x00000310
+#	define		NV04_IMAGE_PATTERN_MONOCHROME_COLOR1			0x00000314
+#	define		NV04_IMAGE_PATTERN_MONOCHROME_PATTERN0			0x00000318
+#	define		NV04_IMAGE_PATTERN_MONOCHROME_PATTERN1			0x0000031c
+
+/****************************************** 
+Object NV20_SWIZZLED_SURFACE used on: NV20 NV30 NV40 G70
+*/
+#define			NV20_SWIZZLED_SURFACE					0x0000009e
+#	define		NV20_SWIZZLED_SURFACE_SET_OBJECT0			0x00000180
+#	define		NV20_SWIZZLED_SURFACE_SET_OBJECT1			0x00000184
+#	define		NV20_SWIZZLED_SURFACE_FORMAT				0x00000300	/* Parameters: log2(height) log2(width) color */
+#	define		NV20_SWIZZLED_SURFACE_OFFSET				0x00000304
+
+/****************************************** 
+Object NV20_TCL_PRIMITIVE_3D used on: NV20 
+*/
+#define			NV20_TCL_PRIMITIVE_3D					0x00000097
+#	define		NV20_TCL_PRIMITIVE_3D_NOP				0x00000100
+#	define		NV20_TCL_PRIMITIVE_3D_NOTIFY				0x00000104
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT0			0x00000180
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT1			0x00000184
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT2			0x00000188
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT3			0x00000194
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT4			0x00000198
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT5			0x0000019c
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT6			0x000001a0
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT7			0x000001a4
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT8			0x000001a8
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT9			0x000001ac
+#	define		NV20_TCL_PRIMITIVE_3D_SET_OBJECT10			0x000001b0
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_HORIZ			0x00000200	/* Parameters: width x */
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_VERT			0x00000204	/* Parameters: height y */
+#	define		NV20_TCL_PRIMITIVE_3D_BUFFER_FORMAT			0x00000208	/* Parameters: type color */
+#	define		NV20_TCL_PRIMITIVE_3D_BUFFER_PITCH			0x0000020c	/* Parameters: depth/stencil buffer pitch color buffer pitch */
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_OFFSET			0x00000210
+#	define		NV20_TCL_PRIMITIVE_3D_DEPTH_OFFSET			0x00000214
+#	define		NV20_TCL_PRIMITIVE_3D_LMA_DEPTH_BUFFER_PITCH		0x0000022c	/* Parameters: pitch */
+#	define		NV20_TCL_PRIMITIVE_3D_LMA_DEPTH_BUFFER_OFFSET		0x00000230
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_CONTROL			0x00000294
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MATERIAL_CONTROL		0x00000298	/* Parameters: back_specular back_ambient back_diffuse back_emission front_specular front_ambient front_diffuse front_emission */
+#	define		NV20_TCL_PRIMITIVE_3D_FOG_MODE				0x0000029c
+#	define		NV20_TCL_PRIMITIVE_3D_FOG_COORD_DIST			0x000002a0
+#	define		NV20_TCL_PRIMITIVE_3D_FOG_ENABLE			0x000002a4
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_HORIZ(d)		(0x000002c0 + d * 0x0004)	/* Parameters: x2 x1 */
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_VERT(d)		(0x000002e0 + d * 0x0004)	/* Parameters: y2 y1 */
+#	define		NV20_TCL_PRIMITIVE_3D_ALPHA_FUNC_ENABLE			0x00000300
+#	define		NV20_TCL_PRIMITIVE_3D_BLEND_FUNC_ENABLE			0x00000304
+#	define		NV20_TCL_PRIMITIVE_3D_CULL_FACE_ENABLE			0x00000308
+#	define		NV20_TCL_PRIMITIVE_3D_DEPTH_TEST_ENABLE			0x0000030c
+#	define		NV20_TCL_PRIMITIVE_3D_DITHER_ENABLE			0x00000310
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHTING_ENABLE			0x00000314
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_SMOOTH_ENABLE		0x0000031c
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_PARAMETERS_ENABLE		0x00000318
+#	define		NV20_TCL_PRIMITIVE_3D_LINE_SMOOTH_ENABLE		0x00000320
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_SMOOTH_ENABLE		0x00000324
+#	define		NV20_TCL_PRIMITIVE_3D_STENCIL_ENABLE			0x0000032c
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_OFFSET_POINT_ENABLE	0x00000330
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_OFFSET_LINE_ENABLE	0x00000334
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FILL_ENABLE	0x00000338
+#	define		NV20_TCL_PRIMITIVE_3D_ALPHA_FUNC_FUNC			0x0000033c
+#	define		NV20_TCL_PRIMITIVE_3D_ALPHA_FUNC_REF			0x00000340
+#	define		NV20_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC			0x00000344
+#	define		NV20_TCL_PRIMITIVE_3D_BLEND_FUNC_DST			0x00000348
+#	define		NV20_TCL_PRIMITIVE_3D_BLEND_COLOR			0x0000034c	/* Parameters: a r g b */
+#	define		NV20_TCL_PRIMITIVE_3D_BLEND_EQUATION			0x00000350
+#	define		NV20_TCL_PRIMITIVE_3D_DEPTH_FUNC			0x00000354
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MASK			0x00000358	/* Parameters: a r g b */
+#	define		NV20_TCL_PRIMITIVE_3D_DEPTH_WRITE_ENABLE		0x0000035c
+#	define		NV20_TCL_PRIMITIVE_3D_STENCIL_MASK			0x00000360
+#	define		NV20_TCL_PRIMITIVE_3D_STENCIL_FUNC_FUNC			0x00000364
+#	define		NV20_TCL_PRIMITIVE_3D_STENCIL_FUNC_REF			0x00000368
+#	define		NV20_TCL_PRIMITIVE_3D_STENCIL_FUNC_MASK			0x0000036c
+#	define		NV20_TCL_PRIMITIVE_3D_STENCIL_OP_FAIL			0x00000370
+#	define		NV20_TCL_PRIMITIVE_3D_STENCIL_OP_ZFAIL			0x00000374
+#	define		NV20_TCL_PRIMITIVE_3D_STENCIL_OP_ZPASS			0x00000378
+#	define		NV20_TCL_PRIMITIVE_3D_SHADE_MODEL			0x0000037c
+#	define		NV20_TCL_PRIMITIVE_3D_LINE_WIDTH			0x00000380
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FACTOR		0x00000384
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_OFFSET_UNITS		0x00000388
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_MODE_FRONT		0x0000038c
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_MODE_BACK			0x00000390
+#	define		NV20_TCL_PRIMITIVE_3D_DEPTH_RANGE_NEAR			0x00000394
+#	define		NV20_TCL_PRIMITIVE_3D_DEPTH_RANGE_FAR			0x00000398
+#	define		NV20_TCL_PRIMITIVE_3D_CULL_FACE				0x0000039c
+#	define		NV20_TCL_PRIMITIVE_3D_FRONT_FACE			0x000003a0
+#	define		NV20_TCL_PRIMITIVE_3D_NORMALIZE_ENABLE			0x000003a4
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MATERIAL_FRONT_R		0x000003a8
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MATERIAL_FRONT_G		0x000003ac
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MATERIAL_FRONT_B		0x000003b0
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MATERIAL_FRONT_A		0x000003b4
+#	define		NV20_TCL_PRIMITIVE_3D_SEPARATE_SPECULAR_ENABLE		0x000003b8
+#	define		NV20_TCL_PRIMITIVE_3D_ENABLED_LIGHTS			0x000003bc	/* Parameters: light 7 light 6 light 5 light 4 light 3 light 2 light 1 light 0 */
+#	define		NV20_TCL_PRIMITIVE_3D_CLIP_PLANE_ENABLE(d)		(0x000003c0 + d * 0x0004)
+#	define		NV20_TCL_PRIMITIVE_3D_TX_MATRIX_ENABLE(d)		(0x00000420 + d * 0x0004)
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_SIZE			0x0000043c
+#	define		NV20_TCL_PRIMITIVE_3D_MODELVIEW_MATRIX( d)		(0x00000480 + d * 0x0004)
+#	define		NV20_TCL_PRIMITIVE_3D_INVERSE_MODELVIEW_MATRIX( d)	(0x00000580 + d * 0x0004)
+#	define		NV20_TCL_PRIMITIVE_3D_PROJECTION_MATRIX( d)		(0x00000680 + d * 0x0004)
+#	define		NV20_TCL_PRIMITIVE_3D_TX_MATRIX(x,y)			(0x000006c0 + y * 0x0010 + x * 0x0004)
+#	define		NV20_TCL_PRIMITIVE_3D_CLIP_PLANE_A(d)			(0x00000840 + d * 0x0010)
+#	define		NV20_TCL_PRIMITIVE_3D_CLIP_PLANE_B(d)			(0x00000844 + d * 0x0010)
+#	define		NV20_TCL_PRIMITIVE_3D_CLIP_PLANE_C(d)			(0x00000848 + d * 0x0010)
+#	define		NV20_TCL_PRIMITIVE_3D_CLIP_PLANE_D(d)			(0x0000084c + d * 0x0010)
+#	define		NV20_TCL_PRIMITIVE_3D_FOG_EQUATION_CONSTANT		0x000009c0
+#	define		NV20_TCL_PRIMITIVE_3D_FOG_EQUATION_LINEAR		0x000009c4
+#	define		NV20_TCL_PRIMITIVE_3D_FOG_EQUATION_QUADRATIC		0x000009c8
+#	define		NV20_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_A	0x000009e0
+#	define		NV20_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_B	0x000009e4
+#	define		NV20_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_C	0x000009e8
+#	define		NV20_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_D	0x000009ec
+#	define		NV20_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_E	0x000009f0
+#	define		NV20_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_F	0x000009f4
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_SPRITE			0x00000a1c	/* Parameters: coord_replace r_mode enable */
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_OX			0x00000a20
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_OY			0x00000a24
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_DEPTH_AVG_S		0x00000a28
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_UNKNOWN_A		0x00000a2c
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_PARAMETER_A			0x00000a30
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_PARAMETER_B			0x00000a34
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_PARAMETER_C			0x00000a38
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_PARAMETER_D			0x00000a3c
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_PARAMETER_E			0x00000a40
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_PARAMETER_F			0x00000a44
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_PARAMETER_G			0x00000a48
+#	define		NV20_TCL_PRIMITIVE_3D_POINT_PARAMETER_H			0x00000a4c
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_PX_DIV2			0x00000af0
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_PY_DIV2			0x00000af4
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_DEPTH_HALF_S		0x00000af8
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_UNKNOWN_B		0x00000afc
+#	define		NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_INST0			0x00000b00
+#	define		NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_INST1			0x00000b04
+#	define		NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_INST2			0x00000b08
+#	define		NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_INST3			0x00000b0c
+#	define		NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_X			0x00000b80
+#	define		NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_Y			0x00000b84
+#	define		NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_Z			0x00000b88
+#	define		NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_W			0x00000b8c
+#	define		NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_ID		0x00001ea4
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R	0x00000a10
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G	0x00000a14
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B	0x00000a18
+#	define		NV20_TCL_PRIMITIVE_3D_TX_OFFSET(d)			(0x00001b00 + d * 0x0040)
+#	define		NV20_TCL_PRIMITIVE_3D_TX_FORMAT(d)			(0x00001b04 + d * 0x0040)	/* Parameters: log2(height) log2(width) lod format cube_map */
+#	define		NV20_TCL_PRIMITIVE_3D_TX_WRAP(d)			(0x00001b08 + d * 0x0040)	/* Parameters: wrap_s wrap_t wrap_r */
+#	define		NV20_TCL_PRIMITIVE_3D_TX_ENABLE(d)			(0x00001b0c + d * 0x0040)	/* Parameters: enable anisotropy */
+#	define		NV20_TCL_PRIMITIVE_3D_TX_NPOT_PITCH(d)			(0x00001b10 + d * 0x0040)	/* Parameters: pitch */
+#	define		NV20_TCL_PRIMITIVE_3D_TX_FILTER(d)			(0x00001b14 + d * 0x0040)	/* Parameters: mag_filter min_filter */
+#	define		NV20_TCL_PRIMITIVE_3D_TX_NPOT_SIZE(d)			(0x00001b1c + d * 0x0040)	/* Parameters: width height */
+#	define		NV20_TCL_PRIMITIVE_3D_TX_PALETTE_OFFSET(d)		(0x00001b20 + d * 0x0040)
+#	define		NV20_TCL_PRIMITIVE_3D_RC_ENABLE				0x00001e60	/* Parameters: number of rc enabled */
+#	define		NV20_TCL_PRIMITIVE_3D_TX_SHADER_OP			0x00001e70	/* Parameters: op0 op1 op2 op3 */
+#	define		NV20_TCL_PRIMITIVE_3D_TX_SHADER_CULL_MODE		0x000017f8	/* Parameters: cull0 cull1 cull2 cull3 */
+#	define		NV20_TCL_PRIMITIVE_3D_TX_SHADER_PREVIOUS		0x00001e78	/* Parameters: prev2 prev3 */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_COLOR0				0x00001e20	/* Parameters: a r g b */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_COLOR1				0x00001e24	/* Parameters: a r g b */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_FINAL0				0x00000288	/* Parameters: vara_mapping vara_component_usage vara_input varb_mapping varb_component_usage varb_input varc_mapping varc_component_usage varc_input vard_mapping vard_component_usage vard_input */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_FINAL1				0x0000028c	/* Parameters: vare_mapping vare_component_usage vare_input varf_mapping varf_component_usage varf_input varg_mapping varg_component_usage varg_input color_sum_clamp */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_IN_ALPHA(d)			(0x00000260 + d * 0x0004)	/* Parameters: vara_mapping vara_component_usage vara_input varb_mapping varb_component_usage varb_input varc_mapping varc_component_usage varc_input vard_mapping vard_component_usage vard_input */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_IN_RGB(d)			(0x00000ac0 + d * 0x0004)	/* Parameters: vara_mapping vara_component_usage vara_input varb_mapping varb_component_usage varb_input varc_mapping varc_component_usage varc_input vard_mapping vard_component_usage vard_input */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_CONSTANT_COLOR0(d)		(0x00000a60 + d * 0x0004)	/* Parameters: a r g b */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_CONSTANT_COLOR1(d)		(0x00000a80 + d * 0x0004)	/* Parameters: a r g b */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_OUT_ALPHA(d)			(0x00000aa0 + d * 0x0004)	/* Parameters: scale bias mux_sum ab_dot_product cd_dot_product sum_output ab_output cd_output */
+#	define		NV20_TCL_PRIMITIVE_3D_RC_OUT_RGB(d)			(0x00001e40 + d * 0x0004)	/* Parameters: scale bias mux_sum ab_dot_product cd_dot_product sum_output ab_output cd_output */
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_POSITION_X(d)		(0x0000105c + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_POSITION_Y(d)		(0x00001060 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_POSITION_Z(d)		(0x00001064 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_HALF_VECTOR_X(d)		(0x00001028 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_HALF_VECTOR_Y(d)		(0x0000102c + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_HALF_VECTOR_Z(d)		(0x00001030 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_DIRECTION_X(d)		(0x00001034 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_DIRECTION_Y(d)		(0x00001038 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_DIRECTION_Z(d)		(0x0000103c + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R(d)	(0x00001000 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_G(d)	(0x00001004 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_B(d)	(0x00001008 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R(d)	(0x0000100c + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_G(d)	(0x00001010 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_B(d)	(0x00001014 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R(d)	(0x00001018 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_G(d)	(0x0000101c + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_B(d)	(0x00001020 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_BACK_SIDE_PRODUCT_AMBIENT(d)	(0x00000c00 + d * 0x0040)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_BACK_SIDE_PRODUCT_DIFFUSE(d)	(0x00000c0c + d * 0x0040)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_BACK_SIDE_PRODUCT_SPECULAR(d)	(0x00000c18 + d * 0x0040)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_CONSTANT_ATTENUATION(d)	(0x00001068 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_LINEAR_ATTENUATION(d)	(0x0000106c + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_QUADRATIC_ATTENUATION(d)	(0x00001070 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_A(d)		(0x00001040 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_EXPONENT(d)		(0x00001044 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_B(d)		(0x00001048 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_X(d)		(0x0000104c + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_Y(d)		(0x00001050 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_Z(d)		(0x00001054 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_C(d)		(0x00001058 + d * 0x0080)
+#	define		NV20_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_A		0x00001e28
+#	define		NV20_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_B		0x00001e2c
+#	define		NV20_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_C		0x00001e30
+#	define		NV20_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_D		0x00001e34
+#	define		NV20_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_E		0x00001e38
+#	define		NV20_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_F		0x00001e3c
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_ENABLE		0x0000147c
+#	define		NV20_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_PATTERN(d)	(0x00001480 + d * 0x0004)
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_POS_3F_X			0x00001500
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_POS_3F_Y			0x00001504
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_POS_3F_Z			0x00001508
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_POS_4F_X			0x00001518
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_POS_4F_Y			0x0000151c
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_POS_4F_Z			0x00001520
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_POS_4F_W			0x00001524
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_POS_4I_XY			0x00001528	/* Parameters: y x */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_POS_4I_ZW			0x0000152c	/* Parameters: w z */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_NOR_3F_X			0x00001530
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_NOR_3F_Y			0x00001534
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_NOR_3F_Z			0x00001538
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_NOR_3I_XY			0x00001540	/* Parameters: y x */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_NOR_3I_Z			0x00001544	/* Parameters: z */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL_4F_R			0x00001550
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL_4F_G			0x00001554
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL_4F_B			0x00001558
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL_4F_A			0x0000155c
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL_3F_R			0x00001560
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL_3F_G			0x00001564
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL_3F_B			0x00001568
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL_4I			0x0000156c	/* Parameters: a b g r */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL2_3F_R			0x00001580
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL2_3F_G			0x00001584
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL2_3F_B			0x00001588
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_COL2_3I			0x0000158c	/* Parameters: a b g r */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX0_2F_S			0x00001590
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX0_2F_T			0x00001594
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX0_2I			0x00001598	/* Parameters: t s */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_S			0x000015a0
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_T			0x000015a4
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_R			0x000015a8
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_Q			0x000015ac
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX0_4I_ST			0x000015b0	/* Parameters: t s */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX0_4I_RQ			0x000015b4	/* Parameters: q r */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX1_2F_S			0x000015b8
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX1_2F_T			0x000015bc
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX1_2I			0x000015c0	/* Parameters: t s */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_S			0x000015c8
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_T			0x000015cc
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_R			0x000015d0
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_Q			0x000015d4
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX1_4I_ST			0x000015d8	/* Parameters: t s */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX1_4I_RQ			0x000015dc	/* Parameters: q r */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX2_2F_S			0x000015e0
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX2_2F_T			0x000015e4
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX2_2I			0x000015e8	/* Parameters: t s */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX2_4F_S			0x000015f0
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX2_4F_T			0x000015f4
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX2_4F_R			0x000015f8
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX2_4F_Q			0x000015fc
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX2_4I_ST			0x00001600	/* Parameters: t s */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX2_4I_RQ			0x00001604	/* Parameters: q r */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX3_2F_S			0x00001608
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX3_2F_T			0x0000160c
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX3_2I			0x00001610	/* Parameters: t s */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX3_4F_S			0x00001620
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX3_4F_T			0x00001624
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX3_4F_R			0x00001628
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX3_4F_Q			0x0000162c
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX3_4I_ST			0x00001630	/* Parameters: t s */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_TX3_4I_RQ			0x00001634	/* Parameters: q r */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_FOG_1F			0x00001698
+#	define		NV20_TCL_PRIMITIVE_3D_EDGE_FLAG				0x000016bc
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR0_POS		0x00001720	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR1_WGH		0x00001724	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR2_NOR		0x00001728	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR3_COL		0x0000172c	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR4_COL2		0x00001730	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR5_FOG		0x00001734	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR6			0x00001738	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR7			0x0000173c	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR8_TX0		0x00001740	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR9_TX1		0x00001744	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR10_TX2		0x00001748	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR11_TX3		0x0000174c	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR12_TX4		0x00001750	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR13_TX5		0x00001754	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR14_TX6		0x00001758	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_POINTER_ATTR15_TX7		0x0000175c	/* Parameters: enabled? offset */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR( d)			(0x00001760 + d * 0x0004)
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR0_POS			0x00001760	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR1_WGH			0x00001764	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR2_NOR			0x00001768	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR3_COL			0x0000176c	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR4_COL2			0x00001770	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR5_FOG			0x00001774	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR6			0x00001778	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR7			0x0000177c	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR8_TX0			0x00001780	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR9_TX1			0x00001784	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR10_TX2			0x00001788	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR11_TX3			0x0000178c	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR12_TX4			0x00001790	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR13_TX5			0x00001794	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR14_TX6			0x00001798	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR15_TX7			0x0000179c	/* Parameters: stride fields type */
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R	0x000017a0
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G	0x000017a4
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B	0x000017a8
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MATERIAL_BACK_A		0x000017ac
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MATERIAL_BACK_R		0x000017b0
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MATERIAL_BACK_G		0x000017b4
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_MATERIAL_BACK_B		0x000017b8
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_ENABLE		0x000017bc
+#	define		NV20_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_OP			0x000017c0
+#	define		NV20_TCL_PRIMITIVE_3D_LIGHT_MODEL_TWO_SIDE_ENABLE	0x000017c4
+#	define		NV20_TCL_PRIMITIVE_3D_BEGIN_END				0x000017fc
+#	define		NV20_TCL_PRIMITIVE_3D_SCISSOR_X2_X1			0x00001c30	/* Parameters: x2 x1 */
+#	define		NV20_TCL_PRIMITIVE_3D_SCISSOR_Y2_Y1			0x00001c50	/* Parameters: y2 y1 */
+#	define		NV20_TCL_PRIMITIVE_3D_CLEAR_VALUE_DEPTH			0x00001d8c
+#	define		NV20_TCL_PRIMITIVE_3D_CLEAR_VALUE_ARGB			0x00001d90
+#	define		NV20_TCL_PRIMITIVE_3D_CLEAR_WHICH_BUFFERS		0x00001d94	/* Parameters: clear color a clear color b clear color g clear color r clear depth clear stencil */
+#	define		NV20_TCL_PRIMITIVE_3D_INDEX_DATA			0x00001800	/* Parameters: index1 index0 */
+#	define		NV20_TCL_PRIMITIVE_3D_VB_VERTEX_BATCH			0x00001810	/* Parameters: count_vertices offset_vertices */
+#	define		NV20_TCL_PRIMITIVE_3D_VERTEX_DATA			0x00001818
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_X			0x00001f00
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_Y			0x00001f04
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_Z			0x00001f08
+#	define		NV20_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_W			0x00001f0c
+
+/****************************************** 
+Object NV30_TCL_PRIMITIVE_3D used on: NV30 NV40 G70
+*/
+#define			NV30_TCL_PRIMITIVE_3D					0x00000097
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT0			0x00000180
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT1			0x00000184
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT2			0x00000188
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT3			0x0000018c
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT4			0x00000194
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT5			0x00000198
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT6			0x000001a4
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT7			0x000001a8
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT8			0x000001ac
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT9			0x000001b4
+#	define		NV30_TCL_PRIMITIVE_3D_SET_OBJECT10			0x000001b8
+#	define		NV30_TCL_PRIMITIVE_3D_SET_VB_SRC0_OBJECT		0x0000019c
+#	define		NV30_TCL_PRIMITIVE_3D_SET_VB_SRC1_OBJECT		0x000001a0
+#	define		NV30_TCL_PRIMITIVE_3D_BUFFER0_PITCH			0x0000020c	/* Parameters: depth/stencil buffer pitch color0 buffer pitch */
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR0_OFFSET			0x00000210
+#	define		NV30_TCL_PRIMITIVE_3D_DEPTH_OFFSET			0x00000214
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR1_OFFSET			0x00000218
+#	define		NV30_TCL_PRIMITIVE_3D_BUFFER1_PITCH			0x0000021c	/* Parameters: color1 buffer pitch */
+#	define		NV30_TCL_PRIMITIVE_3D_LMA_DEPTH_BUFFER_PITCH		0x0000022c	/* Parameters: pitch */
+#	define		NV30_TCL_PRIMITIVE_3D_LMA_DEPTH_BUFFER_OFFSET		0x00000230
+#	define		NV30_TCL_PRIMITIVE_3D_TX_MATRIX_ENABLE(d)		(0x00000240 + d * 0x0004)
+#	define		NV30_TCL_PRIMITIVE_3D_BUFFER2_PITCH			0x00000280
+#	define		NV30_TCL_PRIMITIVE_3D_BUFFER3_PITCH			0x00000284
+#	define		NV30_TCL_PRIMITIVE_3D_BUFFER2_OFFSET			0x00000288
+#	define		NV30_TCL_PRIMITIVE_3D_BUFFER3_OFFSET			0x0000028c
+#	define		NV30_TCL_PRIMITIVE_3D_DITHER_ENABLE			0x00000300
+#	define		NV30_TCL_PRIMITIVE_3D_ALPHA_FUNC_ENABLE			0x00000304
+#	define		NV30_TCL_PRIMITIVE_3D_ALPHA_FUNC_FUNC			0x00000308
+#	define		NV30_TCL_PRIMITIVE_3D_ALPHA_FUNC_REF			0x0000030c
+#	define		NV30_TCL_PRIMITIVE_3D_BLEND_FUNC_ENABLE			0x00000310
+#	define		NV30_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC			0x00000314
+#	define		NV30_TCL_PRIMITIVE_3D_BLEND_FUNC_DST			0x00000318
+#	define		NV30_TCL_PRIMITIVE_3D_BLEND_COLOR			0x0000031c	/* Parameters: a r g b */
+#	define		NV30_TCL_PRIMITIVE_3D_BLEND_EQUATION			0x00000320
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_MASK			0x00000324	/* Parameters: a r g b */
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_ENABLE		0x00000328
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_MASK			0x0000032c
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_FUNC_FUNC		0x00000330
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_FUNC_REF		0x00000334
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_FUNC_MASK		0x00000338
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_OP_FAIL		0x0000033c
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_OP_ZFAIL		0x00000340
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_OP_ZPASS		0x00000344
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_ENABLE		0x00000348
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_MASK		0x0000034c
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_FUNC		0x00000350
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_REF		0x00000354
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_MASK		0x00000358
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_OP_FAIL		0x0000035c
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_OP_ZFAIL		0x00000360
+#	define		NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_OP_ZPASS		0x00000364
+#	define		NV30_TCL_PRIMITIVE_3D_SHADE_MODEL			0x00000368
+#	define		NV30_TCL_PRIMITIVE_3D_FOG_ENABLE			0x0000036c
+#	define		NV30_TCL_PRIMITIVE_3D_FOG_COLOR				0x00000370
+#	define		NV40_TCL_PRIMITIVE_3D_COLOR_MASK_BUFFER123		0x00000370	/* Parameters: buffer3 b buffer3 g buffer3 r buffer3 a buffer2 b buffer2 g buffer2 r buffer2 a buffer1 b buffer1 g buffer1 r buffer1 a */
+#	define		NV30_TCL_PRIMITIVE_3D_NORMALIZE_ENABLE			0x0000037c
+#	define		NV30_TCL_PRIMITIVE_3D_DEPTH_RANGE_NEAR			0x00000394
+#	define		NV30_TCL_PRIMITIVE_3D_DEPTH_RANGE_FAR			0x00000398
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_MATERIAL_FRONT_R		0x000003a0
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_MATERIAL_FRONT_G		0x000003a4
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_MATERIAL_FRONT_B		0x000003a8
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_MATERIAL_FRONT_A		0x000003b4
+#	define		NV30_TCL_PRIMITIVE_3D_LINE_WIDTH_SMOOTH			0x000003b8
+#	define		NV30_TCL_PRIMITIVE_3D_LINE_SMOOTH_ENABLE		0x000003bc
+#	define		NV30_TCL_PRIMITIVE_3D_CLIP_PLANE_ENABLE(d)		(0x00000400 + d * 0x0004)
+#	define		NV30_TCL_PRIMITIVE_3D_MODELVIEW_MATRIX( d)		(0x00000480 + d * 0x0004)
+#	define		NV30_TCL_PRIMITIVE_3D_INVERSE_MODELVIEW_MATRIX( d)	(0x00000580 + d * 0x0004)
+#	define		NV30_TCL_PRIMITIVE_3D_PROJECTION_MATRIX( d)		(0x00000680 + d * 0x0004)
+#	define		NV30_TCL_PRIMITIVE_3D_TX_MATRIX(x,y)			(0x000006c0 + y * 0x0010 + x * 0x0004)
+#	define		NV30_TCL_PRIMITIVE_3D_FP_ACTIVE_PROGRAM			0x000008e4
+#	define		NV30_TCL_PRIMITIVE_3D_FOG_COORD_DIST			0x000008c8
+#	define		NV30_TCL_PRIMITIVE_3D_FOG_MODE				0x000008cc
+#	define		NV30_TCL_PRIMITIVE_3D_FOG_EQUATION_CONSTANT		0x000008d0
+#	define		NV30_TCL_PRIMITIVE_3D_FOG_EQUATION_LINEAR		0x000008d4
+#	define		NV30_TCL_PRIMITIVE_3D_FOG_EQUATION_QUADRATIC		0x000008d8
+#	define		NV30_TCL_PRIMITIVE_3D_RC_COLOR0				0x000008ec	/* Parameters: a r g b */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_COLOR1				0x000008f0	/* Parameters: a r g b */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_FINAL0				0x000008f4	/* Parameters: vara_mapping vara_component_usage vara_input varb_mapping varb_component_usage varb_input varc_mapping varc_component_usage varc_input vard_mapping vard_component_usage vard_input */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_FINAL1				0x000008f8	/* Parameters: vare_mapping vare_component_usage vare_input varf_mapping varf_component_usage varf_input varg_mapping varg_component_usage varg_input color_sum_clamp */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_ENABLE				0x000008fc	/* Parameters: number of rc enabled */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_IN_ALPHA(d)			(0x00000900 + d * 0x0020)	/* Parameters: vara_mapping vara_component_usage vara_input varb_mapping varb_component_usage varb_input varc_mapping varc_component_usage varc_input vard_mapping vard_component_usage vard_input */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_IN_RGB(d)			(0x00000904 + d * 0x0020)	/* Parameters: vara_mapping vara_component_usage vara_input varb_mapping varb_component_usage varb_input varc_mapping varc_component_usage varc_input vard_mapping vard_component_usage vard_input */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_CONSTANT_COLOR0(d)		(0x00000908 + d * 0x0020)	/* Parameters: a r g b */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_CONSTANT_COLOR1(d)		(0x0000090c + d * 0x0020)	/* Parameters: a r g b */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_OUT_ALPHA(d)			(0x00000910 + d * 0x0020)	/* Parameters: scale bias mux_sum ab_dot_product cd_dot_product sum_output ab_output cd_output */
+#	define		NV30_TCL_PRIMITIVE_3D_RC_OUT_RGB(d)			(0x00000914 + d * 0x0020)	/* Parameters: scale bias mux_sum ab_dot_product cd_dot_product sum_output ab_output cd_output */
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_COLOR_BUFFER_DIM0	0x00000200	/* Parameters: width x_offset */
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_COLOR_BUFFER_DIM1	0x00000204	/* Parameters: height y_offset */
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_COLOR_BUFFER_OFS0	0x000002c0	/* Parameters: width x_offset */
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_COLOR_BUFFER_OFS1	0x000002c4	/* Parameters: height y_offset */
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_DIMS_0			0x00000a00	/* Parameters: width x_offset */
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_DIMS_1			0x00000a04	/* Parameters: height y_offset */
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R	0x00000a10
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G	0x00000a14
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_MODEL_FRONT_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B	0x00000a18
+#	define		NV30_TCL_PRIMITIVE_3D_SCISSOR_WIDTH_XPOS		0x000008c0	/* Parameters: width x_offset */
+#	define		NV30_TCL_PRIMITIVE_3D_SCISSOR_HEIGHT_YPOS		0x000008c4	/* Parameters: height y_offset */
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_SPRITE			0x00001ee8	/* Parameters: coord_replace r_mode enable */
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_SIZE			0x00001ee0
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_PARAMETER_A			0x00001ec0
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_PARAMETER_B			0x00001ec4
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_PARAMETER_C			0x00001ec8
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_PARAMETER_D			0x00001ecc
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_PARAMETER_E			0x00001ed0
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_PARAMETER_F			0x00001ed4
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_PARAMETER_G			0x00001ed8
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_PARAMETER_H			0x00001edc
+#	define		NV30_TCL_PRIMITIVE_3D_POINT_PARAMETERS_ENABLE		0x00001ee4
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_OX			0x00000a20
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_OY			0x00000a24
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_NPF_DIV2		0x00000a28
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_UNK0_0x0		0x00000a2c
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_PX_DIV2		0x00000a30
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_PY_DIV2		0x00000a34
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_FMN_DIV2		0x00000a38
+#	define		NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_UNK1_0x0		0x00000a3c
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FILL_ENABLE	0x00000a60
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_OFFSET_LINE_ENABLE	0x00000a64
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_OFFSET_POINT_ENABLE	0x00000a68
+#	define		NV30_TCL_PRIMITIVE_3D_DEPTH_FUNC			0x00000a6c
+#	define		NV30_TCL_PRIMITIVE_3D_DEPTH_WRITE_ENABLE		0x00000a70
+#	define		NV30_TCL_PRIMITIVE_3D_DEPTH_TEST_ENABLE			0x00000a74
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FACTOR		0x00000a78
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_OFFSET_UNITS		0x00000a7c
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_INST0			0x00000b80
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_INST1			0x00000b84
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_INST2			0x00000b88
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_INST3			0x00000b8c
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_MATERIAL_BACK_R		0x000017b0
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_MATERIAL_BACK_G		0x000017b4
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_MATERIAL_BACK_B		0x000017b8
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_MATERIAL_BACK_A		0x000017c0
+#	define		NV30_TCL_PRIMITIVE_3D_OCC_QUERY_OR_COLOR_BUFF_ENABLE	0x000017c8
+#	define		NV30_TCL_PRIMITIVE_3D_STORE_RESULT			0x00001800
+#	define		NV30_TCL_PRIMITIVE_3D_CLIP_PLANE_A(d)			(0x00000e00 + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_CLIP_PLANE_B(d)			(0x00000e04 + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_CLIP_PLANE_C(d)			(0x00000e08 + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_CLIP_PLANE_D(d)			(0x00000e0c + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R(d)	(0x00001000 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_G(d)	(0x00001004 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_B(d)	(0x00001008 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R(d)	(0x0000100c + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_G(d)	(0x00001010 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_B(d)	(0x00001014 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R(d)	(0x00001018 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_G(d)	(0x0000101c + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_B(d)	(0x00001020 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_HALF_VECTOR_X(d)		(0x00001028 + d * 0x0080)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_HALF_VECTOR_Y(d)		(0x0000102c + d * 0x0080)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_HALF_VECTOR_Z(d)		(0x00001030 + d * 0x0080)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_DIRECTION_X(d)		(0x00001034 + d * 0x0080)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_DIRECTION_Y(d)		(0x00001038 + d * 0x0080)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_DIRECTION_Z(d)		(0x0000103c + d * 0x0080)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_CONSTANT_ATTENUATION(d)	(0x00001228 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_LINEAR_ATTENUATION(d)	(0x0000122c + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_QUADRATIC_ATTENUATION(d)	(0x00001230 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_A(d)		(0x00001200 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_EXPONENT(d)		(0x00001204 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_B(d)		(0x00001208 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_X(d)		(0x0000120c + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_Y(d)		(0x00001210 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_Z(d)		(0x00001214 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_C(d)		(0x00001218 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_POSITION_X(d)		(0x0000121c + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_POSITION_Y(d)		(0x00001220 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_POSITION_Z(d)		(0x00001224 + d * 0x0040)
+#	define		NV30_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_A	0x00001400
+#	define		NV30_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_B	0x00001404
+#	define		NV30_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_C	0x00001408
+#	define		NV30_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_D	0x0000140c
+#	define		NV30_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_E	0x00001410
+#	define		NV30_TCL_PRIMITIVE_3D_FRONT_MATERIAL_SHININESS_F	0x00001414
+#	define		NV30_TCL_PRIMITIVE_3D_ENABLED_LIGHTS			0x00001420	/* Parameters: light 7 light 6 light 5 light 4 light 3 light 2 light 1 light 0 */
+#	define		NV30_TCL_PRIMITIVE_3D_UNK1D6C_OFFSET			0x00001d6c
+#	define		NV30_TCL_PRIMITIVE_3D_UNK1D70_VALUE			0x00001d70
+#	define		NV30_TCL_PRIMITIVE_3D_LINE_STIPPLE_ENABLE		0x00001db4
+#	define		NV30_TCL_PRIMITIVE_3D_LINE_STIPPLE_PATTERN		0x00001db8	/* Parameters: factor pattern */
+#	define		NV30_TCL_PRIMITIVE_3D_BEGIN_END				0x00001808
+#	define		NV30_TCL_PRIMITIVE_3D_CULL_FACE				0x00001830
+#	define		NV30_TCL_PRIMITIVE_3D_FRONT_FACE			0x00001834
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_SMOOTH_ENABLE		0x00001838
+#	define		NV30_TCL_PRIMITIVE_3D_CULL_FACE_ENABLE			0x0000183c
+#	define		NV30_TCL_PRIMITIVE_3D_CLEAR_VALUE_DEPTH			0x00001d8c
+#	define		NV30_TCL_PRIMITIVE_3D_CLEAR_VALUE_ARGB			0x00001d90	/* Parameters: a r g b */
+#	define		NV30_TCL_PRIMITIVE_3D_CLEAR_WHICH_BUFFERS		0x00001d94
+#	define		NV30_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_A		0x00001e20
+#	define		NV30_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_B		0x00001e24
+#	define		NV30_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_C		0x00001e28
+#	define		NV30_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_D		0x00001e2c
+#	define		NV30_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_E		0x00001e30
+#	define		NV30_TCL_PRIMITIVE_3D_BACK_MATERIAL_SHININESS_F		0x00001e34
+#	define		NV30_TCL_PRIMITIVE_3D_DO_VERTICES			0x00001dac
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_FROM_ID			0x00001e9c
+#	define		NV30_TCL_PRIMITIVE_3D_VP_PROGRAM_START_ID		0x00001ea0
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_ID		0x00001efc
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P0_X		0x00001f00
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P0_Y		0x00001f04
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P0_Z		0x00001f08
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P0_W		0x00001f0c
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P1_X		0x00001f10
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P1_Y		0x00001f14
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P1_Z		0x00001f18
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P1_W		0x00001f1c
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P2_X		0x00001f20
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P2_Y		0x00001f24
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P2_Z		0x00001f28
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P2_W		0x00001f2c
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P3_X		0x00001f30
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P3_Y		0x00001f34
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P3_Z		0x00001f38
+#	define		NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_P3_W		0x00001f3c
+#	define		NV30_TCL_PRIMITIVE_3D_VTX_ATTR_3X(d)			(0x00001500 + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_VTX_ATTR_3Y(d)			(0x00001504 + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_VTX_ATTR_3Z(d)			(0x00001508 + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_VTX_ATTR_3W(d)			(0x0000150c + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_VTX_ATTR_4X(d)			(0x00001c00 + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_VTX_ATTR_4Y(d)			(0x00001c04 + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_VTX_ATTR_4Z(d)			(0x00001c08 + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_VTX_ATTR_4W(d)			(0x00001c0c + d * 0x0010)
+#	define		NV30_TCL_PRIMITIVE_3D_VB_POINTER_ATTR(d)		(0x00001680 + d * 0x0004)	/* Parameters: source: offset */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_NOR_3I_XY			0x00000a90	/* Parameters: y x */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_NOR_3I_Z			0x00000a94	/* Parameters: z */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX0_2F_S			0x000018c0
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX0_2F_T			0x000018c4
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX1_2F_S			0x000018c8
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX1_2F_T			0x000018cc
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX2_2F_S			0x000018d0
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX2_2F_T			0x000018d4
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX3_2F_S			0x000018d8
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX3_2F_T			0x000018dc
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX0_2I			0x00001920	/* Parameters: t s */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX1_2I			0x00001924	/* Parameters: t s */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX2_2I			0x00001928	/* Parameters: t s */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX3_2I			0x0000192c	/* Parameters: t s */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_COL_4I			0x0000194c	/* Parameters: a b g r */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_COL2_3I			0x00001950	/* Parameters: a b g r */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX0_4I_ST			0x000019c0	/* Parameters: t s */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX0_4I_RQ			0x000019c4	/* Parameters: q r */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX1_4I_ST			0x000019c8	/* Parameters: t s */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX1_4I_RQ			0x000019cc	/* Parameters: q r */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX2_4I_ST			0x000019d0	/* Parameters: t s */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX2_4I_RQ			0x000019d4	/* Parameters: q r */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX3_4I_ST			0x000019d8	/* Parameters: t s */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_TX3_4I_RQ			0x000019dc	/* Parameters: q r */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_FOG_1F			0x00001e54
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_UNK_0			0x00001718
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR( d)			(0x00001740 + d * 0x0004)
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR0_POS			0x00001740	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR1_WGH			0x00001744	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR2_NOR			0x00001748	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR3_COL			0x0000174c	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR4_COL2			0x00001750	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR5_FOG			0x00001754	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR6			0x00001758	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR7			0x0000175c	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR8_TX0			0x00001760	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR9_TX1			0x00001764	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR10_TX2			0x00001768	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR11_TX3			0x0000176c	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR12_TX4			0x00001770	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR13_TX5			0x00001774	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR14_TX6			0x00001778	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR15_TX7			0x0000177c	/* Parameters: stride fields type */
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_R	0x000017a0
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_G	0x000017a4
+#	define		NV30_TCL_PRIMITIVE_3D_LIGHT_MODEL_BACK_SIDE_PRODUCT_AMBIENT_PLUS_EMISSION_B	0x000017a8
+#	define		NV30_TCL_PRIMITIVE_3D_FP_ACTIVE_PROGRAM			0x000008e4
+#	define		NV30_TCL_PRIMITIVE_3D_TX_ADDRESS_UNIT(d)		(0x00001a00 + d * 0x0020)
+#	define		NV30_TCL_PRIMITIVE_3D_TX_FORMAT_UNIT(d)			(0x00001a04 + d * 0x0020)	/* Parameters: mipmap type format ncomp cubic */
+#	define		NV30_TCL_PRIMITIVE_3D_TX_WRAP_UNIT(d)			(0x00001a08 + d * 0x0020)	/* Parameters: wrap_s wrap_t wrap_r */
+#	define		NV30_TCL_PRIMITIVE_3D_TX_ENABLE_UNIT(d)			(0x00001a0c + d * 0x0020)	/* Parameters: nv40_enable nv30_enable anisotropy */
+#	define		NV30_TCL_PRIMITIVE_3D_TX_SWIZZLE_UNIT(d)		(0x00001a10 + d * 0x0020)
+#	define		NV30_TCL_PRIMITIVE_3D_TX_FILTER_UNIT(d)			(0x00001a14 + d * 0x0020)	/* Parameters: filter_min filter_mag */
+#	define		NV30_TCL_PRIMITIVE_3D_TX_XY_DIM_UNIT(d)			(0x00001a18 + d * 0x0020)	/* Parameters: width height */
+#	define		NV30_TCL_PRIMITIVE_3D_TX_UNK07_UNIT(d)			(0x00001a1c + d * 0x0020)
+#	define		NV30_TCL_PRIMITIVE_3D_TX_DEPTH_UNIT(d)			(0x00001840 + d * 0x0004)	/* Parameters: depth NPOT pitch */
+#	define		NV30_TCL_PRIMITIVE_3D_VB_VERTEX_BATCH			0x00001814	/* Parameters: count_vertices offset_vertices */
+#	define		NV30_TCL_PRIMITIVE_3D_VB_ELEMENT_U16			0x0000180c	/* Parameters: 1: 0: */
+#	define		NV30_TCL_PRIMITIVE_3D_VB_ELEMENT_U32			0x00001810
+#	define		NV30_TCL_PRIMITIVE_3D_VERTEX_DATA			0x00001818
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_ENABLE		0x00000374
+#	define		NV30_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_OP			0x00000378
+#	define		NV30_TCL_PRIMITIVE_3D_SET_DISPLAY_LIST_MEM_OFFSET	0x0000181c
+#	define		NV30_TCL_PRIMITIVE_3D_EXECUTE_DISPLAY_LIST		0x00001824	/* Parameters: length start offset */
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_MODE_FRONT		0x00001828
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_MODE_BACK			0x0000182c
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_ENABLE		0x0000147c
+#	define		NV30_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_PATTERN( d)	(0x00001480 + d * 0x0004)
+#	define		NV30_TCL_PRIMITIVE_3D_SET_CLIPPING_PLANES		0x00001478
+#	define		NV30_TCL_PRIMITIVE_3D_VP_IN_REG				0x00001ff0	/* Parameters: vertex pos weight normal primary color secondary color fogcoord texture coords 0 texture ccords 1 texture coords 2 texture coords 3 texture coords 4 texture coords 5 texture coords 6 texture coords 7 */
+#	define		NV30_TCL_PRIMITIVE_3D_VP_OUT_REG			0x00001ff4	/* Parameters: primary color secondary color backface primary color backface secondary color fogcoord pointsize clip plane 0 clip plane 1 clip plane 2 clip plane 3 clip plane 4 clip plane 5 texture coords 0 texture coords 1 texture coords 2 texture coords 3 texture coords 4 texture coords 5 texture coords 6 texture coords 7 */
+
+/****************************************** 
+Object NV30_CLEAR_BUFFER used on: NV30 NV40 G70
+*/
+#define			NV30_CLEAR_BUFFER					0x00000066
+#	define		NV30_CLEAR_BUFFER_SET_DMA_NOTIFY			0x00000180
+#	define		NV30_CLEAR_BUFFER_SET_IMAGE_PATTERN			0x00000188
+#	define		NV30_CLEAR_BUFFER_SET_RASTER_OP				0x0000018c
+#	define		NV30_CLEAR_BUFFER_SET_CONTEXT_SURFACE_2D		0x00000198
+#	define		NV30_CLEAR_BUFFER_UNK002fc				0x000002fc
+
+/****************************************** 
+Object NV50_TCL_PRIMITIVE_3D used on: 
+*/
+#define			NV50_TCL_PRIMITIVE_3D					0x00000097
+#	define		NV50_TCL_PRIMITIVE_3D_SET_OBJECT_0( d)			(0x00000180 + d * 0x0004)
+#	define		NV50_TCL_PRIMITIVE_3D_SET_OBJECT_1( d)			(0x000001c0 + d * 0x0004)
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_FOG_1F			0x00000314
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_2F_X			0x00000380
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_2F_Y			0x00000384
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX0_2F_S			0x000003c0
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX0_2F_T			0x000003c4
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX1_2F_S			0x000003c8
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX1_2F_T			0x000003cc
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX2_2F_S			0x000003d0
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX2_2F_T			0x000003d4
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX3_2F_S			0x000003d8
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX3_2F_T			0x000003dc
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_3F_X			0x00000400
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_3F_Y			0x00000404
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_3F_Z			0x00000408
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_NOR_3F_X			0x00000420
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_NOR_3F_Y			0x00000424
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_NOR_3F_Z			0x00000428
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL_3F_R			0x00000430
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL_3F_G			0x00000434
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL_3F_B			0x00000438
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL2_3F_R			0x00000440
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL2_3F_G			0x00000444
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL2_3F_B			0x00000448
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_4F_X			0x00000500
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_4F_Y			0x00000504
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_4F_Z			0x00000508
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_4F_W			0x0000050c
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL_4F_R			0x00000530
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL_4F_G			0x00000534
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL_4F_B			0x00000538
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL_4F_A			0x0000053c
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_S			0x00000580
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_T			0x00000584
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_R			0x00000588
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX0_4F_Q			0x0000058c
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_S			0x00000590
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_T			0x00000594
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_R			0x00000598
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX1_4F_Q			0x0000059c
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX2_4F_S			0x000005a0
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX2_4F_T			0x000005a4
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX2_4F_R			0x000005a8
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX2_4F_Q			0x000005ac
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX3_4F_S			0x000005b0
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX3_4F_T			0x000005b4
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX3_4F_R			0x000005b8
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX3_4F_Q			0x000005bc
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX0_2I			0x000006a0	/* Parameters: t s */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX1_2I			0x000006a4	/* Parameters: t s */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX2_2I			0x000006a8	/* Parameters: t s */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX3_2I			0x000006ac	/* Parameters: t s */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_4I_XY			0x00000700	/* Parameters: y x */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_POS_4I_ZW			0x00000704	/* Parameters: w z */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX0_4I_ST			0x00000740	/* Parameters: t s */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX0_4I_RQ			0x00000744	/* Parameters: q r */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX1_4I_ST			0x00000748	/* Parameters: t s */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX1_4I_RQ			0x0000074c	/* Parameters: q r */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX2_4I_ST			0x00000750	/* Parameters: t s */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX2_4I_RQ			0x00000754	/* Parameters: q r */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX3_4I_ST			0x00000758	/* Parameters: t s */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_TX3_4I_RQ			0x0000075c	/* Parameters: q r */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_NOR_3I_XY			0x00000790	/* Parameters: y x */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_NOR_3I_Z			0x00000794	/* Parameters: z */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL_4I			0x0000088c	/* Parameters: a b g r */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_COL2_3I			0x00000890	/* Parameters: a b g r */
+#	define		NV50_TCL_PRIMITIVE_3D_VIEWPORT_UNK0_X			0x00000a00
+#	define		NV50_TCL_PRIMITIVE_3D_VIEWPORT_UNK0_Y			0x00000a04
+#	define		NV50_TCL_PRIMITIVE_3D_VIEWPORT_UNK0_Z			0x00000a08
+#	define		NV50_TCL_PRIMITIVE_3D_VIEWPORT_UNK1_X			0x00000a0c
+#	define		NV50_TCL_PRIMITIVE_3D_VIEWPORT_UNK1_Y			0x00000a10
+#	define		NV50_TCL_PRIMITIVE_3D_VIEWPORT_UNK1_Z			0x00000a14
+#	define		NV50_TCL_PRIMITIVE_3D_DEPTH_RANGE_NEAR			0x00000c08
+#	define		NV50_TCL_PRIMITIVE_3D_DEPTH_RANGE_FAR			0x00000c0c
+#	define		NV50_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_HORIZ(d)		(0x00000d00 + d * 0x0008)	/* Parameters: x2 x1 */
+#	define		NV50_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_VERT(d)		(0x00000d04 + d * 0x0008)	/* Parameters: y2 y1 */
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_BUFFER_FIRST		0x00000d74
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_BUFFER_COUNT		0x00000d78
+#	define		NV50_TCL_PRIMITIVE_3D_CLEAR_COLOR_R			0x00000d80
+#	define		NV50_TCL_PRIMITIVE_3D_CLEAR_COLOR_G			0x00000d84
+#	define		NV50_TCL_PRIMITIVE_3D_CLEAR_COLOR_B			0x00000d88
+#	define		NV50_TCL_PRIMITIVE_3D_CLEAR_COLOR_A			0x00000d8c
+#	define		NV50_TCL_PRIMITIVE_3D_CLEAR_DEPTH			0x00000d90
+#	define		NV50_TCL_PRIMITIVE_3D_CLEAR_STENCIL			0x00000da0
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_MODE_FRONT		0x00000dac
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_MODE_BACK			0x00000db0
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_SMOOTH_ENABLE		0x00000db4
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_POINT_ENABLE	0x00000dc0
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_LINE_ENABLE	0x00000dc4
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FILL_ENABLE	0x00000dc8
+#	define		NV50_TCL_PRIMITIVE_3D_SCISSOR_WIDTH_XPOS		0x00000e04	/* Parameters: w x */
+#	define		NV50_TCL_PRIMITIVE_3D_SCISSOR_HEIGHT_YPOS		0x00000e08	/* Parameters: h y */
+#	define		NV50_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_ID		0x00000f00
+#	define		NV50_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_X			0x00000f04
+#	define		NV50_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_Y			0x00000f08
+#	define		NV50_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_Z			0x00000f0c
+#	define		NV50_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_W			0x00000f10
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_REF		0x00000f54
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_MASK		0x00000f58
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_MASK		0x00000f5c
+#	define		NV50_TCL_PRIMITIVE_3D_DEPTH_TEST_ENABLE			0x000012cc
+#	define		NV50_TCL_PRIMITIVE_3D_SHADE_MODEL			0x000012d4
+#	define		NV50_TCL_PRIMITIVE_3D_DEPTH_WRITE_ENABLE		0x000012e8
+#	define		NV50_TCL_PRIMITIVE_3D_ALPHA_FUNC_ENABLE			0x000012ec
+#	define		NV50_TCL_PRIMITIVE_3D_DEPTH_FUNC			0x0000130c
+#	define		NV50_TCL_PRIMITIVE_3D_ALPHA_FUNC_REF			0x00001310
+#	define		NV50_TCL_PRIMITIVE_3D_ALPHA_FUNC_FUNC			0x00001314
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_COLOR_R			0x0000131c
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_COLOR_G			0x00001320
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_COLOR_B			0x00001324
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_COLOR_A			0x00001328
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_EQUATION_RGB		0x00001340
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC_RGB		0x00001344
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_FUNC_DST_RGB		0x00001348
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_EQUATION_ALPHA		0x0000134c
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC_ALPHA		0x00001350
+#	define		NV50_TCL_PRIMITIVE_3D_BLEND_FUNC_DST_ALPHA		0x00001358
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_ENABLE		0x00001380
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_OP_FAIL		0x00001384
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_OP_ZFAIL		0x00001388
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_OP_ZPASS		0x0000138c
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_FUNC_FUNC		0x00001390
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_FUNC_REF		0x00001394
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_MASK			0x00001398
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_FUNC_MASK		0x0000139c
+#	define		NV50_TCL_PRIMITIVE_3D_LINE_WIDTH			0x000013b0
+#	define		NV50_TCL_PRIMITIVE_3D_POINT_SIZE			0x00001518
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FACTOR		0x0000156c
+#	define		NV50_TCL_PRIMITIVE_3D_LINE_SMOOTH_ENABLE		0x00001570
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_ENABLE		0x00001594
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_OP_FAIL		0x00001598
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_OP_ZFAIL		0x0000159c
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_OP_ZPASS		0x000015a0
+#	define		NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_FUNC		0x000015a4
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_UNITS		0x000015bc
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_BEGIN			0x000015dc
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_END			0x000015e0
+#	define		NV50_TCL_PRIMITIVE_3D_VERTEX_DATA			0x00001640
+#	define		NV50_TCL_PRIMITIVE_3D_LINE_STIPPLE_ENABLE		0x0000166c
+#	define		NV50_TCL_PRIMITIVE_3D_LINE_STIPPLE_PATTERN		0x00001680	/* Parameters: pattern factor */
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_ENABLE		0x0000168c
+#	define		NV50_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_PATTERN( d)	(0x00001700 + d * 0x0004)
+#	define		NV50_TCL_PRIMITIVE_3D_CULL_FACE_ENABLE			0x00001918
+#	define		NV50_TCL_PRIMITIVE_3D_FRONT_FACE			0x0000191c
+#	define		NV50_TCL_PRIMITIVE_3D_CULL_FACE				0x00001920
+#	define		NV50_TCL_PRIMITIVE_3D_LOGIC_OP_ENABLE			0x000019c4
+#	define		NV50_TCL_PRIMITIVE_3D_LOGIC_OP_OP			0x000019c8
+#	define		NV50_TCL_PRIMITIVE_3D_CLEAR_BUFFERS			0x000019d0	/* Parameters: color stencil depth */
+#	define		NV50_TCL_PRIMITIVE_3D_COLOR_MASK( d)			(0x00001a00 + d * 0x0004)	/* Parameters: a b g r */
+
+/****************************************** 
+Object NV_DMA_FROM_MEMORY used on: NV03 NV04 NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV_DMA_FROM_MEMORY					0x00000002
+
+/****************************************** 
+Object NV_DMA_TO_MEMORY used on: NV03 NV04 NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV_DMA_TO_MEMORY					0x00000003
+
+/****************************************** 
+Object NV_DMA_IN_MEMORY used on: NV03 NV04 NV10 NV15 NV20 NV30 NV40 G70
+*/
+#define			NV_DMA_IN_MEMORY					0x0000003d
+
+/****************************************** 
+Object NvType0046 used on: NV04 
+*/
+#define			NvType0046						0x00000046
+#	define		NvType0046_DMA_NOTIFY					0x00000180
+#	define		NvType0046_DMA_MEM_1					0x00000184
+#	define		NvType0046_DMA_MEM_2					0x00000188
+#	define		NvType0046_DMA_3					0x0000018c
+#	define		NvType0046_DMA_4					0x00000190
+#	define		NvType0046_OBJ_5					0x00000194
+#	define		NvType0046_OBJ_6					0x00000198
+#	define		NvType0046_PITCH1					0x00000304
+#	define		NvType0046_PITCH2					0x0000030c
+#	define		NvType0046_SIZE						0x00000340	/* Parameters: width height */
+#	define		NvType0046_WIDTH					0x00000344	/* Parameters: visible_width blank_width */
+#	define		NvType0046_HSYNC					0x00000348	/* Parameters: hsync_start hsync_len */
+#	define		NvType0046_HEIGHT					0x0000034c	/* Parameters: visible_height blank_height */
+#	define		NvType0046_VSYNC					0x00000350	/* Parameters: vsync_start vsync_len */
+#	define		NvType0046_FULL_SIZE					0x00000354	/* Parameters: full_width full_height */
+#	define		NvType0046_PIXEL_CLK					0x00000358
+#	define		NvType0046_FLAGS					0x0000035c	/* Parameters: doublescan -hsync -vsync depth */
+
+/****************************************** 
+Object NvType0047 used on: NV04 
+*/
+#define			NvType0047						0x00000047
+#	define		NvType0047_DMA_NOTIFY					0x00000180
+#	define		NvType0047_UNK19C					0x0000019c
+#	define		NvType0047_UNK1A0					0x000001a0
+
+/****************************************** 
+Object NvType0049 used on: NV04 
+*/
+#define			NvType0049						0x00000049
+#	define		NvType0049_DMA_NOTIFY					0x00000180
+#	define		NvType0049_DMA_MEM_1					0x00000184
+#	define		NvType0049_DMA_MEM_2					0x00000188
+
+/****************************************** 
+Object NvType004D used on: NV04 
+*/
+#define			NvType004D						0x0000004d
+#	define		NvType004D_DMA_NOTIFY					0x00000180
+#	define		NvType004D_DMA_MEM_1					0x00000184
+#	define		NvType004D_DMA_MEM_2					0x00000188
+#	define		NvType004D_DMA_MEM_3					0x0000018c
+#	define		NvType004D_DMA_MEM_4					0x00000190
+
+#endif /* _NOUVEAU_REG_H */
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_screen.c b/src/mesa/drivers/dri/nouveau/nouveau_screen.c
new file mode 100644
index 00000000000..881b20149f0
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_screen.c
@@ -0,0 +1,380 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#include "glheader.h"
+#include "imports.h"
+#include "mtypes.h"
+#include "framebuffer.h"
+#include "renderbuffer.h"
+
+#include "nouveau_context.h"
+#include "nouveau_screen.h"
+#include "nouveau_object.h"
+#include "nouveau_span.h"
+
+#include "utils.h"
+#include "context.h"
+#include "vblank.h"
+#include "drirenderbuffer.h"
+
+#include "GL/internal/dri_interface.h"
+
+#include "xmlpool.h"
+
+PUBLIC const char __driConfigOptions[] =
+DRI_CONF_BEGIN
+    DRI_CONF_SECTION_DEBUG
+        DRI_CONF_NO_RAST(false)
+    DRI_CONF_SECTION_END
+DRI_CONF_END;
+static const GLuint __driNConfigOptions = 1;
+
+extern const struct dri_extension common_extensions[];
+extern const struct dri_extension nv10_extensions[];
+extern const struct dri_extension nv20_extensions[];
+extern const struct dri_extension nv30_extensions[];
+extern const struct dri_extension nv40_extensions[];
+extern const struct dri_extension nv50_extensions[];
+
+static nouveauScreenPtr nouveauCreateScreen(__DRIscreenPrivate *sPriv)
+{
+	nouveauScreenPtr screen;
+	NOUVEAUDRIPtr dri_priv=(NOUVEAUDRIPtr)sPriv->pDevPriv;
+
+	/* allocate screen */
+	screen = (nouveauScreenPtr) CALLOC( sizeof(*screen) );
+	if ( !screen ) {         
+		__driUtilMessage("%s: Could not allocate memory for screen structure",__FUNCTION__);
+		return NULL;
+	}
+	
+	screen->card=nouveau_card_lookup(dri_priv->device_id);
+	if (!screen->card) {
+		__driUtilMessage("%s: Unknown card type 0x%04x:0x%04x\n",
+			__func__, dri_priv->device_id >> 16, dri_priv->device_id & 0xFFFF);
+		FREE(screen);
+		return NULL;
+	}
+
+	/* parse information in __driConfigOptions */
+	driParseOptionInfo (&screen->optionCache,__driConfigOptions, __driNConfigOptions);
+
+	screen->fbFormat    = dri_priv->bpp / 8;
+	screen->frontOffset = dri_priv->front_offset;
+	screen->frontPitch  = dri_priv->front_pitch;
+	screen->backOffset  = dri_priv->back_offset;
+	screen->backPitch   = dri_priv->back_pitch;
+	screen->depthOffset = dri_priv->depth_offset;
+	screen->depthPitch  = dri_priv->depth_pitch;
+
+	screen->driScreen = sPriv;
+	return screen;
+}
+
+static void
+nouveauDestroyScreen(__DRIscreenPrivate *sPriv)
+{
+	nouveauScreenPtr screen = (nouveauScreenPtr)sPriv->private;
+
+	if (!screen) return;
+
+	/* free all option information */
+	driDestroyOptionInfo (&screen->optionCache);
+
+	FREE(screen);
+	sPriv->private = NULL;
+}
+
+static GLboolean nouveauInitDriver(__DRIscreenPrivate *sPriv)
+{
+	sPriv->private = (void *) nouveauCreateScreen( sPriv );
+	if ( !sPriv->private ) {
+		nouveauDestroyScreen( sPriv );
+		return GL_FALSE;
+	}
+
+	return GL_TRUE;
+}
+
+/**
+ * Create the Mesa framebuffer and renderbuffers for a given window/drawable.
+ *
+ * \todo This function (and its interface) will need to be updated to support
+ * pbuffers.
+ */
+static GLboolean
+nouveauCreateBuffer(__DRIscreenPrivate *driScrnPriv,
+                    __DRIdrawablePrivate *driDrawPriv,
+                    const __GLcontextModes *mesaVis,
+                    GLboolean isPixmap)
+{
+	nouveauScreenPtr screen = (nouveauScreenPtr) driScrnPriv->private;
+	nouveau_renderbuffer  *nrb;
+	struct gl_framebuffer *fb;
+	const GLboolean swAccum = mesaVis->accumRedBits > 0;
+	const GLboolean swStencil = mesaVis->stencilBits > 0 && mesaVis->depthBits != 24;
+	GLenum color_format = screen->fbFormat == 4 ? GL_RGBA8 : GL_RGB5;
+
+	if (isPixmap)
+		return GL_FALSE; /* not implemented */
+
+	fb = _mesa_create_framebuffer(mesaVis);
+	if (!fb)
+		return GL_FALSE;
+
+	/* Front buffer */
+	nrb = nouveau_renderbuffer_new(color_format,
+				       driScrnPriv->pFB + screen->frontOffset,
+				       screen->frontOffset,
+				       screen->frontPitch * screen->fbFormat,
+				       driDrawPriv);
+	nouveauSpanSetFunctions(nrb, mesaVis);
+	_mesa_add_renderbuffer(fb, BUFFER_FRONT_LEFT, &nrb->mesa);
+
+	if (0 /* unified buffers if we choose to support them.. */) {
+	} else {
+		if (mesaVis->doubleBufferMode) {
+			nrb = nouveau_renderbuffer_new(color_format, NULL,
+						       0, 0,
+						       NULL);
+			nouveauSpanSetFunctions(nrb, mesaVis);
+			_mesa_add_renderbuffer(fb, BUFFER_BACK_LEFT, &nrb->mesa);
+		}
+
+		if (mesaVis->depthBits == 24 && mesaVis->stencilBits == 8) {
+			nrb = nouveau_renderbuffer_new(GL_DEPTH24_STENCIL8_EXT, NULL,
+						       0, 0,
+						       NULL);
+			nouveauSpanSetFunctions(nrb, mesaVis);
+			_mesa_add_renderbuffer(fb, BUFFER_DEPTH, &nrb->mesa);
+			_mesa_add_renderbuffer(fb, BUFFER_STENCIL, &nrb->mesa);
+		} else if (mesaVis->depthBits == 24) {
+			nrb = nouveau_renderbuffer_new(GL_DEPTH_COMPONENT24, NULL,
+						       0, 0,
+						       NULL);
+			nouveauSpanSetFunctions(nrb, mesaVis);
+			_mesa_add_renderbuffer(fb, BUFFER_DEPTH, &nrb->mesa);
+		} else if (mesaVis->depthBits == 16) {
+			nrb = nouveau_renderbuffer_new(GL_DEPTH_COMPONENT16, NULL,
+						       0, 0,
+						       NULL);
+			nouveauSpanSetFunctions(nrb, mesaVis);
+			_mesa_add_renderbuffer(fb, BUFFER_DEPTH, &nrb->mesa);
+		}
+	}
+
+	_mesa_add_soft_renderbuffers(fb,
+				     GL_FALSE, /* color */
+				     GL_FALSE, /* depth */
+				     swStencil,
+				     swAccum,
+				     GL_FALSE, /* alpha */
+				     GL_FALSE  /* aux */);
+
+	driDrawPriv->driverPrivate = (void *) fb;
+	return (driDrawPriv->driverPrivate != NULL);
+}
+
+
+static void
+nouveauDestroyBuffer(__DRIdrawablePrivate *driDrawPriv)
+{
+	_mesa_destroy_framebuffer((GLframebuffer *) (driDrawPriv->driverPrivate));
+}
+
+static int
+nouveauGetSwapInfo(__DRIdrawablePrivate *dpriv, __DRIswapInfo *sInfo)
+{
+	return -1;
+}
+
+static const struct __DriverAPIRec nouveauAPI = {
+	.InitDriver      = nouveauInitDriver,
+	.DestroyScreen   = nouveauDestroyScreen,
+	.CreateContext   = nouveauCreateContext,
+	.DestroyContext  = nouveauDestroyContext,
+	.CreateBuffer    = nouveauCreateBuffer,
+	.DestroyBuffer   = nouveauDestroyBuffer,
+	.SwapBuffers     = nouveauSwapBuffers,
+	.MakeCurrent     = nouveauMakeCurrent,
+	.UnbindContext   = nouveauUnbindContext,
+	.GetSwapInfo     = nouveauGetSwapInfo,
+	.GetMSC          = driGetMSC32,
+	.WaitForMSC      = driWaitForMSC32,
+	.WaitForSBC      = NULL,
+	.SwapBuffersMSC  = NULL,
+	.CopySubBuffer   = nouveauCopySubBuffer
+};
+
+
+static __GLcontextModes *
+nouveauFillInModes( unsigned pixel_bits, unsigned depth_bits,
+		 unsigned stencil_bits, GLboolean have_back_buffer )
+{
+	__GLcontextModes * modes;
+	__GLcontextModes * m;
+	unsigned num_modes;
+	unsigned depth_buffer_factor;
+	unsigned back_buffer_factor;
+	int i;
+
+	static const struct {
+		GLenum format;
+		GLenum type;
+	} fb_format_array[] = {
+		{ GL_RGB , GL_UNSIGNED_SHORT_5_6_5     },
+		{ GL_BGRA, GL_UNSIGNED_INT_8_8_8_8_REV },
+		{ GL_BGR , GL_UNSIGNED_INT_8_8_8_8_REV },
+	};
+
+	/* GLX_SWAP_COPY_OML is only supported because the Intel driver doesn't
+	 * support pageflipping at all.
+	 */
+	static const GLenum back_buffer_modes[] = {
+		GLX_NONE, GLX_SWAP_UNDEFINED_OML, GLX_SWAP_COPY_OML
+	};
+
+	u_int8_t depth_bits_array[4]   = { 0, 16, 24, 24 };
+	u_int8_t stencil_bits_array[4] = { 0,  0,  0,  8 };
+
+	depth_buffer_factor = 4;
+	back_buffer_factor  = (have_back_buffer) ? 3 : 1;
+
+	num_modes = ((pixel_bits==16) ? 1 : 2) *
+		depth_buffer_factor * back_buffer_factor * 4;
+	modes = (*dri_interface->createContextModes)(num_modes,
+						     sizeof(__GLcontextModes));
+	m = modes;
+
+	for (i=((pixel_bits==16)?0:1);i<((pixel_bits==16)?1:3);i++) {
+		if (!driFillInModes(&m, fb_format_array[i].format,
+					fb_format_array[i].type,
+					depth_bits_array,
+					stencil_bits_array,
+					depth_buffer_factor,
+					back_buffer_modes,
+					back_buffer_factor,
+					GLX_TRUE_COLOR)) {
+		fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
+				__func__, __LINE__ );
+		return NULL;
+		}
+
+		if (!driFillInModes(&m, fb_format_array[i].format,
+					fb_format_array[i].type,
+					depth_bits_array,
+					stencil_bits_array,
+					depth_buffer_factor,
+					back_buffer_modes,
+					back_buffer_factor,
+					GLX_DIRECT_COLOR)) {
+		fprintf( stderr, "[%s:%u] Error creating FBConfig!\n",
+				__func__, __LINE__ );
+		return NULL;
+		}
+	}
+
+	return modes;
+}
+
+
+/**
+ * This is the bootstrap function for the driver.  libGL supplies all of the
+ * requisite information about the system, and the driver initializes itself.
+ * This routine also fills in the linked list pointed to by \c driver_modes
+ * with the \c __GLcontextModes that the driver can support for windows or
+ * pbuffers.
+ * 
+ * \return A pointer to a \c __DRIscreenPrivate on success, or \c NULL on 
+ *         failure.
+ */
+PUBLIC
+void * __driCreateNewScreen_20050727( __DRInativeDisplay *dpy, int scrn, __DRIscreen *psc,
+				     const __GLcontextModes * modes,
+				     const __DRIversion * ddx_version,
+				     const __DRIversion * dri_version,
+				     const __DRIversion * drm_version,
+				     const __DRIframebuffer * frame_buffer,
+				     drmAddress pSAREA, int fd, 
+				     int internal_api_version,
+				     const __DRIinterfaceMethods * interface,
+				     __GLcontextModes ** driver_modes)
+			     
+{
+	__DRIscreenPrivate *psp;
+	static const __DRIversion ddx_expected = { 1, 2, 0 };
+	static const __DRIversion dri_expected = { 4, 0, 0 };
+	static const __DRIversion drm_expected = { 0, 0, 3 };
+
+	dri_interface = interface;
+
+	if (!driCheckDriDdxDrmVersions2("nouveau",
+					dri_version, & dri_expected,
+					ddx_version, & ddx_expected,
+					drm_version, & drm_expected)) {
+		return NULL;
+	}
+
+	// temporary lock step versioning
+	if (drm_expected.patch!=drm_version->patch) {
+		__driUtilMessage("%s: wrong DRM version, expected %d, got %d\n",
+				__func__,
+				drm_expected.patch, drm_version->patch);
+		return NULL;
+	}
+
+	psp = __driUtilCreateNewScreen(dpy, scrn, psc, NULL,
+				       ddx_version, dri_version, drm_version,
+				       frame_buffer, pSAREA, fd,
+				       internal_api_version, &nouveauAPI);
+	if ( psp != NULL ) {
+		NOUVEAUDRIPtr dri_priv = (NOUVEAUDRIPtr)psp->pDevPriv;
+
+		*driver_modes = nouveauFillInModes(dri_priv->bpp,
+						   (dri_priv->bpp == 16) ? 16 : 24,
+						   (dri_priv->bpp == 16) ? 0  : 8,
+						   1
+						   );
+
+		/* Calling driInitExtensions here, with a NULL context pointer, does not actually
+		 * enable the extensions.  It just makes sure that all the dispatch offsets for all
+		 * the extensions that *might* be enables are known.  This is needed because the
+		 * dispatch offsets need to be known when _mesa_context_create is called, but we can't
+		 * enable the extensions until we have a context pointer.
+		 * 
+		 * Hello chicken.  Hello egg.  How are you two today?
+		 */
+		driInitExtensions( NULL, common_extensions, GL_FALSE );
+		driInitExtensions( NULL,   nv10_extensions, GL_FALSE );
+		driInitExtensions( NULL,   nv10_extensions, GL_FALSE );
+		driInitExtensions( NULL,   nv30_extensions, GL_FALSE );
+		driInitExtensions( NULL,   nv40_extensions, GL_FALSE );
+		driInitExtensions( NULL,   nv50_extensions, GL_FALSE );
+	}
+
+	return (void *) psp;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_screen.h b/src/mesa/drivers/dri/nouveau/nouveau_screen.h
new file mode 100644
index 00000000000..decdafa86d5
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_screen.h
@@ -0,0 +1,61 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+#ifndef __NOUVEAU_SCREEN_H__
+#define __NOUVEAU_SCREEN_H__
+
+#include "xmlconfig.h"
+
+#include "nouveau_dri.h"
+#include "nouveau_card.h"
+
+typedef struct {
+	nouveau_card* card;
+	u_int32_t bus_type;
+	u_int32_t agp_mode;
+
+        GLint fbFormat;
+
+	GLuint frontOffset;
+	GLuint frontPitch;
+	GLuint backOffset;
+	GLuint backPitch;
+
+	GLuint depthOffset;
+	GLuint depthPitch;
+	GLuint spanOffset;
+
+	__DRIscreenPrivate *driScreen;
+	unsigned int sarea_priv_offset;
+
+	/* Configuration cache with default values for all contexts */
+	driOptionCache optionCache;
+
+} nouveauScreenRec, *nouveauScreenPtr;
+
+
+#endif /* __NOUVEAU_SCREEN_H__ */
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_shader.c b/src/mesa/drivers/dri/nouveau/nouveau_shader.c
new file mode 100644
index 00000000000..ba471325aaa
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_shader.c
@@ -0,0 +1,818 @@
+/*
+ * Copyright (C) 2006 Ben Skeggs.
+ *
+ * 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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
+ *   Ben Skeggs <darktama@iinet.net.au>
+ */
+
+#include "glheader.h"
+#include "macros.h"
+#include "enums.h"
+#include "extensions.h"
+
+#include "program.h"
+#include "tnl/tnl.h"
+#include "shader/arbprogparse.h"
+
+#include "nouveau_context.h"
+#include "nouveau_shader.h"
+
+/*****************************************************************************
+ * Mesa entry points
+ */
+static void
+nouveauBindProgram(GLcontext *ctx, GLenum target, struct gl_program *prog)
+{
+}
+
+static struct gl_program *
+nouveauNewProgram(GLcontext *ctx, GLenum target, GLuint id)
+{
+   nouveauShader *nvs;
+
+   nvs = CALLOC_STRUCT(_nouveauShader);
+   switch (target) {
+   case GL_VERTEX_PROGRAM_ARB:
+      return _mesa_init_vertex_program(ctx, &nvs->mesa.vp, target, id);
+   case GL_FRAGMENT_PROGRAM_ARB:
+      return _mesa_init_fragment_program(ctx, &nvs->mesa.fp, target, id);
+   default:
+      _mesa_problem(ctx, "Unsupported shader target");
+      break;
+   }
+
+   FREE(nvs);
+   return NULL;
+}
+
+static void
+nouveauDeleteProgram(GLcontext *ctx, struct gl_program *prog)
+{
+   nouveauShader *nvs = (nouveauShader *)prog;
+
+   if (nvs->translated)
+      FREE(nvs->program);
+   _mesa_delete_program(ctx, prog);
+}
+
+static void
+nouveauProgramStringNotify(GLcontext *ctx, GLenum target,
+      			   struct gl_program *prog)
+{
+   nouveauShader *nvs = (nouveauShader *)prog;
+
+   if (nvs->translated)
+      FREE(nvs->program);
+   nvs->translated = 0;
+
+   _tnl_program_string(ctx, target, prog);
+}
+
+static GLboolean
+nouveauIsProgramNative(GLcontext * ctx, GLenum target, struct gl_program *prog)
+{
+   nouveauShader *nvs = (nouveauShader *)prog;
+
+   return nvs->translated;
+}
+
+GLboolean
+nvsUpdateShader(GLcontext *ctx, nouveauShader *nvs)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   struct gl_program_parameter_list *plist;
+   int i;
+
+   /* Translate to HW format now if necessary */
+   if (!nvs->translated) {
+      /* Mesa ASM shader -> nouveauShader */
+      if (!nouveau_shader_pass0(ctx, nvs))
+	 return GL_FALSE;
+      /* Basic dead code elimination + register usage info */
+      if (!nouveau_shader_pass1(nvs))
+	 return GL_FALSE;
+      /* nouveauShader -> HW bytecode, HW register alloc */
+      if (!nouveau_shader_pass2(nvs))
+	 return GL_FALSE;
+      assert(nvs->translated);
+      assert(nvs->program);
+   }
+   
+   /* Update state parameters */
+   plist = nvs->mesa.vp.Base.Parameters;
+   _mesa_load_state_parameters(ctx, plist);
+   for (i=0; i<nvs->param_high; i++) {
+      if (!nvs->params[i].in_use)
+	 continue;
+
+      if (!nvs->on_hardware) {
+	 /* if we've been kicked off the hardware there's no guarantee our
+	  * consts are still there.. reupload them all
+	  */
+	 nvs->func->UpdateConst(ctx, nvs, i);
+      } else if (nvs->params[i].source_val) {
+	 /* update any changed state parameters */
+	 if (!TEST_EQ_4V(nvs->params[i].val, nvs->params[i].source_val))
+	    nvs->func->UpdateConst(ctx, nvs, i);
+      }
+   }
+
+   /* Upload program to hardware, this must come after state param update
+    * as >=NV30 fragprogs inline consts into the bytecode.
+    */
+   if (!nvs->on_hardware) {
+      nouveauShader **current;
+
+      if (nvs->mesa.vp.Base.Target == GL_VERTEX_PROGRAM_ARB)
+	 current = &nmesa->current_vertprog;
+      else
+	 current = &nmesa->current_fragprog;
+      if (*current) (*current)->on_hardware = 0;
+
+      nvs->func->UploadToHW(ctx, nvs);
+      nvs->on_hardware = 1;
+
+      *current = nvs;
+   }
+
+   return GL_TRUE;
+}
+
+nouveauShader *
+nvsBuildTextShader(GLcontext *ctx, GLenum target, const char *text)
+{
+   nouveauShader *nvs;
+
+   nvs = CALLOC_STRUCT(_nouveauShader);
+   if (!nvs)
+      return NULL;
+
+   if (target == GL_VERTEX_PROGRAM_ARB) {
+      _mesa_init_vertex_program(ctx, &nvs->mesa.vp, GL_VERTEX_PROGRAM_ARB, 0);
+      _mesa_parse_arb_vertex_program(ctx,
+	    			     GL_VERTEX_PROGRAM_ARB,
+				     text,
+				     strlen(text),
+				     &nvs->mesa.vp);
+   } else if (target == GL_FRAGMENT_PROGRAM_ARB) {
+      _mesa_init_fragment_program(ctx, &nvs->mesa.fp, GL_FRAGMENT_PROGRAM_ARB, 0);
+      _mesa_parse_arb_fragment_program(ctx,
+	    			       GL_FRAGMENT_PROGRAM_ARB,
+				       text,
+				       strlen(text),
+				       &nvs->mesa.fp);
+   }
+
+   nouveau_shader_pass0(ctx, nvs);
+   nouveau_shader_pass1(nvs);
+   nouveau_shader_pass2(nvs);
+
+   return nvs;
+}
+
+static void
+nvsBuildPassthroughVP(GLcontext *ctx)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+   const char *vp_text =
+      "!!ARBvp1.0\n"
+      "OPTION ARB_position_invariant;"
+      ""
+      "MOV result.color, vertex.color;\n"
+      "MOV result.texcoord[0], vertex.texcoord[0];\n"
+      "MOV result.texcoord[1], vertex.texcoord[1];\n"
+      "MOV result.texcoord[2], vertex.texcoord[2];\n"
+      "MOV result.texcoord[3], vertex.texcoord[3];\n"
+      "MOV result.texcoord[4], vertex.texcoord[4];\n"
+      "MOV result.texcoord[5], vertex.texcoord[5];\n"
+      "MOV result.texcoord[6], vertex.texcoord[6];\n"
+      "MOV result.texcoord[7], vertex.texcoord[7];\n"
+      "END";
+
+   nmesa->passthrough_vp = nvsBuildTextShader(ctx,
+	 				      GL_VERTEX_PROGRAM_ARB,
+					      vp_text);
+}
+
+static void
+nvsBuildPassthroughFP(GLcontext *ctx)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	const char *fp_text =
+		"!!ARBfp1.0\n"
+		"MOV result.color, fragment.color;\n"
+		"END";
+
+	nmesa->passthrough_fp = nvsBuildTextShader(ctx,
+						   GL_FRAGMENT_PROGRAM_ARB,
+						   fp_text);
+}
+
+void
+nouveauShaderInitFuncs(GLcontext * ctx)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+   switch (nmesa->screen->card->type) {
+   case NV_20:
+      NV20VPInitShaderFuncs(&nmesa->VPfunc);
+      break;
+   case NV_30:
+      NV30VPInitShaderFuncs(&nmesa->VPfunc);
+      NV30FPInitShaderFuncs(&nmesa->FPfunc);
+      break;
+   case NV_40:
+   case NV_44:
+      NV40VPInitShaderFuncs(&nmesa->VPfunc);
+      NV40FPInitShaderFuncs(&nmesa->FPfunc);
+      break;
+   case NV_50:
+   default:
+      return;
+   }
+
+   /* Build a vertex program that simply passes through all attribs.
+    * Needed to do swtcl on nv40
+    */
+   if (nmesa->screen->card->type >= NV_40)
+      nvsBuildPassthroughVP(ctx);
+
+   /* Needed on NV30, even when using swtcl, if you want to get colours */
+   if (nmesa->screen->card->type >= NV_30)
+      nvsBuildPassthroughFP(ctx);
+
+   ctx->Const.VertexProgram.MaxNativeInstructions    = nmesa->VPfunc.MaxInst;
+   ctx->Const.VertexProgram.MaxNativeAluInstructions = nmesa->VPfunc.MaxInst;
+   ctx->Const.VertexProgram.MaxNativeTexInstructions = nmesa->VPfunc.MaxInst;
+   ctx->Const.VertexProgram.MaxNativeTexIndirections =
+      ctx->Const.VertexProgram.MaxNativeTexInstructions;
+   ctx->Const.VertexProgram.MaxNativeAttribs         = nmesa->VPfunc.MaxAttrib;
+   ctx->Const.VertexProgram.MaxNativeTemps           = nmesa->VPfunc.MaxTemp;
+   ctx->Const.VertexProgram.MaxNativeAddressRegs     = nmesa->VPfunc.MaxAddress;
+   ctx->Const.VertexProgram.MaxNativeParameters      = nmesa->VPfunc.MaxConst;
+
+   if (nmesa->screen->card->type >= NV_30) {
+      ctx->Const.FragmentProgram.MaxNativeInstructions    = nmesa->FPfunc.MaxInst;
+      ctx->Const.FragmentProgram.MaxNativeAluInstructions = nmesa->FPfunc.MaxInst;
+      ctx->Const.FragmentProgram.MaxNativeTexInstructions = nmesa->FPfunc.MaxInst;
+      ctx->Const.FragmentProgram.MaxNativeTexIndirections =
+	 ctx->Const.FragmentProgram.MaxNativeTexInstructions;
+      ctx->Const.FragmentProgram.MaxNativeAttribs         = nmesa->FPfunc.MaxAttrib;
+      ctx->Const.FragmentProgram.MaxNativeTemps           = nmesa->FPfunc.MaxTemp;
+      ctx->Const.FragmentProgram.MaxNativeAddressRegs     = nmesa->FPfunc.MaxAddress;
+      ctx->Const.FragmentProgram.MaxNativeParameters      = nmesa->FPfunc.MaxConst;
+   }
+
+   ctx->Driver.NewProgram		= nouveauNewProgram;
+   ctx->Driver.BindProgram		= nouveauBindProgram;
+   ctx->Driver.DeleteProgram		= nouveauDeleteProgram;
+   ctx->Driver.ProgramStringNotify	= nouveauProgramStringNotify;
+   ctx->Driver.IsProgramNative		= nouveauIsProgramNative;
+}
+
+
+/*****************************************************************************
+ * Disassembly support structs
+ */
+#define CHECK_RANGE(idx, arr) ((idx)<sizeof(_##arr)/sizeof(const char *)) \
+	? _##arr[(idx)] : #arr"_OOB"
+
+#define NODS      (1<<0)
+#define BRANCH_TR (1<<1)
+#define BRANCH_EL (1<<2)
+#define BRANCH_EN (1<<3)
+#define BRANCH_RE (1<<4)
+#define BRANCH_ALL (BRANCH_TR|BRANCH_EL|BRANCH_EN)
+#define COUNT_INC (1<<4)
+#define COUNT_IND (1<<5)
+#define COUNT_NUM (1<<6)
+#define COUNT_ALL (COUNT_INC|COUNT_IND|COUNT_NUM)
+#define TI_UNIT   (1<<7)
+struct _opcode_info
+{
+   const char *name;
+   int numsrc;
+   int flags;
+};
+
+static struct _opcode_info ops[] = {
+   [NVS_OP_ABS] = {"ABS", 1, 0},
+   [NVS_OP_ADD] = {"ADD", 2, 0},
+   [NVS_OP_ARA] = {"ARA", 1, 0},
+   [NVS_OP_ARL] = {"ARL", 1, 0},
+   [NVS_OP_ARR] = {"ARR", 1, 0},
+   [NVS_OP_BRA] = {"BRA", 0, NODS | BRANCH_TR},
+   [NVS_OP_BRK] = {"BRK", 0, NODS},
+   [NVS_OP_CAL] = {"CAL", 0, NODS | BRANCH_TR},
+   [NVS_OP_CMP] = {"CMP", 2, 0},
+   [NVS_OP_COS] = {"COS", 1, 0},
+   [NVS_OP_DIV] = {"DIV", 2, 0},
+   [NVS_OP_DDX] = {"DDX", 1, 0},
+   [NVS_OP_DDY] = {"DDY", 1, 0},
+   [NVS_OP_DP2] = {"DP2", 2, 0},
+   [NVS_OP_DP2A] = {"DP2A", 3, 0},
+   [NVS_OP_DP3] = {"DP3", 2, 0},
+   [NVS_OP_DP4] = {"DP4", 2, 0},
+   [NVS_OP_DPH] = {"DPH", 2, 0},
+   [NVS_OP_DST] = {"DST", 2, 0},
+   [NVS_OP_EX2] = {"EX2", 1, 0},
+   [NVS_OP_EXP] = {"EXP", 1, 0},
+   [NVS_OP_FLR] = {"FLR", 1, 0},
+   [NVS_OP_FRC] = {"FRC", 1, 0},
+   [NVS_OP_IF] = {"IF", 0, NODS | BRANCH_EL | BRANCH_EN},
+   [NVS_OP_KIL] = {"KIL", 1, 0},
+   [NVS_OP_LG2] = {"LG2", 1, 0},
+   [NVS_OP_LIT] = {"LIT", 1, 0},
+   [NVS_OP_LOG] = {"LOG", 1, 0},
+   [NVS_OP_LOOP] = {"LOOP", 0, NODS | COUNT_ALL | BRANCH_EN},
+   [NVS_OP_LRP] = {"LRP", 3, 0},
+   [NVS_OP_MAD] = {"MAD", 3, 0},
+   [NVS_OP_MAX] = {"MAX", 2, 0},
+   [NVS_OP_MIN] = {"MIN", 2, 0},
+   [NVS_OP_MOV] = {"MOV", 1, 0},
+   [NVS_OP_MUL] = {"MUL", 2, 0},
+   [NVS_OP_NRM] = {"NRM", 1, 0},
+   [NVS_OP_PK2H] = {"PK2H", 1, 0},
+   [NVS_OP_PK2US] = {"PK2US", 1, 0},
+   [NVS_OP_PK4B] = {"PK4B", 1, 0},
+   [NVS_OP_PK4UB] = {"PK4UB", 1, 0},
+   [NVS_OP_POW] = {"POW", 2, 0},
+   [NVS_OP_POPA] = {"POPA", 0, 0},
+   [NVS_OP_PUSHA] = {"PUSHA", 1, NODS},
+   [NVS_OP_RCC] = {"RCC", 1, 0},
+   [NVS_OP_RCP] = {"RCP", 1, 0},
+   [NVS_OP_REP] = {"REP", 0, NODS | BRANCH_EN | COUNT_NUM},
+   [NVS_OP_RET] = {"RET", 0, NODS},
+   [NVS_OP_RFL] = {"RFL", 1, 0},
+   [NVS_OP_RSQ] = {"RSQ", 1, 0},
+   [NVS_OP_SCS] = {"SCS", 1, 0},
+   [NVS_OP_SEQ] = {"SEQ", 2, 0},
+   [NVS_OP_SFL] = {"SFL", 2, 0},
+   [NVS_OP_SGE] = {"SGE", 2, 0},
+   [NVS_OP_SGT] = {"SGT", 2, 0},
+   [NVS_OP_SIN] = {"SIN", 1, 0},
+   [NVS_OP_SLE] = {"SLE", 2, 0},
+   [NVS_OP_SLT] = {"SLT", 2, 0},
+   [NVS_OP_SNE] = {"SNE", 2, 0},
+   [NVS_OP_SSG] = {"SSG", 1, 0},
+   [NVS_OP_STR] = {"STR", 2, 0},
+   [NVS_OP_SUB] = {"SUB", 2, 0},
+   [NVS_OP_TEX] = {"TEX", 1, TI_UNIT},
+   [NVS_OP_TXB] = {"TXB", 1, TI_UNIT},
+   [NVS_OP_TXD] = {"TXD", 3, TI_UNIT},
+   [NVS_OP_TXL] = {"TXL", 1, TI_UNIT},
+   [NVS_OP_TXP] = {"TXP", 1, TI_UNIT},
+   [NVS_OP_UP2H] = {"UP2H", 1, 0},
+   [NVS_OP_UP2US] = {"UP2US", 1, 0},
+   [NVS_OP_UP4B] = {"UP4B", 1, 0},
+   [NVS_OP_UP4UB] = {"UP4UB", 1, 0},
+   [NVS_OP_X2D] = {"X2D", 3, 0},
+   [NVS_OP_XPD] = {"XPD", 2, 0},
+   [NVS_OP_NOP] = {"NOP", 0, NODS},
+};
+
+static struct _opcode_info *
+_get_op_info(int op)
+{
+   if (op >= (sizeof(ops) / sizeof(struct _opcode_info)))
+      return NULL;
+   if (ops[op].name == NULL)
+      return NULL;
+   return &ops[op];
+}
+
+static const char *_SFR_STRING[] = {
+   [NVS_FR_POSITION] = "position",
+   [NVS_FR_WEIGHT] = "weight",
+   [NVS_FR_NORMAL] = "normal",
+   [NVS_FR_COL0] = "color",
+   [NVS_FR_COL1] = "color.secondary",
+   [NVS_FR_BFC0] = "bfc",
+   [NVS_FR_BFC1] = "bfc.secondary",
+   [NVS_FR_FOGCOORD] = "fogcoord",
+   [NVS_FR_POINTSZ] = "pointsize",
+   [NVS_FR_TEXCOORD0] = "texcoord[0]",
+   [NVS_FR_TEXCOORD1] = "texcoord[1]",
+   [NVS_FR_TEXCOORD2] = "texcoord[2]",
+   [NVS_FR_TEXCOORD3] = "texcoord[3]",
+   [NVS_FR_TEXCOORD4] = "texcoord[4]",
+   [NVS_FR_TEXCOORD5] = "texcoord[5]",
+   [NVS_FR_TEXCOORD6] = "texcoord[6]",
+   [NVS_FR_TEXCOORD7] = "texcoord[7]",
+   [NVS_FR_FRAGDATA0] = "data[0]",
+   [NVS_FR_FRAGDATA1] = "data[1]",
+   [NVS_FR_FRAGDATA2] = "data[2]",
+   [NVS_FR_FRAGDATA3] = "data[3]",
+   [NVS_FR_CLIP0] = "clip_plane[0]",
+   [NVS_FR_CLIP1] = "clip_plane[1]",
+   [NVS_FR_CLIP2] = "clip_plane[2]",
+   [NVS_FR_CLIP3] = "clip_plane[3]",
+   [NVS_FR_CLIP4] = "clip_plane[4]",
+   [NVS_FR_CLIP5] = "clip_plane[5]",
+   [NVS_FR_CLIP6] = "clip_plane[6]",
+   [NVS_FR_FACING] = "facing",
+};
+
+#define SFR_STRING(idx) CHECK_RANGE((idx), SFR_STRING)
+
+static const char *_SWZ_STRING[] = {
+   [NVS_SWZ_X] = "x",
+   [NVS_SWZ_Y] = "y",
+   [NVS_SWZ_Z] = "z",
+   [NVS_SWZ_W] = "w"
+};
+
+#define SWZ_STRING(idx) CHECK_RANGE((idx), SWZ_STRING)
+
+static const char *_NVS_PREC_STRING[] = {
+   [NVS_PREC_FLOAT32] = "R",
+   [NVS_PREC_FLOAT16] = "H",
+   [NVS_PREC_FIXED12] = "X",
+   [NVS_PREC_UNKNOWN] = "?"
+};
+
+#define NVS_PREC_STRING(idx) CHECK_RANGE((idx), NVS_PREC_STRING)
+
+static const char *_NVS_COND_STRING[] = {
+   [NVS_COND_FL] = "FL",
+   [NVS_COND_LT] = "LT",
+   [NVS_COND_EQ] = "EQ",
+   [NVS_COND_LE] = "LE",
+   [NVS_COND_GT] = "GT",
+   [NVS_COND_NE] = "NE",
+   [NVS_COND_GE] = "GE",
+   [NVS_COND_TR] = "TR",
+   [NVS_COND_UNKNOWN] = "??"
+};
+
+#define NVS_COND_STRING(idx) CHECK_RANGE((idx), NVS_COND_STRING)
+
+/*****************************************************************************
+ * ShaderFragment dumping
+ */
+static void
+nvsDumpIndent(int lvl)
+{
+   while (lvl--)
+      printf("  ");
+}
+
+static void
+nvsDumpSwizzle(nvsSwzComp *swz)
+{
+   printf(".%s%s%s%s",
+	  SWZ_STRING(swz[0]),
+	  SWZ_STRING(swz[1]), SWZ_STRING(swz[2]), SWZ_STRING(swz[3])
+      );
+}
+
+static void
+nvsDumpReg(nvsInstruction * inst, nvsRegister * reg)
+{
+   if (reg->negate)
+      printf("-");
+   if (reg->abs)
+      printf("abs(");
+
+   switch (reg->file) {
+   case NVS_FILE_TEMP:
+      printf("R%d", reg->index);
+      nvsDumpSwizzle(reg->swizzle);
+      break;
+   case NVS_FILE_ATTRIB:
+      printf("attrib.%s", SFR_STRING(reg->index));
+      nvsDumpSwizzle(reg->swizzle);
+      break;
+   case NVS_FILE_ADDRESS:
+      printf("A%d", reg->index);
+      break;
+   case NVS_FILE_CONST:
+      if (reg->indexed)
+	 printf("const[A%d.%s + %d]",
+		reg->addr_reg, SWZ_STRING(reg->addr_comp), reg->index);
+      else
+	 printf("const[%d]", reg->index);
+      nvsDumpSwizzle(reg->swizzle);
+      break;
+   default:
+      printf("UNKNOWN_FILE");
+      break;
+   }
+
+   if (reg->abs)
+      printf(")");
+}
+
+static void
+nvsDumpInstruction(nvsInstruction * inst, int slot, int lvl)
+{
+   struct _opcode_info *opr = &ops[inst->op];
+   int i;
+
+   nvsDumpIndent(lvl);
+   printf("%s ", opr->name);
+
+   if (!opr->flags & NODS) {
+      switch (inst->dest.file) {
+      case NVS_FILE_RESULT:
+	 printf("result.%s", SFR_STRING(inst->dest.index));
+	 break;
+      case NVS_FILE_TEMP:
+	 printf("R%d", inst->dest.index);
+	 break;
+      case NVS_FILE_ADDRESS:
+	 printf("A%d", inst->dest.index);
+	 break;
+      default:
+	 printf("UNKNOWN_DST_FILE");
+	 break;
+      }
+
+      if (inst->mask != SMASK_ALL) {
+	 printf(".");
+	 if (inst->mask & SMASK_X)
+	    printf("x");
+	 if (inst->mask & SMASK_Y)
+	    printf("y");
+	 if (inst->mask & SMASK_Z)
+	    printf("z");
+	 if (inst->mask & SMASK_W)
+	    printf("w");
+      }
+
+      if (opr->numsrc)
+	 printf(", ");
+   }
+
+   for (i = 0; i < opr->numsrc; i++) {
+      nvsDumpReg(inst, &inst->src[i]);
+      if (i != opr->numsrc - 1)
+	 printf(", ");
+   }
+   if (opr->flags & TI_UNIT)
+      printf(", texture[%d]", inst->tex_unit);
+
+   printf("\n");
+}
+
+void
+nvsDumpFragmentList(nvsFragmentHeader *f, int lvl)
+{
+   while (f) {
+      switch (f->type) {
+      case NVS_INSTRUCTION:
+	 nvsDumpInstruction((nvsInstruction*)f, 0, lvl);
+	 break;
+      default:
+	 fprintf(stderr, "%s: Only NVS_INSTRUCTION fragments can be in"
+	                 "nvsFragmentList!\n", __func__);
+	 return;
+      }
+      f = f->next;
+   }
+}
+
+/*****************************************************************************
+ * HW shader disassembly
+ */
+static void
+nvsDisasmHWShaderOp(nvsFunc * shader, int merged)
+{
+   struct _opcode_info *opi;
+   nvsOpcode op;
+   nvsRegFile file;
+   nvsSwzComp swz[4];
+   int i;
+
+   op = shader->GetOpcode(shader, merged);
+   opi = _get_op_info(op);
+   if (!opi) {
+      printf("NO OPINFO!");
+      return;
+   }
+
+   printf("%s", opi->name);
+   if (shader->GetPrecision &&
+       (!(opi->flags & BRANCH_ALL)) && (!(opi->flags * NODS)) &&
+       (op != NVS_OP_NOP))
+      printf("%s", NVS_PREC_STRING(shader->GetPrecision(shader)));
+   if (shader->SupportsConditional && shader->SupportsConditional(shader)) {
+      if (shader->GetConditionUpdate(shader)) {
+	 printf("C%d", shader->GetCondRegID(shader));
+      }
+   }
+   if (shader->GetSaturate && shader->GetSaturate(shader))
+      printf("_SAT");
+
+   if (!(opi->flags & NODS)) {
+      int mask = shader->GetDestMask(shader, merged);
+
+      switch (shader->GetDestFile(shader, merged)) {
+      case NVS_FILE_ADDRESS:
+	 printf(" A%d", shader->GetDestID(shader, merged));
+	 break;
+      case NVS_FILE_TEMP:
+	 printf(" R%d", shader->GetDestID(shader, merged));
+	 break;
+      case NVS_FILE_RESULT:
+	 printf(" result.%s", (SFR_STRING(shader->GetDestID(shader, merged))));
+	 break;
+      default:
+	 printf(" BAD_RESULT_FILE");
+	 break;
+      }
+
+      if (mask != SMASK_ALL) {
+	 printf(".");
+	 if (mask & SMASK_X) printf("x");
+	 if (mask & SMASK_Y) printf("y");
+	 if (mask & SMASK_Z) printf("z");
+	 if (mask & SMASK_W) printf("w");
+      }
+   }
+
+   if (shader->SupportsConditional && shader->SupportsConditional(shader) &&
+       shader->GetConditionTest(shader)) {
+      shader->GetCondRegSwizzle(shader, swz);
+
+      printf(" (%s%d.%s%s%s%s)",
+	    NVS_COND_STRING(shader->GetCondition(shader)),
+	    shader->GetCondRegID(shader),
+	    SWZ_STRING(swz[NVS_SWZ_X]),
+	    SWZ_STRING(swz[NVS_SWZ_Y]),
+	    SWZ_STRING(swz[NVS_SWZ_Z]),
+	    SWZ_STRING(swz[NVS_SWZ_W])
+	    );
+   }
+
+   /* looping */
+   if (opi->flags & COUNT_ALL) {
+      printf(" { ");
+      if (opi->flags & COUNT_NUM) {
+	 printf("%d", shader->GetLoopCount(shader));
+      }
+      if (opi->flags & COUNT_IND) {
+	 printf(", %d", shader->GetLoopInitial(shader));
+      }
+      if (opi->flags & COUNT_INC) {
+	 printf(", %d", shader->GetLoopIncrement(shader));
+      }
+      printf(" }");
+   }
+
+   /* branching */
+   if (opi->flags & BRANCH_TR)
+      printf(" %d", shader->GetBranch(shader));
+   if (opi->flags & BRANCH_EL)
+      printf(" ELSE %d", shader->GetBranchElse(shader));
+   if (opi->flags & BRANCH_EN)
+      printf(" END %d", shader->GetBranchEnd(shader));
+
+   if (!(opi->flags & NODS) && opi->numsrc)
+      printf(",");
+   printf(" ");
+
+   for (i = 0; i < opi->numsrc; i++) {
+      if (shader->GetSourceAbs(shader, merged, i))
+	 printf("abs(");
+      if (shader->GetSourceNegate(shader, merged, i))
+	 printf("-");
+
+      file = shader->GetSourceFile(shader, merged, i);
+      switch (file) {
+      case NVS_FILE_TEMP:
+	 printf("R%d", shader->GetSourceID(shader, merged, i));
+	 break;
+      case NVS_FILE_CONST:
+	 if (shader->GetSourceIndexed(shader, merged, i)) {
+	    printf("c[A%d.%s + 0x%x]",
+		     shader->GetRelAddressRegID(shader),
+		     SWZ_STRING(shader->GetRelAddressSwizzle(shader)),
+		     shader->GetSourceID(shader, merged, i)
+		  );
+	 } else {
+	    float val[4];
+
+	    if (shader->GetSourceConstVal) {
+	       shader->GetSourceConstVal(shader, merged, i, val);
+	       printf("{ %.02f, %.02f, %.02f, %.02f }",
+		     val[0], val[1], val[2], val[3]);
+	    } else {
+	       printf("c[0x%x]", shader->GetSourceID(shader, merged, i));
+	    }
+	 }
+	 break;
+      case NVS_FILE_ATTRIB:
+	 if (shader->GetSourceIndexed(shader, merged, i)) {
+	    printf("attrib[A%d.%s + %d]",
+		     shader->GetRelAddressRegID(shader),
+		     SWZ_STRING(shader->GetRelAddressSwizzle(shader)),
+		     shader->GetSourceID(shader, merged, i)
+		  );
+	 }
+	 else {
+	    printf("attrib.%s",
+		     SFR_STRING(shader->GetSourceID(shader, merged, i))
+		  );
+	 }
+	 break;
+      case NVS_FILE_ADDRESS:
+	 printf("A%d", shader->GetRelAddressRegID(shader));
+	 break;
+      default:
+	 printf("UNKNOWN_SRC_FILE");
+	 break;
+      }
+
+      shader->GetSourceSwizzle(shader, merged, i, swz);
+      if (file != NVS_FILE_ADDRESS &&
+	  (swz[NVS_SWZ_X] != NVS_SWZ_X || swz[NVS_SWZ_Y] != NVS_SWZ_Y ||
+	   swz[NVS_SWZ_Z] != NVS_SWZ_Z || swz[NVS_SWZ_W] != NVS_SWZ_W)) {
+	 printf(".%s%s%s%s", SWZ_STRING(swz[NVS_SWZ_X]),
+	       		     SWZ_STRING(swz[NVS_SWZ_Y]),
+			     SWZ_STRING(swz[NVS_SWZ_Z]),
+			     SWZ_STRING(swz[NVS_SWZ_W]));
+      }
+
+      if (shader->GetSourceAbs(shader, merged, i))
+	 printf(")");
+      if (shader->GetSourceScale) {
+	 int scale = shader->GetSourceScale(shader, merged, i);
+	 if (scale > 1)
+	    printf("{scaled %dx}", scale);
+      }
+      if (i < (opi->numsrc - 1))
+	 printf(", ");
+   }
+
+   if (shader->IsLastInst(shader))
+      printf(" + END");
+}
+
+void
+nvsDisasmHWShader(nvsPtr nvs)
+{
+   nvsFunc *shader = nvs->func;
+   unsigned int iaddr = 0;
+
+   if (!nvs->program) {
+      fprintf(stderr, "No HW program present");
+      return;
+   }
+
+   shader->inst = nvs->program;
+   while (1) {
+      if (shader->inst >= (nvs->program + nvs->program_size)) {
+	 fprintf(stderr, "Reached end of program, but HW inst has no END");
+	 break;
+      }
+
+      printf("\t0x%08x:\n", shader->inst[0]);
+      printf("\t0x%08x:\n", shader->inst[1]);
+      printf("\t0x%08x:\n", shader->inst[2]);
+      printf("\t0x%08x:", shader->inst[3]);
+
+      printf("\n\t\tINST %d.0: ", iaddr);
+      nvsDisasmHWShaderOp(shader, 0);
+      if (shader->HasMergedInst(shader)) {
+	 printf("\n\t\tINST %d.1: ", iaddr);
+	 nvsDisasmHWShaderOp(shader, 1);
+      }
+      printf("\n");
+
+      if (shader->IsLastInst(shader))
+	 break;
+
+      shader->inst += shader->GetOffsetNext(shader);
+      iaddr++;
+   }
+
+   printf("\n");
+}
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_shader.h b/src/mesa/drivers/dri/nouveau/nouveau_shader.h
new file mode 100644
index 00000000000..b2df3546f62
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_shader.h
@@ -0,0 +1,441 @@
+#ifndef __SHADER_COMMON_H__
+#define __SHADER_COMMON_H__
+
+#include "mtypes.h"
+#include "bufferobj.h"
+
+typedef struct _nvsFunc nvsFunc;
+
+#define NVS_MAX_TEMPS   32
+#define NVS_MAX_ATTRIBS 16
+#define NVS_MAX_CONSTS  256
+#define NVS_MAX_ADDRESS 2
+#define NVS_MAX_INSNS   4096
+
+typedef struct _nvs_fragment_header {
+   struct _nvs_fragment_header *parent;
+   struct _nvs_fragment_header *prev;
+   struct _nvs_fragment_header *next;
+   enum {
+      NVS_INSTRUCTION,
+      NVS_BRANCH,
+      NVS_LOOP,
+      NVS_SUBROUTINE
+   } type;
+} nvsFragmentHeader;
+
+typedef union {
+	struct {
+		GLboolean uses_kil;
+		GLuint    num_regs;
+	} NV30FP;
+	struct {
+		uint32_t vp_in_reg;
+		uint32_t vp_out_reg;
+		uint32_t clip_enables;
+	} NV30VP;
+} nvsCardPriv;
+
+typedef struct _nouveauShader {
+   union {
+      struct gl_vertex_program vp;
+      struct gl_fragment_program fp;
+   } mesa;
+   GLcontext *ctx;
+   nvsFunc *func;
+
+   /* State of the final program */
+   GLboolean translated;
+   GLboolean on_hardware;
+   unsigned int *program;
+   unsigned int program_size;
+   unsigned int program_alloc_size;
+   unsigned int program_start_id;
+   unsigned int program_current;
+   struct gl_buffer_object *program_buffer;
+   int		inst_count;
+
+   nvsCardPriv card_priv;
+   int		vp_attrib_map[NVS_MAX_ATTRIBS];
+
+   struct {
+      GLboolean in_use;
+
+      GLfloat  *source_val;	/* NULL if invariant */
+      float	val[4];
+      /* Hardware-specific tracking, currently only nv30_fragprog
+       * makes use of it.
+       */
+      int	*hw_index;
+      int        hw_index_cnt;
+   } params[NVS_MAX_CONSTS];
+   int param_high;
+
+   /* Pass-private data */
+   void *pass_rec;
+
+   nvsFragmentHeader *program_tree;
+} nouveauShader, *nvsPtr;
+
+typedef enum {
+   NVS_FILE_NONE,
+   NVS_FILE_TEMP,
+   NVS_FILE_ATTRIB,
+   NVS_FILE_CONST,
+   NVS_FILE_RESULT,
+   NVS_FILE_ADDRESS,
+   NVS_FILE_UNKNOWN
+} nvsRegFile;
+
+typedef enum {
+   NVS_OP_UNKNOWN = 0,
+   NVS_OP_NOP,
+   NVS_OP_ABS, NVS_OP_ADD, NVS_OP_ARA, NVS_OP_ARL, NVS_OP_ARR,
+   NVS_OP_BRA, NVS_OP_BRK,
+   NVS_OP_CAL, NVS_OP_CMP, NVS_OP_COS,
+   NVS_OP_DDX, NVS_OP_DDY, NVS_OP_DIV, NVS_OP_DP2, NVS_OP_DP2A, NVS_OP_DP3,
+   NVS_OP_DP4, NVS_OP_DPH, NVS_OP_DST,
+   NVS_OP_EX2, NVS_OP_EXP,
+   NVS_OP_FLR, NVS_OP_FRC,
+   NVS_OP_IF,
+   NVS_OP_KIL,
+   NVS_OP_LG2, NVS_OP_LIT, NVS_OP_LOG, NVS_OP_LOOP, NVS_OP_LRP,
+   NVS_OP_MAD, NVS_OP_MAX, NVS_OP_MIN, NVS_OP_MOV, NVS_OP_MUL,
+   NVS_OP_NRM,
+   NVS_OP_PK2H, NVS_OP_PK2US, NVS_OP_PK4B, NVS_OP_PK4UB, NVS_OP_POW,
+   NVS_OP_POPA, NVS_OP_PUSHA,
+   NVS_OP_RCC, NVS_OP_RCP, NVS_OP_REP, NVS_OP_RET, NVS_OP_RFL, NVS_OP_RSQ,
+   NVS_OP_SCS, NVS_OP_SEQ, NVS_OP_SFL, NVS_OP_SGE, NVS_OP_SGT, NVS_OP_SIN,
+   NVS_OP_SLE, NVS_OP_SLT, NVS_OP_SNE, NVS_OP_SSG, NVS_OP_STR, NVS_OP_SUB,
+   NVS_OP_SWZ,
+   NVS_OP_TEX, NVS_OP_TXB, NVS_OP_TXD, NVS_OP_TXL, NVS_OP_TXP,
+   NVS_OP_UP2H, NVS_OP_UP2US, NVS_OP_UP4B, NVS_OP_UP4UB,
+   NVS_OP_X2D, NVS_OP_XPD,
+   NVS_OP_EMUL
+} nvsOpcode;
+
+typedef enum {
+   NVS_PREC_FLOAT32,
+   NVS_PREC_FLOAT16,
+   NVS_PREC_FIXED12,
+   NVS_PREC_UNKNOWN
+} nvsPrecision;
+
+typedef enum {
+   NVS_SWZ_X = 0,
+   NVS_SWZ_Y = 1,
+   NVS_SWZ_Z = 2,
+   NVS_SWZ_W = 3
+} nvsSwzComp;
+
+typedef enum {
+   NVS_FR_POSITION	= 0,
+   NVS_FR_WEIGHT	= 1,
+   NVS_FR_NORMAL	= 2,
+   NVS_FR_COL0		= 3,
+   NVS_FR_COL1		= 4,
+   NVS_FR_FOGCOORD	= 5,
+   NVS_FR_TEXCOORD0	= 8,
+   NVS_FR_TEXCOORD1	= 9,
+   NVS_FR_TEXCOORD2	= 10,
+   NVS_FR_TEXCOORD3	= 11,
+   NVS_FR_TEXCOORD4	= 12,
+   NVS_FR_TEXCOORD5	= 13,
+   NVS_FR_TEXCOORD6	= 14,
+   NVS_FR_TEXCOORD7	= 15,
+   NVS_FR_BFC0		= 16,
+   NVS_FR_BFC1		= 17,
+   NVS_FR_POINTSZ	= 18,
+   NVS_FR_FRAGDATA0	= 19,
+   NVS_FR_FRAGDATA1	= 20,
+   NVS_FR_FRAGDATA2	= 21,
+   NVS_FR_FRAGDATA3	= 22,
+   NVS_FR_CLIP0		= 23,
+   NVS_FR_CLIP1		= 24,
+   NVS_FR_CLIP2		= 25,
+   NVS_FR_CLIP3		= 26,
+   NVS_FR_CLIP4		= 27,
+   NVS_FR_CLIP5		= 28,
+   NVS_FR_CLIP6		= 29,
+   NVS_FR_FACING	= 30,
+   NVS_FR_UNKNOWN
+} nvsFixedReg;
+
+typedef enum {
+   NVS_COND_FL, NVS_COND_LT, NVS_COND_EQ, NVS_COND_LE, NVS_COND_GT,
+   NVS_COND_NE, NVS_COND_GE, NVS_COND_TR, NVS_COND_UN,
+   NVS_COND_UNKNOWN
+} nvsCond;
+
+typedef struct {
+   nvsRegFile	file;
+   unsigned int		index;
+
+   unsigned int		indexed;
+   unsigned int		addr_reg;
+   nvsSwzComp		addr_comp;
+
+   nvsSwzComp		swizzle[4];
+   int			negate;
+   int			abs;
+} nvsRegister;
+
+static const nvsRegister nvr_unused = {
+   .file	= NVS_FILE_ATTRIB,
+   .index	= 0,
+   .indexed	= 0,
+   .addr_reg	= 0,
+   .addr_comp	= NVS_SWZ_X,
+   .swizzle	= {NVS_SWZ_X, NVS_SWZ_Y, NVS_SWZ_Z, NVS_SWZ_W},
+   .negate	= 0,
+   .abs		= 0,
+};
+
+typedef enum {
+   NVS_TEX_TARGET_1D,
+   NVS_TEX_TARGET_2D,
+   NVS_TEX_TARGET_3D,
+   NVS_TEX_TARGET_CUBE,
+   NVS_TEX_TARGET_RECT,
+   NVS_TEX_TARGET_UNKNOWN = 0
+} nvsTexTarget;
+
+typedef enum {
+	NVS_SCALE_1X	 = 0,
+	NVS_SCALE_2X	 = 1,
+	NVS_SCALE_4X	 = 2,
+	NVS_SCALE_8X	 = 3,
+	NVS_SCALE_INV_2X = 5,
+	NVS_SCALE_INV_4X = 6,
+	NVS_SCALE_INV_8X = 7,
+} nvsScale;
+
+/* Arith/TEX instructions */
+typedef struct nvs_instruction {
+   nvsFragmentHeader header;
+
+   nvsOpcode	op;
+   unsigned int saturate;
+
+   nvsRegister	dest;
+   unsigned int	mask;
+   nvsScale	dest_scale;
+
+   nvsRegister	src[3];
+
+   unsigned int tex_unit;
+   nvsTexTarget tex_target;
+
+   nvsCond	cond;
+   nvsSwzComp	cond_swizzle[4];
+   int		cond_reg;
+   int		cond_test;
+   int		cond_update;
+} nvsInstruction;
+
+/* BRA, CAL, IF */
+typedef struct nvs_branch {
+	nvsFragmentHeader  header;
+
+	nvsOpcode	op;
+
+	nvsCond		cond;
+	nvsSwzComp	cond_swizzle[4];
+	int		cond_test;
+
+	nvsFragmentHeader *target_head;
+	nvsFragmentHeader *target_tail;
+	nvsFragmentHeader *else_head;
+	nvsFragmentHeader *else_tail;
+} nvsBranch;
+
+/* LOOP+ENDLOOP */
+typedef struct {
+	nvsFragmentHeader  header;
+
+	int                count;
+	int                initial;
+	int                increment;
+
+	nvsFragmentHeader *insn_head;
+	nvsFragmentHeader *insn_tail;
+} nvsLoop;
+
+/* label+following instructions */
+typedef struct nvs_subroutine {
+	nvsFragmentHeader  header;
+
+	char *             label;
+	nvsFragmentHeader *insn_head;
+	nvsFragmentHeader *insn_tail;
+} nvsSubroutine;
+
+#define SMASK_X (1<<0)
+#define SMASK_Y (1<<1)
+#define SMASK_Z (1<<2)
+#define SMASK_W (1<<3)
+#define SMASK_ALL (SMASK_X|SMASK_Y|SMASK_Z|SMASK_W)
+
+#define SPOS_ADDRESS 3
+struct _op_xlat {
+	unsigned int	NV;
+	nvsOpcode	SOP;
+	int		srcpos[3];
+};
+#define MOD_OPCODE(t,hw,sop,s0,s1,s2) do { \
+	t[hw].NV = hw; \
+	t[hw].SOP = sop; \
+	t[hw].srcpos[0] = s0; \
+	t[hw].srcpos[1] = s1; \
+	t[hw].srcpos[2] = s2; \
+} while(0)
+
+extern unsigned int NVVP_TX_VOP_COUNT;
+extern unsigned int NVVP_TX_NVS_OP_COUNT;
+extern struct _op_xlat NVVP_TX_VOP[];
+extern struct _op_xlat NVVP_TX_SOP[];
+
+extern unsigned int NVFP_TX_AOP_COUNT;
+extern unsigned int NVFP_TX_BOP_COUNT;
+extern struct _op_xlat NVFP_TX_AOP[];
+extern struct _op_xlat NVFP_TX_BOP[];
+
+extern void NV20VPTXSwizzle(int hwswz, nvsSwzComp *swz);
+extern nvsSwzComp NV20VP_TX_SWIZZLE[4];
+
+#define SCAP_SRC_ABS	(1<<0)
+
+struct _nvsFunc {
+   nvsCardPriv *card_priv;
+
+   unsigned int	MaxInst;
+   unsigned int	MaxAttrib;
+   unsigned int	MaxTemp;
+   unsigned int	MaxAddress;
+   unsigned int	MaxConst;
+   unsigned int	caps;
+
+   unsigned int *inst;
+   void		(*UploadToHW)		(GLcontext *, nouveauShader *);
+   void		(*UpdateConst)		(GLcontext *, nouveauShader *, int);
+
+   struct _op_xlat*(*GetOPTXRec)	(nvsFunc *, int merged);
+   struct _op_xlat*(*GetOPTXFromSOP)	(nvsOpcode, int *id);
+
+   void		(*InitInstruction)	(nvsFunc *);
+   int		(*SupportsOpcode)	(nvsFunc *, nvsOpcode);
+   int		(*SupportsResultScale)	(nvsFunc *, nvsScale);
+   void		(*SetOpcode)		(nvsFunc *, unsigned int opcode,
+	 				 int slot);
+   void		(*SetCCUpdate)		(nvsFunc *);
+   void		(*SetCondition)		(nvsFunc *, int on, nvsCond, int reg,
+	 				 nvsSwzComp *swizzle);
+   void		(*SetResult)		(nvsFunc *, nvsRegister *,
+	 				 unsigned int mask, int slot);
+   void		(*SetResultScale)	(nvsFunc *, nvsScale);
+   void		(*SetSource)		(nvsFunc *, nvsRegister *, int pos);
+   void		(*SetTexImageUnit)	(nvsFunc *, int unit);
+   void		(*SetSaturate)		(nvsFunc *);
+   void		(*SetLastInst)		(nvsFunc *);
+
+   void		(*SetBranchTarget)	(nvsFunc *, int addr);
+   void		(*SetBranchElse)	(nvsFunc *, int addr);
+   void		(*SetBranchEnd)		(nvsFunc *, int addr);
+   void		(*SetLoopParams)	(nvsFunc *, int cnt, int init, int inc);
+
+   int		(*HasMergedInst)	(nvsFunc *);
+   int		(*IsLastInst)		(nvsFunc *);
+   int		(*GetOffsetNext)	(nvsFunc *);
+
+   int		(*GetOpcodeSlot)	(nvsFunc *, int merged);
+   unsigned int	(*GetOpcodeHW)		(nvsFunc *, int slot);
+   nvsOpcode	(*GetOpcode)		(nvsFunc *, int merged);
+
+   nvsPrecision	(*GetPrecision)		(nvsFunc *);
+   int		(*GetSaturate)		(nvsFunc *);
+
+   nvsRegFile	(*GetDestFile)		(nvsFunc *, int merged);
+   unsigned int	(*GetDestID)		(nvsFunc *, int merged);
+   unsigned int	(*GetDestMask)		(nvsFunc *, int merged);
+
+   unsigned int	(*GetSourceHW)		(nvsFunc *, int merged, int pos);
+   nvsRegFile	(*GetSourceFile)	(nvsFunc *, int merged, int pos);
+   int		(*GetSourceID)		(nvsFunc *, int merged, int pos);
+   int		(*GetTexImageUnit)	(nvsFunc *);
+   int		(*GetSourceNegate)	(nvsFunc *, int merged, int pos);
+   int		(*GetSourceAbs)		(nvsFunc *, int merged, int pos);
+   void		(*GetSourceSwizzle)	(nvsFunc *, int merged, int pos,
+	 				 nvsSwzComp *swz);
+   int		(*GetSourceIndexed)	(nvsFunc *, int merged, int pos);
+   void		(*GetSourceConstVal)	(nvsFunc *, int merged, int pos,
+	 				 float *val);
+   int		(*GetSourceScale)	(nvsFunc *, int merged, int pos);
+
+   int		(*GetRelAddressRegID)	(nvsFunc *);
+   nvsSwzComp	(*GetRelAddressSwizzle)	(nvsFunc *);
+
+   int		(*SupportsConditional)	(nvsFunc *);
+   int		(*GetConditionUpdate)	(nvsFunc *);
+   int		(*GetConditionTest)	(nvsFunc *);
+   nvsCond	(*GetCondition)		(nvsFunc *);
+   void		(*GetCondRegSwizzle)	(nvsFunc *, nvsSwzComp *swz);
+   int		(*GetCondRegID)		(nvsFunc *);
+   int		(*GetBranch)		(nvsFunc *);
+   int		(*GetBranchElse)	(nvsFunc *);
+   int		(*GetBranchEnd)		(nvsFunc *);
+
+   int		(*GetLoopCount)		(nvsFunc *);
+   int		(*GetLoopInitial)	(nvsFunc *);
+   int		(*GetLoopIncrement)	(nvsFunc *);
+};
+
+static inline nvsRegister
+nvsNegate(nvsRegister reg)
+{
+   reg.negate = !reg.negate;
+   return reg;
+}
+
+static inline nvsRegister
+nvsAbs(nvsRegister reg)
+{
+   reg.abs = 1;
+   return reg;
+}
+
+static inline nvsRegister
+nvsSwizzle(nvsRegister reg, nvsSwzComp x, nvsSwzComp y,
+      	   nvsSwzComp z, nvsSwzComp w)
+{
+   nvsSwzComp sc[4] = { x, y, z, w };
+   nvsSwzComp oc[4];
+   int i;
+
+   for (i=0;i<4;i++)
+      oc[i] = reg.swizzle[i];
+   for (i=0;i<4;i++)
+      reg.swizzle[i] = oc[sc[i]];
+   return reg;
+}
+
+extern GLboolean nvsUpdateShader(GLcontext *ctx, nouveauShader *nvs);
+extern void nvsDisasmHWShader(nvsPtr);
+extern void nvsDumpFragmentList(nvsFragmentHeader *f, int lvl);
+extern nouveauShader *nvsBuildTextShader(GLcontext *ctx, GLenum target,
+      					 const char *text);
+
+extern void NV20VPInitShaderFuncs(nvsFunc *);
+extern void NV30VPInitShaderFuncs(nvsFunc *);
+extern void NV40VPInitShaderFuncs(nvsFunc *);
+
+extern void NV30FPInitShaderFuncs(nvsFunc *);
+extern void NV40FPInitShaderFuncs(nvsFunc *);
+
+extern void nouveauShaderInitFuncs(GLcontext *ctx);
+
+extern GLboolean nouveau_shader_pass0(GLcontext *ctx, nouveauShader *nvs);
+extern GLboolean nouveau_shader_pass1(nvsPtr nvs);
+extern GLboolean nouveau_shader_pass2(nvsPtr nvs);
+
+#endif
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_shader_0.c b/src/mesa/drivers/dri/nouveau/nouveau_shader_0.c
new file mode 100644
index 00000000000..73c1f7c2a53
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_shader_0.c
@@ -0,0 +1,1006 @@
+/*
+ * Copyright (C) 2006 Ben Skeggs.
+ *
+ * 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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
+ *   Ben Skeggs <darktama@iinet.net.au>
+ */
+
+#include "glheader.h"
+#include "macros.h"
+#include "enums.h"
+
+#include "program.h"
+#include "programopt.h"
+#include "program_instruction.h"
+
+#include "nouveau_context.h"
+#include "nouveau_shader.h"
+#include "nouveau_msg.h"
+
+static nvsFixedReg _tx_mesa_vp_dst_reg[VERT_RESULT_MAX] = {
+   NVS_FR_POSITION, NVS_FR_COL0, NVS_FR_COL1, NVS_FR_FOGCOORD,
+   NVS_FR_TEXCOORD0, NVS_FR_TEXCOORD1, NVS_FR_TEXCOORD2, NVS_FR_TEXCOORD3,
+   NVS_FR_TEXCOORD4, NVS_FR_TEXCOORD5, NVS_FR_TEXCOORD6, NVS_FR_TEXCOORD7,
+   NVS_FR_POINTSZ, NVS_FR_BFC0, NVS_FR_BFC1, NVS_FR_UNKNOWN /* EDGE */
+};
+
+static nvsFixedReg _tx_mesa_fp_dst_reg[FRAG_RESULT_MAX] = {
+   NVS_FR_FRAGDATA0 /* COLR */, NVS_FR_FRAGDATA0 /* COLH */,
+   NVS_FR_UNKNOWN /* DEPR */
+};
+
+static nvsFixedReg _tx_mesa_fp_src_reg[FRAG_ATTRIB_MAX] = {
+   NVS_FR_POSITION, NVS_FR_COL0, NVS_FR_COL1, NVS_FR_FOGCOORD,
+   NVS_FR_TEXCOORD0, NVS_FR_TEXCOORD1, NVS_FR_TEXCOORD2, NVS_FR_TEXCOORD3,
+   NVS_FR_TEXCOORD4, NVS_FR_TEXCOORD5, NVS_FR_TEXCOORD6, NVS_FR_TEXCOORD7
+};
+
+static nvsSwzComp _tx_mesa_swizzle[4] = {
+   NVS_SWZ_X, NVS_SWZ_Y, NVS_SWZ_Z, NVS_SWZ_W
+};
+
+static nvsOpcode _tx_mesa_opcode[] = {
+   [OPCODE_ABS] = NVS_OP_ABS, [OPCODE_ADD] = NVS_OP_ADD,
+   [OPCODE_ARA] = NVS_OP_ARA, [OPCODE_ARL] = NVS_OP_ARL,
+   [OPCODE_ARL_NV] = NVS_OP_ARL, [OPCODE_ARR] = NVS_OP_ARR,
+   [OPCODE_CMP] = NVS_OP_CMP, [OPCODE_COS] = NVS_OP_COS,
+   [OPCODE_DDX] = NVS_OP_DDX, [OPCODE_DDY] = NVS_OP_DDY,
+   [OPCODE_DP3] = NVS_OP_DP3, [OPCODE_DP4] = NVS_OP_DP4,
+   [OPCODE_DPH] = NVS_OP_DPH, [OPCODE_DST] = NVS_OP_DST,
+   [OPCODE_EX2] = NVS_OP_EX2, [OPCODE_EXP] = NVS_OP_EXP,
+   [OPCODE_FLR] = NVS_OP_FLR, [OPCODE_FRC] = NVS_OP_FRC,
+   [OPCODE_KIL] = NVS_OP_EMUL, [OPCODE_KIL_NV] = NVS_OP_KIL,
+   [OPCODE_LG2] = NVS_OP_LG2, [OPCODE_LIT] = NVS_OP_LIT,
+   [OPCODE_LOG] = NVS_OP_LOG,
+   [OPCODE_LRP] = NVS_OP_LRP,
+   [OPCODE_MAD] = NVS_OP_MAD, [OPCODE_MAX] = NVS_OP_MAX,
+   [OPCODE_MIN] = NVS_OP_MIN, [OPCODE_MOV] = NVS_OP_MOV,
+   [OPCODE_MUL] = NVS_OP_MUL,
+   [OPCODE_PK2H] = NVS_OP_PK2H, [OPCODE_PK2US] = NVS_OP_PK2US,
+   [OPCODE_PK4B] = NVS_OP_PK4B, [OPCODE_PK4UB] = NVS_OP_PK4UB,
+   [OPCODE_POW] = NVS_OP_POW, [OPCODE_POPA] = NVS_OP_POPA,
+   [OPCODE_PUSHA] = NVS_OP_PUSHA,
+   [OPCODE_RCC] = NVS_OP_RCC, [OPCODE_RCP] = NVS_OP_RCP,
+   [OPCODE_RFL] = NVS_OP_RFL, [OPCODE_RSQ] = NVS_OP_RSQ,
+   [OPCODE_SCS] = NVS_OP_SCS, [OPCODE_SEQ] = NVS_OP_SEQ,
+   [OPCODE_SFL] = NVS_OP_SFL, [OPCODE_SGE] = NVS_OP_SGE,
+   [OPCODE_SGT] = NVS_OP_SGT, [OPCODE_SIN] = NVS_OP_SIN,
+   [OPCODE_SLE] = NVS_OP_SLE, [OPCODE_SLT] = NVS_OP_SLT,
+   [OPCODE_SNE] = NVS_OP_SNE, [OPCODE_SSG] = NVS_OP_SSG,
+   [OPCODE_STR] = NVS_OP_STR, [OPCODE_SUB] = NVS_OP_SUB,
+   [OPCODE_SWZ] = NVS_OP_MOV,
+   [OPCODE_TEX] = NVS_OP_TEX, [OPCODE_TXB] = NVS_OP_TXB,
+   [OPCODE_TXD] = NVS_OP_TXD,
+   [OPCODE_TXL] = NVS_OP_TXL, [OPCODE_TXP] = NVS_OP_TXP,
+   [OPCODE_TXP_NV] = NVS_OP_TXP,
+   [OPCODE_UP2H] = NVS_OP_UP2H, [OPCODE_UP2US] = NVS_OP_UP2US,
+   [OPCODE_UP4B] = NVS_OP_UP4B, [OPCODE_UP4UB] = NVS_OP_UP4UB,
+   [OPCODE_X2D] = NVS_OP_X2D,
+   [OPCODE_XPD] = NVS_OP_XPD
+};
+
+static nvsCond _tx_mesa_condmask[] = {
+   NVS_COND_TR, /* workaround mesa not filling a valid value */
+   NVS_COND_GT, NVS_COND_LT, NVS_COND_UN, NVS_COND_GE,
+   NVS_COND_LE, NVS_COND_NE, NVS_COND_NE, NVS_COND_TR, NVS_COND_FL
+};
+
+struct pass0_rec {
+   int nvs_ipos;
+   int next_temp;
+
+   int mesa_const_base;
+   int mesa_const_last;
+
+   int swzconst_done;
+   int swzconst_id;
+   nvsRegister const_half;
+};
+
+#define X NVS_SWZ_X
+#define Y NVS_SWZ_Y
+#define Z NVS_SWZ_Z
+#define W NVS_SWZ_W
+
+#define FILL_CONDITION_FLAGS(fragment) do {                                    \
+	(fragment)->cond =                                                     \
+	pass0_make_condmask(inst->DstReg.CondMask);                            \
+	if ((fragment)->cond != NVS_COND_TR)                                   \
+		(fragment)->cond_test = 1;                                     \
+	(fragment)->cond_reg = inst->CondDst;                                  \
+	pass0_make_swizzle((fragment)->cond_swizzle, inst->DstReg.CondSwizzle);\
+} while(0)
+
+#define ARITH(op,dest,mask,sat,s0,s1,s2) do {               \
+	nvsinst = pass0_emit(nvs, parent, fpos, (op),       \
+		   (dest), (mask), (sat), (s0), (s1), (s2));\
+	FILL_CONDITION_FLAGS(nvsinst);                      \
+} while(0)
+
+#define ARITHu(op,dest,mask,sat,s0,s1,s2) do {              \
+	nvsinst = pass0_emit(nvs, parent, fpos, (op),       \
+		   (dest), (mask), (sat), (s0), (s1), (s2));\
+} while(0)
+
+static void
+pass0_append_fragment(nvsFragmentHeader *parent,
+		      nvsFragmentHeader *fragment,
+		      int pos)
+{
+	nvsFragmentHeader **head, **tail;
+	assert(parent && fragment);
+
+	switch (parent->type) {
+	case NVS_BRANCH:
+		if (pos == 0) {
+			head = &((nvsBranch *)parent)->target_head;
+			tail = &((nvsBranch *)parent)->target_tail;
+		} else {
+			head = &((nvsBranch *)parent)->else_head;
+			tail = &((nvsBranch *)parent)->else_tail;
+		}
+		break;
+	case NVS_LOOP:
+		head = &((nvsLoop *)parent)->insn_head;
+		tail = &((nvsLoop *)parent)->insn_tail;
+		break;
+	case NVS_SUBROUTINE:
+		head = &((nvsSubroutine *)parent)->insn_head;
+		tail = &((nvsSubroutine *)parent)->insn_tail;
+		break;
+	default:
+		assert(0);
+		break;
+	}
+
+	fragment->parent = parent;
+	fragment->prev   = *tail;
+	fragment->next   = NULL;
+	if (!(*head))
+		*head = fragment;
+	else
+		(*tail)->next = fragment;
+	*tail = fragment;
+
+}
+
+static nvsSubroutine *
+pass0_create_subroutine(nouveauShader *nvs, const char *label)
+{
+	nvsSubroutine *sub;
+
+	sub = CALLOC_STRUCT(nvs_subroutine);
+	if (sub) {
+		sub->header.type = NVS_SUBROUTINE;
+		sub->label	 = strdup(label);
+		if (!nvs->program_tree)
+			nvs->program_tree = &sub->header;
+		else
+			pass0_append_fragment(nvs->program_tree,
+					      &sub->header, 0);
+	}
+
+	return sub;
+}
+
+static void
+pass0_make_reg(nouveauShader *nvs, nvsRegister *reg,
+	       nvsRegFile file, unsigned int index)
+{
+	struct pass0_rec *rec = nvs->pass_rec;
+
+	/* defaults */
+	*reg = nvr_unused;
+	/* -1 == quick-and-dirty temp alloc */
+	if (file == NVS_FILE_TEMP && index == -1) {
+		index = rec->next_temp++;
+		assert(index < NVS_MAX_TEMPS);
+	}
+	reg->file   = file;
+	reg->index  = index;
+}
+
+static void
+pass0_make_swizzle(nvsSwzComp *swz, unsigned int mesa)
+{
+	int i;
+
+	for (i=0;i<4;i++)
+		swz[i] = _tx_mesa_swizzle[GET_SWZ(mesa, i)];
+}
+
+static nvsOpcode
+pass0_make_opcode(enum prog_opcode op)
+{
+	if (op > MAX_OPCODE)
+		return NVS_OP_UNKNOWN;
+	return _tx_mesa_opcode[op];
+}
+
+static nvsCond
+pass0_make_condmask(GLuint mesa)
+{
+	if (mesa > COND_FL)
+		return NVS_COND_UNKNOWN;
+	return _tx_mesa_condmask[mesa];
+}
+
+static unsigned int
+pass0_make_mask(GLuint mesa_mask)
+{
+	unsigned int mask = 0;
+
+	if (mesa_mask & WRITEMASK_X) mask |= SMASK_X;
+	if (mesa_mask & WRITEMASK_Y) mask |= SMASK_Y;
+	if (mesa_mask & WRITEMASK_Z) mask |= SMASK_Z;
+	if (mesa_mask & WRITEMASK_W) mask |= SMASK_W;
+
+	return mask;
+}
+
+static nvsTexTarget
+pass0_make_tex_target(GLuint mesa)
+{
+	switch (mesa) {
+	case TEXTURE_1D_INDEX: return NVS_TEX_TARGET_1D;
+	case TEXTURE_2D_INDEX: return NVS_TEX_TARGET_2D;
+	case TEXTURE_3D_INDEX: return NVS_TEX_TARGET_3D;
+	case TEXTURE_CUBE_INDEX: return NVS_TEX_TARGET_CUBE;
+	case TEXTURE_RECT_INDEX: return NVS_TEX_TARGET_RECT;
+	default:
+		return NVS_TEX_TARGET_UNKNOWN;
+	}
+}
+
+static void
+pass0_make_dst_reg(nvsPtr nvs, nvsRegister *reg,
+      		   struct prog_dst_register *dst)
+{
+	struct gl_program *mesa = (struct gl_program*)&nvs->mesa.vp;
+	nvsFixedReg sfr;
+
+	switch (dst->File) {
+	case PROGRAM_OUTPUT:
+		if (mesa->Target == GL_VERTEX_PROGRAM_ARB) {
+			sfr = (dst->Index < VERT_RESULT_MAX) ?
+				_tx_mesa_vp_dst_reg[dst->Index] :
+				NVS_FR_UNKNOWN;
+		} else {
+			sfr = (dst->Index < FRAG_RESULT_MAX) ?
+				_tx_mesa_fp_dst_reg[dst->Index] :
+				NVS_FR_UNKNOWN;
+		}
+		pass0_make_reg(nvs, reg, NVS_FILE_RESULT, sfr);
+		break;
+	case PROGRAM_TEMPORARY:
+		pass0_make_reg(nvs, reg, NVS_FILE_TEMP, dst->Index);
+		break;
+	case PROGRAM_ADDRESS:
+		pass0_make_reg(nvs, reg, NVS_FILE_ADDRESS, dst->Index);
+		break;
+	default:
+		fprintf(stderr, "Unknown dest file %d\n", dst->File);
+		assert(0);
+	}
+}
+
+static void
+pass0_make_src_reg(nvsPtr nvs, nvsRegister *reg, struct prog_src_register *src)
+{
+	struct pass0_rec *rec = nvs->pass_rec;
+	struct gl_program *mesa = (struct gl_program *)&nvs->mesa.vp.Base;
+	int i;
+
+	*reg = nvr_unused;
+
+	switch (src->File) {
+	case PROGRAM_INPUT:
+		reg->file = NVS_FILE_ATTRIB;
+		if (mesa->Target == GL_VERTEX_PROGRAM_ARB) {
+			for (i=0; i<NVS_MAX_ATTRIBS; i++) {
+				if (nvs->vp_attrib_map[i] == src->Index) {
+					reg->index = i;
+					break;
+				}
+			}
+			if (i==NVS_MAX_ATTRIBS)
+				reg->index = NVS_FR_UNKNOWN;
+		} else {
+			reg->index = (src->Index < FRAG_ATTRIB_MAX) ?
+				_tx_mesa_fp_src_reg[src->Index] :
+				NVS_FR_UNKNOWN;
+		}
+		break;
+	case PROGRAM_STATE_VAR:
+	case PROGRAM_NAMED_PARAM:
+	case PROGRAM_CONSTANT:
+		reg->file    = NVS_FILE_CONST;
+		reg->index   = src->Index + rec->mesa_const_base;
+		reg->indexed = src->RelAddr;
+		if (reg->indexed) {
+			reg->addr_reg  = 0;
+			reg->addr_comp = NVS_SWZ_X;
+		}
+		break;
+	case PROGRAM_TEMPORARY:
+		reg->file  = NVS_FILE_TEMP;
+		reg->index = src->Index;
+		break;
+	default:
+		fprintf(stderr, "Unknown source type %d\n", src->File);
+		assert(0);
+	}
+
+	/* per-component negate handled elsewhere */
+	reg->negate = src->NegateBase != 0;
+	reg->abs    = src->Abs;
+	pass0_make_swizzle(reg->swizzle, src->Swizzle);
+}
+
+static nvsInstruction *
+pass0_emit(nouveauShader *nvs, nvsFragmentHeader *parent, int fpos,
+	   nvsOpcode op, nvsRegister dst,
+	   unsigned int mask, int saturate,
+	   nvsRegister src0, nvsRegister src1, nvsRegister src2)
+{
+	nvsInstruction *sif;
+
+	sif = CALLOC_STRUCT(nvs_instruction);
+	if (!sif)
+		return NULL;
+
+	/* Seems mesa doesn't explicitly 0 this.. */
+	if (nvs->mesa.vp.Base.Target == GL_VERTEX_PROGRAM_ARB)
+		saturate = 0;
+
+	sif->op		= op;
+	sif->saturate	= saturate;
+	sif->dest	= dst;
+	sif->mask	= mask;
+	sif->dest_scale	= NVS_SCALE_1X;
+	sif->src[0]	= src0;
+	sif->src[1]	= src1;
+	sif->src[2]	= src2;
+	sif->cond	= COND_TR;
+	sif->cond_reg	= 0;
+	sif->cond_test	= 0;
+	sif->cond_update= 0;
+	pass0_make_swizzle(sif->cond_swizzle, SWIZZLE_NOOP);
+	pass0_append_fragment(parent, &sif->header, fpos);
+
+	return sif;
+}
+
+static void
+pass0_fixup_swizzle(nvsPtr nvs, nvsFragmentHeader *parent, int fpos,
+		    struct prog_src_register *src,
+		    unsigned int sm1,
+		    unsigned int sm2)
+{
+	static const float sc[4] = { 1.0, 0.0, -1.0, 0.0 };
+	struct pass0_rec *rec = nvs->pass_rec;
+	int fixup_1, fixup_2;
+	nvsInstruction *nvsinst;
+	nvsRegister sr, dr = nvr_unused;
+	nvsRegister sm1const, sm2const;
+
+	if (!rec->swzconst_done) {
+		struct gl_program *prog = &nvs->mesa.vp.Base;
+		rec->swzconst_id = _mesa_add_unnamed_constant(prog->Parameters,
+							      sc, 4);
+		rec->swzconst_done = 1;
+		COPY_4V(nvs->params[rec->swzconst_id].val, sc);
+	}
+
+	fixup_1 = (sm1 != MAKE_SWIZZLE4(0,0,0,0) &&
+			sm2 != MAKE_SWIZZLE4(2,2,2,2));
+	fixup_2 = (sm2 != MAKE_SWIZZLE4(2,2,2,2));
+
+	if (src->File != PROGRAM_TEMPORARY && src->File != PROGRAM_INPUT) {
+		/* We can't use more than one const in an instruction,
+		 * so move the const into a temp, and swizzle from there.
+		 *
+		 * TODO: should just emit the swizzled const, instead of
+		 *       swizzling it in the shader.. would need to reswizzle
+		 *       any state params when they change however..
+		 */
+		pass0_make_reg(nvs, &dr, NVS_FILE_TEMP, -1);
+		pass0_make_src_reg(nvs, &sr, src);
+		ARITHu(NVS_OP_MOV, dr, SMASK_ALL, 0,
+				 sr, nvr_unused, nvr_unused);
+		pass0_make_reg(nvs, &sr, NVS_FILE_TEMP, dr.index);
+	} else {
+		if (fixup_1)
+			src->NegateBase = 0;
+		pass0_make_src_reg(nvs, &sr, src);
+		pass0_make_reg(nvs, &dr, NVS_FILE_TEMP, -1);
+	}
+
+	pass0_make_reg(nvs, &sm1const, NVS_FILE_CONST, rec->swzconst_id);
+	pass0_make_swizzle(sm1const.swizzle, sm1);
+	if (fixup_1 && fixup_2) {
+		/* Any combination with SWIZZLE_ONE */
+		pass0_make_reg(nvs, &sm2const,
+				    NVS_FILE_CONST, rec->swzconst_id);
+		pass0_make_swizzle(sm2const.swizzle, sm2);
+		ARITHu(NVS_OP_MAD, dr, SMASK_ALL, 0, sr, sm1const, sm2const);
+	} else {
+		/* SWIZZLE_ZERO || arbitrary negate */
+		ARITHu(NVS_OP_MUL, dr, SMASK_ALL, 0, sr, sm1const, nvr_unused);
+	}
+
+	src->File	= PROGRAM_TEMPORARY;
+	src->Index	= dr.index;
+	src->Swizzle	= SWIZZLE_NOOP;
+}
+
+#define SET_SWZ(fs, cp, c) fs = (fs & ~(0x7<<(cp*3))) | (c<<(cp*3))
+static void
+pass0_check_sources(nvsPtr nvs, nvsFragmentHeader *parent, int fpos,
+		    struct prog_instruction *inst)
+{
+	unsigned int insrc = -1, constsrc = -1;
+	int i;
+
+	for (i=0;i<_mesa_num_inst_src_regs(inst->Opcode);i++) {
+		struct prog_src_register *src = &inst->SrcReg[i];
+		unsigned int sm_1 = 0, sm_2 = 0;
+		nvsRegister sr, dr;
+		int do_mov = 0, c;
+
+		/* Build up swizzle masks as if we were going to use
+		 * "MAD new, src, const1, const2" to support arbitrary negation
+		 * and SWIZZLE_ZERO/SWIZZLE_ONE.
+		 */
+		for (c=0;c<4;c++) {
+			if (GET_SWZ(src->Swizzle, c) == SWIZZLE_ZERO) {
+				SET_SWZ(sm_1, c, SWIZZLE_Y); /* 0.0 */
+				SET_SWZ(sm_2, c, SWIZZLE_Y);
+				SET_SWZ(src->Swizzle, c, SWIZZLE_X);
+			} else if (GET_SWZ(src->Swizzle, c) == SWIZZLE_ONE) {
+				SET_SWZ(sm_1, c, SWIZZLE_Y);
+				if (src->NegateBase & (1<<c))
+					SET_SWZ(sm_2, c, SWIZZLE_Z); /* -1.0 */
+				else
+					SET_SWZ(sm_2, c, SWIZZLE_X); /* 1.0 */
+				SET_SWZ(src->Swizzle, c, SWIZZLE_X);
+			} else {
+				if (src->NegateBase & (1<<c))
+					SET_SWZ(sm_1, c, SWIZZLE_Z); /* -[xyzw] */
+				else
+					SET_SWZ(sm_1, c, SWIZZLE_X); /*[xyzw]*/
+				SET_SWZ(sm_2, c, SWIZZLE_Y);
+			}
+		}
+
+		/* Unless we're multiplying by 1.0 or -1.0 on all components,
+		 * and we're adding nothing to any component we have to
+		 * emulate the swizzle.
+		 */
+		if ((sm_1 != MAKE_SWIZZLE4(0,0,0,0) &&
+					sm_1 != MAKE_SWIZZLE4(2,2,2,2)) ||
+				sm_2 != MAKE_SWIZZLE4(1,1,1,1)) {
+			pass0_fixup_swizzle(nvs, parent, fpos, src, sm_1, sm_2);
+			/* The source is definitely in a temp now, so don't
+			 * bother checking  for multiple ATTRIB/CONST regs.
+			 */
+			continue;
+		}
+
+		/* HW can't use more than one ATTRIB or PARAM in a single
+		 * instruction */
+		switch (src->File) {
+		case PROGRAM_INPUT:
+			if (insrc != -1 && insrc != src->Index) 
+				do_mov = 1;
+			else insrc = src->Index;
+			break;
+		case PROGRAM_STATE_VAR:
+			if (constsrc != -1 && constsrc != src->Index)
+				do_mov = 1;
+			else constsrc = src->Index;
+			break;
+		default:
+			break;
+		}
+
+		/* Emit any extra ATTRIB/CONST to a temp, and modify the Mesa
+		 * instruction to point at the temp.
+		 */
+		if (do_mov) {
+			pass0_make_src_reg(nvs, &sr, src);
+			pass0_make_reg(nvs, &dr, NVS_FILE_TEMP, -1);
+			pass0_emit(nvs, parent, fpos, NVS_OP_MOV,
+					dr, SMASK_ALL, 0,
+					sr, nvr_unused, nvr_unused);
+
+			src->File	= PROGRAM_TEMPORARY;
+			src->Index	= dr.index;
+			src->Swizzle= SWIZZLE_NOOP;
+		}
+	}
+}
+
+static GLboolean
+pass0_emulate_instruction(nouveauShader *nvs,
+			  nvsFragmentHeader *parent, int fpos,
+			  struct prog_instruction *inst)
+{
+	nvsFunc *shader = nvs->func;
+	nvsRegister src[3], dest, temp;
+	nvsInstruction *nvsinst;
+	unsigned int mask = pass0_make_mask(inst->DstReg.WriteMask);
+	int i, sat;
+
+	sat = (inst->SaturateMode == SATURATE_ZERO_ONE);
+
+	/* Build all the "real" regs for the instruction */
+	for (i=0; i<_mesa_num_inst_src_regs(inst->Opcode); i++)
+		pass0_make_src_reg(nvs, &src[i], &inst->SrcReg[i]);
+	if (inst->Opcode != OPCODE_KIL)
+		pass0_make_dst_reg(nvs, &dest, &inst->DstReg);
+
+	switch (inst->Opcode) {
+	case OPCODE_ABS:
+		if (shader->caps & SCAP_SRC_ABS)
+			ARITH(NVS_OP_MOV, dest, mask, sat,
+					nvsAbs(src[0]), nvr_unused, nvr_unused);
+		else
+			ARITH(NVS_OP_MAX, dest, mask, sat,
+					src[0], nvsNegate(src[0]), nvr_unused);
+		break;
+	case OPCODE_CMP:
+		/*XXX: this will clobber CC0... */
+		ARITH (NVS_OP_MOV, dest, mask, sat,
+				src[2], nvr_unused, nvr_unused);
+		pass0_make_reg(nvs, &temp, NVS_FILE_TEMP, -1);
+		ARITHu(NVS_OP_MOV, temp, SMASK_ALL, 0,
+				src[0], nvr_unused, nvr_unused);
+		nvsinst->cond_update = 1;
+		nvsinst->cond_reg    = 0;
+		ARITH (NVS_OP_MOV, dest, mask, sat,
+				src[1], nvr_unused, nvr_unused);
+		nvsinst->cond      = COND_LT;
+		nvsinst->cond_reg  = 0;
+		nvsinst->cond_test = 1;
+		break;
+	case OPCODE_DPH:
+		pass0_make_reg(nvs, &temp, NVS_FILE_TEMP, -1);
+		ARITHu(NVS_OP_DP3, temp, SMASK_X, 0,
+				 src[0], src[1], nvr_unused);
+		ARITH (NVS_OP_ADD, dest, mask, sat,
+				 nvsSwizzle(temp, X, X, X, X),
+				 nvsSwizzle(src[1], W, W, W, W),
+				 nvr_unused);
+		break;
+	case OPCODE_KIL:
+		/* This is only in ARB shaders, so we don't have to worry
+		 * about clobbering a CC reg as they aren't supported anyway.
+		 *XXX: might have to worry with GLSL however...
+		 */
+		/* MOVC0 temp, src */
+		pass0_make_reg(nvs, &temp, NVS_FILE_TEMP, -1);
+		ARITHu(NVS_OP_MOV, temp, SMASK_ALL, 0,
+				 src[0], nvr_unused, nvr_unused);
+		nvsinst->cond_update = 1;
+		nvsinst->cond_reg    = 0;
+		/* KIL_NV (LT0.xyzw) temp */
+		ARITHu(NVS_OP_KIL, nvr_unused, 0, 0,
+				 nvr_unused, nvr_unused, nvr_unused);
+		nvsinst->cond      = COND_LT;
+		nvsinst->cond_reg  = 0;
+		nvsinst->cond_test = 1;
+		pass0_make_swizzle(nvsinst->cond_swizzle, SWIZZLE_NOOP);
+		break;
+	case OPCODE_LRP:
+		pass0_make_reg(nvs, &temp, NVS_FILE_TEMP, -1);
+		ARITHu(NVS_OP_MAD, temp, mask, 0,
+				 nvsNegate(src[0]), src[2], src[2]);
+		ARITH (NVS_OP_MAD, dest, mask, sat, src[0], src[1], temp);
+		break;
+	case OPCODE_POW:
+		if (shader->SupportsOpcode(shader, NVS_OP_LG2) &&
+				shader->SupportsOpcode(shader, NVS_OP_EX2)) {
+			pass0_make_reg(nvs, &temp, NVS_FILE_TEMP, -1);
+			/* LG2 temp.x, src0.c */
+			ARITHu(NVS_OP_LG2, temp, SMASK_X, 0,
+					 nvsSwizzle(src[0], X, X, X, X),
+					 nvr_unused, nvr_unused);
+			/* MUL temp.x, temp.x, src1.c */
+			ARITHu(NVS_OP_MUL, temp, SMASK_X, 0,
+					 nvsSwizzle(temp, X, X, X, X),
+					 nvsSwizzle(src[1], X, X, X, X),
+					 nvr_unused);
+			/* EX2 dest, temp.x */
+			ARITH (NVS_OP_EX2, dest, mask, sat,
+					 nvsSwizzle(temp, X, X, X, X),
+					 nvr_unused, nvr_unused);
+		} else {
+			/* can we use EXP/LOG instead of EX2/LG2?? */
+			fprintf(stderr, "Implement POW for NV20 vtxprog!\n");
+			return GL_FALSE;
+		}
+		break;
+	case OPCODE_RSQ:
+		pass0_make_reg(nvs, &temp, NVS_FILE_TEMP, -1);
+		ARITHu(NVS_OP_LG2, temp, SMASK_X, 0,
+				 nvsAbs(nvsSwizzle(src[0], X, X, X, X)),
+				 nvr_unused, nvr_unused);
+		nvsinst->dest_scale = NVS_SCALE_INV_2X;
+		ARITH (NVS_OP_EX2, dest, mask, sat,
+				 nvsNegate(nvsSwizzle(temp, X, X, X, X)),
+				 nvr_unused, nvr_unused);
+		break;
+	case OPCODE_SCS:
+		if (mask & SMASK_X)
+			ARITH(NVS_OP_COS, dest, SMASK_X, sat,
+					nvsSwizzle(src[0], X, X, X, X),
+					nvr_unused, nvr_unused);
+		if (mask & SMASK_Y)
+			ARITH(NVS_OP_SIN, dest, SMASK_Y, sat,
+					nvsSwizzle(src[0], X, X, X, X),
+					nvr_unused, nvr_unused);
+		break;
+	case OPCODE_SUB:
+		ARITH(NVS_OP_ADD, dest, mask, sat,
+				src[0], nvsNegate(src[1]), nvr_unused);
+		break;
+	case OPCODE_XPD:
+		pass0_make_reg(nvs, &temp, NVS_FILE_TEMP, -1);
+		ARITHu(NVS_OP_MUL, temp, SMASK_ALL, 0,
+				 nvsSwizzle(src[0], Z, X, Y, Y),
+				 nvsSwizzle(src[1], Y, Z, X, X),
+				 nvr_unused);
+		ARITH (NVS_OP_MAD, dest, (mask & ~SMASK_W), sat,
+				 nvsSwizzle(src[0], Y, Z, X, X),
+				 nvsSwizzle(src[1], Z, X, Y, Y),
+				 nvsNegate(temp));
+		break;
+	default:
+		WARN_ONCE("hw doesn't support opcode \"%s\","
+			  "and no emulation found\n",
+			  _mesa_opcode_string(inst->Opcode));
+		return GL_FALSE;
+	}
+
+	return GL_TRUE;
+}
+
+static GLboolean
+pass0_translate_arith(nouveauShader *nvs, struct gl_program *prog,
+					  int ipos, int fpos,
+					  nvsFragmentHeader *parent)
+{
+	struct prog_instruction *inst = &prog->Instructions[ipos];
+	nvsFunc *shader = nvs->func;
+	nvsInstruction *nvsinst;
+	GLboolean ret;
+
+	/* Deal with multiple ATTRIB/PARAM in a single instruction */
+	pass0_check_sources(nvs, parent, fpos, inst);
+
+	/* Now it's safe to do the prog_instruction->nvsInstruction
+	 * conversion
+	 */
+	if (shader->SupportsOpcode(shader,
+				   pass0_make_opcode(inst->Opcode))) {
+		nvsRegister src[3], dest;
+		int i;
+
+		for (i=0; i<_mesa_num_inst_src_regs(inst->Opcode); i++)
+			pass0_make_src_reg(nvs, &src[i], &inst->SrcReg[i]);
+		pass0_make_dst_reg(nvs, &dest, &inst->DstReg);
+
+		ARITH(pass0_make_opcode(inst->Opcode), dest,
+				pass0_make_mask(inst->DstReg.WriteMask),
+				(inst->SaturateMode != SATURATE_OFF),
+				src[0], src[1], src[2]);
+		nvsinst->tex_unit   = inst->TexSrcUnit;
+		nvsinst->tex_target = pass0_make_tex_target(inst->TexSrcTarget);
+
+		ret = GL_TRUE;
+	} else
+		ret = pass0_emulate_instruction(nvs, parent, fpos, inst);
+
+	return ret;
+}
+
+static GLboolean
+pass0_translate_instructions(nouveauShader *nvs, int ipos, int fpos,
+						 nvsFragmentHeader *parent)
+{
+	struct gl_program *prog = (struct gl_program *)&nvs->mesa.vp;
+
+	while (1) {
+		struct prog_instruction *inst = &prog->Instructions[ipos];
+
+		switch (inst->Opcode) {
+		case OPCODE_END:
+			return GL_TRUE;
+		case OPCODE_BRA:
+		case OPCODE_CAL:
+		case OPCODE_RET:
+		//case OPCODE_LOOP:
+		//case OPCODE_ENDLOOP:
+		//case OPCODE_IF:
+		//case OPCODE_ELSE:
+		//case OPCODE_ENDIF:
+			WARN_ONCE("branch ops unimplemented\n");
+			return GL_FALSE;
+			break;
+		default:
+			if (!pass0_translate_arith(nvs, prog,
+						   ipos, fpos, parent))
+				return GL_FALSE;
+			break;
+		}
+
+		ipos++;
+	}
+
+	return GL_TRUE;
+}
+
+static void
+pass0_build_attrib_map(nouveauShader *nvs, struct gl_vertex_program *vp)
+{
+	GLuint inputs_read = vp->Base.InputsRead;
+	GLuint input_alloc = ~0xFFFF;
+	int i;
+
+	for (i=0; i<NVS_MAX_ATTRIBS; i++)
+		nvs->vp_attrib_map[i] = -1;
+
+	while (inputs_read) {
+		int in = ffs(inputs_read) - 1;
+		int hw;
+		inputs_read &= ~(1<<in);
+
+		if (vp->IsNVProgram) {
+			/* NVvp: must alias */
+			if (in >= VERT_ATTRIB_GENERIC0)
+				hw = in - VERT_ATTRIB_GENERIC0;
+			else
+				hw = in;
+		} else {
+			/* ARBvp: may alias (but we won't)
+			 * GL2.0: must not alias
+			 */
+			if (in >= VERT_ATTRIB_GENERIC0)
+				hw = ffs(~input_alloc) - 1;
+			else 
+				hw = in;
+			input_alloc |= (1<<hw);
+		}
+
+		nvs->vp_attrib_map[hw] = in;
+	}
+
+	if (NOUVEAU_DEBUG & DEBUG_SHADERS) {
+		printf("vtxprog attrib map:\n");
+		for (i=0; i<NVS_MAX_ATTRIBS; i++) {
+			printf(" hw:%d = attrib:%d\n",
+					i, nvs->vp_attrib_map[i]);
+		}
+	}
+}
+
+static void
+pass0_vp_insert_ff_clip_planes(GLcontext *ctx, nouveauShader *nvs)
+{
+	struct gl_program *prog = &nvs->mesa.vp.Base;
+	nvsFragmentHeader *parent = nvs->program_tree;
+	nvsInstruction *nvsinst;
+	GLuint fpos = 0;
+	nvsRegister opos, epos, eqn, mv[4];
+	GLint tokens[6] = { STATE_MATRIX, STATE_MODELVIEW, 0, 0, 0, 0 };
+	GLint id;
+	int i;
+
+	/* modelview transform */
+	pass0_make_reg(nvs, &opos, NVS_FILE_ATTRIB, NVS_FR_POSITION);
+	pass0_make_reg(nvs, &epos, NVS_FILE_TEMP  , -1);
+	for (i=0; i<4; i++) {
+		tokens[3] = tokens[4] = i;
+		id = _mesa_add_state_reference(prog->Parameters, tokens);
+		pass0_make_reg(nvs, &mv[i], NVS_FILE_CONST, id);
+	}
+	ARITHu(NVS_OP_DP4, epos, SMASK_X, 0, opos, mv[0], nvr_unused);
+	ARITHu(NVS_OP_DP4, epos, SMASK_Y, 0, opos, mv[1], nvr_unused);
+	ARITHu(NVS_OP_DP4, epos, SMASK_Z, 0, opos, mv[2], nvr_unused);
+	ARITHu(NVS_OP_DP4, epos, SMASK_W, 0, opos, mv[3], nvr_unused);
+
+	/* Emit code to emulate fixed-function glClipPlane */
+	for (i=0; i<6; i++) {
+		GLuint clipmask = SMASK_X;
+		nvsRegister clip;
+
+		if (!(ctx->Transform.ClipPlanesEnabled & (1<<i)))
+			continue;
+
+		/* Point a const at a user clipping plane */
+		tokens[0] = STATE_CLIPPLANE;
+		tokens[1] = i;
+		id = _mesa_add_state_reference(prog->Parameters, tokens);
+		pass0_make_reg(nvs, &eqn , NVS_FILE_CONST , id);
+		pass0_make_reg(nvs, &clip, NVS_FILE_RESULT, NVS_FR_CLIP0 + i);
+
+		/*XXX: something else needs to take care of modifying the
+		 *     instructions to write to the correct hw clip register.
+		 */
+		switch (i) {
+		case 0: case 3:	clipmask = SMASK_Y; break;
+		case 1: case 4: clipmask = SMASK_Z; break;
+		case 2: case 5: clipmask = SMASK_W; break;
+		}
+
+		/* Emit transform */
+		ARITHu(NVS_OP_DP4, clip, clipmask, 0, epos, eqn, nvr_unused);
+	}
+}
+
+static void
+pass0_rebase_mesa_consts(nouveauShader *nvs)
+{
+	struct pass0_rec *rec = nvs->pass_rec;
+	struct gl_program *prog = &nvs->mesa.vp.Base;
+	struct prog_instruction *inst = prog->Instructions;
+	int i;
+
+	/*XXX: not a good idea, params->hw_index is malloc'd */
+	memset(nvs->params, 0x00, sizeof(nvs->params));
+
+	/* When doing relative addressing on constants, the hardware needs us
+	 * to fill the "const id" field with a positive value.  Determine the
+	 * most negative index that is used so that all accesses to a
+	 * mesa-provided constant can be rebased to a positive index.
+	 */
+	while (inst->Opcode != OPCODE_END) {
+		for (i=0; i<_mesa_num_inst_src_regs(inst->Opcode); i++) {
+			struct prog_src_register *src = &inst->SrcReg[i];
+
+			switch (src->File) {
+			case PROGRAM_STATE_VAR:
+			case PROGRAM_CONSTANT:
+			case PROGRAM_NAMED_PARAM:
+				if (src->RelAddr && src->Index < 0) {
+					int base = src->Index * -1;
+					if (rec->mesa_const_base < base)
+						rec->mesa_const_base = base;
+				}
+				break;
+			default:
+				break;
+			}
+		}
+
+		inst++;
+	}
+}
+
+static void
+pass0_resolve_mesa_consts(nouveauShader *nvs)
+{
+	struct pass0_rec *rec = nvs->pass_rec;
+	struct gl_program *prog = &nvs->mesa.vp.Base;
+	struct gl_program_parameter_list *plist = prog->Parameters;
+	int i;
+
+	/* Init all const tracking/alloc info from the parameter list, rather
+	 * than doing it as we translate the program.  Otherwise:
+	 *   1) we can't get at the correct constant info when relative
+	 *      addressing is being used due to src->Index not pointing
+	 *      at the exact const;
+	 *   2) as we add extra consts to the program, mesa will call realloc()
+	 *      and we get invalid pointers to the const data.
+	 */
+	rec->mesa_const_last = plist->NumParameters + rec->mesa_const_base;
+	nvs->param_high = rec->mesa_const_last;
+	for (i=0; i<plist->NumParameters; i++) {
+		int hw = rec->mesa_const_base + i;
+
+		switch (plist->Parameters[i].Type) {
+		case PROGRAM_NAMED_PARAM:
+		case PROGRAM_STATE_VAR:
+			nvs->params[hw].in_use     = GL_TRUE;
+			nvs->params[hw].source_val = plist->ParameterValues[i];
+			COPY_4V(nvs->params[hw].val, plist->ParameterValues[i]);
+			break;
+		case PROGRAM_CONSTANT:
+			nvs->params[hw].in_use     = GL_TRUE;
+			nvs->params[hw].source_val = NULL;
+			COPY_4V(nvs->params[hw].val, plist->ParameterValues[i]);
+			break;
+		default:
+			assert(0);
+			break;
+		}
+	}
+}
+
+GLboolean
+nouveau_shader_pass0(GLcontext *ctx, nouveauShader *nvs)
+{
+	nouveauContextPtr           nmesa = NOUVEAU_CONTEXT(ctx);
+	struct gl_program          *prog  = (struct gl_program*)nvs;
+	struct gl_vertex_program   *vp    = (struct gl_vertex_program *)prog;
+	struct gl_fragment_program *fp    = (struct gl_fragment_program *)prog;
+	struct pass0_rec *rec;
+	int ret = GL_FALSE;
+
+	rec = CALLOC_STRUCT(pass0_rec);
+	if (!rec)
+		return GL_FALSE;
+
+	rec->next_temp = prog->NumTemporaries;
+	nvs->pass_rec  = rec;
+		
+	nvs->program_tree = (nvsFragmentHeader*)
+		pass0_create_subroutine(nvs, "program body");
+	if (!nvs->program_tree) {
+		FREE(rec);
+		return GL_FALSE;
+	}
+
+	switch (prog->Target) {
+	case GL_VERTEX_PROGRAM_ARB:
+		nvs->func = &nmesa->VPfunc;
+
+		if (vp->IsPositionInvariant)
+			_mesa_insert_mvp_code(ctx, vp);
+		pass0_rebase_mesa_consts(nvs);
+
+		if (!prog->String && ctx->Transform.ClipPlanesEnabled)
+			pass0_vp_insert_ff_clip_planes(ctx, nvs);
+
+		pass0_build_attrib_map(nvs, vp);
+		break;
+	case GL_FRAGMENT_PROGRAM_ARB:
+		nvs->func = &nmesa->FPfunc;
+
+		if (fp->FogOption != GL_NONE)
+			_mesa_append_fog_code(ctx, fp);
+		pass0_rebase_mesa_consts(nvs);
+		break;
+	default:
+		fprintf(stderr, "Unknown program type %d", prog->Target);
+		FREE(rec);
+		/* DESTROY TREE!! */
+		return GL_FALSE;
+	}
+	nvs->func->card_priv = &nvs->card_priv;
+
+	ret = pass0_translate_instructions(nvs, 0, 0, nvs->program_tree);
+	if (ret)
+		pass0_resolve_mesa_consts(nvs);
+	/*XXX: if (!ret) DESTROY TREE!!! */
+
+	FREE(rec);
+	return ret;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_shader_1.c b/src/mesa/drivers/dri/nouveau/nouveau_shader_1.c
new file mode 100644
index 00000000000..90c57d38078
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_shader_1.c
@@ -0,0 +1,14 @@
+#include "glheader.h"
+#include "macros.h"
+#include "enums.h"
+
+#include "nouveau_shader.h"
+
+GLboolean
+nouveau_shader_pass1(nvsPtr nvs)
+{
+
+   return GL_TRUE;
+}
+
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_shader_2.c b/src/mesa/drivers/dri/nouveau/nouveau_shader_2.c
new file mode 100644
index 00000000000..b043f877e42
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_shader_2.c
@@ -0,0 +1,261 @@
+/*
+ * Copyright (C) 2006 Ben Skeggs.
+ *
+ * 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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
+ *   Ben Skeggs <darktama@iinet.net.au>
+ */
+
+#include "glheader.h"
+#include "macros.h"
+#include "enums.h"
+
+#include "program.h"
+
+#include "nouveau_context.h"
+#include "nouveau_shader.h"
+#include "nouveau_msg.h"
+
+struct pass2_rec {
+	/* Map nvsRegister temp ID onto hw temp ID */
+	unsigned int temps[NVS_MAX_TEMPS];
+	/* Track free hw registers */
+	unsigned int hw_temps[NVS_MAX_TEMPS];
+};
+
+static int
+pass2_alloc_hw_temp(nvsPtr nvs)
+{
+	struct pass2_rec *rec = nvs->pass_rec;
+	int i;
+
+	for (i=0; i<nvs->func->MaxTemp; i++) {
+		/* This is a *horrible* hack.. R0 is both temp0 and result.color
+		 * in NV30/40 fragprogs, we can use R0 as a temp before result
+		 * is written however..
+		 */
+		if (nvs->mesa.vp.Base.Target == GL_FRAGMENT_PROGRAM_ARB && i==0)
+			continue;
+		if (rec->hw_temps[i] == 0) {
+			rec->hw_temps[i] = 1;
+			return i;
+		}
+	}
+
+	return -1;
+}
+
+static nvsRegister
+pass2_mangle_reg(nvsPtr nvs, nvsInstruction *inst, nvsRegister reg)
+{
+	struct pass2_rec *rec = nvs->pass_rec;
+
+	if (reg.file == NVS_FILE_TEMP) {
+		if (rec->temps[reg.index] == -1)
+			rec->temps[reg.index] = pass2_alloc_hw_temp(nvs);
+		reg.index = rec->temps[reg.index];
+	}
+
+	return reg;
+}
+
+static void
+pass2_add_instruction(nvsPtr nvs, nvsInstruction *inst, 
+		      struct _op_xlat *op, int slot)
+{
+	nvsSwzComp default_swz[4] = { NVS_SWZ_X, NVS_SWZ_Y,
+				      NVS_SWZ_Z, NVS_SWZ_W };
+	nvsFunc *shader = nvs->func;
+	nvsRegister reg;
+	int i;
+
+	shader->SetOpcode(shader, op->NV, slot);
+	if (inst->saturate	) shader->SetSaturate(shader);
+	if (inst->cond_update	) shader->SetCCUpdate(shader);
+	if (inst->cond_test	) shader->SetCondition(shader, 1, inst->cond,
+						       inst->cond_reg,
+						       inst->cond_swizzle);
+	else			  shader->SetCondition(shader, 0, NVS_COND_TR,
+						       0,
+						       default_swz);
+	switch (inst->op) {
+	case NVS_OP_TEX:
+	case NVS_OP_TXB:
+	case NVS_OP_TXL:
+	case NVS_OP_TXP:
+	case NVS_OP_TXD:
+		shader->SetTexImageUnit(shader, inst->tex_unit);
+		break;
+	default:
+		break;
+	}
+
+	for (i = 0; i < 3; i++) {
+		if (op->srcpos[i] != -1) {
+			reg = pass2_mangle_reg(nvs, inst, inst->src[i]);
+
+			shader->SetSource(shader, &reg, op->srcpos[i]);
+
+			if (reg.file == NVS_FILE_CONST &&
+					shader->GetSourceConstVal) {
+				int idx_slot =
+					nvs->params[reg.index].hw_index_cnt++;
+				nvs->params[reg.index].hw_index = realloc(
+						nvs->params[reg.index].hw_index,
+						sizeof(int) * idx_slot+1);
+				nvs->params[reg.index].hw_index[idx_slot] =
+					nvs->program_current + 4;
+			}
+		}
+	}
+
+	reg = pass2_mangle_reg(nvs, inst, inst->dest);
+	shader->SetResult(shader, &reg, inst->mask, slot);
+
+	if (inst->dest_scale != NVS_SCALE_1X) {
+		shader->SetResultScale(shader, inst->dest_scale);
+	}
+}
+
+static int
+pass2_assemble_instruction(nvsPtr nvs, nvsInstruction *inst, int last)
+{
+	nvsFunc *shader = nvs->func;
+	struct _op_xlat *op;
+	unsigned int hw_inst[8];
+	int slot;
+	int instsz;
+	int i;
+
+	shader->inst = hw_inst;
+
+	/* Assemble this instruction */
+	if (!(op = shader->GetOPTXFromSOP(inst->op, &slot)))
+		return 0;
+	shader->InitInstruction(shader);
+	pass2_add_instruction(nvs, inst, op, slot);
+	if (last)
+		shader->SetLastInst(shader);
+
+	instsz = shader->GetOffsetNext(nvs->func);
+	if (nvs->program_size + instsz >= nvs->program_alloc_size) {
+		nvs->program_alloc_size *= 2;
+		nvs->program = realloc(nvs->program,
+				       nvs->program_alloc_size *
+				       sizeof(uint32_t));
+	}
+
+	for (i=0; i<instsz; i++)
+		nvs->program[nvs->program_current++] = hw_inst[i];
+	nvs->program_size = nvs->program_current;
+	return 1;
+}
+
+static GLboolean
+pass2_translate(nvsPtr nvs, nvsFragmentHeader *f)
+{
+	nvsFunc *shader = nvs->func;
+	GLboolean last;
+
+	while (f) {
+		last = (f == ((nvsSubroutine*)nvs->program_tree)->insn_tail);
+
+		switch (f->type) {
+		case NVS_INSTRUCTION:
+			if (!pass2_assemble_instruction(nvs,
+							(nvsInstruction *)f,
+							last))
+				return GL_FALSE;
+			break;
+		default:
+			WARN_ONCE("Unimplemented fragment type\n");
+			return GL_FALSE;
+		}
+
+		f = f->next;
+	}
+
+	return GL_TRUE;
+}
+
+/* Translate program into hardware format */
+GLboolean
+nouveau_shader_pass2(nvsPtr nvs)
+{
+	struct pass2_rec *rec;
+	int i;
+
+	rec = calloc(1, sizeof(struct pass2_rec));
+	for (i=0; i<NVS_MAX_TEMPS; i++)
+		rec->temps[i] = -1;
+	nvs->pass_rec = rec;
+
+	/* Start off with allocating 4 uint32_t's for each inst, will be grown
+	 * if necessary..
+	 */
+	nvs->program_alloc_size = nvs->mesa.vp.Base.NumInstructions * 4;
+	nvs->program = calloc(nvs->program_alloc_size, sizeof(uint32_t));
+	nvs->program_size    = 0;
+	nvs->program_current = 0;
+
+	if (!pass2_translate(nvs,
+			     ((nvsSubroutine*)nvs->program_tree)->insn_head)) {
+		free(nvs->program);
+		nvs->program = NULL;
+		return GL_FALSE;
+	}
+
+	/* Shrink allocated memory to only what we need */
+	nvs->program = realloc(nvs->program,
+			       nvs->program_size * sizeof(uint32_t));
+	nvs->program_alloc_size = nvs->program_size;
+
+	nvs->translated  = 1;
+	nvs->on_hardware = 0;
+
+	if (NOUVEAU_DEBUG & DEBUG_SHADERS) {
+		fflush(stdout); fflush(stderr);
+		fprintf(stderr, "-----------MESA PROGRAM target=%s, id=0x%x\n",
+				_mesa_lookup_enum_by_nr(
+					nvs->mesa.vp.Base.Target),
+				nvs->mesa.vp.Base.Id);
+		fflush(stdout); fflush(stderr);
+		_mesa_print_program(&nvs->mesa.vp.Base);
+		fflush(stdout); fflush(stderr);
+		fprintf(stderr, "^^^^^^^^^^^^^^^^MESA PROGRAM\n");
+		fflush(stdout); fflush(stderr);
+		fprintf(stderr, "----------------NV PROGRAM\n");
+		fflush(stdout); fflush(stderr);
+		nvsDisasmHWShader(nvs);
+		fflush(stdout); fflush(stderr);
+		fprintf(stderr, "^^^^^^^^^^^^^^^^NV PROGRAM\n");
+		fflush(stdout); fflush(stderr);
+	}
+
+	return GL_TRUE;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_span.c b/src/mesa/drivers/dri/nouveau/nouveau_span.c
new file mode 100644
index 00000000000..74dec66afcf
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_span.c
@@ -0,0 +1,125 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+#include "nouveau_context.h"
+#include "nouveau_span.h"
+#include "nouveau_fifo.h"
+#include "nouveau_lock.h"
+
+#include "swrast/swrast.h"
+
+#define HAVE_HW_DEPTH_SPANS	0
+#define HAVE_HW_DEPTH_PIXELS	0
+#define HAVE_HW_STENCIL_SPANS	0
+#define HAVE_HW_STENCIL_PIXELS	0
+
+#define HW_CLIPLOOP()							\
+   do {									\
+      int _nc = nmesa->numClipRects;					\
+      while ( _nc-- ) {							\
+	 int minx = nmesa->pClipRects[_nc].x1 - nmesa->drawX;		\
+	 int miny = nmesa->pClipRects[_nc].y1 - nmesa->drawY;		\
+	 int maxx = nmesa->pClipRects[_nc].x2 - nmesa->drawX;		\
+	 int maxy = nmesa->pClipRects[_nc].y2 - nmesa->drawY;
+
+#define LOCAL_VARS							\
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);			\
+   nouveau_renderbuffer *nrb = (nouveau_renderbuffer *)rb;		\
+   GLuint height = nrb->mesa.Height;					\
+   GLubyte *map = (GLubyte *)(nrb->map ? nrb->map : nrb->mem->map) +    \
+	 (nmesa->drawY * nrb->pitch) + (nmesa->drawX * nrb->cpp);       \
+   GLuint p;								\
+   (void) p;
+
+#define Y_FLIP( _y )            (height - _y - 1)
+
+#define HW_LOCK()
+
+#define HW_UNLOCK()
+
+
+
+/* ================================================================
+ * Color buffers
+ */
+
+/* RGB565 */ 
+#define SPANTMP_PIXEL_FMT GL_RGB
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_SHORT_5_6_5
+
+#define TAG(x)    nouveau##x##_RGB565
+#define TAG2(x,y) nouveau##x##_RGB565##y
+#define GET_PTR(X,Y) (map + (Y)*nrb->pitch + (X)*nrb->cpp)
+#include "spantmp2.h"
+
+
+/* ARGB8888 */
+#define SPANTMP_PIXEL_FMT GL_BGRA
+#define SPANTMP_PIXEL_TYPE GL_UNSIGNED_INT_8_8_8_8_REV
+
+#define TAG(x)    nouveau##x##_ARGB8888
+#define TAG2(x,y) nouveau##x##_ARGB8888##y
+#define GET_PTR(X,Y) (map + (Y)*nrb->pitch + (X)*nrb->cpp)
+#include "spantmp2.h"
+
+static void
+nouveauSpanRenderStart( GLcontext *ctx )
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   FIRE_RING();
+   LOCK_HARDWARE(nmesa);
+   nouveauWaitForIdleLocked( nmesa );
+}
+
+static void
+nouveauSpanRenderFinish( GLcontext *ctx )
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   _swrast_flush( ctx );
+   nouveauWaitForIdleLocked( nmesa );
+   UNLOCK_HARDWARE( nmesa );
+}
+
+void nouveauSpanInitFunctions( GLcontext *ctx )
+{
+   struct swrast_device_driver *swdd = _swrast_GetDeviceDriverReference(ctx);
+   swdd->SpanRenderStart	= nouveauSpanRenderStart;
+   swdd->SpanRenderFinish	= nouveauSpanRenderFinish;
+}
+
+
+/**
+ * Plug in the Get/Put routines for the given driRenderbuffer.
+ */
+void
+nouveauSpanSetFunctions(nouveau_renderbuffer *nrb, const GLvisual *vis)
+{
+   if (nrb->mesa._ActualFormat == GL_RGBA8)
+      nouveauInitPointers_ARGB8888(&nrb->mesa);
+   else if (nrb->mesa._ActualFormat == GL_RGB5)
+      nouveauInitPointers_RGB565(&nrb->mesa);
+}
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_span.h b/src/mesa/drivers/dri/nouveau/nouveau_span.h
new file mode 100644
index 00000000000..bc39ecd17b5
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_span.h
@@ -0,0 +1,39 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+
+#ifndef __NOUVEAU_SPAN_H__
+#define __NOUVEAU_SPAN_H__
+
+#include "drirenderbuffer.h"
+#include "nouveau_buffers.h"
+
+extern void nouveauSpanInitFunctions( GLcontext *ctx );
+extern void nouveauSpanSetFunctions(nouveau_renderbuffer *nrb, const GLvisual *vis);
+
+#endif /* __NOUVEAU_SPAN_H__ */
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state.c b/src/mesa/drivers/dri/nouveau/nouveau_state.c
new file mode 100644
index 00000000000..e9fd188d73e
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_state.c
@@ -0,0 +1,345 @@
+/**************************************************************************
+
+Copyright 2006 Jeremy Kolb
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#include "nouveau_context.h"
+#include "nouveau_state.h"
+#include "nouveau_swtcl.h"
+#include "nouveau_fifo.h"
+
+#include "swrast/swrast.h"
+#include "tnl/tnl.h"
+#include "swrast_setup/swrast_setup.h"
+
+#include "tnl/t_pipeline.h"
+
+#include "mtypes.h"
+#include "colormac.h"
+
+static __inline__ GLuint nouveauPackColor(GLuint format,
+				       GLubyte r, GLubyte g,
+				       GLubyte b, GLubyte a)
+{
+   switch (format) {
+   case 2:
+      return PACK_COLOR_565( r, g, b );
+   case 4:
+      return PACK_COLOR_8888( r, g, b, a);
+   default:
+      fprintf(stderr, "unknown format %d\n", (int)format);
+      return 0;
+   }
+}
+
+static void nouveauCalcViewport(GLcontext *ctx)
+{
+    /* Calculate the Viewport Matrix */
+    
+    nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+    const GLfloat *v = ctx->Viewport._WindowMap.m;
+    GLfloat *m = nmesa->viewport.m;
+    GLfloat xoffset = nmesa->drawX, yoffset = nmesa->drawY;
+  
+    nmesa->depth_scale = 1.0 / ctx->DrawBuffer->_DepthMaxF;
+
+    m[MAT_SX] =   v[MAT_SX];
+    m[MAT_TX] =   v[MAT_TX] + xoffset + SUBPIXEL_X;
+    m[MAT_SY] = - v[MAT_SY];
+    m[MAT_TY] =   v[MAT_TY] + yoffset + SUBPIXEL_Y;
+    m[MAT_SZ] =   v[MAT_SZ] * nmesa->depth_scale;
+    m[MAT_TZ] =   v[MAT_TZ] * nmesa->depth_scale;
+
+    nmesa->hw_func.WindowMoved(nmesa);
+}
+
+static void nouveauViewport(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
+{
+    /* 
+     * Need to send (at least on an nv35 the following:
+     * cons = 4 (this may be bytes per pixel)
+     *
+     * The viewport:
+     * 445   0x0000bee0   {size: 0x0   channel: 0x1   cmd: 0x00009ee0} <-- VIEWPORT_SETUP/HEADER ?
+     * 446   0x00000000   {size: 0x0   channel: 0x0   cmd: 0x00000000} <-- x * cons
+     * 447   0x00000c80   {size: 0x0   channel: 0x0   cmd: 0x00000c80} <-- (height + x) * cons
+     * 448   0x00000000   {size: 0x0   channel: 0x0   cmd: 0x00000000} <-- y * cons
+     * 449   0x00000960   {size: 0x0   channel: 0x0   cmd: 0x00000960} <-- (width + y) * cons
+     * 44a   0x00082a00   {size: 0x2   channel: 0x1   cmd: 0x00000a00} <-- VIEWPORT_DIMS
+     * 44b   0x04000000  <-- (Width_from_glViewport << 16) | x
+     * 44c   0x03000000  <-- (Height_from_glViewport << 16) | (win_height - height - y)
+     *
+     */
+    
+    nouveauCalcViewport(ctx);
+}
+
+static void nouveauDepthRange(GLcontext *ctx, GLclampd near, GLclampd far)
+{
+    nouveauCalcViewport(ctx);
+}
+
+static void nouveauDDUpdateHWState(GLcontext *ctx)
+{
+    nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+    int new_state = nmesa->new_state;
+
+    if ( new_state || nmesa->new_render_state & _NEW_TEXTURE )
+    {
+        nmesa->new_state = 0;
+
+        /* Update the various parts of the context's state.
+        */
+        /*
+        if ( new_state & NOUVEAU_NEW_ALPHA )
+            nouveauUpdateAlphaMode( ctx );
+
+        if ( new_state & NOUVEAU_NEW_DEPTH )
+            nouveauUpdateZMode( ctx );
+
+        if ( new_state & NOUVEAU_NEW_FOG )
+            nouveauUpdateFogAttrib( ctx );
+
+        if ( new_state & NOUVEAU_NEW_CLIP )
+            nouveauUpdateClipping( ctx );
+
+        if ( new_state & NOUVEAU_NEW_CULL )
+            nouveauUpdateCull( ctx );
+
+        if ( new_state & NOUVEAU_NEW_MASKS )
+            nouveauUpdateMasks( ctx );
+
+        if ( new_state & NOUVEAU_NEW_WINDOW )
+            nouveauUpdateWindow( ctx );
+
+        if ( nmesa->new_render_state & _NEW_TEXTURE ) {
+            nouveauUpdateTextureState( ctx );
+        }*/
+    }
+}
+
+static void nouveauDDInvalidateState(GLcontext *ctx, GLuint new_state)
+{
+    _swrast_InvalidateState( ctx, new_state );
+    _swsetup_InvalidateState( ctx, new_state );
+    _vbo_InvalidateState( ctx, new_state );
+    _tnl_InvalidateState( ctx, new_state );
+    NOUVEAU_CONTEXT(ctx)->new_render_state |= new_state;
+}
+
+/* Initialize the context's hardware state. */
+void nouveauDDInitState(nouveauContextPtr nmesa)
+{
+    uint32_t type = nmesa->screen->card->type;
+    switch(type)
+    {
+        case NV_03:
+            /* Unimplemented */
+            break;
+        case NV_04:
+        case NV_05:
+            nv04InitStateFuncs(nmesa->glCtx, &nmesa->glCtx->Driver);
+            break;
+        case NV_10:
+            nv10InitStateFuncs(nmesa->glCtx, &nmesa->glCtx->Driver);
+            break;
+        case NV_20:
+            nv20InitStateFuncs(nmesa->glCtx, &nmesa->glCtx->Driver);
+            break;
+        case NV_30:
+        case NV_40:
+        case NV_44:
+            nv30InitStateFuncs(nmesa->glCtx, &nmesa->glCtx->Driver);
+            break;
+        case NV_50:
+            nv50InitStateFuncs(nmesa->glCtx, &nmesa->glCtx->Driver);
+            break;
+        default:
+            break;
+    }
+    nouveau_state_cache_init(nmesa);
+}
+
+/* Initialize the driver's state functions */
+void nouveauDDInitStateFuncs(GLcontext *ctx)
+{
+   ctx->Driver.UpdateState		= nouveauDDInvalidateState;
+
+   ctx->Driver.ClearIndex		= NULL;
+   ctx->Driver.ClearColor		= NULL; //nouveauDDClearColor;
+   ctx->Driver.ClearStencil		= NULL; //nouveauDDClearStencil;
+   ctx->Driver.DrawBuffer		= NULL; //nouveauDDDrawBuffer;
+   ctx->Driver.ReadBuffer		= NULL; //nouveauDDReadBuffer;
+
+   ctx->Driver.IndexMask		= NULL;
+   ctx->Driver.ColorMask		= NULL; //nouveauDDColorMask;
+   ctx->Driver.AlphaFunc		= NULL; //nouveauDDAlphaFunc;
+   ctx->Driver.BlendEquationSeparate	= NULL; //nouveauDDBlendEquationSeparate;
+   ctx->Driver.BlendFuncSeparate	= NULL; //nouveauDDBlendFuncSeparate;
+   ctx->Driver.ClearDepth		= NULL; //nouveauDDClearDepth;
+   ctx->Driver.CullFace			= NULL; //nouveauDDCullFace;
+   ctx->Driver.FrontFace		= NULL; //nouveauDDFrontFace;
+   ctx->Driver.DepthFunc		= NULL; //nouveauDDDepthFunc;
+   ctx->Driver.DepthMask		= NULL; //nouveauDDDepthMask;
+   ctx->Driver.Enable			= NULL; //nouveauDDEnable;
+   ctx->Driver.Fogfv			= NULL; //nouveauDDFogfv;
+   ctx->Driver.Hint			= NULL;
+   ctx->Driver.Lightfv			= NULL;
+   ctx->Driver.LightModelfv		= NULL; //nouveauDDLightModelfv;
+   ctx->Driver.LogicOpcode		= NULL; //nouveauDDLogicOpCode;
+   ctx->Driver.PolygonMode		= NULL;
+   ctx->Driver.PolygonStipple		= NULL; //nouveauDDPolygonStipple;
+   ctx->Driver.RenderMode		= NULL; //nouveauDDRenderMode;
+   ctx->Driver.Scissor			= NULL; //nouveauDDScissor;
+   ctx->Driver.ShadeModel		= NULL; //nouveauDDShadeModel;
+   ctx->Driver.StencilFuncSeparate	= NULL; //nouveauDDStencilFuncSeparate;
+   ctx->Driver.StencilMaskSeparate	= NULL; //nouveauDDStencilMaskSeparate;
+   ctx->Driver.StencilOpSeparate	= NULL; //nouveauDDStencilOpSeparate;
+
+   ctx->Driver.DepthRange               = nouveauDepthRange;
+   ctx->Driver.Viewport                 = nouveauViewport;
+
+   /* Pixel path fallbacks.
+    */
+   ctx->Driver.Accum = _swrast_Accum;
+   ctx->Driver.Bitmap = _swrast_Bitmap;
+   ctx->Driver.CopyPixels = _swrast_CopyPixels;
+   ctx->Driver.DrawPixels = _swrast_DrawPixels;
+   ctx->Driver.ReadPixels = _swrast_ReadPixels;
+
+   /* Swrast hooks for imaging extensions:
+    */
+   ctx->Driver.CopyColorTable = _swrast_CopyColorTable;
+   ctx->Driver.CopyColorSubTable = _swrast_CopyColorSubTable;
+   ctx->Driver.CopyConvolutionFilter1D = _swrast_CopyConvolutionFilter1D;
+   ctx->Driver.CopyConvolutionFilter2D = _swrast_CopyConvolutionFilter2D;
+}
+
+#define STATE_INIT(a) if (ctx->Driver.a) ctx->Driver.a
+
+void nouveauInitState(GLcontext *ctx)
+{
+    /*
+     * Mesa should do this for us:
+     */
+
+    STATE_INIT(AlphaFunc)( ctx, 
+            ctx->Color.AlphaFunc,
+            ctx->Color.AlphaRef);
+
+    STATE_INIT(BlendColor)( ctx,
+            ctx->Color.BlendColor );
+
+    STATE_INIT(BlendEquationSeparate)( ctx, 
+            ctx->Color.BlendEquationRGB,
+            ctx->Color.BlendEquationA);
+
+    STATE_INIT(BlendFuncSeparate)( ctx,
+            ctx->Color.BlendSrcRGB,
+            ctx->Color.BlendDstRGB,
+            ctx->Color.BlendSrcA,
+            ctx->Color.BlendDstA);
+
+    STATE_INIT(ClearColor)( ctx, ctx->Color.ClearColor);
+    STATE_INIT(ClearDepth)( ctx, ctx->Depth.Clear);
+    STATE_INIT(ClearStencil)( ctx, ctx->Stencil.Clear);
+
+    STATE_INIT(ColorMask)( ctx, 
+            ctx->Color.ColorMask[RCOMP],
+            ctx->Color.ColorMask[GCOMP],
+            ctx->Color.ColorMask[BCOMP],
+            ctx->Color.ColorMask[ACOMP]);
+
+    STATE_INIT(CullFace)( ctx, ctx->Polygon.CullFaceMode );
+    STATE_INIT(DepthFunc)( ctx, ctx->Depth.Func );
+    STATE_INIT(DepthMask)( ctx, ctx->Depth.Mask );
+
+    STATE_INIT(Enable)( ctx, GL_ALPHA_TEST, ctx->Color.AlphaEnabled );
+    STATE_INIT(Enable)( ctx, GL_BLEND, ctx->Color.BlendEnabled );
+    STATE_INIT(Enable)( ctx, GL_COLOR_LOGIC_OP, ctx->Color.ColorLogicOpEnabled );
+    STATE_INIT(Enable)( ctx, GL_COLOR_SUM, ctx->Fog.ColorSumEnabled );
+    STATE_INIT(Enable)( ctx, GL_CULL_FACE, ctx->Polygon.CullFlag );
+    STATE_INIT(Enable)( ctx, GL_DEPTH_TEST, ctx->Depth.Test );
+    STATE_INIT(Enable)( ctx, GL_DITHER, ctx->Color.DitherFlag );
+    STATE_INIT(Enable)( ctx, GL_FOG, ctx->Fog.Enabled );
+    STATE_INIT(Enable)( ctx, GL_LIGHTING, ctx->Light.Enabled );
+    STATE_INIT(Enable)( ctx, GL_LINE_SMOOTH, ctx->Line.SmoothFlag );
+    STATE_INIT(Enable)( ctx, GL_LINE_STIPPLE, ctx->Line.StippleFlag );
+    STATE_INIT(Enable)( ctx, GL_POINT_SMOOTH, ctx->Point.SmoothFlag );
+    STATE_INIT(Enable)( ctx, GL_POLYGON_OFFSET_FILL, ctx->Polygon.OffsetFill);
+    STATE_INIT(Enable)( ctx, GL_POLYGON_OFFSET_LINE, ctx->Polygon.OffsetLine);
+    STATE_INIT(Enable)( ctx, GL_POLYGON_OFFSET_POINT, ctx->Polygon.OffsetPoint);
+    STATE_INIT(Enable)( ctx, GL_POLYGON_SMOOTH, ctx->Polygon.SmoothFlag );
+    STATE_INIT(Enable)( ctx, GL_POLYGON_STIPPLE, ctx->Polygon.StippleFlag );
+    STATE_INIT(Enable)( ctx, GL_SCISSOR_TEST, ctx->Scissor.Enabled );
+    STATE_INIT(Enable)( ctx, GL_STENCIL_TEST, ctx->Stencil.Enabled );
+    STATE_INIT(Enable)( ctx, GL_TEXTURE_1D, GL_FALSE );
+    STATE_INIT(Enable)( ctx, GL_TEXTURE_2D, GL_FALSE );
+    STATE_INIT(Enable)( ctx, GL_TEXTURE_RECTANGLE_NV, GL_FALSE );
+    STATE_INIT(Enable)( ctx, GL_TEXTURE_3D, GL_FALSE );
+    STATE_INIT(Enable)( ctx, GL_TEXTURE_CUBE_MAP, GL_FALSE );
+
+    STATE_INIT(Fogfv)( ctx, GL_FOG_COLOR, ctx->Fog.Color );
+    STATE_INIT(Fogfv)( ctx, GL_FOG_MODE, 0 );
+    STATE_INIT(Fogfv)( ctx, GL_FOG_DENSITY, &ctx->Fog.Density );
+    STATE_INIT(Fogfv)( ctx, GL_FOG_START, &ctx->Fog.Start );
+    STATE_INIT(Fogfv)( ctx, GL_FOG_END, &ctx->Fog.End );
+
+    STATE_INIT(FrontFace)( ctx, ctx->Polygon.FrontFace );
+
+    {
+        GLfloat f = (GLfloat)ctx->Light.Model.ColorControl;
+        STATE_INIT(LightModelfv)( ctx, GL_LIGHT_MODEL_COLOR_CONTROL, &f );
+    }
+
+    STATE_INIT(LineStipple)( ctx, ctx->Line.StippleFactor, ctx->Line.StipplePattern );
+    STATE_INIT(LineWidth)( ctx, ctx->Line.Width );
+    STATE_INIT(LogicOpcode)( ctx, ctx->Color.LogicOp );
+    STATE_INIT(PointSize)( ctx, ctx->Point.Size );
+    STATE_INIT(PolygonMode)( ctx, GL_FRONT, ctx->Polygon.FrontMode );
+    STATE_INIT(PolygonMode)( ctx, GL_BACK, ctx->Polygon.BackMode );
+    STATE_INIT(PolygonOffset)( ctx,
+	    ctx->Polygon.OffsetFactor,
+	    ctx->Polygon.OffsetUnits );
+    STATE_INIT(PolygonStipple)( ctx, (const GLubyte *)ctx->PolygonStipple );
+    STATE_INIT(ShadeModel)( ctx, ctx->Light.ShadeModel );
+    STATE_INIT(StencilFuncSeparate)( ctx, GL_FRONT,
+            ctx->Stencil.Function[0],
+            ctx->Stencil.Ref[0],
+            ctx->Stencil.ValueMask[0] );
+    STATE_INIT(StencilFuncSeparate)( ctx, GL_BACK,
+            ctx->Stencil.Function[1],
+            ctx->Stencil.Ref[1],
+            ctx->Stencil.ValueMask[1] );
+    STATE_INIT(StencilMaskSeparate)( ctx, GL_FRONT, ctx->Stencil.WriteMask[0] );
+    STATE_INIT(StencilMaskSeparate)( ctx, GL_BACK, ctx->Stencil.WriteMask[1] );
+    STATE_INIT(StencilOpSeparate)( ctx, GL_FRONT,
+            ctx->Stencil.FailFunc[0],
+            ctx->Stencil.ZFailFunc[0],
+            ctx->Stencil.ZPassFunc[0]);
+    STATE_INIT(StencilOpSeparate)( ctx, GL_BACK,
+            ctx->Stencil.FailFunc[1],
+            ctx->Stencil.ZFailFunc[1],
+            ctx->Stencil.ZPassFunc[1]);
+}
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state.h b/src/mesa/drivers/dri/nouveau/nouveau_state.h
new file mode 100644
index 00000000000..dbac71760b6
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_state.h
@@ -0,0 +1,50 @@
+/**************************************************************************
+
+Copyright 2006 Jeremy Kolb
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#ifndef __NOUVEAU_STATE_H__
+#define __NOUVEAU_STATE_H__
+
+#include "nouveau_context.h"
+
+extern void nouveauDDInitState(nouveauContextPtr nmesa);
+extern void nouveauDDInitStateFuncs(GLcontext *ctx);
+
+extern void nv04InitStateFuncs(GLcontext *ctx, struct dd_function_table *func);
+extern void nv10InitStateFuncs(GLcontext *ctx, struct dd_function_table *func);
+extern void nv20InitStateFuncs(GLcontext *ctx, struct dd_function_table *func);
+extern void nv30InitStateFuncs(GLcontext *ctx, struct dd_function_table *func);
+extern void nv50InitStateFuncs(GLcontext *ctx, struct dd_function_table *func);
+
+extern void nouveauInitState(GLcontext *ctx);
+
+/*
+extern void nouveauDDUpdateState(GLcontext *ctx);
+extern void nouveauDDUpdateHWState(GLcontext *ctx);
+
+extern void nouveauEmitHwStateLocked(nouveauContextPtr nmesa);
+*/
+#endif
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state_cache.c b/src/mesa/drivers/dri/nouveau/nouveau_state_cache.c
new file mode 100644
index 00000000000..cb4b9d30270
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_state_cache.c
@@ -0,0 +1,69 @@
+
+#include "nouveau_state_cache.h"
+#include "nouveau_context.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+
+#define BEGIN_RING_NOFLUSH(subchannel,tag,size) do {					\
+	if (nmesa->fifo.free <= (size))							\
+		WAIT_RING(nmesa,(size));						\
+	OUT_RING( ((size)<<18) | ((subchannel) << 13) | (tag));				\
+	nmesa->fifo.free -= ((size) + 1);                                               \
+}while(0)
+
+// flush all the dirty state
+void nouveau_state_cache_flush(nouveauContextPtr nmesa)
+{
+	int i=0;
+	int run=0;
+
+	// fast-path no state changes
+	if (!nmesa->state_cache.dirty)
+		return;
+	nmesa->state_cache.dirty=0;
+
+	do
+	{
+		// jump to a dirty state
+		while((nmesa->state_cache.hdirty[i/NOUVEAU_STATE_CACHE_HIER_SIZE]==0)&&(i<NOUVEAU_STATE_CACHE_ENTRIES))
+			i=(i&~(NOUVEAU_STATE_CACHE_HIER_SIZE-1))+NOUVEAU_STATE_CACHE_HIER_SIZE;
+		while((nmesa->state_cache.atoms[i].dirty==0)&&(i<NOUVEAU_STATE_CACHE_ENTRIES))
+			i++;
+
+		// figure out a run of dirty values
+		run=0;
+		while((nmesa->state_cache.atoms[i+run].dirty)&&(i+run<NOUVEAU_STATE_CACHE_ENTRIES))
+			run++;
+
+		// output everything as a single run
+		if (run>0) {
+			int j;
+
+			BEGIN_RING_NOFLUSH(NvSub3D, i*4, run);
+			for(j=0;j<run;j++)
+			{
+				OUT_RING(nmesa->state_cache.atoms[i+j].value);
+				nmesa->state_cache.atoms[i+j].dirty=0;
+				if ((i+j)%NOUVEAU_STATE_CACHE_HIER_SIZE==0)
+					nmesa->state_cache.hdirty[(i+j)/NOUVEAU_STATE_CACHE_HIER_SIZE-1]=0;
+			}
+			i+=run;
+		}
+	}
+	while(i<NOUVEAU_STATE_CACHE_ENTRIES);
+	nmesa->state_cache.hdirty[NOUVEAU_STATE_CACHE_HIER_SIZE/NOUVEAU_STATE_CACHE_HIER_SIZE-1]=0;
+}
+
+
+// inits the state cache
+void nouveau_state_cache_init(nouveauContextPtr nmesa)
+{
+	int i;
+	for(i=0;i<NOUVEAU_STATE_CACHE_ENTRIES;i++)
+	{
+		nmesa->state_cache.atoms[i].dirty=0;
+		nmesa->state_cache.atoms[i].value=0xDEADBEEF; // nvidia cards like beef
+	}
+	nmesa->state_cache.dirty=0;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_state_cache.h b/src/mesa/drivers/dri/nouveau/nouveau_state_cache.h
new file mode 100644
index 00000000000..5f9d426450b
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_state_cache.h
@@ -0,0 +1,29 @@
+
+#ifndef __NOUVEAU_STATE_CACHE_H__
+#define __NOUVEAU_STATE_CACHE_H__
+
+#include "mtypes.h"
+
+#define NOUVEAU_STATE_CACHE_ENTRIES 2048
+// size of a dirty requests block
+// you can play with that and tune the value to increase/decrease performance
+// but keep it a power of 2 !
+#define NOUVEAU_STATE_CACHE_HIER_SIZE  32
+
+typedef struct nouveau_state_atom_t{
+	uint32_t value;
+	uint32_t dirty;
+}nouveau_state_atom;
+
+typedef struct nouveau_state_cache_t{
+	nouveau_state_atom atoms[NOUVEAU_STATE_CACHE_ENTRIES];
+	uint32_t current_pos;
+	// hierarchical dirty flags
+	uint8_t hdirty[NOUVEAU_STATE_CACHE_ENTRIES/NOUVEAU_STATE_CACHE_HIER_SIZE];
+	// master dirty flag
+	uint8_t dirty;
+}nouveau_state_cache;
+
+
+#endif
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_swtcl.c b/src/mesa/drivers/dri/nouveau/nouveau_swtcl.c
new file mode 100644
index 00000000000..8a013bd9995
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_swtcl.c
@@ -0,0 +1,127 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+/* Common software TCL code */
+
+#include "nouveau_context.h"
+#include "nouveau_swtcl.h"
+#include "nv10_swtcl.h"
+#include "nouveau_span.h"
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "tnl/tnl.h"
+#include "tnl/t_pipeline.h"
+
+/* Common tri functions */
+
+/* The fallbacks */
+void nouveau_fallback_tri(struct nouveau_context *nmesa,
+		nouveauVertex *v0,
+		nouveauVertex *v1,
+		nouveauVertex *v2)
+{    
+	GLcontext *ctx = nmesa->glCtx;
+	SWvertex v[3];
+	_swsetup_Translate(ctx, v0, &v[0]);
+	_swsetup_Translate(ctx, v1, &v[1]);
+	_swsetup_Translate(ctx, v2, &v[2]);
+	_swrast_Triangle(ctx, &v[0], &v[1], &v[2]);
+}
+
+
+void nouveau_fallback_line(struct nouveau_context *nmesa,
+		nouveauVertex *v0,
+		nouveauVertex *v1)
+{
+	GLcontext *ctx = nmesa->glCtx;
+	SWvertex v[2];
+	_swsetup_Translate(ctx, v0, &v[0]);
+	_swsetup_Translate(ctx, v1, &v[1]);
+	_swrast_Line(ctx, &v[0], &v[1]);
+}
+
+
+void nouveau_fallback_point(struct nouveau_context *nmesa,
+		nouveauVertex *v0)
+{
+	GLcontext *ctx = nmesa->glCtx;
+	SWvertex v[1];
+	_swsetup_Translate(ctx, v0, &v[0]);
+	_swrast_Point(ctx, &v[0]);
+}
+
+void nouveauFallback(struct nouveau_context *nmesa, GLuint bit, GLboolean mode)
+{
+	GLcontext *ctx = nmesa->glCtx;
+	GLuint oldfallback = nmesa->Fallback;
+
+	if (mode) {
+		nmesa->Fallback |= bit;
+		if (oldfallback == 0) {
+			if (nmesa->screen->card->type<NV_10) {
+				//nv04FinishPrimitive(nmesa);
+			} else {
+				//nv10FinishPrimitive(nmesa);
+			}
+
+			_swsetup_Wakeup(ctx);
+			nmesa->render_index = ~0;
+		}
+	}
+	else {
+		nmesa->Fallback &= ~bit;
+		if (oldfallback == bit) {
+			_swrast_flush( ctx );
+
+			if (nmesa->screen->card->type<NV_10) {
+				nv04TriInitFunctions(ctx);
+			} else {
+				nv10TriInitFunctions(ctx);
+			}
+
+			_tnl_invalidate_vertex_state( ctx, ~0 );
+			_tnl_invalidate_vertices( ctx, ~0 );
+			_tnl_install_attrs( ctx, 
+					nmesa->vertex_attrs, 
+					nmesa->vertex_attr_count,
+					nmesa->viewport.m, 0 ); 
+		}
+	}    
+}
+
+
+void nouveauRunPipeline( GLcontext *ctx )
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (nmesa->new_state) {
+		nmesa->new_render_state |= nmesa->new_state;
+	}
+
+	_tnl_run_pipeline( ctx );
+}
+
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_swtcl.h b/src/mesa/drivers/dri/nouveau/nouveau_swtcl.h
new file mode 100644
index 00000000000..ba4d8725a65
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_swtcl.h
@@ -0,0 +1,55 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+
+#ifndef __NOUVEAU_SWTCL_H__
+#define __NOUVEAU_SWTCL_H__
+
+#include "nouveau_context.h"
+
+extern void nouveau_fallback_tri(struct nouveau_context *nmesa,
+		nouveauVertex *v0,
+		nouveauVertex *v1,
+		nouveauVertex *v2);
+
+extern void nouveau_fallback_line(struct nouveau_context *nmesa,
+		nouveauVertex *v0,
+		nouveauVertex *v1);
+
+extern void nouveau_fallback_point(struct nouveau_context *nmesa, 
+		nouveauVertex *v0);
+
+extern void nouveauFallback(struct nouveau_context *nmesa, GLuint bit, GLboolean mode);
+
+extern void nouveauRunPipeline( GLcontext *ctx );
+
+extern void nouveauTriInitFunctions( GLcontext *ctx );
+
+
+#endif /* __NOUVEAU_SWTCL_H__ */
+
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_sync.c b/src/mesa/drivers/dri/nouveau/nouveau_sync.c
new file mode 100644
index 00000000000..c47ff3a985d
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_sync.c
@@ -0,0 +1,195 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ *
+ * 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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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.
+ *
+ */
+
+#include "vblank.h" /* for DO_USLEEP */
+
+#include "nouveau_context.h"
+#include "nouveau_buffers.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+#include "nouveau_msg.h"
+#include "nouveau_sync.h"
+
+nouveau_notifier *
+nouveau_notifier_new(GLcontext *ctx, GLuint handle, GLuint count)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nouveau_notifier *notifier;
+
+#ifdef NOUVEAU_RING_DEBUG
+	return NULL;
+#endif
+
+	notifier = CALLOC_STRUCT(nouveau_notifier_t);
+	if (!notifier)
+		return NULL;
+
+	notifier->mem = nouveau_mem_alloc(ctx,
+					  NOUVEAU_MEM_FB | NOUVEAU_MEM_MAPPED,
+					  count * NV_NOTIFIER_SIZE,
+					  0);
+	if (!notifier->mem) {
+		FREE(notifier);
+		return NULL;
+	}
+
+	if (!nouveauCreateDmaObject(nmesa, handle, notifier->mem->offset,
+						   notifier->mem->size,
+						   0 /* NV_DMA_TARGET_FB */,
+						   0 /* NV_DMA_ACCESS_RW */)) {
+		nouveau_mem_free(ctx, notifier->mem);
+		FREE(notifier);
+		return NULL;
+	}
+
+	notifier->handle = handle;
+	return notifier;
+}
+
+void
+nouveau_notifier_destroy(GLcontext *ctx, nouveau_notifier *notifier)
+{
+	/*XXX: free DMA object.. */
+	nouveau_mem_free(ctx, notifier->mem);
+	FREE(notifier);
+}
+
+void
+nouveau_notifier_reset(nouveau_notifier *notifier, GLuint id)
+{
+	volatile GLuint *n = notifier->mem->map + (id * NV_NOTIFIER_SIZE);
+
+#ifdef NOUVEAU_RING_DEBUG
+	return;
+#endif
+
+	n[NV_NOTIFY_TIME_0      /4] = 0x00000000;
+	n[NV_NOTIFY_TIME_1      /4] = 0x00000000;
+	n[NV_NOTIFY_RETURN_VALUE/4] = 0x00000000;
+	n[NV_NOTIFY_STATE       /4] = (NV_NOTIFY_STATE_STATUS_IN_PROCESS <<
+				       NV_NOTIFY_STATE_STATUS_SHIFT);
+}
+
+GLuint
+nouveau_notifier_status(nouveau_notifier *notifier, GLuint id)
+{
+	volatile GLuint *n = notifier->mem->map + (id * NV_NOTIFIER_SIZE);
+
+	return n[NV_NOTIFY_STATE/4] >> NV_NOTIFY_STATE_STATUS_SHIFT;
+}
+
+GLuint
+nouveau_notifier_return_val(nouveau_notifier *notifier, GLuint id)
+{
+	volatile GLuint *n = notifier->mem->map + (id * NV_NOTIFIER_SIZE);
+
+	return n[NV_NOTIFY_RETURN_VALUE/4];
+}
+
+GLboolean
+nouveau_notifier_wait_status(nouveau_notifier *notifier, GLuint id,
+			     GLuint status, GLuint timeout)
+{
+	volatile GLuint *n = notifier->mem->map + (id * NV_NOTIFIER_SIZE);
+	unsigned int time = 0;
+
+#ifdef NOUVEAU_RING_DEBUG
+	return GL_TRUE;
+#endif
+
+	while (time <= timeout) {
+		if (n[NV_NOTIFY_STATE/4] & NV_NOTIFY_STATE_ERROR_CODE_MASK) {
+			MESSAGE("Notifier returned error: 0x%04x\n",
+					n[NV_NOTIFY_STATE] &
+					NV_NOTIFY_STATE_ERROR_CODE_MASK);
+			return GL_FALSE;
+		}
+
+		if (((n[NV_NOTIFY_STATE/4] & NV_NOTIFY_STATE_STATUS_MASK) >>
+				NV_NOTIFY_STATE_STATUS_SHIFT) == status)
+			return GL_TRUE;
+
+		if (timeout) {
+			DO_USLEEP(1);
+			time++;
+		}
+	}
+
+	MESSAGE("Notifier timed out\n");
+	return GL_FALSE;
+}
+
+void
+nouveau_notifier_wait_nop(GLcontext *ctx, nouveau_notifier *notifier,
+					  GLuint subc)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLboolean ret;
+
+	nouveau_notifier_reset(notifier, 0);
+
+	BEGIN_RING_SIZE(subc, NV_NOTIFY, 1);
+	OUT_RING       (NV_NOTIFY_STYLE_WRITE_ONLY);
+	BEGIN_RING_SIZE(subc, NV_NOP, 1);
+	OUT_RING       (0);
+	FIRE_RING();
+
+	ret = nouveau_notifier_wait_status(notifier, 0,
+					   NV_NOTIFY_STATE_STATUS_COMPLETED,
+					   0 /* no timeout */);
+	if (ret == GL_FALSE) MESSAGE("wait on notifier failed\n");
+}
+
+GLboolean nouveauSyncInitFuncs(GLcontext *ctx)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+#ifdef NOUVEAU_RING_DEBUG
+	return GL_TRUE;
+#endif
+
+	nmesa->syncNotifier = nouveau_notifier_new(ctx, NvSyncNotify, 1);
+	if (!nmesa->syncNotifier) {
+		MESSAGE("Failed to create channel sync notifier\n");
+		return GL_FALSE;
+	}
+
+	/* 0x180 is SET_DMA_NOTIFY, should be correct for all supported 3D
+	 * object classes
+	 */
+	BEGIN_RING_CACHE(NvSub3D, 0x180, 1);
+	OUT_RING_CACHE  (NvSyncNotify);
+#ifdef ALLOW_MULTI_SUBCHANNEL
+	BEGIN_RING_SIZE(NvSubMemFormat,
+	      		NV_MEMORY_TO_MEMORY_FORMAT_DMA_NOTIFY, 1);
+	OUT_RING       (NvSyncNotify);
+#endif
+
+	return GL_TRUE;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_sync.h b/src/mesa/drivers/dri/nouveau/nouveau_sync.h
new file mode 100644
index 00000000000..019d5f6629b
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_sync.h
@@ -0,0 +1,67 @@
+/*
+ * Copyright (C) 2007 Ben Skeggs.
+ *
+ * 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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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.
+ *
+ */
+
+#ifndef __NOUVEAU_SYNC_H__
+#define __NOUVEAU_SYNC_H__
+
+#include "nouveau_buffers.h"
+
+#define NV_NOTIFIER_SIZE                                                      32
+#define NV_NOTIFY_TIME_0                                              0x00000000
+#define NV_NOTIFY_TIME_1                                              0x00000004
+#define NV_NOTIFY_RETURN_VALUE                                        0x00000008
+#define NV_NOTIFY_STATE                                               0x0000000C
+#define NV_NOTIFY_STATE_STATUS_MASK                                   0xFF000000
+#define NV_NOTIFY_STATE_STATUS_SHIFT                                          24
+#define NV_NOTIFY_STATE_STATUS_COMPLETED                                    0x00
+#define NV_NOTIFY_STATE_STATUS_IN_PROCESS                                   0x01
+#define NV_NOTIFY_STATE_ERROR_CODE_MASK                               0x0000FFFF
+#define NV_NOTIFY_STATE_ERROR_CODE_SHIFT                                       0
+
+/* Methods that (hopefully) all objects have */
+#define NV_NOP                                                        0x00000100
+#define NV_NOTIFY                                                     0x00000104
+#define NV_NOTIFY_STYLE_WRITE_ONLY                                             0
+
+typedef struct nouveau_notifier_t {
+	GLuint       handle;
+	nouveau_mem *mem;
+} nouveau_notifier;
+
+extern nouveau_notifier *nouveau_notifier_new(GLcontext *, GLuint handle,
+					      GLuint count);
+extern void nouveau_notifier_destroy(GLcontext *, nouveau_notifier *);
+extern void nouveau_notifier_reset(nouveau_notifier *, GLuint id);
+extern GLuint nouveau_notifier_status(nouveau_notifier *, GLuint id);
+extern GLuint nouveau_notifier_return_val(nouveau_notifier *, GLuint id);
+extern GLboolean nouveau_notifier_wait_status(nouveau_notifier *r, GLuint id,
+					      GLuint status, GLuint timeout);
+extern void nouveau_notifier_wait_nop(GLcontext *ctx,
+      				      nouveau_notifier *, GLuint subc);
+
+extern GLboolean nouveauSyncInitFuncs(GLcontext *ctx);
+#endif
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_tex.c b/src/mesa/drivers/dri/nouveau/nouveau_tex.c
new file mode 100644
index 00000000000..0a8d2796695
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_tex.c
@@ -0,0 +1,49 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+#include "nouveau_tex.h"
+
+// XXX needs some love
+void nouveauTexInitFunctions( struct dd_function_table *functions )
+{
+/*
+   functions->TexEnv			= nouveauTexEnv;
+   functions->ChooseTextureFormat	= nouveauChooseTextureFormat;
+   functions->TexImage1D		= nouveauTexImage1D;
+   functions->TexSubImage1D		= nouveauTexSubImage1D;
+   functions->TexImage2D		= nouveauTexImage2D;
+   functions->TexSubImage2D		= nouveauTexSubImage2D;
+   functions->TexParameter		= nouveauTexParameter;
+   functions->BindTexture		= nouveauBindTexture;
+   functions->NewTextureObject		= nouveauNewTextureObject;
+   functions->DeleteTexture		= nouveauDeleteTexture;
+   functions->IsTextureResident		= driIsTextureResident;
+
+   driInitTextureFormats();
+*/
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nouveau_tex.h b/src/mesa/drivers/dri/nouveau/nouveau_tex.h
new file mode 100644
index 00000000000..7ac71f8300b
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nouveau_tex.h
@@ -0,0 +1,38 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+#ifndef __NOUVEAU_TEX_H__
+#define __NOUVEAU_TEX_H__
+
+#include <errno.h>
+#include "mtypes.h"
+#include "macros.h"
+#include "dd.h"
+
+extern void nouveauTexInitFunctions( struct dd_function_table *functions );
+
+#endif /* __NOUVEAU_TEX_H__ */
diff --git a/src/mesa/drivers/dri/nouveau/nv04_state.c b/src/mesa/drivers/dri/nouveau/nv04_state.c
new file mode 100644
index 00000000000..4129ecc50ea
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv04_state.c
@@ -0,0 +1,538 @@
+/**************************************************************************
+
+Copyright 2007 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#include "nouveau_context.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+#include "nouveau_msg.h"
+
+#include "tnl/t_pipeline.h"
+
+#include "mtypes.h"
+#include "colormac.h"
+
+static uint32_t nv04_compare_func(GLuint f)
+{
+	switch ( f ) {
+		case GL_NEVER:		return 1;
+		case GL_LESS:		return 2;
+		case GL_EQUAL:		return 3;
+		case GL_LEQUAL:		return 4;
+		case GL_GREATER:	return 5;
+		case GL_NOTEQUAL:	return 6;
+		case GL_GEQUAL:		return 7;
+		case GL_ALWAYS:		return 8;
+	}
+	WARN_ONCE("Unable to find the function\n");
+	return 0;
+}
+
+static uint32_t nv04_blend_func(GLuint f)
+{
+	switch ( f ) {
+		case GL_ZERO:			return 0x1;
+		case GL_ONE:			return 0x2;
+		case GL_SRC_COLOR:		return 0x3;
+		case GL_ONE_MINUS_SRC_COLOR:	return 0x4;
+		case GL_SRC_ALPHA:		return 0x5;
+		case GL_ONE_MINUS_SRC_ALPHA:	return 0x6;
+		case GL_DST_ALPHA:		return 0x7;
+		case GL_ONE_MINUS_DST_ALPHA:	return 0x8;
+		case GL_DST_COLOR:		return 0x9;
+		case GL_ONE_MINUS_DST_COLOR:	return 0xA;
+		case GL_SRC_ALPHA_SATURATE:	return 0xB;
+	}
+	WARN_ONCE("Unable to find the function 0x%x\n",f);
+	return 0;
+}
+
+static void nv04_emit_control(GLcontext *ctx)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	uint32_t control,cull;
+	GLubyte alpha_ref;
+
+	CLAMPED_FLOAT_TO_UBYTE(alpha_ref, ctx->Color.AlphaRef);
+	control=alpha_ref;
+	control|=(nv04_compare_func(ctx->Color.AlphaFunc)<<8);
+	control|=(ctx->Color.AlphaEnabled<<12);
+	control|=(1<<13);
+	control|=(ctx->Depth.Test<<14);
+	control|=(nv04_compare_func(ctx->Depth.Func)<<16);
+	if ((ctx->Polygon.CullFlag)&&(ctx->Polygon.CullFaceMode!=GL_FRONT_AND_BACK))
+	{
+		if ((ctx->Polygon.FrontFace==GL_CW)&&(ctx->Polygon.CullFaceMode==GL_FRONT))
+			cull=2;
+		if ((ctx->Polygon.FrontFace==GL_CW)&&(ctx->Polygon.CullFaceMode==GL_BACK))
+			cull=3;
+		if ((ctx->Polygon.FrontFace==GL_CCW)&&(ctx->Polygon.CullFaceMode==GL_FRONT))
+			cull=3;
+		if ((ctx->Polygon.FrontFace==GL_CCW)&&(ctx->Polygon.CullFaceMode==GL_BACK))
+			cull=2;
+	}
+	else
+		if (ctx->Polygon.CullFaceMode==GL_FRONT_AND_BACK)
+			cull=0;
+		else
+			cull=1;
+	control|=(cull<<20);
+	control|=(ctx->Color.DitherFlag<<22);
+	if ((ctx->Depth.Test)&&(ctx->Depth.Mask))
+		control|=(1<<24);
+
+	control|=(1<<30); // integer zbuffer format
+
+	BEGIN_RING_CACHE(NvSub3D, NV04_DX5_TEXTURED_TRIANGLE_CONTROL, 1);
+	OUT_RING_CACHE(control);
+}
+
+static void nv04_emit_blend(GLcontext *ctx)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	uint32_t blend;
+
+	blend=0x4; // texture MODULATE_ALPHA
+	blend|=0x20; // alpha is MSB
+	switch(ctx->Light.ShadeModel) {
+		case GL_SMOOTH:blend|=(1<<6);break;
+		case GL_FLAT:  blend|=(2<<6);break;
+		default:break;
+	}
+	if (ctx->Hint.PerspectiveCorrection!=GL_FASTEST)
+		blend|=(1<<8);
+	blend|=(ctx->Fog.Enabled<<16);
+	blend|=(ctx->Color.BlendEnabled<<20);
+	blend|=(nv04_blend_func(ctx->Color.BlendSrcRGB)<<24);
+	blend|=(nv04_blend_func(ctx->Color.BlendDstRGB)<<28);
+
+	BEGIN_RING_CACHE(NvSub3D, NV04_DX5_TEXTURED_TRIANGLE_BLEND, 1);
+	OUT_RING_CACHE(blend);
+}
+
+static void nv04_emit_fog_color(GLcontext *ctx)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte c[4];
+	c[0] = FLOAT_TO_UBYTE( ctx->Fog.Color[0] );
+	c[1] = FLOAT_TO_UBYTE( ctx->Fog.Color[1] );
+	c[2] = FLOAT_TO_UBYTE( ctx->Fog.Color[2] );
+	c[3] = FLOAT_TO_UBYTE( ctx->Fog.Color[3] );
+	BEGIN_RING_CACHE(NvSub3D, NV04_DX5_TEXTURED_TRIANGLE_FOG_COLOR, 1);
+	OUT_RING_CACHE(PACK_COLOR_8888_REV(c[0],c[1],c[2],c[3]));
+}
+
+static void nv04AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref)
+{
+	nv04_emit_control(ctx);
+}
+
+static void nv04BlendColor(GLcontext *ctx, const GLfloat color[4])
+{
+	nv04_emit_blend(ctx);
+}
+
+static void nv04BlendEquationSeparate(GLcontext *ctx, GLenum modeRGB, GLenum modeA)
+{
+	nv04_emit_blend(ctx);
+}
+
+
+static void nv04BlendFuncSeparate(GLcontext *ctx, GLenum sfactorRGB, GLenum dfactorRGB,
+		GLenum sfactorA, GLenum dfactorA)
+{
+	nv04_emit_blend(ctx);
+}
+
+static void nv04Clear(GLcontext *ctx, GLbitfield mask)
+{
+	/* TODO */
+}
+
+static void nv04ClearColor(GLcontext *ctx, const GLfloat color[4])
+{
+	/* TODO */
+}
+
+static void nv04ClearDepth(GLcontext *ctx, GLclampd d)
+{
+	/* TODO */
+}
+
+static void nv04ClearStencil(GLcontext *ctx, GLint s)
+{
+	/* TODO */
+}
+
+static void nv04ClipPlane(GLcontext *ctx, GLenum plane, const GLfloat *equation)
+{
+	/* TODO */
+}
+
+static void nv04ColorMask(GLcontext *ctx, GLboolean rmask, GLboolean gmask,
+		GLboolean bmask, GLboolean amask )
+{
+	/* TODO */
+}
+
+static void nv04ColorMaterial(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	/* TODO I need love */
+}
+
+static void nv04CullFace(GLcontext *ctx, GLenum mode)
+{
+	nv04_emit_control(ctx);
+}
+
+static void nv04FrontFace(GLcontext *ctx, GLenum mode)
+{
+	/* TODO */
+}
+
+static void nv04DepthFunc(GLcontext *ctx, GLenum func)
+{
+	nv04_emit_control(ctx);
+}
+
+static void nv04DepthMask(GLcontext *ctx, GLboolean flag)
+{
+	/* TODO */
+}
+
+static void nv04DepthRange(GLcontext *ctx, GLclampd nearval, GLclampd farval)
+{
+	/* TODO */
+}
+
+/** Specify the current buffer for writing */
+//void (*DrawBuffer)( GLcontext *ctx, GLenum buffer );
+/** Specify the buffers for writing for fragment programs*/
+//void (*DrawBuffers)( GLcontext *ctx, GLsizei n, const GLenum *buffers );
+
+static void nv04Enable(GLcontext *ctx, GLenum cap, GLboolean state)
+{
+	switch(cap)
+	{
+		case GL_ALPHA_TEST:
+			nv04_emit_control(ctx);
+			break;
+//		case GL_AUTO_NORMAL:
+		case GL_BLEND:
+			nv04_emit_blend(ctx);
+			break;
+//		case GL_CLIP_PLANE0:
+//		case GL_CLIP_PLANE1:
+//		case GL_CLIP_PLANE2:
+//		case GL_CLIP_PLANE3:
+//		case GL_CLIP_PLANE4:
+//		case GL_CLIP_PLANE5:
+//		case GL_COLOR_LOGIC_OP:
+//		case GL_COLOR_MATERIAL:
+//		case GL_COLOR_SUM_EXT:
+//		case GL_COLOR_TABLE:
+//		case GL_CONVOLUTION_1D:
+//		case GL_CONVOLUTION_2D:
+		case GL_CULL_FACE:
+			nv04_emit_control(ctx);
+			break;
+		case GL_DEPTH_TEST:
+			nv04_emit_control(ctx);
+			break;
+		case GL_DITHER:
+			nv04_emit_control(ctx);
+			break;
+		case GL_FOG:
+			nv04_emit_blend(ctx);
+			nv04_emit_fog_color(ctx);
+			break;
+//		case GL_HISTOGRAM:
+//		case GL_INDEX_LOGIC_OP:
+//		case GL_LIGHT0:
+//		case GL_LIGHT1:
+//		case GL_LIGHT2:
+//		case GL_LIGHT3:
+//		case GL_LIGHT4:
+//		case GL_LIGHT5:
+//		case GL_LIGHT6:
+//		case GL_LIGHT7:
+//		case GL_LIGHTING:
+//		case GL_LINE_SMOOTH:
+//		case GL_LINE_STIPPLE:
+//		case GL_MAP1_COLOR_4:
+//		case GL_MAP1_INDEX:
+//		case GL_MAP1_NORMAL:
+//		case GL_MAP1_TEXTURE_COORD_1:
+//		case GL_MAP1_TEXTURE_COORD_2:
+//		case GL_MAP1_TEXTURE_COORD_3:
+//		case GL_MAP1_TEXTURE_COORD_4:
+//		case GL_MAP1_VERTEX_3:
+//		case GL_MAP1_VERTEX_4:
+//		case GL_MAP2_COLOR_4:
+//		case GL_MAP2_INDEX:
+//		case GL_MAP2_NORMAL:
+//		case GL_MAP2_TEXTURE_COORD_1:
+//		case GL_MAP2_TEXTURE_COORD_2:
+//		case GL_MAP2_TEXTURE_COORD_3:
+//		case GL_MAP2_TEXTURE_COORD_4:
+//		case GL_MAP2_VERTEX_3:
+//		case GL_MAP2_VERTEX_4:
+//		case GL_MINMAX:
+//		case GL_NORMALIZE:
+//		case GL_POINT_SMOOTH:
+//		case GL_POLYGON_OFFSET_POINT:
+//		case GL_POLYGON_OFFSET_LINE:
+//		case GL_POLYGON_OFFSET_FILL:
+//		case GL_POLYGON_SMOOTH:
+//		case GL_POLYGON_STIPPLE:
+//		case GL_POST_COLOR_MATRIX_COLOR_TABLE:
+//		case GL_POST_CONVOLUTION_COLOR_TABLE:
+//		case GL_RESCALE_NORMAL:
+//		case GL_SCISSOR_TEST:
+//		case GL_SEPARABLE_2D:
+//		case GL_STENCIL_TEST:
+//		case GL_TEXTURE_GEN_Q:
+//		case GL_TEXTURE_GEN_R:
+//		case GL_TEXTURE_GEN_S:
+//		case GL_TEXTURE_GEN_T:
+//		case GL_TEXTURE_1D:
+//		case GL_TEXTURE_2D:
+//		case GL_TEXTURE_3D:
+	}
+}
+
+static void nv04Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+	nv04_emit_blend(ctx);
+	nv04_emit_fog_color(ctx);
+}
+   
+static void nv04Hint(GLcontext *ctx, GLenum target, GLenum mode)
+{
+	switch(target)
+	{
+		case GL_PERSPECTIVE_CORRECTION_HINT:nv04_emit_blend(ctx);break;
+		default:break;
+	}
+}
+
+static void nv04LineStipple(GLcontext *ctx, GLint factor, GLushort pattern )
+{
+	/* TODO not even in your dreams */
+}
+
+static void nv04LineWidth(GLcontext *ctx, GLfloat width)
+{
+	/* TODO */
+}
+
+static void nv04LogicOpcode(GLcontext *ctx, GLenum opcode)
+{
+	/* TODO */
+}
+
+static void nv04PointParameterfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+	/* TODO */
+}
+
+static void nv04PointSize(GLcontext *ctx, GLfloat size)
+{
+	/* TODO */
+}
+
+static void nv04PolygonMode(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	/* TODO */
+}
+
+/** Set the scale and units used to calculate depth values */
+static void nv04PolygonOffset(GLcontext *ctx, GLfloat factor, GLfloat units)
+{
+	/* TODO */
+}
+
+/** Set the polygon stippling pattern */
+static void nv04PolygonStipple(GLcontext *ctx, const GLubyte *mask )
+{
+	/* TODO */
+}
+
+/* Specifies the current buffer for reading */
+void (*ReadBuffer)( GLcontext *ctx, GLenum buffer );
+/** Set rasterization mode */
+void (*RenderMode)(GLcontext *ctx, GLenum mode );
+
+/** Define the scissor box */
+static void nv04Scissor(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
+{
+	/* TODO */
+}
+
+/** Select flat or smooth shading */
+static void nv04ShadeModel(GLcontext *ctx, GLenum mode)
+{
+	nv04_emit_blend(ctx);
+}
+
+/** OpenGL 2.0 two-sided StencilFunc */
+static void nv04StencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func,
+		GLint ref, GLuint mask)
+{
+	/* TODO */
+}
+
+/** OpenGL 2.0 two-sided StencilMask */
+static void nv04StencilMaskSeparate(GLcontext *ctx, GLenum face, GLuint mask)
+{
+	/* TODO */
+}
+
+/** OpenGL 2.0 two-sided StencilOp */
+static void nv04StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail,
+		GLenum zfail, GLenum zpass)
+{
+	/* TODO */
+}
+
+/** Control the generation of texture coordinates */
+void (*TexGen)(GLcontext *ctx, GLenum coord, GLenum pname,
+		const GLfloat *params);
+/** Set texture environment parameters */
+void (*TexEnv)(GLcontext *ctx, GLenum target, GLenum pname,
+		const GLfloat *param);
+/** Set texture parameters */
+void (*TexParameter)(GLcontext *ctx, GLenum target,
+		struct gl_texture_object *texObj,
+		GLenum pname, const GLfloat *params);
+
+/* Update anything that depends on the window position/size */
+static void nv04WindowMoved(nouveauContextPtr nmesa)
+{
+}
+
+/* Initialise any card-specific non-GL related state */
+static GLboolean nv04InitCard(nouveauContextPtr nmesa)
+{
+	nouveauObjectOnSubchannel(nmesa, NvSub3D, Nv3D);
+	nouveauObjectOnSubchannel(nmesa, NvSubCtxSurf3D, NvCtxSurf3D);
+
+	BEGIN_RING_SIZE(NvSubCtxSurf3D, NV04_CONTEXT_SURFACES_3D_DMA_NOTIFY, 3);
+	OUT_RING(NvDmaFB);
+	OUT_RING(NvDmaFB);
+	OUT_RING(NvDmaFB);
+	BEGIN_RING_SIZE(NvSub3D, NV04_DX5_TEXTURED_TRIANGLE_SURFACE, 1);
+	OUT_RING(NvCtxSurf3D);
+	return GL_TRUE;
+}
+
+/* Update buffer offset/pitch/format */
+static GLboolean nv04BindBuffers(nouveauContextPtr nmesa, int num_color,
+		nouveau_renderbuffer **color,
+		nouveau_renderbuffer *depth)
+{
+	GLuint x, y, w, h;
+
+	w = color[0]->mesa.Width;
+	h = color[0]->mesa.Height;
+	x = nmesa->drawX;
+	y = nmesa->drawY;
+
+	/* FIXME pitches have to be aligned ! */
+	BEGIN_RING_SIZE(NvSubCtxSurf3D, NV04_CONTEXT_SURFACES_3D_PITCH, 2);
+	OUT_RING(color[0]->pitch|(depth?(depth->pitch<<16):0));
+	OUT_RING(color[0]->offset);
+
+	if (depth) {
+		BEGIN_RING_SIZE(NvSubCtxSurf3D, NV04_CONTEXT_SURFACES_3D_OFFSET_ZETA, 1);
+		OUT_RING(depth->offset);
+	}
+
+	BEGIN_RING_SIZE(NvSubCtxSurf3D, NV04_CONTEXT_SURFACES_3D_CLIP_HORIZONTAL, 2);
+	OUT_RING((w<<16)|x);
+	OUT_RING((h<<16)|y);
+
+	/* FIXME not sure... */
+	BEGIN_RING_SIZE(NvSubCtxSurf3D, NV04_CONTEXT_SURFACES_3D_CLIP_SIZE, 1);
+	OUT_RING((h<<16)|w);
+
+	BEGIN_RING_SIZE(NvSubCtxSurf3D, NV04_CONTEXT_SURFACES_3D_FORMAT, 1);
+	if (color[0]->mesa._ActualFormat == GL_RGBA8)
+		OUT_RING(108/*A8R8G8B8*/);
+	else
+		OUT_RING(103/*R5G6B5*/);
+
+	return GL_TRUE;
+}
+
+void nv04InitStateFuncs(GLcontext *ctx, struct dd_function_table *func)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	func->AlphaFunc			= nv04AlphaFunc;
+	func->BlendColor		= nv04BlendColor;
+	func->BlendEquationSeparate	= nv04BlendEquationSeparate;
+	func->BlendFuncSeparate		= nv04BlendFuncSeparate;
+	func->Clear			= nv04Clear;
+	func->ClearColor		= nv04ClearColor;
+	func->ClearDepth		= nv04ClearDepth;
+	func->ClearStencil		= nv04ClearStencil;
+	func->ClipPlane			= nv04ClipPlane;
+	func->ColorMask			= nv04ColorMask;
+	func->ColorMaterial		= nv04ColorMaterial;
+	func->CullFace			= nv04CullFace;
+	func->FrontFace			= nv04FrontFace;
+	func->DepthFunc			= nv04DepthFunc;
+	func->DepthMask			= nv04DepthMask;
+	func->DepthRange		= nv04DepthRange;
+	func->Enable			= nv04Enable;
+	func->Fogfv			= nv04Fogfv;
+	func->Hint			= nv04Hint;
+/*	func->Lightfv			= nv04Lightfv;*/
+/*	func->LightModelfv		= nv04LightModelfv; */
+	func->LineStipple		= nv04LineStipple;		/* Not for NV04 */
+	func->LineWidth			= nv04LineWidth;
+	func->LogicOpcode		= nv04LogicOpcode;
+	func->PointParameterfv		= nv04PointParameterfv;
+	func->PointSize			= nv04PointSize;
+	func->PolygonMode		= nv04PolygonMode;
+	func->PolygonOffset		= nv04PolygonOffset;
+	func->PolygonStipple		= nv04PolygonStipple;		/* Not for NV04 */
+/*	func->ReadBuffer		= nv04ReadBuffer;*/
+/*	func->RenderMode		= nv04RenderMode;*/
+	func->Scissor			= nv04Scissor;
+	func->ShadeModel		= nv04ShadeModel;
+	func->StencilFuncSeparate	= nv04StencilFuncSeparate;
+	func->StencilMaskSeparate	= nv04StencilMaskSeparate;
+	func->StencilOpSeparate		= nv04StencilOpSeparate;
+/*	func->TexGen			= nv04TexGen;*/
+/*	func->TexParameter		= nv04TexParameter;*/
+/*	func->TextureMatrix		= nv04TextureMatrix;*/
+
+	nmesa->hw_func.InitCard		= nv04InitCard;
+	nmesa->hw_func.BindBuffers	= nv04BindBuffers;
+	nmesa->hw_func.WindowMoved	= nv04WindowMoved;
+}
diff --git a/src/mesa/drivers/dri/nouveau/nv04_swtcl.c b/src/mesa/drivers/dri/nouveau/nv04_swtcl.c
new file mode 100644
index 00000000000..9b5332b77a7
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv04_swtcl.c
@@ -0,0 +1,618 @@
+/*
+ * Copyright 2007 Stephane Marchesin. 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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sub license,
+ * 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ * VIA, S3 GRAPHICS, AND/OR ITS SUPPLIERS 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.
+ */
+
+/* Software TCL for NV04, NV05, NV06 */
+
+#include <stdio.h>
+#include <math.h>
+
+#include "glheader.h"
+#include "context.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "colormac.h"
+#include "enums.h"
+
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "tnl/t_context.h"
+#include "tnl/t_pipeline.h"
+
+#include "nouveau_swtcl.h"
+#include "nv04_swtcl.h"
+#include "nouveau_context.h"
+#include "nouveau_span.h"
+#include "nouveau_reg.h"
+#include "nouveau_tex.h"
+#include "nouveau_fifo.h"
+#include "nouveau_msg.h"
+#include "nouveau_object.h"
+
+static void nv04RasterPrimitive( GLcontext *ctx, GLenum rprim, GLuint hwprim );
+static void nv04RenderPrimitive( GLcontext *ctx, GLenum prim );
+static void nv04ResetLineStipple( GLcontext *ctx );
+
+
+static inline void nv04_2triangles(struct nouveau_context *nmesa,nouveauVertex* v0,nouveauVertex* v1,nouveauVertex* v2,nouveauVertex* v3,nouveauVertex* v4,nouveauVertex* v5)
+{
+	BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0xA),49);
+	OUT_RINGp(v0,8);
+	OUT_RINGp(v1,8);
+	OUT_RINGp(v2,8);
+	OUT_RINGp(v3,8);
+	OUT_RINGp(v4,8);
+	OUT_RINGp(v5,8);
+	OUT_RING(0xFEDCBA);
+}
+
+static inline void nv04_1triangle(struct nouveau_context *nmesa,nouveauVertex* v0,nouveauVertex* v1,nouveauVertex* v2)
+{
+	BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0xD),25);
+	OUT_RINGp(v0,8);
+	OUT_RINGp(v1,8);
+	OUT_RINGp(v2,8);
+	OUT_RING(0xFED);
+}
+
+static inline void nv04_1quad(struct nouveau_context *nmesa,nouveauVertex* v0,nouveauVertex* v1,nouveauVertex* v2,nouveauVertex* v3)
+{
+	BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0xC),33);
+	OUT_RINGp(v0,8);
+	OUT_RINGp(v1,8);
+	OUT_RINGp(v2,8);
+	OUT_RINGp(v3,8);
+	OUT_RING(0xFECEDC);
+}
+
+static inline void nv04_render_points(GLcontext *ctx,GLuint first,GLuint last)
+{
+	WARN_ONCE("Unimplemented\n");
+}
+
+static inline void nv04_render_line(GLcontext *ctx,GLuint v1,GLuint v2)
+{
+	WARN_ONCE("Unimplemented\n");
+}
+
+static inline void nv04_render_triangle(GLcontext *ctx,GLuint v1,GLuint v2,GLuint v3)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+
+	nv04_1triangle(nmesa,
+			(nouveauVertex*)(vertptr+v1*vertsize),
+			(nouveauVertex*)(vertptr+v2*vertsize),
+			(nouveauVertex*)(vertptr+v3*vertsize)
+		  );
+}
+
+static inline void nv04_render_quad(GLcontext *ctx,GLuint v1,GLuint v2,GLuint v3,GLuint v4)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+
+	nv04_1quad(nmesa,
+			(nouveauVertex*)(vertptr+v1*vertsize),
+			(nouveauVertex*)(vertptr+v2*vertsize),
+			(nouveauVertex*)(vertptr+v3*vertsize),
+			(nouveauVertex*)(vertptr+v4*vertsize)
+		  );
+}
+
+/**********************************************************************/
+/*               Render unclipped begin/end objects                   */
+/**********************************************************************/
+
+static void nv04_render_points_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	// erm
+}
+
+static void nv04_render_lines_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	// umm
+}
+
+static void nv04_render_line_strip_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	// yeah
+}
+
+static void nv04_render_line_loop_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	// right
+}
+
+static void nv04_render_triangles_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	int i;
+
+	for(i=start;i<count-5;i+=6)
+		nv04_2triangles(nmesa,
+				(nouveauVertex*)(vertptr+(i+0)*vertsize),
+				(nouveauVertex*)(vertptr+(i+1)*vertsize),
+				(nouveauVertex*)(vertptr+(i+2)*vertsize),
+				(nouveauVertex*)(vertptr+(i+3)*vertsize),
+				(nouveauVertex*)(vertptr+(i+4)*vertsize),
+				(nouveauVertex*)(vertptr+(i+5)*vertsize)
+			       );
+	if (i!=count)
+	{
+		nv04_1triangle(nmesa,
+				(nouveauVertex*)(vertptr+(i+0)*vertsize),
+				(nouveauVertex*)(vertptr+(i+1)*vertsize),
+				(nouveauVertex*)(vertptr+(i+2)*vertsize)
+			       );
+		i+=3;
+	}
+	if (i!=count)
+		printf("oops\n");
+}
+
+static void nv04_render_tri_strip_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	uint32_t striptbl[]={0x321210,0x543432,0x765654,0x987876,0xBA9A98,0xDCBCBA,0xFEDEDC};
+	int i,j;
+
+	for(i=start;i<count;i+=14)
+	{
+		int numvert=MIN2(16,count-i);
+		int numtri=numvert-2;
+		if (numvert<3)
+			break;
+
+		BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x0),numvert*8);
+		for(j=0;j<numvert;j++)
+			OUT_RINGp((nouveauVertex*)(vertptr+(i+j)*vertsize),8);
+
+		BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_DRAW|NONINC_METHOD,(numtri+1)/2);
+		for(j=0;j<numtri/2;j++)
+			OUT_RING(striptbl[j]);
+		if (numtri%2)
+			OUT_RING(striptbl[numtri/2]&0xFFF);
+	}
+}
+
+static void nv04_render_tri_fan_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	uint32_t fantbl[]={0x320210,0x540430,0x760650,0x980870,0xBA0A90,0xDC0CB0,0xFE0ED0};
+	int i,j;
+
+	BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x0),8);
+	OUT_RINGp((nouveauVertex*)(vertptr+start*vertsize),8);
+
+	for(i=start+1;i<count;i+=14)
+	{
+		int numvert=MIN2(15,count-i);
+		int numtri=numvert-1;
+		if (numvert<3)
+			break;
+
+		BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x1),numvert*8);
+
+		for(j=0;j<numvert;j++)
+			OUT_RINGp((nouveauVertex*)(vertptr+(i+j)*vertsize),8);
+
+		BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_DRAW|NONINC_METHOD,(numtri+1)/2);
+		for(j=0;j<numtri/2;j++)
+			OUT_RING(fantbl[j]);
+		if (numtri%2)
+			OUT_RING(fantbl[numtri/2]&0xFFF);
+	}
+}
+
+static void nv04_render_quads_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	int i;
+
+	for(i=start;i<count;i+=4)
+		nv04_1quad(nmesa,
+				(nouveauVertex*)(vertptr+(i+0)*vertsize),
+				(nouveauVertex*)(vertptr+(i+1)*vertsize),
+				(nouveauVertex*)(vertptr+(i+2)*vertsize),
+				(nouveauVertex*)(vertptr+(i+3)*vertsize)
+			       );
+}
+
+static void nv04_render_noop_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+}
+
+static void (*nv04_render_tab_verts[GL_POLYGON+2])(GLcontext *,
+							   GLuint,
+							   GLuint,
+							   GLuint) =
+{
+   nv04_render_points_verts,
+   nv04_render_lines_verts,
+   nv04_render_line_loop_verts,
+   nv04_render_line_strip_verts,
+   nv04_render_triangles_verts,
+   nv04_render_tri_strip_verts,
+   nv04_render_tri_fan_verts,
+   nv04_render_quads_verts,
+   nv04_render_tri_strip_verts,  //nv04_render_quad_strip_verts
+   nv04_render_tri_fan_verts,    //nv04_render_poly_verts
+   nv04_render_noop_verts,
+};
+
+
+static void nv04_render_points_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	// erm
+}
+
+static void nv04_render_lines_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	// umm
+}
+
+static void nv04_render_line_strip_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	// yeah
+}
+
+static void nv04_render_line_loop_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	// right
+}
+
+static void nv04_render_triangles_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;
+	int i;
+
+	for(i=start;i<count-5;i+=6)
+		nv04_2triangles(nmesa,
+				(nouveauVertex*)(vertptr+elt[i+0]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+1]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+2]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+3]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+4]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+5]*vertsize)
+			       );
+	if (i!=count)
+	{
+		nv04_1triangle(nmesa,
+				(nouveauVertex*)(vertptr+elt[i+0]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+1]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+2]*vertsize)
+			       );
+		i+=3;
+	}
+	if (i!=count)
+		printf("oops\n");
+}
+
+static void nv04_render_tri_strip_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	uint32_t striptbl[]={0x321210,0x543432,0x765654,0x987876,0xBA9A98,0xDCBCBA,0xFEDEDC};
+	const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;
+	int i,j;
+
+	for(i=start;i<count;i+=14)
+	{
+		int numvert=MIN2(16,count-i);
+		int numtri=numvert-2;
+		if (numvert<3)
+			break;
+
+		BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x0),numvert*8);
+		for(j=0;j<numvert;j++)
+			OUT_RINGp((nouveauVertex*)(vertptr+elt[i+j]*vertsize),8);
+
+		BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_DRAW|NONINC_METHOD,(numtri+1)/2);
+		for(j=0;j<numtri/2;j++)
+			OUT_RING(striptbl[j]);
+		if (numtri%2)
+			OUT_RING(striptbl[numtri/2]&0xFFF);
+	}
+}
+
+static void nv04_render_tri_fan_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	uint32_t fantbl[]={0x320210,0x540430,0x760650,0x980870,0xBA0A90,0xDC0CB0,0xFE0ED0};
+	const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;
+	int i,j;
+
+	BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x0),8);
+	OUT_RINGp((nouveauVertex*)(vertptr+elt[start]*vertsize),8);
+
+	for(i=start+1;i<count;i+=14)
+	{
+		int numvert=MIN2(15,count-i);
+		int numtri=numvert-2;
+		if (numvert<3)
+			break;
+
+		BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_TLVERTEX_SX(0x1),numvert*8);
+
+		for(j=0;j<numvert;j++)
+			OUT_RINGp((nouveauVertex*)(vertptr+elt[i+j]*vertsize),8);
+
+		BEGIN_RING_SIZE(NvSub3D,NV04_DX5_TEXTURED_TRIANGLE_DRAW|NONINC_METHOD,(numtri+1)/2);
+		for(j=0;j<numtri/2;j++)
+			OUT_RING(fantbl[j]);
+		if (numtri%2)
+			OUT_RING(fantbl[numtri/2]&0xFFF);
+	}
+}
+
+static void nv04_render_quads_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;
+	int i;
+
+	for(i=start;i<count;i+=4)
+		nv04_1quad(nmesa,
+				(nouveauVertex*)(vertptr+elt[i+0]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+1]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+2]*vertsize),
+				(nouveauVertex*)(vertptr+elt[i+3]*vertsize)
+			       );
+}
+
+static void nv04_render_noop_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+}
+
+static void (*nv04_render_tab_elts[GL_POLYGON+2])(GLcontext *,
+							   GLuint,
+							   GLuint,
+							   GLuint) =
+{
+   nv04_render_points_elts,
+   nv04_render_lines_elts,
+   nv04_render_line_loop_elts,
+   nv04_render_line_strip_elts,
+   nv04_render_triangles_elts,
+   nv04_render_tri_strip_elts,
+   nv04_render_tri_fan_elts,
+   nv04_render_quads_elts,
+   nv04_render_tri_strip_elts,   // nv04_render_quad_strip_elts,
+   nv04_render_tri_fan_elts,     // nv04_render_poly_elts,
+   nv04_render_noop_elts,
+};
+
+
+/**********************************************************************/
+/*                    Choose render functions                         */
+/**********************************************************************/
+
+
+#define EMIT_ATTR( ATTR, STYLE )					\
+do {									\
+   nmesa->vertex_attrs[nmesa->vertex_attr_count].attrib = (ATTR);	\
+   nmesa->vertex_attrs[nmesa->vertex_attr_count].format = (STYLE);	\
+   nmesa->vertex_attr_count++;						\
+} while (0)
+
+#define EMIT_PAD( N )							\
+do {									\
+   nmesa->vertex_attrs[nmesa->vertex_attr_count].attrib = 0;		\
+   nmesa->vertex_attrs[nmesa->vertex_attr_count].format = EMIT_PAD;	\
+   nmesa->vertex_attrs[nmesa->vertex_attr_count].offset = (N);		\
+   nmesa->vertex_attr_count++;						\
+} while (0)
+
+static void nv04_render_clipped_line(GLcontext *ctx,GLuint ii,GLuint jj)
+{
+}
+
+static void nv04_render_clipped_poly(GLcontext *ctx,const GLuint *elts,GLuint n)
+{
+}
+
+static void nv04ChooseRenderState(GLcontext *ctx)
+{
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+
+	tnl->Driver.Render.PrimTabVerts = nv04_render_tab_verts;
+	tnl->Driver.Render.PrimTabElts = nv04_render_tab_elts;
+	tnl->Driver.Render.ClippedLine = nv04_render_clipped_line;
+	tnl->Driver.Render.ClippedPolygon = nv04_render_clipped_poly;
+	tnl->Driver.Render.Points = nv04_render_points;
+	tnl->Driver.Render.Line = nv04_render_line;
+	tnl->Driver.Render.Triangle = nv04_render_triangle;
+	tnl->Driver.Render.Quad = nv04_render_quad;
+}
+
+
+
+static inline void nv04OutputVertexFormat(struct nouveau_context* nmesa)
+{
+	GLcontext* ctx=nmesa->glCtx;
+	DECLARE_RENDERINPUTS(index);
+
+	/*
+	 * Tell t_vertex about the vertex format
+	 */
+	RENDERINPUTS_COPY(index, nmesa->render_inputs_bitset);
+
+	// SX SY SZ INVW
+	// FIXME : we use W instead of INVW, but since W=1 it doesn't matter
+	if (RENDERINPUTS_TEST(index, _TNL_ATTRIB_POS))
+		EMIT_ATTR(_TNL_ATTRIB_POS,EMIT_4F_VIEWPORT);
+	else
+		EMIT_PAD(4*sizeof(float));
+
+	// COLOR
+	if (RENDERINPUTS_TEST(index, _TNL_ATTRIB_COLOR0))
+		EMIT_ATTR(_TNL_ATTRIB_COLOR0,EMIT_4UB_4F_ABGR);
+	else
+		EMIT_PAD(4);
+
+	// SPECULAR
+	if (RENDERINPUTS_TEST(index, _TNL_ATTRIB_COLOR1))
+		EMIT_ATTR(_TNL_ATTRIB_COLOR1,EMIT_4UB_4F_ABGR);
+	else
+		EMIT_PAD(4);
+
+	// TEXTURE
+	if (RENDERINPUTS_TEST(index, _TNL_ATTRIB_TEX0))
+		EMIT_ATTR(_TNL_ATTRIB_TEX0,EMIT_2F);
+	else
+		EMIT_PAD(2*sizeof(float));
+
+	nmesa->vertex_size=_tnl_install_attrs( ctx,
+			nmesa->vertex_attrs, 
+			nmesa->vertex_attr_count,
+			ctx->Viewport._WindowMap.m, 0 );
+}
+
+
+static void nv04ChooseVertexState( GLcontext *ctx )
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+	DECLARE_RENDERINPUTS(index);
+
+	RENDERINPUTS_COPY(index, tnl->render_inputs_bitset);
+	if (!RENDERINPUTS_EQUAL(index, nmesa->render_inputs_bitset))
+	{
+		RENDERINPUTS_COPY(nmesa->render_inputs_bitset, index);
+		nv04OutputVertexFormat(nmesa);
+	}
+}
+
+
+/**********************************************************************/
+/*                 High level hooks for t_vb_render.c                 */
+/**********************************************************************/
+
+
+static void nv04RenderStart(GLcontext *ctx)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (nmesa->new_state) {
+		nmesa->new_render_state |= nmesa->new_state;
+	}
+
+	if (nmesa->new_render_state) {
+		nv04ChooseVertexState(ctx);
+		nv04ChooseRenderState(ctx);
+		nmesa->new_render_state = 0;
+	}
+}
+
+static void nv04RenderFinish(GLcontext *ctx)
+{
+}
+
+
+/* System to flush dma and emit state changes based on the rasterized
+ * primitive.
+ */
+void nv04RasterPrimitive(GLcontext *ctx,
+		GLenum glprim,
+		GLuint hwprim)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+
+	assert (!nmesa->new_state);
+
+	if (hwprim != nmesa->current_primitive)
+	{
+		nmesa->current_primitive=hwprim;
+		
+	}
+}
+
+static const GLuint hw_prim[GL_POLYGON+1] = {
+	GL_POINTS+1,
+	GL_LINES+1,
+	GL_LINE_STRIP+1,
+	GL_LINE_LOOP+1,
+	GL_TRIANGLES+1,
+	GL_TRIANGLE_STRIP+1,
+	GL_TRIANGLE_FAN+1,
+	GL_QUADS+1,
+	GL_QUAD_STRIP+1,
+	GL_POLYGON+1
+};
+
+/* Callback for mesa:
+ */
+static void nv04RenderPrimitive( GLcontext *ctx, GLuint prim )
+{
+	nv04RasterPrimitive( ctx, prim, hw_prim[prim] );
+}
+
+static void nv04ResetLineStipple( GLcontext *ctx )
+{
+	/* FIXME do something here */
+	WARN_ONCE("Unimplemented nv04ResetLineStipple\n");
+}
+
+
+/**********************************************************************/
+/*                            Initialization.                         */
+/**********************************************************************/
+
+void nv04TriInitFunctions(GLcontext *ctx)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+
+	tnl->Driver.RunPipeline = nouveauRunPipeline;
+	tnl->Driver.Render.Start = nv04RenderStart;
+	tnl->Driver.Render.Finish = nv04RenderFinish;
+	tnl->Driver.Render.PrimitiveNotify = nv04RenderPrimitive;
+	tnl->Driver.Render.ResetLineStipple = nv04ResetLineStipple;
+	tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
+	tnl->Driver.Render.CopyPV = _tnl_copy_pv;
+	tnl->Driver.Render.Interp = _tnl_interp;
+
+	_tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 32 );
+
+	nmesa->verts = (GLubyte *)tnl->clipspace.vertex_buf;
+}
+
+
diff --git a/src/mesa/drivers/dri/nouveau/nv04_swtcl.h b/src/mesa/drivers/dri/nouveau/nv04_swtcl.h
new file mode 100644
index 00000000000..42dde5383eb
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv04_swtcl.h
@@ -0,0 +1,12 @@
+#ifndef __NV04_SWTCL_H__
+#define __NV04_SWTCL_H__
+
+#include "mtypes.h"
+
+extern void nv04Fallback( GLcontext *ctx, GLuint bit, GLboolean mode );
+extern void nv04FinishPrimitive(struct nouveau_context *nmesa);
+extern void nv04TriInitFunctions(GLcontext *ctx);
+#define FALLBACK( nmesa, bit, mode ) nouveauFallback( nmesa->glCtx, bit, mode )
+
+#endif /* __NV04_SWTCL_H__ */
+
diff --git a/src/mesa/drivers/dri/nouveau/nv10_state.c b/src/mesa/drivers/dri/nouveau/nv10_state.c
new file mode 100644
index 00000000000..5f304ccab92
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv10_state.c
@@ -0,0 +1,798 @@
+/**************************************************************************
+
+Copyright 2006 Nouveau
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#include "nouveau_context.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+
+#include "tnl/t_pipeline.h"
+
+#include "mtypes.h"
+#include "colormac.h"
+
+static void nv10ViewportScale(nouveauContextPtr nmesa)
+{
+	GLcontext *ctx = nmesa->glCtx;
+	GLuint w = ctx->Viewport.Width;
+	GLuint h = ctx->Viewport.Height;
+
+	GLfloat max_depth = (ctx->Viewport.Near + ctx->Viewport.Far) * 0.5;
+/*	if (ctx->DrawBuffer) {
+		switch (ctx->DrawBuffer->_DepthBuffer->DepthBits) {
+			case 16:
+				max_depth *= 32767.0;
+				break;
+			case 24:
+				max_depth *= 16777215.0;
+				break;
+		}
+	} else {*/
+		/* Default to 24 bits range */	
+		max_depth *= 16777215.0;
+/*	}*/
+
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_VIEWPORT_SCALE_X, 4);
+	OUT_RING_CACHEf ((((GLfloat) w) * 0.5) - 2048.0);
+	OUT_RING_CACHEf ((((GLfloat) h) * 0.5) - 2048.0);
+	OUT_RING_CACHEf (max_depth);
+	OUT_RING_CACHEf (0.0);
+}
+
+static void nv10AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte ubRef;
+	CLAMPED_FLOAT_TO_UBYTE(ubRef, ref);
+
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_ALPHA_FUNC_FUNC, 2);
+	OUT_RING_CACHE(func);
+	OUT_RING_CACHE(ubRef);
+}
+
+static void nv10BlendColor(GLcontext *ctx, const GLfloat color[4])
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); 
+	GLubyte cf[4];
+
+	CLAMPED_FLOAT_TO_UBYTE(cf[0], color[0]);
+	CLAMPED_FLOAT_TO_UBYTE(cf[1], color[1]);
+	CLAMPED_FLOAT_TO_UBYTE(cf[2], color[2]);
+	CLAMPED_FLOAT_TO_UBYTE(cf[3], color[3]);
+
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_BLEND_COLOR, 1);
+	OUT_RING_CACHE(PACK_COLOR_8888(cf[3], cf[1], cf[2], cf[0]));
+}
+
+static void nv10BlendEquationSeparate(GLcontext *ctx, GLenum modeRGB, GLenum modeA)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	assert( modeRGB == modeA );
+
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_BLEND_EQUATION, 1);
+	OUT_RING_CACHE(modeRGB);
+}
+
+
+static void nv10BlendFuncSeparate(GLcontext *ctx, GLenum sfactorRGB, GLenum dfactorRGB,
+		GLenum sfactorA, GLenum dfactorA)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	assert( sfactorRGB == sfactorA );
+	assert( dfactorRGB == dfactorA );
+
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC, 2);
+	OUT_RING_CACHE(sfactorRGB);
+	OUT_RING_CACHE(dfactorRGB);
+}
+
+static void nv10Clear(GLcontext *ctx, GLbitfield mask)
+{
+	/* TODO */
+}
+
+static void nv10ClearColor(GLcontext *ctx, const GLfloat color[4])
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte c[4];
+	UNCLAMPED_FLOAT_TO_RGBA_CHAN(c,color);
+	nmesa->clear_color_value = PACK_COLOR_8888(c[3],c[0],c[1],c[2]);
+}
+
+static void nv10ClearDepth(GLcontext *ctx, GLclampd d)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+/*	switch (ctx->DrawBuffer->_DepthBuffer->DepthBits) {
+		case 16:
+			nmesa->clear_value = (uint32_t)(d*0x7FFF);
+			break;
+		case 24:*/
+			nmesa->clear_value = ((nmesa->clear_value&0x000000FF) |
+				(((uint32_t)(d*0xFFFFFF))<<8));
+/*			break;
+	}*/
+}
+
+static void nv10ClearStencil(GLcontext *ctx, GLint s)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+/*	if (ctx->DrawBuffer->_DepthBuffer->DepthBits == 24) {*/
+		nmesa->clear_value = ((nmesa->clear_value&0xFFFFFF00)|
+			(s&0x000000FF));
+/*	}*/
+}
+
+static void nv10ClipPlane(GLcontext *ctx, GLenum plane, const GLfloat *equation)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_CLIP_PLANE_A(plane), 4);
+	OUT_RING_CACHEf(equation[0]);
+	OUT_RING_CACHEf(equation[1]);
+	OUT_RING_CACHEf(equation[2]);
+	OUT_RING_CACHEf(equation[3]);
+}
+
+static void nv10ColorMask(GLcontext *ctx, GLboolean rmask, GLboolean gmask,
+		GLboolean bmask, GLboolean amask )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_COLOR_MASK, 1);
+	OUT_RING_CACHE(((amask && 0x01) << 24) | ((rmask && 0x01) << 16) | ((gmask && 0x01)<< 8) | ((bmask && 0x01) << 0));
+}
+
+static void nv10ColorMaterial(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	/* TODO I need love */
+}
+
+static void nv10CullFace(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_CULL_FACE, 1);
+	OUT_RING_CACHE(mode);
+}
+
+static void nv10FrontFace(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_FRONT_FACE, 1);
+	OUT_RING_CACHE(mode);
+}
+
+static void nv10DepthFunc(GLcontext *ctx, GLenum func)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_DEPTH_FUNC, 1);
+	OUT_RING_CACHE(func);
+}
+
+static void nv10DepthMask(GLcontext *ctx, GLboolean flag)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_DEPTH_WRITE_ENABLE, 1);
+	OUT_RING_CACHE(flag);
+}
+
+static void nv10DepthRange(GLcontext *ctx, GLclampd nearval, GLclampd farval)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	GLfloat depth_scale = 16777216.0;
+	if (ctx->DrawBuffer->_DepthBuffer->DepthBits == 16) {
+		depth_scale = 32768.0;
+	}
+
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_DEPTH_RANGE_NEAR, 2);
+	OUT_RING_CACHEf(nearval * depth_scale);
+	OUT_RING_CACHEf(farval * depth_scale);
+
+	nv10ViewportScale(nmesa);
+}
+
+/** Specify the current buffer for writing */
+//void (*DrawBuffer)( GLcontext *ctx, GLenum buffer );
+/** Specify the buffers for writing for fragment programs*/
+//void (*DrawBuffers)( GLcontext *ctx, GLsizei n, const GLenum *buffers );
+
+static void nv10Enable(GLcontext *ctx, GLenum cap, GLboolean state)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	switch(cap)
+	{
+		case GL_ALPHA_TEST:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_ALPHA_FUNC_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_AUTO_NORMAL:
+		case GL_BLEND:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_BLEND_FUNC_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_CLIP_PLANE0:
+		case GL_CLIP_PLANE1:
+		case GL_CLIP_PLANE2:
+		case GL_CLIP_PLANE3:
+		case GL_CLIP_PLANE4:
+		case GL_CLIP_PLANE5:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_CLIP_PLANE_ENABLE(cap-GL_CLIP_PLANE0), 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_COLOR_LOGIC_OP:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_COLOR_MATERIAL:
+//		case GL_COLOR_SUM_EXT:
+//		case GL_COLOR_TABLE:
+//		case GL_CONVOLUTION_1D:
+//		case GL_CONVOLUTION_2D:
+		case GL_CULL_FACE:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_CULL_FACE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_DEPTH_TEST:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_DEPTH_TEST_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_DITHER:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_DITHER_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_FOG:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_FOG_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_HISTOGRAM:
+//		case GL_INDEX_LOGIC_OP:
+		case GL_LIGHT0:
+		case GL_LIGHT1:
+		case GL_LIGHT2:
+		case GL_LIGHT3:
+		case GL_LIGHT4:
+		case GL_LIGHT5:
+		case GL_LIGHT6:
+		case GL_LIGHT7:
+			{
+			uint32_t mask=1<<(2*(cap-GL_LIGHT0));
+			nmesa->enabled_lights=((nmesa->enabled_lights&mask)|(mask*state));
+			if (nmesa->lighting_enabled)
+			{
+				BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_ENABLED_LIGHTS, 1);
+				OUT_RING_CACHE(nmesa->enabled_lights);
+			}
+			break;
+			}
+		case GL_LIGHTING:
+			nmesa->lighting_enabled=state;
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_ENABLED_LIGHTS, 1);
+			if (nmesa->lighting_enabled)
+				OUT_RING_CACHE(nmesa->enabled_lights);
+			else
+				OUT_RING_CACHE(0x0);
+			break;
+		case GL_LINE_SMOOTH:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LINE_SMOOTH_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_LINE_STIPPLE:
+//		case GL_MAP1_COLOR_4:
+//		case GL_MAP1_INDEX:
+//		case GL_MAP1_NORMAL:
+//		case GL_MAP1_TEXTURE_COORD_1:
+//		case GL_MAP1_TEXTURE_COORD_2:
+//		case GL_MAP1_TEXTURE_COORD_3:
+//		case GL_MAP1_TEXTURE_COORD_4:
+//		case GL_MAP1_VERTEX_3:
+//		case GL_MAP1_VERTEX_4:
+//		case GL_MAP2_COLOR_4:
+//		case GL_MAP2_INDEX:
+//		case GL_MAP2_NORMAL:
+//		case GL_MAP2_TEXTURE_COORD_1:
+//		case GL_MAP2_TEXTURE_COORD_2:
+//		case GL_MAP2_TEXTURE_COORD_3:
+//		case GL_MAP2_TEXTURE_COORD_4:
+//		case GL_MAP2_VERTEX_3:
+//		case GL_MAP2_VERTEX_4:
+//		case GL_MINMAX:
+		case GL_NORMALIZE:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_NORMALIZE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_POINT_SMOOTH:
+		case GL_POLYGON_OFFSET_POINT:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_POLYGON_OFFSET_POINT_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_OFFSET_LINE:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_POLYGON_OFFSET_LINE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_OFFSET_FILL:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FILL_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_SMOOTH:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_POLYGON_SMOOTH_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_POLYGON_STIPPLE:
+//		case GL_POST_COLOR_MATRIX_COLOR_TABLE:
+//		case GL_POST_CONVOLUTION_COLOR_TABLE:
+//		case GL_RESCALE_NORMAL:
+//		case GL_SCISSOR_TEST:
+//		case GL_SEPARABLE_2D:
+		case GL_STENCIL_TEST:
+			// TODO BACK and FRONT ?
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_STENCIL_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_TEXTURE_GEN_Q:
+//		case GL_TEXTURE_GEN_R:
+//		case GL_TEXTURE_GEN_S:
+//		case GL_TEXTURE_GEN_T:
+//		case GL_TEXTURE_1D:
+//		case GL_TEXTURE_2D:
+//		case GL_TEXTURE_3D:
+	}
+}
+
+static void nv10Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+    nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+    switch(pname)
+    {
+        case GL_FOG_MODE:
+            //BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_FOG_MODE, 1);
+            //OUT_RING_CACHE (params);
+            break;
+            /* TODO: unsure about the rest.*/
+        default:
+            break;
+    }
+
+}
+   
+static void nv10Hint(GLcontext *ctx, GLenum target, GLenum mode)
+{
+	/* TODO I need love (fog and line_smooth hints) */
+}
+
+// void (*IndexMask)(GLcontext *ctx, GLuint mask);
+
+enum {
+	SPOTLIGHT_NO_UPDATE,
+	SPOTLIGHT_UPDATE_EXPONENT,
+	SPOTLIGHT_UPDATE_DIRECTION,
+	SPOTLIGHT_UPDATE_ALL
+};
+
+static void nv10Lightfv(GLcontext *ctx, GLenum light, GLenum pname, const GLfloat *params )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLint p = light - GL_LIGHT0;
+	struct gl_light *l = &ctx->Light.Light[p];
+	int spotlight_update = SPOTLIGHT_NO_UPDATE;
+
+	switch(pname)
+	{
+		case GL_AMBIENT:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_DIFFUSE:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_SPECULAR:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_POSITION:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_POSITION_X(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_SPOT_DIRECTION:
+			spotlight_update = SPOTLIGHT_UPDATE_DIRECTION;
+			break;
+		case GL_SPOT_EXPONENT:
+			spotlight_update = SPOTLIGHT_UPDATE_EXPONENT;
+			break;
+		case GL_SPOT_CUTOFF:
+			spotlight_update = SPOTLIGHT_UPDATE_ALL;
+			break;
+		case GL_CONSTANT_ATTENUATION:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_CONSTANT_ATTENUATION(p), 1);
+			OUT_RING_CACHEf(*params);
+			break;
+		case GL_LINEAR_ATTENUATION:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_LINEAR_ATTENUATION(p), 1);
+			OUT_RING_CACHEf(*params);
+			break;
+		case GL_QUADRATIC_ATTENUATION:
+			BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_QUADRATIC_ATTENUATION(p), 1);
+			OUT_RING_CACHEf(*params);
+			break;
+		default:
+			break;
+	}
+
+	switch(spotlight_update) {
+		case SPOTLIGHT_UPDATE_DIRECTION:
+			{
+				GLfloat x,y,z;
+				GLfloat spot_light_coef_a = 1.0 / (l->_CosCutoff - 1.0);
+				x = spot_light_coef_a * l->_NormDirection[0];
+				y = spot_light_coef_a * l->_NormDirection[1];
+				z = spot_light_coef_a * l->_NormDirection[2];
+				BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_X(p), 3);
+				OUT_RING_CACHEf(x);
+				OUT_RING_CACHEf(y);
+				OUT_RING_CACHEf(z);
+			}
+			break;
+		case SPOTLIGHT_UPDATE_EXPONENT:
+			{
+				GLfloat cc,lc,qc;
+				cc = 1.0;	/* FIXME: These need to be correctly computed */
+				lc = 0.0;
+				qc = 2.0;
+				BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_A(p), 3);
+				OUT_RING_CACHEf(cc);
+				OUT_RING_CACHEf(lc);
+				OUT_RING_CACHEf(qc);
+			}
+			break;
+		case SPOTLIGHT_UPDATE_ALL:
+			{
+				GLfloat cc,lc,qc, x,y,z, c;
+				GLfloat spot_light_coef_a = 1.0 / (l->_CosCutoff - 1.0);
+				cc = 1.0;	/* FIXME: These need to be correctly computed */
+				lc = 0.0;
+				qc = 2.0;
+				x = spot_light_coef_a * l->_NormDirection[0];
+				y = spot_light_coef_a * l->_NormDirection[1];
+				z = spot_light_coef_a * l->_NormDirection[2];
+				c = spot_light_coef_a + 1.0;
+				BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_A(p), 7);
+				OUT_RING_CACHEf(cc);
+				OUT_RING_CACHEf(lc);
+				OUT_RING_CACHEf(qc);
+				OUT_RING_CACHEf(x);
+				OUT_RING_CACHEf(y);
+				OUT_RING_CACHEf(z);
+				OUT_RING_CACHEf(c);
+			}
+			break;
+		default:
+			break;
+	}
+}
+
+/** Set the lighting model parameters */
+static void (*LightModelfv)(GLcontext *ctx, GLenum pname, const GLfloat *params);
+
+
+static void nv10LineStipple(GLcontext *ctx, GLint factor, GLushort pattern )
+{
+	/* Not for NV10 */
+}
+
+static void nv10LineWidth(GLcontext *ctx, GLfloat width)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_LINE_WIDTH, 1);
+	OUT_RING_CACHE(((int) (width * 8.0)) & -4);
+}
+
+static void nv10LogicOpcode(GLcontext *ctx, GLenum opcode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_OP, 1);
+	OUT_RING_CACHE(opcode);
+}
+
+static void nv10PointParameterfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+	/*TODO: not sure what goes here. */
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	
+}
+
+static void nv10PointSize(GLcontext *ctx, GLfloat size)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_POINT_SIZE, 1);
+	OUT_RING_CACHE(((int) (size * 8.0)) & -4);
+}
+
+static void nv10PolygonMode(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_POLYGON_MODE_FRONT, 1);
+		OUT_RING_CACHE(mode);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_POLYGON_MODE_BACK, 1);
+		OUT_RING_CACHE(mode);
+	}
+}
+
+/** Set the scale and units used to calculate depth values */
+static void nv10PolygonOffset(GLcontext *ctx, GLfloat factor, GLfloat units)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+        BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FACTOR, 2);
+        OUT_RING_CACHEf(factor);
+        OUT_RING_CACHEf(units);
+}
+
+/** Set the polygon stippling pattern */
+static void nv10PolygonStipple(GLcontext *ctx, const GLubyte *mask )
+{
+	/* Not for NV10 */
+}
+
+/* Specifies the current buffer for reading */
+void (*ReadBuffer)( GLcontext *ctx, GLenum buffer );
+/** Set rasterization mode */
+void (*RenderMode)(GLcontext *ctx, GLenum mode );
+
+/** Define the scissor box */
+static void nv10Scissor(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
+{
+}
+
+/** Select flat or smooth shading */
+static void nv10ShadeModel(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_SHADE_MODEL, 1);
+	OUT_RING_CACHE(mode);
+}
+
+/** OpenGL 2.0 two-sided StencilFunc */
+static void nv10StencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func,
+		GLint ref, GLuint mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	/* NV10 do not have separate FRONT and BACK stencils */
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_STENCIL_FUNC_FUNC, 3);
+	OUT_RING_CACHE(func);
+	OUT_RING_CACHE(ref);
+	OUT_RING_CACHE(mask);
+}
+
+/** OpenGL 2.0 two-sided StencilMask */
+static void nv10StencilMaskSeparate(GLcontext *ctx, GLenum face, GLuint mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	/* NV10 do not have separate FRONT and BACK stencils */
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_STENCIL_MASK, 1);
+	OUT_RING_CACHE(mask);
+}
+
+/** OpenGL 2.0 two-sided StencilOp */
+static void nv10StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail,
+		GLenum zfail, GLenum zpass)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	/* NV10 do not have separate FRONT and BACK stencils */
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_STENCIL_OP_FAIL, 3);
+	OUT_RING_CACHE(fail);
+	OUT_RING_CACHE(zfail);
+	OUT_RING_CACHE(zpass);
+}
+
+/** Control the generation of texture coordinates */
+void (*TexGen)(GLcontext *ctx, GLenum coord, GLenum pname,
+		const GLfloat *params);
+/** Set texture environment parameters */
+void (*TexEnv)(GLcontext *ctx, GLenum target, GLenum pname,
+		const GLfloat *param);
+/** Set texture parameters */
+void (*TexParameter)(GLcontext *ctx, GLenum target,
+		struct gl_texture_object *texObj,
+		GLenum pname, const GLfloat *params);
+
+static void nv10TextureMatrix(GLcontext *ctx, GLuint unit, const GLmatrix *mat)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+        BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_TX_MATRIX(unit, 0), 16);
+        /*XXX: This SHOULD work.*/
+        OUT_RING_CACHEp(mat->m, 16);
+}
+
+/* Update anything that depends on the window position/size */
+static void nv10WindowMoved(nouveauContextPtr nmesa)
+{
+	GLcontext *ctx = nmesa->glCtx;
+	GLfloat *v = nmesa->viewport.m;
+	GLuint w = ctx->Viewport.Width;
+	GLuint h = ctx->Viewport.Height;
+	GLuint x = ctx->Viewport.X + nmesa->drawX;
+	GLuint y = ctx->Viewport.Y + nmesa->drawY;
+	int i;
+
+        BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_VIEWPORT_HORIZ, 2);
+        OUT_RING_CACHE((w << 16) | x);
+        OUT_RING_CACHE((h << 16) | y);
+
+	/* something to do with clears, possibly doesn't belong here */
+	BEGIN_RING_SIZE(NvSub3D, 0x02b4, 1);
+	OUT_RING(0);
+
+	BEGIN_RING_CACHE(NvSub3D,
+	      NV10_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_HORIZ(0), 1);
+        OUT_RING_CACHE(((w+x-1) << 16) | x | 0x08000800);
+	BEGIN_RING_CACHE(NvSub3D,
+	      NV10_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_VERT(0), 1);
+        OUT_RING_CACHE(((h+y-1) << 16) | y | 0x08000800);
+	for (i=1; i<8; i++) {
+		BEGIN_RING_CACHE(NvSub3D,
+		      NV10_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_HORIZ(i), 1);
+        	OUT_RING_CACHE(0);
+		BEGIN_RING_CACHE(NvSub3D,
+		      NV10_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_VERT(i), 1);
+	        OUT_RING_CACHE(0);
+	}
+
+	nv10ViewportScale(nmesa);
+}
+
+/* Initialise any card-specific non-GL related state */
+static GLboolean nv10InitCard(nouveauContextPtr nmesa)
+{
+	nouveauObjectOnSubchannel(nmesa, NvSub3D, Nv3D);
+
+	BEGIN_RING_SIZE(NvSub3D, NV10_TCL_PRIMITIVE_3D_SET_DMA_IN_MEMORY0, 2);
+	OUT_RING(NvDmaFB);	/* 184 dma_in_memory0 */
+	OUT_RING(NvDmaFB);	/* 188 dma_in_memory1 */
+	BEGIN_RING_SIZE(NvSub3D, NV10_TCL_PRIMITIVE_3D_SET_DMA_IN_MEMORY2, 2);
+	OUT_RING(NvDmaFB);	/* 194 dma_in_memory2 */
+	OUT_RING(NvDmaFB);	/* 198 dma_in_memory3 */
+
+	BEGIN_RING_SIZE(NvSub3D, 0x0290, 1);
+	OUT_RING(0x00100001);
+	BEGIN_RING_SIZE(NvSub3D, 0x03f4, 1);
+	OUT_RING(0);
+
+	/* not for nv10, only for >= nv11 */
+	if ((nmesa->screen->card->id>>4) >= 0x11) {
+	        BEGIN_RING_SIZE(NvSub3D, 0x120, 3);
+        	OUT_RING(0);
+	        OUT_RING(1);
+	        OUT_RING(2);
+	}
+
+	return GL_TRUE;
+}
+
+/* Update buffer offset/pitch/format */
+static GLboolean nv10BindBuffers(nouveauContextPtr nmesa, int num_color,
+				 nouveau_renderbuffer **color,
+				 nouveau_renderbuffer *depth)
+{
+	GLuint x, y, w, h;
+	GLuint pitch, format, depth_pitch;
+
+	w = color[0]->mesa.Width;
+	h = color[0]->mesa.Height;
+	x = nmesa->drawX;
+	y = nmesa->drawY;
+
+	if (num_color != 1)
+		return GL_FALSE;
+
+        BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_VIEWPORT_HORIZ, 6);
+        OUT_RING_CACHE((w << 16) | x);
+        OUT_RING_CACHE((h << 16) | y);
+	depth_pitch = (depth ? depth->pitch : color[0]->pitch);
+	pitch = (depth_pitch<<16) | color[0]->pitch;
+	format = 0x108;
+	if (color[0]->mesa._ActualFormat != GL_RGBA8) {
+		format = 0x103; /* R5G6B5 color buffer */
+	}
+	OUT_RING_CACHE(format);
+	OUT_RING_CACHE(pitch);
+	OUT_RING_CACHE(color[0]->offset);
+	OUT_RING_CACHE(depth ? depth->offset : color[0]->offset);
+
+	/* Always set to bottom left of buffer */
+	BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_X, 4);
+	OUT_RING_CACHEf (0.0);
+	OUT_RING_CACHEf ((GLfloat) h);
+	OUT_RING_CACHEf (0.0);
+	OUT_RING_CACHEf (0.0);
+
+	return GL_TRUE;
+}
+
+void nv10InitStateFuncs(GLcontext *ctx, struct dd_function_table *func)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	func->AlphaFunc			= nv10AlphaFunc;
+	func->BlendColor		= nv10BlendColor;
+	func->BlendEquationSeparate	= nv10BlendEquationSeparate;
+	func->BlendFuncSeparate		= nv10BlendFuncSeparate;
+	func->Clear			= nv10Clear;
+	func->ClearColor		= nv10ClearColor;
+	func->ClearDepth		= nv10ClearDepth;
+	func->ClearStencil		= nv10ClearStencil;
+	func->ClipPlane			= nv10ClipPlane;
+	func->ColorMask			= nv10ColorMask;
+	func->ColorMaterial		= nv10ColorMaterial;
+	func->CullFace			= nv10CullFace;
+	func->FrontFace			= nv10FrontFace;
+	func->DepthFunc			= nv10DepthFunc;
+	func->DepthMask			= nv10DepthMask;
+	func->DepthRange		= nv10DepthRange;
+	func->Enable			= nv10Enable;
+	func->Fogfv			= nv10Fogfv;
+	func->Hint			= nv10Hint;
+	func->Lightfv			= nv10Lightfv;
+/*	func->LightModelfv		= nv10LightModelfv; */
+	func->LineStipple		= nv10LineStipple;		/* Not for NV10 */
+	func->LineWidth			= nv10LineWidth;
+	func->LogicOpcode		= nv10LogicOpcode;
+	func->PointParameterfv		= nv10PointParameterfv;
+	func->PointSize			= nv10PointSize;
+	func->PolygonMode		= nv10PolygonMode;
+	func->PolygonOffset		= nv10PolygonOffset;
+	func->PolygonStipple		= nv10PolygonStipple;		/* Not for NV10 */
+/*	func->ReadBuffer		= nv10ReadBuffer;*/
+/*	func->RenderMode		= nv10RenderMode;*/
+	func->Scissor			= nv10Scissor;
+	func->ShadeModel		= nv10ShadeModel;
+	func->StencilFuncSeparate	= nv10StencilFuncSeparate;
+	func->StencilMaskSeparate	= nv10StencilMaskSeparate;
+	func->StencilOpSeparate		= nv10StencilOpSeparate;
+/*	func->TexGen			= nv10TexGen;*/
+/*	func->TexParameter		= nv10TexParameter;*/
+	func->TextureMatrix		= nv10TextureMatrix;
+
+	nmesa->hw_func.InitCard		= nv10InitCard;
+	nmesa->hw_func.BindBuffers	= nv10BindBuffers;
+	nmesa->hw_func.WindowMoved	= nv10WindowMoved;
+}
diff --git a/src/mesa/drivers/dri/nouveau/nv10_swtcl.c b/src/mesa/drivers/dri/nouveau/nv10_swtcl.c
new file mode 100644
index 00000000000..32da40661b4
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv10_swtcl.c
@@ -0,0 +1,647 @@
+/*
+ * Copyright 1998-2003 VIA Technologies, Inc. All Rights Reserved.
+ * Copyright 2001-2003 S3 Graphics, Inc. All Rights Reserved.
+ * Copyright 2006 Stephane Marchesin. 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"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sub license,
+ * 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ * VIA, S3 GRAPHICS, AND/OR ITS SUPPLIERS 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.
+ */
+
+/* Software TCL for NV10, NV20, NV30, NV40, NV50 */
+
+#include <stdio.h>
+#include <math.h>
+
+#include "glheader.h"
+#include "context.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "colormac.h"
+#include "enums.h"
+
+#include "swrast/swrast.h"
+#include "swrast_setup/swrast_setup.h"
+#include "tnl/t_context.h"
+#include "tnl/t_pipeline.h"
+
+#include "nouveau_swtcl.h"
+#include "nv10_swtcl.h"
+#include "nouveau_context.h"
+#include "nouveau_span.h"
+#include "nouveau_reg.h"
+#include "nouveau_tex.h"
+#include "nouveau_fifo.h"
+#include "nouveau_msg.h"
+#include "nouveau_object.h"
+
+static void nv10RasterPrimitive( GLcontext *ctx, GLenum rprim, GLuint hwprim );
+static void nv10RenderPrimitive( GLcontext *ctx, GLenum prim );
+static void nv10ResetLineStipple( GLcontext *ctx );
+
+
+
+static inline void nv10StartPrimitive(struct nouveau_context* nmesa,uint32_t primitive,uint32_t size)
+{
+	if (nmesa->screen->card->type==NV_10)
+		BEGIN_RING_SIZE(NvSub3D,NV10_TCL_PRIMITIVE_3D_BEGIN_END,1);
+	else if (nmesa->screen->card->type==NV_20)
+		BEGIN_RING_SIZE(NvSub3D,NV20_TCL_PRIMITIVE_3D_BEGIN_END,1);
+	else
+		BEGIN_RING_SIZE(NvSub3D,NV30_TCL_PRIMITIVE_3D_BEGIN_END,1);
+	OUT_RING(primitive);
+
+	if (nmesa->screen->card->type==NV_10)
+		BEGIN_RING_SIZE(NvSub3D,NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_DATA|NONINC_METHOD,size);
+	else if (nmesa->screen->card->type==NV_20)
+		BEGIN_RING_SIZE(NvSub3D,NV20_TCL_PRIMITIVE_3D_VERTEX_DATA|NONINC_METHOD,size);
+	else
+		BEGIN_RING_SIZE(NvSub3D,NV30_TCL_PRIMITIVE_3D_VERTEX_DATA|NONINC_METHOD,size);
+}
+
+inline void nv10FinishPrimitive(struct nouveau_context *nmesa)
+{
+	if (nmesa->screen->card->type==NV_10)
+		BEGIN_RING_SIZE(NvSub3D,NV10_TCL_PRIMITIVE_3D_BEGIN_END,1);
+	else if (nmesa->screen->card->type==NV_20)
+		BEGIN_RING_SIZE(NvSub3D,NV20_TCL_PRIMITIVE_3D_BEGIN_END,1);
+	else
+		BEGIN_RING_SIZE(NvSub3D,NV30_TCL_PRIMITIVE_3D_BEGIN_END,1);
+	OUT_RING(0x0);
+	FIRE_RING();
+}
+
+
+static inline void nv10ExtendPrimitive(struct nouveau_context* nmesa, int size)
+{
+	/* make sure there's enough room. if not, wait */
+	if (RING_AVAILABLE()<size)
+	{
+		WAIT_RING(nmesa,size);
+	}
+}
+
+/**********************************************************************/
+/*               Render unclipped begin/end objects                   */
+/**********************************************************************/
+
+static inline void nv10_render_generic_primitive_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags,GLuint prim)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	GLuint size_dword = vertsize*(count-start)/4;
+
+	nv10ExtendPrimitive(nmesa, size_dword);
+	nv10StartPrimitive(nmesa,prim+1,size_dword);
+	OUT_RINGp((nouveauVertex*)(vertptr+(start*vertsize)),size_dword);
+	nv10FinishPrimitive(nmesa);
+}
+
+static void nv10_render_points_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_POINTS);
+}
+
+static void nv10_render_lines_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_LINES);
+}
+
+static void nv10_render_line_strip_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_LINE_STRIP);
+}
+
+static void nv10_render_line_loop_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_LINE_LOOP);
+}
+
+static void nv10_render_triangles_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_TRIANGLES);
+}
+
+static void nv10_render_tri_strip_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_TRIANGLE_STRIP);
+}
+
+static void nv10_render_tri_fan_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_TRIANGLE_FAN);
+}
+
+static void nv10_render_quads_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_QUADS);
+}
+
+static void nv10_render_quad_strip_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_QUAD_STRIP);
+}
+
+static void nv10_render_poly_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_verts(ctx,start,count,flags,GL_POLYGON);
+}
+
+static void nv10_render_noop_verts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+}
+
+static void (*nv10_render_tab_verts[GL_POLYGON+2])(GLcontext *,
+							   GLuint,
+							   GLuint,
+							   GLuint) =
+{
+   nv10_render_points_verts,
+   nv10_render_lines_verts,
+   nv10_render_line_loop_verts,
+   nv10_render_line_strip_verts,
+   nv10_render_triangles_verts,
+   nv10_render_tri_strip_verts,
+   nv10_render_tri_fan_verts,
+   nv10_render_quads_verts,
+   nv10_render_quad_strip_verts,
+   nv10_render_poly_verts,
+   nv10_render_noop_verts,
+};
+
+
+static inline void nv10_render_generic_primitive_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags,GLuint prim)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	GLuint size_dword = vertsize*(count-start)/4;
+	const GLuint * const elt = TNL_CONTEXT(ctx)->vb.Elts;
+	GLuint j;
+
+	nv10ExtendPrimitive(nmesa, size_dword);
+	nv10StartPrimitive(nmesa,prim+1,size_dword);
+	for (j=start; j<count; j++ ) {
+		OUT_RINGp((nouveauVertex*)(vertptr+(elt[j]*vertsize)),vertsize/4);
+	}
+	nv10FinishPrimitive(nmesa);
+}
+
+static void nv10_render_points_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_POINTS);
+}
+
+static void nv10_render_lines_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_LINES);
+}
+
+static void nv10_render_line_strip_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_LINE_STRIP);
+}
+
+static void nv10_render_line_loop_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_LINE_LOOP);
+}
+
+static void nv10_render_triangles_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_TRIANGLES);
+}
+
+static void nv10_render_tri_strip_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_TRIANGLE_STRIP);
+}
+
+static void nv10_render_tri_fan_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_TRIANGLE_FAN);
+}
+
+static void nv10_render_quads_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_QUADS);
+}
+
+static void nv10_render_quad_strip_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_QUAD_STRIP);
+}
+
+static void nv10_render_poly_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+	nv10_render_generic_primitive_elts(ctx,start,count,flags,GL_POLYGON);
+}
+
+static void nv10_render_noop_elts(GLcontext *ctx,GLuint start,GLuint count,GLuint flags)
+{
+}
+
+static void (*nv10_render_tab_elts[GL_POLYGON+2])(GLcontext *,
+							   GLuint,
+							   GLuint,
+							   GLuint) =
+{
+   nv10_render_points_elts,
+   nv10_render_lines_elts,
+   nv10_render_line_loop_elts,
+   nv10_render_line_strip_elts,
+   nv10_render_triangles_elts,
+   nv10_render_tri_strip_elts,
+   nv10_render_tri_fan_elts,
+   nv10_render_quads_elts,
+   nv10_render_quad_strip_elts,
+   nv10_render_poly_elts,
+   nv10_render_noop_elts,
+};
+
+
+/**********************************************************************/
+/*                    Choose render functions                         */
+/**********************************************************************/
+
+
+#define EMIT_ATTR( ATTR, STYLE )					\
+do {									\
+   nmesa->vertex_attrs[nmesa->vertex_attr_count].attrib = (ATTR);	\
+   nmesa->vertex_attrs[nmesa->vertex_attr_count].format = (STYLE);	\
+   nmesa->vertex_attr_count++;						\
+} while (0)
+
+static void nv10_render_clipped_line(GLcontext *ctx,GLuint ii,GLuint jj)
+{
+
+}
+
+static void nv10_render_clipped_poly(GLcontext *ctx,const GLuint *elts,GLuint n)
+{
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+	struct vertex_buffer *VB = &tnl->vb;
+	GLuint *tmp = VB->Elts;
+	VB->Elts = (GLuint *)elts;
+	nv10_render_generic_primitive_elts( ctx, 0, n, PRIM_BEGIN|PRIM_END,GL_POLYGON );
+	VB->Elts = tmp;
+}
+
+static inline void nv10_render_points(GLcontext *ctx,GLuint first,GLuint last)
+{
+	WARN_ONCE("Unimplemented\n");
+}
+
+static inline void nv10_render_line(GLcontext *ctx,GLuint v1,GLuint v2)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	GLuint size_dword = vertsize*(2)/4;
+
+	nv10ExtendPrimitive(nmesa, size_dword);
+	nv10StartPrimitive(nmesa,GL_LINES+1,size_dword);
+	OUT_RINGp((nouveauVertex*)(vertptr+(v1*vertsize)),vertsize);
+	OUT_RINGp((nouveauVertex*)(vertptr+(v2*vertsize)),vertsize);
+	nv10FinishPrimitive(nmesa);
+}
+
+static inline void nv10_render_triangle(GLcontext *ctx,GLuint v1,GLuint v2,GLuint v3)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	GLuint size_dword = vertsize*(3)/4;
+
+	nv10ExtendPrimitive(nmesa, size_dword);
+	nv10StartPrimitive(nmesa,GL_TRIANGLES+1,size_dword);
+	OUT_RINGp((nouveauVertex*)(vertptr+(v1*vertsize)),vertsize);
+	OUT_RINGp((nouveauVertex*)(vertptr+(v2*vertsize)),vertsize);
+	OUT_RINGp((nouveauVertex*)(vertptr+(v3*vertsize)),vertsize);
+	nv10FinishPrimitive(nmesa);
+}
+
+static inline void nv10_render_quad(GLcontext *ctx,GLuint v1,GLuint v2,GLuint v3,GLuint v4)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte *vertptr = (GLubyte *)nmesa->verts;
+	GLuint vertsize = nmesa->vertex_size;
+	GLuint size_dword = vertsize*(4)/4;
+
+	nv10ExtendPrimitive(nmesa, size_dword);
+	nv10StartPrimitive(nmesa,GL_QUADS+1,size_dword);
+	OUT_RINGp((nouveauVertex*)(vertptr+(v1*vertsize)),vertsize);
+	OUT_RINGp((nouveauVertex*)(vertptr+(v2*vertsize)),vertsize);
+	OUT_RINGp((nouveauVertex*)(vertptr+(v3*vertsize)),vertsize);
+	OUT_RINGp((nouveauVertex*)(vertptr+(v4*vertsize)),vertsize);
+	nv10FinishPrimitive(nmesa);
+}
+
+
+
+static void nv10ChooseRenderState(GLcontext *ctx)
+{
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+
+	tnl->Driver.Render.PrimTabVerts = nv10_render_tab_verts;
+	tnl->Driver.Render.PrimTabElts = nv10_render_tab_elts;
+	tnl->Driver.Render.ClippedLine = nv10_render_clipped_line;
+	tnl->Driver.Render.ClippedPolygon = nv10_render_clipped_poly;
+	tnl->Driver.Render.Points = nv10_render_points;
+	tnl->Driver.Render.Line = nv10_render_line;
+	tnl->Driver.Render.Triangle = nv10_render_triangle;
+	tnl->Driver.Render.Quad = nv10_render_quad;
+}
+
+
+
+static inline void nv10OutputVertexFormat(struct nouveau_context* nmesa)
+{
+	GLcontext* ctx=nmesa->glCtx;
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+	DECLARE_RENDERINPUTS(index);
+	struct vertex_buffer *VB = &tnl->vb;
+	int attr_size[16];
+	int default_attr_size[8]={3,3,3,4,3,1,4,4};
+	int i;
+	int slots=0;
+	int total_size=0;
+	/* t_vertex_generic dereferences a NULL pointer if we
+	 * pass NULL as the vp transform...
+	 */
+	const GLfloat ident_vp[16] = {
+	   1.0, 0.0, 0.0, 0.0,
+	   0.0, 1.0, 0.0, 0.0,
+	   0.0, 0.0, 1.0, 0.0,
+	   0.0, 0.0, 0.0, 1.0
+	};
+
+	nmesa->vertex_attr_count = 0;
+	RENDERINPUTS_COPY(index, nmesa->render_inputs_bitset);
+
+	/*
+	 * Determine attribute sizes
+	 */
+	for(i=0;i<8;i++)
+	{
+		if (RENDERINPUTS_TEST(index, i))
+			attr_size[i]=default_attr_size[i];
+		else
+			attr_size[i]=0;
+	}
+	for(i=8;i<16;i++)
+	{
+		if (RENDERINPUTS_TEST(index, i))
+			attr_size[i]=VB->TexCoordPtr[i-8]->size;
+		else
+			attr_size[i]=0;
+	}
+
+	/*
+	 * Tell t_vertex about the vertex format
+	 */
+	for(i=0;i<16;i++)
+	{
+		if (RENDERINPUTS_TEST(index, i))
+		{
+			slots=i+1;
+			if (i==_TNL_ATTRIB_POS)
+			{
+				/* special-case POS */
+				EMIT_ATTR(_TNL_ATTRIB_POS,EMIT_3F_VIEWPORT);
+			}
+			else
+			{
+				switch(attr_size[i])
+				{
+					case 1:
+						EMIT_ATTR(i,EMIT_1F);
+						break;
+					case 2:
+						EMIT_ATTR(i,EMIT_2F);
+						break;
+					case 3:
+						EMIT_ATTR(i,EMIT_3F);
+						break;
+					case 4:
+						EMIT_ATTR(i,EMIT_4F);
+						break;
+				}
+			}
+			if (i==_TNL_ATTRIB_COLOR0)
+				nmesa->color_offset=total_size;
+			if (i==_TNL_ATTRIB_COLOR1)
+				nmesa->specular_offset=total_size;
+			total_size+=attr_size[i];
+		}
+	}
+
+	nmesa->vertex_size=_tnl_install_attrs( ctx,
+			nmesa->vertex_attrs, 
+			nmesa->vertex_attr_count,
+			ident_vp, 0 );
+	assert(nmesa->vertex_size==total_size*4);
+
+	/* 
+	 * Tell the hardware about the vertex format
+	 */
+	if (nmesa->screen->card->type==NV_10) {
+		int size;
+
+#define NV_VERTEX_ATTRIBUTE_TYPE_FLOAT 2
+
+#define NV10_SET_VERTEX_ATTRIB(i,j) \
+	do {	\
+		size = attr_size[j] << 4;	\
+		size |= (attr_size[j]*4) << 8;	\
+		size |= NV_VERTEX_ATTRIBUTE_TYPE_FLOAT;	\
+		BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_VERTEX_ATTR(i),1);	\
+		OUT_RING_CACHE(size);	\
+	} while (0)
+
+		NV10_SET_VERTEX_ATTRIB(0, _TNL_ATTRIB_POS);
+		NV10_SET_VERTEX_ATTRIB(1, _TNL_ATTRIB_COLOR0);
+		NV10_SET_VERTEX_ATTRIB(2, _TNL_ATTRIB_COLOR1);
+		NV10_SET_VERTEX_ATTRIB(3, _TNL_ATTRIB_TEX0);
+		NV10_SET_VERTEX_ATTRIB(4, _TNL_ATTRIB_TEX1);
+		NV10_SET_VERTEX_ATTRIB(5, _TNL_ATTRIB_NORMAL);
+		NV10_SET_VERTEX_ATTRIB(6, _TNL_ATTRIB_WEIGHT);
+		NV10_SET_VERTEX_ATTRIB(7, _TNL_ATTRIB_FOG);
+
+		BEGIN_RING_CACHE(NvSub3D, NV10_TCL_PRIMITIVE_3D_VERTEX_ARRAY_VALIDATE,1);
+		OUT_RING_CACHE(0);
+	} else if (nmesa->screen->card->type==NV_20) {
+		for(i=0;i<16;i++)
+		{
+			int size=attr_size[i];
+			BEGIN_RING_CACHE(NvSub3D,NV20_TCL_PRIMITIVE_3D_VERTEX_ATTR(i),1);
+			OUT_RING_CACHE(NV_VERTEX_ATTRIBUTE_TYPE_FLOAT|(size*0x10));
+		}
+	} else {
+		BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_DO_VERTICES, 1);
+		OUT_RING(0);
+		BEGIN_RING_CACHE(NvSub3D,NV30_TCL_PRIMITIVE_3D_VERTEX_ATTR0_POS,slots);
+		for(i=0;i<slots;i++)
+		{
+			int size=attr_size[i];
+			OUT_RING_CACHE(NV_VERTEX_ATTRIBUTE_TYPE_FLOAT|(size*0x10));
+		}
+		// FIXME this is probably not needed
+		BEGIN_RING_SIZE(NvSub3D,NV30_TCL_PRIMITIVE_3D_VERTEX_UNK_0,1);
+		OUT_RING(0);
+		BEGIN_RING_SIZE(NvSub3D,NV30_TCL_PRIMITIVE_3D_VERTEX_UNK_0,1);
+		OUT_RING(0);
+		BEGIN_RING_SIZE(NvSub3D,NV30_TCL_PRIMITIVE_3D_VERTEX_UNK_0,1);
+		OUT_RING(0);
+	}
+}
+
+
+static void nv10ChooseVertexState( GLcontext *ctx )
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+	DECLARE_RENDERINPUTS(index);
+	
+	RENDERINPUTS_COPY(index, tnl->render_inputs_bitset);
+	if (!RENDERINPUTS_EQUAL(index, nmesa->render_inputs_bitset))
+	{
+		RENDERINPUTS_COPY(nmesa->render_inputs_bitset, index);
+		nv10OutputVertexFormat(nmesa);
+	}
+
+	if (nmesa->screen->card->type == NV_30) {
+		nouveauShader *fp;
+		
+		if (ctx->FragmentProgram.Enabled) {
+			fp = (nouveauShader *) ctx->FragmentProgram.Current;
+			nvsUpdateShader(ctx, fp);
+		} else
+			nvsUpdateShader(ctx, nmesa->passthrough_fp);
+	}
+
+	if (nmesa->screen->card->type >= NV_40) {
+		/* Ensure passthrough shader is being used, and mvp matrix
+		 * is up to date
+		 */
+		nvsUpdateShader(ctx, nmesa->passthrough_vp);
+
+		/* Update texenv shader / user fragprog */
+		nvsUpdateShader(ctx, (nouveauShader*)ctx->FragmentProgram._Current);
+	}
+}
+
+
+/**********************************************************************/
+/*                 High level hooks for t_vb_render.c                 */
+/**********************************************************************/
+
+
+static void nv10RenderStart(GLcontext *ctx)
+{
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (nmesa->new_state) {
+		nmesa->new_render_state |= nmesa->new_state;
+	}
+
+	if (nmesa->new_render_state) {
+		nv10ChooseVertexState(ctx);
+		nv10ChooseRenderState(ctx);
+		nmesa->new_render_state = 0;
+	}
+}
+
+static void nv10RenderFinish(GLcontext *ctx)
+{
+}
+
+
+/* System to flush dma and emit state changes based on the rasterized
+ * primitive.
+ */
+void nv10RasterPrimitive(GLcontext *ctx,
+		GLenum glprim,
+		GLuint hwprim)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+
+	assert (!nmesa->new_state);
+
+	if (hwprim != nmesa->current_primitive)
+	{
+		nmesa->current_primitive=hwprim;
+		
+	}
+}
+
+static const GLuint hw_prim[GL_POLYGON+1] = {
+	GL_POINTS+1,
+	GL_LINES+1,
+	GL_LINE_STRIP+1,
+	GL_LINE_LOOP+1,
+	GL_TRIANGLES+1,
+	GL_TRIANGLE_STRIP+1,
+	GL_TRIANGLE_FAN+1,
+	GL_QUADS+1,
+	GL_QUAD_STRIP+1,
+	GL_POLYGON+1
+};
+
+/* Callback for mesa:
+ */
+static void nv10RenderPrimitive( GLcontext *ctx, GLuint prim )
+{
+	nv10RasterPrimitive( ctx, prim, hw_prim[prim] );
+}
+
+static void nv10ResetLineStipple( GLcontext *ctx )
+{
+	/* FIXME do something here */
+	WARN_ONCE("Unimplemented nv10ResetLineStipple\n");
+}
+
+
+/**********************************************************************/
+/*                            Initialization.                         */
+/**********************************************************************/
+
+void nv10TriInitFunctions(GLcontext *ctx)
+{
+	struct nouveau_context *nmesa = NOUVEAU_CONTEXT(ctx);
+	TNLcontext *tnl = TNL_CONTEXT(ctx);
+
+	tnl->Driver.RunPipeline = nouveauRunPipeline;
+	tnl->Driver.Render.Start = nv10RenderStart;
+	tnl->Driver.Render.Finish = nv10RenderFinish;
+	tnl->Driver.Render.PrimitiveNotify = nv10RenderPrimitive;
+	tnl->Driver.Render.ResetLineStipple = nv10ResetLineStipple;
+	tnl->Driver.Render.BuildVertices = _tnl_build_vertices;
+	tnl->Driver.Render.CopyPV = _tnl_copy_pv;
+	tnl->Driver.Render.Interp = _tnl_interp;
+
+	_tnl_init_vertices( ctx, ctx->Const.MaxArrayLockSize + 12, 
+			64 * sizeof(GLfloat) );
+
+	nmesa->verts = (GLubyte *)tnl->clipspace.vertex_buf;
+}
+
+
diff --git a/src/mesa/drivers/dri/nouveau/nv10_swtcl.h b/src/mesa/drivers/dri/nouveau/nv10_swtcl.h
new file mode 100644
index 00000000000..7c854addd2d
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv10_swtcl.h
@@ -0,0 +1,40 @@
+/**************************************************************************
+
+Copyright 2006 Stephane Marchesin
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+
+
+#ifndef __NV10_SWTCL_H__
+#define __NV10_SWTCL_H__
+
+#include "mtypes.h"
+
+extern void nv10Fallback( GLcontext *ctx, GLuint bit, GLboolean mode );
+extern void nv10FinishPrimitive(struct nouveau_context *nmesa);
+extern void nv10TriInitFunctions(GLcontext *ctx);
+#define FALLBACK( nmesa, bit, mode ) nouveauFallback( nmesa->glCtx, bit, mode )
+
+#endif /* __NV10_SWTCL_H__ */
+
diff --git a/src/mesa/drivers/dri/nouveau/nv20_shader.h b/src/mesa/drivers/dri/nouveau/nv20_shader.h
new file mode 100644
index 00000000000..7d2e29db668
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv20_shader.h
@@ -0,0 +1,121 @@
+/* NV20_TCL_PRIMITIVE_3D_0x0B00 */
+#define NV20_VP_INST_0B00	0x00000000 /* always 0? */
+#define NV20_VP_INST0_KNOWN	0
+
+/* NV20_TCL_PRIMITIVE_3D_0x0B04 */
+#define NV20_VP_INST_SCA_OPCODE_SHIFT				25
+#define NV20_VP_INST_SCA_OPCODE_MASK				(0x0F << 25)
+#define NV20_VP_INST_OPCODE_RCP	0x2
+#define NV20_VP_INST_OPCODE_RCC	0x3
+#define NV20_VP_INST_OPCODE_RSQ	0x4
+#define NV20_VP_INST_OPCODE_EXP	0x5
+#define NV20_VP_INST_OPCODE_LOG	0x6
+#define NV20_VP_INST_OPCODE_LIT	0x7
+#define NV20_VP_INST_VEC_OPCODE_SHIFT				21
+#define NV20_VP_INST_VEC_OPCODE_MASK				(0x0F << 21)
+#define NV20_VP_INST_OPCODE_NOP	0x0 /* guess */
+#define NV20_VP_INST_OPCODE_MOV	0x1
+#define NV20_VP_INST_OPCODE_MUL	0x2
+#define NV20_VP_INST_OPCODE_ADD	0x3
+#define NV20_VP_INST_OPCODE_MAD	0x4
+#define NV20_VP_INST_OPCODE_DP3	0x5
+#define NV20_VP_INST_OPCODE_DPH	0x6
+#define NV20_VP_INST_OPCODE_DP4	0x7
+#define NV20_VP_INST_OPCODE_DST	0x8
+#define NV20_VP_INST_OPCODE_MIN	0x9
+#define NV20_VP_INST_OPCODE_MAX	0xA
+#define NV20_VP_INST_OPCODE_SLT	0xB
+#define NV20_VP_INST_OPCODE_SGE	0xC
+#define NV20_VP_INST_OPCODE_ARL	0xD
+#define NV20_VP_INST_CONST_SRC_SHIFT				13
+#define NV20_VP_INST_CONST_SRC_MASK				(0xFF << 13)
+#define NV20_VP_INST_INPUT_SRC_SHIFT				9
+#define NV20_VP_INST_INPUT_SRC_MASK				(0xF << 9) /* guess */
+#define NV20_VP_INST_INPUT_SRC_POS	0
+#define NV20_VP_INST_INPUT_SRC_COL0	3
+#define NV20_VP_INST_INPUT_SRC_COL1	4
+#define NV20_VP_INST_INPUT_SRC_TC(n)	(9+n)
+#define NV20_VP_INST_SRC0H_SHIFT				0
+#define NV20_VP_INST_SRC0H_MASK					(0x1FF << 0)
+#define NV20_VP_INST1_KNOWN	( \
+	NV20_VP_INST_OPCODE_MASK | \
+	NV20_VP_INST_CONST_SRC_MASK | \
+	NV20_VP_INST_INPUT_SRC_MASK | \
+	NV20_VP_INST_SRC0H_MASK \
+	)
+
+/* NV20_TCL_PRIMITIVE_3D_0x0B08 */
+#define NV20_VP_INST_SRC0L_SHIFT				26
+#define NV20_VP_INST_SRC0L_MASK					(0x3F  <<26)
+#define NV20_VP_INST_SRC1_SHIFT					11
+#define NV20_VP_INST_SRC1_MASK					(0x7FFF<<11)
+#define NV20_VP_INST_SRC2H_SHIFT				0
+#define NV20_VP_INST_SRC2H_MASK					(0x7FF << 0)
+
+/* NV20_TCL_PRIMITIVE_3D_0x0B0C */
+#define NV20_VP_INST_SRC2L_SHIFT				28
+#define NV20_VP_INST_SRC2L_MASK					(0x0F  <<28)
+#define NV20_VP_INST_VTEMP_WRITEMASK_SHIFT			24
+#define NV20_VP_INST_VTEMP_WRITEMASK_MASK			(0x0F  <<24)
+#    define NV20_VP_INST_TEMP_WRITEMASK_X	(1<<27)
+#    define NV20_VP_INST_TEMP_WRITEMASK_Y	(1<<26)
+#    define NV20_VP_INST_TEMP_WRITEMASK_Z	(1<<25)
+#    define NV20_VP_INST_TEMP_WRITEMASK_W	(1<<24)
+#define NV20_VP_INST_DEST_TEMP_ID_SHIFT				20
+#define NV20_VP_INST_DEST_TEMP_ID_MASK				(0x0F  <<20)
+#define NV20_VP_INST_STEMP_WRITEMASK_SHIFT			16
+#define NV20_VP_INST_STEMP_WRITEMASK_MASK			(0x0F  <<16)
+#    define NV20_VP_INST_STEMP_WRITEMASK_X	(1<<19)
+#    define NV20_VP_INST_STEMP_WRITEMASK_Y	(1<<18)
+#    define NV20_VP_INST_STEMP_WRITEMASK_Z	(1<<17)
+#    define NV20_VP_INST_STEMP_WRITEMASK_W	(1<<16)
+#define NV20_VP_INST_DEST_WRITEMASK_SHIFT			12
+#define NV20_VP_INST_DEST_WRITEMASK_MASK			(0x0F  <<12)
+#    define NV20_VP_INST_DEST_WRITEMASK_X	(1<<15)
+#    define NV20_VP_INST_DEST_WRITEMASK_Y	(1<<14)
+#    define NV20_VP_INST_DEST_WRITEMASK_Z	(1<<13)
+#    define NV20_VP_INST_DEST_WRITEMASK_W	(1<<12)
+#define NV20_VP_INST_DEST_SHIFT					3
+#define NV20_VP_INST_DEST_MASK					(0xF << 3) /* guess */
+#define NV20_VP_INST_DEST_POS	0
+#define NV20_VP_INST_DEST_COL0	3
+#define NV20_VP_INST_DEST_COL1	4
+#define NV20_VP_INST_DEST_TC(n)	(9+n)
+#define NV20_VP_INST_INDEX_CONST				(1<<1)
+#define NV20_VP_INST3_KNOWN ( \
+	NV20_VP_INST_SRC2L_MASK | \
+	NV20_VP_INST_TEMP_WRITEMASK_MASK | \
+	NV20_VP_INST_DEST_TEMP_ID_MASK | \
+	NV20_VP_INST_STEMP_WRITEMASK_MASK | \
+	NV20_VP_INST_DEST_WRITEMASK_MASK | \
+	NV20_VP_INST_DEST_MASK | \
+	NV20_VP_INST_INDEX_CONST \
+	)
+
+/* Useful to split the source selection regs into their pieces */
+#define NV20_VP_SRC0_HIGH_SHIFT 6
+#define NV20_VP_SRC0_HIGH_MASK  0x00007FC0
+#define NV20_VP_SRC0_LOW_MASK   0x0000003F
+#define NV20_VP_SRC2_HIGH_SHIFT 4
+#define NV20_VP_SRC2_HIGH_MASK  0x00007FF0
+#define NV20_VP_SRC2_LOW_MASK   0x0000000F
+
+#define NV20_VP_SRC_REG_NEGATE					(1<<14)
+#define NV20_VP_SRC_REG_SWZ_X_SHIFT				12
+#define NV20_VP_SRC_REG_SWZ_X_MASK				(0x03  <<12)
+#define NV20_VP_SRC_REG_SWZ_Y_SHIFT				10
+#define NV20_VP_SRC_REG_SWZ_Y_MASK				(0x03  <<10)
+#define NV20_VP_SRC_REG_SWZ_Z_SHIFT				8
+#define NV20_VP_SRC_REG_SWZ_Z_MASK				(0x03  << 8)
+#define NV20_VP_SRC_REG_SWZ_W_SHIFT				6
+#define NV20_VP_SRC_REG_SWZ_W_MASK				(0x03  << 6)
+#define NV20_VP_SRC_REG_SWZ_ALL_SHIFT				6
+#define NV20_VP_SRC_REG_SWZ_ALL_MASK				(0xFF  << 6)
+#define NV20_VP_SRC_REG_TEMP_ID_SHIFT				2
+#define NV20_VP_SRC_REG_TEMP_ID_MASK				(0x0F  << 0)
+#define NV20_VP_SRC_REG_TYPE_SHIFT				0
+#define NV20_VP_SRC_REG_TYPE_MASK				(0x03  << 0)
+#define NV20_VP_SRC_REG_TYPE_TEMP	1
+#define NV20_VP_SRC_REG_TYPE_INPUT	2
+#define NV20_VP_SRC_REG_TYPE_CONST	3	/* guess */
+
diff --git a/src/mesa/drivers/dri/nouveau/nv20_state.c b/src/mesa/drivers/dri/nouveau/nv20_state.c
new file mode 100644
index 00000000000..030713c0dbb
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv20_state.c
@@ -0,0 +1,811 @@
+/**************************************************************************
+
+Copyright 2006 Nouveau
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#include "nouveau_context.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+
+#include "tnl/t_pipeline.h"
+
+#include "mtypes.h"
+#include "colormac.h"
+
+static void nv20AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte ubRef;
+	CLAMPED_FLOAT_TO_UBYTE(ubRef, ref);
+
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_ALPHA_FUNC_FUNC, 2);
+	OUT_RING_CACHE(func);
+	OUT_RING_CACHE(ubRef);
+}
+
+static void nv20BlendColor(GLcontext *ctx, const GLfloat color[4])
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); 
+	GLubyte cf[4];
+
+	CLAMPED_FLOAT_TO_UBYTE(cf[0], color[0]);
+	CLAMPED_FLOAT_TO_UBYTE(cf[1], color[1]);
+	CLAMPED_FLOAT_TO_UBYTE(cf[2], color[2]);
+	CLAMPED_FLOAT_TO_UBYTE(cf[3], color[3]);
+
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_BLEND_COLOR, 1);
+	OUT_RING_CACHE(PACK_COLOR_8888(cf[3], cf[1], cf[2], cf[0]));
+}
+
+static void nv20BlendEquationSeparate(GLcontext *ctx, GLenum modeRGB, GLenum modeA)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_BLEND_EQUATION, 1);
+	OUT_RING_CACHE((modeA<<16) | modeRGB);
+}
+
+
+static void nv20BlendFuncSeparate(GLcontext *ctx, GLenum sfactorRGB, GLenum dfactorRGB,
+		GLenum sfactorA, GLenum dfactorA)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC, 2);
+	OUT_RING_CACHE((sfactorA<<16) | sfactorRGB);
+	OUT_RING_CACHE((dfactorA<<16) | dfactorRGB);
+}
+
+static void nv20Clear(GLcontext *ctx, GLbitfield mask)
+{
+	/* TODO */
+}
+
+static void nv20ClearColor(GLcontext *ctx, const GLfloat color[4])
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte c[4];
+	UNCLAMPED_FLOAT_TO_RGBA_CHAN(c,color);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_CLEAR_VALUE_ARGB, 1);
+	OUT_RING_CACHE(PACK_COLOR_8888(c[3],c[0],c[1],c[2]));
+}
+
+static void nv20ClearDepth(GLcontext *ctx, GLclampd d)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nmesa->clear_value=((nmesa->clear_value&0x000000FF)|(((uint32_t)(d*0xFFFFFF))<<8));
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_CLEAR_VALUE_DEPTH, 1);
+	OUT_RING_CACHE(nmesa->clear_value);
+}
+
+/* we're don't support indexed buffers
+   void (*ClearIndex)(GLcontext *ctx, GLuint index)
+ */
+
+static void nv20ClearStencil(GLcontext *ctx, GLint s)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nmesa->clear_value=((nmesa->clear_value&0xFFFFFF00)|(s&0x000000FF));
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_CLEAR_VALUE_DEPTH, 1);
+	OUT_RING_CACHE(nmesa->clear_value);
+}
+
+static void nv20ClipPlane(GLcontext *ctx, GLenum plane, const GLfloat *equation)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_CLIP_PLANE_A(plane), 4);
+	OUT_RING_CACHEf(equation[0]);
+	OUT_RING_CACHEf(equation[1]);
+	OUT_RING_CACHEf(equation[2]);
+	OUT_RING_CACHEf(equation[3]);
+}
+
+static void nv20ColorMask(GLcontext *ctx, GLboolean rmask, GLboolean gmask,
+		GLboolean bmask, GLboolean amask )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_COLOR_MASK, 1);
+	OUT_RING_CACHE(((amask && 0x01) << 24) | ((rmask && 0x01) << 16) | ((gmask && 0x01)<< 8) | ((bmask && 0x01) << 0));
+}
+
+static void nv20ColorMaterial(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	// TODO I need love
+}
+
+static void nv20CullFace(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_CULL_FACE, 1);
+	OUT_RING_CACHE(mode);
+}
+
+static void nv20FrontFace(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_FRONT_FACE, 1);
+	OUT_RING_CACHE(mode);
+}
+
+static void nv20DepthFunc(GLcontext *ctx, GLenum func)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_DEPTH_FUNC, 1);
+	OUT_RING_CACHE(func);
+}
+
+static void nv20DepthMask(GLcontext *ctx, GLboolean flag)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_DEPTH_WRITE_ENABLE, 1);
+	OUT_RING_CACHE(flag);
+}
+
+static void nv20DepthRange(GLcontext *ctx, GLclampd nearval, GLclampd farval)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_DEPTH_RANGE_NEAR, 2);
+	OUT_RING_CACHEf(nearval);
+	OUT_RING_CACHEf(farval);
+}
+
+/** Specify the current buffer for writing */
+//void (*DrawBuffer)( GLcontext *ctx, GLenum buffer );
+/** Specify the buffers for writing for fragment programs*/
+//void (*DrawBuffers)( GLcontext *ctx, GLsizei n, const GLenum *buffers );
+
+static void nv20Enable(GLcontext *ctx, GLenum cap, GLboolean state)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	switch(cap)
+	{
+		case GL_ALPHA_TEST:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_ALPHA_FUNC_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_AUTO_NORMAL:
+		case GL_BLEND:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_BLEND_FUNC_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_CLIP_PLANE0:
+		case GL_CLIP_PLANE1:
+		case GL_CLIP_PLANE2:
+		case GL_CLIP_PLANE3:
+		case GL_CLIP_PLANE4:
+		case GL_CLIP_PLANE5:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_CLIP_PLANE_ENABLE(cap-GL_CLIP_PLANE0), 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_COLOR_LOGIC_OP:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_COLOR_MATERIAL:
+//		case GL_COLOR_SUM_EXT:
+//		case GL_COLOR_TABLE:
+//		case GL_CONVOLUTION_1D:
+//		case GL_CONVOLUTION_2D:
+		case GL_CULL_FACE:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_CULL_FACE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_DEPTH_TEST:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_DEPTH_TEST_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_DITHER:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_DITHER_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_FOG:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_FOG_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_HISTOGRAM:
+//		case GL_INDEX_LOGIC_OP:
+		case GL_LIGHT0:
+		case GL_LIGHT1:
+		case GL_LIGHT2:
+		case GL_LIGHT3:
+		case GL_LIGHT4:
+		case GL_LIGHT5:
+		case GL_LIGHT6:
+		case GL_LIGHT7:
+			{
+			uint32_t mask=0x11<<(2*(cap-GL_LIGHT0));
+			nmesa->enabled_lights=((nmesa->enabled_lights&mask)|(mask*state));
+			if (nmesa->lighting_enabled)
+			{
+				BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_ENABLED_LIGHTS, 1);
+				OUT_RING_CACHE(nmesa->enabled_lights);
+			}
+			break;
+			}
+		case GL_LIGHTING:
+			nmesa->lighting_enabled=state;
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_ENABLED_LIGHTS, 1);
+			if (nmesa->lighting_enabled)
+				OUT_RING_CACHE(nmesa->enabled_lights);
+			else
+				OUT_RING_CACHE(0x0);
+			break;
+		case GL_LINE_SMOOTH:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LINE_SMOOTH_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_LINE_STIPPLE:
+//		case GL_MAP1_COLOR_4:
+//		case GL_MAP1_INDEX:
+//		case GL_MAP1_NORMAL:
+//		case GL_MAP1_TEXTURE_COORD_1:
+//		case GL_MAP1_TEXTURE_COORD_2:
+//		case GL_MAP1_TEXTURE_COORD_3:
+//		case GL_MAP1_TEXTURE_COORD_4:
+//		case GL_MAP1_VERTEX_3:
+//		case GL_MAP1_VERTEX_4:
+//		case GL_MAP2_COLOR_4:
+//		case GL_MAP2_INDEX:
+//		case GL_MAP2_NORMAL:
+//		case GL_MAP2_TEXTURE_COORD_1:
+//		case GL_MAP2_TEXTURE_COORD_2:
+//		case GL_MAP2_TEXTURE_COORD_3:
+//		case GL_MAP2_TEXTURE_COORD_4:
+//		case GL_MAP2_VERTEX_3:
+//		case GL_MAP2_VERTEX_4:
+//		case GL_MINMAX:
+		case GL_NORMALIZE:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_NORMALIZE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_POINT_SMOOTH:
+		case GL_POLYGON_OFFSET_POINT:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POLYGON_OFFSET_POINT_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_OFFSET_LINE:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POLYGON_OFFSET_LINE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_OFFSET_FILL:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FILL_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_SMOOTH:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POLYGON_SMOOTH_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_STIPPLE:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_POST_COLOR_MATRIX_COLOR_TABLE:
+//		case GL_POST_CONVOLUTION_COLOR_TABLE:
+//		case GL_RESCALE_NORMAL:
+		case GL_SCISSOR_TEST:
+			/* No enable bit, nv20Scissor will adjust to max range */
+			ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
+			      			 ctx->Scissor.Width, ctx->Scissor.Height);
+			break;
+//		case GL_SEPARABLE_2D:
+		case GL_STENCIL_TEST:
+			// TODO BACK and FRONT ?
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_STENCIL_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_TEXTURE_GEN_Q:
+//		case GL_TEXTURE_GEN_R:
+//		case GL_TEXTURE_GEN_S:
+//		case GL_TEXTURE_GEN_T:
+//		case GL_TEXTURE_1D:
+//		case GL_TEXTURE_2D:
+//		case GL_TEXTURE_3D:
+	}
+}
+
+static void nv20Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+    nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+    switch(pname)
+    {
+        case GL_FOG_MODE:
+            BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_FOG_MODE, 1);
+            //OUT_RING_CACHE (params);
+            break;
+            /* TODO: unsure about the rest.*/
+        default:
+            break;
+    }
+
+}
+   
+static void nv20Hint(GLcontext *ctx, GLenum target, GLenum mode)
+{
+	// TODO I need love (fog and line_smooth hints)
+}
+
+// void (*IndexMask)(GLcontext *ctx, GLuint mask);
+
+enum {
+	SPOTLIGHT_NO_UPDATE,
+	SPOTLIGHT_UPDATE_EXPONENT,
+	SPOTLIGHT_UPDATE_DIRECTION,
+	SPOTLIGHT_UPDATE_ALL
+};
+
+static void nv20Lightfv(GLcontext *ctx, GLenum light, GLenum pname, const GLfloat *params )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLint p = light - GL_LIGHT0;
+	struct gl_light *l = &ctx->Light.Light[p];
+	int spotlight_update = SPOTLIGHT_NO_UPDATE;
+
+	/* not sure where the fourth param value goes...*/
+	switch(pname)
+	{
+		case GL_AMBIENT:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_DIFFUSE:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_SPECULAR:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_POSITION:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_POSITION_X(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_SPOT_DIRECTION:
+			spotlight_update = SPOTLIGHT_UPDATE_DIRECTION;
+			break;
+		case GL_SPOT_EXPONENT:
+			spotlight_update = SPOTLIGHT_UPDATE_EXPONENT;
+			break;
+		case GL_SPOT_CUTOFF:
+			spotlight_update = SPOTLIGHT_UPDATE_ALL;
+			break;
+		case GL_CONSTANT_ATTENUATION:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_CONSTANT_ATTENUATION(p), 1);
+			OUT_RING_CACHEf(*params);
+			break;
+		case GL_LINEAR_ATTENUATION:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_LINEAR_ATTENUATION(p), 1);
+			OUT_RING_CACHEf(*params);
+			break;
+		case GL_QUADRATIC_ATTENUATION:
+			BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_QUADRATIC_ATTENUATION(p), 1);
+			OUT_RING_CACHEf(*params);
+			break;
+		default:
+			break;
+	}
+
+	switch(spotlight_update) {
+		case SPOTLIGHT_UPDATE_DIRECTION:
+			{
+				GLfloat x,y,z;
+				GLfloat spot_light_coef_a = 1.0 / (l->_CosCutoff - 1.0);
+				x = spot_light_coef_a * l->_NormDirection[0];
+				y = spot_light_coef_a * l->_NormDirection[1];
+				z = spot_light_coef_a * l->_NormDirection[2];
+				BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_X(p), 3);
+				OUT_RING_CACHEf(x);
+				OUT_RING_CACHEf(y);
+				OUT_RING_CACHEf(z);
+			}
+			break;
+		case SPOTLIGHT_UPDATE_EXPONENT:
+			{
+				GLfloat cc,lc,qc;
+				cc = 1.0;	/* FIXME: These need to be correctly computed */
+				lc = 0.0;
+				qc = 2.0;
+				BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_A(p), 3);
+				OUT_RING_CACHEf(cc);
+				OUT_RING_CACHEf(lc);
+				OUT_RING_CACHEf(qc);
+			}
+			break;
+		case SPOTLIGHT_UPDATE_ALL:
+			{
+				GLfloat cc,lc,qc, x,y,z, c;
+				GLfloat spot_light_coef_a = 1.0 / (l->_CosCutoff - 1.0);
+				cc = 1.0;	/* FIXME: These need to be correctly computed */
+				lc = 0.0;
+				qc = 2.0;
+				x = spot_light_coef_a * l->_NormDirection[0];
+				y = spot_light_coef_a * l->_NormDirection[1];
+				z = spot_light_coef_a * l->_NormDirection[2];
+				c = spot_light_coef_a + 1.0;
+				BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_A(p), 7);
+				OUT_RING_CACHEf(cc);
+				OUT_RING_CACHEf(lc);
+				OUT_RING_CACHEf(qc);
+				OUT_RING_CACHEf(x);
+				OUT_RING_CACHEf(y);
+				OUT_RING_CACHEf(z);
+				OUT_RING_CACHEf(c);
+			}
+			break;
+		default:
+			break;
+	}
+}
+
+/** Set the lighting model parameters */
+static void (*LightModelfv)(GLcontext *ctx, GLenum pname, const GLfloat *params);
+
+
+static void nv20LineStipple(GLcontext *ctx, GLint factor, GLushort pattern )
+{
+/*	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LINE_STIPPLE_PATTERN, 1);
+	OUT_RING_CACHE((pattern << 16) | factor);*/
+}
+
+static void nv20LineWidth(GLcontext *ctx, GLfloat width)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LINE_WIDTH, 1);
+	OUT_RING_CACHEf(width);
+}
+
+static void nv20LogicOpcode(GLcontext *ctx, GLenum opcode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_OP, 1);
+	OUT_RING_CACHE(opcode);
+}
+
+static void nv20PointParameterfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+	/*TODO: not sure what goes here. */
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	
+}
+
+/** Specify the diameter of rasterized points */
+static void nv20PointSize(GLcontext *ctx, GLfloat size)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POINT_SIZE, 1);
+	OUT_RING_CACHEf(size);
+}
+
+/** Select a polygon rasterization mode */
+static void nv20PolygonMode(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POLYGON_MODE_FRONT, 1);
+		OUT_RING_CACHE(mode);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POLYGON_MODE_BACK, 1);
+		OUT_RING_CACHE(mode);
+	}
+}
+
+/** Set the scale and units used to calculate depth values */
+static void nv20PolygonOffset(GLcontext *ctx, GLfloat factor, GLfloat units)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FACTOR, 2);
+        OUT_RING_CACHEf(factor);
+        OUT_RING_CACHEf(units);
+}
+
+/** Set the polygon stippling pattern */
+static void nv20PolygonStipple(GLcontext *ctx, const GLubyte *mask )
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_PATTERN(0), 32);
+        OUT_RING_CACHEp(mask, 32);
+}
+
+/* Specifies the current buffer for reading */
+void (*ReadBuffer)( GLcontext *ctx, GLenum buffer );
+/** Set rasterization mode */
+void (*RenderMode)(GLcontext *ctx, GLenum mode );
+
+/** Define the scissor box */
+static void nv20Scissor(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	/* There's no scissor enable bit, so adjust the scissor to cover the
+	 * maximum draw buffer bounds
+	 */
+	if (!ctx->Scissor.Enabled) {
+	   x = y = 0;
+	   w = h = 4095;
+	} else {
+	   x += nmesa->drawX;
+	   y += nmesa->drawY;
+	}
+
+        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SCISSOR_X2_X1, 2);
+        OUT_RING_CACHE(((x+w-1) << 16) | x);
+        OUT_RING_CACHE(((y+h-1) << 16) | y);
+}
+
+/** Select flat or smooth shading */
+static void nv20ShadeModel(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SHADE_MODEL, 1);
+	OUT_RING_CACHE(mode);
+}
+
+/** OpenGL 2.0 two-sided StencilFunc */
+static void nv20StencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func,
+		GLint ref, GLuint mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_STENCIL_FUNC_FUNC, 3);
+	OUT_RING_CACHE(func);
+	OUT_RING_CACHE(ref);
+	OUT_RING_CACHE(mask);
+}
+
+/** OpenGL 2.0 two-sided StencilMask */
+static void nv20StencilMaskSeparate(GLcontext *ctx, GLenum face, GLuint mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_STENCIL_MASK, 1);
+	OUT_RING_CACHE(mask);
+}
+
+/** OpenGL 2.0 two-sided StencilOp */
+static void nv20StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail,
+		GLenum zfail, GLenum zpass)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_STENCIL_OP_FAIL, 1);
+	OUT_RING_CACHE(fail);
+	OUT_RING_CACHE(zfail);
+	OUT_RING_CACHE(zpass);
+}
+
+/** Control the generation of texture coordinates */
+void (*TexGen)(GLcontext *ctx, GLenum coord, GLenum pname,
+		const GLfloat *params);
+/** Set texture environment parameters */
+void (*TexEnv)(GLcontext *ctx, GLenum target, GLenum pname,
+		const GLfloat *param);
+/** Set texture parameters */
+void (*TexParameter)(GLcontext *ctx, GLenum target,
+		struct gl_texture_object *texObj,
+		GLenum pname, const GLfloat *params);
+
+static void nv20TextureMatrix(GLcontext *ctx, GLuint unit, const GLmatrix *mat)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_TX_MATRIX(unit, 0), 16);
+        /*XXX: This SHOULD work.*/
+        OUT_RING_CACHEp(mat->m, 16);
+}
+
+/* Update anything that depends on the window position/size */
+static void nv20WindowMoved(nouveauContextPtr nmesa)
+{
+	GLcontext *ctx = nmesa->glCtx;
+	GLfloat *v = nmesa->viewport.m;
+	GLuint w = ctx->Viewport.Width;
+	GLuint h = ctx->Viewport.Height;
+	GLuint x = ctx->Viewport.X + nmesa->drawX;
+	GLuint y = ctx->Viewport.Y + nmesa->drawY;
+	int i;
+
+        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_VIEWPORT_HORIZ, 2);
+        OUT_RING_CACHE((w << 16) | x);
+        OUT_RING_CACHE((h << 16) | y);
+
+	BEGIN_RING_SIZE(NvSub3D, 0x02b4, 1);
+	OUT_RING(0);
+
+	BEGIN_RING_CACHE(NvSub3D,
+	      NV20_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_HORIZ(0), 1);
+        OUT_RING_CACHE((4095 << 16) | 0);
+	BEGIN_RING_CACHE(NvSub3D,
+	      NV20_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_VERT(0), 1);
+        OUT_RING_CACHE((4095 << 16) | 0);
+	for (i=1; i<8; i++) {
+		BEGIN_RING_CACHE(NvSub3D,
+		      NV20_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_HORIZ(i), 1);
+        	OUT_RING_CACHE(0);
+		BEGIN_RING_CACHE(NvSub3D,
+		      NV20_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_VERT(i), 1);
+	        OUT_RING_CACHE(0);
+	}
+
+	ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
+	      		    ctx->Scissor.Width, ctx->Scissor.Height);
+
+	/* TODO: recalc viewport scale coefs */
+}
+
+/* Initialise any card-specific non-GL related state */
+static GLboolean nv20InitCard(nouveauContextPtr nmesa)
+{
+	nouveauObjectOnSubchannel(nmesa, NvSub3D, Nv3D);
+
+	BEGIN_RING_SIZE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SET_OBJECT1, 2);
+	OUT_RING(NvDmaFB);	/* 184 dma_object1 */
+	OUT_RING(NvDmaFB);	/* 188 dma_object2 */
+	BEGIN_RING_SIZE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SET_OBJECT3, 2);
+	OUT_RING(NvDmaFB);	/* 194 dma_object3 */
+	OUT_RING(NvDmaFB);	/* 198 dma_object4 */
+	BEGIN_RING_SIZE(NvSub3D, NV20_TCL_PRIMITIVE_3D_SET_OBJECT8, 1);
+	OUT_RING(NvDmaFB);	/* 1a8 dma_object8 */
+
+	BEGIN_RING_SIZE(NvSub3D, 0x17e0, 3);
+	OUT_RINGf(0.0);
+	OUT_RINGf(0.0);
+	OUT_RINGf(1.0);
+
+	BEGIN_RING_SIZE(NvSub3D, 0x1e6c, 1);
+	OUT_RING(0x0db6);
+	BEGIN_RING_SIZE(NvSub3D, 0x0290, 1);
+	OUT_RING(0x00100001);
+	BEGIN_RING_SIZE(NvSub3D, 0x09fc, 1);
+	OUT_RING(0);
+	BEGIN_RING_SIZE(NvSub3D, 0x1d80, 1);
+	OUT_RING(1);
+	BEGIN_RING_SIZE(NvSub3D, 0x09f8, 1);
+	OUT_RING(4);
+
+	BEGIN_RING_SIZE(NvSub3D, 0x17ec, 3);
+	OUT_RINGf(0.0);
+	OUT_RINGf(1.0);
+	OUT_RINGf(0.0);
+
+	BEGIN_RING_SIZE(NvSub3D, 0x1d88, 1);
+	OUT_RING(3);
+
+	/* FIXME: More dma objects to setup ? */
+
+	BEGIN_RING_SIZE(NvSub3D, 0x1e98, 1);
+	OUT_RING(0);
+
+	BEGIN_RING_SIZE(NvSub3D, 0x120, 3);
+	OUT_RING(0);
+	OUT_RING(1);
+	OUT_RING(2);
+
+	return GL_TRUE;
+}
+
+/* Update buffer offset/pitch/format */
+static GLboolean nv20BindBuffers(nouveauContextPtr nmesa, int num_color,
+				 nouveau_renderbuffer **color,
+				 nouveau_renderbuffer *depth)
+{
+	GLuint x, y, w, h;
+	GLuint pitch, format, depth_pitch;
+
+	w = color[0]->mesa.Width;
+	h = color[0]->mesa.Height;
+	x = nmesa->drawX;
+	y = nmesa->drawY;
+
+	if (num_color != 1)
+		return GL_FALSE;
+
+        BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_VIEWPORT_HORIZ, 6);
+        OUT_RING_CACHE((w << 16) | x);
+        OUT_RING_CACHE((h << 16) | y);
+	depth_pitch = (depth ? depth->pitch : color[0]->pitch);
+	pitch = (depth_pitch<<16) | color[0]->pitch;
+	format = 0x128;
+	if (color[0]->mesa._ActualFormat != GL_RGBA8) {
+		format = 0x123; /* R5G6B5 color buffer */
+	}
+	OUT_RING_CACHE(format);
+	OUT_RING_CACHE(pitch);
+	OUT_RING_CACHE(color[0]->offset);
+	OUT_RING_CACHE(depth ? depth->offset : color[0]->offset);
+
+	if (depth) {
+		BEGIN_RING_SIZE(NvSub3D, NV20_TCL_PRIMITIVE_3D_LMA_DEPTH_BUFFER_PITCH, 2);
+		/* TODO: use a different buffer */
+		OUT_RING(depth->pitch);
+		OUT_RING(depth->offset);
+	}
+
+	/* Always set to bottom left of buffer */
+	BEGIN_RING_CACHE(NvSub3D, NV20_TCL_PRIMITIVE_3D_VIEWPORT_ORIGIN_X, 4);
+	OUT_RING_CACHEf (0.0);
+	OUT_RING_CACHEf ((GLfloat) h);
+	OUT_RING_CACHEf (0.0);
+	OUT_RING_CACHEf (0.0);
+
+	return GL_TRUE;
+}
+
+void nv20InitStateFuncs(GLcontext *ctx, struct dd_function_table *func)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	func->AlphaFunc			= nv20AlphaFunc;
+	func->BlendColor		= nv20BlendColor;
+	func->BlendEquationSeparate	= nv20BlendEquationSeparate;
+	func->BlendFuncSeparate		= nv20BlendFuncSeparate;
+	func->Clear			= nv20Clear;
+	func->ClearColor		= nv20ClearColor;
+	func->ClearDepth		= nv20ClearDepth;
+	func->ClearStencil		= nv20ClearStencil;
+	func->ClipPlane			= nv20ClipPlane;
+	func->ColorMask			= nv20ColorMask;
+	func->ColorMaterial		= nv20ColorMaterial;
+	func->CullFace			= nv20CullFace;
+	func->FrontFace			= nv20FrontFace;
+	func->DepthFunc			= nv20DepthFunc;
+	func->DepthMask			= nv20DepthMask;
+	func->DepthRange		= nv20DepthRange;
+	func->Enable			= nv20Enable;
+	func->Fogfv			= nv20Fogfv;
+	func->Hint			= nv20Hint;
+	func->Lightfv			= nv20Lightfv;
+/*	func->LightModelfv		= nv20LightModelfv; */
+	func->LineStipple		= nv20LineStipple;
+	func->LineWidth			= nv20LineWidth;
+	func->LogicOpcode		= nv20LogicOpcode;
+	func->PointParameterfv		= nv20PointParameterfv;
+	func->PointSize			= nv20PointSize;
+	func->PolygonMode		= nv20PolygonMode;
+	func->PolygonOffset		= nv20PolygonOffset;
+	func->PolygonStipple		= nv20PolygonStipple;
+/*	func->ReadBuffer		= nv20ReadBuffer;*/
+/*	func->RenderMode		= nv20RenderMode;*/
+	func->Scissor			= nv20Scissor;
+	func->ShadeModel		= nv20ShadeModel;
+	func->StencilFuncSeparate	= nv20StencilFuncSeparate;
+	func->StencilMaskSeparate	= nv20StencilMaskSeparate;
+	func->StencilOpSeparate		= nv20StencilOpSeparate;
+/*	func->TexGen			= nv20TexGen;*/
+/*	func->TexParameter		= nv20TexParameter;*/
+	func->TextureMatrix		= nv20TextureMatrix;
+
+	nmesa->hw_func.InitCard		= nv20InitCard;
+	nmesa->hw_func.BindBuffers	= nv20BindBuffers;
+	nmesa->hw_func.WindowMoved	= nv20WindowMoved;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nv20_vertprog.c b/src/mesa/drivers/dri/nouveau/nv20_vertprog.c
new file mode 100644
index 00000000000..60cfcd70561
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv20_vertprog.c
@@ -0,0 +1,447 @@
+#include "nouveau_context.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+
+#include "nouveau_shader.h"
+#include "nv20_shader.h"
+
+unsigned int NVVP_TX_VOP_COUNT = 16;
+unsigned int NVVP_TX_NVS_OP_COUNT = 16;
+struct _op_xlat NVVP_TX_VOP[32];
+struct _op_xlat NVVP_TX_SOP[32];
+
+nvsSwzComp NV20VP_TX_SWIZZLE[4] = { NVS_SWZ_X, NVS_SWZ_Y, NVS_SWZ_Z, NVS_SWZ_W };
+
+/*****************************************************************************
+ * Support routines
+ */
+static void
+NV20VPUploadToHW(GLcontext *ctx, nouveauShader *nvs)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   int i;
+
+   /* XXX: missing a way to say what insn we're uploading from, and possible
+    *      the program start position (if NV20 has one) */
+   for (i=0; i<nvs->program_size; i+=4) {
+      BEGIN_RING_SIZE(NvSub3D, NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_INST0, 4);
+      OUT_RING(nvs->program[i + 0]);
+      OUT_RING(nvs->program[i + 1]);
+      OUT_RING(nvs->program[i + 2]);
+      OUT_RING(nvs->program[i + 3]);
+   }
+}
+
+static void
+NV20VPUpdateConst(GLcontext *ctx, nouveauShader *nvs, int id)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+   /* Worth checking if the value *actually* changed? Mesa doesn't tell us this
+    * as far as I know..
+    */
+   BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_ID, 1);
+   OUT_RING (id);
+   BEGIN_RING_SIZE(NvSub3D, NV20_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_X, 4);
+   OUT_RINGf(nvs->params[id].source_val[0]);
+   OUT_RINGf(nvs->params[id].source_val[1]);
+   OUT_RINGf(nvs->params[id].source_val[2]);
+   OUT_RINGf(nvs->params[id].source_val[3]);
+}
+
+/*****************************************************************************
+ * Assembly routines
+ */
+
+/*****************************************************************************
+ * Disassembly routines
+ */
+void
+NV20VPTXSwizzle(int hwswz, nvsSwzComp *swz)
+{
+   swz[NVS_SWZ_X] = NV20VP_TX_SWIZZLE[(hwswz & 0xC0) >> 6];
+   swz[NVS_SWZ_Y] = NV20VP_TX_SWIZZLE[(hwswz & 0x30) >> 4];
+   swz[NVS_SWZ_Z] = NV20VP_TX_SWIZZLE[(hwswz & 0x0C) >> 2];
+   swz[NVS_SWZ_W] = NV20VP_TX_SWIZZLE[(hwswz & 0x03) >> 0];
+}
+
+static int
+NV20VPHasMergedInst(nvsFunc * shader)
+{
+   if (shader->GetOpcodeHW(shader, 0) != NV20_VP_INST_OPCODE_NOP &&
+       shader->GetOpcodeHW(shader, 1) != NV20_VP_INST_OPCODE_NOP)
+      printf
+	 ("\n\n*****both opcode fields have values - PLEASE REPORT*****\n");
+   return 0;
+}
+
+static int
+NV20VPIsLastInst(nvsFunc * shader)
+{
+   return ((shader->inst[3] & (1 << 0)) ? 1 : 0);
+}
+
+static int
+NV20VPGetOffsetNext(nvsFunc * shader)
+{
+   return 4;
+}
+
+static struct _op_xlat *
+NV20VPGetOPTXRec(nvsFunc * shader, int merged)
+{
+   struct _op_xlat *opr;
+   int op;
+
+   if (shader->GetOpcodeSlot(shader, merged)) {
+      opr = NVVP_TX_SOP;
+      op = shader->GetOpcodeHW(shader, 1);
+      if (op >= NVVP_TX_NVS_OP_COUNT)
+	 return NULL;
+   }
+   else {
+      opr = NVVP_TX_VOP;
+      op = shader->GetOpcodeHW(shader, 0);
+      if (op >= NVVP_TX_VOP_COUNT)
+	 return NULL;
+   }
+
+   if (opr[op].SOP == NVS_OP_UNKNOWN)
+      return NULL;
+   return &opr[op];
+}
+
+static struct _op_xlat *
+NV20VPGetOPTXFromSOP(nvsOpcode sop, int *id)
+{
+   int i;
+
+   for (i=0;i<NVVP_TX_VOP_COUNT;i++) {
+      if (NVVP_TX_VOP[i].SOP == sop) {
+	 if (id) *id = 0;
+	 return &NVVP_TX_VOP[i];
+      }
+   }
+
+   for (i=0;i<NVVP_TX_NVS_OP_COUNT;i++) {
+      if (NVVP_TX_SOP[i].SOP == sop) {
+	 if (id) *id = 1;
+	 return &NVVP_TX_SOP[i];
+      }
+   }
+
+   return NULL;
+}
+
+static int
+NV20VPGetOpcodeSlot(nvsFunc * shader, int merged)
+{
+   if (shader->HasMergedInst(shader))
+      return merged;
+   if (shader->GetOpcodeHW(shader, 0) == NV20_VP_INST_OPCODE_NOP)
+      return 1;
+   return 0;
+}
+
+static nvsOpcode
+NV20VPGetOpcode(nvsFunc * shader, int merged)
+{
+   struct _op_xlat *opr;
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr)
+      return NVS_OP_UNKNOWN;
+
+   return opr->SOP;
+}
+
+static nvsOpcode
+NV20VPGetOpcodeHW(nvsFunc * shader, int slot)
+{
+   if (slot)
+      return (shader->inst[1] & NV20_VP_INST_SCA_OPCODE_MASK)
+	 >> NV20_VP_INST_SCA_OPCODE_SHIFT;
+   return (shader->inst[1] & NV20_VP_INST_VEC_OPCODE_MASK)
+      >> NV20_VP_INST_VEC_OPCODE_SHIFT;
+}
+
+static nvsRegFile
+NV20VPGetDestFile(nvsFunc * shader, int merged)
+{
+   switch (shader->GetOpcode(shader, merged)) {
+   case NVS_OP_ARL:
+      return NVS_FILE_ADDRESS;
+   default:
+      /*FIXME: This probably isn't correct.. */
+      if ((shader->inst[3] & NV20_VP_INST_DEST_WRITEMASK_MASK) == 0)
+	 return NVS_FILE_TEMP;
+      return NVS_FILE_RESULT;
+   }
+}
+
+static unsigned int
+NV20VPGetDestID(nvsFunc * shader, int merged)
+{
+   int id;
+
+   switch (shader->GetDestFile(shader, merged)) {
+   case NVS_FILE_RESULT:
+      id = ((shader->inst[3] & NV20_VP_INST_DEST_MASK)
+	    >> NV20_VP_INST_DEST_SHIFT);
+      switch (id) {
+      case NV20_VP_INST_DEST_POS  : return NVS_FR_POSITION;
+      case NV20_VP_INST_DEST_COL0 : return NVS_FR_COL0;
+      case NV20_VP_INST_DEST_COL1 : return NVS_FR_COL1;
+      case NV20_VP_INST_DEST_TC(0): return NVS_FR_TEXCOORD0;
+      case NV20_VP_INST_DEST_TC(1): return NVS_FR_TEXCOORD1;
+      case NV20_VP_INST_DEST_TC(2): return NVS_FR_TEXCOORD2;
+      case NV20_VP_INST_DEST_TC(3): return NVS_FR_TEXCOORD3;
+      default:
+	 return -1;
+      }
+   case NVS_FILE_ADDRESS:
+      return 0;
+   case NVS_FILE_TEMP:
+      id = ((shader->inst[3] & NV20_VP_INST_DEST_TEMP_ID_MASK)
+	    >> NV20_VP_INST_DEST_TEMP_ID_SHIFT);
+      return id;
+   default:
+      return -1;
+   }
+}
+
+static unsigned int
+NV20VPGetDestMask(nvsFunc * shader, int merged)
+{
+   int hwmask, mask = 0;
+
+   /* Special handling for ARL - hardware only supports a
+    * 1-component address reg
+    */
+   if (shader->GetOpcode(shader, merged) == NVS_OP_ARL)
+      return SMASK_X;
+
+   if (shader->GetDestFile(shader, merged) == NVS_FILE_RESULT)
+      hwmask = (shader->inst[3] & NV20_VP_INST_DEST_WRITEMASK_MASK)
+	 >> NV20_VP_INST_DEST_WRITEMASK_SHIFT;
+   else if (shader->GetOpcodeSlot(shader, merged))
+      hwmask = (shader->inst[3] & NV20_VP_INST_STEMP_WRITEMASK_MASK)
+	 >> NV20_VP_INST_STEMP_WRITEMASK_SHIFT;
+   else
+      hwmask = (shader->inst[3] & NV20_VP_INST_VTEMP_WRITEMASK_MASK)
+	 >> NV20_VP_INST_VTEMP_WRITEMASK_SHIFT;
+
+   if (hwmask & (1 << 3)) mask |= SMASK_X;
+   if (hwmask & (1 << 2)) mask |= SMASK_Y;
+   if (hwmask & (1 << 1)) mask |= SMASK_Z;
+   if (hwmask & (1 << 0)) mask |= SMASK_W;
+
+   return mask;
+}
+
+static unsigned int
+NV20VPGetSourceHW(nvsFunc * shader, int merged, int pos)
+{
+   struct _op_xlat *opr;
+   unsigned int src;
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr)
+      return -1;
+
+   switch (opr->srcpos[pos]) {
+   case 0:
+      src = ((shader->inst[1] & NV20_VP_INST_SRC0H_MASK)
+	     >> NV20_VP_INST_SRC0H_SHIFT)
+	 << NV20_VP_SRC0_HIGH_SHIFT;
+      src |= ((shader->inst[2] & NV20_VP_INST_SRC0L_MASK)
+	      >> NV20_VP_INST_SRC0L_SHIFT);
+      break;
+   case 1:
+      src = ((shader->inst[2] & NV20_VP_INST_SRC1_MASK)
+	     >> NV20_VP_INST_SRC1_SHIFT);
+      break;
+   case 2:
+      src = ((shader->inst[2] & NV20_VP_INST_SRC2H_MASK)
+	     >> NV20_VP_INST_SRC2H_SHIFT)
+	 << NV20_VP_SRC2_HIGH_SHIFT;
+      src |= ((shader->inst[3] & NV20_VP_INST_SRC2L_MASK)
+	      >> NV20_VP_INST_SRC2L_SHIFT);
+      break;
+   default:
+      src = -1;
+   }
+
+   return src;
+}
+
+static nvsRegFile
+NV20VPGetSourceFile(nvsFunc * shader, int merged, int pos)
+{
+   unsigned int src;
+   struct _op_xlat *opr;
+   int file;
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr || opr->srcpos[pos] == -1)
+      return -1;
+
+   switch (opr->srcpos[pos]) {
+   case SPOS_ADDRESS:
+      return NVS_FILE_ADDRESS;
+   default:
+      src = NV20VPGetSourceHW(shader, merged, pos);
+      file = (src & NV20_VP_SRC_REG_TYPE_MASK) >> NV20_VP_SRC_REG_TYPE_SHIFT;
+
+      switch (file) {
+      case NV20_VP_SRC_REG_TYPE_TEMP : return NVS_FILE_TEMP;
+      case NV20_VP_SRC_REG_TYPE_INPUT: return NVS_FILE_ATTRIB;
+      case NV20_VP_SRC_REG_TYPE_CONST: return NVS_FILE_CONST;
+      default:
+	 return NVS_FILE_UNKNOWN;
+      }
+   }
+}
+
+static int
+NV20VPGetSourceID(nvsFunc * shader, int merged, int pos)
+{
+   unsigned int src;
+
+   switch (shader->GetSourceFile(shader, merged, pos)) {
+   case NVS_FILE_TEMP:
+      src = shader->GetSourceHW(shader, merged, pos);
+      return ((src & NV20_VP_SRC_REG_TEMP_ID_MASK) >>
+	      NV20_VP_SRC_REG_TEMP_ID_SHIFT);
+   case NVS_FILE_CONST:
+      return ((shader->inst[1] & NV20_VP_INST_CONST_SRC_MASK)
+	      >> NV20_VP_INST_CONST_SRC_SHIFT);
+   case NVS_FILE_ATTRIB:
+      src = ((shader->inst[1] & NV20_VP_INST_INPUT_SRC_MASK)
+	     >> NV20_VP_INST_INPUT_SRC_SHIFT);
+      switch (src) {
+      case NV20_VP_INST_INPUT_SRC_POS  : return NVS_FR_POSITION;
+      case NV20_VP_INST_INPUT_SRC_COL0 : return NVS_FR_COL0;
+      case NV20_VP_INST_INPUT_SRC_COL1 : return NVS_FR_COL1;
+      case NV20_VP_INST_INPUT_SRC_TC(0): return NVS_FR_TEXCOORD0;
+      case NV20_VP_INST_INPUT_SRC_TC(1): return NVS_FR_TEXCOORD1;
+      case NV20_VP_INST_INPUT_SRC_TC(2): return NVS_FR_TEXCOORD2;
+      case NV20_VP_INST_INPUT_SRC_TC(3): return NVS_FR_TEXCOORD3;
+      default:
+	 return NVS_FR_UNKNOWN;
+      }
+   default:
+      return -1;
+   }
+}
+
+static int
+NV20VPGetSourceNegate(nvsFunc * shader, int merged, int pos)
+{
+   unsigned int src;
+
+   src = shader->GetSourceHW(shader, merged, pos);
+
+   return ((src & NV20_VP_SRC_REG_NEGATE) ? 1 : 0);
+}
+
+static int
+NV20VPGetSourceAbs(nvsFunc * shader, int merged, int pos)
+{
+   /* NV20 can't do ABS on sources?  Appears to be emulated with
+    *      MAX reg, reg, -reg
+    */
+   return 0;
+}
+
+static void
+NV20VPGetSourceSwizzle(nvsFunc * shader, int merged, int pos, nvsSwzComp *swz)
+{
+   unsigned int src;
+   int swzbits;
+
+   src = shader->GetSourceHW(shader, merged, pos);
+   swzbits  =
+      (src & NV20_VP_SRC_REG_SWZ_ALL_MASK) >> NV20_VP_SRC_REG_SWZ_ALL_SHIFT;
+   return NV20VPTXSwizzle(swzbits, swz);
+}
+
+static int
+NV20VPGetSourceIndexed(nvsFunc * shader, int merged, int pos)
+{
+   /* I don't think NV20 can index into attribs, at least no GL
+    * extension is exposed that will allow it.
+    */
+   if (shader->GetSourceFile(shader, merged, pos) != NVS_FILE_CONST)
+      return 0;
+   if (shader->inst[3] & NV20_VP_INST_INDEX_CONST)
+      return 1;
+   return 0;
+}
+
+static int
+NV20VPGetAddressRegID(nvsFunc * shader)
+{
+   /* Only 1 address reg */
+   return 0;
+}
+
+static nvsSwzComp
+NV20VPGetAddressRegSwizzle(nvsFunc * shader)
+{
+   /* Only A0.x available */
+   return NVS_SWZ_X;
+}
+
+void
+NV20VPInitShaderFuncs(nvsFunc * shader)
+{
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_NOP, NVS_OP_NOP, -1, -1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_MOV, NVS_OP_MOV,  0, -1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_MUL, NVS_OP_MUL,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_ADD, NVS_OP_ADD,  0,  2, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_MAD, NVS_OP_MAD,  0,  1,  2);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_DP3, NVS_OP_DP3,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_DPH, NVS_OP_DPH,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_DP4, NVS_OP_DP4,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_DST, NVS_OP_DST,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_MIN, NVS_OP_MIN,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_MAX, NVS_OP_MAX,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_SLT, NVS_OP_SLT,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_SGE, NVS_OP_SGE,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV20_VP_INST_OPCODE_ARL, NVS_OP_ARL,  0, -1, -1);
+
+   MOD_OPCODE(NVVP_TX_SOP, NV20_VP_INST_OPCODE_NOP, NVS_OP_NOP, -1, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV20_VP_INST_OPCODE_RCP, NVS_OP_RCP,  2, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV20_VP_INST_OPCODE_RCC, NVS_OP_RCC,  2, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV20_VP_INST_OPCODE_RSQ, NVS_OP_RSQ,  2, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV20_VP_INST_OPCODE_EXP, NVS_OP_EXP,  2, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV20_VP_INST_OPCODE_LOG, NVS_OP_LOG,  2, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV20_VP_INST_OPCODE_LIT, NVS_OP_LIT,  2, -1, -1);
+
+   shader->UploadToHW		= NV20VPUploadToHW;
+   shader->UpdateConst		= NV20VPUpdateConst;
+
+   shader->GetOPTXRec		= NV20VPGetOPTXRec;
+   shader->GetOPTXFromSOP	= NV20VPGetOPTXFromSOP;
+
+   shader->HasMergedInst	= NV20VPHasMergedInst;
+   shader->IsLastInst		= NV20VPIsLastInst;
+   shader->GetOffsetNext	= NV20VPGetOffsetNext;
+   shader->GetOpcodeSlot	= NV20VPGetOpcodeSlot;
+   shader->GetOpcode		= NV20VPGetOpcode;
+   shader->GetOpcodeHW		= NV20VPGetOpcodeHW;
+   shader->GetDestFile		= NV20VPGetDestFile;
+   shader->GetDestID		= NV20VPGetDestID;
+   shader->GetDestMask		= NV20VPGetDestMask;
+   shader->GetSourceHW		= NV20VPGetSourceHW;
+   shader->GetSourceFile	= NV20VPGetSourceFile;
+   shader->GetSourceID		= NV20VPGetSourceID;
+   shader->GetSourceNegate	= NV20VPGetSourceNegate;
+   shader->GetSourceAbs		= NV20VPGetSourceAbs;
+   shader->GetSourceSwizzle	= NV20VPGetSourceSwizzle;
+   shader->GetSourceIndexed	= NV20VPGetSourceIndexed;
+   shader->GetRelAddressRegID	= NV20VPGetAddressRegID;
+   shader->GetRelAddressSwizzle	= NV20VPGetAddressRegSwizzle;
+}
diff --git a/src/mesa/drivers/dri/nouveau/nv30_fragprog.c b/src/mesa/drivers/dri/nouveau/nv30_fragprog.c
new file mode 100644
index 00000000000..e32452361e8
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv30_fragprog.c
@@ -0,0 +1,742 @@
+#include <stdint.h>
+
+#include "glheader.h"
+#include "macros.h"
+
+#include "nouveau_context.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+#include "nouveau_drm.h"
+#include "nouveau_shader.h"
+#include "nouveau_object.h"
+#include "nouveau_msg.h"
+#include "nouveau_bufferobj.h"
+#include "nv30_shader.h"
+
+unsigned int NVFP_TX_AOP_COUNT = 64;
+struct _op_xlat NVFP_TX_AOP[64];
+
+/*******************************************************************************
+ * Support routines
+ */
+
+static void
+NV30FPUploadToHW(GLcontext *ctx, nouveauShader *nvs)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   nvsCardPriv *priv = &nvs->card_priv;
+   uint32_t offset;
+
+   if (!nvs->program_buffer)
+      nvs->program_buffer = ctx->Driver.NewBufferObject(ctx, 0,
+							GL_ARRAY_BUFFER_ARB);
+
+   /* Should use STATIC_DRAW_ARB if shader doesn't use changable params */
+   nouveau_bo_init_storage(ctx, NOUVEAU_BO_VRAM_OK,
+	 		  nvs->program_size * sizeof(uint32_t),
+			  (const GLvoid *)nvs->program,
+			  GL_DYNAMIC_DRAW_ARB,
+			  nvs->program_buffer);
+
+   offset = nouveau_bo_gpu_ref(ctx, nvs->program_buffer);
+
+   /* Not using state cache here, updated programs at the same address don't
+    * seem to take effect unless the ACTIVE_PROGRAM method is called again.
+    * HW caches the program somewhere?
+    */
+   BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_FP_ACTIVE_PROGRAM, 1);
+   OUT_RING       (offset | 1);
+   if (nmesa->screen->card->type == NV_30) {  
+	   BEGIN_RING_SIZE(NvSub3D,
+			   0x1d60 /*NV30_TCL_PRIMITIVE_3D_FP_CONTROL*/, 1);
+	   OUT_RING       ((priv->NV30FP.uses_kil << 7));
+	   BEGIN_RING_SIZE(NvSub3D, 0x1450, 1);
+	   OUT_RING       (priv->NV30FP.num_regs << 16);
+   } else {
+	   BEGIN_RING_SIZE(NvSub3D,
+			   0x1d60 /*NV30_TCL_PRIMITIVE_3D_FP_CONTROL*/, 1);
+	   OUT_RING       ((priv->NV30FP.uses_kil <<  7) |
+			   (priv->NV30FP.num_regs << 24));
+   }
+}
+
+static void
+NV30FPUpdateConst(GLcontext *ctx, nouveauShader *nvs, int id)
+{
+   uint32_t *new     = nvs->params[id].source_val ?
+      (uint32_t*)nvs->params[id].source_val : (uint32_t*)nvs->params[id].val;
+   uint32_t *current;
+   int i;
+
+   for (i=0; i<nvs->params[id].hw_index_cnt; i++) {
+      current = nvs->program + nvs->params[id].hw_index[i];
+      COPY_4V(current, new);
+   }
+   nvs->on_hardware = 0;
+}
+
+/*******************************************************************************
+ * Assembly helpers
+ */
+static struct _op_xlat *
+NV30FPGetOPTXFromSOP(nvsOpcode op, int *id)
+{
+   int i;
+
+   for (i=0; i<NVFP_TX_AOP_COUNT; i++) {
+      if (NVFP_TX_AOP[i].SOP == op) {
+	 if (id) *id = 0;
+	 return &NVFP_TX_AOP[i];
+      }
+   }
+
+   return NULL;
+}
+
+static int
+NV30FPSupportsOpcode(nvsFunc *shader, nvsOpcode op)
+{
+   if (shader->GetOPTXFromSOP(op, NULL))
+      return 1;
+   return 0;
+}
+
+static void
+NV30FPSetOpcode(nvsFunc *shader, unsigned int opcode, int slot)
+{
+   if (opcode == NV30_FP_OP_OPCODE_KIL)
+      shader->card_priv->NV30FP.uses_kil = GL_TRUE;
+   shader->inst[0] &= ~NV30_FP_OP_OPCODE_MASK;
+   shader->inst[0] |= (opcode << NV30_FP_OP_OPCODE_SHIFT);
+}
+
+static void
+NV30FPSetCCUpdate(nvsFunc *shader)
+{
+   shader->inst[0] |= NV30_FP_OP_COND_WRITE_ENABLE;
+}
+
+static void
+NV30FPSetCondition(nvsFunc *shader, int on, nvsCond cond, int reg,
+      		  nvsSwzComp *swz)
+{
+   nvsSwzComp default_swz[4] = { NVS_SWZ_X, NVS_SWZ_Y, NVS_SWZ_Z, NVS_SWZ_W };
+   unsigned int hwcond;
+
+   /* cond masking is always enabled */
+   if (!on) {
+      cond = NVS_COND_TR;
+      reg  = 0;
+      swz  = default_swz;
+   }
+
+   switch (cond) {
+   case NVS_COND_TR: hwcond = NV30_FP_OP_COND_TR; break;
+   case NVS_COND_FL: hwcond = NV30_FP_OP_COND_FL; break;
+   case NVS_COND_LT: hwcond = NV30_FP_OP_COND_LT; break;
+   case NVS_COND_GT: hwcond = NV30_FP_OP_COND_GT; break;
+   case NVS_COND_LE: hwcond = NV30_FP_OP_COND_LE; break;
+   case NVS_COND_GE: hwcond = NV30_FP_OP_COND_GE; break;
+   case NVS_COND_EQ: hwcond = NV30_FP_OP_COND_EQ; break;
+   case NVS_COND_NE: hwcond = NV30_FP_OP_COND_NE; break;
+   default:
+	WARN_ONCE("unknown fp condmask=%d\n", cond);
+	hwcond = NV30_FP_OP_COND_TR;
+	break;
+   }
+
+   shader->inst[1] &= ~NV30_FP_OP_COND_MASK;
+   shader->inst[1] |= (hwcond << NV30_FP_OP_COND_SHIFT);
+
+   shader->inst[1] &= ~NV30_FP_OP_COND_SWZ_ALL_MASK;
+   shader->inst[1] |= (swz[NVS_SWZ_X] << NV30_FP_OP_COND_SWZ_X_SHIFT);
+   shader->inst[1] |= (swz[NVS_SWZ_Y] << NV30_FP_OP_COND_SWZ_Y_SHIFT);
+   shader->inst[1] |= (swz[NVS_SWZ_Z] << NV30_FP_OP_COND_SWZ_Z_SHIFT);
+   shader->inst[1] |= (swz[NVS_SWZ_W] << NV30_FP_OP_COND_SWZ_W_SHIFT);
+}
+
+static void
+NV30FPSetHighReg(nvsFunc *shader, int id)
+{
+   if (shader->card_priv->NV30FP.num_regs < (id+1)) {
+      if (id == 0)
+	 id = 1; /* necessary? */
+      shader->card_priv->NV30FP.num_regs = (id+1);
+   }
+}
+
+static void
+NV30FPSetResult(nvsFunc *shader, nvsRegister *reg, unsigned int mask, int slot)
+{
+   unsigned int hwreg;
+
+   if (mask & SMASK_X) shader->inst[0] |= NV30_FP_OP_OUT_X;
+   if (mask & SMASK_Y) shader->inst[0] |= NV30_FP_OP_OUT_Y;
+   if (mask & SMASK_Z) shader->inst[0] |= NV30_FP_OP_OUT_Z;
+   if (mask & SMASK_W) shader->inst[0] |= NV30_FP_OP_OUT_W;
+
+   if (reg->file == NVS_FILE_RESULT) {
+      hwreg = 0; /* FIXME: this is only fragment.color */
+      /* This is *not* correct, I have no idea what it is either */
+      shader->inst[0] |= NV30_FP_OP_UNK0_7;
+   } else {
+      shader->inst[0] &= ~NV30_FP_OP_UNK0_7;
+      hwreg = reg->index;
+   }
+   NV30FPSetHighReg(shader, hwreg);
+   shader->inst[0] &= ~NV30_FP_OP_OUT_REG_SHIFT;
+   shader->inst[0] |= (hwreg  << NV30_FP_OP_OUT_REG_SHIFT);
+}
+
+static void
+NV30FPSetSource(nvsFunc *shader, nvsRegister *reg, int pos)
+{
+   unsigned int hwsrc = 0;
+
+   switch (reg->file) {
+   case NVS_FILE_TEMP:
+      hwsrc |= (NV30_FP_REG_TYPE_TEMP << NV30_FP_REG_TYPE_SHIFT);
+      hwsrc |= (reg->index << NV30_FP_REG_SRC_SHIFT);
+      NV30FPSetHighReg(shader, reg->index);
+      break;
+   case NVS_FILE_ATTRIB:
+      {
+	 unsigned int hwin;
+
+	 switch (reg->index) {
+	 case NVS_FR_POSITION : hwin = NV30_FP_OP_INPUT_SRC_POSITION; break;
+	 case NVS_FR_COL0     : hwin = NV30_FP_OP_INPUT_SRC_COL0; break;
+	 case NVS_FR_COL1     : hwin = NV30_FP_OP_INPUT_SRC_COL1; break;
+	 case NVS_FR_FOGCOORD : hwin = NV30_FP_OP_INPUT_SRC_FOGC; break;
+	 case NVS_FR_TEXCOORD0: hwin = NV30_FP_OP_INPUT_SRC_TC(0); break;
+	 case NVS_FR_TEXCOORD1: hwin = NV30_FP_OP_INPUT_SRC_TC(1); break;
+	 case NVS_FR_TEXCOORD2: hwin = NV30_FP_OP_INPUT_SRC_TC(2); break;
+	 case NVS_FR_TEXCOORD3: hwin = NV30_FP_OP_INPUT_SRC_TC(3); break;
+	 case NVS_FR_TEXCOORD4: hwin = NV30_FP_OP_INPUT_SRC_TC(4); break;
+	 case NVS_FR_TEXCOORD5: hwin = NV30_FP_OP_INPUT_SRC_TC(5); break;
+	 case NVS_FR_TEXCOORD6: hwin = NV30_FP_OP_INPUT_SRC_TC(6); break;
+	 case NVS_FR_TEXCOORD7: hwin = NV30_FP_OP_INPUT_SRC_TC(7); break;
+	 default:
+		WARN_ONCE("unknown fp input %d\n", reg->index);
+		hwin = NV30_FP_OP_INPUT_SRC_COL0;
+		break;
+	 }
+	 shader->inst[0] &= ~NV30_FP_OP_INPUT_SRC_MASK;
+	 shader->inst[0] |= (hwin << NV30_FP_OP_INPUT_SRC_SHIFT);
+	 hwsrc |= (hwin << NV30_FP_REG_SRC_SHIFT);
+      }
+      hwsrc |= (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT);
+      break;
+   case NVS_FILE_CONST:
+      /* consts are inlined after the inst */
+      hwsrc |= (NV30_FP_REG_TYPE_CONST << NV30_FP_REG_TYPE_SHIFT);
+      break;
+   default:
+      assert(0);
+      break;
+   }
+
+   if (reg->negate)
+      hwsrc |= NV30_FP_REG_NEGATE;
+   if (reg->abs)
+      shader->inst[1] |= (1 << (29+pos));
+   hwsrc |= (reg->swizzle[NVS_SWZ_X] << NV30_FP_REG_SWZ_X_SHIFT);
+   hwsrc |= (reg->swizzle[NVS_SWZ_Y] << NV30_FP_REG_SWZ_Y_SHIFT);
+   hwsrc |= (reg->swizzle[NVS_SWZ_Z] << NV30_FP_REG_SWZ_Z_SHIFT);
+   hwsrc |= (reg->swizzle[NVS_SWZ_W] << NV30_FP_REG_SWZ_W_SHIFT);
+
+   shader->inst[pos+1] &= ~NV30_FP_REG_ALL_MASK;
+   shader->inst[pos+1] |= hwsrc;
+}
+
+static void
+NV30FPSetTexImageUnit(nvsFunc *shader, int unit)
+{
+   shader->inst[0] &= ~NV30_FP_OP_TEX_UNIT_SHIFT;
+   shader->inst[0] |= (unit << NV30_FP_OP_TEX_UNIT_SHIFT);
+}
+
+static void
+NV30FPSetSaturate(nvsFunc *shader)
+{
+   shader->inst[0] |= NV30_FP_OP_OUT_SAT;
+}
+
+static void
+NV30FPInitInstruction(nvsFunc *shader)
+{
+   unsigned int hwsrc;
+
+   shader->inst[0] = 0;
+
+   hwsrc = (NV30_FP_REG_TYPE_INPUT << NV30_FP_REG_TYPE_SHIFT) |
+      	   (NVS_SWZ_X << NV30_FP_REG_SWZ_X_SHIFT) |
+	   (NVS_SWZ_Y << NV30_FP_REG_SWZ_Y_SHIFT) |
+	   (NVS_SWZ_Z << NV30_FP_REG_SWZ_Z_SHIFT) |
+	   (NVS_SWZ_W << NV30_FP_REG_SWZ_W_SHIFT);
+   shader->inst[1] = hwsrc;
+   shader->inst[2] = hwsrc;
+   shader->inst[3] = hwsrc;
+}
+
+static void
+NV30FPSetLastInst(nvsFunc *shader)
+{
+   shader->inst[0] |= 1; 
+}
+
+/*******************************************************************************
+ * Disassembly helpers
+ */
+static struct _op_xlat *
+NV30FPGetOPTXRec(nvsFunc * shader, int merged)
+{
+   int op;
+
+   op = shader->GetOpcodeHW(shader, 0);
+   if (op > NVFP_TX_AOP_COUNT)
+      return NULL;
+   if (NVFP_TX_AOP[op].SOP == NVS_OP_UNKNOWN)
+      return NULL;
+   return &NVFP_TX_AOP[op];
+}
+
+static int
+NV30FPHasMergedInst(nvsFunc * shader)
+{
+   return 0;
+}
+
+static int
+NV30FPIsLastInst(nvsFunc * shader)
+{
+   return ((shader->inst[0] & NV30_FP_OP_PROGRAM_END) ? 1 : 0);
+}
+
+static int
+NV30FPGetOffsetNext(nvsFunc * shader)
+{
+   int i;
+
+   for (i = 0; i < 3; i++)
+      if (shader->GetSourceFile(shader, 0, i) == NVS_FILE_CONST)
+	 return 8;
+   return 4;
+}
+
+static nvsOpcode
+NV30FPGetOpcode(nvsFunc * shader, int merged)
+{
+   struct _op_xlat *opr;
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr)
+      return NVS_OP_UNKNOWN;
+
+   return opr->SOP;
+}
+
+static unsigned int
+NV30FPGetOpcodeHW(nvsFunc * shader, int slot)
+{
+   int op;
+
+   op = (shader->inst[0] & NV30_FP_OP_OPCODE_MASK) >> NV30_FP_OP_OPCODE_SHIFT;
+
+   return op;
+}
+
+static nvsRegFile
+NV30FPGetDestFile(nvsFunc * shader, int merged)
+{
+   /* Result regs overlap temporary regs */
+   return NVS_FILE_TEMP;
+}
+
+static unsigned int
+NV30FPGetDestID(nvsFunc * shader, int merged)
+{
+   int id;
+
+   switch (shader->GetDestFile(shader, merged)) {
+   case NVS_FILE_TEMP:
+      id = ((shader->inst[0] & NV30_FP_OP_OUT_REG_MASK)
+	    >> NV30_FP_OP_OUT_REG_SHIFT);
+      return id;
+   default:
+      return -1;
+   }
+}
+
+static unsigned int
+NV30FPGetDestMask(nvsFunc * shader, int merged)
+{
+   unsigned int mask = 0;
+
+   if (shader->inst[0] & NV30_FP_OP_OUT_X) mask |= SMASK_X;
+   if (shader->inst[0] & NV30_FP_OP_OUT_Y) mask |= SMASK_Y;
+   if (shader->inst[0] & NV30_FP_OP_OUT_Z) mask |= SMASK_Z;
+   if (shader->inst[0] & NV30_FP_OP_OUT_W) mask |= SMASK_W;
+
+   return mask;
+}
+
+static unsigned int
+NV30FPGetSourceHW(nvsFunc * shader, int merged, int pos)
+{
+   struct _op_xlat *opr;
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr || opr->srcpos[pos] == -1)
+      return -1;
+
+   return shader->inst[opr->srcpos[pos] + 1];
+}
+
+static nvsRegFile
+NV30FPGetSourceFile(nvsFunc * shader, int merged, int pos)
+{
+   unsigned int src;
+   struct _op_xlat *opr;
+   int file;
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr || opr->srcpos[pos] == -1)
+      return NVS_FILE_UNKNOWN;
+
+   switch (opr->srcpos[pos]) {
+   case SPOS_ADDRESS: return NVS_FILE_ADDRESS;
+   default:
+      src = shader->GetSourceHW(shader, merged, pos);
+      file = (src & NV30_FP_REG_TYPE_MASK) >> NV30_FP_REG_TYPE_SHIFT;
+
+      switch (file) {
+      case NV30_FP_REG_TYPE_TEMP : return NVS_FILE_TEMP;
+      case NV30_FP_REG_TYPE_INPUT: return NVS_FILE_ATTRIB;
+      case NV30_FP_REG_TYPE_CONST: return NVS_FILE_CONST;
+      default:
+	 return NVS_FILE_UNKNOWN;
+      }
+   }
+}
+
+static int
+NV30FPGetSourceID(nvsFunc * shader, int merged, int pos)
+{
+   switch (shader->GetSourceFile(shader, merged, pos)) {
+   case NVS_FILE_ATTRIB:
+      switch ((shader->inst[0] & NV30_FP_OP_INPUT_SRC_MASK)
+	      >> NV30_FP_OP_INPUT_SRC_SHIFT) {
+      case NV30_FP_OP_INPUT_SRC_POSITION: return NVS_FR_POSITION;
+      case NV30_FP_OP_INPUT_SRC_COL0    : return NVS_FR_COL0;
+      case NV30_FP_OP_INPUT_SRC_COL1    : return NVS_FR_COL1;
+      case NV30_FP_OP_INPUT_SRC_FOGC    : return NVS_FR_FOGCOORD;
+      case NV30_FP_OP_INPUT_SRC_TC(0)   : return NVS_FR_TEXCOORD0;
+      case NV30_FP_OP_INPUT_SRC_TC(1)   : return NVS_FR_TEXCOORD1;
+      case NV30_FP_OP_INPUT_SRC_TC(2)   : return NVS_FR_TEXCOORD2;
+      case NV30_FP_OP_INPUT_SRC_TC(3)   : return NVS_FR_TEXCOORD3;
+      case NV30_FP_OP_INPUT_SRC_TC(4)   : return NVS_FR_TEXCOORD4;
+      case NV30_FP_OP_INPUT_SRC_TC(5)   : return NVS_FR_TEXCOORD5;
+      case NV30_FP_OP_INPUT_SRC_TC(6)   : return NVS_FR_TEXCOORD6;
+      case NV30_FP_OP_INPUT_SRC_TC(7)   : return NVS_FR_TEXCOORD7;
+      default:
+	 return -1;
+      }
+      break;
+   case NVS_FILE_TEMP:
+      {
+	 unsigned int src;
+
+	 src = shader->GetSourceHW(shader, merged, pos);
+	 return ((src & NV30_FP_REG_SRC_MASK) >> NV30_FP_REG_SRC_SHIFT);
+      }
+   case NVS_FILE_CONST:		/* inlined into fragprog */
+   default:
+      return -1;
+   }
+}
+
+static int
+NV30FPGetTexImageUnit(nvsFunc *shader)
+{
+      return ((shader->inst[0] & NV30_FP_OP_TEX_UNIT_MASK)
+	      >> NV30_FP_OP_TEX_UNIT_SHIFT);
+}
+
+static int
+NV30FPGetSourceNegate(nvsFunc * shader, int merged, int pos)
+{
+   unsigned int src;
+
+   src = shader->GetSourceHW(shader, merged, pos);
+
+   if (src == -1)
+      return -1;
+   return ((src & NV30_FP_REG_NEGATE) ? 1 : 0);
+}
+
+static int
+NV30FPGetSourceAbs(nvsFunc * shader, int merged, int pos)
+{
+   struct _op_xlat *opr;
+   static unsigned int abspos[3] = {
+      NV30_FP_OP_OUT_ABS,
+      (1 << 30),		/* guess */
+      (1 << 31)			/* guess */
+   };
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr || opr->srcpos[pos] == -1)
+      return -1;
+
+   return ((shader->inst[1] & abspos[opr->srcpos[pos]]) ? 1 : 0);
+}
+
+nvsSwzComp NV30FP_TX_SWIZZLE[4] = {NVS_SWZ_X, NVS_SWZ_Y, NVS_SWZ_Z, NVS_SWZ_W };
+
+static void
+NV30FPTXSwizzle(int hwswz, nvsSwzComp *swz)
+{
+   swz[NVS_SWZ_W] = NV30FP_TX_SWIZZLE[(hwswz & 0xC0) >> 6];
+   swz[NVS_SWZ_Z] = NV30FP_TX_SWIZZLE[(hwswz & 0x30) >> 4];
+   swz[NVS_SWZ_Y] = NV30FP_TX_SWIZZLE[(hwswz & 0x0C) >> 2];
+   swz[NVS_SWZ_X] = NV30FP_TX_SWIZZLE[(hwswz & 0x03) >> 0];
+}
+
+static void
+NV30FPGetSourceSwizzle(nvsFunc * shader, int merged, int pos, nvsSwzComp *swz)
+{
+   unsigned int src;
+   int swzbits;
+
+   src = shader->GetSourceHW(shader, merged, pos);
+   swzbits = (src & NV30_FP_REG_SWZ_ALL_MASK) >> NV30_FP_REG_SWZ_ALL_SHIFT;
+   NV30FPTXSwizzle(swzbits, swz);
+}
+
+static int
+NV30FPGetSourceIndexed(nvsFunc * shader, int merged, int pos)
+{
+   switch (shader->GetSourceFile(shader, merged, pos)) {
+   case NVS_FILE_ATTRIB:
+      return ((shader->inst[3] & NV30_FP_OP_INDEX_INPUT) ? 1 : 0);
+   default:
+      return 0;
+   }
+}
+
+static void
+NV30FPGetSourceConstVal(nvsFunc * shader, int merged, int pos, float *val)
+{
+   val[0] = *(float *) &(shader->inst[4]);
+   val[1] = *(float *) &(shader->inst[5]);
+   val[2] = *(float *) &(shader->inst[6]);
+   val[3] = *(float *) &(shader->inst[7]);
+}
+
+static int
+NV30FPGetSourceScale(nvsFunc * shader, int merged, int pos)
+{
+/*FIXME: is this per-source, only for a specific source, or all sources??*/
+   return (1 << ((shader->inst[2] & NV30_FP_OP_SRC_SCALE_MASK)
+		 >> NV30_FP_OP_SRC_SCALE_SHIFT));
+}
+
+static int
+NV30FPGetAddressRegID(nvsFunc * shader)
+{
+   return 0;
+}
+
+static nvsSwzComp
+NV30FPGetAddressRegSwizzle(nvsFunc * shader)
+{
+   return NVS_SWZ_X;
+}
+
+static int
+NV30FPSupportsConditional(nvsFunc * shader)
+{
+   /*FIXME: Is this true of all ops? */
+   return 1;
+}
+
+static int
+NV30FPGetConditionUpdate(nvsFunc * shader)
+{
+   return ((shader->inst[0] & NV30_FP_OP_COND_WRITE_ENABLE) ? 1 : 0);
+}
+
+static int
+NV30FPGetConditionTest(nvsFunc * shader)
+{
+   /*FIXME: always? */
+   return 1;
+}
+
+static nvsCond
+NV30FPGetCondition(nvsFunc * shader)
+{
+   int cond;
+
+   cond = ((shader->inst[1] & NV30_FP_OP_COND_MASK)
+	   >> NV30_FP_OP_COND_SHIFT);
+
+   switch (cond) {
+   case NV30_FP_OP_COND_FL: return NVS_COND_FL;
+   case NV30_FP_OP_COND_LT: return NVS_COND_LT;
+   case NV30_FP_OP_COND_EQ: return NVS_COND_EQ;
+   case NV30_FP_OP_COND_LE: return NVS_COND_LE;
+   case NV30_FP_OP_COND_GT: return NVS_COND_GT;
+   case NV30_FP_OP_COND_NE: return NVS_COND_NE;
+   case NV30_FP_OP_COND_GE: return NVS_COND_GE;
+   case NV30_FP_OP_COND_TR: return NVS_COND_TR;
+   default:
+      return NVS_COND_UNKNOWN;
+   }
+}
+
+static void
+NV30FPGetCondRegSwizzle(nvsFunc * shader, nvsSwzComp *swz)
+{
+   int swzbits;
+
+   swzbits = (shader->inst[1] & NV30_FP_OP_COND_SWZ_ALL_MASK)
+      >> NV30_FP_OP_COND_SWZ_ALL_SHIFT;
+   NV30FPTXSwizzle(swzbits, swz);
+}
+
+static int
+NV30FPGetCondRegID(nvsFunc * shader)
+{
+   return 0;
+}
+
+static nvsPrecision
+NV30FPGetPrecision(nvsFunc * shader)
+{
+   int p;
+
+   p = (shader->inst[0] & NV30_FP_OP_PRECISION_MASK)
+      >> NV30_FP_OP_PRECISION_SHIFT;
+
+   switch (p) {
+   case NV30_FP_PRECISION_FP32: return NVS_PREC_FLOAT32;
+   case NV30_FP_PRECISION_FP16: return NVS_PREC_FLOAT16;
+   case NV30_FP_PRECISION_FX12: return NVS_PREC_FIXED12;
+   default:
+      return NVS_PREC_UNKNOWN;
+   }
+}
+
+static int
+NV30FPGetSaturate(nvsFunc * shader)
+{
+   return ((shader->inst[0] & NV30_FP_OP_OUT_SAT) ? 1 : 0);
+}
+
+/*******************************************************************************
+ * Init
+ */
+void
+NV30FPInitShaderFuncs(nvsFunc * shader)
+{
+   /* These are probably bogus, I made them up... */
+   shader->MaxInst	= 1024;
+   shader->MaxAttrib	= 16;
+   shader->MaxTemp	= 32;
+   shader->MaxAddress	= 1;
+   shader->MaxConst	= 256;
+   shader->caps		= SCAP_SRC_ABS;
+
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_MOV, NVS_OP_MOV, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_MUL, NVS_OP_MUL, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_ADD, NVS_OP_ADD, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_MAD, NVS_OP_MAD, 0, 1, 2);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_DP3, NVS_OP_DP3, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_DP4, NVS_OP_DP4, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_DST, NVS_OP_DST, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_MIN, NVS_OP_MIN, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_MAX, NVS_OP_MAX, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_SLT, NVS_OP_SLT, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_SGE, NVS_OP_SGE, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_FRC, NVS_OP_FRC, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_FLR, NVS_OP_FLR, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_TEX, NVS_OP_TEX, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_TXD, NVS_OP_TXD, 0, 1, 2);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_TXP, NVS_OP_TXP, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_TXB, NVS_OP_TXB, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_SEQ, NVS_OP_SEQ, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_SGT, NVS_OP_SGT, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_SLE, NVS_OP_SLE, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_SNE, NVS_OP_SNE, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_RCP, NVS_OP_RCP, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_LG2, NVS_OP_LG2, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_EX2, NVS_OP_EX2, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_COS, NVS_OP_COS, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_SIN, NVS_OP_SIN, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_NOP, NVS_OP_NOP, -1, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_DDX, NVS_OP_DDX, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_DDY, NVS_OP_DDY, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_KIL, NVS_OP_KIL, -1, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_PK4B, NVS_OP_PK4B, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_UP4B, NVS_OP_UP4B, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_PK2H, NVS_OP_PK2H, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_UP2H, NVS_OP_UP2H, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_PK4UB, NVS_OP_PK4UB, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_UP4UB, NVS_OP_UP4UB, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_PK2US, NVS_OP_PK2US, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_UP2US, NVS_OP_UP2US, 0, -1, -1);
+   /*FIXME: Haven't confirmed the source positions for the below opcodes */
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_LIT, NVS_OP_LIT, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_LRP, NVS_OP_LRP, 0, 1, 2);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_POW, NVS_OP_POW, 0, 1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_RSQ, NVS_OP_RSQ, 0, -1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV30_FP_OP_OPCODE_RFL, NVS_OP_RFL, 0, 1, -1);
+
+   shader->GetOPTXRec		= NV30FPGetOPTXRec;
+   shader->GetOPTXFromSOP	= NV30FPGetOPTXFromSOP;
+
+   shader->UploadToHW		= NV30FPUploadToHW;
+   shader->UpdateConst		= NV30FPUpdateConst;
+
+   shader->InitInstruction	= NV30FPInitInstruction;
+   shader->SupportsOpcode	= NV30FPSupportsOpcode;
+   shader->SetOpcode		= NV30FPSetOpcode;
+   shader->SetCCUpdate		= NV30FPSetCCUpdate;
+   shader->SetCondition		= NV30FPSetCondition;
+   shader->SetResult		= NV30FPSetResult;
+   shader->SetSource		= NV30FPSetSource;
+   shader->SetTexImageUnit	= NV30FPSetTexImageUnit;
+   shader->SetSaturate		= NV30FPSetSaturate;
+   shader->SetLastInst		= NV30FPSetLastInst;
+
+   shader->HasMergedInst	= NV30FPHasMergedInst;
+   shader->IsLastInst		= NV30FPIsLastInst;
+   shader->GetOffsetNext	= NV30FPGetOffsetNext;
+   shader->GetOpcode		= NV30FPGetOpcode;
+   shader->GetOpcodeHW		= NV30FPGetOpcodeHW;
+   shader->GetDestFile		= NV30FPGetDestFile;
+   shader->GetDestID		= NV30FPGetDestID;
+   shader->GetDestMask		= NV30FPGetDestMask;
+   shader->GetSourceHW		= NV30FPGetSourceHW;
+   shader->GetSourceFile	= NV30FPGetSourceFile;
+   shader->GetSourceID		= NV30FPGetSourceID;
+   shader->GetTexImageUnit	= NV30FPGetTexImageUnit;
+   shader->GetSourceNegate	= NV30FPGetSourceNegate;
+   shader->GetSourceAbs		= NV30FPGetSourceAbs;
+   shader->GetSourceSwizzle	= NV30FPGetSourceSwizzle;
+   shader->GetSourceIndexed	= NV30FPGetSourceIndexed;
+   shader->GetSourceConstVal	= NV30FPGetSourceConstVal;
+   shader->GetSourceScale	= NV30FPGetSourceScale;
+   shader->GetRelAddressRegID	= NV30FPGetAddressRegID;
+   shader->GetRelAddressSwizzle = NV30FPGetAddressRegSwizzle;
+   shader->GetPrecision		= NV30FPGetPrecision;
+   shader->GetSaturate		= NV30FPGetSaturate;
+   shader->SupportsConditional	= NV30FPSupportsConditional;
+   shader->GetConditionUpdate	= NV30FPGetConditionUpdate;
+   shader->GetConditionTest	= NV30FPGetConditionTest;
+   shader->GetCondition		= NV30FPGetCondition;
+   shader->GetCondRegSwizzle	= NV30FPGetCondRegSwizzle;
+   shader->GetCondRegID		= NV30FPGetCondRegID;
+}
diff --git a/src/mesa/drivers/dri/nouveau/nv30_shader.h b/src/mesa/drivers/dri/nouveau/nv30_shader.h
new file mode 100644
index 00000000000..7a027dd4273
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv30_shader.h
@@ -0,0 +1,379 @@
+#ifndef __NV30_SHADER_H__
+#define __NV30_SHADER_H__
+
+/* Vertex programs instruction set
+ *
+ * 128bit opcodes, split into 4 32-bit ones for ease of use.
+ *
+ * Non-native instructions
+ *	 ABS - MOV + NV40_VP_INST0_DEST_ABS
+ *	 POW - EX2 + MUL + LG2
+ *	 SUB - ADD, second source negated
+ *	 SWZ - MOV
+ *	 XPD -  
+ *
+ * Register access
+ *	 - Only one INPUT can be accessed per-instruction (move extras into TEMPs)
+ *	 - Only one CONST can be accessed per-instruction (move extras into TEMPs)
+ *
+ * Relative Addressing
+ *	 According to the value returned for MAX_PROGRAM_NATIVE_ADDRESS_REGISTERS_ARB
+ *	 there are only two address registers available.  The destination in the ARL
+ *	 instruction is set to TEMP <n> (The temp isn't actually written).
+ *
+ *	 When using vanilla ARB_v_p, the proprietary driver will squish both the available
+ *	 ADDRESS regs into the first hardware reg in the X and Y components.
+ *
+ *	 To use an address reg as an index into consts, the CONST_SRC is set to
+ *	 (const_base + offset) and INDEX_CONST is set.
+ *
+ *	 To access the second address reg use ADDR_REG_SELECT_1. A particular component
+ *	 of the address regs is selected with ADDR_SWZ.
+ *
+ *	 Only one address register can be accessed per instruction.
+ *
+ * Conditional execution (see NV_vertex_program{2,3} for details)
+ *	 Conditional execution of an instruction is enabled by setting COND_TEST_ENABLE, and
+ *	 selecting the condition which will allow the test to pass with COND_{FL,LT,...}.
+ *	 It is possible to swizzle the values in the condition register, which allows for
+ *	 testing against an individual component.
+ *
+ * Branching
+ *	 The BRA/CAL instructions seem to follow a slightly different opcode layout.  The
+ *	 destination instruction ID (IADDR) overlaps a source field.  Instruction ID's seem to
+ *	 be numbered based on the UPLOAD_FROM_ID FIFO command, and is incremented automatically
+ *	 on each UPLOAD_INST FIFO command.
+ *
+ *	 Conditional branching is achieved by using the condition tests described above.
+ *	 There doesn't appear to be dedicated looping instructions, but this can be done
+ *	 using a temp reg + conditional branching.
+ *
+ *	 Subroutines may be uploaded before the main program itself, but the first executed
+ *	 instruction is determined by the PROGRAM_START_ID FIFO command.
+ *
+ */
+
+/* DWORD 0 */
+#define NV30_VP_INST_ADDR_REG_SELECT_1				(1 << 24)
+#define NV30_VP_INST_SRC2_ABS			 		(1 << 23) /* guess */
+#define NV30_VP_INST_SRC1_ABS			 		(1 << 22) /* guess */
+#define NV30_VP_INST_SRC0_ABS			 		(1 << 21) /* guess */
+#define NV30_VP_INST_OUT_RESULT					(1 << 20)
+#define NV30_VP_INST_DEST_TEMP_ID_SHIFT				16
+#define NV30_VP_INST_DEST_TEMP_ID_MASK				(0x0F << 16)
+#define NV30_VP_INST_COND_UPDATE_ENABLE				(1<<15)
+#define NV30_VP_INST_COND_TEST_ENABLE				(1<<14)
+#define NV30_VP_INST_COND_SHIFT					11
+#define NV30_VP_INST_COND_MASK					(0x07 << 11)
+#	define NV30_VP_INST_COND_FL	0 /* guess */	
+#	define NV30_VP_INST_COND_LT	1	
+#	define NV30_VP_INST_COND_EQ	2
+#	define NV30_VP_INST_COND_LE	3
+#	define NV30_VP_INST_COND_GT	4
+#	define NV30_VP_INST_COND_NE	5
+#	define NV30_VP_INST_COND_GE	6
+#	define NV30_VP_INST_COND_TR	7 /* guess */
+#define NV30_VP_INST_COND_SWZ_X_SHIFT				9
+#define NV30_VP_INST_COND_SWZ_X_MASK				(0x03 <<  9)
+#define NV30_VP_INST_COND_SWZ_Y_SHIFT				7
+#define NV30_VP_INST_COND_SWZ_Y_MASK				(0x03 <<  7)
+#define NV30_VP_INST_COND_SWZ_Z_SHIFT				5
+#define NV30_VP_INST_COND_SWZ_Z_MASK				(0x03 <<  5)
+#define NV30_VP_INST_COND_SWZ_W_SHIFT				3
+#define NV30_VP_INST_COND_SWZ_W_MASK				(0x03 <<  3)
+#define NV30_VP_INST_COND_SWZ_ALL_SHIFT				3
+#define NV30_VP_INST_COND_SWZ_ALL_MASK				(0xFF <<  3)
+#define NV30_VP_INST_ADDR_SWZ_SHIFT				1
+#define NV30_VP_INST_ADDR_SWZ_MASK				(0x03 <<  1)
+#define NV30_VP_INST_SCA_OPCODEH_SHIFT				0
+#define NV30_VP_INST_SCA_OPCODEH_MASK				(0x01 <<  0)
+
+/* DWORD 1 */
+#define NV30_VP_INST_SCA_OPCODEL_SHIFT				28
+#define NV30_VP_INST_SCA_OPCODEL_MASK				(0x0F << 28)
+#	define NV30_VP_INST_OP_NOP	0x00
+#	define NV30_VP_INST_OP_RCP	0x02
+#	define NV30_VP_INST_OP_RCC	0x03
+#	define NV30_VP_INST_OP_RSQ	0x04
+#	define NV30_VP_INST_OP_EXP	0x05
+#	define NV30_VP_INST_OP_LOG	0x06
+#	define NV30_VP_INST_OP_LIT	0x07
+#	define NV30_VP_INST_OP_BRA	0x09
+#	define NV30_VP_INST_OP_CAL	0x0B
+#	define NV30_VP_INST_OP_RET	0x0C
+#	define NV30_VP_INST_OP_LG2	0x0D
+#	define NV30_VP_INST_OP_EX2	0x0E
+#	define NV30_VP_INST_OP_SIN	0x0F
+#	define NV30_VP_INST_OP_COS	0x10
+#define NV30_VP_INST_VEC_OPCODE_SHIFT				23
+#define NV30_VP_INST_VEC_OPCODE_MASK				(0x1F << 23)
+#	define NV30_VP_INST_OP_NOPV	0x00
+#	define NV30_VP_INST_OP_MOV	0x01
+#	define NV30_VP_INST_OP_MUL	0x02
+#	define NV30_VP_INST_OP_ADD	0x03
+#	define NV30_VP_INST_OP_MAD	0x04
+#	define NV30_VP_INST_OP_DP3	0x05
+#	define NV30_VP_INST_OP_DP4	0x07
+#	define NV30_VP_INST_OP_DPH	0x06
+#	define NV30_VP_INST_OP_DST	0x08
+#	define NV30_VP_INST_OP_MIN	0x09
+#	define NV30_VP_INST_OP_MAX	0x0A
+#	define NV30_VP_INST_OP_SLT	0x0B
+#	define NV30_VP_INST_OP_SGE	0x0C
+#	define NV30_VP_INST_OP_ARL	0x0D
+#	define NV30_VP_INST_OP_FRC	0x0E
+#	define NV30_VP_INST_OP_FLR	0x0F
+#	define NV30_VP_INST_OP_SEQ	0x10
+#	define NV30_VP_INST_OP_SFL	0x11
+#	define NV30_VP_INST_OP_SGT	0x12
+#	define NV30_VP_INST_OP_SLE	0x13
+#	define NV30_VP_INST_OP_SNE	0x14
+#	define NV30_VP_INST_OP_STR	0x15
+#	define NV30_VP_INST_OP_SSG	0x16
+#	define NV30_VP_INST_OP_ARR	0x17
+#	define NV30_VP_INST_OP_ARA	0x18
+#define NV30_VP_INST_CONST_SRC_SHIFT				14
+#define NV30_VP_INST_CONST_SRC_MASK				(0xFF << 14)
+#define NV30_VP_INST_INPUT_SRC_SHIFT				9		/*NV20*/
+#define NV30_VP_INST_INPUT_SRC_MASK				(0x0F <<  9)	/*NV20*/
+#	define NV30_VP_INST_IN_POS	0	  /* These seem to match the bindings specified in */
+#	define NV30_VP_INST_IN_WEIGHT	1	  /* the ARB_v_p spec (2.14.3.1) */
+#	define NV30_VP_INST_IN_NORMAL	2	  
+#	define NV30_VP_INST_IN_COL0	3	  /* Should probably confirm them all though */
+#	define NV30_VP_INST_IN_COL1	4
+#	define NV30_VP_INST_IN_FOGC	5
+#	define NV30_VP_INST_IN_TC0	8
+#	define NV30_VP_INST_IN_TC(n)	(8+n)
+#define NV30_VP_INST_SRC0H_SHIFT				0		/*NV20*/
+#define NV30_VP_INST_SRC0H_MASK					(0x1FF << 0)	/*NV20*/
+
+/* DWORD 2 */
+#define NV30_VP_INST_SRC0L_SHIFT				26		/*NV20*/
+#define NV30_VP_INST_SRC0L_MASK					(0x3F  <<26)	/*NV20*/
+#define NV30_VP_INST_SRC1_SHIFT					11		/*NV20*/
+#define NV30_VP_INST_SRC1_MASK					(0x7FFF<<11)	/*NV20*/
+#define NV30_VP_INST_SRC2H_SHIFT				0		/*NV20*/
+#define NV30_VP_INST_SRC2H_MASK					(0x7FF << 0)	/*NV20*/
+#define NV30_VP_INST_IADDR_SHIFT				2
+#define NV30_VP_INST_IADDR_MASK					(0xFF <<  2) 	/* guess */
+
+/* DWORD 3 */
+#define NV30_VP_INST_SRC2L_SHIFT				28		/*NV20*/
+#define NV30_VP_INST_SRC2L_MASK					(0x0F  <<28)	/*NV20*/
+#define NV30_VP_INST_STEMP_WRITEMASK_SHIFT			24
+#define NV30_VP_INST_STEMP_WRITEMASK_MASK			(0x0F << 24)
+#define NV30_VP_INST_VTEMP_WRITEMASK_SHIFT			20
+#define NV30_VP_INST_VTEMP_WRITEMASK_MASK			(0x0F << 20)
+#define NV30_VP_INST_SDEST_WRITEMASK_SHIFT			16
+#define NV30_VP_INST_SDEST_WRITEMASK_MASK			(0x0F << 16)
+#define NV30_VP_INST_VDEST_WRITEMASK_SHIFT			12		/*NV20*/
+#define NV30_VP_INST_VDEST_WRITEMASK_MASK			(0x0F << 12)	/*NV20*/
+#define NV30_VP_INST_DEST_ID_SHIFT				2
+#define NV30_VP_INST_DEST_ID_MASK				(0x0F <<  2)
+#	define NV30_VP_INST_DEST_POS	0
+#	define NV30_VP_INST_DEST_COL0	3
+#	define NV30_VP_INST_DEST_COL1	4
+#	define NV30_VP_INST_DEST_TC(n)	(8+n)
+
+/* Source-register definition - matches NV20 exactly */
+#define NV30_VP_SRC_REG_NEGATE					(1<<14)
+#define NV30_VP_SRC_REG_SWZ_X_SHIFT				12
+#define NV30_VP_SRC_REG_SWZ_X_MASK				(0x03  <<12)
+#define NV30_VP_SRC_REG_SWZ_Y_SHIFT				10
+#define NV30_VP_SRC_REG_SWZ_Y_MASK				(0x03  <<10)
+#define NV30_VP_SRC_REG_SWZ_Z_SHIFT				8
+#define NV30_VP_SRC_REG_SWZ_Z_MASK				(0x03  << 8)
+#define NV30_VP_SRC_REG_SWZ_W_SHIFT				6
+#define NV30_VP_SRC_REG_SWZ_W_MASK				(0x03  << 6)
+#define NV30_VP_SRC_REG_SWZ_ALL_SHIFT				6
+#define NV30_VP_SRC_REG_SWZ_ALL_MASK				(0xFF  << 6)
+#define NV30_VP_SRC_REG_TEMP_ID_SHIFT				2
+#define NV30_VP_SRC_REG_TEMP_ID_MASK				(0x0F  << 0)
+#define NV30_VP_SRC_REG_TYPE_SHIFT				0
+#define NV30_VP_SRC_REG_TYPE_MASK				(0x03  << 0)
+#define NV30_VP_SRC_REG_TYPE_TEMP	1
+#define NV30_VP_SRC_REG_TYPE_INPUT	2
+#define NV30_VP_SRC_REG_TYPE_CONST	3 /* guess */
+
+/*
+ * Each fragment program opcode appears to be comprised of 4 32-bit values.
+ *
+ *	 0 - Opcode, output reg/mask, ATTRIB source
+ *	 1 - Source 0
+ *	 2 - Source 1
+ *	 3 - Source 2
+ *
+ * There appears to be no special difference between result regs and temp regs.
+ * 		result.color == R0.xyzw
+ * 		result.depth == R1.z
+ * When the fragprog contains instructions to write depth, NV30_TCL_PRIMITIVE_3D_UNK1D78=0
+ * otherwise it is set to 1.
+ *
+ * Constants are inserted directly after the instruction that uses them.
+ * 
+ * It appears that it's not possible to use two input registers in one
+ * instruction as the input sourcing is done in the instruction dword
+ * and not the source selection dwords.  As such instructions such as:
+ * 
+ *		 ADD result.color, fragment.color, fragment.texcoord[0];
+ *
+ * must be split into two MOV's and then an ADD (nvidia does this) but
+ * I'm not sure why it's not just one MOV and then source the second input
+ * in the ADD instruction..
+ *
+ * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary
+ * negation requires multiplication with a const.
+ *
+ * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO/SWIZZLE_ONE
+ * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as SWIZZLE_ZERO
+ * is implemented simply by not writing to the relevant components of the destination.
+ *
+ * Conditional execution
+ *   TODO
+ * 
+ * Non-native instructions:
+ *	 LIT
+ *	 LRP - MAD+MAD
+ *	 SUB - ADD, negate second source
+ *	 RSQ - LG2 + EX2
+ *	 POW - LG2 + MUL + EX2
+ *	 SCS - COS + SIN
+ *	 XPD
+ */
+
+//== Opcode / Destination selection ==
+#define NV30_FP_OP_PROGRAM_END					(1 << 0)
+#define NV30_FP_OP_OUT_REG_SHIFT				1
+#define NV30_FP_OP_OUT_REG_MASK					(31 << 1)	/* uncertain */
+/* Needs to be set when writing outputs to get expected result.. */
+#define NV30_FP_OP_UNK0_7					(1 << 7)
+#define NV30_FP_OP_COND_WRITE_ENABLE				(1 << 8)
+#define NV30_FP_OP_OUTMASK_SHIFT				9
+#define NV30_FP_OP_OUTMASK_MASK					(0xF << 9)
+#	define NV30_FP_OP_OUT_X	(1<<9)
+#	define NV30_FP_OP_OUT_Y	(1<<10)
+#	define NV30_FP_OP_OUT_Z	(1<<11)
+#	define NV30_FP_OP_OUT_W	(1<<12)
+/* Uncertain about these, especially the input_src values.. it's possible that
+ * they can be dynamically changed.
+ */
+#define NV30_FP_OP_INPUT_SRC_SHIFT				13
+#define NV30_FP_OP_INPUT_SRC_MASK				(15 << 13)
+#	define NV30_FP_OP_INPUT_SRC_POSITION	0x0
+#	define NV30_FP_OP_INPUT_SRC_COL0	0x1
+#	define NV30_FP_OP_INPUT_SRC_COL1	0x2
+#	define NV30_FP_OP_INPUT_SRC_FOGC	0x3
+#	define NV30_FP_OP_INPUT_SRC_TC0		0x4
+#	define NV30_FP_OP_INPUT_SRC_TC(n)	(0x4 + n)
+#define NV30_FP_OP_TEX_UNIT_SHIFT				17
+#define NV30_FP_OP_TEX_UNIT_MASK				(0xF << 17) /* guess */
+#define NV30_FP_OP_PRECISION_SHIFT				22
+#define NV30_FP_OP_PRECISION_MASK				(3 << 22)
+#   define NV30_FP_PRECISION_FP32	0
+#   define NV30_FP_PRECISION_FP16	1
+#   define NV30_FP_PRECISION_FX12	2
+#define NV30_FP_OP_OPCODE_SHIFT					24
+#define NV30_FP_OP_OPCODE_MASK					(0x3F << 24)
+#	define NV30_FP_OP_OPCODE_NOP	0x00
+#	define NV30_FP_OP_OPCODE_MOV	0x01
+#	define NV30_FP_OP_OPCODE_MUL	0x02
+#	define NV30_FP_OP_OPCODE_ADD	0x03
+#	define NV30_FP_OP_OPCODE_MAD	0x04
+#	define NV30_FP_OP_OPCODE_DP3	0x05
+#	define NV30_FP_OP_OPCODE_DP4	0x06
+#	define NV30_FP_OP_OPCODE_DST	0x07
+#	define NV30_FP_OP_OPCODE_MIN	0x08
+#	define NV30_FP_OP_OPCODE_MAX	0x09
+#	define NV30_FP_OP_OPCODE_SLT	0x0A
+#	define NV30_FP_OP_OPCODE_SGE	0x0B
+#	define NV30_FP_OP_OPCODE_SLE	0x0C
+#	define NV30_FP_OP_OPCODE_SGT	0x0D
+#	define NV30_FP_OP_OPCODE_SNE	0x0E
+#	define NV30_FP_OP_OPCODE_SEQ	0x0F
+#	define NV30_FP_OP_OPCODE_FRC	0x10
+#	define NV30_FP_OP_OPCODE_FLR	0x11
+#	define NV30_FP_OP_OPCODE_KIL	0x12
+#	define NV30_FP_OP_OPCODE_PK4B   0x13
+#	define NV30_FP_OP_OPCODE_UP4B   0x14
+#	define NV30_FP_OP_OPCODE_DDX	0x15 /* can only write XY */
+#	define NV30_FP_OP_OPCODE_DDY	0x16 /* can only write XY */
+#	define NV30_FP_OP_OPCODE_TEX	0x17
+#	define NV30_FP_OP_OPCODE_TXP	0x18
+#	define NV30_FP_OP_OPCODE_TXD	0x19
+#	define NV30_FP_OP_OPCODE_RCP	0x1A
+#	define NV30_FP_OP_OPCODE_RSQ	0x1B
+#	define NV30_FP_OP_OPCODE_EX2	0x1C
+#	define NV30_FP_OP_OPCODE_LG2	0x1D
+#	define NV30_FP_OP_OPCODE_LIT	0x1E
+#	define NV30_FP_OP_OPCODE_LRP	0x1F
+#	define NV30_FP_OP_OPCODE_COS	0x22
+#	define NV30_FP_OP_OPCODE_SIN	0x23
+#	define NV30_FP_OP_OPCODE_PK2H   0x24
+#	define NV30_FP_OP_OPCODE_UP2H   0x25
+#	define NV30_FP_OP_OPCODE_POW	0x26
+#	define NV30_FP_OP_OPCODE_PK4UB  0x27
+#	define NV30_FP_OP_OPCODE_UP4UB  0x28
+#	define NV30_FP_OP_OPCODE_PK2US  0x29
+#	define NV30_FP_OP_OPCODE_UP2US  0x2A
+#	define NV30_FP_OP_OPCODE_DP2A   0x2E
+#	define NV30_FP_OP_OPCODE_TXB	0x31
+#	define NV30_FP_OP_OPCODE_RFL	0x36
+#define NV30_FP_OP_OUT_SAT					(1 << 31)
+
+/* high order bits of SRC0 */
+#define NV30_FP_OP_OUT_ABS					(1 << 29)
+#define NV30_FP_OP_COND_SWZ_W_SHIFT				27
+#define NV30_FP_OP_COND_SWZ_W_MASK				(3 << 27)
+#define NV30_FP_OP_COND_SWZ_Z_SHIFT				25
+#define NV30_FP_OP_COND_SWZ_Z_MASK				(3 << 25)
+#define NV30_FP_OP_COND_SWZ_Y_SHIFT				23
+#define NV30_FP_OP_COND_SWZ_Y_MASK				(3 << 23)
+#define NV30_FP_OP_COND_SWZ_X_SHIFT				21
+#define NV30_FP_OP_COND_SWZ_X_MASK				(3 << 21)
+#define NV30_FP_OP_COND_SWZ_ALL_SHIFT				21
+#define NV30_FP_OP_COND_SWZ_ALL_MASK				(0xFF << 21)
+#define NV30_FP_OP_COND_SHIFT					18
+#define NV30_FP_OP_COND_MASK					(0x07 << 18)
+#	define NV30_FP_OP_COND_FL	0
+#	define NV30_FP_OP_COND_LT	1
+#	define NV30_FP_OP_COND_EQ	2
+#	define NV30_FP_OP_COND_LE	3
+#	define NV30_FP_OP_COND_GT	4
+#	define NV30_FP_OP_COND_NE	5
+#	define NV30_FP_OP_COND_GE	6
+#	define NV30_FP_OP_COND_TR	7
+
+/* high order bits of SRC1 */
+#define NV30_FP_OP_SRC_SCALE_SHIFT				28
+#define NV30_FP_OP_SRC_SCALE_MASK				(3 << 28)
+
+/* high order bits of SRC2 */
+#define NV30_FP_OP_INDEX_INPUT					(1 << 30)
+
+//== Register selection ==
+#define NV30_FP_REG_ALL_MASK					(0x1FFFF<<0)
+#define NV30_FP_REG_TYPE_SHIFT					0
+#define NV30_FP_REG_TYPE_MASK					(3 << 0)
+#	define NV30_FP_REG_TYPE_TEMP	0
+#	define NV30_FP_REG_TYPE_INPUT	1
+#	define NV30_FP_REG_TYPE_CONST	2
+#define NV30_FP_REG_SRC_SHIFT					2 /* uncertain */
+#define NV30_FP_REG_SRC_MASK					(31 << 2)
+#define NV30_FP_REG_UNK_0					(1 << 8)
+#define NV30_FP_REG_SWZ_ALL_SHIFT				9
+#define NV30_FP_REG_SWZ_ALL_MASK				(255 << 9)
+#define NV30_FP_REG_SWZ_X_SHIFT					9
+#define NV30_FP_REG_SWZ_X_MASK					(3 << 9)
+#define NV30_FP_REG_SWZ_Y_SHIFT					11
+#define NV30_FP_REG_SWZ_Y_MASK					(3 << 11)
+#define NV30_FP_REG_SWZ_Z_SHIFT					13
+#define NV30_FP_REG_SWZ_Z_MASK					(3 << 13)
+#define NV30_FP_REG_SWZ_W_SHIFT					15
+#define NV30_FP_REG_SWZ_W_MASK					(3 << 15)
+#	define NV30_FP_SWIZZLE_X	0
+#	define NV30_FP_SWIZZLE_Y	1
+#	define NV30_FP_SWIZZLE_Z	2
+#	define NV30_FP_SWIZZLE_W	3
+#define NV30_FP_REG_NEGATE					(1 << 17)
+
+#endif
diff --git a/src/mesa/drivers/dri/nouveau/nv30_state.c b/src/mesa/drivers/dri/nouveau/nv30_state.c
new file mode 100644
index 00000000000..ad21fa27302
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv30_state.c
@@ -0,0 +1,980 @@
+/**************************************************************************
+
+Copyright 2006 Nouveau
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#include "nouveau_context.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+#include "nouveau_state.h"
+
+#include "tnl/t_pipeline.h"
+
+#include "mtypes.h"
+#include "colormac.h"
+
+#define NOUVEAU_CARD_USING_SHADERS (nmesa->screen->card->type >= NV_40)
+
+static void nv30AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte ubRef;
+	CLAMPED_FLOAT_TO_UBYTE(ubRef, ref);
+
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_ALPHA_FUNC_FUNC, 2);
+	OUT_RING_CACHE(func);     /* NV30_TCL_PRIMITIVE_3D_ALPHA_FUNC_FUNC */
+	OUT_RING_CACHE(ubRef);    /* NV30_TCL_PRIMITIVE_3D_ALPHA_FUNC_REF  */
+}
+
+static void nv30BlendColor(GLcontext *ctx, const GLfloat color[4])
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); 
+	GLubyte cf[4];
+
+	CLAMPED_FLOAT_TO_UBYTE(cf[0], color[0]);
+	CLAMPED_FLOAT_TO_UBYTE(cf[1], color[1]);
+	CLAMPED_FLOAT_TO_UBYTE(cf[2], color[2]);
+	CLAMPED_FLOAT_TO_UBYTE(cf[3], color[3]);
+
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_BLEND_COLOR, 1);
+	OUT_RING_CACHE(PACK_COLOR_8888(cf[3], cf[1], cf[2], cf[0]));
+}
+
+static void nv30BlendEquationSeparate(GLcontext *ctx, GLenum modeRGB, GLenum modeA)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_BLEND_EQUATION, 1);
+	OUT_RING_CACHE((modeA<<16) | modeRGB);
+}
+
+
+static void nv30BlendFuncSeparate(GLcontext *ctx, GLenum sfactorRGB, GLenum dfactorRGB,
+		GLenum sfactorA, GLenum dfactorA)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC, 2);
+	OUT_RING_CACHE((sfactorA<<16) | sfactorRGB);
+	OUT_RING_CACHE((dfactorA<<16) | dfactorRGB);
+}
+
+static void nv30Clear(GLcontext *ctx, GLbitfield mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLuint hw_bufs = 0;
+
+	if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT))
+		hw_bufs |= 0xf0;
+	if (mask & (BUFFER_BIT_DEPTH))
+		hw_bufs |= 0x03;
+
+	if (hw_bufs) {
+		BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_CLEAR_WHICH_BUFFERS, 1);
+		OUT_RING(hw_bufs);
+	}
+}
+
+static void nv30ClearColor(GLcontext *ctx, const GLfloat color[4])
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte c[4];
+	UNCLAMPED_FLOAT_TO_RGBA_CHAN(c,color);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_CLEAR_VALUE_ARGB, 1);
+	OUT_RING_CACHE(PACK_COLOR_8888(c[3],c[0],c[1],c[2]));
+}
+
+static void nv30ClearDepth(GLcontext *ctx, GLclampd d)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nmesa->clear_value=((nmesa->clear_value&0x000000FF)|(((uint32_t)(d*0xFFFFFF))<<8));
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_CLEAR_VALUE_DEPTH, 1);
+	OUT_RING_CACHE(nmesa->clear_value);
+}
+
+/* we're don't support indexed buffers
+   void (*ClearIndex)(GLcontext *ctx, GLuint index)
+ */
+
+static void nv30ClearStencil(GLcontext *ctx, GLint s)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	nmesa->clear_value=((nmesa->clear_value&0xFFFFFF00)|(s&0x000000FF));
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_CLEAR_VALUE_DEPTH, 1);
+	OUT_RING_CACHE(nmesa->clear_value);
+}
+
+static void nv30ClipPlane(GLcontext *ctx, GLenum plane, const GLfloat *equation)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (NOUVEAU_CARD_USING_SHADERS)
+		return;
+
+	plane -= GL_CLIP_PLANE0;
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_CLIP_PLANE_A(plane), 4);
+	OUT_RING_CACHEf(equation[0]);
+	OUT_RING_CACHEf(equation[1]);
+	OUT_RING_CACHEf(equation[2]);
+	OUT_RING_CACHEf(equation[3]);
+}
+
+static void nv30ColorMask(GLcontext *ctx, GLboolean rmask, GLboolean gmask,
+		GLboolean bmask, GLboolean amask )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_COLOR_MASK, 1);
+	OUT_RING_CACHE(((amask && 0x01) << 24) | ((rmask && 0x01) << 16) | ((gmask && 0x01)<< 8) | ((bmask && 0x01) << 0));
+}
+
+static void nv30ColorMaterial(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	// TODO I need love
+}
+
+static void nv30CullFace(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_CULL_FACE, 1);
+	OUT_RING_CACHE(mode);
+}
+
+static void nv30FrontFace(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_FRONT_FACE, 1);
+	OUT_RING_CACHE(mode);
+}
+
+static void nv30DepthFunc(GLcontext *ctx, GLenum func)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_DEPTH_FUNC, 1);
+	OUT_RING_CACHE(func);
+}
+
+static void nv30DepthMask(GLcontext *ctx, GLboolean flag)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_DEPTH_WRITE_ENABLE, 1);
+	OUT_RING_CACHE(flag);
+}
+
+static void nv30DepthRange(GLcontext *ctx, GLclampd nearval, GLclampd farval)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_DEPTH_RANGE_NEAR, 2);
+	OUT_RING_CACHEf(nearval);
+	OUT_RING_CACHEf(farval);
+}
+
+/** Specify the current buffer for writing */
+//void (*DrawBuffer)( GLcontext *ctx, GLenum buffer );
+/** Specify the buffers for writing for fragment programs*/
+//void (*DrawBuffers)( GLcontext *ctx, GLsizei n, const GLenum *buffers );
+
+static void nv30Enable(GLcontext *ctx, GLenum cap, GLboolean state)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	switch(cap)
+	{
+		case GL_ALPHA_TEST:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_ALPHA_FUNC_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_AUTO_NORMAL:
+		case GL_BLEND:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_BLEND_FUNC_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_CLIP_PLANE0:
+		case GL_CLIP_PLANE1:
+		case GL_CLIP_PLANE2:
+		case GL_CLIP_PLANE3:
+		case GL_CLIP_PLANE4:
+		case GL_CLIP_PLANE5:
+			if (NOUVEAU_CARD_USING_SHADERS) {
+				nouveauShader *nvs = (nouveauShader *)ctx->VertexProgram._Current;
+				if (nvs)
+					nvs->translated = GL_FALSE;
+			} else {
+				BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_CLIP_PLANE_ENABLE(cap-GL_CLIP_PLANE0), 1);
+				OUT_RING_CACHE(state);
+			}
+			break;
+		case GL_COLOR_LOGIC_OP:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_COLOR_MATERIAL:
+//		case GL_COLOR_SUM_EXT:
+//		case GL_COLOR_TABLE:
+//		case GL_CONVOLUTION_1D:
+//		case GL_CONVOLUTION_2D:
+		case GL_CULL_FACE:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_CULL_FACE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_DEPTH_TEST:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_DEPTH_TEST_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_DITHER:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_DITHER_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_FOG:
+			if (NOUVEAU_CARD_USING_SHADERS)
+				break;
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_FOG_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_HISTOGRAM:
+//		case GL_INDEX_LOGIC_OP:
+		case GL_LIGHT0:
+		case GL_LIGHT1:
+		case GL_LIGHT2:
+		case GL_LIGHT3:
+		case GL_LIGHT4:
+		case GL_LIGHT5:
+		case GL_LIGHT6:
+		case GL_LIGHT7:
+			{
+			uint32_t mask=0x11<<(2*(cap-GL_LIGHT0));
+
+			if (NOUVEAU_CARD_USING_SHADERS)
+				break;
+
+			nmesa->enabled_lights=((nmesa->enabled_lights&mask)|(mask*state));
+			if (nmesa->lighting_enabled)
+			{
+				BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_ENABLED_LIGHTS, 1);
+				OUT_RING_CACHE(nmesa->enabled_lights);
+			}
+			break;
+			}
+		case GL_LIGHTING:
+			if (NOUVEAU_CARD_USING_SHADERS)
+				break;
+
+			nmesa->lighting_enabled=state;
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_ENABLED_LIGHTS, 1);
+			if (nmesa->lighting_enabled)
+				OUT_RING_CACHE(nmesa->enabled_lights);
+			else
+				OUT_RING_CACHE(0x0);
+			break;
+//		case GL_LINE_SMOOTH:
+//		case GL_LINE_STIPPLE:
+//		case GL_MAP1_COLOR_4:
+//		case GL_MAP1_INDEX:
+//		case GL_MAP1_NORMAL:
+//		case GL_MAP1_TEXTURE_COORD_1:
+//		case GL_MAP1_TEXTURE_COORD_2:
+//		case GL_MAP1_TEXTURE_COORD_3:
+//		case GL_MAP1_TEXTURE_COORD_4:
+//		case GL_MAP1_VERTEX_3:
+//		case GL_MAP1_VERTEX_4:
+//		case GL_MAP2_COLOR_4:
+//		case GL_MAP2_INDEX:
+//		case GL_MAP2_NORMAL:
+//		case GL_MAP2_TEXTURE_COORD_1:
+//		case GL_MAP2_TEXTURE_COORD_2:
+//		case GL_MAP2_TEXTURE_COORD_3:
+//		case GL_MAP2_TEXTURE_COORD_4:
+//		case GL_MAP2_VERTEX_3:
+//		case GL_MAP2_VERTEX_4:
+//		case GL_MINMAX:
+		case GL_NORMALIZE:
+			if (nmesa->screen->card->type != NV_44) {
+				BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_NORMALIZE_ENABLE, 1);
+				OUT_RING_CACHE(state);
+			}
+			break;
+//		case GL_POINT_SMOOTH:
+		case GL_POLYGON_OFFSET_POINT:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POLYGON_OFFSET_POINT_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_OFFSET_LINE:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POLYGON_OFFSET_LINE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_OFFSET_FILL:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FILL_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_SMOOTH:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POLYGON_SMOOTH_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_STIPPLE:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_POST_COLOR_MATRIX_COLOR_TABLE:
+//		case GL_POST_CONVOLUTION_COLOR_TABLE:
+//		case GL_RESCALE_NORMAL:
+		case GL_SCISSOR_TEST:
+			/* No enable bit, nv30Scissor will adjust to max range */
+			ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
+			      			 ctx->Scissor.Width, ctx->Scissor.Height);
+			break;
+//		case GL_SEPARABLE_2D:
+		case GL_STENCIL_TEST:
+			// TODO BACK and FRONT ?
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_TEXTURE_GEN_Q:
+//		case GL_TEXTURE_GEN_R:
+//		case GL_TEXTURE_GEN_S:
+//		case GL_TEXTURE_GEN_T:
+//		case GL_TEXTURE_1D:
+//		case GL_TEXTURE_2D:
+//		case GL_TEXTURE_3D:
+	}
+}
+
+static void nv30Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+    nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+    if (NOUVEAU_CARD_USING_SHADERS)
+        return;
+
+    switch(pname)
+    {
+    case GL_FOG_MODE:
+    {
+        int mode = 0;
+        /* The modes are different in GL and the card.  */
+        switch(ctx->Fog.Mode)
+        {
+        case GL_LINEAR:
+            mode = 0x804;
+            break;
+        case GL_EXP:
+            mode = 0x802;
+            break;
+        case GL_EXP2:
+            mode = 0x803;
+            break;
+        }
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_FOG_MODE, 1);
+	OUT_RING_CACHE (mode);
+	break;
+    }
+    case GL_FOG_COLOR:
+    {
+	GLubyte c[4];
+	UNCLAMPED_FLOAT_TO_RGBA_CHAN(c,params);
+        BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_FOG_COLOR, 1);
+        /* nvidia ignores the alpha channel */
+	OUT_RING_CACHE(PACK_COLOR_8888_REV(c[0],c[1],c[2],c[3]));
+        break;
+    }
+    case GL_FOG_DENSITY:
+    case GL_FOG_START:
+    case GL_FOG_END:
+    {
+        GLfloat f=0., c=0.;
+        switch(ctx->Fog.Mode)
+        {
+        case GL_LINEAR:
+            f = -1.0/(ctx->Fog.End - ctx->Fog.Start);
+            c = ctx->Fog.Start/(ctx->Fog.End - ctx->Fog.Start) + 2.001953;
+            break;
+        case GL_EXP:
+            f = -0.090168*ctx->Fog.Density;
+            c = 1.5;
+        case GL_EXP2:
+            f = -0.212330*ctx->Fog.Density;
+            c = 1.5;
+        }
+        BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_FOG_EQUATION_LINEAR, 1);
+        OUT_RING_CACHE(f);
+        BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_FOG_EQUATION_CONSTANT, 1);
+        OUT_RING_CACHE(c);
+        BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_FOG_EQUATION_QUADRATIC, 1);
+        OUT_RING_CACHE(0); /* Is this always the same? */
+        break;
+    }
+//    case GL_FOG_COORD_SRC:
+    default:
+        break;
+    }
+}
+   
+static void nv30Hint(GLcontext *ctx, GLenum target, GLenum mode)
+{
+	// TODO I need love (fog and line_smooth hints)
+}
+
+// void (*IndexMask)(GLcontext *ctx, GLuint mask);
+
+enum {
+	SPOTLIGHT_NO_UPDATE,
+	SPOTLIGHT_UPDATE_EXPONENT,
+	SPOTLIGHT_UPDATE_DIRECTION,
+	SPOTLIGHT_UPDATE_ALL
+};
+
+static void nv30Lightfv(GLcontext *ctx, GLenum light, GLenum pname, const GLfloat *params )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLint p = light - GL_LIGHT0;
+	struct gl_light *l = &ctx->Light.Light[p];
+	int spotlight_update = SPOTLIGHT_NO_UPDATE;
+
+	if (NOUVEAU_CARD_USING_SHADERS)
+	   return;
+
+	/* not sure where the fourth param value goes...*/
+	switch(pname)
+	{
+		case GL_AMBIENT:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_AMBIENT_R(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_DIFFUSE:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_DIFFUSE_R(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_SPECULAR:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_FRONT_SIDE_PRODUCT_SPECULAR_R(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_POSITION:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_POSITION_X(p), 3);
+			OUT_RING_CACHEf(params[0]);
+			OUT_RING_CACHEf(params[1]);
+			OUT_RING_CACHEf(params[2]);
+			break;
+		case GL_SPOT_DIRECTION:
+			spotlight_update = SPOTLIGHT_UPDATE_DIRECTION;
+			break;
+		case GL_SPOT_EXPONENT:
+			spotlight_update = SPOTLIGHT_UPDATE_EXPONENT;
+			break;
+		case GL_SPOT_CUTOFF:
+			spotlight_update = SPOTLIGHT_UPDATE_ALL;
+			break;
+		case GL_CONSTANT_ATTENUATION:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_CONSTANT_ATTENUATION(p), 1);
+			OUT_RING_CACHEf(*params);
+			break;
+		case GL_LINEAR_ATTENUATION:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_LINEAR_ATTENUATION(p), 1);
+			OUT_RING_CACHEf(*params);
+			break;
+		case GL_QUADRATIC_ATTENUATION:
+			BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_QUADRATIC_ATTENUATION(p), 1);
+			OUT_RING_CACHEf(*params);
+			break;
+		default:
+			break;
+	}
+
+	switch(spotlight_update) {
+		case SPOTLIGHT_UPDATE_DIRECTION:
+			{
+				GLfloat x,y,z;
+				GLfloat spot_light_coef_a = 1.0 / (l->_CosCutoff - 1.0);
+				x = spot_light_coef_a * l->_NormDirection[0];
+				y = spot_light_coef_a * l->_NormDirection[1];
+				z = spot_light_coef_a * l->_NormDirection[2];
+				BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_DIR_X(p), 3);
+				OUT_RING_CACHEf(x);
+				OUT_RING_CACHEf(y);
+				OUT_RING_CACHEf(z);
+			}
+			break;
+		case SPOTLIGHT_UPDATE_EXPONENT:
+			{
+				GLfloat cc,lc,qc;
+				cc = 1.0;	/* FIXME: These need to be correctly computed */
+				lc = 0.0;
+				qc = 2.0;
+				BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_A(p), 3);
+				OUT_RING_CACHEf(cc);
+				OUT_RING_CACHEf(lc);
+				OUT_RING_CACHEf(qc);
+			}
+			break;
+		case SPOTLIGHT_UPDATE_ALL:
+			{
+				GLfloat cc,lc,qc, x,y,z, c;
+				GLfloat spot_light_coef_a = 1.0 / (l->_CosCutoff - 1.0);
+				cc = 1.0;	/* FIXME: These need to be correctly computed */
+				lc = 0.0;
+				qc = 2.0;
+				x = spot_light_coef_a * l->_NormDirection[0];
+				y = spot_light_coef_a * l->_NormDirection[1];
+				z = spot_light_coef_a * l->_NormDirection[2];
+				c = spot_light_coef_a + 1.0;
+				BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LIGHT_SPOT_CUTOFF_A(p), 7);
+				OUT_RING_CACHEf(cc);
+				OUT_RING_CACHEf(lc);
+				OUT_RING_CACHEf(qc);
+				OUT_RING_CACHEf(x);
+				OUT_RING_CACHEf(y);
+				OUT_RING_CACHEf(z);
+				OUT_RING_CACHEf(c);
+			}
+			break;
+		default:
+			break;
+	}
+}
+
+/** Set the lighting model parameters */
+void (*LightModelfv)(GLcontext *ctx, GLenum pname, const GLfloat *params);
+
+
+static void nv30LineStipple(GLcontext *ctx, GLint factor, GLushort pattern )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LINE_STIPPLE_PATTERN, 1);
+	OUT_RING_CACHE((pattern << 16) | factor);
+}
+
+static void nv30LineWidth(GLcontext *ctx, GLfloat width)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte ubWidth;
+
+	ubWidth = (GLubyte)(width * 8.0) & 0xFF;
+
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LINE_WIDTH_SMOOTH, 1);
+	OUT_RING_CACHE(ubWidth);
+}
+
+static void nv30LogicOpcode(GLcontext *ctx, GLenum opcode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_COLOR_LOGIC_OP_OP, 1);
+	OUT_RING_CACHE(opcode);
+}
+
+static void nv30PointParameterfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+	/*TODO: not sure what goes here. */
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	
+}
+
+/** Specify the diameter of rasterized points */
+static void nv30PointSize(GLcontext *ctx, GLfloat size)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POINT_SIZE, 1);
+	OUT_RING_CACHEf(size);
+}
+
+/** Select a polygon rasterization mode */
+static void nv30PolygonMode(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POLYGON_MODE_FRONT, 1);
+		OUT_RING_CACHE(mode);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POLYGON_MODE_BACK, 1);
+		OUT_RING_CACHE(mode);
+	}
+}
+
+/** Set the scale and units used to calculate depth values */
+static void nv30PolygonOffset(GLcontext *ctx, GLfloat factor, GLfloat units)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+        BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FACTOR, 2);
+        OUT_RING_CACHEf(factor);
+
+        /* Looks like we always multiply units by 2.0... according to the dumps.*/
+        OUT_RING_CACHEf(units * 2.0);
+}
+
+/** Set the polygon stippling pattern */
+static void nv30PolygonStipple(GLcontext *ctx, const GLubyte *mask )
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+        BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_PATTERN(0), 32);
+        OUT_RING_CACHEp(mask, 32);
+}
+
+/* Specifies the current buffer for reading */
+void (*ReadBuffer)( GLcontext *ctx, GLenum buffer );
+/** Set rasterization mode */
+void (*RenderMode)(GLcontext *ctx, GLenum mode );
+
+/** Define the scissor box */
+static void nv30Scissor(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	/* There's no scissor enable bit, so adjust the scissor to cover the
+	 * maximum draw buffer bounds
+	 */
+	if (!ctx->Scissor.Enabled) {
+	   x = y = 0;
+	   w = h = 4095;
+	} else {
+	   x += nmesa->drawX;
+	   y += nmesa->drawY;
+	}
+
+        BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SCISSOR_WIDTH_XPOS, 2);
+        OUT_RING_CACHE(((w) << 16) | x);
+        OUT_RING_CACHE(((h) << 16) | y);
+}
+
+/** Select flat or smooth shading */
+static void nv30ShadeModel(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SHADE_MODEL, 1);
+	OUT_RING_CACHE(mode);
+}
+
+/** OpenGL 2.0 two-sided StencilFunc */
+static void nv30StencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func,
+		GLint ref, GLuint mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_FUNC, 3);
+		OUT_RING_CACHE(func);
+		OUT_RING_CACHE(ref);
+		OUT_RING_CACHE(mask);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_FUNC_FUNC, 3);
+		OUT_RING_CACHE(func);
+		OUT_RING_CACHE(ref);
+		OUT_RING_CACHE(mask);
+	}
+}
+
+/** OpenGL 2.0 two-sided StencilMask */
+static void nv30StencilMaskSeparate(GLcontext *ctx, GLenum face, GLuint mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_MASK, 1);
+		OUT_RING_CACHE(mask);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_MASK, 1);
+		OUT_RING_CACHE(mask);
+	}
+}
+
+/** OpenGL 2.0 two-sided StencilOp */
+static void nv30StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail,
+		GLenum zfail, GLenum zpass)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_STENCIL_FRONT_OP_FAIL, 3);
+		OUT_RING_CACHE(fail);
+		OUT_RING_CACHE(zfail);
+		OUT_RING_CACHE(zpass);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_STENCIL_BACK_OP_FAIL, 3);
+		OUT_RING_CACHE(fail);
+		OUT_RING_CACHE(zfail);
+		OUT_RING_CACHE(zpass);
+	}
+}
+
+/** Control the generation of texture coordinates */
+void (*TexGen)(GLcontext *ctx, GLenum coord, GLenum pname,
+		const GLfloat *params);
+/** Set texture environment parameters */
+void (*TexEnv)(GLcontext *ctx, GLenum target, GLenum pname,
+		const GLfloat *param);
+/** Set texture parameters */
+void (*TexParameter)(GLcontext *ctx, GLenum target,
+		struct gl_texture_object *texObj,
+		GLenum pname, const GLfloat *params);
+
+static void nv30TextureMatrix(GLcontext *ctx, GLuint unit, const GLmatrix *mat)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (!NOUVEAU_CARD_USING_SHADERS) {
+		BEGIN_RING_CACHE(NvSub3D,
+				 NV30_TCL_PRIMITIVE_3D_TX_MATRIX(unit, 0), 16);
+		/*XXX: This SHOULD work.*/
+		OUT_RING_CACHEp(mat->m, 16);
+	}
+}
+
+static void nv30WindowMoved(nouveauContextPtr nmesa)
+{
+	GLcontext *ctx = nmesa->glCtx;
+	GLfloat *v = nmesa->viewport.m;
+	GLuint w = ctx->Viewport.Width;
+	GLuint h = ctx->Viewport.Height;
+	GLuint x = ctx->Viewport.X + nmesa->drawX;
+	GLuint y = ctx->Viewport.Y + nmesa->drawY;
+
+        BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VIEWPORT_DIMS_0, 2);
+        OUT_RING_CACHE((w << 16) | x);
+        OUT_RING_CACHE((h << 16) | y);
+	/* something to do with clears, possibly doesn't belong here */
+	BEGIN_RING_CACHE(NvSub3D,
+	      NV30_TCL_PRIMITIVE_3D_VIEWPORT_COLOR_BUFFER_OFS0, 2);
+        OUT_RING_CACHE(((w+x) << 16) | x);
+        OUT_RING_CACHE(((h+y) << 16) | y);
+	/* viewport transform */
+	BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VIEWPORT_XFRM_OX, 8);
+	OUT_RING_CACHEf (v[MAT_TX]);
+	OUT_RING_CACHEf (v[MAT_TY]);
+	OUT_RING_CACHEf (v[MAT_TZ]);
+	OUT_RING_CACHEf (0.0);
+	OUT_RING_CACHEf (v[MAT_SX]);
+	OUT_RING_CACHEf (v[MAT_SY]);
+	OUT_RING_CACHEf (v[MAT_SZ]);
+	OUT_RING_CACHEf (0.0);
+
+	ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
+	      		    ctx->Scissor.Width, ctx->Scissor.Height);
+}
+
+static GLboolean nv30InitCard(nouveauContextPtr nmesa)
+{
+	int i;
+	nouveauObjectOnSubchannel(nmesa, NvSub3D, Nv3D);
+
+	BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_OBJECT1, 3);
+	OUT_RING(NvDmaFB);
+	OUT_RING(NvDmaAGP);
+        OUT_RING(NvDmaFB);
+	BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_OBJECT8, 1);
+	OUT_RING(NvDmaFB);
+	BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_OBJECT4, 2);
+	OUT_RING(NvDmaFB);
+	OUT_RING(NvDmaFB);
+        BEGIN_RING_SIZE(NvSub3D, 0x1b0, 1); /* SET_OBJECT8B*/
+        OUT_RING(NvDmaFB);
+
+        for(i = 0x2c8; i <= 0x2fc; i += 4)
+        {
+            BEGIN_RING_SIZE(NvSub3D, i, 1);
+            OUT_RING(0x0);
+        }
+
+	BEGIN_RING_SIZE(NvSub3D, 0x0220, 1);
+	OUT_RING(1);
+
+	BEGIN_RING_SIZE(NvSub3D, 0x03b0, 1);
+	OUT_RING(0x00100000);
+	BEGIN_RING_SIZE(NvSub3D, 0x1454, 1);
+	OUT_RING(0);
+	BEGIN_RING_SIZE(NvSub3D, 0x1d80, 1);
+	OUT_RING(3);
+	
+	/* NEW */
+       	BEGIN_RING_SIZE(NvSub3D, 0x1e98, 1);
+        OUT_RING(0);
+        BEGIN_RING_SIZE(NvSub3D, 0x17e0, 3);
+        OUT_RING(0);
+        OUT_RING(0);
+        OUT_RING(0x3f800000);
+        BEGIN_RING_SIZE(NvSub3D, 0x1f80, 16);
+        OUT_RING(0); OUT_RING(0); OUT_RING(0); OUT_RING(0); 
+        OUT_RING(0); OUT_RING(0); OUT_RING(0); OUT_RING(0); 
+        OUT_RING(0x0000ffff);
+        OUT_RING(0); OUT_RING(0); OUT_RING(0); OUT_RING(0); 
+        OUT_RING(0); OUT_RING(0); OUT_RING(0); 
+/*
+        BEGIN_RING_SIZE(NvSub3D, 0x100, 2);
+        OUT_RING(0);
+        OUT_RING(0);
+*/
+        BEGIN_RING_SIZE(NvSub3D, 0x120, 3);
+        OUT_RING(0);
+        OUT_RING(1);
+        OUT_RING(2);
+
+        BEGIN_RING_SIZE(NvSub3D, 0x1d88, 1);
+        OUT_RING(0x00001200);
+
+	BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_RC_ENABLE, 1);
+	OUT_RING       (0);
+
+	return GL_TRUE;
+}
+
+static GLboolean nv40InitCard(nouveauContextPtr nmesa)
+{
+	nouveauObjectOnSubchannel(nmesa, NvSub3D, Nv3D);
+
+	BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_OBJECT1, 2);
+	OUT_RING(NvDmaFB);
+	OUT_RING(NvDmaFB);
+	BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_OBJECT8, 1);
+	OUT_RING(NvDmaFB);
+	BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_OBJECT4, 2);
+	OUT_RING(NvDmaFB);
+	OUT_RING(NvDmaFB);
+	BEGIN_RING_SIZE(NvSub3D, 0x0220, 1);
+	OUT_RING(1);
+
+	BEGIN_RING_SIZE(NvSub3D, 0x1ea4, 3);
+	OUT_RING(0x00000010);
+	OUT_RING(0x01000100);
+	OUT_RING(0xff800006);
+	BEGIN_RING_SIZE(NvSub3D, 0x1fc4, 1);
+	OUT_RING(0x06144321);
+	BEGIN_RING_SIZE(NvSub3D, 0x1fc8, 2);
+	OUT_RING(0xedcba987);
+	OUT_RING(0x00000021);
+	BEGIN_RING_SIZE(NvSub3D, 0x1fd0, 1);
+	OUT_RING(0x00171615);
+	BEGIN_RING_SIZE(NvSub3D, 0x1fd4, 1);
+	OUT_RING(0x001b1a19);
+
+	BEGIN_RING_SIZE(NvSub3D, 0x1ef8, 1);
+	OUT_RING(0x0020ffff);
+	BEGIN_RING_SIZE(NvSub3D, 0x1d64, 1);
+	OUT_RING(0x00d30000);
+	BEGIN_RING_SIZE(NvSub3D, 0x1e94, 1);
+	OUT_RING(0x00000001);
+
+	return GL_TRUE;
+}
+
+static GLboolean nv30BindBuffers(nouveauContextPtr nmesa, int num_color,
+		nouveau_renderbuffer **color,
+		nouveau_renderbuffer *depth)
+{
+	GLuint x, y, w, h;
+
+	w = color[0]->mesa.Width;
+	h = color[0]->mesa.Height;
+	x = nmesa->drawX;
+	y = nmesa->drawY;
+
+	if (num_color != 1)
+		return GL_FALSE;
+	BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VIEWPORT_COLOR_BUFFER_DIM0, 5);
+	OUT_RING        (((w+x)<<16)|x);
+	OUT_RING        (((h+y)<<16)|y);
+	if (color[0]->mesa._ActualFormat == GL_RGBA8)
+		OUT_RING        (0x148);
+	else
+		OUT_RING        (0x143);
+	if (nmesa->screen->card->type >= NV_40)
+		OUT_RING        (color[0]->pitch);
+	else
+		OUT_RING        (color[0]->pitch | (depth ? (depth->pitch << 16): 0));
+	OUT_RING        (color[0]->offset);
+
+	if (depth) {
+		BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_DEPTH_OFFSET, 1);
+		OUT_RING        (depth->offset);
+		if (nmesa->screen->card->type >= NV_40) {
+			BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_LMA_DEPTH_BUFFER_PITCH, 1);
+			OUT_RING        (depth->pitch);
+		}
+	}
+
+	return GL_TRUE;
+}
+
+void nv30InitStateFuncs(GLcontext *ctx, struct dd_function_table *func)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	func->AlphaFunc			= nv30AlphaFunc;
+	func->BlendColor		= nv30BlendColor;
+	func->BlendEquationSeparate	= nv30BlendEquationSeparate;
+	func->BlendFuncSeparate		= nv30BlendFuncSeparate;
+	func->Clear			= nv30Clear;
+	func->ClearColor		= nv30ClearColor;
+	func->ClearDepth		= nv30ClearDepth;
+	func->ClearStencil		= nv30ClearStencil;
+	func->ClipPlane			= nv30ClipPlane;
+	func->ColorMask			= nv30ColorMask;
+	func->ColorMaterial		= nv30ColorMaterial;
+	func->CullFace			= nv30CullFace;
+	func->FrontFace			= nv30FrontFace;
+	func->DepthFunc			= nv30DepthFunc;
+	func->DepthMask			= nv30DepthMask;
+	func->DepthRange                = nv30DepthRange;
+	func->Enable			= nv30Enable;
+	func->Fogfv			= nv30Fogfv;
+	func->Hint			= nv30Hint;
+	func->Lightfv			= nv30Lightfv;
+/*	func->LightModelfv		= nv30LightModelfv; */
+	func->LineStipple		= nv30LineStipple;
+	func->LineWidth			= nv30LineWidth;
+	func->LogicOpcode		= nv30LogicOpcode;
+	func->PointParameterfv		= nv30PointParameterfv;
+	func->PointSize			= nv30PointSize;
+	func->PolygonMode		= nv30PolygonMode;
+	func->PolygonOffset		= nv30PolygonOffset;
+	func->PolygonStipple		= nv30PolygonStipple;
+#if 0
+	func->ReadBuffer		= nv30ReadBuffer;
+	func->RenderMode		= nv30RenderMode;
+#endif
+	func->Scissor			= nv30Scissor;
+	func->ShadeModel		= nv30ShadeModel;
+	func->StencilFuncSeparate	= nv30StencilFuncSeparate;
+	func->StencilMaskSeparate	= nv30StencilMaskSeparate;
+	func->StencilOpSeparate		= nv30StencilOpSeparate;
+#if 0
+	func->TexGen			= nv30TexGen;
+	func->TexParameter		= nv30TexParameter;
+#endif
+	func->TextureMatrix		= nv30TextureMatrix;
+
+
+	if (nmesa->screen->card->type >= NV_40)
+	   nmesa->hw_func.InitCard	= nv40InitCard;
+	else 
+	   nmesa->hw_func.InitCard	= nv30InitCard;
+	nmesa->hw_func.BindBuffers	= nv30BindBuffers;
+	nmesa->hw_func.WindowMoved	= nv30WindowMoved;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nv30_vertprog.c b/src/mesa/drivers/dri/nouveau/nv30_vertprog.c
new file mode 100644
index 00000000000..d023e8439e3
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv30_vertprog.c
@@ -0,0 +1,367 @@
+#include "nouveau_context.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+
+#include "nouveau_shader.h"
+#include "nv30_shader.h"
+
+/*****************************************************************************
+ * Support routines
+ */
+static void
+NV30VPUploadToHW(GLcontext *ctx, nouveauShader *nvs)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   int i;
+
+   /* We can do better here and keep more than one VP on the hardware, and
+    * switch between them with PROGRAM_START_ID..
+    */
+   BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_FROM_ID, 1);
+   OUT_RING(0);
+   for (i=0; i<nvs->program_size; i+=4) {
+      BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_INST0, 4);
+      OUT_RING(nvs->program[i + 0]);
+      OUT_RING(nvs->program[i + 1]);
+      OUT_RING(nvs->program[i + 2]);
+      OUT_RING(nvs->program[i + 3]);
+   }
+   BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VP_PROGRAM_START_ID, 1);
+   OUT_RING(0);
+
+   BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VP_IN_REG, 2);
+   OUT_RING(nvs->card_priv.NV30VP.vp_in_reg);
+   OUT_RING(nvs->card_priv.NV30VP.vp_out_reg);
+
+   BEGIN_RING_CACHE(NvSub3D, NV30_TCL_PRIMITIVE_3D_SET_CLIPPING_PLANES, 1);
+   OUT_RING_CACHE  (nvs->card_priv.NV30VP.clip_enables);
+}
+
+static void
+NV30VPUpdateConst(GLcontext *ctx, nouveauShader *nvs, int id)
+{
+   nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+   GLfloat *val;
+
+   val = nvs->params[id].source_val ?
+      nvs->params[id].source_val : nvs->params[id].val;
+
+   BEGIN_RING_SIZE(NvSub3D, NV30_TCL_PRIMITIVE_3D_VP_UPLOAD_CONST_ID, 5);
+   OUT_RING (id);
+   OUT_RINGp(val, 4);
+}
+
+/*****************************************************************************
+ * Assembly routines
+ */
+static void
+NV30VPSetBranchTarget(nvsFunc *shader, int addr)
+{
+	shader->inst[2] &= ~NV30_VP_INST_IADDR_MASK;
+	shader->inst[2] |= (addr << NV30_VP_INST_IADDR_SHIFT);
+}
+
+/*****************************************************************************
+ * Disassembly routines
+ */
+static unsigned int
+NV30VPGetOpcodeHW(nvsFunc * shader, int slot)
+{
+   int op;
+
+   if (slot) {
+      op = (shader->inst[1] & NV30_VP_INST_SCA_OPCODEL_MASK)
+	 >> NV30_VP_INST_SCA_OPCODEL_SHIFT;
+      op |= ((shader->inst[0] & NV30_VP_INST_SCA_OPCODEH_MASK)
+	     >> NV30_VP_INST_SCA_OPCODEH_SHIFT) << 4;
+   }
+   else {
+      op = (shader->inst[1] & NV30_VP_INST_VEC_OPCODE_MASK)
+	 >> NV30_VP_INST_VEC_OPCODE_SHIFT;
+   }
+
+   return op;
+}
+
+static nvsRegFile
+NV30VPGetDestFile(nvsFunc * shader, int merged)
+{
+   switch (shader->GetOpcode(shader, merged)) {
+   case NVS_OP_ARL:
+   case NVS_OP_ARR:
+   case NVS_OP_ARA:
+      return NVS_FILE_ADDRESS;
+   default:
+      /*FIXME: This probably isn't correct.. */
+      if ((shader->inst[3] & NV30_VP_INST_VDEST_WRITEMASK_MASK) != 0)
+	 return NVS_FILE_RESULT;
+      if ((shader->inst[3] & NV30_VP_INST_SDEST_WRITEMASK_MASK) != 0)
+	 return NVS_FILE_RESULT;
+      return NVS_FILE_TEMP;
+   }
+}
+
+static unsigned int
+NV30VPGetDestID(nvsFunc * shader, int merged)
+{
+   int id;
+
+   switch (shader->GetDestFile(shader, merged)) {
+   case NVS_FILE_RESULT:
+      id = ((shader->inst[3] & NV30_VP_INST_DEST_ID_MASK)
+	    >> NV30_VP_INST_DEST_ID_SHIFT);
+      switch (id) {
+      case NV30_VP_INST_DEST_POS  : return NVS_FR_POSITION;
+      case NV30_VP_INST_DEST_COL0 : return NVS_FR_COL0;
+      case NV30_VP_INST_DEST_COL1 : return NVS_FR_COL1;
+      case NV30_VP_INST_DEST_TC(0): return NVS_FR_TEXCOORD0;
+      case NV30_VP_INST_DEST_TC(1): return NVS_FR_TEXCOORD1;
+      case NV30_VP_INST_DEST_TC(2): return NVS_FR_TEXCOORD2;
+      case NV30_VP_INST_DEST_TC(3): return NVS_FR_TEXCOORD3;
+      case NV30_VP_INST_DEST_TC(4): return NVS_FR_TEXCOORD4;
+      case NV30_VP_INST_DEST_TC(5): return NVS_FR_TEXCOORD5;
+      case NV30_VP_INST_DEST_TC(6): return NVS_FR_TEXCOORD6;
+      case NV30_VP_INST_DEST_TC(7): return NVS_FR_TEXCOORD7;
+      default:
+	 return -1;
+      }
+   case NVS_FILE_ADDRESS:
+   case NVS_FILE_TEMP:
+      return (shader->inst[0] & NV30_VP_INST_DEST_TEMP_ID_MASK)
+	 >> NV30_VP_INST_DEST_TEMP_ID_SHIFT;
+   default:
+      return -1;
+   }
+}
+
+static unsigned int
+NV30VPGetDestMask(nvsFunc * shader, int merged)
+{
+   int hwmask, mask = 0;
+
+   if (shader->GetDestFile(shader, merged) == NVS_FILE_RESULT)
+      if (shader->GetOpcodeSlot(shader, merged))
+	 hwmask = (shader->inst[3] & NV30_VP_INST_SDEST_WRITEMASK_MASK)
+	    >> NV30_VP_INST_SDEST_WRITEMASK_SHIFT;
+      else
+	 hwmask = (shader->inst[3] & NV30_VP_INST_VDEST_WRITEMASK_MASK)
+	    >> NV30_VP_INST_VDEST_WRITEMASK_SHIFT;
+   else if (shader->GetOpcodeSlot(shader, merged))
+      hwmask = (shader->inst[3] & NV30_VP_INST_STEMP_WRITEMASK_MASK)
+	 >> NV30_VP_INST_STEMP_WRITEMASK_SHIFT;
+   else
+      hwmask = (shader->inst[3] & NV30_VP_INST_VTEMP_WRITEMASK_MASK)
+	 >> NV30_VP_INST_VTEMP_WRITEMASK_SHIFT;
+
+   if (hwmask & (1 << 3)) mask |= SMASK_X;
+   if (hwmask & (1 << 2)) mask |= SMASK_Y;
+   if (hwmask & (1 << 1)) mask |= SMASK_Z;
+   if (hwmask & (1 << 0)) mask |= SMASK_W;
+
+   return mask;
+}
+
+static int
+NV30VPGetSourceID(nvsFunc * shader, int merged, int pos)
+{
+   unsigned int src;
+
+   switch (shader->GetSourceFile(shader, merged, pos)) {
+   case NVS_FILE_TEMP:
+      src = shader->GetSourceHW(shader, merged, pos);
+      return ((src & NV30_VP_SRC_REG_TEMP_ID_MASK) >>
+	      NV30_VP_SRC_REG_TEMP_ID_SHIFT);
+   case NVS_FILE_CONST:
+      return ((shader->inst[1] & NV30_VP_INST_CONST_SRC_MASK)
+	      >> NV30_VP_INST_CONST_SRC_SHIFT);
+   case NVS_FILE_ATTRIB:
+      src = ((shader->inst[1] & NV30_VP_INST_INPUT_SRC_MASK)
+	     >> NV30_VP_INST_INPUT_SRC_SHIFT);
+      switch (src) {
+      case NV30_VP_INST_IN_POS  : return NVS_FR_POSITION;
+      case NV30_VP_INST_IN_COL0 : return NVS_FR_COL0;
+      case NV30_VP_INST_IN_COL1 : return NVS_FR_COL1;
+      case NV30_VP_INST_IN_TC(0): return NVS_FR_TEXCOORD0;
+      case NV30_VP_INST_IN_TC(1): return NVS_FR_TEXCOORD1;
+      case NV30_VP_INST_IN_TC(2): return NVS_FR_TEXCOORD2;
+      case NV30_VP_INST_IN_TC(3): return NVS_FR_TEXCOORD3;
+      case NV30_VP_INST_IN_TC(4): return NVS_FR_TEXCOORD4;
+      case NV30_VP_INST_IN_TC(5): return NVS_FR_TEXCOORD5;
+      case NV30_VP_INST_IN_TC(6): return NVS_FR_TEXCOORD6;
+      case NV30_VP_INST_IN_TC(7): return NVS_FR_TEXCOORD7;
+      default:
+	 return NVS_FR_UNKNOWN;
+      }
+   default:
+      return -1;
+   }
+}
+
+static int
+NV30VPGetSourceAbs(nvsFunc * shader, int merged, int pos)
+{
+   struct _op_xlat *opr;
+   static unsigned int abspos[3] = {
+      NV30_VP_INST_SRC0_ABS,
+      NV30_VP_INST_SRC1_ABS,
+      NV30_VP_INST_SRC2_ABS,
+   };
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr || opr->srcpos[pos] == -1 || opr->srcpos[pos] > 2)
+      return 0;
+
+   return ((shader->inst[0] & abspos[opr->srcpos[pos]]) ? 1 : 0);
+}
+
+static int
+NV30VPGetRelAddressRegID(nvsFunc * shader)
+{
+   return ((shader->inst[0] & NV30_VP_INST_ADDR_REG_SELECT_1) ? 1 : 0);
+}
+
+static nvsSwzComp
+NV30VPGetRelAddressSwizzle(nvsFunc * shader)
+{
+   nvsSwzComp swz;
+
+   swz = NV20VP_TX_SWIZZLE[(shader->inst[0] & NV30_VP_INST_ADDR_SWZ_MASK)
+			   >> NV30_VP_INST_ADDR_SWZ_SHIFT];
+   return swz;
+}
+
+static int
+NV30VPSupportsConditional(nvsFunc * shader)
+{
+   /*FIXME: Is this true of all ops? */
+   return 1;
+}
+
+static int
+NV30VPGetConditionUpdate(nvsFunc * shader)
+{
+   return ((shader->inst[0] & NV30_VP_INST_COND_UPDATE_ENABLE) ? 1 : 0);
+}
+
+static int
+NV30VPGetConditionTest(nvsFunc * shader)
+{
+   int op;
+
+   /* The condition test is unconditionally enabled on some
+    * instructions. ie: the condition test bit does *NOT* have
+    * to be set.
+    *
+    * FIXME: check other relevant ops for this situation.
+    */
+   op = shader->GetOpcodeHW(shader, 1);
+   switch (op) {
+   case NV30_VP_INST_OP_BRA:
+      return 1;
+   default:
+      return ((shader->inst[0] & NV30_VP_INST_COND_TEST_ENABLE) ? 1 : 0);
+   }
+}
+
+static nvsCond
+NV30VPGetCondition(nvsFunc * shader)
+{
+   int cond;
+
+   cond = ((shader->inst[0] & NV30_VP_INST_COND_MASK)
+	   >> NV30_VP_INST_COND_SHIFT);
+
+   switch (cond) {
+   case NV30_VP_INST_COND_FL: return NVS_COND_FL;
+   case NV30_VP_INST_COND_LT: return NVS_COND_LT;
+   case NV30_VP_INST_COND_EQ: return NVS_COND_EQ;
+   case NV30_VP_INST_COND_LE: return NVS_COND_LE;
+   case NV30_VP_INST_COND_GT: return NVS_COND_GT;
+   case NV30_VP_INST_COND_NE: return NVS_COND_NE;
+   case NV30_VP_INST_COND_GE: return NVS_COND_GE;
+   case NV30_VP_INST_COND_TR: return NVS_COND_TR;
+   default:
+      return NVS_COND_UNKNOWN;
+   }
+}
+
+static void
+NV30VPGetCondRegSwizzle(nvsFunc * shader, nvsSwzComp *swz)
+{
+   int swzbits;
+
+   swzbits = (shader->inst[0] & NV30_VP_INST_COND_SWZ_ALL_MASK)
+      >> NV30_VP_INST_COND_SWZ_ALL_SHIFT;
+   NV20VPTXSwizzle(swzbits, swz);
+}
+
+static int
+NV30VPGetCondRegID(nvsFunc * shader)
+{
+   return 0;
+}
+
+
+static int
+NV30VPGetBranch(nvsFunc * shader)
+{
+   return ((shader->inst[2] & NV30_VP_INST_IADDR_MASK)
+	   >> NV30_VP_INST_IADDR_SHIFT);
+}
+
+void
+NV30VPInitShaderFuncs(nvsFunc * shader)
+{
+   /* Inherit NV20 code, a lot of it is the same */
+   NV20VPInitShaderFuncs(shader);
+
+   /* Increase max valid opcode ID, and add new instructions */
+   NVVP_TX_VOP_COUNT = NVVP_TX_NVS_OP_COUNT = 32;
+
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_FRC, NVS_OP_FRC,  0, -1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_FLR, NVS_OP_FLR,  0, -1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_SEQ, NVS_OP_SEQ,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_SFL, NVS_OP_SFL,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_SGT, NVS_OP_SGT,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_SLE, NVS_OP_SLE,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_SNE, NVS_OP_SNE,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_STR, NVS_OP_STR,  0,  1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_SSG, NVS_OP_SSG,  0, -1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_ARR, NVS_OP_ARR,  0, -1, -1);
+   MOD_OPCODE(NVVP_TX_VOP, NV30_VP_INST_OP_ARA, NVS_OP_ARA,  3, -1, -1);
+
+   MOD_OPCODE(NVVP_TX_SOP, NV30_VP_INST_OP_BRA, NVS_OP_BRA, -1, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV30_VP_INST_OP_CAL, NVS_OP_CAL, -1, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV30_VP_INST_OP_RET, NVS_OP_RET, -1, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV30_VP_INST_OP_LG2, NVS_OP_LG2,  2, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV30_VP_INST_OP_EX2, NVS_OP_EX2,  2, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV30_VP_INST_OP_SIN, NVS_OP_SIN,  2, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV30_VP_INST_OP_COS, NVS_OP_COS,  2, -1, -1);
+
+   shader->UploadToHW		= NV30VPUploadToHW;
+   shader->UpdateConst		= NV30VPUpdateConst;
+
+   shader->GetOpcodeHW		= NV30VPGetOpcodeHW;
+
+   shader->GetDestFile		= NV30VPGetDestFile;
+   shader->GetDestID		= NV30VPGetDestID;
+   shader->GetDestMask		= NV30VPGetDestMask;
+
+   shader->GetSourceID		= NV30VPGetSourceID;
+   shader->GetSourceAbs		= NV30VPGetSourceAbs;
+
+   shader->GetRelAddressRegID	= NV30VPGetRelAddressRegID;
+   shader->GetRelAddressSwizzle	= NV30VPGetRelAddressSwizzle;
+
+   shader->SupportsConditional	= NV30VPSupportsConditional;
+   shader->GetConditionUpdate	= NV30VPGetConditionUpdate;
+   shader->GetConditionTest	= NV30VPGetConditionTest;
+   shader->GetCondition		= NV30VPGetCondition;
+   shader->GetCondRegSwizzle	= NV30VPGetCondRegSwizzle;
+   shader->GetCondRegID		= NV30VPGetCondRegID;
+
+   shader->GetBranch		= NV30VPGetBranch;
+   shader->SetBranchTarget	= NV30VPSetBranchTarget;
+}
+
diff --git a/src/mesa/drivers/dri/nouveau/nv40_fragprog.c b/src/mesa/drivers/dri/nouveau/nv40_fragprog.c
new file mode 100644
index 00000000000..3e4ae0496e4
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv40_fragprog.c
@@ -0,0 +1,224 @@
+#include "nouveau_shader.h"
+#include "nv40_shader.h"
+
+/* branching ops */
+unsigned int NVFP_TX_BOP_COUNT = 5;
+struct _op_xlat NVFP_TX_BOP[64];
+
+
+/*****************************************************************************
+ * Assembly routines
+ * 	- These extend the NV30 routines, which are almost identical.  NV40
+ * 	  just has branching hacked into the instruction set.
+ */
+static int
+NV40FPSupportsResultScale(nvsFunc *shader, nvsScale scale)
+{
+	switch (scale) {
+	case NVS_SCALE_1X:
+	case NVS_SCALE_2X:
+	case NVS_SCALE_4X:
+	case NVS_SCALE_8X:
+	case NVS_SCALE_INV_2X:
+	case NVS_SCALE_INV_4X:
+	case NVS_SCALE_INV_8X:
+		return 1;
+	default:
+		return 0;
+	}
+}
+
+static void
+NV40FPSetResultScale(nvsFunc *shader, nvsScale scale)
+{
+	shader->inst[2] &= ~NV40_FP_OP_DST_SCALE_MASK;
+	shader->inst[2] |= ((unsigned int)scale << NV40_FP_OP_DST_SCALE_SHIFT);
+}
+
+static void
+NV40FPSetBranchTarget(nvsFunc *shader, int addr)
+{
+	shader->inst[2] &= ~NV40_FP_OP_IADDR_MASK;
+	shader->inst[2] |= (addr << NV40_FP_OP_IADDR_SHIFT);
+}
+
+static void
+NV40FPSetBranchElse(nvsFunc *shader, int addr)
+{
+	shader->inst[2] &= ~NV40_FP_OP_ELSE_ID_MASK;
+	shader->inst[2] |= (addr << NV40_FP_OP_ELSE_ID_SHIFT);
+}
+
+static void
+NV40FPSetBranchEnd(nvsFunc *shader, int addr)
+{
+	shader->inst[3] &= ~NV40_FP_OP_END_ID_MASK;
+	shader->inst[3] |= (addr << NV40_FP_OP_END_ID_SHIFT);
+}
+
+static void
+NV40FPSetLoopParams(nvsFunc *shader, int count, int initial, int increment)
+{
+	shader->inst[2] &= ~(NV40_FP_OP_LOOP_COUNT_MASK |
+			     NV40_FP_OP_LOOP_INDEX_MASK |
+			     NV40_FP_OP_LOOP_INCR_MASK);
+	shader->inst[2] |= ((count     << NV40_FP_OP_LOOP_COUNT_SHIFT) |
+			    (initial   << NV40_FP_OP_LOOP_INDEX_SHIFT) |
+			    (increment << NV40_FP_OP_LOOP_INCR_SHIFT));
+}
+
+/*****************************************************************************
+ * Disassembly routines
+ */
+static struct _op_xlat *
+NV40FPGetOPTXRec(nvsFunc * shader, int merged)
+{
+   struct _op_xlat *opr;
+   int op;
+
+   op = shader->GetOpcodeHW(shader, 0);
+   if (shader->inst[2] & NV40_FP_OP_OPCODE_IS_BRANCH) {
+      opr = NVFP_TX_BOP;
+      op &= ~NV40_FP_OP_OPCODE_IS_BRANCH;
+      if (op > NVFP_TX_BOP_COUNT)
+	 return NULL;
+   }
+   else {
+      opr = NVFP_TX_AOP;
+      if (op > NVFP_TX_AOP_COUNT)
+	 return NULL;
+   }
+
+   if (opr[op].SOP == NVS_OP_UNKNOWN)
+      return NULL;
+   return &opr[op];
+}
+
+static int
+NV40FPGetSourceID(nvsFunc * shader, int merged, int pos)
+{
+   switch (shader->GetSourceFile(shader, merged, pos)) {
+   case NVS_FILE_ATTRIB:
+      switch ((shader->inst[0] & NV40_FP_OP_INPUT_SRC_MASK)
+	      >> NV40_FP_OP_INPUT_SRC_SHIFT) {
+      case NV40_FP_OP_INPUT_SRC_POSITION: return NVS_FR_POSITION;
+      case NV40_FP_OP_INPUT_SRC_COL0    : return NVS_FR_COL0;
+      case NV40_FP_OP_INPUT_SRC_COL1    : return NVS_FR_COL1;
+      case NV40_FP_OP_INPUT_SRC_FOGC    : return NVS_FR_FOGCOORD;
+      case NV40_FP_OP_INPUT_SRC_TC(0)   : return NVS_FR_TEXCOORD0;
+      case NV40_FP_OP_INPUT_SRC_TC(1)   : return NVS_FR_TEXCOORD1;
+      case NV40_FP_OP_INPUT_SRC_TC(2)   : return NVS_FR_TEXCOORD2;
+      case NV40_FP_OP_INPUT_SRC_TC(3)   : return NVS_FR_TEXCOORD3;
+      case NV40_FP_OP_INPUT_SRC_TC(4)   : return NVS_FR_TEXCOORD4;
+      case NV40_FP_OP_INPUT_SRC_TC(5)   : return NVS_FR_TEXCOORD5;
+      case NV40_FP_OP_INPUT_SRC_TC(6)   : return NVS_FR_TEXCOORD6;
+      case NV40_FP_OP_INPUT_SRC_TC(7)   : return NVS_FR_TEXCOORD7;
+      case NV40_FP_OP_INPUT_SRC_FACING  : return NVS_FR_FACING;
+      default:
+	 return -1;
+      }
+      break;
+   case NVS_FILE_TEMP:
+      {
+	 unsigned int src;
+
+	 src = shader->GetSourceHW(shader, merged, pos);
+	 return ((src & NV40_FP_REG_SRC_MASK) >> NV40_FP_REG_SRC_SHIFT);
+      }
+   case NVS_FILE_CONST:		/* inlined into fragprog */
+   default:
+      return -1;
+   }
+}
+
+static int
+NV40FPGetBranch(nvsFunc * shader)
+{
+   return ((shader->inst[2] & NV40_FP_OP_IADDR_MASK)
+	   >> NV40_FP_OP_IADDR_SHIFT);;
+}
+
+static int
+NV40FPGetBranchElse(nvsFunc * shader)
+{
+   return ((shader->inst[2] & NV40_FP_OP_ELSE_ID_MASK)
+	   >> NV40_FP_OP_ELSE_ID_SHIFT);
+}
+
+static int
+NV40FPGetBranchEnd(nvsFunc * shader)
+{
+   return ((shader->inst[3] & NV40_FP_OP_END_ID_MASK)
+	   >> NV40_FP_OP_END_ID_SHIFT);
+}
+
+static int
+NV40FPGetLoopCount(nvsFunc * shader)
+{
+   return ((shader->inst[2] & NV40_FP_OP_LOOP_COUNT_MASK)
+	   >> NV40_FP_OP_LOOP_COUNT_SHIFT);
+}
+
+static int
+NV40FPGetLoopInitial(nvsFunc * shader)
+{
+   return ((shader->inst[2] & NV40_FP_OP_LOOP_INDEX_MASK)
+	   >> NV40_FP_OP_LOOP_INDEX_SHIFT);
+}
+
+static int
+NV40FPGetLoopIncrement(nvsFunc * shader)
+{
+   return ((shader->inst[2] & NV40_FP_OP_LOOP_INCR_MASK)
+	   >> NV40_FP_OP_LOOP_INCR_SHIFT);
+}
+
+void
+NV40FPInitShaderFuncs(nvsFunc * shader)
+{
+   /* Inherit NV30 FP code, it's mostly the same */
+   NV30FPInitShaderFuncs(shader);
+
+   /* Kill off opcodes seen on NV30, but not seen on NV40 - need to find
+    * out if these actually work or not.
+    *
+    * update: either LIT/RSQ don't work on nv40, or I generate bad code for
+    *         them.  haven't tested the others yet
+    */
+   MOD_OPCODE(NVFP_TX_AOP, 0x1B, NVS_OP_UNKNOWN, -1, -1, -1);	/* NV30 RSQ */
+   MOD_OPCODE(NVFP_TX_AOP, 0x1E, NVS_OP_UNKNOWN, -1, -1, -1);	/* NV30 LIT */
+   MOD_OPCODE(NVFP_TX_AOP, 0x1F, NVS_OP_UNKNOWN, -1, -1, -1);	/* NV30 LRP */
+   MOD_OPCODE(NVFP_TX_AOP, 0x26, NVS_OP_UNKNOWN, -1, -1, -1);	/* NV30 POW */
+   MOD_OPCODE(NVFP_TX_AOP, 0x36, NVS_OP_UNKNOWN, -1, -1, -1);	/* NV30 RFL */
+
+   /* Extra opcodes supported on NV40 */
+   MOD_OPCODE(NVFP_TX_AOP, NV40_FP_OP_OPCODE_DIV     , NVS_OP_DIV ,  0,  1, -1);
+   MOD_OPCODE(NVFP_TX_AOP, NV40_FP_OP_OPCODE_DP2A    , NVS_OP_DP2A,  0,  1,  2);
+   MOD_OPCODE(NVFP_TX_AOP, NV40_FP_OP_OPCODE_TXL     , NVS_OP_TXL ,  0, -1, -1);
+
+   MOD_OPCODE(NVFP_TX_BOP, NV40_FP_OP_BRA_OPCODE_BRK , NVS_OP_BRK , -1, -1, -1);
+   MOD_OPCODE(NVFP_TX_BOP, NV40_FP_OP_BRA_OPCODE_CAL , NVS_OP_CAL , -1, -1, -1);
+   MOD_OPCODE(NVFP_TX_BOP, NV40_FP_OP_BRA_OPCODE_IF  , NVS_OP_IF  , -1, -1, -1);
+   MOD_OPCODE(NVFP_TX_BOP, NV40_FP_OP_BRA_OPCODE_LOOP, NVS_OP_LOOP, -1, -1, -1);
+   MOD_OPCODE(NVFP_TX_BOP, NV40_FP_OP_BRA_OPCODE_REP , NVS_OP_REP , -1, -1, -1);
+   MOD_OPCODE(NVFP_TX_BOP, NV40_FP_OP_BRA_OPCODE_RET , NVS_OP_RET , -1, -1, -1);
+
+   shader->SupportsResultScale	= NV40FPSupportsResultScale;
+   shader->SetResultScale	= NV40FPSetResultScale;
+
+   /* fragment.facing */
+   shader->GetSourceID		= NV40FPGetSourceID;
+
+   /* branching */
+   shader->GetOPTXRec		= NV40FPGetOPTXRec;
+   shader->GetBranch		= NV40FPGetBranch;
+   shader->GetBranchElse	= NV40FPGetBranchElse;
+   shader->GetBranchEnd		= NV40FPGetBranchEnd;
+   shader->GetLoopCount		= NV40FPGetLoopCount;
+   shader->GetLoopInitial	= NV40FPGetLoopInitial;
+   shader->GetLoopIncrement	= NV40FPGetLoopIncrement;
+   shader->SetBranchTarget	= NV40FPSetBranchTarget;
+   shader->SetBranchElse	= NV40FPSetBranchElse;
+   shader->SetBranchEnd		= NV40FPSetBranchEnd;
+   shader->SetLoopParams	= NV40FPSetLoopParams;
+}
diff --git a/src/mesa/drivers/dri/nouveau/nv40_shader.h b/src/mesa/drivers/dri/nouveau/nv40_shader.h
new file mode 100644
index 00000000000..584f4c23e08
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv40_shader.h
@@ -0,0 +1,467 @@
+#ifndef __NV40_SHADER_H__
+#define __NV40_SHADER_H__
+
+/* Vertex programs instruction set
+ *
+ * The NV40 instruction set is very similar to NV30.  Most fields are in
+ * a slightly different position in the instruction however.
+ *
+ * Merged instructions
+ *     In some cases it is possible to put two instructions into one opcode
+ *     slot.  The rules for when this is OK is not entirely clear to me yet.
+ *
+ *     There are separate writemasks and dest temp register fields for each
+ *     grouping of instructions.  There is however only one field with the
+ *     ID of a result register.  Writing to temp/result regs is selected by
+ *     setting VEC_RESULT/SCA_RESULT.
+ *
+ * Temporary registers
+ *     The source/dest temp register fields have been extended by 1 bit, to
+ *     give a total of 32 temporary registers.
+ *
+ * Relative Addressing
+ *     NV40 can use an address register to index into vertex attribute regs.
+ *     This is done by putting the offset value into INPUT_SRC and setting
+ *     the INDEX_INPUT flag.
+ *
+ * Conditional execution (see NV_vertex_program{2,3} for details)
+ *     There is a second condition code register on NV40, it's use is enabled
+ *     by setting the COND_REG_SELECT_1 flag.
+ *
+ * Texture lookup
+ *     TODO
+ */
+
+/* ---- OPCODE BITS 127:96 / data DWORD 0 --- */
+#define NV40_VP_INST_VEC_RESULT                                        (1 << 30)
+/* uncertain.. */
+#define NV40_VP_INST_COND_UPDATE_ENABLE                        ((1 << 14)|1<<29)
+/* use address reg as index into attribs */
+#define NV40_VP_INST_INDEX_INPUT                                       (1 << 27)
+#define NV40_VP_INST_COND_REG_SELECT_1                                 (1 << 25)
+#define NV40_VP_INST_ADDR_REG_SELECT_1                                 (1 << 24)
+#define NV40_VP_INST_SRC2_ABS                                          (1 << 23)
+#define NV40_VP_INST_SRC1_ABS                                          (1 << 22)
+#define NV40_VP_INST_SRC0_ABS                                          (1 << 21)
+#define NV40_VP_INST_VEC_DEST_TEMP_SHIFT                                      15
+#define NV40_VP_INST_VEC_DEST_TEMP_MASK                             (0x1F << 15)
+#define NV40_VP_INST_COND_TEST_ENABLE                                  (1 << 13)
+#define NV40_VP_INST_COND_SHIFT                                               10
+#define NV40_VP_INST_COND_MASK                                       (0x7 << 10)
+#    define NV40_VP_INST_COND_FL                                               0
+#    define NV40_VP_INST_COND_LT                                               1
+#    define NV40_VP_INST_COND_EQ                                               2
+#    define NV40_VP_INST_COND_LE                                               3
+#    define NV40_VP_INST_COND_GT                                               4
+#    define NV40_VP_INST_COND_NE                                               5
+#    define NV40_VP_INST_COND_GE                                               6
+#    define NV40_VP_INST_COND_TR                                               7
+#define NV40_VP_INST_COND_SWZ_X_SHIFT                                          8
+#define NV40_VP_INST_COND_SWZ_X_MASK                                    (3 << 8)
+#define NV40_VP_INST_COND_SWZ_Y_SHIFT                                          6
+#define NV40_VP_INST_COND_SWZ_Y_MASK                                    (3 << 6)
+#define NV40_VP_INST_COND_SWZ_Z_SHIFT                                          4
+#define NV40_VP_INST_COND_SWZ_Z_MASK                                    (3 << 4)
+#define NV40_VP_INST_COND_SWZ_W_SHIFT                                          2
+#define NV40_VP_INST_COND_SWZ_W_MASK                                    (3 << 2)
+#define NV40_VP_INST_COND_SWZ_ALL_SHIFT                                        2
+#define NV40_VP_INST_COND_SWZ_ALL_MASK                               (0xFF << 2)
+#define NV40_VP_INST_ADDR_SWZ_SHIFT                                            0
+#define NV40_VP_INST_ADDR_SWZ_MASK                                   (0x03 << 0)
+#define NV40_VP_INST0_KNOWN ( \
+                NV40_VP_INST_INDEX_INPUT | \
+                NV40_VP_INST_COND_REG_SELECT_1 | \
+                NV40_VP_INST_ADDR_REG_SELECT_1 | \
+                NV40_VP_INST_SRC2_ABS | \
+                NV40_VP_INST_SRC1_ABS | \
+                NV40_VP_INST_SRC0_ABS | \
+                NV40_VP_INST_VEC_DEST_TEMP_MASK | \
+                NV40_VP_INST_COND_TEST_ENABLE | \
+                NV40_VP_INST_COND_MASK | \
+                NV40_VP_INST_COND_SWZ_ALL_MASK | \
+                NV40_VP_INST_ADDR_SWZ_MASK)
+
+/* ---- OPCODE BITS 95:64 / data DWORD 1 --- */
+#define NV40_VP_INST_VEC_OPCODE_SHIFT                                         22
+#define NV40_VP_INST_VEC_OPCODE_MASK                                (0x1F << 22)
+#    define NV40_VP_INST_OP_NOP                                             0x00
+#    define NV40_VP_INST_OP_MOV                                             0x01
+#    define NV40_VP_INST_OP_MUL                                             0x02
+#    define NV40_VP_INST_OP_ADD                                             0x03
+#    define NV40_VP_INST_OP_MAD                                             0x04
+#    define NV40_VP_INST_OP_DP3                                             0x05
+#    define NV40_VP_INST_OP_DP4                                             0x07
+#    define NV40_VP_INST_OP_DPH                                             0x06
+#    define NV40_VP_INST_OP_DST                                             0x08
+#    define NV40_VP_INST_OP_MIN                                             0x09
+#    define NV40_VP_INST_OP_MAX                                             0x0A
+#    define NV40_VP_INST_OP_SLT                                             0x0B
+#    define NV40_VP_INST_OP_SGE                                             0x0C
+#    define NV40_VP_INST_OP_ARL                                             0x0D
+#    define NV40_VP_INST_OP_FRC                                             0x0E
+#    define NV40_VP_INST_OP_FLR                                             0x0F
+#    define NV40_VP_INST_OP_SEQ                                             0x10
+#    define NV40_VP_INST_OP_SFL                                             0x11
+#    define NV40_VP_INST_OP_SGT                                             0x12
+#    define NV40_VP_INST_OP_SLE                                             0x13
+#    define NV40_VP_INST_OP_SNE                                             0x14
+#    define NV40_VP_INST_OP_STR                                             0x15
+#    define NV40_VP_INST_OP_SSG                                             0x16
+#    define NV40_VP_INST_OP_ARR                                             0x17
+#    define NV40_VP_INST_OP_ARA                                             0x18
+#    define NV40_VP_INST_OP_TXWHAT                                          0x19
+#define NV40_VP_INST_SCA_OPCODE_SHIFT                                         27
+#define NV40_VP_INST_SCA_OPCODE_MASK                                (0x1F << 27)
+#    define NV40_VP_INST_OP_RCP                                             0x02
+#    define NV40_VP_INST_OP_RCC                                             0x03
+#    define NV40_VP_INST_OP_RSQ                                             0x04
+#    define NV40_VP_INST_OP_EXP                                             0x05
+#    define NV40_VP_INST_OP_LOG                                             0x06
+#    define NV40_VP_INST_OP_LIT                                             0x07
+#    define NV40_VP_INST_OP_BRA                                             0x09
+#    define NV40_VP_INST_OP_CAL                                             0x0B
+#    define NV40_VP_INST_OP_RET                                             0x0C
+#    define NV40_VP_INST_OP_LG2                                             0x0D
+#    define NV40_VP_INST_OP_EX2                                             0x0E
+#    define NV40_VP_INST_OP_SIN                                             0x0F
+#    define NV40_VP_INST_OP_COS                                             0x10
+#    define NV40_VP_INST_OP_PUSHA                                           0x13
+#    define NV40_VP_INST_OP_POPA                                            0x14
+#define NV40_VP_INST_CONST_SRC_SHIFT                                          12
+#define NV40_VP_INST_CONST_SRC_MASK                                 (0xFF << 12)
+#define NV40_VP_INST_INPUT_SRC_SHIFT                                           8
+#define NV40_VP_INST_INPUT_SRC_MASK                                  (0x0F << 8)
+#    define NV40_VP_INST_IN_POS                                                0
+#    define NV40_VP_INST_IN_WEIGHT                                             1
+#    define NV40_VP_INST_IN_NORMAL                                             2
+#    define NV40_VP_INST_IN_COL0                                               3
+#    define NV40_VP_INST_IN_COL1                                               4
+#    define NV40_VP_INST_IN_FOGC                                               5
+#    define NV40_VP_INST_IN_TC0                                                8
+#    define NV40_VP_INST_IN_TC(n)                                          (8+n)
+#define NV40_VP_INST_SRC0H_SHIFT                                               0
+#define NV40_VP_INST_SRC0H_MASK                                      (0xFF << 0)
+#define NV40_VP_INST1_KNOWN ( \
+                NV40_VP_INST_VEC_OPCODE_MASK | \
+                NV40_VP_INST_SCA_OPCODE_MASK | \
+                NV40_VP_INST_CONST_SRC_MASK  | \
+                NV40_VP_INST_INPUT_SRC_MASK  | \
+                NV40_VP_INST_SRC0H_MASK \
+                )
+
+/* ---- OPCODE BITS 63:32 / data DWORD 2 --- */
+#define NV40_VP_INST_SRC0L_SHIFT                                              23
+#define NV40_VP_INST_SRC0L_MASK                                    (0x1FF << 23)
+#define NV40_VP_INST_SRC1_SHIFT                                                6
+#define NV40_VP_INST_SRC1_MASK                                    (0x1FFFF << 6)
+#define NV40_VP_INST_SRC2H_SHIFT                                               0
+#define NV40_VP_INST_SRC2H_MASK                                      (0x3F << 0)
+#define NV40_VP_INST_IADDRH_SHIFT                                              0
+#define NV40_VP_INST_IADDRH_MASK                                     (0x1F << 0)
+
+/* ---- OPCODE BITS 31:0 / data DWORD 3 --- */
+#define NV40_VP_INST_IADDRL_SHIFT                                             29
+#define NV40_VP_INST_IADDRL_MASK                                       (7 << 29)
+#define NV40_VP_INST_SRC2L_SHIFT                                              21
+#define NV40_VP_INST_SRC2L_MASK                                    (0x7FF << 21)
+#define NV40_VP_INST_SCA_WRITEMASK_SHIFT                                      17
+#define NV40_VP_INST_SCA_WRITEMASK_MASK                              (0xF << 17)
+#    define NV40_VP_INST_SCA_WRITEMASK_X                               (1 << 20)
+#    define NV40_VP_INST_SCA_WRITEMASK_Y                               (1 << 19)
+#    define NV40_VP_INST_SCA_WRITEMASK_Z                               (1 << 18)
+#    define NV40_VP_INST_SCA_WRITEMASK_W                               (1 << 17)
+#define NV40_VP_INST_VEC_WRITEMASK_SHIFT                                      13
+#define NV40_VP_INST_VEC_WRITEMASK_MASK                              (0xF << 13)
+#    define NV40_VP_INST_VEC_WRITEMASK_X                               (1 << 16)
+#    define NV40_VP_INST_VEC_WRITEMASK_Y                               (1 << 15)
+#    define NV40_VP_INST_VEC_WRITEMASK_Z                               (1 << 14)
+#    define NV40_VP_INST_VEC_WRITEMASK_W                               (1 << 13)
+#define NV40_VP_INST_SCA_RESULT                                        (1 << 12)
+#define NV40_VP_INST_SCA_DEST_TEMP_SHIFT                                       7
+#define NV40_VP_INST_SCA_DEST_TEMP_MASK                              (0x1F << 7)
+#define NV40_VP_INST_DEST_SHIFT                                                2
+#define NV40_VP_INST_DEST_MASK                                         (31 << 2)
+#    define NV40_VP_INST_DEST_POS                                              0
+#    define NV40_VP_INST_DEST_COL0                                             1
+#    define NV40_VP_INST_DEST_COL1                                             2
+#    define NV40_VP_INST_DEST_BFC0                                             3
+#    define NV40_VP_INST_DEST_BFC1                                             4
+#    define NV40_VP_INST_DEST_FOGC                                             5
+#    define NV40_VP_INST_DEST_PSZ                                              6
+#    define NV40_VP_INST_DEST_TC0                                              7
+#    define NV40_VP_INST_DEST_TC(n)                                        (7+n)
+#    define NV40_VP_INST_DEST_TEMP                                          0x1F
+#define NV40_VP_INST_INDEX_CONST                                        (1 << 1)
+#define NV40_VP_INST_LAST                                               (1 << 0)
+#define NV40_VP_INST3_KNOWN ( \
+                NV40_VP_INST_SRC2L_MASK |\
+                NV40_VP_INST_SCA_WRITEMASK_MASK |\
+                NV40_VP_INST_VEC_WRITEMASK_MASK |\
+                NV40_VP_INST_SCA_DEST_TEMP_MASK |\
+                NV40_VP_INST_DEST_MASK |\
+                NV40_VP_INST_INDEX_CONST)
+
+/* Useful to split the source selection regs into their pieces */
+#define NV40_VP_SRC0_HIGH_SHIFT                                                9
+#define NV40_VP_SRC0_HIGH_MASK                                        0x0001FE00
+#define NV40_VP_SRC0_LOW_MASK                                         0x000001FF
+#define NV40_VP_SRC2_HIGH_SHIFT                                               11
+#define NV40_VP_SRC2_HIGH_MASK                                        0x0001F800
+#define NV40_VP_SRC2_LOW_MASK                                         0x000007FF
+
+/* Source selection - these are the bits you fill NV40_VP_INST_SRCn with */
+#define NV40_VP_SRC_NEGATE                                             (1 << 16)
+#define NV40_VP_SRC_SWZ_X_SHIFT                                               14
+#define NV40_VP_SRC_SWZ_X_MASK                                         (3 << 14)
+#define NV40_VP_SRC_SWZ_Y_SHIFT                                               12
+#define NV40_VP_SRC_SWZ_Y_MASK                                         (3 << 12)
+#define NV40_VP_SRC_SWZ_Z_SHIFT                                               10
+#define NV40_VP_SRC_SWZ_Z_MASK                                         (3 << 10)
+#define NV40_VP_SRC_SWZ_W_SHIFT                                                8
+#define NV40_VP_SRC_SWZ_W_MASK                                          (3 << 8)
+#define NV40_VP_SRC_SWZ_ALL_SHIFT                                              8
+#define NV40_VP_SRC_SWZ_ALL_MASK                                     (0xFF << 8)
+#define NV40_VP_SRC_TEMP_SRC_SHIFT                                             2
+#define NV40_VP_SRC_TEMP_SRC_MASK                                    (0x1F << 2)
+#define NV40_VP_SRC_REG_TYPE_SHIFT                                             0
+#define NV40_VP_SRC_REG_TYPE_MASK                                       (3 << 0)
+#    define NV40_VP_SRC_REG_TYPE_UNK0                                          0
+#    define NV40_VP_SRC_REG_TYPE_TEMP                                          1
+#    define NV40_VP_SRC_REG_TYPE_INPUT                                         2
+#    define NV40_VP_SRC_REG_TYPE_CONST                                         3
+
+
+/*
+ * Each fragment program opcode appears to be comprised of 4 32-bit values.
+ *
+ *         0 - Opcode, output reg/mask, ATTRIB source
+ *         1 - Source 0
+ *         2 - Source 1
+ *         3 - Source 2
+ *
+ * There appears to be no special difference between result regs and temp regs.
+ *                 result.color == R0.xyzw
+ *                 result.depth == R1.z
+ * When the fragprog contains instructions to write depth,
+ * NV30_TCL_PRIMITIVE_3D_UNK1D78=0 otherwise it is set to 1.
+ *
+ * Constants are inserted directly after the instruction that uses them.
+ * 
+ * It appears that it's not possible to use two input registers in one
+ * instruction as the input sourcing is done in the instruction dword
+ * and not the source selection dwords.  As such instructions such as:
+ * 
+ *                 ADD result.color, fragment.color, fragment.texcoord[0];
+ *
+ * must be split into two MOV's and then an ADD (nvidia does this) but
+ * I'm not sure why it's not just one MOV and then source the second input
+ * in the ADD instruction..
+ *
+ * Negation of the full source is done with NV30_FP_REG_NEGATE, arbitrary
+ * negation requires multiplication with a const.
+ *
+ * Arbitrary swizzling is supported with the exception of SWIZZLE_ZERO and
+ * SWIZZLE_ONE.
+ *
+ * The temp/result regs appear to be initialised to (0.0, 0.0, 0.0, 0.0) as
+ * SWIZZLE_ZERO is implemented simply by not writing to the relevant components
+ * of the destination.
+ *
+ * Looping
+ *   Loops appear to be fairly expensive on NV40 at least, the proprietary
+ *   driver goes to a lot of effort to avoid using the native looping
+ *   instructions.  If the total number of *executed* instructions between
+ *   REP/ENDREP or LOOP/ENDLOOP is <=500, the driver will unroll the loop.
+ *   The maximum loop count is 255.
+ *
+ * Conditional execution
+ *   TODO
+ * 
+ * Non-native instructions:
+ *         LIT
+ *         LRP - MAD+MAD
+ *         SUB - ADD, negate second source
+ *         RSQ - LG2 + EX2
+ *         POW - LG2 + MUL + EX2
+ *         SCS - COS + SIN
+ *         XPD
+ *         DP2 - MUL + ADD
+ *         NRM
+ */
+
+//== Opcode / Destination selection ==
+#define NV40_FP_OP_PROGRAM_END                                          (1 << 0)
+#define NV40_FP_OP_OUT_REG_SHIFT                                               1
+#define NV40_FP_OP_OUT_REG_MASK                                        (31 << 1)
+/* Needs to be set when writing outputs to get expected result.. */
+#define NV40_FP_OP_UNK0_7                                               (1 << 7)
+#define NV40_FP_OP_COND_WRITE_ENABLE                                    (1 << 8)
+#define NV40_FP_OP_OUTMASK_SHIFT                                               9
+#define NV40_FP_OP_OUTMASK_MASK                                       (0xF << 9)
+#    define NV40_FP_OP_OUT_X                                            (1 << 9)
+#    define NV40_FP_OP_OUT_Y                                            (1 <<10)
+#    define NV40_FP_OP_OUT_Z                                            (1 <<11)
+#    define NV40_FP_OP_OUT_W                                            (1 <<12)
+/* Uncertain about these, especially the input_src values.. it's possible that
+ * they can be dynamically changed.
+ */
+#define NV40_FP_OP_INPUT_SRC_SHIFT                                            13
+#define NV40_FP_OP_INPUT_SRC_MASK                                     (15 << 13)
+#    define NV40_FP_OP_INPUT_SRC_POSITION                                    0x0
+#    define NV40_FP_OP_INPUT_SRC_COL0                                        0x1
+#    define NV40_FP_OP_INPUT_SRC_COL1                                        0x2
+#    define NV40_FP_OP_INPUT_SRC_FOGC                                        0x3
+#    define NV40_FP_OP_INPUT_SRC_TC0                                         0x4
+#    define NV40_FP_OP_INPUT_SRC_TC(n)                                 (0x4 + n)
+#    define NV40_FP_OP_INPUT_SRC_FACING                                      0xE
+#define NV40_FP_OP_TEX_UNIT_SHIFT                                             17
+#define NV40_FP_OP_TEX_UNIT_MASK                                     (0xF << 17)
+#define NV40_FP_OP_PRECISION_SHIFT                                            22
+#define NV40_FP_OP_PRECISION_MASK                                      (3 << 22)
+#   define NV40_FP_PRECISION_FP32                                              0
+#   define NV40_FP_PRECISION_FP16                                              1
+#   define NV40_FP_PRECISION_FX12                                              2
+#define NV40_FP_OP_OPCODE_SHIFT                                               24
+#define NV40_FP_OP_OPCODE_MASK                                      (0x3F << 24)
+#        define NV40_FP_OP_OPCODE_NOP                                       0x00
+#        define NV40_FP_OP_OPCODE_MOV                                       0x01
+#        define NV40_FP_OP_OPCODE_MUL                                       0x02
+#        define NV40_FP_OP_OPCODE_ADD                                       0x03
+#        define NV40_FP_OP_OPCODE_MAD                                       0x04
+#        define NV40_FP_OP_OPCODE_DP3                                       0x05
+#        define NV40_FP_OP_OPCODE_DP4                                       0x06
+#        define NV40_FP_OP_OPCODE_DST                                       0x07
+#        define NV40_FP_OP_OPCODE_MIN                                       0x08
+#        define NV40_FP_OP_OPCODE_MAX                                       0x09
+#        define NV40_FP_OP_OPCODE_SLT                                       0x0A
+#        define NV40_FP_OP_OPCODE_SGE                                       0x0B
+#        define NV40_FP_OP_OPCODE_SLE                                       0x0C
+#        define NV40_FP_OP_OPCODE_SGT                                       0x0D
+#        define NV40_FP_OP_OPCODE_SNE                                       0x0E
+#        define NV40_FP_OP_OPCODE_SEQ                                       0x0F
+#        define NV40_FP_OP_OPCODE_FRC                                       0x10
+#        define NV40_FP_OP_OPCODE_FLR                                       0x11
+#        define NV40_FP_OP_OPCODE_KIL                                       0x12
+#        define NV40_FP_OP_OPCODE_PK4B                                      0x13
+#        define NV40_FP_OP_OPCODE_UP4B                                      0x14
+/* DDX/DDY can only write to XY */
+#        define NV40_FP_OP_OPCODE_DDX                                       0x15
+#        define NV40_FP_OP_OPCODE_DDY                                       0x16
+#        define NV40_FP_OP_OPCODE_TEX                                       0x17
+#        define NV40_FP_OP_OPCODE_TXP                                       0x18
+#        define NV40_FP_OP_OPCODE_TXD                                       0x19
+#        define NV40_FP_OP_OPCODE_RCP                                       0x1A
+#        define NV40_FP_OP_OPCODE_EX2                                       0x1C
+#        define NV40_FP_OP_OPCODE_LG2                                       0x1D
+#        define NV40_FP_OP_OPCODE_COS                                       0x22
+#        define NV40_FP_OP_OPCODE_SIN                                       0x23
+#        define NV40_FP_OP_OPCODE_PK2H                                      0x24
+#        define NV40_FP_OP_OPCODE_UP2H                                      0x25
+#        define NV40_FP_OP_OPCODE_PK4UB                                     0x27
+#        define NV40_FP_OP_OPCODE_UP4UB                                     0x28
+#        define NV40_FP_OP_OPCODE_PK2US                                     0x29
+#        define NV40_FP_OP_OPCODE_UP2US                                     0x2A
+#        define NV40_FP_OP_OPCODE_DP2A                                      0x2E
+#        define NV40_FP_OP_OPCODE_TXL                                       0x2F
+#        define NV40_FP_OP_OPCODE_TXB                                       0x31
+#        define NV40_FP_OP_OPCODE_DIV                                       0x3A
+/* The use of these instructions appears to be indicated by bit 31 of DWORD 2.*/
+#        define NV40_FP_OP_BRA_OPCODE_BRK                                    0x0
+#        define NV40_FP_OP_BRA_OPCODE_CAL                                    0x1
+#        define NV40_FP_OP_BRA_OPCODE_IF                                     0x2
+#        define NV40_FP_OP_BRA_OPCODE_LOOP                                   0x3
+#        define NV40_FP_OP_BRA_OPCODE_REP                                    0x4
+#        define NV40_FP_OP_BRA_OPCODE_RET                                    0x5
+#define NV40_FP_OP_OUT_SAT                                             (1 << 31)
+
+/* high order bits of SRC0 */
+#define NV40_FP_OP_OUT_ABS                                             (1 << 29)
+#define NV40_FP_OP_COND_SWZ_W_SHIFT                                           27
+#define NV40_FP_OP_COND_SWZ_W_MASK                                     (3 << 27)
+#define NV40_FP_OP_COND_SWZ_Z_SHIFT                                           25
+#define NV40_FP_OP_COND_SWZ_Z_MASK                                     (3 << 25)
+#define NV40_FP_OP_COND_SWZ_Y_SHIFT                                           23
+#define NV40_FP_OP_COND_SWZ_Y_MASK                                     (3 << 23)
+#define NV40_FP_OP_COND_SWZ_X_SHIFT                                           21
+#define NV40_FP_OP_COND_SWZ_X_MASK                                     (3 << 21)
+#define NV40_FP_OP_COND_SWZ_ALL_SHIFT                                         21
+#define NV40_FP_OP_COND_SWZ_ALL_MASK                                (0xFF << 21)
+#define NV40_FP_OP_COND_SHIFT                                                 18
+#define NV40_FP_OP_COND_MASK                                        (0x07 << 18)
+#        define NV40_FP_OP_COND_FL                                             0
+#        define NV40_FP_OP_COND_LT                                             1
+#        define NV40_FP_OP_COND_EQ                                             2
+#        define NV40_FP_OP_COND_LE                                             3
+#        define NV40_FP_OP_COND_GT                                             4
+#        define NV40_FP_OP_COND_NE                                             5
+#        define NV40_FP_OP_COND_GE                                             6
+#        define NV40_FP_OP_COND_TR                                             7
+
+/* high order bits of SRC1 */
+#define NV40_FP_OP_OPCODE_IS_BRANCH                                      (1<<31)
+#define NV40_FP_OP_DST_SCALE_SHIFT                                            28
+#define NV40_FP_OP_DST_SCALE_MASK                                      (3 << 28)
+
+/* SRC1 LOOP */
+#define NV40_FP_OP_LOOP_INCR_SHIFT                                            19
+#define NV40_FP_OP_LOOP_INCR_MASK                                   (0xFF << 19)
+#define NV40_FP_OP_LOOP_INDEX_SHIFT                                           10
+#define NV40_FP_OP_LOOP_INDEX_MASK                                  (0xFF << 10)
+#define NV40_FP_OP_LOOP_COUNT_SHIFT                                            2
+#define NV40_FP_OP_LOOP_COUNT_MASK                                   (0xFF << 2)
+
+/* SRC1 IF */
+#define NV40_FP_OP_ELSE_ID_SHIFT                                               2
+#define NV40_FP_OP_ELSE_ID_MASK                                      (0xFF << 2)
+
+/* SRC1 CAL */
+#define NV40_FP_OP_IADDR_SHIFT                                                 2
+#define NV40_FP_OP_IADDR_MASK                                        (0xFF << 2)
+
+/* SRC1 REP
+ *   I have no idea why there are 3 count values here..  but they
+ *   have always been filled with the same value in my tests so
+ *   far..
+ */
+#define NV40_FP_OP_REP_COUNT1_SHIFT                                            2
+#define NV40_FP_OP_REP_COUNT1_MASK                                   (0xFF << 2)
+#define NV40_FP_OP_REP_COUNT2_SHIFT                                           10
+#define NV40_FP_OP_REP_COUNT2_MASK                                  (0xFF << 10)
+#define NV40_FP_OP_REP_COUNT3_SHIFT                                           19
+#define NV40_FP_OP_REP_COUNT3_MASK                                  (0xFF << 19)
+
+/* SRC2 REP/IF */
+#define NV40_FP_OP_END_ID_SHIFT                                                2
+#define NV40_FP_OP_END_ID_MASK                                       (0xFF << 2)
+
+// SRC2 high-order
+#define NV40_FP_OP_INDEX_INPUT                                         (1 << 30)
+#define NV40_FP_OP_ADDR_INDEX_SHIFT                                           19
+#define NV40_FP_OP_ADDR_INDEX_MASK                                   (0xF << 19)
+
+//== Register selection ==
+#define NV40_FP_REG_TYPE_SHIFT                                                 0
+#define NV40_FP_REG_TYPE_MASK                                           (3 << 0)
+#        define NV40_FP_REG_TYPE_TEMP                                          0
+#        define NV40_FP_REG_TYPE_INPUT                                         1
+#        define NV40_FP_REG_TYPE_CONST                                         2
+#define NV40_FP_REG_SRC_SHIFT                                                  2
+#define NV40_FP_REG_SRC_MASK                                           (31 << 2)
+#define NV40_FP_REG_UNK_0                                               (1 << 8)
+#define NV40_FP_REG_SWZ_ALL_SHIFT                                              9
+#define NV40_FP_REG_SWZ_ALL_MASK                                      (255 << 9)
+#define NV40_FP_REG_SWZ_X_SHIFT                                                9
+#define NV40_FP_REG_SWZ_X_MASK                                          (3 << 9)
+#define NV40_FP_REG_SWZ_Y_SHIFT                                               11
+#define NV40_FP_REG_SWZ_Y_MASK                                         (3 << 11)
+#define NV40_FP_REG_SWZ_Z_SHIFT                                               13
+#define NV40_FP_REG_SWZ_Z_MASK                                         (3 << 13)
+#define NV40_FP_REG_SWZ_W_SHIFT                                               15
+#define NV40_FP_REG_SWZ_W_MASK                                         (3 << 15)
+#        define NV40_FP_SWIZZLE_X                                              0
+#        define NV40_FP_SWIZZLE_Y                                              1
+#        define NV40_FP_SWIZZLE_Z                                              2
+#        define NV40_FP_SWIZZLE_W                                              3
+#define NV40_FP_REG_NEGATE                                             (1 << 17)
+
+#endif
diff --git a/src/mesa/drivers/dri/nouveau/nv40_vertprog.c b/src/mesa/drivers/dri/nouveau/nv40_vertprog.c
new file mode 100644
index 00000000000..d054140bcd0
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv40_vertprog.c
@@ -0,0 +1,778 @@
+#include "nouveau_shader.h"
+#include "nouveau_msg.h"
+#include "nv40_shader.h"
+
+/*****************************************************************************
+ * Assembly routines
+ */
+static int
+NV40VPSupportsOpcode(nvsFunc * shader, nvsOpcode op)
+{
+   if (shader->GetOPTXFromSOP(op, NULL))
+      return 1;
+   return 0;
+}
+
+static void
+NV40VPSetOpcode(nvsFunc *shader, unsigned int opcode, int slot)
+{
+   if (slot) {
+      shader->inst[1] &= ~NV40_VP_INST_SCA_OPCODE_MASK;
+      shader->inst[1] |= (opcode << NV40_VP_INST_SCA_OPCODE_SHIFT);
+   } else {
+      shader->inst[1] &= ~NV40_VP_INST_VEC_OPCODE_MASK;
+      shader->inst[1] |= (opcode << NV40_VP_INST_VEC_OPCODE_SHIFT);
+   }
+}
+
+static void
+NV40VPSetCCUpdate(nvsFunc *shader)
+{
+   shader->inst[0] |= NV40_VP_INST_COND_UPDATE_ENABLE;
+}
+
+static void
+NV40VPSetCondition(nvsFunc *shader, int on, nvsCond cond, int reg,
+      		   nvsSwzComp *swizzle)
+{
+   unsigned int hwcond;
+
+   if (on ) shader->inst[0] |= NV40_VP_INST_COND_TEST_ENABLE;
+   else     shader->inst[0] &= ~NV40_VP_INST_COND_TEST_ENABLE;
+   if (reg) shader->inst[0] |= NV40_VP_INST_COND_REG_SELECT_1;
+   else     shader->inst[0] &= ~NV40_VP_INST_COND_REG_SELECT_1;
+
+   switch (cond) {
+   case NVS_COND_TR: hwcond = NV40_VP_INST_COND_TR; break;
+   case NVS_COND_FL: hwcond = NV40_VP_INST_COND_FL; break;
+   case NVS_COND_LT: hwcond = NV40_VP_INST_COND_LT; break;
+   case NVS_COND_GT: hwcond = NV40_VP_INST_COND_GT; break;
+   case NVS_COND_NE: hwcond = NV40_VP_INST_COND_NE; break;
+   case NVS_COND_EQ: hwcond = NV40_VP_INST_COND_EQ; break;
+   case NVS_COND_GE: hwcond = NV40_VP_INST_COND_GE; break;
+   case NVS_COND_LE: hwcond = NV40_VP_INST_COND_LE; break;
+   default:
+	WARN_ONCE("unknown vp cond %d\n", cond);
+	hwcond = NV40_VP_INST_COND_TR;
+	break;
+   }
+   shader->inst[0] &= ~NV40_VP_INST_COND_MASK;
+   shader->inst[0] |= (hwcond << NV40_VP_INST_COND_SHIFT);
+
+   shader->inst[0] &= ~NV40_VP_INST_COND_SWZ_ALL_MASK;
+   shader->inst[0] |= (swizzle[NVS_SWZ_X] << NV40_VP_INST_COND_SWZ_X_SHIFT);
+   shader->inst[0] |= (swizzle[NVS_SWZ_Y] << NV40_VP_INST_COND_SWZ_Y_SHIFT);
+   shader->inst[0] |= (swizzle[NVS_SWZ_Z] << NV40_VP_INST_COND_SWZ_Z_SHIFT);
+   shader->inst[0] |= (swizzle[NVS_SWZ_W] << NV40_VP_INST_COND_SWZ_W_SHIFT);
+}
+
+/* these just exist here until nouveau_reg.h has them. */
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_COL0	(1<<0)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_COL1	(1<<1)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_BFC0	(1<<2)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_BFC1	(1<<3)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_FOGC	(1<<4)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_PSZ	(1<<5)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP0	(1<<6)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP1	(1<<7)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP2	(1<<8)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP3	(1<<9)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP4	(1<<10)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP5	(1<<11)
+#define NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_TEX0   (1<<14)
+
+static unsigned int
+NV40VPTranslateResultReg(nvsFunc *shader, nvsFixedReg result,
+					  unsigned int *mask_ret)
+{
+	unsigned int *out_reg = &shader->card_priv->NV30VP.vp_out_reg;
+	unsigned int *clip_en = &shader->card_priv->NV30VP.clip_enables;
+
+	*mask_ret = 0xf;
+
+	switch (result) {
+	case NVS_FR_POSITION:
+		/* out_reg POS implied */
+		return NV40_VP_INST_DEST_POS;
+	case NVS_FR_COL0:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_COL0;
+		return NV40_VP_INST_DEST_COL0;
+	case NVS_FR_COL1:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_COL1;
+		return NV40_VP_INST_DEST_COL1;
+	case NVS_FR_BFC0:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_BFC0;
+		return NV40_VP_INST_DEST_BFC0;
+	case NVS_FR_BFC1:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_BFC1;
+		return NV40_VP_INST_DEST_BFC1;
+	case NVS_FR_FOGCOORD:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_FOGC;
+		*mask_ret = 0x8;
+		return NV40_VP_INST_DEST_FOGC;
+	case NVS_FR_CLIP0:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP0;
+		(*clip_en) |= 0x00000002;
+		*mask_ret = 0x4;
+		return NV40_VP_INST_DEST_FOGC;
+	case NVS_FR_CLIP1:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP1;
+		(*clip_en) |= 0x00000020;
+		*mask_ret = 0x2;
+		return NV40_VP_INST_DEST_FOGC;
+	case NVS_FR_CLIP2:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP2;
+		(*clip_en) |= 0x00000200;
+		*mask_ret = 0x1;
+		return NV40_VP_INST_DEST_FOGC;
+	case NVS_FR_POINTSZ:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_PSZ;
+		*mask_ret = 0x8;
+		return NV40_VP_INST_DEST_PSZ;
+	case NVS_FR_CLIP3:
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP3;
+		(*clip_en) |= 0x00002000;
+		*mask_ret = 0x4;
+		return NV40_VP_INST_DEST_PSZ;
+	case NVS_FR_CLIP4:
+		(*clip_en) |= 0x00020000;
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP4;
+		*mask_ret = 0x2;
+		return NV40_VP_INST_DEST_PSZ;
+	case NVS_FR_CLIP5:
+		(*clip_en) |= 0x00200000;
+		(*out_reg) |= NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_CLP5;
+		*mask_ret = 0x1;
+		return NV40_VP_INST_DEST_PSZ;
+	case NVS_FR_TEXCOORD0:
+	case NVS_FR_TEXCOORD1:
+	case NVS_FR_TEXCOORD2:
+	case NVS_FR_TEXCOORD3:
+	case NVS_FR_TEXCOORD4:
+	case NVS_FR_TEXCOORD5:
+	case NVS_FR_TEXCOORD6:
+	case NVS_FR_TEXCOORD7:
+	{
+		int unit = result - NVS_FR_TEXCOORD0;
+		(*out_reg) |= (NV30_TCL_PRIMITIVE_3D_VP_OUT_REG_TEX0 << unit);
+		return NV40_VP_INST_DEST_TC(unit);
+	}
+	default:
+		WARN_ONCE("unknown vp output %d\n", result);
+		return NV40_VP_INST_DEST_POS;
+	}
+}
+
+static void
+NV40VPSetResult(nvsFunc *shader, nvsRegister * dest, unsigned int mask,
+      		int slot)
+{
+   unsigned int hwmask = 0;
+
+   if (mask & SMASK_X) hwmask |= (1 << 3);
+   if (mask & SMASK_Y) hwmask |= (1 << 2);
+   if (mask & SMASK_Z) hwmask |= (1 << 1);
+   if (mask & SMASK_W) hwmask |= (1 << 0);
+
+   if (dest->file == NVS_FILE_RESULT) {
+      unsigned int valid_mask;
+      int hwidx;
+
+      hwidx = NV40VPTranslateResultReg(shader, dest->index, &valid_mask);
+      if (hwmask & ~valid_mask)
+	      WARN_ONCE("writing invalid components of result reg\n");
+      hwmask &= valid_mask;
+
+      shader->inst[3] &= ~NV40_VP_INST_DEST_MASK;
+      shader->inst[3] |= (hwidx << NV40_VP_INST_DEST_SHIFT);
+
+      if (slot) shader->inst[3] |= NV40_VP_INST_SCA_RESULT;
+      else      shader->inst[0] |= NV40_VP_INST_VEC_RESULT;
+   } else {
+      /* NVS_FILE_TEMP || NVS_FILE_ADDRESS */
+      if (slot) {
+	 shader->inst[3] &= ~NV40_VP_INST_SCA_RESULT;
+	 shader->inst[3] &= ~NV40_VP_INST_SCA_DEST_TEMP_MASK;
+	 shader->inst[3] |= (dest->index << NV40_VP_INST_SCA_DEST_TEMP_SHIFT);
+      } else {
+	 shader->inst[0] &= ~NV40_VP_INST_VEC_RESULT;
+	 shader->inst[0] &= ~(NV40_VP_INST_VEC_DEST_TEMP_MASK | (1<<20));
+	 shader->inst[0] |= (dest->index << NV40_VP_INST_VEC_DEST_TEMP_SHIFT);
+      }
+   }
+
+   if (slot) {
+      shader->inst[3] &= ~NV40_VP_INST_SCA_WRITEMASK_MASK;
+      shader->inst[3] |= (hwmask << NV40_VP_INST_SCA_WRITEMASK_SHIFT);
+   } else {
+      shader->inst[3] &= ~NV40_VP_INST_VEC_WRITEMASK_MASK;
+      shader->inst[3] |= (hwmask << NV40_VP_INST_VEC_WRITEMASK_SHIFT);
+   }
+}
+
+static void
+NV40VPInsertSource(nvsFunc *shader, unsigned int hw, int pos)
+{
+   switch (pos) {
+   case 0:
+      shader->inst[1] &= ~NV40_VP_INST_SRC0H_MASK;
+      shader->inst[2] &= ~NV40_VP_INST_SRC0L_MASK;
+      shader->inst[1] |= ((hw & NV40_VP_SRC0_HIGH_MASK) >>
+	    NV40_VP_SRC0_HIGH_SHIFT)
+	 << NV40_VP_INST_SRC0H_SHIFT;
+      shader->inst[2] |= (hw & NV40_VP_SRC0_LOW_MASK)
+	 << NV40_VP_INST_SRC0L_SHIFT;
+      break;
+   case 1:
+      shader->inst[2] &= ~NV40_VP_INST_SRC1_MASK;
+      shader->inst[2] |= hw
+	 << NV40_VP_INST_SRC1_SHIFT;
+      break;
+   case 2:
+      shader->inst[2] &= ~NV40_VP_INST_SRC2H_MASK;
+      shader->inst[3] &= ~NV40_VP_INST_SRC2L_MASK;
+      shader->inst[2] |= ((hw & NV40_VP_SRC2_HIGH_MASK) >>
+	    NV40_VP_SRC2_HIGH_SHIFT)
+	 << NV40_VP_INST_SRC2H_SHIFT;
+      shader->inst[3] |= (hw & NV40_VP_SRC2_LOW_MASK)
+	 << NV40_VP_INST_SRC2L_SHIFT;
+      break;
+   default:
+      assert(0);
+      break;
+   }
+}
+
+static void
+NV40VPSetSource(nvsFunc *shader, nvsRegister * src, int pos)
+{
+   unsigned int hw = 0;
+
+   switch (src->file) {
+   case NVS_FILE_ADDRESS:
+      break;
+   case NVS_FILE_ATTRIB:
+      hw |= (NV40_VP_SRC_REG_TYPE_INPUT << NV40_VP_SRC_REG_TYPE_SHIFT);
+
+      shader->inst[1] &= ~NV40_VP_INST_INPUT_SRC_MASK;
+      shader->inst[1] |= (src->index << NV40_VP_INST_INPUT_SRC_SHIFT);
+      shader->card_priv->NV30VP.vp_in_reg |= (1 << src->index);
+      if (src->indexed) {
+	 shader->inst[0] |= NV40_VP_INST_INDEX_INPUT;
+	 if (src->addr_reg)
+	    shader->inst[0] |= NV40_VP_INST_ADDR_REG_SELECT_1;
+	 else
+	    shader->inst[0] &= ~NV40_VP_INST_ADDR_REG_SELECT_1;
+	 shader->inst[0] &= ~NV40_VP_INST_ADDR_SWZ_SHIFT;
+	 shader->inst[0] |= (src->addr_comp << NV40_VP_INST_ADDR_SWZ_SHIFT);
+      } else
+	 shader->inst[0] &= ~NV40_VP_INST_INDEX_INPUT;
+      break;
+   case NVS_FILE_CONST:
+      hw |= (NV40_VP_SRC_REG_TYPE_CONST << NV40_VP_SRC_REG_TYPE_SHIFT);
+
+      shader->inst[1] &= ~NV40_VP_INST_CONST_SRC_MASK;
+      shader->inst[1] |= (src->index << NV40_VP_INST_CONST_SRC_SHIFT);
+      if (src->indexed) {
+	 shader->inst[3] |= NV40_VP_INST_INDEX_CONST;
+	 if (src->addr_reg)
+	    shader->inst[0] |= NV40_VP_INST_ADDR_REG_SELECT_1;
+	 else
+	    shader->inst[0] &= ~NV40_VP_INST_ADDR_REG_SELECT_1;
+	 shader->inst[0] &= ~NV40_VP_INST_ADDR_SWZ_MASK;
+	 shader->inst[0] |= (src->addr_comp << NV40_VP_INST_ADDR_SWZ_SHIFT);
+      } else
+	 shader->inst[3] &= ~NV40_VP_INST_INDEX_CONST;
+      break;
+   case NVS_FILE_TEMP:
+      hw |= (NV40_VP_SRC_REG_TYPE_TEMP << NV40_VP_SRC_REG_TYPE_SHIFT);
+      hw |= (src->index << NV40_VP_SRC_TEMP_SRC_SHIFT);
+      break;
+   default:
+      fprintf(stderr, "unknown source file %d\n", src->file);
+      assert(0);
+      break;
+   }
+
+   if (src->file != NVS_FILE_ADDRESS) {
+      if (src->negate)
+	 hw |= NV40_VP_SRC_NEGATE;
+      if (src->abs)
+	 shader->inst[0] |= (1 << (21 + pos));
+      else
+	 shader->inst[0] &= ~(1 << (21 + pos));
+      hw |= (src->swizzle[0] << NV40_VP_SRC_SWZ_X_SHIFT);
+      hw |= (src->swizzle[1] << NV40_VP_SRC_SWZ_Y_SHIFT);
+      hw |= (src->swizzle[2] << NV40_VP_SRC_SWZ_Z_SHIFT);
+      hw |= (src->swizzle[3] << NV40_VP_SRC_SWZ_W_SHIFT);
+
+      NV40VPInsertSource(shader, hw, pos);
+   }
+}
+
+static void
+NV40VPSetBranchTarget(nvsFunc *shader, int addr)
+{
+	shader->inst[2] &= ~NV40_VP_INST_IADDRH_MASK;
+	shader->inst[2] |= ((addr & 0xf8) >> 3) << NV40_VP_INST_IADDRH_SHIFT;
+	shader->inst[3] &= ~NV40_VP_INST_IADDRL_MASK;
+	shader->inst[3] |= ((addr & 0x07) << NV40_VP_INST_IADDRL_SHIFT);
+}
+
+static void
+NV40VPInitInstruction(nvsFunc *shader)
+{
+   unsigned int hwsrc = 0;
+
+   shader->inst[0] = /*NV40_VP_INST_VEC_RESULT | */
+      		     NV40_VP_INST_VEC_DEST_TEMP_MASK | (1<<20);
+   shader->inst[1] = 0;
+   shader->inst[2] = 0;
+   shader->inst[3] = NV40_VP_INST_SCA_RESULT |
+      		     NV40_VP_INST_SCA_DEST_TEMP_MASK |
+		     NV40_VP_INST_DEST_MASK;
+   
+   hwsrc = (NV40_VP_SRC_REG_TYPE_INPUT << NV40_VP_SRC_REG_TYPE_SHIFT) |
+      	   (NVS_SWZ_X << NV40_VP_SRC_SWZ_X_SHIFT) |
+      	   (NVS_SWZ_Y << NV40_VP_SRC_SWZ_Y_SHIFT) |
+      	   (NVS_SWZ_Z << NV40_VP_SRC_SWZ_Z_SHIFT) |
+      	   (NVS_SWZ_W << NV40_VP_SRC_SWZ_W_SHIFT);
+   NV40VPInsertSource(shader, hwsrc, 0);
+   NV40VPInsertSource(shader, hwsrc, 1);
+   NV40VPInsertSource(shader, hwsrc, 2);
+}
+
+static void
+NV40VPSetLastInst(nvsFunc *shader)
+{
+   shader->inst[3] |= 1;
+}
+
+/*****************************************************************************
+ * Disassembly routines
+ */
+static int
+NV40VPHasMergedInst(nvsFunc * shader)
+{
+   if (shader->GetOpcodeHW(shader, 0) != NV40_VP_INST_OP_NOP &&
+       shader->GetOpcodeHW(shader, 1) != NV40_VP_INST_OP_NOP)
+      return 1;
+   return 0;
+}
+
+static unsigned int
+NV40VPGetOpcodeHW(nvsFunc * shader, int slot)
+{
+   int op;
+
+   if (slot)
+      op = (shader->inst[1] & NV40_VP_INST_SCA_OPCODE_MASK)
+	 >> NV40_VP_INST_SCA_OPCODE_SHIFT;
+   else
+      op = (shader->inst[1] & NV40_VP_INST_VEC_OPCODE_MASK)
+	 >> NV40_VP_INST_VEC_OPCODE_SHIFT;
+
+   return op;
+}
+
+static nvsRegFile
+NV40VPGetDestFile(nvsFunc * shader, int merged)
+{
+   nvsOpcode op;
+
+   op = shader->GetOpcode(shader, merged);
+   switch (op) {
+   case NVS_OP_ARL:
+   case NVS_OP_ARR:
+   case NVS_OP_ARA:
+   case NVS_OP_POPA:
+      return NVS_FILE_ADDRESS;
+   default:
+      if (shader->GetOpcodeSlot(shader, merged)) {
+	 if (shader->inst[3] & NV40_VP_INST_SCA_RESULT)
+	    return NVS_FILE_RESULT;
+      }
+      else {
+	 if (shader->inst[0] & NV40_VP_INST_VEC_RESULT)
+	    return NVS_FILE_RESULT;
+      }
+      return NVS_FILE_TEMP;
+   }
+
+}
+
+static unsigned int
+NV40VPGetDestID(nvsFunc * shader, int merged)
+{
+   int id;
+
+   switch (shader->GetDestFile(shader, merged)) {
+   case NVS_FILE_RESULT:
+      id = ((shader->inst[3] & NV40_VP_INST_DEST_MASK)
+	    >> NV40_VP_INST_DEST_SHIFT);
+      switch (id) {
+      case NV40_VP_INST_DEST_POS : return NVS_FR_POSITION;
+      case NV40_VP_INST_DEST_COL0: return NVS_FR_COL0;
+      case NV40_VP_INST_DEST_COL1: return NVS_FR_COL1;
+      case NV40_VP_INST_DEST_BFC0: return NVS_FR_BFC0;
+      case NV40_VP_INST_DEST_BFC1: return NVS_FR_BFC1;
+      case NV40_VP_INST_DEST_FOGC: {
+	    int mask = shader->GetDestMask(shader, merged);
+	    switch (mask) {
+	    case SMASK_X: return NVS_FR_FOGCOORD;
+	    case SMASK_Y: return NVS_FR_CLIP0;
+	    case SMASK_Z: return NVS_FR_CLIP1;
+	    case SMASK_W: return NVS_FR_CLIP2;
+	    default:
+	       printf("more than 1 mask component set in FOGC writemask!\n");
+	       return NVS_FR_UNKNOWN;
+	    }
+	 }
+      case NV40_VP_INST_DEST_PSZ:
+	 {
+	    int mask = shader->GetDestMask(shader, merged);
+	    switch (mask) {
+	    case SMASK_X: return NVS_FR_POINTSZ;
+	    case SMASK_Y: return NVS_FR_CLIP3;
+	    case SMASK_Z: return NVS_FR_CLIP4;
+	    case SMASK_W: return NVS_FR_CLIP5;
+	    default:
+	       printf("more than 1 mask component set in PSZ writemask!\n");
+	       return NVS_FR_UNKNOWN;
+	    }
+	 }
+      case NV40_VP_INST_DEST_TC(0): return NVS_FR_TEXCOORD0;
+      case NV40_VP_INST_DEST_TC(1): return NVS_FR_TEXCOORD1;
+      case NV40_VP_INST_DEST_TC(2): return NVS_FR_TEXCOORD2;
+      case NV40_VP_INST_DEST_TC(3): return NVS_FR_TEXCOORD3;
+      case NV40_VP_INST_DEST_TC(4): return NVS_FR_TEXCOORD4;
+      case NV40_VP_INST_DEST_TC(5): return NVS_FR_TEXCOORD5;
+      case NV40_VP_INST_DEST_TC(6): return NVS_FR_TEXCOORD6;
+      case NV40_VP_INST_DEST_TC(7): return NVS_FR_TEXCOORD7;
+      default:
+	 return -1;
+      }
+   case NVS_FILE_ADDRESS:
+      /* Instructions that write address regs are encoded as if
+       * they would write temps.
+       */
+   case NVS_FILE_TEMP:
+      if (shader->GetOpcodeSlot(shader, merged))
+	 id = ((shader->inst[3] & NV40_VP_INST_SCA_DEST_TEMP_MASK)
+	       >> NV40_VP_INST_SCA_DEST_TEMP_SHIFT);
+      else
+	 id = ((shader->inst[0] & NV40_VP_INST_VEC_DEST_TEMP_MASK)
+	       >> NV40_VP_INST_VEC_DEST_TEMP_SHIFT);
+      return id;
+   default:
+      return -1;
+   }
+}
+
+static unsigned int
+NV40VPGetDestMask(nvsFunc * shader, int merged)
+{
+   unsigned int mask = 0;
+
+   if (shader->GetOpcodeSlot(shader, merged)) {
+      if (shader->inst[3] & NV40_VP_INST_SCA_WRITEMASK_X) mask |= SMASK_X;
+      if (shader->inst[3] & NV40_VP_INST_SCA_WRITEMASK_Y) mask |= SMASK_Y;
+      if (shader->inst[3] & NV40_VP_INST_SCA_WRITEMASK_Z) mask |= SMASK_Z;
+      if (shader->inst[3] & NV40_VP_INST_SCA_WRITEMASK_W) mask |= SMASK_W;
+   } else {
+      if (shader->inst[3] & NV40_VP_INST_VEC_WRITEMASK_X) mask |= SMASK_X;
+      if (shader->inst[3] & NV40_VP_INST_VEC_WRITEMASK_Y) mask |= SMASK_Y;
+      if (shader->inst[3] & NV40_VP_INST_VEC_WRITEMASK_Z) mask |= SMASK_Z;
+      if (shader->inst[3] & NV40_VP_INST_VEC_WRITEMASK_W) mask |= SMASK_W;
+   }
+
+   return mask;
+}
+
+static unsigned int
+NV40VPGetSourceHW(nvsFunc * shader, int merged, int pos)
+{
+   struct _op_xlat *opr;
+   unsigned int src;
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr)
+      return -1;
+
+   switch (opr->srcpos[pos]) {
+   case 0:
+      src = ((shader->inst[1] & NV40_VP_INST_SRC0H_MASK)
+	     >> NV40_VP_INST_SRC0H_SHIFT)
+	 << NV40_VP_SRC0_HIGH_SHIFT;
+      src |= ((shader->inst[2] & NV40_VP_INST_SRC0L_MASK)
+	      >> NV40_VP_INST_SRC0L_SHIFT);
+      break;
+   case 1:
+      src = ((shader->inst[2] & NV40_VP_INST_SRC1_MASK)
+	     >> NV40_VP_INST_SRC1_SHIFT);
+      break;
+   case 2:
+      src = ((shader->inst[2] & NV40_VP_INST_SRC2H_MASK)
+	     >> NV40_VP_INST_SRC2H_SHIFT)
+	 << NV40_VP_SRC2_HIGH_SHIFT;
+      src |= ((shader->inst[3] & NV40_VP_INST_SRC2L_MASK)
+	      >> NV40_VP_INST_SRC2L_SHIFT);
+      break;
+   default:
+      src = -1;
+   }
+
+   return src;
+}
+
+static nvsRegFile
+NV40VPGetSourceFile(nvsFunc * shader, int merged, int pos)
+{
+   unsigned int src;
+   struct _op_xlat *opr;
+   int file;
+
+   opr = shader->GetOPTXRec(shader, merged);
+   if (!opr || opr->srcpos[pos] == -1)
+      return -1;
+
+   switch (opr->srcpos[pos]) {
+   case SPOS_ADDRESS: return NVS_FILE_ADDRESS;
+   default:
+      src = shader->GetSourceHW(shader, merged, pos);
+      file = (src & NV40_VP_SRC_REG_TYPE_MASK) >> NV40_VP_SRC_REG_TYPE_SHIFT;
+
+      switch (file) {
+      case NV40_VP_SRC_REG_TYPE_TEMP : return NVS_FILE_TEMP;
+      case NV40_VP_SRC_REG_TYPE_INPUT: return NVS_FILE_ATTRIB;
+      case NV40_VP_SRC_REG_TYPE_CONST: return NVS_FILE_CONST;
+      default:
+	 return NVS_FILE_UNKNOWN;
+      }
+   }
+}
+
+static int
+NV40VPGetSourceID(nvsFunc * shader, int merged, int pos)
+{
+   switch (shader->GetSourceFile(shader, merged, pos)) {
+   case NVS_FILE_ATTRIB:
+      switch ((shader->inst[1] & NV40_VP_INST_INPUT_SRC_MASK)
+	      >> NV40_VP_INST_INPUT_SRC_SHIFT) {
+      case NV40_VP_INST_IN_POS:		return NVS_FR_POSITION;
+      case NV40_VP_INST_IN_WEIGHT:	return NVS_FR_WEIGHT;
+      case NV40_VP_INST_IN_NORMAL:	return NVS_FR_NORMAL;
+      case NV40_VP_INST_IN_COL0:	return NVS_FR_COL0;
+      case NV40_VP_INST_IN_COL1:	return NVS_FR_COL1;
+      case NV40_VP_INST_IN_FOGC:	return NVS_FR_FOGCOORD;
+      case NV40_VP_INST_IN_TC(0):	return NVS_FR_TEXCOORD0;
+      case NV40_VP_INST_IN_TC(1):	return NVS_FR_TEXCOORD1;
+      case NV40_VP_INST_IN_TC(2):	return NVS_FR_TEXCOORD2;
+      case NV40_VP_INST_IN_TC(3):	return NVS_FR_TEXCOORD3;
+      case NV40_VP_INST_IN_TC(4):	return NVS_FR_TEXCOORD4;
+      case NV40_VP_INST_IN_TC(5):	return NVS_FR_TEXCOORD5;
+      case NV40_VP_INST_IN_TC(6):	return NVS_FR_TEXCOORD6;
+      case NV40_VP_INST_IN_TC(7):	return NVS_FR_TEXCOORD7;
+      default:
+	 return -1;
+      }
+      break;
+   case NVS_FILE_CONST:
+      return ((shader->inst[1] & NV40_VP_INST_CONST_SRC_MASK)
+	      >> NV40_VP_INST_CONST_SRC_SHIFT);
+   case NVS_FILE_TEMP:
+      {
+	 unsigned int src;
+
+	 src = shader->GetSourceHW(shader, merged, pos);
+	 return ((src & NV40_VP_SRC_TEMP_SRC_MASK) >>
+		 NV40_VP_SRC_TEMP_SRC_SHIFT);
+      }
+   default:
+      return -1;
+   }
+}
+
+static int
+NV40VPGetSourceNegate(nvsFunc * shader, int merged, int pos)
+{
+   unsigned int src;
+
+   src = shader->GetSourceHW(shader, merged, pos);
+
+   if (src == -1)
+      return -1;
+   return ((src & NV40_VP_SRC_NEGATE) ? 1 : 0);
+}
+
+static void
+NV40VPGetSourceSwizzle(nvsFunc * shader, int merged, int pos, nvsSwzComp *swz)
+{
+   unsigned int src;
+   int swzbits;
+
+   src = shader->GetSourceHW(shader, merged, pos);
+   swzbits = (src & NV40_VP_SRC_SWZ_ALL_MASK) >> NV40_VP_SRC_SWZ_ALL_SHIFT;
+   NV20VPTXSwizzle(swzbits, swz);
+}
+
+static int
+NV40VPGetSourceIndexed(nvsFunc * shader, int merged, int pos)
+{
+   switch (shader->GetSourceFile(shader, merged, pos)) {
+   case NVS_FILE_ATTRIB:
+      return ((shader->inst[0] & NV40_VP_INST_INDEX_INPUT) ? 1 : 0);
+   case NVS_FILE_CONST:
+      return ((shader->inst[3] & NV40_VP_INST_INDEX_CONST) ? 1 : 0);
+   default:
+      return 0;
+   }
+}
+
+static nvsSwzComp
+NV40VPGetAddressRegSwizzle(nvsFunc * shader)
+{
+   nvsSwzComp swz;
+
+   swz = NV20VP_TX_SWIZZLE[(shader->inst[0] & NV40_VP_INST_ADDR_SWZ_MASK)
+			   >> NV40_VP_INST_ADDR_SWZ_SHIFT];
+   return swz;
+}
+
+static int
+NV40VPSupportsConditional(nvsFunc * shader)
+{
+   /*FIXME: Is this true of all ops? */
+   return 1;
+}
+
+static int
+NV40VPGetConditionUpdate(nvsFunc * shader)
+{
+   return ((shader->inst[0] & NV40_VP_INST_COND_UPDATE_ENABLE) ? 1 : 0);
+}
+
+static int
+NV40VPGetConditionTest(nvsFunc * shader)
+{
+   int op;
+
+   /* The condition test is unconditionally enabled on some
+    * instructions. ie: the condition test bit does *NOT* have
+    * to be set.
+    *
+    * FIXME: check other relevant ops for this situation.
+    */
+   op = shader->GetOpcodeHW(shader, 1);
+   switch (op) {
+   case NV40_VP_INST_OP_BRA:
+      return 1;
+   default:
+      return ((shader->inst[0] & NV40_VP_INST_COND_TEST_ENABLE) ? 1 : 0);
+   }
+}
+
+static nvsCond
+NV40VPGetCondition(nvsFunc * shader)
+{
+   int cond;
+
+   cond = ((shader->inst[0] & NV40_VP_INST_COND_MASK)
+	   >> NV40_VP_INST_COND_SHIFT);
+
+   switch (cond) {
+   case NV40_VP_INST_COND_FL: return NVS_COND_FL;
+   case NV40_VP_INST_COND_LT: return NVS_COND_LT;
+   case NV40_VP_INST_COND_EQ: return NVS_COND_EQ;
+   case NV40_VP_INST_COND_LE: return NVS_COND_LE;
+   case NV40_VP_INST_COND_GT: return NVS_COND_GT;
+   case NV40_VP_INST_COND_NE: return NVS_COND_NE;
+   case NV40_VP_INST_COND_GE: return NVS_COND_GE;
+   case NV40_VP_INST_COND_TR: return NVS_COND_TR;
+   default:
+      return NVS_COND_UNKNOWN;
+   }
+}
+
+static void
+NV40VPGetCondRegSwizzle(nvsFunc * shader, nvsSwzComp *swz)
+{
+   int swzbits;
+
+   swzbits = (shader->inst[0] & NV40_VP_INST_COND_SWZ_ALL_MASK)
+      >> NV40_VP_INST_COND_SWZ_ALL_SHIFT;
+   NV20VPTXSwizzle(swzbits, swz);
+}
+
+static int
+NV40VPGetCondRegID(nvsFunc * shader)
+{
+   return ((shader->inst[0] & NV40_VP_INST_COND_REG_SELECT_1) ? 1 : 0);
+}
+
+static int
+NV40VPGetBranch(nvsFunc * shader)
+{
+   int addr;
+
+   addr = ((shader->inst[2] & NV40_VP_INST_IADDRH_MASK)
+	   >> NV40_VP_INST_IADDRH_SHIFT) << 3;
+   addr |= ((shader->inst[3] & NV40_VP_INST_IADDRL_MASK)
+	    >> NV40_VP_INST_IADDRL_SHIFT);
+   return addr;
+}
+
+void
+NV40VPInitShaderFuncs(nvsFunc * shader)
+{
+   /* Inherit NV30 VP code, we share some of it */
+   NV30VPInitShaderFuncs(shader);
+
+   /* Limits */
+   shader->MaxInst	= 4096;
+   shader->MaxAttrib	= 16;
+   shader->MaxTemp	= 32;
+   shader->MaxAddress	= 2;
+   shader->MaxConst	= 256;
+   shader->caps		= SCAP_SRC_ABS;
+
+   /* Add extra opcodes for NV40+ */
+//      MOD_OPCODE(NVVP_TX_VOP, NV40_VP_INST_OP_TXWHAT, NVS_OP_TEX  ,  0,  4, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV40_VP_INST_OP_PUSHA, NVS_OP_PUSHA,  3, -1, -1);
+   MOD_OPCODE(NVVP_TX_SOP, NV40_VP_INST_OP_POPA , NVS_OP_POPA , -1, -1, -1);
+
+   shader->InitInstruction	= NV40VPInitInstruction;
+   shader->SupportsOpcode	= NV40VPSupportsOpcode;
+   shader->SetOpcode		= NV40VPSetOpcode;
+   shader->SetCCUpdate		= NV40VPSetCCUpdate;
+   shader->SetCondition		= NV40VPSetCondition;
+   shader->SetResult		= NV40VPSetResult;
+   shader->SetSource		= NV40VPSetSource;
+   shader->SetLastInst		= NV40VPSetLastInst;
+   shader->SetBranchTarget	= NV40VPSetBranchTarget;
+
+   shader->HasMergedInst	= NV40VPHasMergedInst;
+   shader->GetOpcodeHW		= NV40VPGetOpcodeHW;
+
+   shader->GetDestFile		= NV40VPGetDestFile;
+   shader->GetDestID		= NV40VPGetDestID;
+   shader->GetDestMask		= NV40VPGetDestMask;
+
+   shader->GetSourceHW		= NV40VPGetSourceHW;
+   shader->GetSourceFile	= NV40VPGetSourceFile;
+   shader->GetSourceID		= NV40VPGetSourceID;
+   shader->GetSourceNegate	= NV40VPGetSourceNegate;
+   shader->GetSourceSwizzle	= NV40VPGetSourceSwizzle;
+   shader->GetSourceIndexed	= NV40VPGetSourceIndexed;
+
+   shader->GetRelAddressSwizzle	= NV40VPGetAddressRegSwizzle;
+
+   shader->SupportsConditional	= NV40VPSupportsConditional;
+   shader->GetConditionUpdate	= NV40VPGetConditionUpdate;
+   shader->GetConditionTest	= NV40VPGetConditionTest;
+   shader->GetCondition		= NV40VPGetCondition;
+   shader->GetCondRegSwizzle	= NV40VPGetCondRegSwizzle;
+   shader->GetCondRegID		= NV40VPGetCondRegID;
+
+   shader->GetBranch		= NV40VPGetBranch;
+}
diff --git a/src/mesa/drivers/dri/nouveau/nv50_state.c b/src/mesa/drivers/dri/nouveau/nv50_state.c
new file mode 100644
index 00000000000..818e268615c
--- /dev/null
+++ b/src/mesa/drivers/dri/nouveau/nv50_state.c
@@ -0,0 +1,641 @@
+/**************************************************************************
+
+Copyright 2006 Nouveau
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ERIC ANHOLT OR SILICON INTEGRATED SYSTEMS CORP 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.
+
+**************************************************************************/
+
+#include "nouveau_context.h"
+#include "nouveau_object.h"
+#include "nouveau_fifo.h"
+#include "nouveau_reg.h"
+#include "nouveau_state.h"
+
+#include "tnl/t_pipeline.h"
+
+#include "mtypes.h"
+#include "colormac.h"
+
+static void nv50AlphaFunc(GLcontext *ctx, GLenum func, GLfloat ref)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLubyte ubRef;
+	CLAMPED_FLOAT_TO_UBYTE(ubRef, ref);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_ALPHA_FUNC_REF, 2);
+	OUT_RING_CACHE(ubRef);
+	OUT_RING_CACHE(func);
+}
+
+static void nv50BlendColor(GLcontext *ctx, const GLfloat color[4])
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx); 
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_BLEND_COLOR_R, 4);
+	OUT_RING_CACHEf(color[0]);
+	OUT_RING_CACHEf(color[1]);
+	OUT_RING_CACHEf(color[2]);
+	OUT_RING_CACHEf(color[3]);
+}
+
+static void nv50BlendEquationSeparate(GLcontext *ctx, GLenum modeRGB, GLenum modeA)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_BLEND_EQUATION_RGB, 1);
+	OUT_RING_CACHE(modeRGB);
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_BLEND_EQUATION_ALPHA, 1);
+	OUT_RING_CACHE(modeA);
+}
+
+
+static void nv50BlendFuncSeparate(GLcontext *ctx, GLenum sfactorRGB, GLenum dfactorRGB,
+		GLenum sfactorA, GLenum dfactorA)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC_RGB, 2);
+	OUT_RING_CACHE(sfactorRGB);	/* FIXME, sometimes has |0x4000 */
+	OUT_RING_CACHE(dfactorRGB);	/* FIXME, sometimes has |0x4000 */
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_BLEND_FUNC_SRC_ALPHA, 2);
+	OUT_RING_CACHE(sfactorA);	/* FIXME, sometimes has |0x4000 */
+	OUT_RING_CACHE(dfactorA);	/* FIXME, sometimes has |0x4000 */
+}
+
+static void nv50Clear(GLcontext *ctx, GLbitfield mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	GLuint hw_bufs = 0;
+
+	if (mask & (BUFFER_BIT_FRONT_LEFT | BUFFER_BIT_BACK_LEFT))
+		hw_bufs |= 0x3c;
+	if (mask & (BUFFER_BIT_STENCIL))
+		hw_bufs |= 0x02;
+	if (mask & (BUFFER_BIT_DEPTH))
+		hw_bufs |= 0x01;
+
+	if (hw_bufs) {
+		BEGIN_RING_SIZE(NvSub3D, NV50_TCL_PRIMITIVE_3D_CLEAR_BUFFERS, 1);
+		OUT_RING(hw_bufs);
+	}
+}
+
+static void nv50ClearColor(GLcontext *ctx, const GLfloat color[4])
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_CLEAR_COLOR_R, 4);
+	OUT_RING_CACHEf(color[0]);
+	OUT_RING_CACHEf(color[1]);
+	OUT_RING_CACHEf(color[2]);
+	OUT_RING_CACHEf(color[3]);
+}
+
+static void nv50ClearDepth(GLcontext *ctx, GLclampd d)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_CLEAR_DEPTH, 1);
+	OUT_RING_CACHEf(d);
+}
+
+/* we're don't support indexed buffers
+   void (*ClearIndex)(GLcontext *ctx, GLuint index)
+ */
+
+static void nv50ClearStencil(GLcontext *ctx, GLint s)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_CLEAR_STENCIL, 1);
+	OUT_RING_CACHE(s);
+}
+
+static void nv50ClipPlane(GLcontext *ctx, GLenum plane, const GLfloat *equation)
+{
+	/* Only using shaders */
+}
+
+static void nv50ColorMask(GLcontext *ctx, GLboolean rmask, GLboolean gmask,
+		GLboolean bmask, GLboolean amask )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	int i;
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_COLOR_MASK(0), 8);
+	for (i=0; i<8; i++) {
+		OUT_RING_CACHE(((amask && 0x01) << 12) | ((bmask && 0x01) << 8) | ((gmask && 0x01)<< 4) | ((rmask && 0x01) << 0));
+	}
+}
+
+static void nv50ColorMaterial(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	// TODO I need love
+}
+
+static void nv50CullFace(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_CULL_FACE, 1);
+	OUT_RING_CACHE(mode);
+}
+
+static void nv50FrontFace(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_FRONT_FACE, 1);
+	OUT_RING_CACHE(mode);
+}
+
+static void nv50DepthFunc(GLcontext *ctx, GLenum func)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_DEPTH_FUNC, 1);
+	OUT_RING_CACHE(func);
+}
+
+static void nv50DepthMask(GLcontext *ctx, GLboolean flag)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_DEPTH_WRITE_ENABLE, 1);
+	OUT_RING_CACHE(flag);
+}
+
+static void nv50DepthRange(GLcontext *ctx, GLclampd nearval, GLclampd farval)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_DEPTH_RANGE_NEAR, 2);
+	OUT_RING_CACHEf(nearval);
+	OUT_RING_CACHEf(farval);
+}
+
+/** Specify the current buffer for writing */
+//void (*DrawBuffer)( GLcontext *ctx, GLenum buffer );
+/** Specify the buffers for writing for fragment programs*/
+//void (*DrawBuffers)( GLcontext *ctx, GLsizei n, const GLenum *buffers );
+
+static void nv50Enable(GLcontext *ctx, GLenum cap, GLboolean state)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	switch(cap)
+	{
+		case GL_ALPHA_TEST:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_ALPHA_FUNC_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_AUTO_NORMAL:
+//		case GL_BLEND:
+//		case GL_CLIP_PLANE0:
+//		case GL_CLIP_PLANE1:
+//		case GL_CLIP_PLANE2:
+//		case GL_CLIP_PLANE3:
+//		case GL_CLIP_PLANE4:
+//		case GL_CLIP_PLANE5:
+		case GL_COLOR_LOGIC_OP:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_LOGIC_OP_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_COLOR_MATERIAL:
+//		case GL_COLOR_SUM_EXT:
+//		case GL_COLOR_TABLE:
+//		case GL_CONVOLUTION_1D:
+//		case GL_CONVOLUTION_2D:
+		case GL_CULL_FACE:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_CULL_FACE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_DEPTH_TEST:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_DEPTH_TEST_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_DITHER:
+//		case GL_FOG:
+//		case GL_HISTOGRAM:
+//		case GL_INDEX_LOGIC_OP:
+//		case GL_LIGHT0:
+//		case GL_LIGHT1:
+//		case GL_LIGHT2:
+//		case GL_LIGHT3:
+//		case GL_LIGHT4:
+//		case GL_LIGHT5:
+//		case GL_LIGHT6:
+//		case GL_LIGHT7:
+//		case GL_LIGHTING:
+		case GL_LINE_SMOOTH:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_LINE_SMOOTH_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_LINE_STIPPLE:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_LINE_STIPPLE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_MAP1_COLOR_4:
+//		case GL_MAP1_INDEX:
+//		case GL_MAP1_NORMAL:
+//		case GL_MAP1_TEXTURE_COORD_1:
+//		case GL_MAP1_TEXTURE_COORD_2:
+//		case GL_MAP1_TEXTURE_COORD_3:
+//		case GL_MAP1_TEXTURE_COORD_4:
+//		case GL_MAP1_VERTEX_3:
+//		case GL_MAP1_VERTEX_4:
+//		case GL_MAP2_COLOR_4:
+//		case GL_MAP2_INDEX:
+//		case GL_MAP2_NORMAL:
+//		case GL_MAP2_TEXTURE_COORD_1:
+//		case GL_MAP2_TEXTURE_COORD_2:
+//		case GL_MAP2_TEXTURE_COORD_3:
+//		case GL_MAP2_TEXTURE_COORD_4:
+//		case GL_MAP2_VERTEX_3:
+//		case GL_MAP2_VERTEX_4:
+//		case GL_MINMAX:
+//		case GL_NORMALIZE:
+//		case GL_POINT_SMOOTH:
+		case GL_POLYGON_OFFSET_POINT:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_POINT_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_OFFSET_LINE:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_LINE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_OFFSET_FILL:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FILL_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_SMOOTH:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_SMOOTH_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+		case GL_POLYGON_STIPPLE:
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_POST_COLOR_MATRIX_COLOR_TABLE:
+//		case GL_POST_CONVOLUTION_COLOR_TABLE:
+//		case GL_RESCALE_NORMAL:
+		case GL_SCISSOR_TEST:
+			/* No enable bit, nv50Scissor will adjust to max range */
+			ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
+			      			 ctx->Scissor.Width, ctx->Scissor.Height);
+			break;
+//		case GL_SEPARABLE_2D:
+		case GL_STENCIL_TEST:
+			// TODO BACK and FRONT ?
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_ENABLE, 1);
+			OUT_RING_CACHE(state);
+			break;
+//		case GL_TEXTURE_GEN_Q:
+//		case GL_TEXTURE_GEN_R:
+//		case GL_TEXTURE_GEN_S:
+//		case GL_TEXTURE_GEN_T:
+//		case GL_TEXTURE_1D:
+//		case GL_TEXTURE_2D:
+//		case GL_TEXTURE_3D:
+	}
+}
+
+static void nv50Fogfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+	/* Only using shaders */
+}
+   
+static void nv50Hint(GLcontext *ctx, GLenum target, GLenum mode)
+{
+	// TODO I need love (fog and line_smooth hints)
+}
+
+// void (*IndexMask)(GLcontext *ctx, GLuint mask);
+
+static void nv50Lightfv(GLcontext *ctx, GLenum light, GLenum pname, const GLfloat *params )
+{
+	/* Only with shaders */
+}
+
+/** Set the lighting model parameters */
+void (*LightModelfv)(GLcontext *ctx, GLenum pname, const GLfloat *params);
+
+
+static void nv50LineStipple(GLcontext *ctx, GLint factor, GLushort pattern )
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_LINE_STIPPLE_PATTERN, 1);
+	OUT_RING_CACHE((pattern << 8) | factor);
+}
+
+static void nv50LineWidth(GLcontext *ctx, GLfloat width)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_LINE_WIDTH, 1);
+	OUT_RING_CACHEf(width);
+}
+
+static void nv50LogicOpcode(GLcontext *ctx, GLenum opcode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_LOGIC_OP_OP, 1);
+	OUT_RING_CACHE(opcode);
+}
+
+static void nv50PointParameterfv(GLcontext *ctx, GLenum pname, const GLfloat *params)
+{
+	/*TODO: not sure what goes here. */
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+	
+}
+
+/** Specify the diameter of rasterized points */
+static void nv50PointSize(GLcontext *ctx, GLfloat size)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POINT_SIZE, 1);
+	OUT_RING_CACHEf(size);
+}
+
+/** Select a polygon rasterization mode */
+static void nv50PolygonMode(GLcontext *ctx, GLenum face, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_MODE_FRONT, 1);
+		OUT_RING_CACHE(mode);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_MODE_BACK, 1);
+		OUT_RING_CACHE(mode);
+	}
+}
+
+/** Set the scale and units used to calculate depth values */
+static void nv50PolygonOffset(GLcontext *ctx, GLfloat factor, GLfloat units)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+ 
+        BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_FACTOR, 1);
+        OUT_RING_CACHEf(factor);
+        BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_OFFSET_UNITS, 1);
+        OUT_RING_CACHEf(units);
+}
+
+/** Set the polygon stippling pattern */
+static void nv50PolygonStipple(GLcontext *ctx, const GLubyte *mask )
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+        BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_POLYGON_STIPPLE_PATTERN(0), 32);
+        OUT_RING_CACHEp(mask, 32);
+}
+
+/* Specifies the current buffer for reading */
+void (*ReadBuffer)( GLcontext *ctx, GLenum buffer );
+/** Set rasterization mode */
+void (*RenderMode)(GLcontext *ctx, GLenum mode );
+
+/** Define the scissor box */
+static void nv50Scissor(GLcontext *ctx, GLint x, GLint y, GLsizei w, GLsizei h)
+{
+        nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	/* There's no scissor enable bit, so adjust the scissor to cover the
+	 * maximum draw buffer bounds
+	 */
+	if (!ctx->Scissor.Enabled) {
+	   x = y = 0;
+	   w = h = 8191;
+	} else {
+	   x += nmesa->drawX;
+	   y += nmesa->drawY;
+	}
+
+        BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_SCISSOR_WIDTH_XPOS, 2);
+        OUT_RING_CACHE(((w) << 16) | x);
+        OUT_RING_CACHE(((h) << 16) | y);
+}
+
+/** Select flat or smooth shading */
+static void nv50ShadeModel(GLcontext *ctx, GLenum mode)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_SHADE_MODEL, 1);
+	OUT_RING_CACHE(mode);
+}
+
+/** OpenGL 2.0 two-sided StencilFunc */
+static void nv50StencilFuncSeparate(GLcontext *ctx, GLenum face, GLenum func,
+		GLint ref, GLuint mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_FUNC, 1);
+		OUT_RING_CACHE(func);
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_REF, 1);
+		OUT_RING_CACHE(ref);
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_FUNC_MASK, 1);
+		OUT_RING_CACHE(mask);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_FUNC_FUNC, 2);
+		OUT_RING_CACHE(func);
+		OUT_RING_CACHE(ref);
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_FUNC_MASK, 1);
+		OUT_RING_CACHE(mask);
+	}
+}
+
+/** OpenGL 2.0 two-sided StencilMask */
+static void nv50StencilMaskSeparate(GLcontext *ctx, GLenum face, GLuint mask)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_MASK, 1);
+		OUT_RING_CACHE(mask);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_MASK, 1);
+		OUT_RING_CACHE(mask);
+	}
+}
+
+/** OpenGL 2.0 two-sided StencilOp */
+static void nv50StencilOpSeparate(GLcontext *ctx, GLenum face, GLenum fail,
+		GLenum zfail, GLenum zpass)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	if (face == GL_FRONT || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_FRONT_OP_FAIL, 3);
+		OUT_RING_CACHE(fail);
+		OUT_RING_CACHE(zfail);
+		OUT_RING_CACHE(zpass);
+	}
+	if (face == GL_BACK || face == GL_FRONT_AND_BACK) {
+		BEGIN_RING_CACHE(NvSub3D, NV50_TCL_PRIMITIVE_3D_STENCIL_BACK_OP_FAIL, 3);
+		OUT_RING_CACHE(fail);
+		OUT_RING_CACHE(zfail);
+		OUT_RING_CACHE(zpass);
+	}
+}
+
+/** Control the generation of texture coordinates */
+void (*TexGen)(GLcontext *ctx, GLenum coord, GLenum pname,
+		const GLfloat *params);
+/** Set texture environment parameters */
+void (*TexEnv)(GLcontext *ctx, GLenum target, GLenum pname,
+		const GLfloat *param);
+/** Set texture parameters */
+void (*TexParameter)(GLcontext *ctx, GLenum target,
+		struct gl_texture_object *texObj,
+		GLenum pname, const GLfloat *params);
+
+static void nv50TextureMatrix(GLcontext *ctx, GLuint unit, const GLmatrix *mat)
+{
+	/* Only with shaders */
+}
+
+static void nv50WindowMoved(nouveauContextPtr nmesa)
+{
+	GLcontext *ctx = nmesa->glCtx;
+	GLfloat *v = nmesa->viewport.m;
+	GLuint w = ctx->Viewport.Width;
+	GLuint h = ctx->Viewport.Height;
+	GLuint x = ctx->Viewport.X + nmesa->drawX;
+	GLuint y = ctx->Viewport.Y + nmesa->drawY;
+	int i;
+
+	BEGIN_RING_CACHE(NvSub3D,
+	      NV50_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_HORIZ(0), 2);
+        OUT_RING_CACHE((8191 << 16) | 0);
+        OUT_RING_CACHE((8191 << 16) | 0);
+	for (i=1; i<8; i++) {
+		BEGIN_RING_CACHE(NvSub3D,
+		      NV50_TCL_PRIMITIVE_3D_VIEWPORT_CLIP_HORIZ(i), 2);
+        	OUT_RING_CACHE(0);
+	        OUT_RING_CACHE(0);
+	}
+
+	ctx->Driver.Scissor(ctx, ctx->Scissor.X, ctx->Scissor.Y,
+	      		    ctx->Scissor.Width, ctx->Scissor.Height);
+}
+
+static GLboolean nv50InitCard(nouveauContextPtr nmesa)
+{
+	int i,j;
+
+	nouveauObjectOnSubchannel(nmesa, NvSub3D, Nv3D);
+
+	BEGIN_RING_SIZE(NvSub3D, 0x1558, 1);
+	OUT_RING(1);
+
+	BEGIN_RING_SIZE(NvSub3D, NV50_TCL_PRIMITIVE_3D_SET_OBJECT_1(0), 8);
+	for (i=0; i<8; i++) {
+		OUT_RING(NvDmaFB);
+	}
+
+	BEGIN_RING_SIZE(NvSub3D, NV50_TCL_PRIMITIVE_3D_SET_OBJECT_0(0), 12);
+	for (i=0; i<12; i++) {
+		OUT_RING(NvDmaFB);
+	}
+
+	BEGIN_RING_SIZE(NvSub3D, 0x121c, 1);
+	OUT_RING(1);
+
+	for (i=0; i<8; i++) {
+		BEGIN_RING_SIZE(NvSub3D, 0x0200 + (i*0x20), 5);
+		for (j=0; j<5; j++) {
+			OUT_RING(0);
+		}
+	}
+
+	BEGIN_RING_SIZE(NvSub3D, 0x0fe0, 5);
+	OUT_RING(0);
+	OUT_RING(0);
+	OUT_RING(0x16);
+	OUT_RING(0);
+	OUT_RING(0);
+
+	return GL_FALSE;
+}
+
+static GLboolean nv50BindBuffers(nouveauContextPtr nmesa, int num_color,
+      				 nouveau_renderbuffer **color,
+				 nouveau_renderbuffer *depth)
+{
+	return GL_FALSE;
+}
+
+void nv50InitStateFuncs(GLcontext *ctx, struct dd_function_table *func)
+{
+	nouveauContextPtr nmesa = NOUVEAU_CONTEXT(ctx);
+
+	func->AlphaFunc			= nv50AlphaFunc;
+	func->BlendColor		= nv50BlendColor;
+	func->BlendEquationSeparate	= nv50BlendEquationSeparate;
+	func->BlendFuncSeparate		= nv50BlendFuncSeparate;
+	func->Clear			= nv50Clear;
+	func->ClearColor		= nv50ClearColor;
+	func->ClearDepth		= nv50ClearDepth;
+	func->ClearStencil		= nv50ClearStencil;
+	func->ClipPlane			= nv50ClipPlane;
+	func->ColorMask			= nv50ColorMask;
+	func->ColorMaterial		= nv50ColorMaterial;
+	func->CullFace			= nv50CullFace;
+	func->FrontFace			= nv50FrontFace;
+	func->DepthFunc			= nv50DepthFunc;
+	func->DepthMask			= nv50DepthMask;
+	func->DepthRange		= nv50DepthRange;
+	func->Enable			= nv50Enable;
+	func->Fogfv			= nv50Fogfv;
+	func->Hint			= nv50Hint;
+	func->Lightfv			= nv50Lightfv;
+/*	func->LightModelfv		= nv50LightModelfv;	 */
+	func->LineStipple		= nv50LineStipple;
+	func->LineWidth			= nv50LineWidth;
+	func->LogicOpcode		= nv50LogicOpcode;
+	func->PointParameterfv		= nv50PointParameterfv;
+	func->PointSize			= nv50PointSize;
+	func->PolygonMode		= nv50PolygonMode;
+	func->PolygonOffset		= nv50PolygonOffset;
+	func->PolygonStipple		= nv50PolygonStipple;
+/*	func->ReadBuffer		= nv50ReadBuffer;	*/
+/*	func->RenderMode		= nv50RenderMode;	*/
+	func->Scissor			= nv50Scissor;
+	func->ShadeModel		= nv50ShadeModel;
+	func->StencilFuncSeparate	= nv50StencilFuncSeparate;
+	func->StencilMaskSeparate	= nv50StencilMaskSeparate;
+	func->StencilOpSeparate		= nv50StencilOpSeparate;
+/*	func->TexGen			= nv50TexGen;		*/
+/*	func->TexParameter		= nv50TexParameter;	*/
+	func->TextureMatrix		= nv50TextureMatrix;
+
+	nmesa->hw_func.InitCard		= nv50InitCard;
+	nmesa->hw_func.BindBuffers	= nv50BindBuffers;
+	nmesa->hw_func.WindowMoved	= nv50WindowMoved;
+}
diff --git a/src/mesa/drivers/dri/r128/r128_context.c b/src/mesa/drivers/dri/r128/r128_context.c
index 8ec027542ad..89ddafa02a5 100644
--- a/src/mesa/drivers/dri/r128/r128_context.c
+++ b/src/mesa/drivers/dri/r128/r128_context.c
@@ -42,7 +42,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -235,7 +235,7 @@ GLboolean r128CreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -293,7 +293,7 @@ void r128DestroyContext( __DRIcontextPrivate *driContextPriv  )
 
       _swsetup_DestroyContext( rmesa->glCtx );
       _tnl_DestroyContext( rmesa->glCtx );
-      _ac_DestroyContext( rmesa->glCtx );
+      _vbo_DestroyContext( rmesa->glCtx );
       _swrast_DestroyContext( rmesa->glCtx );
 
       if ( release_texture_heaps ) {
diff --git a/src/mesa/drivers/dri/r128/r128_state.c b/src/mesa/drivers/dri/r128/r128_state.c
index 1bfd3709377..e476afa5d82 100644
--- a/src/mesa/drivers/dri/r128/r128_state.c
+++ b/src/mesa/drivers/dri/r128/r128_state.c
@@ -44,7 +44,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "enums.h"
 #include "colormac.h"
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1250,7 +1250,7 @@ static void r128DDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    R128_CONTEXT(ctx)->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/r200/Makefile b/src/mesa/drivers/dri/r200/Makefile
index 75c09ff867e..e9144ac75ce 100644
--- a/src/mesa/drivers/dri/r200/Makefile
+++ b/src/mesa/drivers/dri/r200/Makefile
@@ -26,10 +26,6 @@ DRIVER_SOURCES = r200_context.c \
 		 r200_span.c \
 		 r200_maos.c \
 		 r200_sanity.c \
-		 r200_vtxfmt.c \
-		 r200_vtxfmt_c.c \
-		 r200_vtxfmt_sse.c \
-		 r200_vtxfmt_x86.c \
 		 r200_fragshader.c \
 		 r200_vertprog.c \
 		 radeon_screen.c \
@@ -37,7 +33,7 @@ DRIVER_SOURCES = r200_context.c \
 
 C_SOURCES = $(COMMON_SOURCES) $(DRIVER_SOURCES)
 
-X86_SOURCES = r200_vtxtmp_x86.S
+X86_SOURCES = 
 
 DRIVER_DEFINES = -DRADEON_COMMON=1 -DRADEON_COMMON_FOR_R200
 
diff --git a/src/mesa/drivers/dri/r200/r200_context.c b/src/mesa/drivers/dri/r200/r200_context.c
index bb7a16c9327..fc6eb93daa8 100644
--- a/src/mesa/drivers/dri/r200/r200_context.c
+++ b/src/mesa/drivers/dri/r200/r200_context.c
@@ -45,7 +45,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -60,7 +60,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_tex.h"
 #include "r200_swtcl.h"
 #include "r200_tcl.h"
-#include "r200_vtxfmt.h"
 #include "r200_maos.h"
 #include "r200_vertprog.h"
 
@@ -415,7 +414,7 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
    _ae_create_context( ctx );
@@ -424,11 +423,10 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
     */
    _tnl_destroy_pipeline( ctx );
    _tnl_install_pipeline( ctx, r200_pipeline );
-   ctx->Driver.FlushVertices = r200FlushVertices;
 
    /* Try and keep materials and vertices separate:
     */
-   _tnl_isolate_materials( ctx, GL_TRUE );
+/*    _tnl_isolate_materials( ctx, GL_TRUE ); */
 
 
    /* Configure swrast and TNL to match hardware characteristics:
@@ -533,12 +531,6 @@ GLboolean r200CreateContext( const __GLcontextModes *glVisual,
       TCL_FALLBACK(rmesa->glCtx, R200_TCL_FALLBACK_TCL_DISABLE, 1);
    }
 
-   if (rmesa->r200Screen->chip_flags & RADEON_CHIPSET_TCL) {
-      if (tcl_mode >= DRI_CONF_TCL_VTXFMT)
-	 r200VtxfmtInit( ctx, tcl_mode >= DRI_CONF_TCL_CODEGEN );
-
-      _tnl_need_dlist_norm_lengths( ctx, GL_FALSE );
-   }
    return GL_TRUE;
 }
 
@@ -568,7 +560,7 @@ void r200DestroyContext( __DRIcontextPrivate *driContextPriv )
       release_texture_heaps = (rmesa->glCtx->Shared->RefCount == 1);
       _swsetup_DestroyContext( rmesa->glCtx );
       _tnl_DestroyContext( rmesa->glCtx );
-      _ac_DestroyContext( rmesa->glCtx );
+      _vbo_DestroyContext( rmesa->glCtx );
       _swrast_DestroyContext( rmesa->glCtx );
 
       r200DestroySwtcl( rmesa->glCtx );
@@ -579,12 +571,6 @@ void r200DestroyContext( __DRIcontextPrivate *driContextPriv )
 	 r200FlushCmdBuf( rmesa, __FUNCTION__ );
       }
 
-      if (!(rmesa->TclFallback & R200_TCL_FALLBACK_TCL_DISABLE)) {
-	 int tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode");
-	 if (tcl_mode >= DRI_CONF_TCL_VTXFMT)
-	    r200VtxfmtDestroy( rmesa->glCtx );
-      }
-
       if (rmesa->state.scissor.pClipRects) {
 	 FREE(rmesa->state.scissor.pClipRects);
 	 rmesa->state.scissor.pClipRects = NULL;
@@ -700,9 +686,6 @@ r200MakeCurrent( __DRIcontextPrivate *driContextPriv,
 			  (GLframebuffer *) driDrawPriv->driverPrivate,
 			  (GLframebuffer *) driReadPriv->driverPrivate );
 
-      if (newCtx->vb.enabled)
-	 r200VtxfmtMakeCurrent( newCtx->glCtx );
-
       _mesa_update_state( newCtx->glCtx );
       r200ValidateState( newCtx->glCtx );
 
@@ -727,6 +710,5 @@ r200UnbindContext( __DRIcontextPrivate *driContextPriv )
    if (R200_DEBUG & DEBUG_DRI)
       fprintf(stderr, "%s ctx %p\n", __FUNCTION__, (void *)rmesa->glCtx);
 
-   r200VtxfmtUnbindContext( rmesa->glCtx );
    return GL_TRUE;
 }
diff --git a/src/mesa/drivers/dri/r200/r200_context.h b/src/mesa/drivers/dri/r200/r200_context.h
index fa38a78e263..e840a502c0b 100644
--- a/src/mesa/drivers/dri/r200/r200_context.h
+++ b/src/mesa/drivers/dri/r200/r200_context.h
@@ -107,8 +107,7 @@ struct r200_vertex_program {
         VERTEX_SHADER_INSTRUCTION instr[R200_VSF_MAX_INST + 6];
         int pos_end;
         int inputs[VERT_ATTRIB_MAX];
-        int rev_inputs[16];
-        int gen_inputs_mapped;
+        GLubyte inputmap_rev[16];
         int native;
         int fogpidx;
         int fogmode;
@@ -724,8 +723,6 @@ struct r200_store {
 /* r200_tcl.c
  */
 struct r200_tcl_info {
-   GLuint vertex_format;
-   GLint last_offset;
    GLuint hw_primitive;
 
 /* hw can handle 12 components max */
@@ -735,13 +732,7 @@ struct r200_tcl_info {
    GLuint *Elts;
 
    struct r200_dma_region indexed_verts;
-   struct r200_dma_region obj;
-   struct r200_dma_region rgba;
-   struct r200_dma_region spec;
-   struct r200_dma_region fog;
-   struct r200_dma_region tex[R200_MAX_TEXTURE_UNITS];
-   struct r200_dma_region norm;
-   struct r200_dma_region generic[16];
+   struct r200_dma_region vertex_data[15];
 };
 
 
@@ -811,87 +802,6 @@ struct r200_ioctl {
 #define R200_MAX_PRIMS 64
 
 
-/* Want to keep a cache of these around.  Each is parameterized by
- * only a single value which has only a small range.  Only expect a
- * few, so just rescan the list each time?
- */
-struct dynfn {
-   struct dynfn *next, *prev;
-   int key[2];
-   char *code;
-};
-
-struct dfn_lists {
-   struct dynfn Vertex2f;
-   struct dynfn Vertex2fv;
-   struct dynfn Vertex3f;
-   struct dynfn Vertex3fv;
-   struct dynfn Color4ub;
-   struct dynfn Color4ubv;
-   struct dynfn Color3ub;
-   struct dynfn Color3ubv;
-   struct dynfn Color4f;
-   struct dynfn Color4fv;
-   struct dynfn Color3f;
-   struct dynfn Color3fv;
-   struct dynfn SecondaryColor3ubEXT;
-   struct dynfn SecondaryColor3ubvEXT;
-   struct dynfn SecondaryColor3fEXT;
-   struct dynfn SecondaryColor3fvEXT;
-   struct dynfn Normal3f;
-   struct dynfn Normal3fv;
-   struct dynfn TexCoord3f;
-   struct dynfn TexCoord3fv;
-   struct dynfn TexCoord2f;
-   struct dynfn TexCoord2fv;
-   struct dynfn TexCoord1f;
-   struct dynfn TexCoord1fv;
-   struct dynfn MultiTexCoord3fARB;
-   struct dynfn MultiTexCoord3fvARB;
-   struct dynfn MultiTexCoord2fARB;
-   struct dynfn MultiTexCoord2fvARB;
-   struct dynfn MultiTexCoord1fARB;
-   struct dynfn MultiTexCoord1fvARB;
-   struct dynfn FogCoordfEXT;
-   struct dynfn FogCoordfvEXT;
-};
-
-struct dfn_generators {
-   struct dynfn *(*Vertex2f)( GLcontext *, const int * );
-   struct dynfn *(*Vertex2fv)( GLcontext *, const int * );
-   struct dynfn *(*Vertex3f)( GLcontext *, const int * );
-   struct dynfn *(*Vertex3fv)( GLcontext *, const int * );
-   struct dynfn *(*Color4ub)( GLcontext *, const int * );
-   struct dynfn *(*Color4ubv)( GLcontext *, const int * );
-   struct dynfn *(*Color3ub)( GLcontext *, const int * );
-   struct dynfn *(*Color3ubv)( GLcontext *, const int * );
-   struct dynfn *(*Color4f)( GLcontext *, const int * );
-   struct dynfn *(*Color4fv)( GLcontext *, const int * );
-   struct dynfn *(*Color3f)( GLcontext *, const int * );
-   struct dynfn *(*Color3fv)( GLcontext *, const int * );
-   struct dynfn *(*SecondaryColor3ubEXT)( GLcontext *, const int * );
-   struct dynfn *(*SecondaryColor3ubvEXT)( GLcontext *, const int * );
-   struct dynfn *(*SecondaryColor3fEXT)( GLcontext *, const int * );
-   struct dynfn *(*SecondaryColor3fvEXT)( GLcontext *, const int * );
-   struct dynfn *(*Normal3f)( GLcontext *, const int * );
-   struct dynfn *(*Normal3fv)( GLcontext *, const int * );
-   struct dynfn *(*TexCoord3f)( GLcontext *, const int * );
-   struct dynfn *(*TexCoord3fv)( GLcontext *, const int * );
-   struct dynfn *(*TexCoord2f)( GLcontext *, const int * );
-   struct dynfn *(*TexCoord2fv)( GLcontext *, const int * );
-   struct dynfn *(*TexCoord1f)( GLcontext *, const int * );
-   struct dynfn *(*TexCoord1fv)( GLcontext *, const int * );
-   struct dynfn *(*MultiTexCoord3fARB)( GLcontext *, const int * );
-   struct dynfn *(*MultiTexCoord3fvARB)( GLcontext *, const int * );
-   struct dynfn *(*MultiTexCoord2fARB)( GLcontext *, const int * );
-   struct dynfn *(*MultiTexCoord2fvARB)( GLcontext *, const int * );
-   struct dynfn *(*MultiTexCoord1fARB)( GLcontext *, const int * );
-   struct dynfn *(*MultiTexCoord1fvARB)( GLcontext *, const int * );
-   struct dynfn *(*FogCoordfEXT)( GLcontext *, const int * );
-   struct dynfn *(*FogCoordfvEXT)( GLcontext *, const int * );
-};
-
-
 
 struct r200_prim {
    GLuint start;
@@ -912,43 +822,6 @@ struct r200_prim {
 
 #define R200_MAX_VERTEX_SIZE ((3*6)+11)
 
-struct r200_vbinfo {
-   GLint counter, initial_counter;
-   GLint *dmaptr;
-   void (*notify)( void );
-   GLint vertex_size;
-
-   union { float f; int i; r200_color_t color; } vertex[R200_MAX_VERTEX_SIZE];
-
-   GLfloat *normalptr;
-   GLfloat *floatcolorptr;
-   GLfloat *fogptr;
-   r200_color_t *colorptr;
-   GLfloat *floatspecptr;
-   r200_color_t *specptr;
-   GLfloat *texcoordptr[8];	/* 6 (TMU) + 2 for r200_vtxfmt_c.c when GL_TEXTURE6/7 */
-
-
-   GLenum *prim;		/* &ctx->Driver.CurrentExecPrimitive */
-   GLuint primflags;
-   GLboolean enabled;		/* *_NO_VTXFMT / *_NO_TCL env vars */
-   GLboolean installed;
-   GLboolean fell_back;
-   GLboolean recheck;
-   GLint nrverts;
-   GLuint vtxfmt_0, vtxfmt_1;
-
-   GLuint installed_vertex_format;
-   GLuint installed_color_3f_sz;
-
-   struct r200_prim primlist[R200_MAX_PRIMS];
-   int nrprims;
-
-   struct dfn_lists dfn_cache;
-   struct dfn_generators codegen;
-   GLvertexformat vtxfmt;
-};
-
 
 struct r200_context {
    GLcontext *glCtx;			/* Mesa context */
@@ -1040,10 +913,6 @@ struct r200_context {
     */
    struct r200_swtcl_info swtcl;
 
-   /* r200_vtxfmt.c
-    */
-   struct r200_vbinfo vb;
-
    /* Mirrors of some DRI state
     */
    struct r200_dri_mirror dri;
diff --git a/src/mesa/drivers/dri/r200/r200_maos.h b/src/mesa/drivers/dri/r200/r200_maos.h
index b9e4d3c2392..4998f67445c 100644
--- a/src/mesa/drivers/dri/r200/r200_maos.h
+++ b/src/mesa/drivers/dri/r200/r200_maos.h
@@ -38,7 +38,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "r200_context.h"
 
-extern void r200EmitArrays( GLcontext *ctx, GLuint inputs );
+extern void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev );
 extern void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs );
 
 #endif
diff --git a/src/mesa/drivers/dri/r200/r200_maos_arrays.c b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
index 39c1f68911f..6a6c30a2b0d 100644
--- a/src/mesa/drivers/dri/r200/r200_maos_arrays.c
+++ b/src/mesa/drivers/dri/r200/r200_maos_arrays.c
@@ -376,7 +376,7 @@ static void emit_vector( GLcontext *ctx,
 /* Emit any changed arrays to new GART memory, re-emit a packet to
  * update the arrays.  
  */
-void r200EmitArrays( GLcontext *ctx, GLuint inputs )
+void r200EmitArrays( GLcontext *ctx, GLubyte *vimap_rev )
 {
    r200ContextPtr rmesa = R200_CONTEXT( ctx );
    struct vertex_buffer *VB = &TNL_CONTEXT( ctx )->vb;
@@ -384,318 +384,130 @@ void r200EmitArrays( GLcontext *ctx, GLuint inputs )
    GLuint nr = 0;
    GLuint vfmt0 = 0, vfmt1 = 0;
    GLuint count = VB->Count;
-   GLuint i;
-   GLuint generic_in_mapped = 0;
-   struct r200_vertex_program *vp = NULL;
-
-   /* this looks way more complicated than necessary... */
-   if (ctx->VertexProgram._Enabled) {
-      vp = rmesa->curr_vp_hw;
-      generic_in_mapped = vp->gen_inputs_mapped;
-   }
-
-   if (inputs & VERT_BIT_POS) {
-      if (!rmesa->tcl.obj.buf) 
-	 emit_vector( ctx, 
-		      &rmesa->tcl.obj, 
-		      (char *)VB->ObjPtr->data,
-		      VB->ObjPtr->size,
-		      VB->ObjPtr->stride,
-		      count);
-
-      switch( VB->ObjPtr->size ) {
-      case 4: vfmt0 |= R200_VTX_W0;
-      case 3: vfmt0 |= R200_VTX_Z0;
-      case 2: 
-      default:
-	 break;
-      }
-      component[nr++] = &rmesa->tcl.obj;
-   }
-   else if (generic_in_mapped & (1 << 0)) {
-      int geninput = vp->rev_inputs[0] - VERT_ATTRIB_GENERIC0;
-      if (!rmesa->tcl.generic[geninput].buf) {
-         emit_vector( ctx,
-		      &(rmesa->tcl.generic[geninput]),
-		      (char *)VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->data,
-		      4,
-		      VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->stride,
-		      count );
-      }
-      component[nr++] = &rmesa->tcl.generic[geninput];
-      vfmt0 |=  R200_VTX_W0 | R200_VTX_Z0;
-   }
-
-   if (inputs & VERT_BIT_NORMAL) {
-      if (!rmesa->tcl.norm.buf)
-	 emit_vector( ctx, 
-		      &(rmesa->tcl.norm), 
-		      (char *)VB->NormalPtr->data,
-		      3,
-		      VB->NormalPtr->stride,
-		      count);
-
-      vfmt0 |= R200_VTX_N0;
-      component[nr++] = &rmesa->tcl.norm;
-   }
-
-   if (inputs & VERT_BIT_FOG) {
-      if (!rmesa->tcl.fog.buf) {
-	 if (ctx->VertexProgram._Enabled)
-	    emit_vector( ctx, 
-			 &(rmesa->tcl.fog), 
-			 (char *)VB->FogCoordPtr->data,
+   GLuint i, emitsize;
+
+   for ( i = 0; i < 15; i++ ) {
+      GLubyte attrib = vimap_rev[i];
+      if (attrib != 255) {
+	 switch (i) {
+	 case 0:
+	    emitsize = (VB->AttribPtr[attrib]->size);
+	    switch (emitsize) {
+	    case 4:
+	       vfmt0 |= R200_VTX_W0;
+	       /* fallthrough */
+	    case 3:
+	       vfmt0 |= R200_VTX_Z0;
+	       break;
+	    case 2:
+	       break;
+	    default: assert(0);
+	    }
+	    break;
+	 case 1:
+	    assert(attrib == VERT_ATTRIB_WEIGHT);
+	    emitsize = (VB->AttribPtr[attrib]->size);
+	    vfmt0 |= emitsize << R200_VTX_WEIGHT_COUNT_SHIFT;
+	    break;
+	 case 2:
+	    assert(attrib == VERT_ATTRIB_NORMAL);
+	    emitsize = 3;
+	    vfmt0 |= R200_VTX_N0;
+	    break;
+	 case 3:
+	    /* special handling to fix up fog. Will get us into trouble with vbos...*/
+	    assert(attrib == VERT_ATTRIB_FOG);
+	    if (!rmesa->tcl.vertex_data[i].buf) {
+	       if (ctx->VertexProgram._Enabled)
+		  emit_vector( ctx,
+			 &(rmesa->tcl.vertex_data[attrib]),
+			 (char *)VB->AttribPtr[attrib]->data,
 			 1,
-			 VB->FogCoordPtr->stride,
+			 VB->AttribPtr[attrib]->stride,
 			 count);
-	 else
-	    emit_vecfog( ctx, 
-			 &(rmesa->tcl.fog), 
-			 (char *)VB->FogCoordPtr->data,
-			 VB->FogCoordPtr->stride,
+	       else
+		  emit_vecfog( ctx,
+			 &(rmesa->tcl.vertex_data[attrib]),
+			 (char *)VB->AttribPtr[attrib]->data,
+			 VB->AttribPtr[attrib]->stride,
 			 count);
-      }
-
-      vfmt0 |= R200_VTX_DISCRETE_FOG;
-      component[nr++] = &rmesa->tcl.fog;
-   }
- 
-   if (inputs & VERT_BIT_COLOR0) {
-      int emitsize;
-
-      if (VB->ColorPtr[0]->size == 4 &&
-	  (VB->ColorPtr[0]->stride != 0 ||
-	   VB->ColorPtr[0]->data[0][3] != 1.0)) { 
-	 vfmt0 |= R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT; 
-	 emitsize = 4;
-      }
-      else { 
-	 vfmt0 |= R200_VTX_FP_RGB << R200_VTX_COLOR_0_SHIFT; 
-	 emitsize = 3;
-      }
-
-      if (!rmesa->tcl.rgba.buf)
-	 emit_vector( ctx, 
-		      &(rmesa->tcl.rgba), 
-		      (char *)VB->ColorPtr[0]->data,
-		      emitsize,
-		      VB->ColorPtr[0]->stride,
-		      count);
-
-      component[nr++] = &rmesa->tcl.rgba;
-   }
-/*	vfmt0 |= R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT;
-	emit_ubyte_rgba( ctx, &rmesa->tcl.rgba, 
-		(char *)VB->ColorPtr[0]->data, 4,
-		      VB->ColorPtr[0]->stride, count);*/
-   else if (generic_in_mapped & (1 << 2)) {
-      int geninput = vp->rev_inputs[2] - VERT_ATTRIB_GENERIC0;
-      if (!rmesa->tcl.generic[geninput].buf) {
-         emit_vector( ctx,
-		      &(rmesa->tcl.generic[geninput]),
-		      (char *)VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->data,
-		      4,
-		      VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->stride,
-		      count );
-      }
-      component[nr++] = &rmesa->tcl.generic[geninput];
-      vfmt0 |= R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT;
-   }
-
-
-   if (inputs & VERT_BIT_COLOR1) {
-      if (!rmesa->tcl.spec.buf) {
-	 emit_vector( ctx, 
-		      &rmesa->tcl.spec, 
-		      (char *)VB->SecondaryColorPtr[0]->data,
-		      3,
-		      VB->SecondaryColorPtr[0]->stride,
-		      count);
-      }
-
-      /* How does this work?
-       */
-      vfmt0 |= R200_VTX_FP_RGB << R200_VTX_COLOR_1_SHIFT; 
-      component[nr++] = &rmesa->tcl.spec;
-   }
-   else if (generic_in_mapped & (1 << 3)) {
-      int geninput = vp->rev_inputs[3] - VERT_ATTRIB_GENERIC0;
-      if (!rmesa->tcl.generic[geninput].buf) {
-         emit_vector( ctx,
-		      &(rmesa->tcl.generic[geninput]),
-		      (char *)VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->data,
-		      4,
-		      VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->stride,
-		      count );
-      }
-      component[nr++] = &rmesa->tcl.generic[geninput];
-      vfmt0 |= R200_VTX_FP_RGBA << R200_VTX_COLOR_1_SHIFT;
-   }
-
-   if (generic_in_mapped & (1 << 4)) {
-      int geninput = vp->rev_inputs[4] - VERT_ATTRIB_GENERIC0;
-      if (!rmesa->tcl.generic[geninput].buf) {
-         emit_vector( ctx,
-		      &(rmesa->tcl.generic[geninput]),
-		      (char *)VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->data,
-		      4,
-		      VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->stride,
-		      count );
-      }
-      component[nr++] = &rmesa->tcl.generic[geninput];
-      vfmt0 |= R200_VTX_FP_RGBA << R200_VTX_COLOR_2_SHIFT;
-   }
-
-   if (generic_in_mapped & (1 << 5)) {
-      int geninput = vp->rev_inputs[5] - VERT_ATTRIB_GENERIC0;
-      if (!rmesa->tcl.generic[geninput].buf) {
-         emit_vector( ctx,
-		      &(rmesa->tcl.generic[geninput]),
-		      (char *)VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->data,
-		      4,
-		      VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->stride,
-		      count );
-      }
-      component[nr++] = &rmesa->tcl.generic[geninput];
-      vfmt0 |= R200_VTX_FP_RGBA << R200_VTX_COLOR_3_SHIFT;
-   }
-
-   for ( i = 0 ; i < 6 ; i++ ) {
-      if (inputs & (VERT_BIT_TEX0 << i)) {
-	 if (!rmesa->tcl.tex[i].buf)
-	     emit_vector( ctx, 
-			  &(rmesa->tcl.tex[i]),
-			  (char *)VB->TexCoordPtr[i]->data,
-			  VB->TexCoordPtr[i]->size,
-			  VB->TexCoordPtr[i]->stride,
-			  count );
-
-	 vfmt1 |= VB->TexCoordPtr[i]->size << (i * 3);
-	 component[nr++] = &rmesa->tcl.tex[i];
-      }
-      else if (generic_in_mapped & (1 << (i + 6))) {
-	 int geninput = vp->rev_inputs[i + 6] - VERT_ATTRIB_GENERIC0;
-	 if (!rmesa->tcl.generic[geninput].buf) {
-            emit_vector( ctx,
-			 &(rmesa->tcl.generic[geninput]),
-			(char *)VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->data,
-			4,
-			VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->stride,
-			count );
-	 }
-	 component[nr++] = &rmesa->tcl.generic[geninput];
-	 vfmt1 |= 4 << (R200_VTX_TEX0_COMP_CNT_SHIFT + (i * 3));
-      }
-   }
-
-   if (generic_in_mapped & (1 << 13)) {
-      int geninput = vp->rev_inputs[13] - VERT_ATTRIB_GENERIC0;
-      if (!rmesa->tcl.generic[geninput].buf) {
-         emit_vector( ctx,
-		      &(rmesa->tcl.generic[geninput]),
-		      (char *)VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->data,
-		      4,
-		      VB->AttribPtr[geninput + VERT_ATTRIB_GENERIC0]->stride,
-		      count );
-      }
-      component[nr++] = &rmesa->tcl.generic[geninput];
-      vfmt0 |= R200_VTX_XY1 | R200_VTX_Z1 | R200_VTX_W1;
-   }
-
-/* doesn't work. Wrong order with mixed generic & conventional! */
-/*
-   if (ctx->VertexProgram._Enabled) {
-      int *vp_inputs = rmesa->curr_vp_hw->inputs;
-      for ( i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++ ) {
-	 if (inputs & (1 << i)) {
-	    int geninput = i - VERT_ATTRIB_GENERIC0;
-	    if (!rmesa->tcl.generic[geninput].buf) {
-	       emit_vector( ctx,
-			  &(rmesa->tcl.generic[geninput]),
-			  (char *)VB->AttribPtr[i]->data,
-			  4,
-			  VB->AttribPtr[i]->stride,
-			  count );
 	    }
-	    component[nr++] = &rmesa->tcl.generic[geninput];
-	    switch (vp_inputs[i]) {
-	    case 0:
-	       vfmt0 |=  R200_VTX_W0 | R200_VTX_Z0;
-	       break;
+	    vfmt0 |= R200_VTX_DISCRETE_FOG;
+	    goto after_emit;
+	    break;
+	 case 4:
+	 case 5:
+	 case 6:
+	 case 7:
+	    if (VB->AttribPtr[attrib]->size == 4 &&
+	       (VB->AttribPtr[attrib]->stride != 0 ||
+		VB->AttribPtr[attrib]->data[0][3] != 1.0)) emitsize = 4;
+	    else emitsize = 3;
+	    if (emitsize == 4)
+	       vfmt0 |= R200_VTX_FP_RGBA << (R200_VTX_COLOR_0_SHIFT + (i - 4) * 2);
+	    else {
+	       vfmt0 |= R200_VTX_FP_RGB << (R200_VTX_COLOR_0_SHIFT + (i - 4) * 2);
+	    }
+	    break;
+	 case 8:
+	 case 9:
+	 case 10:
+	 case 11:
+	 case 12:
+	 case 13:
+	    emitsize = VB->AttribPtr[attrib]->size;
+	    vfmt1 |= emitsize << (R200_VTX_TEX0_COMP_CNT_SHIFT + (i - 8) * 3);
+	    break;
+	 case 14:
+	    emitsize = VB->AttribPtr[attrib]->size >= 2 ? VB->AttribPtr[attrib]->size : 2;
+	    switch (emitsize) {
 	    case 2:
+	       vfmt0 |= R200_VTX_XY1;
+	       /* fallthrough */
 	    case 3:
+	       vfmt0 |= R200_VTX_Z1;
+	       /* fallthrough */
 	    case 4:
-	    case 5:
-	       vfmt0 |= R200_VTX_FP_RGBA << (R200_VTX_COLOR_0_SHIFT + (vp_inputs[i] - 2) * 2);
-	       break;
-	    case 6:
-	    case 7:
-	    case 8:
-	    case 9:
-	    case 10:
-	    case 11:
-	       vfmt1 |= 4 << (R200_VTX_TEX0_COMP_CNT_SHIFT + (vp_inputs[i] - 6) * 3);
-	       break;
-	    case 13:
-	       vfmt0 |= R200_VTX_XY1 | R200_VTX_Z1 | R200_VTX_W1;
-	       break;
-	    case 1:
-	    case 12:
-	    default:
-	       assert(0);
+	       vfmt0 |= R200_VTX_W1;
+	    break;
 	    }
+	 default:
+	    assert(0);
+	 }
+	 if (!rmesa->tcl.vertex_data[i].buf) {
+	    emit_vector( ctx,
+			 &(rmesa->tcl.vertex_data[i]),
+			 (char *)VB->AttribPtr[attrib]->data,
+			 emitsize,
+			 VB->AttribPtr[attrib]->stride,
+			 count );
 	 }
+after_emit:
+	 assert(nr < 12);
+	 component[nr++] = &rmesa->tcl.vertex_data[i];
       }
    }
-*/
 
    if (vfmt0 != rmesa->hw.vtx.cmd[VTX_VTXFMT_0] ||
-       vfmt1 != rmesa->hw.vtx.cmd[VTX_VTXFMT_1]) { 
-      R200_STATECHANGE( rmesa, vtx ); 
+       vfmt1 != rmesa->hw.vtx.cmd[VTX_VTXFMT_1]) {
+      R200_STATECHANGE( rmesa, vtx );
       rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = vfmt0;
       rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = vfmt1;
-   } 
+   }
 
    rmesa->tcl.nr_aos_components = nr;
-   rmesa->tcl.vertex_format = vfmt0;
 }
 
 
 void r200ReleaseArrays( GLcontext *ctx, GLuint newinputs )
 {
-   GLuint unit;
    r200ContextPtr rmesa = R200_CONTEXT( ctx );
 
-/*    if (R200_DEBUG & DEBUG_VERTS)  */
-/*       _tnl_print_vert_flags( __FUNCTION__, newinputs ); */
-
-   if (newinputs & VERT_BIT_POS) 
-     r200ReleaseDmaRegion( rmesa, &rmesa->tcl.obj, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_NORMAL) 
-      r200ReleaseDmaRegion( rmesa, &rmesa->tcl.norm, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_FOG) 
-      r200ReleaseDmaRegion( rmesa, &rmesa->tcl.fog, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_COLOR0) 
-      r200ReleaseDmaRegion( rmesa, &rmesa->tcl.rgba, __FUNCTION__ );
-
-   if (newinputs & VERT_BIT_COLOR1) 
-      r200ReleaseDmaRegion( rmesa, &rmesa->tcl.spec, __FUNCTION__ );
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (newinputs & VERT_BIT_TEX(unit))
-	 r200ReleaseDmaRegion( rmesa, &rmesa->tcl.tex[unit], __FUNCTION__ );
-   }
-
-   if (ctx->VertexProgram._Enabled) {
-      int i;
-      for (i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++) {
-	 if (newinputs & (1 << i))
-	    r200ReleaseDmaRegion( rmesa,
-	       &rmesa->tcl.generic[i - VERT_ATTRIB_GENERIC0], __FUNCTION__ );
-      }
+   /* only do it for changed inputs ? */
+   int i;
+   for (i = 0; i < 15; i++) {
+      if (newinputs & (1 << i))
+	 r200ReleaseDmaRegion( rmesa,
+	    &rmesa->tcl.vertex_data[i], __FUNCTION__ );
    }
-
 }
diff --git a/src/mesa/drivers/dri/r200/r200_state.c b/src/mesa/drivers/dri/r200/r200_state.c
index dc1fbef72ee..bdb487f2b93 100644
--- a/src/mesa/drivers/dri/r200/r200_state.c
+++ b/src/mesa/drivers/dri/r200/r200_state.c
@@ -43,7 +43,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "framebuffer.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -54,7 +54,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_tcl.h"
 #include "r200_tex.h"
 #include "r200_swtcl.h"
-#include "r200_vtxfmt.h"
 #include "r200_vertprog.h"
 
 #include "drirenderbuffer.h"
@@ -2536,23 +2535,24 @@ static void r200InvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _ae_invalidate_state( ctx, new_state );
    R200_CONTEXT(ctx)->NewGLState |= new_state;
-   r200VtxfmtInvalidate( ctx );
 }
 
 /* A hack.  The r200 can actually cope just fine with materials
- * between begin/ends, so fix this. But how ?
+ * between begin/ends, so fix this.
+ * Should map to inputs just like the generic vertex arrays for vertex progs.
+ * In theory there could still be too many and we'd still need a fallback.
  */
 static GLboolean check_material( GLcontext *ctx )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    GLint i;
 
-   for (i = _TNL_ATTRIB_MAT_FRONT_AMBIENT; 
-	i < _TNL_ATTRIB_MAT_BACK_INDEXES; 
+   for (i = _TNL_ATTRIB_MAT_FRONT_AMBIENT;
+	i < _TNL_ATTRIB_MAT_BACK_INDEXES;
 	i++)
       if (tnl->vb.AttribPtr[i] &&
 	  tnl->vb.AttribPtr[i]->stride)
@@ -2560,7 +2560,7 @@ static GLboolean check_material( GLcontext *ctx )
 
    return GL_FALSE;
 }
-      
+
 static void r200WrapRunPipeline( GLcontext *ctx )
 {
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
@@ -2574,7 +2574,7 @@ static void r200WrapRunPipeline( GLcontext *ctx )
    if (rmesa->NewGLState)
       r200ValidateState( ctx );
 
-   has_material = (ctx->Light.Enabled && check_material( ctx ));
+   has_material = !ctx->VertexProgram._Enabled && ctx->Light.Enabled && check_material( ctx );
 
    if (has_material) {
       TCL_FALLBACK( ctx, R200_TCL_FALLBACK_MATERIAL, GL_TRUE );
diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c
index d95a80c7bbc..b40d0bdcb7c 100644
--- a/src/mesa/drivers/dri/r200/r200_state_init.c
+++ b/src/mesa/drivers/dri/r200/r200_state_init.c
@@ -39,7 +39,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "api_arrayelt.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -50,7 +50,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_tcl.h"
 #include "r200_tex.h"
 #include "r200_swtcl.h"
-#include "r200_vtxfmt.h"
 
 #include "xmlpool.h"
 
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.c b/src/mesa/drivers/dri/r200/r200_swtcl.c
index c14a275f7a3..25d229d8ed6 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.c
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.c
@@ -48,7 +48,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
-#include "tnl/t_vtx_api.h"
 
 #include "r200_context.h"
 #include "r200_ioctl.h"
@@ -936,13 +935,6 @@ r200PointsBitmap( GLcontext *ctx, GLint px, GLint py,
 }
 
 
-void r200FlushVertices( GLcontext *ctx, GLuint flags )
-{
-   _tnl_FlushVertices( ctx, flags );
-
-   if (flags & FLUSH_STORED_VERTICES)
-      R200_NEWPRIM( R200_CONTEXT( ctx ) );
-}
 
 /**********************************************************************/
 /*                            Initialization.                         */
diff --git a/src/mesa/drivers/dri/r200/r200_swtcl.h b/src/mesa/drivers/dri/r200/r200_swtcl.h
index ce2b6b5f06b..ccf817988c5 100644
--- a/src/mesa/drivers/dri/r200/r200_swtcl.h
+++ b/src/mesa/drivers/dri/r200/r200_swtcl.h
@@ -42,7 +42,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 extern void r200InitSwtcl( GLcontext *ctx );
 extern void r200DestroySwtcl( GLcontext *ctx );
 
-extern void r200FlushVertices( GLcontext *ctx, GLuint flags );
 extern void r200ChooseRenderState( GLcontext *ctx );
 extern void r200ChooseVertexState( GLcontext *ctx );
 
diff --git a/src/mesa/drivers/dri/r200/r200_tcl.c b/src/mesa/drivers/dri/r200/r200_tcl.c
index 86edf7132ce..e0c32b26d92 100644
--- a/src/mesa/drivers/dri/r200/r200_tcl.c
+++ b/src/mesa/drivers/dri/r200/r200_tcl.c
@@ -40,7 +40,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "colormac.h"
 #include "light.h"
 
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 
@@ -384,8 +384,14 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
    r200ContextPtr rmesa = R200_CONTEXT(ctx);
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &tnl->vb;
-   GLuint inputs = 0;
    GLuint i;
+   GLubyte *vimap_rev;
+/* use hw fixed order for simplicity, pos 0, weight 1, normal 2, fog 3, 
+   color0 - color3 4-7, texcoord0 - texcoord5 8-13, pos 1 14. Must not use
+   more than 12 of those at the same time. */
+   GLubyte map_rev_fixed[15] = {255, 255, 255, 255, 255, 255, 255, 255,
+			    255, 255, 255, 255, 255, 255, 255};
+
 
    /* TODO: separate this from the swtnl pipeline 
     */
@@ -404,30 +410,40 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
       r200ValidateState( ctx );
 
    if (!ctx->VertexProgram._Enabled) {
-      inputs = VERT_BIT_POS | VERT_BIT_COLOR0;
    /* NOTE: inputs != tnl->render_inputs - these are the untransformed
     * inputs.
     */
+      map_rev_fixed[0] = VERT_ATTRIB_POS;
+      /* technically there is no reason we always need VA_COLOR0. In theory
+         could disable it depending on lighting, color materials, texturing... */
+      map_rev_fixed[4] = VERT_ATTRIB_COLOR0;
+
       if (ctx->Light.Enabled) {
-	 inputs |= VERT_BIT_NORMAL;
+	 map_rev_fixed[2] = VERT_ATTRIB_NORMAL;
       }
 
+      /* this also enables VA_COLOR1 when using separate specular
+         lighting model, which is unnecessary.
+         FIXME: OTOH, we're missing the case where a ATI_fragment_shader accesses
+         the secondary color (if lighting is disabled). The chip seems
+         misconfigured for that though elsewhere (tcl output, might lock up) */
       if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
-	 inputs |= VERT_BIT_COLOR1;
+	 map_rev_fixed[5] = VERT_ATTRIB_COLOR1;
       }
 
       if ( (ctx->Fog.FogCoordinateSource == GL_FOG_COORD) && ctx->Fog.Enabled ) {
-	 inputs |= VERT_BIT_FOG;
+	 map_rev_fixed[3] = VERT_ATTRIB_FOG;
       }
 
       for (i = 0 ; i < ctx->Const.MaxTextureUnits; i++) {
 	 if (ctx->Texture.Unit[i]._ReallyEnabled) {
 	    if (rmesa->TexGenNeedNormals[i]) {
-	       inputs |= VERT_BIT_NORMAL;
+	       map_rev_fixed[2] = VERT_ATTRIB_NORMAL;
 	    }
-	    inputs |= VERT_BIT_TEX(i);
+	    map_rev_fixed[8 + i] = VERT_ATTRIB_TEX0 + i;
 	 }
       }
+      vimap_rev = &map_rev_fixed[0];
    }
    else {
       /* vtx_tcl_output_vtxfmt_0/1 need to match configuration of "fragment
@@ -437,14 +453,8 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
 	 We only need to change compsel. */
       GLuint out_compsel = 0;
       GLuint vp_out = rmesa->curr_vp_hw->mesa_program.Base.OutputsWritten;
-#if 0
-      /* can't handle other inputs, generic attribs etc. currently - should never arrive here */
-      assert ((rmesa->curr_vp_hw->mesa_program.Base.InputsRead &
-	 ~(VERT_BIT_POS | VERT_BIT_NORMAL | VERT_BIT_COLOR0 | VERT_BIT_COLOR1 |
-	  VERT_BIT_FOG | VERT_BIT_TEX0 | VERT_BIT_TEX1 | VERT_BIT_TEX2 |
-	  VERT_BIT_TEX3 | VERT_BIT_TEX4 | VERT_BIT_TEX5)) == 0);
-#endif
-      inputs |= rmesa->curr_vp_hw->mesa_program.Base.InputsRead;
+
+      vimap_rev = &rmesa->curr_vp_hw->inputmap_rev[0];
       assert(vp_out & (1 << VERT_RESULT_HPOS));
       out_compsel = R200_OUTPUT_XYZW;
       if (vp_out & (1 << VERT_RESULT_COL0)) {
@@ -473,7 +483,7 @@ static GLboolean r200_run_tcl_render( GLcontext *ctx,
    /* Do the actual work:
     */
    r200ReleaseArrays( ctx, ~0 /* stage->changed_inputs */ );
-   r200EmitArrays( ctx, inputs );
+   r200EmitArrays( ctx, vimap_rev );
 
    rmesa->tcl.Elts = VB->Elts;
 
diff --git a/src/mesa/drivers/dri/r200/r200_vertprog.c b/src/mesa/drivers/dri/r200/r200_vertprog.c
index 899e84caa0f..4960d481d5d 100644
--- a/src/mesa/drivers/dri/r200/r200_vertprog.c
+++ b/src/mesa/drivers/dri/r200/r200_vertprog.c
@@ -405,7 +405,6 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
    int dofogfix = 0;
    int fog_temp_i = 0;
    int free_inputs;
-   int free_inputs_conv;
    int array_count = 0;
 
    vp->native = GL_FALSE;
@@ -461,7 +460,7 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
 
    /* for fogc, can't change mesa_vp, as it would hose swtnl, and exp with
       base e isn't directly available neither. */
-   if (mesa_vp->Base.OutputsWritten & VERT_RESULT_FOGC && !vp->fogpidx) {
+   if ((mesa_vp->Base.OutputsWritten & (1 << VERT_RESULT_FOGC)) && !vp->fogpidx) {
       struct gl_program_parameter_list *paramList;
       GLint tokens[6] = { STATE_FOG_PARAMS, 0, 0, 0, 0, 0 };
       paramList = mesa_vp->Base.Parameters;
@@ -477,6 +476,8 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
 
    for(i = 0; i < VERT_ATTRIB_MAX; i++)
       vp->inputs[i] = -1;
+   for(i = 0; i < 15; i++)
+      vp->inputmap_rev[i] = 255;
    free_inputs = 0x2ffd;
 
 /* fglrx uses fixed inputs as follows for conventional attribs.
@@ -499,39 +500,45 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
 /* may look different when using idx buf / input_route instead of se_vtx_fmt? */
    if (mesa_vp->Base.InputsRead & VERT_BIT_POS) {
       vp->inputs[VERT_ATTRIB_POS] = 0;
+      vp->inputmap_rev[0] = VERT_ATTRIB_POS;
       free_inputs &= ~(1 << 0);
       array_count++;
    }
    if (mesa_vp->Base.InputsRead & VERT_BIT_WEIGHT) {
-   /* we don't actually handle that later. Then again, we don't have to... */
       vp->inputs[VERT_ATTRIB_WEIGHT] = 12;
+      vp->inputmap_rev[1] = VERT_ATTRIB_WEIGHT;
       array_count++;
    }
    if (mesa_vp->Base.InputsRead & VERT_BIT_NORMAL) {
       vp->inputs[VERT_ATTRIB_NORMAL] = 1;
+      vp->inputmap_rev[2] = VERT_ATTRIB_NORMAL;
       array_count++;
    }
    if (mesa_vp->Base.InputsRead & VERT_BIT_COLOR0) {
       vp->inputs[VERT_ATTRIB_COLOR0] = 2;
+      vp->inputmap_rev[4] = VERT_ATTRIB_COLOR0;
       free_inputs &= ~(1 << 2);
       array_count++;
    }
    if (mesa_vp->Base.InputsRead & VERT_BIT_COLOR1) {
       vp->inputs[VERT_ATTRIB_COLOR1] = 3;
+      vp->inputmap_rev[5] = VERT_ATTRIB_COLOR1;
       free_inputs &= ~(1 << 3);
       array_count++;
    }
    if (mesa_vp->Base.InputsRead & VERT_BIT_FOG) {
       vp->inputs[VERT_ATTRIB_FOG] = 15; array_count++;
+      vp->inputmap_rev[3] = VERT_ATTRIB_FOG;
+      array_count++;
    }
    for (i = VERT_ATTRIB_TEX0; i <= VERT_ATTRIB_TEX5; i++) {
       if (mesa_vp->Base.InputsRead & (1 << i)) {
 	 vp->inputs[i] = i - VERT_ATTRIB_TEX0 + 6;
+	 vp->inputmap_rev[8 + i - VERT_ATTRIB_TEX0] = i;
 	 free_inputs &= ~(1 << (i - VERT_ATTRIB_TEX0 + 6));
 	 array_count++;
       }
    }
-   free_inputs_conv = free_inputs;
    /* using VERT_ATTRIB_TEX6/7 would be illegal */
    /* completely ignore aliasing? */
    for (i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++) {
@@ -550,13 +557,14 @@ static GLboolean r200_translate_vertex_program(GLcontext *ctx, struct r200_verte
 	    if (free_inputs & (1 << j)) {
 	       free_inputs &= ~(1 << j);
 	       vp->inputs[i] = j;
-	       vp->rev_inputs[j] = i;
+	       if (j == 0) vp->inputmap_rev[j] = i; /* mapped to pos */
+	       else if (j < 12) vp->inputmap_rev[j + 2] = i; /* mapped to col/tex */
+	       else vp->inputmap_rev[j + 1] = i; /* mapped to pos1 */
 	       break;
 	    }
 	 }
       }
    }
-   vp->gen_inputs_mapped = free_inputs ^ free_inputs_conv;
 
    if (!(mesa_vp->Base.OutputsWritten & (1 << VERT_RESULT_HPOS))) {
       if (R200_DEBUG & DEBUG_FALLBACKS) {
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt.c b/src/mesa/drivers/dri/r200/r200_vtxfmt.c
deleted file mode 100644
index d73fbbafd5f..00000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt.c
+++ /dev/null
@@ -1,1234 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt.c,v 1.4 2003/05/06 23:52:08 daenzer Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "r200_context.h"
-#include "r200_state.h"
-#include "r200_ioctl.h"
-#include "r200_tex.h"
-#include "r200_tcl.h"
-#include "r200_swtcl.h"
-#include "r200_vtxfmt.h"
-
-#include "api_noop.h"
-#include "api_arrayelt.h"
-#include "context.h"
-#include "mtypes.h"
-#include "enums.h"
-#include "glapi.h"
-#include "colormac.h"
-#include "light.h"
-#include "state.h"
-#include "vtxfmt.h"
-
-#include "tnl/tnl.h"
-#include "tnl/t_context.h"
-#include "tnl/t_array_api.h"
-#include "tnl/t_save_api.h"
-
-#include "dispatch.h"
-
-static void r200VtxFmtFlushVertices( GLcontext *, GLuint );
-
-static void count_func( const char *name,  struct dynfn *l )
-{
-   int i = 0;
-   struct dynfn *f;
-   foreach (f, l) i++;
-   if (i) fprintf(stderr, "%s: %d\n", name, i );
-}
-
-static void count_funcs( r200ContextPtr rmesa )
-{
-   count_func( "Vertex2f", &rmesa->vb.dfn_cache.Vertex2f );
-   count_func( "Vertex2fv", &rmesa->vb.dfn_cache.Vertex2fv );
-   count_func( "Vertex3f", &rmesa->vb.dfn_cache.Vertex3f );
-   count_func( "Vertex3fv", &rmesa->vb.dfn_cache.Vertex3fv );
-   count_func( "Color4ub", &rmesa->vb.dfn_cache.Color4ub ); 
-   count_func( "Color4ubv", &rmesa->vb.dfn_cache.Color4ubv ); 
-   count_func( "Color3ub", &rmesa->vb.dfn_cache.Color3ub ); 
-   count_func( "Color3ubv", &rmesa->vb.dfn_cache.Color3ubv ); 
-   count_func( "Color4f", &rmesa->vb.dfn_cache.Color4f );
-   count_func( "Color4fv", &rmesa->vb.dfn_cache.Color4fv );
-   count_func( "Color3f", &rmesa->vb.dfn_cache.Color3f );
-   count_func( "Color3fv", &rmesa->vb.dfn_cache.Color3fv );
-   count_func( "SecondaryColor3f", &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   count_func( "SecondaryColor3fv", &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   count_func( "SecondaryColor3ub", &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   count_func( "SecondaryColor3ubv", &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   count_func( "Normal3f", &rmesa->vb.dfn_cache.Normal3f );
-   count_func( "Normal3fv", &rmesa->vb.dfn_cache.Normal3fv );
-   count_func( "TexCoord3f", &rmesa->vb.dfn_cache.TexCoord3f );
-   count_func( "TexCoord3fv", &rmesa->vb.dfn_cache.TexCoord3fv );
-   count_func( "TexCoord2f", &rmesa->vb.dfn_cache.TexCoord2f );
-   count_func( "TexCoord2fv", &rmesa->vb.dfn_cache.TexCoord2fv );
-   count_func( "TexCoord1f", &rmesa->vb.dfn_cache.TexCoord1f );
-   count_func( "TexCoord1fv", &rmesa->vb.dfn_cache.TexCoord1fv );
-   count_func( "MultiTexCoord3fARB", &rmesa->vb.dfn_cache.MultiTexCoord3fARB );
-   count_func( "MultiTexCoord3fvARB", &rmesa->vb.dfn_cache.MultiTexCoord3fvARB );
-   count_func( "MultiTexCoord2fARB", &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   count_func( "MultiTexCoord2fvARB", &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   count_func( "MultiTexCoord1fARB", &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   count_func( "MultiTexCoord1fvARB", &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-/*   count_func( "FogCoordfEXT", &rmesa->vb.dfn_cache.FogCoordfEXT );
-   count_func( "FogCoordfvEXT", &rmesa->vb.dfn_cache.FogCoordfvEXT );*/
-}
-
-static void r200NewList( GLcontext *ctx, GLuint list, GLenum mode )
-{
-   VFMT_FALLBACK( __FUNCTION__ );
-   _tnl_NewList( ctx, list, mode );
-   return;
-}
-
-void r200_copy_to_current( GLcontext *ctx ) 
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   unsigned i;
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT);
-
-   if (rmesa->vb.vtxfmt_0 & R200_VTX_N0) {
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][0] = rmesa->vb.normalptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][1] = rmesa->vb.normalptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][2] = rmesa->vb.normalptr[2];
-   }
-
-   if (rmesa->vb.vtxfmt_0 & R200_VTX_DISCRETE_FOG) {
-      ctx->Current.Attrib[VERT_ATTRIB_FOG][0] = rmesa->vb.fogptr[0];
-   }
-
-   switch( VTX_COLOR(rmesa->vb.vtxfmt_0, 0) ) {
-   case R200_VTX_PK_RGBA:
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->red );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->green );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->blue );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->alpha );
-      break;
-
-   case R200_VTX_FP_RGB:
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = rmesa->vb.floatcolorptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = rmesa->vb.floatcolorptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = rmesa->vb.floatcolorptr[2];
-      break;
-
-   case R200_VTX_FP_RGBA:
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = rmesa->vb.floatcolorptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = rmesa->vb.floatcolorptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = rmesa->vb.floatcolorptr[2];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = rmesa->vb.floatcolorptr[3];
-      break;
-      
-   default:
-      break;
-   }
-      
-   if (VTX_COLOR(rmesa->vb.vtxfmt_0, 1) == R200_VTX_PK_RGBA) {
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][0] = UBYTE_TO_FLOAT( rmesa->vb.specptr->red );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][1] = UBYTE_TO_FLOAT( rmesa->vb.specptr->green );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][2] = UBYTE_TO_FLOAT( rmesa->vb.specptr->blue );
-   } 
-
-   for ( i = 0 ; i < ctx->Const.MaxTextureUnits ; i++ ) {
-      const unsigned count = VTX_TEXn_COUNT( rmesa->vb.vtxfmt_1, i );
-      GLfloat * const src = rmesa->vb.texcoordptr[i];
-
-      if ( count != 0 ) {
-	 switch( count ) {
-	 case 3:
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][1] = src[1];
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][2] = src[2];
-	    break;
-	 case 2:
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][1] = src[1];
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][2] = 0.0F;
-	    break;
-	 case 1:
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][1] = 0.0F;
-	    ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][2] = 0.0F;
-	    break;
-	 }
-
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][0] = src[0];
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX0+i][3] = 1.0F;
-      }
-   }
-
-   ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-}
-
-static GLboolean discreet_gl_prim[GL_POLYGON+1] = {
-   1,				/* 0 points */
-   1,				/* 1 lines */
-   0,				/* 2 line_strip */
-   0,				/* 3 line_loop */
-   1,				/* 4 tris */
-   0,				/* 5 tri_fan */
-   0,				/* 6 tri_strip */
-   1,				/* 7 quads */
-   0,				/* 8 quadstrip */
-   0,				/* 9 poly */
-};
-
-static void flush_prims( r200ContextPtr rmesa )
-{
-   int i,j;
-   struct r200_dma_region tmp = rmesa->dma.current;
-   
-   tmp.buf->refcount++;
-   tmp.aos_size = rmesa->vb.vertex_size;
-   tmp.aos_stride = rmesa->vb.vertex_size;
-   tmp.aos_start = GET_START(&tmp);
-
-   rmesa->dma.current.ptr = rmesa->dma.current.start += 
-      (rmesa->vb.initial_counter - rmesa->vb.counter) * 
-      rmesa->vb.vertex_size * 4; 
-
-   rmesa->tcl.vertex_format = rmesa->vb.vtxfmt_0;
-   rmesa->tcl.aos_components[0] = &tmp;
-   rmesa->tcl.nr_aos_components = 1;
-   rmesa->dma.flush = NULL;
-
-   /* Optimize the primitive list:
-    */
-   if (rmesa->vb.nrprims > 1) {
-      for (j = 0, i = 1 ; i < rmesa->vb.nrprims; i++) {
-	 int pj = rmesa->vb.primlist[j].prim & 0xf;
-	 int pi = rmesa->vb.primlist[i].prim & 0xf;
-      
-	 if (pj == pi && discreet_gl_prim[pj] &&
-	     rmesa->vb.primlist[i].start == rmesa->vb.primlist[j].end) {
-	    rmesa->vb.primlist[j].end = rmesa->vb.primlist[i].end;
-	 }
-	 else {
-	    j++;
-	    if (j != i) rmesa->vb.primlist[j] = rmesa->vb.primlist[i];
-	 }
-      }
-      rmesa->vb.nrprims = j+1;
-   }
-
-   if (rmesa->vb.vtxfmt_0 != rmesa->hw.vtx.cmd[VTX_VTXFMT_0] ||
-       rmesa->vb.vtxfmt_1 != rmesa->hw.vtx.cmd[VTX_VTXFMT_1]) { 
-      R200_STATECHANGE( rmesa, vtx ); 
-      rmesa->hw.vtx.cmd[VTX_VTXFMT_0] = rmesa->vb.vtxfmt_0;
-      rmesa->hw.vtx.cmd[VTX_VTXFMT_1] = rmesa->vb.vtxfmt_1;
-   } 
-
-
-   for (i = 0 ; i < rmesa->vb.nrprims; i++) {
-      if (R200_DEBUG & DEBUG_PRIMS)
-	 fprintf(stderr, "vtxfmt prim %d: %s %d..%d\n", i,
-		 _mesa_lookup_enum_by_nr( rmesa->vb.primlist[i].prim & 
-					  PRIM_MODE_MASK ),
-		 rmesa->vb.primlist[i].start,
-		 rmesa->vb.primlist[i].end);
-
-      if (rmesa->vb.primlist[i].start < rmesa->vb.primlist[i].end)
-	 r200EmitPrimitive( rmesa->glCtx,
-			    rmesa->vb.primlist[i].start,
-			    rmesa->vb.primlist[i].end,
-			    rmesa->vb.primlist[i].prim );
-   }
-
-   rmesa->vb.nrprims = 0;
-   r200ReleaseDmaRegion( rmesa, &tmp, __FUNCTION__ );
-}
-
-
-static void start_prim( r200ContextPtr rmesa, GLuint mode )
-{
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, 
-	      rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   rmesa->vb.primlist[rmesa->vb.nrprims].start = 
-      rmesa->vb.initial_counter - rmesa->vb.counter;
-   rmesa->vb.primlist[rmesa->vb.nrprims].prim = mode;
-}
-
-static void note_last_prim( r200ContextPtr rmesa, GLuint flags )
-{
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, 
-	      rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      rmesa->vb.primlist[rmesa->vb.nrprims].prim |= flags;
-      rmesa->vb.primlist[rmesa->vb.nrprims].end = 
-	 rmesa->vb.initial_counter - rmesa->vb.counter;
-
-      if (++(rmesa->vb.nrprims) == R200_MAX_PRIMS)
-	 flush_prims( rmesa );
-   }
-}
-
-
-static void copy_vertex( r200ContextPtr rmesa, GLuint n, GLfloat *dst )
-{
-   GLuint i;
-   GLfloat *src = (GLfloat *)(rmesa->dma.current.address + 
-			      rmesa->dma.current.ptr + 
-			      (rmesa->vb.primlist[rmesa->vb.nrprims].start + n) * 
-			      rmesa->vb.vertex_size * 4);
-
-   if (R200_DEBUG & DEBUG_VFMT) 
-      fprintf(stderr, "copy_vertex %d\n", rmesa->vb.primlist[rmesa->vb.nrprims].start + n);
-
-   for (i = 0 ; i < rmesa->vb.vertex_size; i++) {
-      dst[i] = src[i];
-   }
-}
-
-/* NOTE: This actually reads the copied vertices back from uncached
- * memory.  Could also use the counter/notify mechanism to populate
- * tmp on the fly as vertices are generated.  
- */
-static GLuint copy_dma_verts( r200ContextPtr rmesa, GLfloat (*tmp)[R200_MAX_VERTEX_SIZE] )
-{
-   GLuint ovf, i;
-   GLuint nr = (rmesa->vb.initial_counter - rmesa->vb.counter) - 
-      rmesa->vb.primlist[rmesa->vb.nrprims].start;
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d verts\n", __FUNCTION__, nr);
-
-   switch( rmesa->vb.prim[0] )
-   {
-   case GL_POINTS:
-      return 0;
-   case GL_LINES:
-      ovf = nr&1;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_TRIANGLES:
-      ovf = nr%3;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_QUADS:
-      ovf = nr&3;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_LINE_STRIP:
-      if (nr == 0) 
-	 return 0;
-      copy_vertex( rmesa, nr-1, tmp[0] );
-      return 1;
-   case GL_LINE_LOOP:
-   case GL_TRIANGLE_FAN:
-   case GL_POLYGON:
-      if (nr == 0) 
-	 return 0;
-      else if (nr == 1) {
-	 copy_vertex( rmesa, 0, tmp[0] );
-	 return 1;
-      } else {
-	 copy_vertex( rmesa, 0, tmp[0] );
-	 copy_vertex( rmesa, nr-1, tmp[1] );
-	 return 2;
-      }
-   case GL_TRIANGLE_STRIP:
-      ovf = MIN2( nr, 2 );
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_QUAD_STRIP:
-      switch (nr) {
-      case 0: ovf = 0; break;
-      case 1: ovf = 1; break;
-      default: ovf = 2 + (nr&1); break;
-      }
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
-static void VFMT_FALLBACK_OUTSIDE_BEGIN_END( const char *caller )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & (DEBUG_VFMT|DEBUG_FALLBACKS))
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-   if (ctx->Driver.NeedFlush) 
-      r200VtxFmtFlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   if (ctx->NewState)
-      _mesa_update_state( ctx ); /* clear state so fell_back sticks */
-
-   _tnl_wakeup_exec( ctx );
-   ctx->Driver.FlushVertices = r200FlushVertices;
-   ctx->Driver.NewList = _tnl_NewList;
-
-   assert( rmesa->dma.flush == 0 );
-   rmesa->vb.fell_back = GL_TRUE;
-   rmesa->vb.installed = GL_FALSE;
-}
-
-
-/**
- * \todo
- * An interesting optimization of this function would be to have 3 element
- * table with the dispatch offsets of the TexCoord?fv functions, use count
- * to look-up the table, and a specialized version of GL_CALL that used the
- * offset number instead of the name.
- */
-static void dispatch_multitexcoord( GLuint count, GLuint unit, GLfloat * f )
-{
-   switch( count ) {
-   case 3:
-      CALL_MultiTexCoord3fvARB(GET_DISPATCH(), (GL_TEXTURE0+unit, f));
-      break;
-   case 2:
-      CALL_MultiTexCoord2fvARB(GET_DISPATCH(), (GL_TEXTURE0+unit, f));
-      break;
-   case 1:
-      CALL_MultiTexCoord1fvARB(GET_DISPATCH(), (GL_TEXTURE0+unit, f));
-      break;
-   default:
-      assert( count == 0 );
-      break;
-   }
-}
-
-void VFMT_FALLBACK( const char *caller )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat tmp[3][R200_MAX_VERTEX_SIZE];
-   GLuint i, prim;
-   GLuint ind0 = rmesa->vb.vtxfmt_0;
-   GLuint ind1 = rmesa->vb.vtxfmt_1;
-   GLuint nrverts;
-   GLfloat alpha = 1.0;
-   GLuint count;
-   GLuint unit;
-
-   if (R200_DEBUG & (DEBUG_FALLBACKS|DEBUG_VFMT))
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) {
-      VFMT_FALLBACK_OUTSIDE_BEGIN_END( __FUNCTION__ );
-      return;
-   }
-
-   /* Copy vertices out of dma:
-    */
-   nrverts = copy_dma_verts( rmesa, tmp );
-
-   /* Finish the prim at this point:
-    */
-   note_last_prim( rmesa, 0 );
-   flush_prims( rmesa );
-
-   /* Update ctx->Driver.CurrentExecPrimitive and swap in swtnl. 
-    */
-   prim = rmesa->vb.prim[0];
-   ctx->Driver.CurrentExecPrimitive = GL_POLYGON+1;
-   _tnl_wakeup_exec( ctx );
-   ctx->Driver.FlushVertices = r200FlushVertices;
-
-   assert(rmesa->dma.flush == 0);
-   rmesa->vb.fell_back = GL_TRUE;
-   rmesa->vb.installed = GL_FALSE;
-   CALL_Begin(GET_DISPATCH(), (prim));
-
-   if (rmesa->vb.installed_color_3f_sz == 4)
-      alpha = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3];
-
-   /* Replay saved vertices
-    */
-   for (i = 0 ; i < nrverts; i++) {
-      GLuint offset = 3;
-
-      if (ind0 & R200_VTX_N0) {
-	 CALL_Normal3fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset += 3;
-      }
-
-      if (ind0 & R200_VTX_DISCRETE_FOG) {
-	 CALL_FogCoordfvEXT(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset++;
-      }
-
-      if (VTX_COLOR(ind0, 0) == R200_VTX_PK_RGBA) {
-	 CALL_Color4ubv(GET_DISPATCH(), ((GLubyte *)&tmp[i][offset]));
-	 offset++;
-      }
-      else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGBA) {
-	 CALL_Color4fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset+=4;
-      } 
-      else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGB) {
-	 CALL_Color3fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset+=3;
-      }
-
-      if (VTX_COLOR(ind0, 1) == R200_VTX_PK_RGBA) {
-	 CALL_SecondaryColor3ubvEXT(GET_DISPATCH(), ((GLubyte *)&tmp[i][offset]));
-	 offset++;
-      }
-
-      for ( unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++ ) {
-	 count = VTX_TEXn_COUNT( ind1, unit );
-	 dispatch_multitexcoord( count, unit, &tmp[i][offset] );
-	 offset += count;
-      }
-
-      CALL_Vertex3fv(GET_DISPATCH(), (&tmp[i][0]));
-   }
-
-   /* Replay current vertex
-    */
-   if (ind0 & R200_VTX_N0) 
-      CALL_Normal3fv(GET_DISPATCH(), (rmesa->vb.normalptr));
-   if (ind0 & R200_VTX_DISCRETE_FOG) {
-      CALL_FogCoordfvEXT(GET_DISPATCH(), (rmesa->vb.fogptr));
-   }
-
-   if (VTX_COLOR(ind0, 0) == R200_VTX_PK_RGBA) {
-      CALL_Color4ub(GET_DISPATCH(), (rmesa->vb.colorptr->red,
-				     rmesa->vb.colorptr->green,
-				     rmesa->vb.colorptr->blue,
-				     rmesa->vb.colorptr->alpha));
-   }
-   else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGBA) {
-      CALL_Color4fv(GET_DISPATCH(), (rmesa->vb.floatcolorptr));
-   }
-   else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGB) {
-      if (rmesa->vb.installed_color_3f_sz == 4 && alpha != 1.0) {
-	 CALL_Color4f(GET_DISPATCH(), (rmesa->vb.floatcolorptr[0],
-				       rmesa->vb.floatcolorptr[1],
-				       rmesa->vb.floatcolorptr[2],
-				       alpha));
-      }
-      else {
-	 CALL_Color3fv(GET_DISPATCH(), (rmesa->vb.floatcolorptr));
-      }
-   }
-
-   if (VTX_COLOR(ind0, 1) == R200_VTX_PK_RGBA) 
-       CALL_SecondaryColor3ubEXT(GET_DISPATCH(), (rmesa->vb.specptr->red, 
-						  rmesa->vb.specptr->green,
-						  rmesa->vb.specptr->blue));
-
-   for ( unit = 0 ; unit < ctx->Const.MaxTextureUnits ; unit++ ) {
-      count = VTX_TEXn_COUNT( ind1, unit );
-      dispatch_multitexcoord( count, unit, rmesa->vb.texcoordptr[unit] );
-   }
-}
-
-
-
-static void wrap_buffer( void )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat tmp[3][R200_MAX_VERTEX_SIZE];
-   GLuint i, nrverts;
-
-   if (R200_DEBUG & (DEBUG_VFMT|DEBUG_PRIMS))
-      fprintf(stderr, "%s %d\n", __FUNCTION__,
-	      rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   /* Don't deal with parity.
-    */
-   if ((((rmesa->vb.initial_counter - rmesa->vb.counter) -  
-	 rmesa->vb.primlist[rmesa->vb.nrprims].start) & 1)) {
-      rmesa->vb.counter++;
-      rmesa->vb.initial_counter++;
-      return;
-   }
-
-   /* Copy vertices out of dma:
-    */
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) 
-      nrverts = 0;
-   else {
-      nrverts = copy_dma_verts( rmesa, tmp );
-
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%d vertices to copy\n", nrverts);
-   
-      /* Finish the prim at this point:
-       */
-      note_last_prim( rmesa, 0 );
-   }
-
-   /* Fire any buffered primitives
-    */
-   flush_prims( rmesa );
-
-   /* Get new buffer
-    */
-   r200RefillCurrentDmaRegion( rmesa );
-
-   /* Reset counter, dmaptr
-    */
-   rmesa->vb.dmaptr = (int *)(rmesa->dma.current.ptr + rmesa->dma.current.address);
-   rmesa->vb.counter = (rmesa->dma.current.end - rmesa->dma.current.ptr) / 
-      (rmesa->vb.vertex_size * 4);
-   rmesa->vb.counter--;
-   rmesa->vb.initial_counter = rmesa->vb.counter;
-   rmesa->vb.notify = wrap_buffer;
-
-   rmesa->dma.flush = flush_prims;
-
-   /* Restart wrapped primitive:
-    */
-   if (rmesa->vb.prim[0] != GL_POLYGON+1)
-      start_prim( rmesa, rmesa->vb.prim[0] );
-
-
-   /* Reemit saved vertices
-    */
-   for (i = 0 ; i < nrverts; i++) {
-      if (R200_DEBUG & DEBUG_VERTS) {
-	 int j;
-	 fprintf(stderr, "re-emit vertex %d to %p\n", i, 
-		 (void *)rmesa->vb.dmaptr);
-	 if (R200_DEBUG & DEBUG_VERBOSE)
-	    for (j = 0 ; j < rmesa->vb.vertex_size; j++) 
-	       fprintf(stderr, "\t%08x/%f\n", *(int*)&tmp[i][j], tmp[i][j]);
-      }
-
-      memcpy( rmesa->vb.dmaptr, tmp[i], rmesa->vb.vertex_size * 4 );
-      rmesa->vb.dmaptr += rmesa->vb.vertex_size;
-      rmesa->vb.counter--;
-   }
-}
-
-
-/**
- * Determines the hardware vertex format based on the current state vector.
- * 
- * \returns
- * If the hardware TCL unit is capable of handling the current state vector,
- * \c GL_TRUE is returned.  Otherwise, \c GL_FALSE is returned.
- *
- * \todo
- * Make this color format selection data driven.  If we receive only ubytes,
- * send color as ubytes.  Also check if converting (with free checking for
- * overflow) is cheaper than sending floats directly.
- *
- * \todo
- * When intializing texture coordinates, it might be faster to just copy the
- * entire \c VERT_ATTRIB_TEX0 vector into the vertex buffer.  It may mean that
- * some of the data (i.e., the last texture coordinate components) get copied
- * over, but that still may be faster than the conditional branching.  If
- * nothing else, the code will be smaller and easier to follow.
- */
-static GLboolean check_vtx_fmt( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLuint ind0 = R200_VTX_Z0;
-   GLuint ind1 = 0;
-   GLuint i;
-   GLuint count[R200_MAX_TEXTURE_UNITS];
-
-   if (rmesa->TclFallback || rmesa->vb.fell_back || ctx->CompileFlag ||
-      (ctx->Fog.Enabled && (ctx->Fog.FogCoordinateSource == GL_FOG_COORD)) ||
-      /* TODO: set tcl out fmt/compsel and reenable vtxfmt code */
-      ctx->VertexProgram._Enabled)
-      return GL_FALSE;
-
-   if (ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) 
-      ctx->Driver.FlushVertices( ctx, FLUSH_UPDATE_CURRENT );
-   
-   /* Make all this event-driven:
-    */
-   if (ctx->Light.Enabled) {
-      ind0 |= R200_VTX_N0;
-
-      if (ctx->Light.ColorMaterialEnabled) 
-	 ind0 |= R200_VTX_FP_RGBA << R200_VTX_COLOR_0_SHIFT;
-      else
-	 ind0 |= R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT;
-   }
-   else {
-      /* TODO: make this data driven?
-       */
-      ind0 |= R200_VTX_PK_RGBA << R200_VTX_COLOR_0_SHIFT;
-	 
-      if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
-	 ind0 |= R200_VTX_PK_RGBA << R200_VTX_COLOR_1_SHIFT;
-      }
-   }
-
-   if ( ctx->Fog.FogCoordinateSource == GL_FOG_COORD ) {
-      ind0 |= R200_VTX_DISCRETE_FOG;
-   }
-
-   for ( i = 0 ; i < ctx->Const.MaxTextureUnits ; i++ ) {
-      count[i] = 0;
-
-      if (ctx->Texture.Unit[i]._ReallyEnabled) {
-	 if (rmesa->TexGenNeedNormals[i]) {
-	    ind0 |= R200_VTX_N0;
-	 }
-	 else {
-	    switch( ctx->Texture.Unit[i]._ReallyEnabled ) {
-	    case TEXTURE_CUBE_BIT:
-	    case TEXTURE_3D_BIT:
-	       count[i] = 3;
-	       break;
-	    case TEXTURE_2D_BIT:
-	    case TEXTURE_RECT_BIT:
-	       count[i] = 2;
-	       break;
-	    case TEXTURE_1D_BIT:
-	       count[i] = 1;
-	       break;
-	    }
-
-	    ind1 |= count[i] << (3 * i);
-	 }
-      }
-   }
-
-   if (R200_DEBUG & (DEBUG_VFMT|DEBUG_STATE))
-      fprintf(stderr, "%s: format: 0x%x, 0x%x\n", __FUNCTION__, ind0, ind1 );
-
-   R200_NEWPRIM(rmesa);
-   rmesa->vb.vtxfmt_0 = ind0;
-   rmesa->vb.vtxfmt_1 = ind1;
-   rmesa->vb.prim = &ctx->Driver.CurrentExecPrimitive;
-
-   rmesa->vb.vertex_size = 3;
-   rmesa->vb.normalptr = ctx->Current.Attrib[VERT_ATTRIB_NORMAL];
-   rmesa->vb.colorptr = NULL;
-   rmesa->vb.floatcolorptr = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
-   rmesa->vb.fogptr = ctx->Current.Attrib[VERT_ATTRIB_FOG];
-   rmesa->vb.specptr = NULL;
-   rmesa->vb.floatspecptr = ctx->Current.Attrib[VERT_ATTRIB_COLOR1];
-   rmesa->vb.texcoordptr[0] = ctx->Current.Attrib[VERT_ATTRIB_TEX0];
-   rmesa->vb.texcoordptr[1] = ctx->Current.Attrib[VERT_ATTRIB_TEX1];
-   rmesa->vb.texcoordptr[2] = ctx->Current.Attrib[VERT_ATTRIB_TEX2];
-   rmesa->vb.texcoordptr[3] = ctx->Current.Attrib[VERT_ATTRIB_TEX3];
-   rmesa->vb.texcoordptr[4] = ctx->Current.Attrib[VERT_ATTRIB_TEX4];
-   rmesa->vb.texcoordptr[5] = ctx->Current.Attrib[VERT_ATTRIB_TEX5];
-   rmesa->vb.texcoordptr[6] = ctx->Current.Attrib[VERT_ATTRIB_TEX0];	/* dummy */
-   rmesa->vb.texcoordptr[7] = ctx->Current.Attrib[VERT_ATTRIB_TEX0];	/* dummy */
-
-   /* Run through and initialize the vertex components in the order
-    * the hardware understands:
-    */
-   if (ind0 & R200_VTX_N0) {
-      rmesa->vb.normalptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 3;
-      rmesa->vb.normalptr[0] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][0];
-      rmesa->vb.normalptr[1] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][1];
-      rmesa->vb.normalptr[2] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][2];
-   }
-
-   if (ind0 & R200_VTX_DISCRETE_FOG) {
-      rmesa->vb.fogptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 1;
-      rmesa->vb.fogptr[0] = ctx->Current.Attrib[VERT_ATTRIB_FOG][0];
-   }
-
-   if (VTX_COLOR(ind0, 0) == R200_VTX_PK_RGBA) {
-      rmesa->vb.colorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].color;
-      rmesa->vb.vertex_size += 1;
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->red,   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->green, ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->blue,  ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->alpha, ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] );
-   }
-   else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGBA) {
-      rmesa->vb.floatcolorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 4;
-      rmesa->vb.floatcolorptr[0] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0];
-      rmesa->vb.floatcolorptr[1] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1];
-      rmesa->vb.floatcolorptr[2] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2];
-      rmesa->vb.floatcolorptr[3] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3];
-   }
-   else if (VTX_COLOR(ind0, 0) == R200_VTX_FP_RGB) {
-      rmesa->vb.floatcolorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 3;
-      rmesa->vb.floatcolorptr[0] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0];
-      rmesa->vb.floatcolorptr[1] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1];
-      rmesa->vb.floatcolorptr[2] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2];
-   }   
-   
-   if (VTX_COLOR(ind0, 1) == R200_VTX_PK_RGBA) {
-      rmesa->vb.specptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].color;
-      rmesa->vb.vertex_size += 1;
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->red,   ctx->Current.Attrib[VERT_ATTRIB_COLOR1][0] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->green, ctx->Current.Attrib[VERT_ATTRIB_COLOR1][1] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->blue,  ctx->Current.Attrib[VERT_ATTRIB_COLOR1][2] );
-   }
-
-
-   for ( i = 0 ; i < ctx->Const.MaxTextureUnits ; i++ ) {
-      if ( count[i] != 0 ) {
-	 float * const attr = ctx->Current.Attrib[VERT_ATTRIB_TEX0+i];
-	 unsigned  j;
-
-	 rmesa->vb.texcoordptr[i] = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-
-	 for ( j = 0 ; j < count[i] ; j++ ) {
-	    rmesa->vb.texcoordptr[i][j] = attr[j];
-	 }
-
-	 rmesa->vb.vertex_size += count[i];
-      }
-   }
-
-   if (rmesa->vb.installed_vertex_format != rmesa->vb.vtxfmt_0) {
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "reinstall on vertex_format change\n");
-      _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-      rmesa->vb.installed_vertex_format = rmesa->vb.vtxfmt_0;
-   }
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s -- success\n", __FUNCTION__);
-
-   return GL_TRUE;
-}
-
-
-void r200VtxfmtInvalidate( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   rmesa->vb.recheck = GL_TRUE;
-   rmesa->vb.fell_back = GL_FALSE;
-}
-
-
-static void r200VtxfmtValidate( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (ctx->Driver.NeedFlush)
-      ctx->Driver.FlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   rmesa->vb.recheck = GL_FALSE;
-
-   if (check_vtx_fmt( ctx )) {
-      if (!rmesa->vb.installed) {
-	 if (R200_DEBUG & DEBUG_VFMT)
-	    fprintf(stderr, "reinstall (new install)\n");
-
-	 _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-	 ctx->Driver.FlushVertices = r200VtxFmtFlushVertices;
-	 ctx->Driver.NewList = r200NewList;
-	 rmesa->vb.installed = GL_TRUE;
-      }
-      else if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: already installed", __FUNCTION__);
-   } 
-   else {
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: failed\n", __FUNCTION__);
-
-      if (rmesa->vb.installed) {
-	 if (rmesa->dma.flush)
-	    rmesa->dma.flush( rmesa );
-	 _tnl_wakeup_exec( ctx );
-	 ctx->Driver.FlushVertices = r200FlushVertices;
-	 ctx->Driver.NewList =_tnl_NewList;
-	 rmesa->vb.installed = GL_FALSE;
-      }
-   }      
-}
-
-
-
-/* Materials:
- */
-static void r200_Materialfv( GLenum face, GLenum pname, 
-			       const GLfloat *params )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      VFMT_FALLBACK( __FUNCTION__ );
-      CALL_Materialfv(GET_DISPATCH(), (face, pname, params));
-      return;
-   }
-   _mesa_noop_Materialfv( face, pname, params );
-   r200UpdateMaterial( ctx );
-}
-
-
-/* Begin/End
- */
-static void r200_Begin( GLenum mode )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s( %s )\n", __FUNCTION__,
-	      _mesa_lookup_enum_by_nr( mode ));
-
-   if (mode > GL_POLYGON) {
-      _mesa_error( ctx, GL_INVALID_ENUM, "glBegin" );
-      return;
-   }
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glBegin" );
-      return;
-   }
-   
-   if (ctx->NewState) 
-      _mesa_update_state( ctx );
-
-   if (rmesa->NewGLState)
-      r200ValidateState( ctx );
-
-   if (rmesa->vb.recheck) 
-      r200VtxfmtValidate( ctx );
-
-   if (!rmesa->vb.installed) {
-      CALL_Begin(GET_DISPATCH(), (mode));
-      return;
-   }
-
-
-   if (rmesa->dma.flush && rmesa->vb.counter < 12) {
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: flush almost-empty buffers\n", __FUNCTION__);
-      flush_prims( rmesa );
-   }
-
-   /* Need to arrange to save vertices here?  Or always copy from dma (yuk)?
-    */
-   if (!rmesa->dma.flush) {
-      if (rmesa->dma.current.ptr + 12*rmesa->vb.vertex_size*4 > 
-	  rmesa->dma.current.end) {
-	 R200_NEWPRIM( rmesa );
-	 r200RefillCurrentDmaRegion( rmesa );
-      }
-
-      rmesa->vb.dmaptr = (int *)(rmesa->dma.current.address + rmesa->dma.current.ptr);
-      rmesa->vb.counter = (rmesa->dma.current.end - rmesa->dma.current.ptr) / 
-	 (rmesa->vb.vertex_size * 4);
-      rmesa->vb.counter--;
-      rmesa->vb.initial_counter = rmesa->vb.counter;
-      rmesa->vb.notify = wrap_buffer;
-      rmesa->dma.flush = flush_prims;
-      ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   }
-   
-   
-   rmesa->vb.prim[0] = mode;
-   start_prim( rmesa, mode | PRIM_BEGIN );
-}
-
-
-
-static void r200_End( void )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) {
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glEnd" );
-      return;
-   }
-
-   note_last_prim( rmesa, PRIM_END );
-   rmesa->vb.prim[0] = GL_POLYGON+1;
-}
-
-
-/* Fallback on difficult entrypoints:
- */
-#define PRE_LOOPBACK( FUNC )			\
-do {						\
-   if (R200_DEBUG & DEBUG_VFMT) 		\
-      fprintf(stderr, "%s\n", __FUNCTION__);	\
-   VFMT_FALLBACK( __FUNCTION__ );		\
-} while (0)
-#define TAG(x) r200_fallback_##x
-#include "vtxfmt_tmp.h"
-
-
-
-static GLboolean r200NotifyBegin( GLcontext *ctx, GLenum p )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-   
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(!rmesa->vb.installed);
-
-   if (ctx->NewState) 
-      _mesa_update_state( ctx );
-
-   if (rmesa->NewGLState)
-      r200ValidateState( ctx );
-
-   if (ctx->Driver.NeedFlush)
-      ctx->Driver.FlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   if (rmesa->vb.recheck) 
-      r200VtxfmtValidate( ctx );
-
-   if (!rmesa->vb.installed) {
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s -- failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   r200_Begin( p );
-   return GL_TRUE;
-}
-
-static void r200VtxFmtFlushVertices( GLcontext *ctx, GLuint flags )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   if (R200_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(rmesa->vb.installed);
-
-   if (flags & FLUSH_UPDATE_CURRENT) {
-      r200_copy_to_current( ctx );
-      if (R200_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "reinstall on update_current\n");
-      _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-      ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-   }
-
-   if (flags & FLUSH_STORED_VERTICES) {
-      assert (rmesa->dma.flush == 0 ||
-	      rmesa->dma.flush == flush_prims);
-      if (rmesa->dma.flush == flush_prims)
-	 flush_prims( rmesa );
-      ctx->Driver.NeedFlush &= ~FLUSH_STORED_VERTICES;
-   }
-}
-
-
-
-/* At this point, don't expect very many versions of each function to
- * be generated, so not concerned about freeing them?
- */
-
-
-/**
- * Called once during context creation.
- */
-void r200VtxfmtInit( GLcontext *ctx, GLboolean useCodegen )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-   GLvertexformat *vfmt = &(rmesa->vb.vtxfmt);
-
-   /* start by initializing to no-op functions */
-   _mesa_noop_vtxfmt_init(vfmt);
-
-   /* Hook in chooser functions for codegen, etc:
-    */
-   r200VtxfmtInitChoosers( vfmt );
-
-   /* Handled fully in supported states, but no codegen:
-    */
-   vfmt->Materialfv = r200_Materialfv;
-   vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = r200_Begin;
-   vfmt->End = r200_End;
-
-   /* Fallback for performance reasons:  (Fix with cva/elt path here and
-    * dmatmp2.h style primitive-merging)
-    *
-    * These should call NotifyBegin(), as should _tnl_EvalMesh, to allow
-    * a driver-hook.
-    */
-   vfmt->DrawArrays = r200_fallback_DrawArrays;
-   vfmt->DrawElements = r200_fallback_DrawElements;
-   vfmt->DrawRangeElements = r200_fallback_DrawRangeElements; 
-
-   /* Active but unsupported -- fallback if we receive these:
-    */
-   vfmt->CallList = r200_fallback_CallList;
-   vfmt->CallLists = r200_fallback_CallLists;
-   vfmt->EvalCoord1f = r200_fallback_EvalCoord1f;
-   vfmt->EvalCoord1fv = r200_fallback_EvalCoord1fv;
-   vfmt->EvalCoord2f = r200_fallback_EvalCoord2f;
-   vfmt->EvalCoord2fv = r200_fallback_EvalCoord2fv;
-   vfmt->EvalMesh1 = r200_fallback_EvalMesh1;
-   vfmt->EvalMesh2 = r200_fallback_EvalMesh2;
-   vfmt->EvalPoint1 = r200_fallback_EvalPoint1;
-   vfmt->EvalPoint2 = r200_fallback_EvalPoint2;
-   vfmt->TexCoord4f = r200_fallback_TexCoord4f;
-   vfmt->TexCoord4fv = r200_fallback_TexCoord4fv;
-   vfmt->MultiTexCoord4fARB = r200_fallback_MultiTexCoord4fARB;
-   vfmt->MultiTexCoord4fvARB = r200_fallback_MultiTexCoord4fvARB;
-   vfmt->Vertex4f = r200_fallback_Vertex4f;
-   vfmt->Vertex4fv = r200_fallback_Vertex4fv;
-   vfmt->VertexAttrib1fNV  = r200_fallback_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = r200_fallback_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV  = r200_fallback_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = r200_fallback_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV  = r200_fallback_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = r200_fallback_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV  = r200_fallback_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = r200_fallback_VertexAttrib4fvNV;
-   vfmt->FogCoordfEXT = r200_fallback_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = r200_fallback_FogCoordfvEXT;
-   
-   (void)r200_fallback_vtxfmt;
-
-   TNL_CONTEXT(ctx)->Driver.NotifyBegin = r200NotifyBegin;
-
-   rmesa->vb.enabled = 1;
-   rmesa->vb.prim = &ctx->Driver.CurrentExecPrimitive;
-   rmesa->vb.primflags = 0;
-
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex2f );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex2fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4ub );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4ubv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3ub );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3ubv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4f );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.Normal3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Normal3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord3f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord2f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord2fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord1f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord1fv );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord3fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord3fvARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-/*   make_empty_list( &rmesa->vb.dfn_cache.FogCoordfEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.FogCoordfvEXT );*/
-
-   r200InitCodegen( &rmesa->vb.codegen, useCodegen );
-}
-
-static void free_funcs( struct dynfn *l )
-{
-   struct dynfn *f, *tmp;
-   foreach_s (f, tmp, l) {
-      remove_from_list( f );
-      _mesa_exec_free( f->code );
-      _mesa_free( f );
-   }
-}
-
-void r200VtxfmtUnbindContext( GLcontext *ctx )
-{
-}
-
-
-void r200VtxfmtMakeCurrent( GLcontext *ctx )
-{
-}
-
-
-void r200VtxfmtDestroy( GLcontext *ctx )
-{
-   r200ContextPtr rmesa = R200_CONTEXT( ctx );
-
-   count_funcs( rmesa );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex2f );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex2fv );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex3f );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex3fv );
-   free_funcs( &rmesa->vb.dfn_cache.Color4ub );
-   free_funcs( &rmesa->vb.dfn_cache.Color4ubv );
-   free_funcs( &rmesa->vb.dfn_cache.Color3ub );
-   free_funcs( &rmesa->vb.dfn_cache.Color3ubv );
-   free_funcs( &rmesa->vb.dfn_cache.Color4f );
-   free_funcs( &rmesa->vb.dfn_cache.Color4fv );
-   free_funcs( &rmesa->vb.dfn_cache.Color3f );
-   free_funcs( &rmesa->vb.dfn_cache.Color3fv );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   free_funcs( &rmesa->vb.dfn_cache.Normal3f );
-   free_funcs( &rmesa->vb.dfn_cache.Normal3fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord3f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord3fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord2f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord2fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord1f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord1fv );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord3fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord3fvARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-/*   free_funcs( &rmesa->vb.dfn_cache.FogCoordfEXT );
-   free_funcs( &rmesa->vb.dfn_cache.FogCoordfvEXT );*/
-}
-
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt.h b/src/mesa/drivers/dri/r200/r200_vtxfmt.h
deleted file mode 100644
index 46999191e04..00000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt.h
+++ /dev/null
@@ -1,123 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt.h,v 1.1 2002/10/30 12:51:53 alanh Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __R200_VTXFMT_H__
-#define __R200_VTXFMT_H__
-
-#include "r200_context.h"
-
-
-
-extern void r200VtxfmtUpdate( GLcontext *ctx );
-extern void r200VtxfmtInit( GLcontext *ctx, GLboolean useCodegen );
-extern void r200VtxfmtInvalidate( GLcontext *ctx );
-extern void r200VtxfmtDestroy( GLcontext *ctx );
-extern void r200VtxfmtInitChoosers( GLvertexformat *vfmt );
-
-extern void r200VtxfmtMakeCurrent( GLcontext *ctx );
-extern void r200VtxfmtUnbindContext( GLcontext *ctx );
-
-extern void r200_copy_to_current( GLcontext *ctx );
-extern void VFMT_FALLBACK( const char *caller );
-
-#define DFN( FUNC, CACHE)				\
-do {							\
-   char *start = (char *)&FUNC;				\
-   char *end = (char *)&FUNC##_end;			\
-   insert_at_head( &CACHE, dfn );			\
-   dfn->key[0] = key[0];				\
-   dfn->key[1] = key[1];				\
-   dfn->code = _mesa_exec_malloc( end - start );	\
-   _mesa_memcpy(dfn->code, start, end - start);		\
-}							\
-while ( 0 )
-
-#define FIXUP( CODE, OFFSET, CHECKVAL, NEWVAL )	\
-do {						\
-   int *icode = (int *)(CODE+OFFSET);		\
-   assert (*icode == CHECKVAL);			\
-   *icode = (int)NEWVAL;			\
-} while (0)
-
-
-/* Useful for figuring out the offsets:
- */
-#define FIXUP2( CODE, OFFSET, CHECKVAL, NEWVAL )		\
-do {								\
-   while (*(int *)(CODE+OFFSET) != CHECKVAL) OFFSET++;		\
-   /*fprintf(stderr, "%s/%d CVAL %x OFFSET %d VAL %x\n", __FUNCTION__,*/ \
-   /*	   __LINE__, CHECKVAL, OFFSET, (int)(NEWVAL));*/		\
-   *(int *)(CODE+OFFSET) = (int)(NEWVAL);				\
-   OFFSET += 4;							\
-} while (0)
-
-/* 
- */
-void r200InitCodegen( struct dfn_generators *gen, GLboolean useCodegen );
-void r200InitX86Codegen( struct dfn_generators *gen );
-void r200InitSSECodegen( struct dfn_generators *gen );
-
-
-
-/* Defined in r200_vtxfmt_x86.c
- */
-struct dynfn *r200_makeX86Vertex2f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Vertex2fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Vertex3f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Vertex3fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color4ub( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color4ubv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color3ub( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color3ubv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color4f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color4fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color3f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Color3fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86SecondaryColor3ubEXT( GLcontext *, const int * );
-struct dynfn *r200_makeX86SecondaryColor3ubvEXT( GLcontext *, const int * );
-struct dynfn *r200_makeX86SecondaryColor3fEXT( GLcontext *, const int * );
-struct dynfn *r200_makeX86SecondaryColor3fvEXT( GLcontext *, const int * );
-struct dynfn *r200_makeX86Normal3f( GLcontext *, const int * );
-struct dynfn *r200_makeX86Normal3fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86TexCoord2f( GLcontext *, const int * );
-struct dynfn *r200_makeX86TexCoord2fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86TexCoord1f( GLcontext *, const int * );
-struct dynfn *r200_makeX86TexCoord1fv( GLcontext *, const int * );
-struct dynfn *r200_makeX86MultiTexCoord2fARB( GLcontext *, const int * );
-struct dynfn *r200_makeX86MultiTexCoord2fvARB( GLcontext *, const int * );
-struct dynfn *r200_makeX86MultiTexCoord1fARB( GLcontext *, const int * );
-struct dynfn *r200_makeX86MultiTexCoord1fvARB( GLcontext *, const int * );
-
-#endif
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt_c.c b/src/mesa/drivers/dri/r200/r200_vtxfmt_c.c
deleted file mode 100644
index 1db5950c8ff..00000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt_c.c
+++ /dev/null
@@ -1,1002 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt_c.c,v 1.2 2002/12/16 16:18:56 dawes Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "mtypes.h"
-#include "colormac.h"
-#include "simple_list.h"
-#include "api_noop.h"
-#include "vtxfmt.h"
-
-#include "r200_vtxfmt.h"
-#include "r200_tcl.h"
-
-#include "dispatch.h"
-
-/* Fallback versions of all the entrypoints for situations where
- * codegen isn't available.  This is still a lot faster than the
- * vb/pipeline implementation in Mesa.
- */
-static void r200_Vertex3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&x;
-   *rmesa->vb.dmaptr++ = *(int *)&y;
-   *rmesa->vb.dmaptr++ = *(int *)&z;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void r200_Vertex3fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&v[0];
-   *rmesa->vb.dmaptr++ = *(int *)&v[1];
-   *rmesa->vb.dmaptr++ = *(int *)&v[2];
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void r200_Vertex2f( GLfloat x, GLfloat y )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   int i;
-   
-   *rmesa->vb.dmaptr++ = *(int *)&x;
-   *rmesa->vb.dmaptr++ = *(int *)&y;
-   *rmesa->vb.dmaptr++ = 0;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void r200_Vertex2fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&v[0];
-   *rmesa->vb.dmaptr++ = *(int *)&v[1];
-   *rmesa->vb.dmaptr++ = 0;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-
-/* Color for ubyte (packed) color formats:
- */
-#if 0
-static void r200_Color3ub_ub( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= 0xff;
-}
-
-static void r200_Color3ubv_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= v[0];
-   dest->green	= v[1];
-   dest->blue	= v[2];
-   dest->alpha	= 0xff;
-}
-
-static void r200_Color4ub_ub( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= a;
-}
-
-static void r200_Color4ubv_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   *(GLuint *)rmesa->vb.colorptr = LE32_TO_CPU(*(GLuint *)v);
-}
-#endif /* 0 */
-
-static void r200_Color3f_ub( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   dest->alpha = 255;
-}
-
-static void r200_Color3fv_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   dest->alpha = 255;
-}
-
-static void r200_Color4f_ub( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, a );
-}
-
-static void r200_Color4fv_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, v[3] );
-}
-
-
-/* Color for float color+alpha formats:
- */
-#if 0
-static void r200_Color3ub_4f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = 1.0;
-}
-
-static void r200_Color3ubv_4f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = 1.0;
-}
-
-static void r200_Color4ub_4f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = UBYTE_TO_FLOAT(a);
-}
-
-static void r200_Color4ubv_4f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = UBYTE_TO_FLOAT(v[3]);
-}
-#endif /* 0 */
-
-
-static void r200_Color3f_4f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = 1.0;		
-}
-
-static void r200_Color3fv_4f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = 1.0;
-}
-
-static void r200_Color4f_4f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = a;
-}
-
-static void r200_Color4fv_4f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = v[3];
-}
-
-
-/* Color for float color formats:
- */
-#if 0
-static void r200_Color3ub_3f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-}
-
-static void r200_Color3ubv_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-}
-
-static void r200_Color4ub_3f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(a);
-}
-
-static void r200_Color4ubv_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(v[3]);
-}
-#endif /* 0 */
-
-
-static void r200_Color3f_3f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-}
-
-static void r200_Color3fv_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-}
-
-static void r200_Color4f_3f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = a;
-}
-
-static void r200_Color4fv_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = v[3]; 
-}
-
-
-/* Secondary Color:
- */
-#if 0
-static void r200_SecondaryColor3ubEXT_ub( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.specptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= 0xff;
-}
-
-static void r200_SecondaryColor3ubvEXT_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.specptr;
-   dest->red	= v[0];
-   dest->green	= v[1];
-   dest->blue	= v[2];
-   dest->alpha	= 0xff;
-}
-#endif /* 0 */
-
-static void r200_SecondaryColor3fEXT_ub( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.specptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   dest->alpha = 255;
-}
-
-static void r200_SecondaryColor3fvEXT_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   r200_color_t *dest = rmesa->vb.specptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   dest->alpha = 255;
-}
-
-#if 0
-static void r200_SecondaryColor3ubEXT_3f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = 1.0;
-}
-
-static void r200_SecondaryColor3ubvEXT_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = 1.0;
-}
-#endif /* 0 */
-
-static void r200_SecondaryColor3fEXT_3f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = 1.0;
-}
-
-static void r200_SecondaryColor3fvEXT_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = 1.0;
-}
-
-
-
-/* Normal
- */
-static void r200_Normal3f( GLfloat n0, GLfloat n1, GLfloat n2 )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.normalptr;
-   dest[0] = n0;
-   dest[1] = n1;
-   dest[2] = n2;
-}
-
-static void r200_Normal3fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.normalptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-}
-
-
-/* FogCoord
- */
-static void r200_FogCoordfEXT( GLfloat f )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.fogptr;
-   dest[0] = r200ComputeFogBlendFactor( ctx, f );
-/*   ctx->Current.Attrib[VERT_ATTRIB_FOG][0] = f;*/
-}
-
-static void r200_FogCoordfvEXT( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.fogptr;
-   dest[0] = r200ComputeFogBlendFactor( ctx, v[0] );
-/*   ctx->Current.Attrib[VERT_ATTRIB_FOG][0] = v[0];*/
-}
-
-
-/* TexCoord
- */
-
-/* \todo maybe (target & 4 ? target & 5 : target & 3) is more save than (target & 7) */
-static void r200_MultiTexCoord1fARB(GLenum target, GLfloat s)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLint unit = (target & 7);
-   GLfloat * const dest = rmesa->vb.texcoordptr[unit];
-
-   switch( ctx->Texture.Unit[unit]._ReallyEnabled ) {
-   case TEXTURE_CUBE_BIT:
-   case TEXTURE_3D_BIT:
-      dest[2] = 0.0;
-      /* FALLTHROUGH */
-   case TEXTURE_2D_BIT:
-   case TEXTURE_RECT_BIT:
-      dest[1] = 0.0;
-      /* FALLTHROUGH */
-   case TEXTURE_1D_BIT:
-      dest[0] = s;
-   }
-}
-
-static void r200_MultiTexCoord2fARB(GLenum target, GLfloat s, GLfloat t)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLint unit = (target & 7);
-   GLfloat * const dest = rmesa->vb.texcoordptr[unit];
-
-   switch( ctx->Texture.Unit[unit]._ReallyEnabled ) {
-   case TEXTURE_CUBE_BIT:
-   case TEXTURE_3D_BIT:
-      dest[2] = 0.0;
-      /* FALLTHROUGH */
-   case TEXTURE_2D_BIT:
-   case TEXTURE_RECT_BIT:
-      dest[1] = t;
-      dest[0] = s;
-      break;
-   default:
-      VFMT_FALLBACK(__FUNCTION__);
-      CALL_MultiTexCoord2fARB(GET_DISPATCH(), (target, s, t));
-      return;	
-   }
-}
-
-static void r200_MultiTexCoord3fARB(GLenum target, GLfloat s, GLfloat t, GLfloat r)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   GLint unit = (target & 7);
-   GLfloat * const dest = rmesa->vb.texcoordptr[unit];
-
-   switch( ctx->Texture.Unit[unit]._ReallyEnabled ) {
-   case TEXTURE_CUBE_BIT:
-   case TEXTURE_3D_BIT:
-      dest[2] = r;
-      dest[1] = t;
-      dest[0] = s;
-      break;
-   default:
-      VFMT_FALLBACK(__FUNCTION__);
-      CALL_MultiTexCoord3fARB(GET_DISPATCH(), (target, s, t, r));
-      return;	
-   }
-}
-
-static void r200_TexCoord1f(GLfloat s)
-{
-   r200_MultiTexCoord1fARB(GL_TEXTURE0, s);
-}
-
-static void r200_TexCoord2f(GLfloat s, GLfloat t)
-{
-   r200_MultiTexCoord2fARB(GL_TEXTURE0, s, t);
-}
-
-static void r200_TexCoord3f(GLfloat s, GLfloat t, GLfloat r)
-{
-   r200_MultiTexCoord3fARB(GL_TEXTURE0, s, t, r);
-}
-
-static void r200_TexCoord1fv(const GLfloat *v)
-{
-   r200_MultiTexCoord1fARB(GL_TEXTURE0, v[0]);
-}
-
-static void r200_TexCoord2fv(const GLfloat *v)
-{
-   r200_MultiTexCoord2fARB(GL_TEXTURE0, v[0], v[1]);
-}
-
-static void r200_TexCoord3fv(const GLfloat *v)
-{
-   r200_MultiTexCoord3fARB(GL_TEXTURE0, v[0], v[1], v[2]);
-}
-
-static void r200_MultiTexCoord1fvARB(GLenum target, const GLfloat *v)
-{
-   r200_MultiTexCoord1fARB(target, v[0]);
-}
-
-static void r200_MultiTexCoord2fvARB(GLenum target, const GLfloat *v)
-{
-   r200_MultiTexCoord2fARB(target, v[0], v[1]);
-}
-
-static void r200_MultiTexCoord3fvARB(GLenum target, const GLfloat *v)
-{
-   r200_MultiTexCoord3fARB(target, v[0], v[1], v[2]);
-}
-
-
-static struct dynfn *lookup( struct dynfn *l, const int *key )
-{
-   struct dynfn *f;
-
-   foreach( f, l ) {
-      if (f->key[0] == key[0] && f->key[1] == key[1]) 
-	 return f;
-   }
-
-   return NULL;
-}
-
-/* Can't use the loopback template for this:
- */
-
-#define CHOOSE(FN, FNTYPE, MASK0, MASK1, ARGS1, ARGS2 )			\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);				\
-   int key[2];								\
-   struct dynfn *dfn;							\
-									\
-   key[0] = rmesa->vb.vtxfmt_0 & MASK0;					\
-   key[1] = rmesa->vb.vtxfmt_1 & MASK1;					\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (dfn == 0)							\
-      dfn = rmesa->vb.codegen.FN( ctx, key );				\
-   else if (R200_DEBUG & DEBUG_CODEGEN)					\
-      fprintf(stderr, "%s -- cached codegen\n", __FUNCTION__ );		\
-									\
-   if (dfn)								\
-      SET_ ## FN (ctx->Exec, (FNTYPE)(dfn->code));			\
-   else {								\
-      if (R200_DEBUG & DEBUG_CODEGEN)					\
-	 fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, r200_##FN);				\
-   }									\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-/* For the _3f case, only allow one color function to be hooked in at
- * a time.  Eventually, use a similar mechanism to allow selecting the
- * color component of the vertex format based on client behaviour.  
- *
- * Note:  Perform these actions even if there is a codegen or cached 
- * codegen version of the chosen function.
- */
-#define CHOOSE_COLOR(FN, FNTYPE, NR, MASK0, MASK1, ARGS1, ARGS2 )	\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);			\
-   int key[2];								\
-   struct dynfn *dfn;							\
-									\
-   key[0] = rmesa->vb.vtxfmt_0 & MASK0;					\
-   key[1] = rmesa->vb.vtxfmt_1 & MASK1;					\
-									\
-   if (VTX_COLOR(rmesa->vb.vtxfmt_0,0) == R200_VTX_PK_RGBA) {		\
-      SET_ ## FN (ctx->Exec, r200_##FN##_ub);				\
-   }									\
-   else if (VTX_COLOR(rmesa->vb.vtxfmt_0,0) == R200_VTX_FP_RGB) {	\
-									\
-      if (rmesa->vb.installed_color_3f_sz != NR) {			\
-         rmesa->vb.installed_color_3f_sz = NR;				\
-         if (NR == 3) ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = 1.0;	\
-         if (ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) {		\
-            r200_copy_to_current( ctx );				\
-            _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );	\
-            CALL_ ## FN (ctx->Exec, ARGS2);				\
-            return;							\
-         }								\
-      }									\
-									\
-      SET_ ## FN (ctx->Exec, r200_##FN##_3f);				\
-   }									\
-   else {								\
-      SET_ ## FN (ctx->Exec, r200_##FN##_4f);				\
-   }									\
-									\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (!dfn) dfn = rmesa->vb.codegen.FN( ctx, key );			\
-									\
-   if (dfn) {								\
-      if (R200_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- codegen version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, (FNTYPE)dfn->code);			\
-   }									\
-   else if (R200_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- 'c' version\n", __FUNCTION__ );		\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-/* Right now there are both _ub and _3f versions of the secondary color
- * functions.  Currently, we only set-up the hardware to use the _ub versions.
- * The _3f versions are needed for the cases where secondary color isn't used
- * in the vertex format, but it still needs to be stored in the context
- * state vector.
- */
-#define CHOOSE_SECONDARY_COLOR(FN, FNTYPE, MASK0, MASK1, ARGS1, ARGS2 )	\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);			\
-   int key[2];								\
-   struct dynfn *dfn;							\
-									\
-   key[0] = rmesa->vb.vtxfmt_0 & MASK0;					\
-   key[1] = rmesa->vb.vtxfmt_1 & MASK1;					\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (dfn == 0)							\
-      dfn = rmesa->vb.codegen.FN( ctx, key );			\
-   else  if (R200_DEBUG & DEBUG_CODEGEN)				\
-      fprintf(stderr, "%s -- cached version\n", __FUNCTION__ );		\
-									\
-   if (dfn)								\
-      SET_ ## FN (ctx->Exec, (FNTYPE)(dfn->code));			\
-   else {								\
-      if (R200_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, (VTX_COLOR(rmesa->vb.vtxfmt_0,1) == R200_VTX_PK_RGBA) \
-	  ? r200_##FN##_ub : r200_##FN##_3f);				\
-   }									\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;		\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-
-
-
-
-/* VTXFMT_0
- */
-#define MASK_XYZW  (R200_VTX_W0|R200_VTX_Z0)
-#define MASK_NORM  (MASK_XYZW|R200_VTX_N0)
-#define MASK_FOG   (MASK_NORM |R200_VTX_DISCRETE_FOG)
-#define MASK_COLOR (MASK_FOG |(R200_VTX_COLOR_MASK<<R200_VTX_COLOR_0_SHIFT))
-#define MASK_SPEC  (MASK_COLOR|(R200_VTX_COLOR_MASK<<R200_VTX_COLOR_1_SHIFT))
-
-/* VTXFMT_1
- */
-#define MASK_ST0 (0x7 << R200_VTX_TEX0_COMP_CNT_SHIFT)
-/* FIXME: maybe something like in the radeon driver is needed here? */
-
-
-typedef void (*p4f)( GLfloat, GLfloat, GLfloat, GLfloat );
-typedef void (*p3f)( GLfloat, GLfloat, GLfloat );
-typedef void (*p2f)( GLfloat, GLfloat );
-typedef void (*p1f)( GLfloat );
-typedef void (*pe3f)( GLenum, GLfloat, GLfloat, GLfloat );
-typedef void (*pe2f)( GLenum, GLfloat, GLfloat );
-typedef void (*pe1f)( GLenum, GLfloat );
-typedef void (*p4ub)( GLubyte, GLubyte, GLubyte, GLubyte );
-typedef void (*p3ub)( GLubyte, GLubyte, GLubyte );
-typedef void (*pfv)( const GLfloat * );
-typedef void (*pefv)( GLenum, const GLfloat * );
-typedef void (*pubv)( const GLubyte * );
-
-
-CHOOSE(Normal3f, p3f, MASK_NORM, 0, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(Normal3fv, pfv, MASK_NORM, 0, 
-       (const GLfloat *v), (v))
-
-#if 0
-CHOOSE_COLOR(Color4ub, p4ub, 4, MASK_COLOR, 0,
-	(GLubyte a,GLubyte b, GLubyte c, GLubyte d), (a,b,c,d))
-CHOOSE_COLOR(Color4ubv, pubv, 4, MASK_COLOR, 0, 
-	(const GLubyte *v), (v))
-CHOOSE_COLOR(Color3ub, p3ub, 3, MASK_COLOR, 0, 
-	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
-CHOOSE_COLOR(Color3ubv, pubv, 3, MASK_COLOR, 0, 
-	(const GLubyte *v), (v))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3ubEXT, p3ub, MASK_SPEC, 0, 
-	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3ubvEXT, pubv, MASK_SPEC, 0, 
-	(const GLubyte *v), (v))
-#endif
-
-CHOOSE_COLOR(Color4f, p4f, 4, MASK_COLOR, 0, 
-	(GLfloat a,GLfloat b, GLfloat c, GLfloat d), (a,b,c,d))
-CHOOSE_COLOR(Color4fv, pfv, 4, MASK_COLOR, 0, 
-	(const GLfloat *v), (v))
-CHOOSE_COLOR(Color3f, p3f, 3, MASK_COLOR, 0,
-	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
-CHOOSE_COLOR(Color3fv, pfv, 3, MASK_COLOR, 0,
-	(const GLfloat *v), (v))
-
-
-CHOOSE_SECONDARY_COLOR(SecondaryColor3fEXT, p3f, MASK_SPEC, 0,
-	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3fvEXT, pfv, MASK_SPEC, 0,
-	(const GLfloat *v), (v))
-
-CHOOSE(TexCoord3f, p3f, ~0, MASK_ST0, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(TexCoord3fv, pfv, ~0, MASK_ST0, 
-       (const GLfloat *v), (v))
-CHOOSE(TexCoord2f, p2f, ~0, MASK_ST0, 
-       (GLfloat a,GLfloat b), (a,b))
-CHOOSE(TexCoord2fv, pfv, ~0, MASK_ST0, 
-       (const GLfloat *v), (v))
-CHOOSE(TexCoord1f, p1f, ~0, MASK_ST0, 
-       (GLfloat a), (a))
-CHOOSE(TexCoord1fv, pfv, ~0, MASK_ST0, 
-       (const GLfloat *v), (v))
-
-CHOOSE(MultiTexCoord3fARB, pe3f, ~0, ~0,
-	 (GLenum u,GLfloat a,GLfloat b,GLfloat c), (u,a,b,c))
-CHOOSE(MultiTexCoord3fvARB, pefv, ~0, ~0,
-	(GLenum u,const GLfloat *v), (u,v))
-CHOOSE(MultiTexCoord2fARB, pe2f, ~0, ~0,
-	 (GLenum u,GLfloat a,GLfloat b), (u,a,b))
-CHOOSE(MultiTexCoord2fvARB, pefv, ~0, ~0,
-	(GLenum u,const GLfloat *v), (u,v))
-CHOOSE(MultiTexCoord1fARB, pe1f, ~0, ~0,
-	 (GLenum u,GLfloat a), (u,a))
-CHOOSE(MultiTexCoord1fvARB, pefv, ~0, ~0,
-	(GLenum u,const GLfloat *v), (u,v))
-
-CHOOSE(Vertex3f, p3f, ~0, ~0, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(Vertex3fv, pfv, ~0, ~0, 
-       (const GLfloat *v), (v))
-CHOOSE(Vertex2f, p2f, ~0, ~0, 
-       (GLfloat a,GLfloat b), (a,b))
-CHOOSE(Vertex2fv, pfv, ~0, ~0, 
-       (const GLfloat *v), (v))
-
-CHOOSE(FogCoordfEXT, p1f, MASK_FOG, ~0, 
-       (GLfloat f), (f))
-CHOOSE(FogCoordfvEXT, pfv, MASK_FOG, ~0, 
-       (const GLfloat *f), (f))
-
-
-
-
-void r200VtxfmtInitChoosers( GLvertexformat *vfmt )
-{
-   vfmt->Color3f = choose_Color3f;
-   vfmt->Color3fv = choose_Color3fv;
-   vfmt->Color4f = choose_Color4f;
-   vfmt->Color4fv = choose_Color4fv;
-   vfmt->SecondaryColor3fEXT = choose_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = choose_SecondaryColor3fvEXT;
-   vfmt->MultiTexCoord1fARB = choose_MultiTexCoord1fARB;
-   vfmt->MultiTexCoord1fvARB = choose_MultiTexCoord1fvARB;
-   vfmt->MultiTexCoord2fARB = choose_MultiTexCoord2fARB;
-   vfmt->MultiTexCoord2fvARB = choose_MultiTexCoord2fvARB;
-   vfmt->MultiTexCoord3fARB = choose_MultiTexCoord3fARB;
-   vfmt->MultiTexCoord3fvARB = choose_MultiTexCoord3fvARB;
-   vfmt->Normal3f = choose_Normal3f;
-   vfmt->Normal3fv = choose_Normal3fv;
-   vfmt->TexCoord1f = choose_TexCoord1f;
-   vfmt->TexCoord1fv = choose_TexCoord1fv;
-   vfmt->TexCoord2f = choose_TexCoord2f;
-   vfmt->TexCoord2fv = choose_TexCoord2fv;
-   vfmt->TexCoord3f = choose_TexCoord3f;
-   vfmt->TexCoord3fv = choose_TexCoord3fv;
-   vfmt->Vertex2f = choose_Vertex2f;
-   vfmt->Vertex2fv = choose_Vertex2fv;
-   vfmt->Vertex3f = choose_Vertex3f;
-   vfmt->Vertex3fv = choose_Vertex3fv;
-/*   vfmt->FogCoordfEXT = choose_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = choose_FogCoordfvEXT;*/
-
-   /* TODO: restore ubyte colors to vtxfmt.
-    */
-#if 0
-   vfmt->Color3ub = choose_Color3ub;
-   vfmt->Color3ubv = choose_Color3ubv;
-   vfmt->Color4ub = choose_Color4ub;
-   vfmt->Color4ubv = choose_Color4ubv;
-   vfmt->SecondaryColor3ubEXT = choose_SecondaryColor3ubEXT;
-   vfmt->SecondaryColor3ubvEXT = choose_SecondaryColor3ubvEXT;
-#endif
-}
-
-
-static struct dynfn *codegen_noop( GLcontext *ctx, const int *key )
-{
-   (void) ctx; (void) key;
-   return NULL;
-}
-
-void r200InitCodegen( struct dfn_generators *gen, GLboolean useCodegen )
-{
-   gen->Vertex3f = codegen_noop;
-   gen->Vertex3fv = codegen_noop;
-   gen->Color4ub = codegen_noop;
-   gen->Color4ubv = codegen_noop;
-   gen->Normal3f = codegen_noop;
-   gen->Normal3fv = codegen_noop;
-
-   gen->TexCoord3f = codegen_noop;
-   gen->TexCoord3fv = codegen_noop;
-   gen->TexCoord2f = codegen_noop;
-   gen->TexCoord2fv = codegen_noop;
-   gen->TexCoord1f = codegen_noop;
-   gen->TexCoord1fv = codegen_noop;
-
-   gen->MultiTexCoord3fARB = codegen_noop;
-   gen->MultiTexCoord3fvARB = codegen_noop;
-   gen->MultiTexCoord2fARB = codegen_noop;
-   gen->MultiTexCoord2fvARB = codegen_noop;
-   gen->MultiTexCoord1fARB = codegen_noop;
-   gen->MultiTexCoord1fvARB = codegen_noop;
-/*   gen->FogCoordfEXT = codegen_noop;
-   gen->FogCoordfvEXT = codegen_noop;*/
-
-   gen->Vertex2f = codegen_noop;
-   gen->Vertex2fv = codegen_noop;
-   gen->Color3ub = codegen_noop;
-   gen->Color3ubv = codegen_noop;
-   gen->Color4f = codegen_noop;
-   gen->Color4fv = codegen_noop;
-   gen->Color3f = codegen_noop;
-   gen->Color3fv = codegen_noop;
-   gen->SecondaryColor3fEXT = codegen_noop;
-   gen->SecondaryColor3fvEXT = codegen_noop;
-   gen->SecondaryColor3ubEXT = codegen_noop;
-   gen->SecondaryColor3ubvEXT = codegen_noop;
-
-   if (useCodegen) {
-#if defined(USE_X86_ASM)
-      r200InitX86Codegen( gen );
-#endif
-
-#if defined(USE_SSE_ASM)
-      r200InitSSECodegen( gen );
-#endif
-   }
-}
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt_sse.c b/src/mesa/drivers/dri/r200/r200_vtxfmt_sse.c
deleted file mode 100644
index 5901730494a..00000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt_sse.c
+++ /dev/null
@@ -1,234 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt_sse.c,v 1.1 2002/10/30 12:51:53 alanh Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "simple_list.h" 
-#include "r200_vtxfmt.h"
-
-#if defined(USE_SSE_ASM)
-#include "x86/common_x86_asm.h"
-
-#define EXTERN( FUNC )		\
-extern const char *FUNC;	\
-extern const char *FUNC##_end
-
-EXTERN( _sse_Attribute2fv );
-EXTERN( _sse_Attribute2f );
-EXTERN( _sse_Attribute3fv );
-EXTERN( _sse_Attribute3f );
-EXTERN( _sse_MultiTexCoord2fv );
-EXTERN( _sse_MultiTexCoord2f );
-EXTERN( _sse_MultiTexCoord2fv_2 );
-EXTERN( _sse_MultiTexCoord2f_2 );
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.
- */
-
-static struct dynfn *r200_makeSSEAttribute2fv( struct dynfn * cache, const int * key,
-					       const char * name, void * dest)
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _sse_Attribute2fv, (*cache) );
-   FIXUP(dfn->code, 10, 0x0, (int)dest);
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSEAttribute2f( struct dynfn * cache, const int * key,
-					      const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _sse_Attribute2f, (*cache) );
-   FIXUP(dfn->code, 8, 0x0, (int)dest); 
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSEAttribute3fv( struct dynfn * cache, const int * key,
-					       const char * name, void * dest)
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _sse_Attribute3fv, (*cache) );
-   FIXUP(dfn->code, 13, 0x0, (int)dest);
-   FIXUP(dfn->code, 18, 0x8, 8+(int)dest);
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSEAttribute3f( struct dynfn * cache, const int * key,
-					      const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _sse_Attribute3f, (*cache) );
-   FIXUP(dfn->code, 12, 0x0, (int)dest); 
-   FIXUP(dfn->code, 17, 0x8, 8+(int)dest); 
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSENormal3fv( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeSSEAttribute3fv( & rmesa->vb.dfn_cache.Normal3fv, key,
-				    __FUNCTION__, rmesa->vb.normalptr );
-}
-
-static struct dynfn *r200_makeSSENormal3f( GLcontext *ctx, const int * key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeSSEAttribute3f( & rmesa->vb.dfn_cache.Normal3f, key,
-				   __FUNCTION__, rmesa->vb.normalptr );
-}
-
-static struct dynfn *r200_makeSSEColor3fv( GLcontext *ctx, const int * key )
-{
-   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
-      return NULL;
-   else
-   {
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      return r200_makeSSEAttribute3fv( & rmesa->vb.dfn_cache.Color3fv, key,
-				       __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-static struct dynfn *r200_makeSSEColor3f( GLcontext *ctx, const int * key )
-{
-   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
-      return NULL;
-   else
-   {
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      return r200_makeSSEAttribute3f( & rmesa->vb.dfn_cache.Color3f, key,
-				      __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-#if 0 /* Temporarily disabled as it is broken w/the new cubemap code. - idr */
-static struct dynfn *r200_makeSSETexCoord2fv( GLcontext *ctx, const int * key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeSSEAttribute2fv( & rmesa->vb.dfn_cache.TexCoord2fv, key,
-				    __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-static struct dynfn *r200_makeSSETexCoord2f( GLcontext *ctx, const int * key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeSSEAttribute2f( & rmesa->vb.dfn_cache.TexCoord2f, key,
-				   __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-static struct dynfn *r200_makeSSEMultiTexCoord2fv( GLcontext *ctx, const int * key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
-
-   if (rmesa->vb.texcoordptr[1] == rmesa->vb.texcoordptr[0]+4) {
-      DFN ( _sse_MultiTexCoord2fv, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 18, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-   } else {
-      DFN ( _sse_MultiTexCoord2fv_2, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 14, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-
-static struct dynfn *r200_makeSSEMultiTexCoord2f( GLcontext *ctx, const int * key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
-
-   if (rmesa->vb.texcoordptr[1] == rmesa->vb.texcoordptr[0]+4) {
-      DFN ( _sse_MultiTexCoord2f, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 16, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-   } else {
-      DFN ( _sse_MultiTexCoord2f_2, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 15, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-#endif
-
-void r200InitSSECodegen( struct dfn_generators *gen )
-{
-   if ( cpu_has_xmm ) {
-      gen->Normal3fv = (void *) r200_makeSSENormal3fv;
-      gen->Normal3f = (void *) r200_makeSSENormal3f;
-      gen->Color3fv = (void *) r200_makeSSEColor3fv;
-      gen->Color3f = (void *) r200_makeSSEColor3f;
-#if 0 /* Temporarily disabled as it is broken w/the new cubemap code. - idr */
-      gen->TexCoord2fv = (void *) r200_makeSSETexCoord2fv;
-      gen->TexCoord2f = (void *) r200_makeSSETexCoord2f;
-      gen->MultiTexCoord2fvARB = (void *) r200_makeSSEMultiTexCoord2fv;
-      gen->MultiTexCoord2fARB = (void *) r200_makeSSEMultiTexCoord2f;
-#endif
-   }
-}
-
-#else 
-
-void r200InitSSECodegen( struct dfn_generators *gen )
-{
-   (void) gen;
-}
-
-#endif
diff --git a/src/mesa/drivers/dri/r200/r200_vtxfmt_x86.c b/src/mesa/drivers/dri/r200/r200_vtxfmt_x86.c
deleted file mode 100644
index b78a55c31f6..00000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxfmt_x86.c
+++ /dev/null
@@ -1,440 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxfmt_x86.c,v 1.2 2002/12/16 16:18:56 dawes Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "simple_list.h" 
-#include "r200_vtxfmt.h"
-
-#if defined(USE_X86_ASM)
-
-#define EXTERN( FUNC )		\
-extern const char *FUNC;	\
-extern const char *FUNC##_end
-
-EXTERN ( _x86_Attribute2fv );
-EXTERN ( _x86_Attribute2f );
-EXTERN ( _x86_Attribute3fv );
-EXTERN ( _x86_Attribute3f );
-EXTERN ( _x86_Vertex3fv_6 );
-EXTERN ( _x86_Vertex3fv_8 );
-EXTERN ( _x86_Vertex3fv );
-EXTERN ( _x86_Vertex3f_4 );
-EXTERN ( _x86_Vertex3f_6 );
-EXTERN ( _x86_Vertex3f );
-EXTERN ( _x86_Color4ubv_ub );
-EXTERN ( _x86_Color4ubv_4f );
-EXTERN ( _x86_Color4ub_ub );
-EXTERN ( _x86_MultiTexCoord2fv );
-EXTERN ( _x86_MultiTexCoord2fv_2 );
-EXTERN ( _x86_MultiTexCoord2f );
-EXTERN ( _x86_MultiTexCoord2f_2 );
-
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.  Generic x86 versions.
- */
-
-struct dynfn *r200_makeX86Vertex3f( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x 0x%08x %d\n", __FUNCTION__, 
-	      key[0], key[1], rmesa->vb.vertex_size );
-
-   switch (rmesa->vb.vertex_size) {
-   case 4: {
-
-      DFN ( _x86_Vertex3f_4, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 2, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 25, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 36, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 46, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 51, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 60, 0x0, (int)&rmesa->vb.notify);
-      break;
-   }
-   case 6: {
-
-      DFN ( _x86_Vertex3f_6, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 3, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 28, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 34, 0x0, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 40, 0x0, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 57, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 63, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 70, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 79, 0x0, (int)&rmesa->vb.notify);
-      break;
-   }
-   default: {
-
-      DFN ( _x86_Vertex3f, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 3, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 9, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 37, 0x0, rmesa->vb.vertex_size-3);
-      FIXUP(dfn->code, 44, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 50, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 56, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x0, (int)&rmesa->vb.notify);
-   break;
-   }
-   }
-
-   return dfn;
-}
-
-
-
-struct dynfn *r200_makeX86Vertex3fv( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x 0x%08x %d\n", __FUNCTION__, 
-	      key[0], key[1], rmesa->vb.vertex_size );
-
-   switch (rmesa->vb.vertex_size) {
-   case 6: {
-
-      DFN ( _x86_Vertex3fv_6, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 1, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 27, 0x0000001c, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 33, 0x00000020, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 45, 0x00000024, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 56, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 61, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 76, 0x00000008, (int)&rmesa->vb.notify);
-      break;
-   }
-   
-
-   case 8: {
-
-      DFN ( _x86_Vertex3fv_8, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 1, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 27, 0x0000001c, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 33, 0x00000020, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 45, 0x0000001c, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 51, 0x00000020, (int)&rmesa->vb.vertex[6]);
-      FIXUP(dfn->code, 63, 0x00000024, (int)&rmesa->vb.vertex[7]);
-      FIXUP(dfn->code, 74, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 79, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 85, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 94, 0x00000008, (int)&rmesa->vb.notify);
-      break;
-   }
-   
-
-
-   default: {
-
-      DFN ( _x86_Vertex3fv, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 8, 0x01010101, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 32, 0x00000006, rmesa->vb.vertex_size-3);
-      FIXUP(dfn->code, 37, 0x00000058, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 45, 0x01010101, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 50, 0x02020202, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 58, 0x02020202, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x0, (int)&rmesa->vb.notify);
-   break;
-   }
-   }
-
-   return dfn;
-}
-
-static struct dynfn *
-r200_makeX86Attribute2fv( struct dynfn * cache, const int *key,
-			  const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _x86_Attribute2fv, (*cache) );
-   FIXUP(dfn->code, 11, 0x0, (int)dest); 
-   FIXUP(dfn->code, 16, 0x4, 4+(int)dest); 
-
-   return dfn;
-}
-
-static struct dynfn *
-r200_makeX86Attribute2f( struct dynfn * cache, const int *key,
-			 const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _x86_Attribute2f, (*cache) );
-   FIXUP(dfn->code, 1, 0x0, (int)dest); 
-
-   return dfn;
-}
-
-
-static struct dynfn *
-r200_makeX86Attribute3fv( struct dynfn * cache, const int *key,
-			  const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _x86_Attribute3fv, (*cache) );
-   FIXUP(dfn->code, 14, 0x0, (int)dest); 
-   FIXUP(dfn->code, 20, 0x4, 4+(int)dest); 
-   FIXUP(dfn->code, 25, 0x8, 8+(int)dest);
-
-   return dfn;
-}
-
-static struct dynfn *
-r200_makeX86Attribute3f( struct dynfn * cache, const int *key,
-			 const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key[0] );
-
-   DFN ( _x86_Attribute3f, (*cache) );
-   FIXUP(dfn->code, 14, 0x0, (int)dest); 
-   FIXUP(dfn->code, 20, 0x4, 4+(int)dest); 
-   FIXUP(dfn->code, 25, 0x8, 8+(int)dest);
-
-   return dfn;
-}
-
-struct dynfn *r200_makeX86Normal3fv( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeX86Attribute3fv( & rmesa->vb.dfn_cache.Normal3fv, key,
-				    __FUNCTION__, rmesa->vb.normalptr );
-}
-
-struct dynfn *r200_makeX86Normal3f( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeX86Attribute3f( & rmesa->vb.dfn_cache.Normal3f, key,
-				   __FUNCTION__, rmesa->vb.normalptr );
-}
-
-struct dynfn *r200_makeX86Color4ubv( GLcontext *ctx, const int *key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
-
-   if (VTX_COLOR(key[0],0) == R200_VTX_PK_RGBA) {
-      DFN ( _x86_Color4ubv_ub, rmesa->vb.dfn_cache.Color4ubv);
-      FIXUP(dfn->code, 5, 0x12345678, (int)rmesa->vb.colorptr); 
-      return dfn;
-   } 
-   else {
-
-      DFN ( _x86_Color4ubv_4f, rmesa->vb.dfn_cache.Color4ubv);
-      FIXUP(dfn->code, 2, 0x00000000, (int)_mesa_ubyte_to_float_color_tab); 
-      FIXUP(dfn->code, 27, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr); 
-      FIXUP(dfn->code, 33, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+4); 
-      FIXUP(dfn->code, 55, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+8); 
-      FIXUP(dfn->code, 61, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+12); 
-      return dfn;
-   }
-}
-
-struct dynfn *r200_makeX86Color4ub( GLcontext *ctx, const int *key )
-{
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key[0] );
-
-   if (VTX_COLOR(key[0],0) == R200_VTX_PK_RGBA) {
-      struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      DFN ( _x86_Color4ub_ub, rmesa->vb.dfn_cache.Color4ub );
-      FIXUP(dfn->code, 18, 0x0, (int)rmesa->vb.colorptr); 
-      FIXUP(dfn->code, 24, 0x0, (int)rmesa->vb.colorptr+1); 
-      FIXUP(dfn->code, 30, 0x0, (int)rmesa->vb.colorptr+2); 
-      FIXUP(dfn->code, 36, 0x0, (int)rmesa->vb.colorptr+3); 
-      return dfn;
-   }
-   else
-      return NULL;
-}
-
-
-struct dynfn *r200_makeX86Color3fv( GLcontext *ctx, const int *key )
-{
-   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
-      return NULL;
-   else
-   {
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      return r200_makeX86Attribute3fv( & rmesa->vb.dfn_cache.Color3fv, key,
-				       __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-struct dynfn *r200_makeX86Color3f( GLcontext *ctx, const int *key )
-{
-   if (VTX_COLOR(key[0],0) != R200_VTX_FP_RGB) 
-      return NULL;
-   else
-   {
-      r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-      return r200_makeX86Attribute3f( & rmesa->vb.dfn_cache.Color3f, key,
-				      __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-
-
-#if 0 /* Temporarily disabled as it is broken w/the new cubemap code. - idr */
-struct dynfn *r200_makeX86TexCoord2fv( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeX86Attribute2fv( & rmesa->vb.dfn_cache.TexCoord2fv, key,
-				    __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-struct dynfn *r200_makeX86TexCoord2f( GLcontext *ctx, const int *key )
-{
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   return r200_makeX86Attribute2f( & rmesa->vb.dfn_cache.TexCoord2f, key,
-				   __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-struct dynfn *r200_makeX86MultiTexCoord2fvARB( GLcontext *ctx, const int *key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x 0x%08x\n", __FUNCTION__, key[0], key[1] );
-
-   if (rmesa->vb.texcoordptr[1] == rmesa->vb.texcoordptr[0]+4) {
-      DFN ( _x86_MultiTexCoord2fv, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 21, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-      FIXUP(dfn->code, 27, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]+4);
-   } else {
-      DFN ( _x86_MultiTexCoord2fv_2, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 14, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-
-struct dynfn *r200_makeX86MultiTexCoord2fARB( GLcontext *ctx, 
-					      const int *key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   r200ContextPtr rmesa = R200_CONTEXT(ctx);
-
-   if (R200_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x 0x%08x\n", __FUNCTION__, key[0], key[1] );
-
-   if (rmesa->vb.texcoordptr[1] == rmesa->vb.texcoordptr[0]+4) {
-      DFN ( _x86_MultiTexCoord2f, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 20, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]); 
-      FIXUP(dfn->code, 26, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]+4); 
-   }
-   else {
-      /* Note: this might get generated multiple times, even though the
-       * actual emitted code is the same.
-       */
-      DFN ( _x86_MultiTexCoord2f_2, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 18, 0x0, (int)rmesa->vb.texcoordptr); 
-   }      
-   return dfn;
-}
-#endif
-
-void r200InitX86Codegen( struct dfn_generators *gen )
-{
-   gen->Vertex3f = r200_makeX86Vertex3f;
-   gen->Vertex3fv = r200_makeX86Vertex3fv;
-   gen->Color4ub = r200_makeX86Color4ub; /* PKCOLOR only */
-   gen->Color4ubv = r200_makeX86Color4ubv; /* PKCOLOR only */
-   gen->Normal3f = r200_makeX86Normal3f;
-   gen->Normal3fv = r200_makeX86Normal3fv;
-#if 0 /* Temporarily disabled as it is broken w/the new cubemap code. - idr */
-   gen->TexCoord2f = r200_makeX86TexCoord2f;
-   gen->TexCoord2fv = r200_makeX86TexCoord2fv;
-   gen->MultiTexCoord2fARB = r200_makeX86MultiTexCoord2fARB;
-   gen->MultiTexCoord2fvARB = r200_makeX86MultiTexCoord2fvARB;
-#endif
-   gen->Color3f = r200_makeX86Color3f;
-   gen->Color3fv = r200_makeX86Color3fv;
-
-   /* Not done:
-    */
-/*     gen->Vertex2f = r200_makeX86Vertex2f; */
-/*     gen->Vertex2fv = r200_makeX86Vertex2fv; */
-/*     gen->Color3ub = r200_makeX86Color3ub; */
-/*     gen->Color3ubv = r200_makeX86Color3ubv; */
-/*     gen->Color4f = r200_makeX86Color4f; */
-/*     gen->Color4fv = r200_makeX86Color4fv; */
-/*     gen->TexCoord1f = r200_makeX86TexCoord1f; */
-/*     gen->TexCoord1fv = r200_makeX86TexCoord1fv; */
-/*     gen->MultiTexCoord1fARB = r200_makeX86MultiTexCoord1fARB; */
-/*     gen->MultiTexCoord1fvARB = r200_makeX86MultiTexCoord1fvARB; */
-}
-
-
-#else 
-
-void r200InitX86Codegen( struct dfn_generators *gen )
-{
-   (void) gen;
-}
-
-#endif
diff --git a/src/mesa/drivers/dri/r200/r200_vtxtmp_x86.S b/src/mesa/drivers/dri/r200/r200_vtxtmp_x86.S
deleted file mode 100644
index 5e33c7bdeea..00000000000
--- a/src/mesa/drivers/dri/r200/r200_vtxtmp_x86.S
+++ /dev/null
@@ -1,499 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_vtxtmp_x86.S,v 1.2 2002/11/07 18:31:59 tsi Exp $ */
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS 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.
-
-**************************************************************************/
-
-#define GLOBL( x )	\
-.globl x;		\
-x:
-
-.data
-.align 4
-
-/*
-	vertex 3f vertex size 4
-*/
-	
-GLOBL ( _x86_Vertex3f_4 )
-	movl	(0), %ecx
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	%eax, (%ecx)
-	movl	%edx, 4(%ecx)
-	movl	12(%esp), %eax
-	movl	(0), %edx
-	movl	%eax, 8(%ecx)
-	movl	%edx, 12(%ecx)
-	movl	(0), %eax
-	addl	$16, %ecx
-	dec 	%eax
-	movl	%ecx, (0)
-	movl	%eax, (0)
-	je	.1 
-	ret
-.1:	jmp	*0
-	
-GLOBL ( _x86_Vertex3f_4_end )
-
-/*
-	vertex 3f vertex size 6
-*/
-GLOBL ( _x86_Vertex3f_6 )
-	push	%edi
-	movl	(0), %edi
-	movl	8(%esp), %eax
-	movl	12(%esp), %edx
-	movl	16(%esp), %ecx
-	movl	%eax, (%edi)
-	movl	%edx, 4(%edi)
-	movl	%ecx, 8(%edi)
-	movl	(0), %eax
-	movl	(0), %edx
-	movl	(0), %ecx
-	movl	%eax, 12(%edi)
-	movl	%edx, 16(%edi)
-	movl	%ecx, 20(%edi)
-	addl	$24, %edi
-	movl	(0), %eax
-	movl	%edi, (0)
-	dec 	%eax
-	pop 	%edi
-	movl	%eax, (0)
-	je	.2
-	ret
-.2:	jmp	*0
-GLOBL ( _x86_Vertex3f_6_end )
-/*
-	vertex 3f generic size
-*/
-GLOBL ( _x86_Vertex3f )
-	push	%edi
-	push	%esi
-	movl	$0, %esi
-	movl	(0), %edi
-	movl	12(%esp), %eax
-	movl	16(%esp), %edx
-	movl	20(%esp), %ecx
-	movl	%eax, (%edi)
-	movl	%edx, 4(%edi)
-	movl	%ecx, 8(%edi)
-	addl	$12, %edi
-	movl	$0, %ecx
-	repz
-	movsl %ds:(%esi), %es:(%edi)
-	movl	(0), %eax
-	movl	%edi, (0)
-	dec 	%eax
-	movl	%eax, (0)
-	pop 	%esi
-	pop 	%edi
-	je  	.3
-	ret
-.3:	jmp	*0
-
-GLOBL ( _x86_Vertex3f_end )
-
-/*
-	Vertex 3fv vertex size 6
-*/
-GLOBL ( _x86_Vertex3fv_6 )
-	movl	(0), %eax
-	movl	4(%esp), %ecx
-	movl	(%ecx), %edx
-	movl	%edx, (%eax)
-	movl	4(%ecx), %edx
-	movl	8(%ecx), %ecx
-	movl	%edx, 4(%eax)
-	movl	%ecx, 8(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 12(%eax)
-	movl	%ecx, 16(%eax)
-	movl	(36), %edx
-	movl	%edx, 20(%eax)
-	addl	$24, %eax
-	movl	%eax, 0
-	movl	4, %eax
-	dec 	%eax
-	movl	%eax, 4
-	je	.4
-	ret
-.4:	jmp    *8
-	
-GLOBL ( _x86_Vertex3fv_6_end )
-
-/*
-	Vertex 3fv vertex size 8
-*/
-GLOBL ( _x86_Vertex3fv_8 )
-	movl	(0), %eax
-	movl	4(%esp), %ecx
-	movl	(%ecx), %edx
-	movl	%edx ,(%eax)
-	movl	4(%ecx) ,%edx
-	movl	8(%ecx) ,%ecx
-	movl	%edx, 4(%eax)
-	movl	%ecx, 8(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 12(%eax)
-	movl	%ecx, 16(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 20(%eax)
-	movl	%ecx, 24(%eax)
-	movl	(36), %edx
-	movl	%edx, 28(%eax)
-	addl	$32, %eax
-	movl	%eax, (0)
-	movl	4, %eax
-	dec	%eax
-	movl    %eax, (4)
-	je	.5
-	ret
-.5:	jmp    *8
-	
-GLOBL ( _x86_Vertex3fv_8_end )
-
-/*
-	Vertex 3fv generic vertex size
-*/
-GLOBL ( _x86_Vertex3fv )
-	movl	4(%esp), %edx
-	push	%edi
-	push	%esi
-	movl	(0x1010101), %edi
-	movl	(%edx), %eax
-	movl	4(%edx), %ecx
-	movl	8(%edx), %esi
-	movl	%eax, (%edi)
-	movl	%ecx, 4(%edi)
-	movl	%esi, 8(%edi)
-	addl	$12, %edi
-	movl	$6, %ecx
-	movl	$0x58, %esi
-	repz
-	movsl %ds:(%esi), %es:(%edi)
-	movl	%edi, (0x1010101)
-	movl	(0x2020202), %eax
-	pop	%esi
-	pop	%edi
-	dec	%eax
-	movl	%eax, (0x2020202)
-	je	.6
-	ret
-.6:	jmp    *0
-GLOBL ( _x86_Vertex3fv_end )
-
-
-/**
- * Generic handler for 2 float format data.  This can be used for
- * TexCoord2f and possibly other functions.
- */
-
-GLOBL ( _x86_Attribute2f )
-	movl	$0x0, %edx
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	movl	%eax, (%edx)
-	movl	%ecx, 4(%edx)
-	ret
-GLOBL ( _x86_Attribute2f_end )
-
-
-/**
- * Generic handler for 2 float vector format data.  This can be used for
- * TexCoord2fv and possibly other functions.
- */
-
-GLOBL( _x86_Attribute2fv)
-	movl 4(%esp), %eax      /* load 'v' off stack */
-	movl (%eax), %ecx       /* load v[0] */
-	movl 4(%eax), %eax      /* load v[1] */
-	movl %ecx, 0      	/* store v[0] to current vertex */
-	movl %eax, 4      	/* store v[1] to current vertex */
-	ret
-GLOBL ( _x86_Attribute2fv_end )
-
-
-/**
- * Generic handler for 3 float format data.  This can be used for
- * Normal3f, Color3f (when the color target is also float), or
- * TexCoord3f.
- */
-
-GLOBL ( _x86_Attribute3f )
-	movl	4(%esp), %ecx
-	movl	8(%esp), %edx
-	movl	12(%esp), %eax
-	movl	%ecx, 0
-	movl	%edx, 4
-	movl	%eax, 8
-	ret
-GLOBL ( _x86_Attribute3f_end )
-
-/**
- * Generic handler for 3 float vector format data.  This can be used for
- * Normal3f, Color3f (when the color target is also float), or
- * TexCoord3f.
- */
-
-GLOBL( _x86_Attribute3fv)
-	movl 4(%esp), %eax      /* load 'v' off stack */
-	movl (%eax), %ecx       /* load v[0] */
-	movl 4(%eax), %edx      /* load v[1] */
-	movl 8(%eax), %eax      /* load v[2] */
-	movl %ecx, 0      	/* store v[0] to current vertex */
-	movl %edx, 4      	/* store v[1] to current vertex */
-	movl %eax, 8      	/* store v[2] to current vertex */
-	ret
-GLOBL ( _x86_Attribute3fv_end )
-
-
-/*
-	Color 4ubv_ub
-*/
-GLOBL ( _x86_Color4ubv_ub )
-	movl 4(%esp), %eax
-	movl $0x12345678, %edx
-	movl (%eax), %eax
-	movl %eax, (%edx)
-	ret
-GLOBL ( _x86_Color4ubv_ub_end )
-
-/*
-	Color 4ubv 4f
-*/
-GLOBL ( _x86_Color4ubv_4f )
-	push	%ebx
-	movl	$0, %edx
-	xor	%eax, %eax
-	xor	%ecx, %ecx
-	movl	8(%esp), %ebx
-	movl	(%ebx), %ebx
-	mov	%bl, %al
-	mov	%bh, %cl
-	movl	(%edx,%eax,4),%eax
-	movl	(%edx,%ecx,4),%ecx
-	movl	%eax, (0xdeadbeaf)
-	movl	%ecx, (0xdeadbeaf)
-	xor	%eax, %eax
-	xor	%ecx, %ecx
-	shr	$16, %ebx
-	mov	%bl, %al
-	mov	%bh, %cl
-	movl	(%edx,%eax,4), %eax
-	movl	(%edx,%ecx,4), %ecx
-	movl	%eax, (0xdeadbeaf)
-	movl	%ecx, (0xdeadbeaf)
-	pop	%ebx
-	ret
-GLOBL ( _x86_Color4ubv_4f_end )
-
-/*
-
-	Color4ub_ub
-*/
-GLOBL( _x86_Color4ub_ub )
-	push	%ebx
-	movl	8(%esp), %eax
-	movl	12(%esp), %edx
-	movl	16(%esp), %ecx
-	movl	20(%esp), %ebx
-	mov	%al, (0)
-	mov	%dl, (0)
-	mov	%cl, (0)
-	mov	%bl, (0)
-	pop	%ebx
-	ret
-GLOBL( _x86_Color4ub_ub_end )
-
-
-/* \todo: change the "and $7, %eax" to something like "target & 4 ? target & 5 : target & 3)" */
-/*
-	MultiTexCoord2fv st0/st1
-*/
-GLOBL( _x86_MultiTexCoord2fv )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$7, %eax
-	movl	(%ecx), %edx
-	shl	$3, %eax
-	movl	4(%ecx), %ecx
-	movl	%edx, 0xdeadbeef(%eax)
-	movl	%ecx, 0xdeadbeef(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2fv_end )
-
-/*
-	MultiTexCoord2fv
-*/
-
-GLOBL( _x86_MultiTexCoord2fv_2 )
-	movl	4(%esp,1), %eax
-	movl	8(%esp,1), %ecx
-	and	$0x7, %eax
-	movl	0(,%eax,4), %edx
-	movl	(%ecx), %eax
-	movl	%eax, (%edx)
-	movl	4(%ecx), %eax
-	movl	%eax, 4(%edx)
-	ret
-GLOBL( _x86_MultiTexCoord2fv_2_end )
-
-/*
-	MultiTexCoord2f st0/st1
-*/
-GLOBL( _x86_MultiTexCoord2f )
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	12(%esp), %ecx
-	and	$7, %eax
-	shl	$3, %eax
-	movl	%edx, 0xdeadbeef(%eax)
-	movl	%ecx, 0xdeadbeef(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2f_end )
-
-/*
-	MultiTexCoord2f
-*/
-GLOBL( _x86_MultiTexCoord2f_2 )
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	12(%esp,1), %ecx
-	and	$7,%eax
-	movl	0(,%eax,4), %eax
-	movl	%edx, (%eax)
-	movl	%ecx, 4(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2f_2_end )
-
-#if defined(USE_SSE_ASM)
-/**
- * This can be used as a template for either Color3fv (when the color
- * target is also a 3f) or Normal3fv.
- */
-
-GLOBL( _sse_Attribute3fv )
-	movl	4(%esp), %eax
-	movlps	(%eax), %xmm0
-	movl	8(%eax), %eax
-	movlps	%xmm0, 0
-	movl	%eax, 8
-	ret
-GLOBL( _sse_Attribute3fv_end )
-
-/**
- * This can be used as a template for either Color3f (when the color
- * target is also a 3f) or Normal3f.
- */
-
-GLOBL( _sse_Attribute3f )
-	movlps	4(%esp), %xmm0
-	movl	12(%esp), %eax
-	movlps	%xmm0, 0
-	movl	%eax, 8
-	ret
-GLOBL( _sse_Attribute3f_end )
-
-
-/**
- * Generic handler for 2 float vector format data.  This can be used for
- * TexCoord2fv and possibly other functions.
- */
-
-GLOBL( _sse_Attribute2fv )
-	movl	4(%esp), %eax
-	movlps	(%eax), %xmm0
-	movlps	%xmm0, 0
-	ret
-GLOBL( _sse_Attribute2fv_end )
-
-
-/**
- * Generic handler for 2 float format data.  This can be used for
- * TexCoord2f and possibly other functions.
- */
-
-GLOBL( _sse_Attribute2f )
-	movlps	4(%esp), %xmm0
-	movlps	%xmm0, 0
-	ret
-GLOBL( _sse_Attribute2f_end )
-
-/*
-	MultiTexCoord2fv st0/st1
-*/
-GLOBL( _sse_MultiTexCoord2fv )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$7, %eax
-	movlps	(%ecx), %xmm0
-	movlps	%xmm0, 0xdeadbeef(,%eax,8)
-	ret
-GLOBL( _sse_MultiTexCoord2fv_end )
-
-/*
-	MultiTexCoord2fv
-*/
-GLOBL( _sse_MultiTexCoord2fv_2 )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$0x7, %eax
-	movl	0(,%eax,4), %edx
-	movlps	(%ecx), %xmm0
-	movlps	%xmm0, (%edx)
-	ret
-GLOBL( _sse_MultiTexCoord2fv_2_end )
-
-/*
-	MultiTexCoord2f st0/st1
-*/
-GLOBL( _sse_MultiTexCoord2f )
-	movl	4(%esp), %eax
-	and	$7, %eax
-	movlps	8(%esp), %xmm0
-	movlps	%xmm0, 0xdeadbeef(,%eax,8)
-	ret
-GLOBL( _sse_MultiTexCoord2f_end )
-
-/*
-	MultiTexCoord2f
-*/
-GLOBL( _sse_MultiTexCoord2f_2 )
-	movl	4(%esp), %eax
-	movlps	8(%esp), %xmm0
-	and	$7,%eax
-	movl	0(,%eax,4), %eax
-	movlps	%xmm0, (%eax)
-	ret
-GLOBL( _sse_MultiTexCoord2f_2_end )
-#endif
-
-#if defined (__ELF__) && defined (__linux__)
-	.section .note.GNU-stack,"",%progbits
-#endif
diff --git a/src/mesa/drivers/dri/r300/Makefile b/src/mesa/drivers/dri/r300/Makefile
index 5abb91d9876..2ee23289341 100644
--- a/src/mesa/drivers/dri/r300/Makefile
+++ b/src/mesa/drivers/dri/r300/Makefile
@@ -90,9 +90,11 @@ COMMON_SYMLINKS = \
 
 include ../Makefile.template
 
-$(SYMLINKS):
+server:
 	mkdir -p server
-	for i in $(SYMLINKS) ; do rm -f $$i && test -f ../radeon/$$i && ln -s ../../radeon/$$i $$i ; done
+
+$(SYMLINKS): server
+	@[ -e $@ ] || ln -sf ../../radeon/$@ server/
 
 $(COMMON_SYMLINKS):
 	@[ -e $@ ] || ln -sf ../radeon/$@ ./
diff --git a/src/mesa/drivers/dri/r300/r200_context.h b/src/mesa/drivers/dri/r300/r200_context.h
deleted file mode 100644
index a06d7152d78..00000000000
--- a/src/mesa/drivers/dri/r300/r200_context.h
+++ /dev/null
@@ -1,822 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_context.h,v 1.2 2002/12/16 16:18:54 dawes Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __R200_CONTEXT_H__
-#define __R200_CONTEXT_H__
-
-#ifdef GLX_DIRECT_RENDERING
-
-#include "tnl/t_vertex.h"
-#include "drm.h"
-#include "radeon_drm.h"
-#include "dri_util.h"
-#include "texmem.h"
-
-#include "macros.h"
-#include "mtypes.h"
-#include "colormac.h"
-#include "r200_reg.h"
-#include "radeon_context.h"
-
-#define ENABLE_HW_3D_TEXTURE 1	/* XXX this is temporary! */
-
-struct r200_context;
-typedef struct r200_context r200ContextRec;
-typedef struct r200_context *r200ContextPtr;
-
-#include "mm.h"
-
-/* The blit width for texture uploads
- */
-#define BLIT_WIDTH_BYTES 1024
-
-/* Use the templated vertex format:
- */
-#define COLOR_IS_RGBA
-#define TAG(x) r200##x
-#include "tnl_dd/t_dd_vertex.h"
-#undef TAG
-
-typedef void (*r200_tri_func) (r200ContextPtr,
-			       r200Vertex *, r200Vertex *, r200Vertex *);
-
-typedef void (*r200_line_func) (r200ContextPtr, r200Vertex *, r200Vertex *);
-
-typedef void (*r200_point_func) (r200ContextPtr, r200Vertex *);
-
-struct r200_depthbuffer_state {
-	GLfloat scale;
-};
-
-struct r200_stencilbuffer_state {
-	GLboolean hwBuffer;
-	GLuint clear;		/* rb3d_stencilrefmask value */
-};
-
-struct r200_stipple_state {
-	GLuint mask[32];
-};
-
-typedef struct r200_tex_obj r200TexObj, *r200TexObjPtr;
-
-/* Texture object in locally shared texture space.
- */
-struct r200_tex_obj {
-	driTextureObject base;
-
-	GLuint bufAddr;		/* Offset to start of locally
-				   shared texture block */
-
-	GLuint dirty_state;	/* Flags (1 per texunit) for
-				   whether or not this texobj
-				   has dirty hardware state
-				   (pp_*) that needs to be
-				   brought into the
-				   texunit. */
-
-	drm_radeon_tex_image_t image[6][RADEON_MAX_TEXTURE_LEVELS];
-	/* Six, for the cube faces */
-
-	GLuint pp_txfilter;	/* hardware register values */
-	GLuint pp_txformat;
-	GLuint pp_txformat_x;
-	GLuint pp_txoffset;	/* Image location in texmem.
-				   All cube faces follow. */
-	GLuint pp_txsize;	/* npot only */
-	GLuint pp_txpitch;	/* npot only */
-	GLuint pp_border_color;
-	GLuint pp_cubic_faces;	/* cube face 1,2,3,4 log2 sizes */
-
-	GLboolean border_fallback;
-};
-
-struct r200_texture_env_state {
-	r200TexObjPtr texobj;
-	GLenum format;
-	GLenum envMode;
-};
-
-#define R200_MAX_TEXTURE_UNITS 6
-
-struct r200_texture_state {
-	struct r200_texture_env_state unit[R200_MAX_TEXTURE_UNITS];
-};
-
-struct r200_state_atom {
-	struct r200_state_atom *next, *prev;
-	const char *name;	/* for debug */
-	int cmd_size;		/* size in bytes */
-	GLuint idx;
-	int *cmd;		/* one or more cmd's */
-	int *lastcmd;		/* one or more cmd's */
-	int *savedcmd;		/* one or more cmd's */
-	GLboolean dirty;
-	 GLboolean(*check) (GLcontext *, int);	/* is this state active? */
-};
-
-/* Trying to keep these relatively short as the variables are becoming
- * extravagently long.  Drop the driver name prefix off the front of
- * everything - I think we know which driver we're in by now, and keep the
- * prefix to 3 letters unless absolutely impossible.
- */
-
-#define CTX_CMD_0             0
-#define CTX_PP_MISC           1
-#define CTX_PP_FOG_COLOR      2
-#define CTX_RE_SOLID_COLOR    3
-#define CTX_RB3D_BLENDCNTL    4
-#define CTX_RB3D_DEPTHOFFSET  5
-#define CTX_RB3D_DEPTHPITCH   6
-#define CTX_RB3D_ZSTENCILCNTL 7
-#define CTX_CMD_1             8
-#define CTX_PP_CNTL           9
-#define CTX_RB3D_CNTL         10
-#define CTX_RB3D_COLOROFFSET  11
-#define CTX_CMD_2             12	/* why */
-#define CTX_RB3D_COLORPITCH   13	/* why */
-#define CTX_STATE_SIZE_OLDDRM 14
-#define CTX_CMD_3             14
-#define CTX_RB3D_BLENDCOLOR   15
-#define CTX_RB3D_ABLENDCNTL   16
-#define CTX_RB3D_CBLENDCNTL   17
-#define CTX_STATE_SIZE_NEWDRM 18
-
-#define SET_CMD_0               0
-#define SET_SE_CNTL             1
-#define SET_RE_CNTL             2	/* replace se_coord_fmt */
-#define SET_STATE_SIZE          3
-
-#define VTE_CMD_0               0
-#define VTE_SE_VTE_CNTL         1
-#define VTE_STATE_SIZE          2
-
-#define LIN_CMD_0               0
-#define LIN_RE_LINE_PATTERN     1
-#define LIN_RE_LINE_STATE       2
-#define LIN_CMD_1               3
-#define LIN_SE_LINE_WIDTH       4
-#define LIN_STATE_SIZE          5
-
-#define MSK_CMD_0               0
-#define MSK_RB3D_STENCILREFMASK 1
-#define MSK_RB3D_ROPCNTL        2
-#define MSK_RB3D_PLANEMASK      3
-#define MSK_STATE_SIZE          4
-
-#define VPT_CMD_0           0
-#define VPT_SE_VPORT_XSCALE          1
-#define VPT_SE_VPORT_XOFFSET         2
-#define VPT_SE_VPORT_YSCALE          3
-#define VPT_SE_VPORT_YOFFSET         4
-#define VPT_SE_VPORT_ZSCALE          5
-#define VPT_SE_VPORT_ZOFFSET         6
-#define VPT_STATE_SIZE      7
-
-#define ZBS_CMD_0               0
-#define ZBS_SE_ZBIAS_FACTOR     1
-#define ZBS_SE_ZBIAS_CONSTANT   2
-#define ZBS_STATE_SIZE          3
-
-#define MSC_CMD_0               0
-#define MSC_RE_MISC             1
-#define MSC_STATE_SIZE          2
-
-#define TAM_CMD_0               0
-#define TAM_DEBUG3              1
-#define TAM_STATE_SIZE          2
-
-#define TEX_CMD_0                   0
-#define TEX_PP_TXFILTER             1	/*2c00 */
-#define TEX_PP_TXFORMAT             2	/*2c04 */
-#define TEX_PP_TXFORMAT_X           3	/*2c08 */
-#define TEX_PP_TXSIZE               4	/*2c0c */
-#define TEX_PP_TXPITCH              5	/*2c10 */
-#define TEX_PP_BORDER_COLOR         6	/*2c14 */
-#define TEX_CMD_1                   7
-#define TEX_PP_TXOFFSET             8	/*2d00 */
-#define TEX_STATE_SIZE              9
-
-#define CUBE_CMD_0                  0	/* 1 register follows */
-#define CUBE_PP_CUBIC_FACES         1	/* 0x2c18 */
-#define CUBE_CMD_1                  2	/* 5 registers follow */
-#define CUBE_PP_CUBIC_OFFSET_F1     3	/* 0x2d04 */
-#define CUBE_PP_CUBIC_OFFSET_F2     4	/* 0x2d08 */
-#define CUBE_PP_CUBIC_OFFSET_F3     5	/* 0x2d0c */
-#define CUBE_PP_CUBIC_OFFSET_F4     6	/* 0x2d10 */
-#define CUBE_PP_CUBIC_OFFSET_F5     7	/* 0x2d14 */
-#define CUBE_STATE_SIZE             8
-
-#define PIX_CMD_0                   0
-#define PIX_PP_TXCBLEND             1
-#define PIX_PP_TXCBLEND2            2
-#define PIX_PP_TXABLEND             3
-#define PIX_PP_TXABLEND2            4
-#define PIX_STATE_SIZE              5
-
-#define TF_CMD_0                    0
-#define TF_TFACTOR_0                1
-#define TF_TFACTOR_1                2
-#define TF_TFACTOR_2                3
-#define TF_TFACTOR_3                4
-#define TF_TFACTOR_4                5
-#define TF_TFACTOR_5                6
-#define TF_STATE_SIZE               7
-
-#define TCL_CMD_0                 0
-#define TCL_LIGHT_MODEL_CTL_0     1
-#define TCL_LIGHT_MODEL_CTL_1     2
-#define TCL_PER_LIGHT_CTL_0       3
-#define TCL_PER_LIGHT_CTL_1       4
-#define TCL_PER_LIGHT_CTL_2       5
-#define TCL_PER_LIGHT_CTL_3       6
-#define TCL_CMD_1                 7
-#define TCL_UCP_VERT_BLEND_CTL    8
-#define TCL_STATE_SIZE            9
-
-#define MSL_CMD_0                     0
-#define MSL_MATRIX_SELECT_0           1
-#define MSL_MATRIX_SELECT_1           2
-#define MSL_MATRIX_SELECT_2           3
-#define MSL_MATRIX_SELECT_3           4
-#define MSL_MATRIX_SELECT_4           5
-#define MSL_STATE_SIZE                6
-
-#define TCG_CMD_0                 0
-#define TCG_TEX_PROC_CTL_2            1
-#define TCG_TEX_PROC_CTL_3            2
-#define TCG_TEX_PROC_CTL_0            3
-#define TCG_TEX_PROC_CTL_1            4
-#define TCG_TEX_CYL_WRAP_CTL      5
-#define TCG_STATE_SIZE            6
-
-#define MTL_CMD_0            0
-#define MTL_EMMISSIVE_RED    1
-#define MTL_EMMISSIVE_GREEN  2
-#define MTL_EMMISSIVE_BLUE   3
-#define MTL_EMMISSIVE_ALPHA  4
-#define MTL_AMBIENT_RED      5
-#define MTL_AMBIENT_GREEN    6
-#define MTL_AMBIENT_BLUE     7
-#define MTL_AMBIENT_ALPHA    8
-#define MTL_DIFFUSE_RED      9
-#define MTL_DIFFUSE_GREEN    10
-#define MTL_DIFFUSE_BLUE     11
-#define MTL_DIFFUSE_ALPHA    12
-#define MTL_SPECULAR_RED     13
-#define MTL_SPECULAR_GREEN   14
-#define MTL_SPECULAR_BLUE    15
-#define MTL_SPECULAR_ALPHA   16
-#define MTL_CMD_1            17
-#define MTL_SHININESS        18
-#define MTL_STATE_SIZE       19
-
-#define VAP_CMD_0                   0
-#define VAP_SE_VAP_CNTL             1
-#define VAP_STATE_SIZE              2
-
-/* Replaces a lot of packet info from radeon
- */
-#define VTX_CMD_0                   0
-#define VTX_VTXFMT_0            1
-#define VTX_VTXFMT_1            2
-#define VTX_TCL_OUTPUT_VTXFMT_0 3
-#define VTX_TCL_OUTPUT_VTXFMT_1 4
-#define VTX_CMD_1               5
-#define VTX_TCL_OUTPUT_COMPSEL  6
-#define VTX_CMD_2               7
-#define VTX_STATE_CNTL          8
-#define VTX_STATE_SIZE          9
-
-#define VTX_COLOR(v,n)   (((v)>>(R200_VTX_COLOR_0_SHIFT+(n)*2))&\
-                         R200_VTX_COLOR_MASK)
-
-/**
- * Given the \c R200_SE_VTX_FMT_1 for the current vertex state, determine
- * how many components are in texture coordinate \c n.
- */
-#define VTX_TEXn_COUNT(v,n)   (((v) >> (3 * n)) & 0x07)
-
-#define MAT_CMD_0              0
-#define MAT_ELT_0              1
-#define MAT_STATE_SIZE         17
-
-#define GRD_CMD_0                  0
-#define GRD_VERT_GUARD_CLIP_ADJ    1
-#define GRD_VERT_GUARD_DISCARD_ADJ 2
-#define GRD_HORZ_GUARD_CLIP_ADJ    3
-#define GRD_HORZ_GUARD_DISCARD_ADJ 4
-#define GRD_STATE_SIZE             5
-
-/* position changes frequently when lighting in modelpos - separate
- * out to new state item?
- */
-#define LIT_CMD_0                  0
-#define LIT_AMBIENT_RED            1
-#define LIT_AMBIENT_GREEN          2
-#define LIT_AMBIENT_BLUE           3
-#define LIT_AMBIENT_ALPHA          4
-#define LIT_DIFFUSE_RED            5
-#define LIT_DIFFUSE_GREEN          6
-#define LIT_DIFFUSE_BLUE           7
-#define LIT_DIFFUSE_ALPHA          8
-#define LIT_SPECULAR_RED           9
-#define LIT_SPECULAR_GREEN         10
-#define LIT_SPECULAR_BLUE          11
-#define LIT_SPECULAR_ALPHA         12
-#define LIT_POSITION_X             13
-#define LIT_POSITION_Y             14
-#define LIT_POSITION_Z             15
-#define LIT_POSITION_W             16
-#define LIT_DIRECTION_X            17
-#define LIT_DIRECTION_Y            18
-#define LIT_DIRECTION_Z            19
-#define LIT_DIRECTION_W            20
-#define LIT_ATTEN_QUADRATIC        21
-#define LIT_ATTEN_LINEAR           22
-#define LIT_ATTEN_CONST            23
-#define LIT_ATTEN_XXX              24
-#define LIT_CMD_1                  25
-#define LIT_SPOT_DCD               26
-#define LIT_SPOT_DCM               27
-#define LIT_SPOT_EXPONENT          28
-#define LIT_SPOT_CUTOFF            29
-#define LIT_SPECULAR_THRESH        30
-#define LIT_RANGE_CUTOFF           31	/* ? */
-#define LIT_ATTEN_CONST_INV        32
-#define LIT_STATE_SIZE             33
-
-/* Fog
- */
-#define FOG_CMD_0      0
-#define FOG_R          1
-#define FOG_C          2
-#define FOG_D          3
-#define FOG_PAD        4
-#define FOG_STATE_SIZE 5
-
-/* UCP
- */
-#define UCP_CMD_0      0
-#define UCP_X          1
-#define UCP_Y          2
-#define UCP_Z          3
-#define UCP_W          4
-#define UCP_STATE_SIZE 5
-
-/* GLT - Global ambient
- */
-#define GLT_CMD_0      0
-#define GLT_RED        1
-#define GLT_GREEN      2
-#define GLT_BLUE       3
-#define GLT_ALPHA      4
-#define GLT_STATE_SIZE 5
-
-/* EYE
- */
-#define EYE_CMD_0          0
-#define EYE_X              1
-#define EYE_Y              2
-#define EYE_Z              3
-#define EYE_RESCALE_FACTOR 4
-#define EYE_STATE_SIZE     5
-
-/* CST - constant state
- */
-#define CST_CMD_0                             0
-#define CST_PP_CNTL_X                         1
-#define CST_CMD_1                             2
-#define CST_RB3D_DEPTHXY_OFFSET               3
-#define CST_CMD_2                             4
-#define CST_RE_AUX_SCISSOR_CNTL               5
-#define CST_CMD_3                             6
-#define CST_RE_SCISSOR_TL_0                   7
-#define CST_RE_SCISSOR_BR_0                   8
-#define CST_CMD_4                             9
-#define CST_SE_VAP_CNTL_STATUS                10
-#define CST_CMD_5                             11
-#define CST_RE_POINTSIZE                      12
-#define CST_CMD_6                             13
-#define CST_SE_TCL_INPUT_VTX_0                14
-#define CST_SE_TCL_INPUT_VTX_1                15
-#define CST_SE_TCL_INPUT_VTX_2                16
-#define CST_SE_TCL_INPUT_VTX_3                17
-#define CST_STATE_SIZE                        18
-
-struct r200_hw_state {
-	/* Head of the linked list of state atoms. */
-	struct r200_state_atom atomlist;
-
-	/* Hardware state, stored as cmdbuf commands:
-	 *   -- Need to doublebuffer for
-	 *           - reviving state after loss of context
-	 *           - eliding noop statechange loops? (except line stipple count)
-	 */
-	struct r200_state_atom ctx;
-	struct r200_state_atom set;
-	struct r200_state_atom vte;
-	struct r200_state_atom lin;
-	struct r200_state_atom msk;
-	struct r200_state_atom vpt;
-	struct r200_state_atom vap;
-	struct r200_state_atom vtx;
-	struct r200_state_atom tcl;
-	struct r200_state_atom msl;
-	struct r200_state_atom tcg;
-	struct r200_state_atom msc;
-	struct r200_state_atom cst;
-	struct r200_state_atom tam;
-	struct r200_state_atom tf;
-	struct r200_state_atom tex[6];
-	struct r200_state_atom cube[6];
-	struct r200_state_atom zbs;
-	struct r200_state_atom mtl[2];
-	struct r200_state_atom mat[9];
-	struct r200_state_atom lit[8];	/* includes vec, scl commands */
-	struct r200_state_atom ucp[6];
-	struct r200_state_atom pix[6];	/* pixshader stages */
-	struct r200_state_atom eye;	/* eye pos */
-	struct r200_state_atom grd;	/* guard band clipping */
-	struct r200_state_atom fog;
-	struct r200_state_atom glt;
-
-	int max_state_size;	/* Number of bytes necessary for a full state emit. */
-	GLboolean is_dirty, all_dirty;
-};
-
-struct r200_colorbuffer_state {
-	int roundEnable;
-};
-
-struct r200_state {
-	/* Derived state for internal purposes:
-	 */
-	struct r200_colorbuffer_state color;
-	struct r200_depthbuffer_state depth;
-	struct r200_stencilbuffer_state stencil;
-	struct r200_stipple_state stipple;
-	struct r200_texture_state texture;
-};
-
-/* Need refcounting on dma buffers:
- */
-struct r200_dma_buffer {
-	int refcount;		/* the number of retained regions in buf */
-	drmBufPtr buf;
-};
-
-#define GET_START(rvb) (rmesa->radeon.radeonScreen->gart_buffer_offset +		\
-			(rvb)->address - rmesa->dma.buf0_address +	\
-			(rvb)->start)
-
-/* A retained region, eg vertices for indexed vertices.
- */
-struct r200_dma_region {
-	struct r200_dma_buffer *buf;
-	char *address;		/* == buf->address */
-	int start, end, ptr;	/* offsets from start of buf */
-	int aos_start;
-	int aos_stride;
-	int aos_size;
-};
-
-struct r200_dma {
-	/* Active dma region.  Allocations for vertices and retained
-	 * regions come from here.  Also used for emitting random vertices,
-	 * these may be flushed by calling flush_current();
-	 */
-	struct r200_dma_region current;
-
-	void (*flush) (r200ContextPtr);
-
-	char *buf0_address;	/* start of buf[0], for index calcs */
-	GLuint nr_released_bufs;	/* flush after so many buffers released */
-};
-
-#define R200_CMD_BUF_SZ  (8*1024)
-
-struct r200_store {
-	GLuint statenr;
-	GLuint primnr;
-	char cmd_buf[R200_CMD_BUF_SZ];
-	int cmd_used;
-	int elts_start;
-};
-
-/* r200_tcl.c
- */
-struct r200_tcl_info {
-	GLuint vertex_format;
-	GLint last_offset;
-	GLuint hw_primitive;
-
-	struct r200_dma_region *aos_components[8];
-	GLuint nr_aos_components;
-
-	GLuint *Elts;
-
-	struct r200_dma_region indexed_verts;
-	struct r200_dma_region obj;
-	struct r200_dma_region rgba;
-	struct r200_dma_region spec;
-	struct r200_dma_region fog;
-	struct r200_dma_region tex[R200_MAX_TEXTURE_UNITS];
-	struct r200_dma_region norm;
-};
-
-/* r200_swtcl.c
- */
-struct r200_swtcl_info {
-	GLuint RenderIndex;
-
-   /**
-    * Size of a hardware vertex.  This is calculated when \c ::vertex_attrs is
-    * installed in the Mesa state vector.
-    */
-	GLuint vertex_size;
-
-   /**
-    * Attributes instructing the Mesa TCL pipeline where / how to put vertex
-    * data in the hardware buffer.
-    */
-	struct tnl_attr_map vertex_attrs[VERT_ATTRIB_MAX];
-
-   /**
-    * Number of elements of \c ::vertex_attrs that are actually used.
-    */
-	GLuint vertex_attr_count;
-
-   /**
-    * Cached pointer to the buffer where Mesa will store vertex data.
-    */
-	GLubyte *verts;
-
-	/* Fallback rasterization functions
-	 */
-	r200_point_func draw_point;
-	r200_line_func draw_line;
-	r200_tri_func draw_tri;
-
-	GLuint hw_primitive;
-	GLenum render_primitive;
-	GLuint numverts;
-
-   /**
-    * Offset of the 4UB color data within a hardware (swtcl) vertex.
-    */
-	GLuint coloroffset;
-
-   /**
-    * Offset of the 3UB specular color data within a hardware (swtcl) vertex.
-    */
-	GLuint specoffset;
-
-   /**
-    * Should Mesa project vertex data or will the hardware do it?
-    */
-	GLboolean needproj;
-
-	struct r200_dma_region indexed_verts;
-};
-
-struct r200_ioctl {
-	GLuint vertex_offset;
-	GLuint vertex_size;
-};
-
-#define R200_MAX_PRIMS 64
-
-/* Want to keep a cache of these around.  Each is parameterized by
- * only a single value which has only a small range.  Only expect a
- * few, so just rescan the list each time?
- */
-struct dynfn {
-	struct dynfn *next, *prev;
-	int key[2];
-	char *code;
-};
-
-struct dfn_lists {
-	struct dynfn Vertex2f;
-	struct dynfn Vertex2fv;
-	struct dynfn Vertex3f;
-	struct dynfn Vertex3fv;
-	struct dynfn Color4ub;
-	struct dynfn Color4ubv;
-	struct dynfn Color3ub;
-	struct dynfn Color3ubv;
-	struct dynfn Color4f;
-	struct dynfn Color4fv;
-	struct dynfn Color3f;
-	struct dynfn Color3fv;
-	struct dynfn SecondaryColor3ubEXT;
-	struct dynfn SecondaryColor3ubvEXT;
-	struct dynfn SecondaryColor3fEXT;
-	struct dynfn SecondaryColor3fvEXT;
-	struct dynfn Normal3f;
-	struct dynfn Normal3fv;
-	struct dynfn TexCoord3f;
-	struct dynfn TexCoord3fv;
-	struct dynfn TexCoord2f;
-	struct dynfn TexCoord2fv;
-	struct dynfn TexCoord1f;
-	struct dynfn TexCoord1fv;
-	struct dynfn MultiTexCoord3fARB;
-	struct dynfn MultiTexCoord3fvARB;
-	struct dynfn MultiTexCoord2fARB;
-	struct dynfn MultiTexCoord2fvARB;
-	struct dynfn MultiTexCoord1fARB;
-	struct dynfn MultiTexCoord1fvARB;
-};
-
-struct dfn_generators {
-	struct dynfn *(*Vertex2f) (GLcontext *, const int *);
-	struct dynfn *(*Vertex2fv) (GLcontext *, const int *);
-	struct dynfn *(*Vertex3f) (GLcontext *, const int *);
-	struct dynfn *(*Vertex3fv) (GLcontext *, const int *);
-	struct dynfn *(*Color4ub) (GLcontext *, const int *);
-	struct dynfn *(*Color4ubv) (GLcontext *, const int *);
-	struct dynfn *(*Color3ub) (GLcontext *, const int *);
-	struct dynfn *(*Color3ubv) (GLcontext *, const int *);
-	struct dynfn *(*Color4f) (GLcontext *, const int *);
-	struct dynfn *(*Color4fv) (GLcontext *, const int *);
-	struct dynfn *(*Color3f) (GLcontext *, const int *);
-	struct dynfn *(*Color3fv) (GLcontext *, const int *);
-	struct dynfn *(*SecondaryColor3ubEXT) (GLcontext *, const int *);
-	struct dynfn *(*SecondaryColor3ubvEXT) (GLcontext *, const int *);
-	struct dynfn *(*SecondaryColor3fEXT) (GLcontext *, const int *);
-	struct dynfn *(*SecondaryColor3fvEXT) (GLcontext *, const int *);
-	struct dynfn *(*Normal3f) (GLcontext *, const int *);
-	struct dynfn *(*Normal3fv) (GLcontext *, const int *);
-	struct dynfn *(*TexCoord3f) (GLcontext *, const int *);
-	struct dynfn *(*TexCoord3fv) (GLcontext *, const int *);
-	struct dynfn *(*TexCoord2f) (GLcontext *, const int *);
-	struct dynfn *(*TexCoord2fv) (GLcontext *, const int *);
-	struct dynfn *(*TexCoord1f) (GLcontext *, const int *);
-	struct dynfn *(*TexCoord1fv) (GLcontext *, const int *);
-	struct dynfn *(*MultiTexCoord3fARB) (GLcontext *, const int *);
-	struct dynfn *(*MultiTexCoord3fvARB) (GLcontext *, const int *);
-	struct dynfn *(*MultiTexCoord2fARB) (GLcontext *, const int *);
-	struct dynfn *(*MultiTexCoord2fvARB) (GLcontext *, const int *);
-	struct dynfn *(*MultiTexCoord1fARB) (GLcontext *, const int *);
-	struct dynfn *(*MultiTexCoord1fvARB) (GLcontext *, const int *);
-};
-
-struct r200_prim {
-	GLuint start;
-	GLuint end;
-	GLuint prim;
-};
-
-   /* A maximum total of 29 elements per vertex:  3 floats for position, 3
-    * floats for normal, 4 floats for color, 4 bytes for secondary color,
-    * 3 floats for each texture unit (18 floats total).
-    *
-    * we maybe need add. 4 to prevent segfault if someone specifies
-    * GL_TEXTURE6/GL_TEXTURE7 (esp. for the codegen-path) (FIXME: )
-    *
-    * The position data is never actually stored here, so 3 elements could be
-    * trimmed out of the buffer.
-    */
-
-#define R200_MAX_VERTEX_SIZE ((3*6)+11)
-
-struct r200_vbinfo {
-	GLint counter, initial_counter;
-	GLint *dmaptr;
-	void (*notify) (void);
-	GLint vertex_size;
-
-	union {
-		float f;
-		int i;
-		r200_color_t color;
-	} vertex[R200_MAX_VERTEX_SIZE];
-
-	GLfloat *normalptr;
-	GLfloat *floatcolorptr;
-	r200_color_t *colorptr;
-	GLfloat *floatspecptr;
-	r200_color_t *specptr;
-	GLfloat *texcoordptr[8];	/* 6 (TMU) + 2 for r200_vtxfmt_c.c when GL_TEXTURE6/7 */
-
-	GLenum *prim;		/* &ctx->Driver.CurrentExecPrimitive */
-	GLuint primflags;
-	GLboolean enabled;	/* *_NO_VTXFMT / *_NO_TCL env vars */
-	GLboolean installed;
-	GLboolean fell_back;
-	GLboolean recheck;
-	GLint nrverts;
-	GLuint vtxfmt_0, vtxfmt_1;
-
-	GLuint installed_vertex_format;
-	GLuint installed_color_3f_sz;
-
-	struct r200_prim primlist[R200_MAX_PRIMS];
-	int nrprims;
-
-	struct dfn_lists dfn_cache;
-	struct dfn_generators codegen;
-	GLvertexformat vtxfmt;
-};
-
-/**
- * R200 context structure.
- */
-struct r200_context {
-	struct radeon_context radeon; /* parent class, must be first */
-
-	/* Driver and hardware state management
-	 */
-	struct r200_hw_state hw;
-	struct r200_state state;
-
-	/* Texture object bookkeeping
-	 */
-	unsigned nr_heaps;
-	driTexHeap *texture_heaps[RADEON_NR_TEX_HEAPS];
-	driTextureObject swapped;
-	int texture_depth;
-	float initialMaxAnisotropy;
-
-	/* Rasterization and vertex state:
-	 */
-	GLuint NewGLState;
-
-	/* Vertex buffers
-	 */
-	struct r200_ioctl ioctl;
-	struct r200_dma dma;
-	struct r200_store store;
-	GLboolean save_on_next_unlock;
-
-	/* Clientdata textures;
-	 */
-	GLuint prefer_gart_client_texturing;
-
-	/* TCL stuff
-	 */
-	GLmatrix TexGenMatrix[R200_MAX_TEXTURE_UNITS];
-	GLboolean recheck_texgen[R200_MAX_TEXTURE_UNITS];
-	GLboolean TexGenNeedNormals[R200_MAX_TEXTURE_UNITS];
-	GLuint TexMatEnabled;
-	GLuint TexMatCompSel;
-	GLuint TexGenEnabled;
-	GLuint TexGenInputs;
-	GLuint TexGenCompSel;
-	GLmatrix tmpmat;
-
-	/* r200_tcl.c
-	 */
-	struct r200_tcl_info tcl;
-
-	/* r200_swtcl.c
-	 */
-	struct r200_swtcl_info swtcl;
-
-	/* r200_vtxfmt.c
-	 */
-	struct r200_vbinfo vb;
-};
-
-#define R200_CONTEXT(ctx)		((r200ContextPtr)(ctx->DriverCtx))
-
-extern void r200DestroyContext(__DRIcontextPrivate * driContextPriv);
-extern GLboolean r200CreateContext(const __GLcontextModes * glVisual,
-				   __DRIcontextPrivate * driContextPriv,
-				   void *sharedContextPrivate);
-
-#endif
-#endif				/* __R200_CONTEXT_H__ */
diff --git a/src/mesa/drivers/dri/r300/r200_ioctl.h b/src/mesa/drivers/dri/r300/r200_ioctl.h
deleted file mode 100644
index db7bd7697f1..00000000000
--- a/src/mesa/drivers/dri/r300/r200_ioctl.h
+++ /dev/null
@@ -1,204 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __R200_IOCTL_H__
-#define __R200_IOCTL_H__
-
-#include "simple_list.h"
-#include "radeon_dri.h"
-#include "radeon_lock.h"
-
-#include "xf86drm.h"
-#include "drm.h"
-#include "r200_context.h"
-#include "radeon_drm.h"
-
-extern void r200EmitState(r200ContextPtr rmesa);
-extern void r200EmitVertexAOS(r200ContextPtr rmesa,
-			      GLuint vertex_size, GLuint offset);
-
-extern void r200EmitVbufPrim(r200ContextPtr rmesa,
-			     GLuint primitive, GLuint vertex_nr);
-
-extern void r200FlushElts(r200ContextPtr rmesa);
-
-extern GLushort *r200AllocEltsOpenEnded(r200ContextPtr rmesa,
-					GLuint primitive, GLuint min_nr);
-
-extern void r200EmitAOS(r200ContextPtr rmesa,
-			struct r200_dma_region **regions,
-			GLuint n, GLuint offset);
-
-extern void r200EmitBlit(r200ContextPtr rmesa,
-			 GLuint color_fmt,
-			 GLuint src_pitch,
-			 GLuint src_offset,
-			 GLuint dst_pitch,
-			 GLuint dst_offset,
-			 GLint srcx, GLint srcy,
-			 GLint dstx, GLint dsty, GLuint w, GLuint h);
-
-extern void r200EmitWait(r200ContextPtr rmesa, GLuint flags);
-
-extern void r200FlushCmdBuf(r200ContextPtr rmesa, const char *);
-extern int r200FlushCmdBufLocked(r200ContextPtr rmesa, const char *caller);
-extern void r200Flush(GLcontext * ctx);
-
-extern void r200RefillCurrentDmaRegion(r200ContextPtr rmesa);
-
-extern void r200AllocDmaRegion(r200ContextPtr rmesa,
-			       struct r200_dma_region *region,
-			       int bytes, int alignment);
-
-extern void r200AllocDmaRegionVerts(r200ContextPtr rmesa,
-				    struct r200_dma_region *region,
-				    int numverts, int vertsize, int alignment);
-
-extern void r200ReleaseDmaRegion(r200ContextPtr rmesa,
-				 struct r200_dma_region *region,
-				 const char *caller);
-
-extern void r200WaitForVBlank(r200ContextPtr rmesa);
-extern void r200InitIoctlFuncs(struct dd_function_table *functions);
-
-extern void *r200AllocateMemoryMESA(__DRInativeDisplay * dpy, int scrn,
-				    GLsizei size, GLfloat readfreq,
-				    GLfloat writefreq, GLfloat priority);
-extern void r200FreeMemoryMESA(__DRInativeDisplay * dpy, int scrn,
-			       GLvoid * pointer);
-extern GLuint r200GetMemoryOffsetMESA(__DRInativeDisplay * dpy, int scrn,
-				      const GLvoid * pointer);
-
-extern GLboolean r200IsGartMemory(r200ContextPtr rmesa, const GLvoid * pointer,
-				  GLint size);
-
-extern GLuint r200GartOffsetFromVirtual(r200ContextPtr rmesa,
-					const GLvoid * pointer);
-
-void r200SaveHwState(r200ContextPtr radeon);
-void r200SetUpAtomList(r200ContextPtr rmesa);
-
-/* ================================================================
- * Helper macros:
- */
-
-/* Close off the last primitive, if it exists.
- */
-#define R200_NEWPRIM( rmesa )			\
-do {						\
-   if ( rmesa->dma.flush )			\
-      rmesa->dma.flush( rmesa );	\
-} while (0)
-
-/* Can accomodate several state changes and primitive changes without
- * actually firing the buffer.
- */
-#define R200_STATECHANGE( rmesa, ATOM )			\
-do {								\
-   R200_NEWPRIM( rmesa );					\
-   rmesa->hw.ATOM.dirty = GL_TRUE;				\
-   rmesa->hw.is_dirty = GL_TRUE;				\
-} while (0)
-
-#define R200_DB_STATE( ATOM )			        \
-   memcpy( rmesa->hw.ATOM.lastcmd, rmesa->hw.ATOM.cmd,	\
-	   rmesa->hw.ATOM.cmd_size * 4)
-
-static __inline int R200_DB_STATECHANGE(r200ContextPtr rmesa,
-					struct r200_state_atom *atom)
-{
-	if (memcmp(atom->cmd, atom->lastcmd, atom->cmd_size * 4)) {
-		int *tmp;
-		R200_NEWPRIM(rmesa);
-		atom->dirty = GL_TRUE;
-		rmesa->hw.is_dirty = GL_TRUE;
-		tmp = atom->cmd;
-		atom->cmd = atom->lastcmd;
-		atom->lastcmd = tmp;
-		return 1;
-	} else
-		return 0;
-}
-
-/* Fire the buffered vertices no matter what.
- */
-#define R200_FIREVERTICES( r200 )			\
-do {							\
-   if ( (r200)->store.cmd_used || (r200)->dma.flush ) {	\
-      radeonFlush( (r200)->radeon.glCtx );		\
-   }							\
-} while (0)
-
-/* Command lengths.  Note that any time you ensure ELTS_BUFSZ or VBUF_BUFSZ
- * are available, you will also be adding an rmesa->state.max_state_size because
- * r200EmitState is called from within r200EmitVbufPrim and r200FlushElts.
- */
-#define AOS_BUFSZ(nr)	((3 + ((nr / 2) * 3) + ((nr & 1) * 2)) * sizeof(int))
-#define VERT_AOS_BUFSZ	(5 * sizeof(int))
-#define ELTS_BUFSZ(nr)	(12 + nr * 2)
-#define VBUF_BUFSZ	(3 * sizeof(int))
-
-/* Ensure that a minimum amount of space is available in the command buffer.
- * This is used to ensure atomicity of state updates with the rendering requests
- * that rely on them.
- *
- * An alternative would be to implement a "soft lock" such that when the buffer
- * wraps at an inopportune time, we grab the lock, flush the current buffer,
- * and hang on to the lock until the critical section is finished and we flush
- * the buffer again and unlock.
- */
-static __inline void r200EnsureCmdBufSpace(r200ContextPtr rmesa, int bytes)
-{
-	if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ)
-		r200FlushCmdBuf(rmesa, __FUNCTION__);
-	assert(bytes <= R200_CMD_BUF_SZ);
-}
-
-/* Alloc space in the command buffer
- */
-static __inline char *r200AllocCmdBuf(r200ContextPtr rmesa,
-				      int bytes, const char *where)
-{
-	char *head;
-
-	if (rmesa->store.cmd_used + bytes > R200_CMD_BUF_SZ)
-		r200FlushCmdBuf(rmesa, where);
-
-	head = rmesa->store.cmd_buf + rmesa->store.cmd_used;
-	rmesa->store.cmd_used += bytes;
-	assert(rmesa->store.cmd_used <= R200_CMD_BUF_SZ);
-	return head;
-}
-
-#endif				/* __R200_IOCTL_H__ */
diff --git a/src/mesa/drivers/dri/r300/r200_reg.h b/src/mesa/drivers/dri/r300/r200_reg.h
deleted file mode 100644
index 1336e961ac6..00000000000
--- a/src/mesa/drivers/dri/r300/r200_reg.h
+++ /dev/null
@@ -1,1423 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/r200/r200_reg.h,v 1.2 2002/12/16 16:18:54 dawes Exp $ */
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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.
-*/
-
-#ifndef _R200_REG_H_
-#define _R200_REG_H_
-
-#define R200_PP_MISC                      0x1c14
-#define     R200_REF_ALPHA_MASK        0x000000ff
-#define     R200_ALPHA_TEST_FAIL       (0 << 8)
-#define     R200_ALPHA_TEST_LESS       (1 << 8)
-#define     R200_ALPHA_TEST_LEQUAL     (2 << 8)
-#define     R200_ALPHA_TEST_EQUAL      (3 << 8)
-#define     R200_ALPHA_TEST_GEQUAL     (4 << 8)
-#define     R200_ALPHA_TEST_GREATER    (5 << 8)
-#define     R200_ALPHA_TEST_NEQUAL     (6 << 8)
-#define     R200_ALPHA_TEST_PASS       (7 << 8)
-#define     R200_ALPHA_TEST_OP_MASK    (7 << 8)
-#define     R200_CHROMA_FUNC_FAIL      (0 << 16)
-#define     R200_CHROMA_FUNC_PASS      (1 << 16)
-#define     R200_CHROMA_FUNC_NEQUAL    (2 << 16)
-#define     R200_CHROMA_FUNC_EQUAL     (3 << 16)
-#define     R200_CHROMA_KEY_NEAREST    (0 << 18)
-#define     R200_CHROMA_KEY_ZERO       (1 << 18)
-#define     R200_RIGHT_HAND_CUBE_D3D   (0 << 24)
-#define     R200_RIGHT_HAND_CUBE_OGL   (1 << 24)
-#define R200_PP_FOG_COLOR                 0x1c18
-#define     R200_FOG_COLOR_MASK        0x00ffffff
-#define     R200_FOG_VERTEX            (0 << 24)
-#define     R200_FOG_TABLE             (1 << 24)
-#define     R200_FOG_USE_DEPTH         (0 << 25)
-#define     R200_FOG_USE_W             (1 << 25)
-#define     R200_FOG_USE_DIFFUSE_ALPHA (2 << 25)
-#define     R200_FOG_USE_SPEC_ALPHA    (3 << 25)
-#define     R200_FOG_USE_VTX_FOG       (4 << 25)
-#define     R200_FOG_USE_MASK          (7 << 25)
-#define R200_RE_SOLID_COLOR               0x1c1c
-#define R200_RB3D_BLENDCNTL               0x1c20
-#define     R200_COMB_FCN_MASK                    (7  << 12)
-#define     R200_COMB_FCN_ADD_CLAMP               (0  << 12)
-#define     R200_COMB_FCN_ADD_NOCLAMP             (1  << 12)
-#define     R200_COMB_FCN_SUB_CLAMP               (2  << 12)
-#define     R200_COMB_FCN_SUB_NOCLAMP             (3  << 12)
-#define     R200_COMB_FCN_MIN                     (4  << 12)
-#define     R200_COMB_FCN_MAX                     (5  << 12)
-#define     R200_COMB_FCN_RSUB_CLAMP              (6  << 12)
-#define     R200_COMB_FCN_RSUB_NOCLAMP            (7  << 12)
-#define       R200_BLEND_GL_ZERO                  (32)
-#define       R200_BLEND_GL_ONE                   (33)
-#define       R200_BLEND_GL_SRC_COLOR             (34)
-#define       R200_BLEND_GL_ONE_MINUS_SRC_COLOR   (35)
-#define       R200_BLEND_GL_DST_COLOR             (36)
-#define       R200_BLEND_GL_ONE_MINUS_DST_COLOR   (37)
-#define       R200_BLEND_GL_SRC_ALPHA             (38)
-#define       R200_BLEND_GL_ONE_MINUS_SRC_ALPHA   (39)
-#define       R200_BLEND_GL_DST_ALPHA             (40)
-#define       R200_BLEND_GL_ONE_MINUS_DST_ALPHA   (41)
-#define       R200_BLEND_GL_SRC_ALPHA_SATURATE    (42)	/* src factor only */
-#define       R200_BLEND_GL_CONST_COLOR           (43)
-#define       R200_BLEND_GL_ONE_MINUS_CONST_COLOR (44)
-#define       R200_BLEND_GL_CONST_ALPHA           (45)
-#define       R200_BLEND_GL_ONE_MINUS_CONST_ALPHA (46)
-#define       R200_BLEND_MASK                     (63)
-#define     R200_SRC_BLEND_SHIFT                  (16)
-#define     R200_DST_BLEND_SHIFT                  (24)
-#define R200_RB3D_DEPTHOFFSET             0x1c24
-#define R200_RB3D_DEPTHPITCH              0x1c28
-#define     R200_DEPTHPITCH_MASK         0x00001ff8
-#define     R200_DEPTH_ENDIAN_NO_SWAP    (0 << 18)
-#define     R200_DEPTH_ENDIAN_WORD_SWAP  (1 << 18)
-#define     R200_DEPTH_ENDIAN_DWORD_SWAP (2 << 18)
-#define R200_RB3D_ZSTENCILCNTL            0x1c2c
-#define     R200_DEPTH_FORMAT_MASK          (0xf << 0)
-#define     R200_DEPTH_FORMAT_16BIT_INT_Z   (0  <<  0)
-#define     R200_DEPTH_FORMAT_24BIT_INT_Z   (2  <<  0)
-#define     R200_DEPTH_FORMAT_24BIT_FLOAT_Z (3  <<  0)
-#define     R200_DEPTH_FORMAT_32BIT_INT_Z   (4  <<  0)
-#define     R200_DEPTH_FORMAT_32BIT_FLOAT_Z (5  <<  0)
-#define     R200_DEPTH_FORMAT_24BIT_FLOAT_W (9  <<  0)
-#define     R200_DEPTH_FORMAT_32BIT_FLOAT_W (11 <<  0)
-#define     R200_Z_TEST_NEVER               (0  <<  4)
-#define     R200_Z_TEST_LESS                (1  <<  4)
-#define     R200_Z_TEST_LEQUAL              (2  <<  4)
-#define     R200_Z_TEST_EQUAL               (3  <<  4)
-#define     R200_Z_TEST_GEQUAL              (4  <<  4)
-#define     R200_Z_TEST_GREATER             (5  <<  4)
-#define     R200_Z_TEST_NEQUAL              (6  <<  4)
-#define     R200_Z_TEST_ALWAYS              (7  <<  4)
-#define     R200_Z_TEST_MASK                (7  <<  4)
-#define     R200_STENCIL_TEST_NEVER         (0  << 12)
-#define     R200_STENCIL_TEST_LESS          (1  << 12)
-#define     R200_STENCIL_TEST_LEQUAL        (2  << 12)
-#define     R200_STENCIL_TEST_EQUAL         (3  << 12)
-#define     R200_STENCIL_TEST_GEQUAL        (4  << 12)
-#define     R200_STENCIL_TEST_GREATER       (5  << 12)
-#define     R200_STENCIL_TEST_NEQUAL        (6  << 12)
-#define     R200_STENCIL_TEST_ALWAYS        (7  << 12)
-#define     R200_STENCIL_TEST_MASK          (0x7 << 12)
-#define     R200_STENCIL_FAIL_KEEP          (0  << 16)
-#define     R200_STENCIL_FAIL_ZERO          (1  << 16)
-#define     R200_STENCIL_FAIL_REPLACE       (2  << 16)
-#define     R200_STENCIL_FAIL_INC           (3  << 16)
-#define     R200_STENCIL_FAIL_DEC           (4  << 16)
-#define     R200_STENCIL_FAIL_INVERT        (5  << 16)
-#define     R200_STENCIL_FAIL_INC_WRAP      (6  << 16)
-#define     R200_STENCIL_FAIL_DEC_WRAP      (7  << 16)
-#define     R200_STENCIL_FAIL_MASK          (0x7 << 16)
-#define     R200_STENCIL_ZPASS_KEEP         (0  << 20)
-#define     R200_STENCIL_ZPASS_ZERO         (1  << 20)
-#define     R200_STENCIL_ZPASS_REPLACE      (2  << 20)
-#define     R200_STENCIL_ZPASS_INC          (3  << 20)
-#define     R200_STENCIL_ZPASS_DEC          (4  << 20)
-#define     R200_STENCIL_ZPASS_INVERT       (5  << 20)
-#define     R200_STENCIL_ZPASS_INC_WRAP     (6  << 20)
-#define     R200_STENCIL_ZPASS_DEC_WRAP     (7  << 20)
-#define     R200_STENCIL_ZPASS_MASK         (0x7 << 20)
-#define     R200_STENCIL_ZFAIL_KEEP         (0  << 24)
-#define     R200_STENCIL_ZFAIL_ZERO         (1  << 24)
-#define     R200_STENCIL_ZFAIL_REPLACE      (2  << 24)
-#define     R200_STENCIL_ZFAIL_INC          (3  << 24)
-#define     R200_STENCIL_ZFAIL_DEC          (4  << 24)
-#define     R200_STENCIL_ZFAIL_INVERT       (5  << 24)
-#define     R200_STENCIL_ZFAIL_INC_WRAP     (6  << 24)
-#define     R200_STENCIL_ZFAIL_DEC_WRAP     (7  << 24)
-#define     R200_STENCIL_ZFAIL_MASK         (0x7 << 24)
-#define     R200_Z_WRITE_ENABLE             (1  << 30)
-/*gap*/
-#define R200_PP_CNTL                      0x1c38
-#define     R200_TEX_0_ENABLE                         0x00000010
-#define     R200_TEX_1_ENABLE                         0x00000020
-#define     R200_TEX_2_ENABLE                         0x00000040
-#define     R200_TEX_3_ENABLE                         0x00000080
-#define     R200_TEX_4_ENABLE                         0x00000100
-#define     R200_TEX_5_ENABLE                         0x00000200
-#define     R200_TEX_ENABLE_MASK                      0x000003f0
-#define     R200_FILTER_ROUND_MODE_MASK               0x00000400
-#define     R200_TEX_BLEND_7_ENABLE                   0x00000800
-#define     R200_TEX_BLEND_0_ENABLE                   0x00001000
-#define     R200_TEX_BLEND_1_ENABLE                   0x00002000
-#define     R200_TEX_BLEND_2_ENABLE                   0x00004000
-#define     R200_TEX_BLEND_3_ENABLE                   0x00008000
-#define     R200_TEX_BLEND_4_ENABLE                   0x00010000
-#define     R200_TEX_BLEND_5_ENABLE                   0x00020000
-#define     R200_TEX_BLEND_6_ENABLE                   0x00040000
-#define     R200_MULTI_PASS_ENABLE                    0x00080000
-#define     R200_SPECULAR_ENABLE                      0x00200000
-#define     R200_FOG_ENABLE                           0x00400000
-#define     R200_ALPHA_TEST_ENABLE                    0x00800000
-#define     R200_ANTI_ALIAS_NONE                       0x00000000
-#define     R200_ANTI_ALIAS_LINE                       0x01000000
-#define     R200_ANTI_ALIAS_POLY                       0x02000000
-#define     R200_ANTI_ALIAS_MASK                       0x03000000
-#define R200_RB3D_CNTL                    0x1c3c
-#define     R200_ALPHA_BLEND_ENABLE       (1  <<  0)
-#define     R200_PLANE_MASK_ENABLE        (1  <<  1)
-#define     R200_DITHER_ENABLE            (1  <<  2)
-#define     R200_ROUND_ENABLE             (1  <<  3)
-#define     R200_SCALE_DITHER_ENABLE      (1  <<  4)
-#define     R200_DITHER_INIT              (1  <<  5)
-#define     R200_ROP_ENABLE               (1  <<  6)
-#define     R200_STENCIL_ENABLE           (1  <<  7)
-#define     R200_Z_ENABLE                 (1  <<  8)
-#define     R200_DEPTH_XZ_OFFEST_ENABLE   (1  <<  9)
-#define     R200_COLOR_FORMAT_ARGB1555    (3  << 10)
-#define     R200_COLOR_FORMAT_RGB565      (4  << 10)
-#define     R200_COLOR_FORMAT_ARGB8888    (6  << 10)
-#define     R200_COLOR_FORMAT_RGB332      (7  << 10)
-#define     R200_COLOR_FORMAT_Y8          (8  << 10)
-#define     R200_COLOR_FORMAT_RGB8        (9  << 10)
-#define     R200_COLOR_FORMAT_YUV422_VYUY (11 << 10)
-#define     R200_COLOR_FORMAT_YUV422_YVYU (12 << 10)
-#define     R200_COLOR_FORMAT_aYUV444     (14 << 10)
-#define     R200_COLOR_FORMAT_ARGB4444    (15 << 10)
-#define     R200_CLRCMP_FLIP_ENABLE       (1  << 14)
-#define     R200_SEPARATE_ALPHA_ENABLE    (1  << 16)
-#define R200_RB3D_COLOROFFSET             0x1c40
-#define     R200_COLOROFFSET_MASK      0xfffffff0
-#define R200_RE_WIDTH_HEIGHT              0x1c44
-#define     R200_RE_WIDTH_SHIFT        0
-#define     R200_RE_HEIGHT_SHIFT       16
-#define R200_RB3D_COLORPITCH              0x1c48
-#define     R200_COLORPITCH_MASK         0x000001ff8
-#define     R200_COLOR_ENDIAN_NO_SWAP    (0 << 18)
-#define     R200_COLOR_ENDIAN_WORD_SWAP  (1 << 18)
-#define     R200_COLOR_ENDIAN_DWORD_SWAP (2 << 18)
-#define R200_SE_CNTL                      0x1c4c
-#define     R200_FFACE_CULL_CW          (0 <<  0)
-#define     R200_FFACE_CULL_CCW         (1 <<  0)
-#define     R200_FFACE_CULL_DIR_MASK    (1 <<  0)
-#define     R200_BFACE_CULL             (0 <<  1)
-#define     R200_BFACE_SOLID            (3 <<  1)
-#define     R200_FFACE_CULL             (0 <<  3)
-#define     R200_FFACE_SOLID            (3 <<  3)
-#define     R200_FFACE_CULL_MASK        (3 <<  3)
-#define     R200_FLAT_SHADE_VTX_0       (0 <<  6)
-#define     R200_FLAT_SHADE_VTX_1       (1 <<  6)
-#define     R200_FLAT_SHADE_VTX_2       (2 <<  6)
-#define     R200_FLAT_SHADE_VTX_LAST    (3 <<  6)
-#define     R200_DIFFUSE_SHADE_SOLID    (0 <<  8)
-#define     R200_DIFFUSE_SHADE_FLAT     (1 <<  8)
-#define     R200_DIFFUSE_SHADE_GOURAUD  (2 <<  8)
-#define     R200_DIFFUSE_SHADE_MASK     (3 <<  8)
-#define     R200_ALPHA_SHADE_SOLID      (0 << 10)
-#define     R200_ALPHA_SHADE_FLAT       (1 << 10)
-#define     R200_ALPHA_SHADE_GOURAUD    (2 << 10)
-#define     R200_ALPHA_SHADE_MASK       (3 << 10)
-#define     R200_SPECULAR_SHADE_SOLID   (0 << 12)
-#define     R200_SPECULAR_SHADE_FLAT    (1 << 12)
-#define     R200_SPECULAR_SHADE_GOURAUD (2 << 12)
-#define     R200_SPECULAR_SHADE_MASK    (3 << 12)
-#define     R200_FOG_SHADE_SOLID        (0 << 14)
-#define     R200_FOG_SHADE_FLAT         (1 << 14)
-#define     R200_FOG_SHADE_GOURAUD      (2 << 14)
-#define     R200_FOG_SHADE_MASK         (3 << 14)
-#define     R200_ZBIAS_ENABLE_POINT     (1 << 16)
-#define     R200_ZBIAS_ENABLE_LINE      (1 << 17)
-#define     R200_ZBIAS_ENABLE_TRI       (1 << 18)
-#define     R200_WIDELINE_ENABLE        (1 << 20)
-#define     R200_VTX_PIX_CENTER_D3D     (0 << 27)
-#define     R200_VTX_PIX_CENTER_OGL     (1 << 27)
-#define     R200_ROUND_MODE_TRUNC       (0 << 28)
-#define     R200_ROUND_MODE_ROUND       (1 << 28)
-#define     R200_ROUND_MODE_ROUND_EVEN  (2 << 28)
-#define     R200_ROUND_MODE_ROUND_ODD   (3 << 28)
-#define     R200_ROUND_PREC_16TH_PIX    (0 << 30)
-#define     R200_ROUND_PREC_8TH_PIX     (1 << 30)
-#define     R200_ROUND_PREC_4TH_PIX     (2 << 30)
-#define     R200_ROUND_PREC_HALF_PIX    (3 << 30)
-#define R200_RE_CNTL                      0x1c50
-#define     R200_STIPPLE_ENABLE                     0x1
-#define     R200_SCISSOR_ENABLE                     0x2
-#define     R200_PATTERN_ENABLE                     0x4
-#define     R200_PERSPECTIVE_ENABLE                 0x8
-#define     R200_POINT_SMOOTH                       0x20
-#define     R200_VTX_STQ0_D3D                       0x00010000
-#define     R200_VTX_STQ1_D3D                       0x00040000
-#define     R200_VTX_STQ2_D3D                       0x00100000
-#define     R200_VTX_STQ3_D3D                       0x00400000
-#define     R200_VTX_STQ4_D3D                       0x01000000
-#define     R200_VTX_STQ5_D3D                       0x04000000
-/* gap */
-#define R200_RE_STIPPLE_ADDR              0x1cc8
-#define R200_RE_STIPPLE_DATA              0x1ccc
-#define R200_RE_LINE_PATTERN              0x1cd0
-#define     R200_LINE_PATTERN_MASK             0x0000ffff
-#define     R200_LINE_REPEAT_COUNT_SHIFT       16
-#define     R200_LINE_PATTERN_START_SHIFT      24
-#define     R200_LINE_PATTERN_LITTLE_BIT_ORDER (0 << 28)
-#define     R200_LINE_PATTERN_BIG_BIT_ORDER    (1 << 28)
-#define     R200_LINE_PATTERN_AUTO_RESET       (1 << 29)
-#define R200_RE_LINE_STATE                0x1cd4
-#define     R200_LINE_CURRENT_PTR_SHIFT       0
-#define     R200_LINE_CURRENT_COUNT_SHIFT     8
-#define R200_RE_SCISSOR_TL_0              0x1cd8
-#define R200_RE_SCISSOR_BR_0              0x1cdc
-#define R200_RE_SCISSOR_TL_1              0x1ce0
-#define R200_RE_SCISSOR_BR_1              0x1ce4
-#define R200_RE_SCISSOR_TL_2              0x1ce8
-#define R200_RE_SCISSOR_BR_2              0x1cec
-/* gap */
-#define R200_RB3D_DEPTHXY_OFFSET          0x1d60
-#define     R200_DEPTHX_SHIFT  0
-#define     R200_DEPTHY_SHIFT  16
-/* gap */
-#define R200_RB3D_STENCILREFMASK          0x1d7c
-#define     R200_STENCIL_REF_SHIFT           0
-#define     R200_STENCIL_REF_MASK            (0xff << 0)
-#define     R200_STENCIL_MASK_SHIFT          16
-#define     R200_STENCIL_VALUE_MASK          (0xff << 16)
-#define     R200_STENCIL_WRITEMASK_SHIFT     24
-#define     R200_STENCIL_WRITE_MASK          (0xff << 24)
-#define R200_RB3D_ROPCNTL                 0x1d80
-#define     R200_ROP_MASK                    (15 << 8)
-#define     R200_ROP_CLEAR                   (0  << 8)
-#define     R200_ROP_NOR                     (1  << 8)
-#define     R200_ROP_AND_INVERTED            (2  << 8)
-#define     R200_ROP_COPY_INVERTED           (3  << 8)
-#define     R200_ROP_AND_REVERSE             (4  << 8)
-#define     R200_ROP_INVERT                  (5  << 8)
-#define     R200_ROP_XOR                     (6  << 8)
-#define     R200_ROP_NAND                    (7  << 8)
-#define     R200_ROP_AND                     (8  << 8)
-#define     R200_ROP_EQUIV                   (9  << 8)
-#define     R200_ROP_NOOP                    (10 << 8)
-#define     R200_ROP_OR_INVERTED             (11 << 8)
-#define     R200_ROP_COPY                    (12 << 8)
-#define     R200_ROP_OR_REVERSE              (13 << 8)
-#define     R200_ROP_OR                      (14 << 8)
-#define     R200_ROP_SET                     (15 << 8)
-#define R200_RB3D_PLANEMASK               0x1d84
-/* gap */
-#define R200_SE_VPORT_XSCALE              0x1d98
-#define R200_SE_VPORT_XOFFSET             0x1d9c
-#define R200_SE_VPORT_YSCALE              0x1da0
-#define R200_SE_VPORT_YOFFSET             0x1da4
-#define R200_SE_VPORT_ZSCALE              0x1da8
-#define R200_SE_VPORT_ZOFFSET             0x1dac
-#define R200_SE_ZBIAS_FACTOR              0x1db0
-#define R200_SE_ZBIAS_CONSTANT            0x1db4
-#define R200_SE_LINE_WIDTH                0x1db8
-#define	    R200_LINE_WIDTH_SHIFT                   0x00000000
-#define	    R200_MINPOINTSIZE_SHIFT                 0x00000010
-/* gap */
-#define R200_SE_VAP_CNTL                           0x2080
-#define     R200_VAP_TCL_ENABLE                       0x00000001
-#define     R200_VAP_SINGLE_BUF_STATE_ENABLE          0x00000010
-#define     R200_VAP_FORCE_W_TO_ONE                   0x00010000
-#define     R200_VAP_D3D_TEX_DEFAULT                  0x00020000
-#define     R200_VAP_VF_MAX_VTX_NUM__SHIFT            18
-#define     R200_VAP_DX_CLIP_SPACE_DEF                0x00400000
-#define R200_SE_VF_CNTL                           0x2084
-#define     R200_VF_PRIM_NONE                         0x00000000
-#define     R200_VF_PRIM_POINTS                       0x00000001
-#define     R200_VF_PRIM_LINES                        0x00000002
-#define     R200_VF_PRIM_LINE_STRIP                   0x00000003
-#define     R200_VF_PRIM_TRIANGLES                    0x00000004
-#define     R200_VF_PRIM_TRIANGLE_FAN                 0x00000005
-#define     R200_VF_PRIM_TRIANGLE_STRIP               0x00000006
-#define     R200_VF_PRIM_RECT_LIST                    0x00000008
-#define     R200_VF_PRIM_3VRT_POINTS                  0x00000009
-#define     R200_VF_PRIM_3VRT_LINES                   0x0000000a
-#define     R200_VF_PRIM_POINT_SPRITES                0x0000000b
-#define     R200_VF_PRIM_LINE_LOOP                    0x0000000c
-#define     R200_VF_PRIM_QUADS                        0x0000000d
-#define     R200_VF_PRIM_QUAD_STRIP                   0x0000000e
-#define     R200_VF_PRIM_POLYGON                      0x0000000f
-#define     R200_VF_PRIM_MASK                         0x0000000f
-#define     R200_VF_PRIM_WALK_IND                     0x00000010
-#define     R200_VF_PRIM_WALK_LIST                    0x00000020
-#define     R200_VF_PRIM_WALK_RING                    0x00000030
-#define     R200_VF_PRIM_WALK_MASK                    0x00000030
-#define     R200_VF_COLOR_ORDER_RGBA                  0x00000040
-#define     R200_VF_TCL_OUTPUT_VTX_ENABLE             0x00000200
-#define     R200_VF_INDEX_SZ_4                        0x00000800
-#define     R200_VF_VERTEX_NUMBER_MASK                0xffff0000
-#define     R200_VF_VERTEX_NUMBER_SHIFT               16
-#define R200_SE_VTX_FMT_0                 0x2088
-#define     R200_VTX_XY                     0	/* always have xy */
-#define     R200_VTX_Z0                     (1<<0)
-#define     R200_VTX_W0                     (1<<1)
-#define     R200_VTX_WEIGHT_COUNT_SHIFT     (2)
-#define     R200_VTX_PV_MATRIX_SEL          (1<<5)
-#define     R200_VTX_N0                     (1<<6)
-#define     R200_VTX_POINT_SIZE             (1<<7)
-#define     R200_VTX_DISCRETE_FOG           (1<<8)
-#define     R200_VTX_SHININESS_0            (1<<9)
-#define     R200_VTX_SHININESS_1            (1<<10)
-#define       R200_VTX_COLOR_NOT_PRESENT      0
-#define       R200_VTX_PK_RGBA          1
-#define       R200_VTX_FP_RGB           2
-#define       R200_VTX_FP_RGBA          3
-#define       R200_VTX_COLOR_MASK             3
-#define     R200_VTX_COLOR_0_SHIFT          11
-#define     R200_VTX_COLOR_1_SHIFT          13
-#define     R200_VTX_COLOR_2_SHIFT          15
-#define     R200_VTX_COLOR_3_SHIFT          17
-#define     R200_VTX_COLOR_4_SHIFT          19
-#define     R200_VTX_COLOR_5_SHIFT          21
-#define     R200_VTX_COLOR_6_SHIFT          23
-#define     R200_VTX_COLOR_7_SHIFT          25
-#define     R200_VTX_XY1                    (1<<28)
-#define     R200_VTX_Z1                     (1<<29)
-#define     R200_VTX_W1                     (1<<30)
-#define     R200_VTX_N1                     (1<<31)
-#define R200_SE_VTX_FMT_1                 0x208c
-#define     R200_VTX_TEX0_COMP_CNT_SHIFT        0
-#define     R200_VTX_TEX1_COMP_CNT_SHIFT        3
-#define     R200_VTX_TEX2_COMP_CNT_SHIFT        6
-#define     R200_VTX_TEX3_COMP_CNT_SHIFT        9
-#define     R200_VTX_TEX4_COMP_CNT_SHIFT        12
-#define     R200_VTX_TEX5_COMP_CNT_SHIFT        15
-#define R200_SE_TCL_OUTPUT_VTX_FMT_0      0x2090
-#define R200_SE_TCL_OUTPUT_VTX_FMT_1      0x2094
-/* gap */
-#define R200_SE_VTE_CNTL                  0x20b0
-#define     R200_VPORT_X_SCALE_ENA                0x00000001
-#define     R200_VPORT_X_OFFSET_ENA               0x00000002
-#define     R200_VPORT_Y_SCALE_ENA                0x00000004
-#define     R200_VPORT_Y_OFFSET_ENA               0x00000008
-#define     R200_VPORT_Z_SCALE_ENA                0x00000010
-#define     R200_VPORT_Z_OFFSET_ENA               0x00000020
-#define     R200_VTX_XY_FMT                       0x00000100
-#define     R200_VTX_Z_FMT                        0x00000200
-#define     R200_VTX_W0_FMT                       0x00000400
-#define     R200_VTX_W0_NORMALIZE                 0x00000800
-#define     R200_VTX_ST_DENORMALIZED              0x00001000
-/* gap */
-#define R200_SE_VTX_NUM_ARRAYS            0x20c0
-#define R200_SE_VTX_AOS_ATTR01            0x20c4
-#define R200_SE_VTX_AOS_ADDR0             0x20c8
-#define R200_SE_VTX_AOS_ADDR1             0x20cc
-#define R200_SE_VTX_AOS_ATTR23            0x20d0
-#define R200_SE_VTX_AOS_ADDR2             0x20d4
-#define R200_SE_VTX_AOS_ADDR3             0x20d8
-#define R200_SE_VTX_AOS_ATTR45            0x20dc
-#define R200_SE_VTX_AOS_ADDR4             0x20e0
-#define R200_SE_VTX_AOS_ADDR5             0x20e4
-#define R200_SE_VTX_AOS_ATTR67            0x20e8
-#define R200_SE_VTX_AOS_ADDR6             0x20ec
-#define R200_SE_VTX_AOS_ADDR7             0x20f0
-#define R200_SE_VTX_AOS_ATTR89            0x20f4
-#define R200_SE_VTX_AOS_ADDR8             0x20f8
-#define R200_SE_VTX_AOS_ADDR9             0x20fc
-#define R200_SE_VTX_AOS_ATTR1011          0x2100
-#define R200_SE_VTX_AOS_ADDR10            0x2104
-#define R200_SE_VTX_AOS_ADDR11            0x2108
-#define R200_SE_VF_MAX_VTX_INDX           0x210c
-#define R200_SE_VF_MIN_VTX_INDX           0x2110
-/* gap */
-#define R200_SE_VAP_CNTL_STATUS           0x2140
-#define     R200_VC_NO_SWAP                  (0 << 0)
-#define     R200_VC_16BIT_SWAP               (1 << 0)
-#define     R200_VC_32BIT_SWAP               (2 << 0)
-/* gap */
-#define R200_SE_VTX_STATE_CNTL                     0x2180
-#define     R200_VSC_COLOR_0_ASSEMBLY_CNTL_SHIFT    0x00000000
-#define     R200_VSC_COLOR_1_ASSEMBLY_CNTL_SHIFT    0x00000002
-#define     R200_VSC_COLOR_2_ASSEMBLY_CNTL_SHIFT    0x00000004
-#define     R200_VSC_COLOR_3_ASSEMBLY_CNTL_SHIFT    0x00000006
-#define     R200_VSC_COLOR_4_ASSEMBLY_CNTL_SHIFT    0x00000008
-#define     R200_VSC_COLOR_5_ASSEMBLY_CNTL_SHIFT    0x0000000a
-#define     R200_VSC_COLOR_6_ASSEMBLY_CNTL_SHIFT    0x0000000c
-#define     R200_VSC_COLOR_7_ASSEMBLY_CNTL_SHIFT    0x0000000e
-#define     R200_VSC_UPDATE_USER_COLOR_0_ENABLE    0x00010000
-#define     R200_VSC_UPDATE_USER_COLOR_1_ENABLE    0x00020000
-/* gap */
-#define R200_SE_TCL_VECTOR_INDX_REG                0x2200
-#define R200_SE_TCL_VECTOR_DATA_REG                0x2204
-#define R200_SE_TCL_SCALAR_INDX_REG                0x2208
-#define R200_SE_TCL_SCALAR_DATA_REG                0x220c
-/* gap */
-#define R200_SE_TCL_MATRIX_SEL_0                   0x2230
-#define     R200_MODELVIEW_0_SHIFT           (0)
-#define     R200_MODELVIEW_1_SHIFT           (8)
-#define     R200_MODELVIEW_2_SHIFT           (16)
-#define     R200_MODELVIEW_3_SHIFT           (24)
-#define R200_SE_TCL_MATRIX_SEL_1                   0x2234
-#define     R200_IT_MODELVIEW_0_SHIFT        (0)
-#define     R200_IT_MODELVIEW_1_SHIFT        (8)
-#define     R200_IT_MODELVIEW_2_SHIFT        (16)
-#define     R200_IT_MODELVIEW_3_SHIFT        (24)
-#define R200_SE_TCL_MATRIX_SEL_2                   0x2238
-#define     R200_MODELPROJECT_0_SHIFT         (0)
-#define     R200_MODELPROJECT_1_SHIFT         (8)
-#define     R200_MODELPROJECT_2_SHIFT         (16)
-#define     R200_MODELPROJECT_3_SHIFT         (24)
-#define R200_SE_TCL_MATRIX_SEL_3                   0x223c
-#define     R200_TEXMAT_0_SHIFT    0
-#define     R200_TEXMAT_1_SHIFT    8
-#define     R200_TEXMAT_2_SHIFT    16
-#define     R200_TEXMAT_3_SHIFT    24
-#define R200_SE_TCL_MATRIX_SEL_4                   0x2240
-#define     R200_TEXMAT_4_SHIFT    0
-#define     R200_TEXMAT_5_SHIFT    8
-/* gap */
-#define R200_SE_TCL_OUTPUT_VTX_COMP_SEL     0x2250
-#define     R200_OUTPUT_XYZW                    (1<<0)
-#define     R200_OUTPUT_COLOR_0                 (1<<8)
-#define     R200_OUTPUT_COLOR_1                 (1<<9)
-#define     R200_OUTPUT_TEX_0                   (1<<16)
-#define     R200_OUTPUT_TEX_1                   (1<<17)
-#define     R200_OUTPUT_TEX_2                   (1<<18)
-#define     R200_OUTPUT_TEX_3                   (1<<19)
-#define     R200_OUTPUT_TEX_4                   (1<<20)
-#define     R200_OUTPUT_TEX_5                   (1<<21)
-#define     R200_OUTPUT_TEX_MASK                (0x3f<<16)
-#define     R200_OUTPUT_DISCRETE_FOG            (1<<24)
-#define     R200_OUTPUT_PT_SIZE                 (1<<25)
-#define     R200_FORCE_INORDER_PROC             (1<<31)
-#define R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_0  0x2254
-#define	    R200_VERTEX_POSITION_ADDR__SHIFT     0x00000000
-#define R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_1  0x2258
-#define	    R200_VTX_COLOR_0_ADDR__SHIFT         0x00000000
-#define	    R200_VTX_COLOR_1_ADDR__SHIFT         0x00000008
-#define R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_2  0x225c
-#define	    R200_VTX_TEX_0_ADDR__SHIFT           0x00000000
-#define	    R200_VTX_TEX_1_ADDR__SHIFT           0x00000008
-#define	    R200_VTX_TEX_2_ADDR__SHIFT           0x00000010
-#define	    R200_VTX_TEX_3_ADDR__SHIFT           0x00000018
-#define R200_SE_TCL_INPUT_VTX_VECTOR_ADDR_3  0x2260
-#define	    R200_VTX_TEX_4_ADDR__SHIFT           0x00000000
-#define	    R200_VTX_TEX_5_ADDR__SHIFT           0x00000008
-
-/* gap */
-#define R200_SE_TCL_LIGHT_MODEL_CTL_0       0x2268
-#define     R200_LIGHTING_ENABLE                (1<<0)
-#define     R200_LIGHT_IN_MODELSPACE            (1<<1)
-#define     R200_LOCAL_VIEWER                   (1<<2)
-#define     R200_NORMALIZE_NORMALS              (1<<3)
-#define     R200_RESCALE_NORMALS                (1<<4)
-#define     R200_SPECULAR_LIGHTS                (1<<5)
-#define     R200_DIFFUSE_SPECULAR_COMBINE       (1<<6)
-#define     R200_LIGHT_ALPHA                    (1<<7)
-#define     R200_LOCAL_LIGHT_VEC_GL             (1<<8)
-#define     R200_LIGHT_NO_NORMAL_AMBIENT_ONLY   (1<<9)
-#define     R200_LIGHT_TWOSIDE                  (1<<10)
-#define     R200_FRONT_SHININESS_SOURCE_SHIFT       (0xb)
-#define     R200_BACK_SHININESS_SOURCE_SHIFT        (0xd)
-#define       R200_LM0_SOURCE_MATERIAL_0           (0)
-#define       R200_LM0_SOURCE_MATERIAL_1           (1)
-#define       R200_LM0_SOURCE_VERTEX_SHININESS_0   (2)
-#define       R200_LM0_SOURCE_VERTEX_SHININESS_1   (3)
-#define R200_SE_TCL_LIGHT_MODEL_CTL_1       0x226c
-#define       R200_LM1_SOURCE_LIGHT_PREMULT        (0)
-#define       R200_LM1_SOURCE_MATERIAL_0           (1)
-#define       R200_LM1_SOURCE_VERTEX_COLOR_0       (2)
-#define       R200_LM1_SOURCE_VERTEX_COLOR_1       (3)
-#define       R200_LM1_SOURCE_VERTEX_COLOR_2       (4)
-#define       R200_LM1_SOURCE_VERTEX_COLOR_3       (5)
-#define       R200_LM1_SOURCE_VERTEX_COLOR_4       (6)
-#define       R200_LM1_SOURCE_VERTEX_COLOR_5       (7)
-#define       R200_LM1_SOURCE_VERTEX_COLOR_6       (8)
-#define       R200_LM1_SOURCE_VERTEX_COLOR_7       (9)
-#define       R200_LM1_SOURCE_MATERIAL_1           (0xf)
-#define     R200_FRONT_EMISSIVE_SOURCE_SHIFT        (0)
-#define     R200_FRONT_AMBIENT_SOURCE_SHIFT         (4)
-#define     R200_FRONT_DIFFUSE_SOURCE_SHIFT         (8)
-#define     R200_FRONT_SPECULAR_SOURCE_SHIFT        (12)
-#define     R200_BACK_EMISSIVE_SOURCE_SHIFT         (16)
-#define     R200_BACK_AMBIENT_SOURCE_SHIFT          (20)
-#define     R200_BACK_DIFFUSE_SOURCE_SHIFT          (24)
-#define     R200_BACK_SPECULAR_SOURCE_SHIFT         (28)
-#define R200_SE_TCL_PER_LIGHT_CTL_0       0x2270
-#define     R200_LIGHT_0_ENABLE                    (1<<0)
-#define     R200_LIGHT_0_ENABLE_AMBIENT            (1<<1)
-#define     R200_LIGHT_0_ENABLE_SPECULAR           (1<<2)
-#define     R200_LIGHT_0_IS_LOCAL                  (1<<3)
-#define     R200_LIGHT_0_IS_SPOT                   (1<<4)
-#define     R200_LIGHT_0_DUAL_CONE                 (1<<5)
-#define     R200_LIGHT_0_ENABLE_RANGE_ATTEN        (1<<6)
-#define     R200_LIGHT_0_CONSTANT_RANGE_ATTEN      (1<<7)
-#define     R200_LIGHT_1_ENABLE                    (1<<16)
-#define     R200_LIGHT_1_ENABLE_AMBIENT            (1<<17)
-#define     R200_LIGHT_1_ENABLE_SPECULAR           (1<<18)
-#define     R200_LIGHT_1_IS_LOCAL                  (1<<19)
-#define     R200_LIGHT_1_IS_SPOT                   (1<<20)
-#define     R200_LIGHT_1_DUAL_CONE                 (1<<21)
-#define     R200_LIGHT_1_ENABLE_RANGE_ATTEN        (1<<22)
-#define     R200_LIGHT_1_CONSTANT_RANGE_ATTEN      (1<<23)
-#define     R200_LIGHT_0_SHIFT                   (0)
-#define     R200_LIGHT_1_SHIFT                   (16)
-#define R200_SE_TCL_PER_LIGHT_CTL_1       0x2274
-#define     R200_LIGHT_2_SHIFT                   (0)
-#define     R200_LIGHT_3_SHIFT                   (16)
-#define R200_SE_TCL_PER_LIGHT_CTL_2       0x2278
-#define     R200_LIGHT_4_SHIFT                   (0)
-#define     R200_LIGHT_5_SHIFT                   (16)
-#define R200_SE_TCL_PER_LIGHT_CTL_3       0x227c
-#define     R200_LIGHT_6_SHIFT                   (0)
-#define     R200_LIGHT_7_SHIFT                   (16)
-/* gap */
-#define R200_SE_TCL_TEX_PROC_CTL_2        0x22a8
-#define     R200_TEXGEN_0_COMP_MASK_SHIFT        (0)
-#define     R200_TEXGEN_1_COMP_MASK_SHIFT        (4)
-#define     R200_TEXGEN_2_COMP_MASK_SHIFT        (8)
-#define     R200_TEXGEN_3_COMP_MASK_SHIFT        (12)
-#define     R200_TEXGEN_4_COMP_MASK_SHIFT        (16)
-#define     R200_TEXGEN_5_COMP_MASK_SHIFT        (20)
-#define R200_SE_TCL_TEX_PROC_CTL_3        0x22ac
-#define     R200_TEXGEN_0_INPUT_TEX_SHIFT        (0)
-#define     R200_TEXGEN_1_INPUT_TEX_SHIFT        (4)
-#define     R200_TEXGEN_2_INPUT_TEX_SHIFT        (8)
-#define     R200_TEXGEN_3_INPUT_TEX_SHIFT        (12)
-#define     R200_TEXGEN_4_INPUT_TEX_SHIFT        (16)
-#define     R200_TEXGEN_5_INPUT_TEX_SHIFT        (20)
-#define R200_SE_TCL_TEX_PROC_CTL_0        0x22b0
-#define     R200_TEXGEN_TEXMAT_0_ENABLE         (1<<0)
-#define     R200_TEXGEN_TEXMAT_1_ENABLE         (1<<1)
-#define     R200_TEXGEN_TEXMAT_2_ENABLE         (1<<2)
-#define     R200_TEXGEN_TEXMAT_3_ENABLE         (1<<3)
-#define     R200_TEXGEN_TEXMAT_4_ENABLE         (1<<4)
-#define     R200_TEXGEN_TEXMAT_5_ENABLE         (1<<5)
-#define     R200_TEXMAT_0_ENABLE                (1<<8)
-#define     R200_TEXMAT_1_ENABLE                (1<<9)
-#define     R200_TEXMAT_2_ENABLE                (1<<10)
-#define     R200_TEXMAT_3_ENABLE                (1<<11)
-#define     R200_TEXMAT_4_ENABLE                (1<<12)
-#define     R200_TEXMAT_5_ENABLE                (1<<13)
-#define     R200_TEXGEN_FORCE_W_TO_ONE          (1<<16)
-#define R200_SE_TCL_TEX_PROC_CTL_1        0x22b4
-#define       R200_TEXGEN_INPUT_MASK           (0xf)
-#define       R200_TEXGEN_INPUT_TEXCOORD_0     (0)
-#define       R200_TEXGEN_INPUT_TEXCOORD_1     (1)
-#define       R200_TEXGEN_INPUT_TEXCOORD_2     (2)
-#define       R200_TEXGEN_INPUT_TEXCOORD_3     (3)
-#define       R200_TEXGEN_INPUT_TEXCOORD_4     (4)
-#define       R200_TEXGEN_INPUT_TEXCOORD_5     (5)
-#define       R200_TEXGEN_INPUT_OBJ            (8)
-#define       R200_TEXGEN_INPUT_EYE            (9)
-#define       R200_TEXGEN_INPUT_EYE_NORMAL     (0xa)
-#define       R200_TEXGEN_INPUT_EYE_REFLECT    (0xb)
-#define       R200_TEXGEN_INPUT_SPHERE         (0xd)
-#define     R200_TEXGEN_0_INPUT_SHIFT        (0)
-#define     R200_TEXGEN_1_INPUT_SHIFT        (4)
-#define     R200_TEXGEN_2_INPUT_SHIFT        (8)
-#define     R200_TEXGEN_3_INPUT_SHIFT        (12)
-#define     R200_TEXGEN_4_INPUT_SHIFT        (16)
-#define     R200_TEXGEN_5_INPUT_SHIFT        (20)
-#define R200_SE_TC_TEX_CYL_WRAP_CTL       0x22b8
-/* gap */
-#define R200_SE_TCL_UCP_VERT_BLEND_CTL    0x22c0
-#define     R200_UCP_IN_CLIP_SPACE              (1<<0)
-#define     R200_UCP_IN_MODEL_SPACE             (1<<1)
-#define     R200_UCP_ENABLE_0                   (1<<2)
-#define     R200_UCP_ENABLE_1                   (1<<3)
-#define     R200_UCP_ENABLE_2                   (1<<4)
-#define     R200_UCP_ENABLE_3                   (1<<5)
-#define     R200_UCP_ENABLE_4                   (1<<6)
-#define     R200_UCP_ENABLE_5                   (1<<7)
-#define     R200_TCL_FOG_MASK                   (3<<8)
-#define     R200_TCL_FOG_DISABLE                (0<<8)
-#define     R200_TCL_FOG_EXP                    (1<<8)
-#define     R200_TCL_FOG_EXP2                   (2<<8)
-#define     R200_TCL_FOG_LINEAR                 (3<<8)
-#define     R200_RNG_BASED_FOG                  (1<<10)
-#define     R200_CLIP_DISABLE                   (1<<11)
-#define     R200_CULL_FRONT_IS_CW               (0<<28)
-#define     R200_CULL_FRONT_IS_CCW              (1<<28)
-#define     R200_CULL_FRONT                     (1<<29)
-#define     R200_CULL_BACK                      (1<<30)
-#define R200_SE_TCL_POINT_SPRITE_CNTL     0x22c4
-/* gap */
-#define R200_SE_VTX_ST_POS_0_X_4                   0x2300
-#define R200_SE_VTX_ST_POS_0_Y_4                   0x2304
-#define R200_SE_VTX_ST_POS_0_Z_4                   0x2308
-#define R200_SE_VTX_ST_POS_0_W_4                   0x230c
-#define R200_SE_VTX_ST_NORM_0_X                    0x2310
-#define R200_SE_VTX_ST_NORM_0_Y                    0x2314
-#define R200_SE_VTX_ST_NORM_0_Z                    0x2318
-#define R200_SE_VTX_ST_PVMS                        0x231c
-#define R200_SE_VTX_ST_CLR_0_R                     0x2320
-#define R200_SE_VTX_ST_CLR_0_G                     0x2324
-#define R200_SE_VTX_ST_CLR_0_B                     0x2328
-#define R200_SE_VTX_ST_CLR_0_A                     0x232c
-#define R200_SE_VTX_ST_CLR_1_R                     0x2330
-#define R200_SE_VTX_ST_CLR_1_G                     0x2334
-#define R200_SE_VTX_ST_CLR_1_B                     0x2338
-#define R200_SE_VTX_ST_CLR_1_A                     0x233c
-#define R200_SE_VTX_ST_CLR_2_R                     0x2340
-#define R200_SE_VTX_ST_CLR_2_G                     0x2344
-#define R200_SE_VTX_ST_CLR_2_B                     0x2348
-#define R200_SE_VTX_ST_CLR_2_A                     0x234c
-#define R200_SE_VTX_ST_CLR_3_R                     0x2350
-#define R200_SE_VTX_ST_CLR_3_G                     0x2354
-#define R200_SE_VTX_ST_CLR_3_B                     0x2358
-#define R200_SE_VTX_ST_CLR_3_A                     0x235c
-#define R200_SE_VTX_ST_CLR_4_R                     0x2360
-#define R200_SE_VTX_ST_CLR_4_G                     0x2364
-#define R200_SE_VTX_ST_CLR_4_B                     0x2368
-#define R200_SE_VTX_ST_CLR_4_A                     0x236c
-#define R200_SE_VTX_ST_CLR_5_R                     0x2370
-#define R200_SE_VTX_ST_CLR_5_G                     0x2374
-#define R200_SE_VTX_ST_CLR_5_B                     0x2378
-#define R200_SE_VTX_ST_CLR_5_A                     0x237c
-#define R200_SE_VTX_ST_CLR_6_R                     0x2380
-#define R200_SE_VTX_ST_CLR_6_G                     0x2384
-#define R200_SE_VTX_ST_CLR_6_B                     0x2388
-#define R200_SE_VTX_ST_CLR_6_A                     0x238c
-#define R200_SE_VTX_ST_CLR_7_R                     0x2390
-#define R200_SE_VTX_ST_CLR_7_G                     0x2394
-#define R200_SE_VTX_ST_CLR_7_B                     0x2398
-#define R200_SE_VTX_ST_CLR_7_A                     0x239c
-#define R200_SE_VTX_ST_TEX_0_S                     0x23a0
-#define R200_SE_VTX_ST_TEX_0_T                     0x23a4
-#define R200_SE_VTX_ST_TEX_0_R                     0x23a8
-#define R200_SE_VTX_ST_TEX_0_Q                     0x23ac
-#define R200_SE_VTX_ST_TEX_1_S                     0x23b0
-#define R200_SE_VTX_ST_TEX_1_T                     0x23b4
-#define R200_SE_VTX_ST_TEX_1_R                     0x23b8
-#define R200_SE_VTX_ST_TEX_1_Q                     0x23bc
-#define R200_SE_VTX_ST_TEX_2_S                     0x23c0
-#define R200_SE_VTX_ST_TEX_2_T                     0x23c4
-#define R200_SE_VTX_ST_TEX_2_R                     0x23c8
-#define R200_SE_VTX_ST_TEX_2_Q                     0x23cc
-#define R200_SE_VTX_ST_TEX_3_S                     0x23d0
-#define R200_SE_VTX_ST_TEX_3_T                     0x23d4
-#define R200_SE_VTX_ST_TEX_3_R                     0x23d8
-#define R200_SE_VTX_ST_TEX_3_Q                     0x23dc
-#define R200_SE_VTX_ST_TEX_4_S                     0x23e0
-#define R200_SE_VTX_ST_TEX_4_T                     0x23e4
-#define R200_SE_VTX_ST_TEX_4_R                     0x23e8
-#define R200_SE_VTX_ST_TEX_4_Q                     0x23ec
-#define R200_SE_VTX_ST_TEX_5_S                     0x23f0
-#define R200_SE_VTX_ST_TEX_5_T                     0x23f4
-#define R200_SE_VTX_ST_TEX_5_R                     0x23f8
-#define R200_SE_VTX_ST_TEX_5_Q                     0x23fc
-#define R200_SE_VTX_ST_PNT_SPRT_SZ                 0x2400
-#define R200_SE_VTX_ST_DISC_FOG                    0x2404
-#define R200_SE_VTX_ST_SHININESS_0                 0x2408
-#define R200_SE_VTX_ST_SHININESS_1                 0x240c
-#define R200_SE_VTX_ST_BLND_WT_0                   0x2410
-#define R200_SE_VTX_ST_BLND_WT_1                   0x2414
-#define R200_SE_VTX_ST_BLND_WT_2                   0x2418
-#define R200_SE_VTX_ST_BLND_WT_3                   0x241c
-#define R200_SE_VTX_ST_POS_1_X                     0x2420
-#define R200_SE_VTX_ST_POS_1_Y                     0x2424
-#define R200_SE_VTX_ST_POS_1_Z                     0x2428
-#define R200_SE_VTX_ST_POS_1_W                     0x242c
-#define R200_SE_VTX_ST_NORM_1_X                    0x2430
-#define R200_SE_VTX_ST_NORM_1_Y                    0x2434
-#define R200_SE_VTX_ST_NORM_1_Z                    0x2438
-#define R200_SE_VTX_ST_USR_CLR_0_R                 0x2440
-#define R200_SE_VTX_ST_USR_CLR_0_G                 0x2444
-#define R200_SE_VTX_ST_USR_CLR_0_B                 0x2448
-#define R200_SE_VTX_ST_USR_CLR_0_A                 0x244c
-#define R200_SE_VTX_ST_USR_CLR_1_R                 0x2450
-#define R200_SE_VTX_ST_USR_CLR_1_G                 0x2454
-#define R200_SE_VTX_ST_USR_CLR_1_B                 0x2458
-#define R200_SE_VTX_ST_USR_CLR_1_A                 0x245c
-#define R200_SE_VTX_ST_CLR_0_PKD                   0x2460
-#define R200_SE_VTX_ST_CLR_1_PKD                   0x2464
-#define R200_SE_VTX_ST_CLR_2_PKD                   0x2468
-#define R200_SE_VTX_ST_CLR_3_PKD                   0x246c
-#define R200_SE_VTX_ST_CLR_4_PKD                   0x2470
-#define R200_SE_VTX_ST_CLR_5_PKD                   0x2474
-#define R200_SE_VTX_ST_CLR_6_PKD                   0x2478
-#define R200_SE_VTX_ST_CLR_7_PKD                   0x247c
-#define R200_SE_VTX_ST_POS_0_X_2                   0x2480
-#define R200_SE_VTX_ST_POS_0_Y_2                   0x2484
-#define R200_SE_VTX_ST_PAR_CLR_LD                  0x2488
-#define R200_SE_VTX_ST_USR_CLR_PKD                 0x248c
-#define R200_SE_VTX_ST_POS_0_X_3                   0x2490
-#define R200_SE_VTX_ST_POS_0_Y_3                   0x2494
-#define R200_SE_VTX_ST_POS_0_Z_3                   0x2498
-#define R200_SE_VTX_ST_END_OF_PKT                  0x249c
-/* gap */
-#define R200_RE_POINTSIZE                          0x2648
-#define     R200_POINTSIZE_SHIFT                       0
-#define     R200_MAXPOINTSIZE_SHIFT                    16
-/* gap */
-#define R200_RE_TOP_LEFT                  0x26c0
-#define     R200_RE_LEFT_SHIFT         0
-#define     R200_RE_TOP_SHIFT          16
-#define R200_RE_MISC                      0x26c4
-#define     R200_STIPPLE_COORD_MASK           0x1f
-#define     R200_STIPPLE_X_OFFSET_SHIFT       0
-#define     R200_STIPPLE_X_OFFSET_MASK        (0x1f << 0)
-#define     R200_STIPPLE_Y_OFFSET_SHIFT       8
-#define     R200_STIPPLE_Y_OFFSET_MASK        (0x1f << 8)
-#define     R200_STIPPLE_LITTLE_BIT_ORDER     (0 << 16)
-#define     R200_STIPPLE_BIG_BIT_ORDER        (1 << 16)
-/* gap */
-#define R200_RE_AUX_SCISSOR_CNTL                   0x26f0
-#define     R200_EXCLUSIVE_SCISSOR_0      0x01000000
-#define     R200_EXCLUSIVE_SCISSOR_1      0x02000000
-#define     R200_EXCLUSIVE_SCISSOR_2      0x04000000
-#define     R200_SCISSOR_ENABLE_0         0x10000000
-#define     R200_SCISSOR_ENABLE_1         0x20000000
-#define     R200_SCISSOR_ENABLE_2         0x40000000
-/* gap */
-#define R200_PP_TXFILTER_0                0x2c00
-#define     R200_MAG_FILTER_NEAREST                   (0  <<  0)
-#define     R200_MAG_FILTER_LINEAR                    (1  <<  0)
-#define     R200_MAG_FILTER_MASK                      (1  <<  0)
-#define     R200_MIN_FILTER_NEAREST                   (0  <<  1)
-#define     R200_MIN_FILTER_LINEAR                    (1  <<  1)
-#define     R200_MIN_FILTER_NEAREST_MIP_NEAREST       (2  <<  1)
-#define     R200_MIN_FILTER_NEAREST_MIP_LINEAR        (3  <<  1)
-#define     R200_MIN_FILTER_LINEAR_MIP_NEAREST        (6  <<  1)
-#define     R200_MIN_FILTER_LINEAR_MIP_LINEAR         (7  <<  1)
-#define     R200_MIN_FILTER_ANISO_NEAREST             (8  <<  1)
-#define     R200_MIN_FILTER_ANISO_LINEAR              (9  <<  1)
-#define     R200_MIN_FILTER_ANISO_NEAREST_MIP_NEAREST (10 <<  1)
-#define     R200_MIN_FILTER_ANISO_NEAREST_MIP_LINEAR  (11 <<  1)
-#define     R200_MIN_FILTER_MASK                      (15 <<  1)
-#define     R200_MAX_ANISO_1_TO_1                     (0  <<  5)
-#define     R200_MAX_ANISO_2_TO_1                     (1  <<  5)
-#define     R200_MAX_ANISO_4_TO_1                     (2  <<  5)
-#define     R200_MAX_ANISO_8_TO_1                     (3  <<  5)
-#define     R200_MAX_ANISO_16_TO_1                    (4  <<  5)
-#define     R200_MAX_ANISO_MASK                       (7  <<  5)
-#define     R200_MAX_MIP_LEVEL_MASK                   (0x0f << 16)
-#define     R200_MAX_MIP_LEVEL_SHIFT                  16
-#define     R200_YUV_TO_RGB                           (1  << 20)
-#define     R200_YUV_TEMPERATURE_COOL                 (0  << 21)
-#define     R200_YUV_TEMPERATURE_HOT                  (1  << 21)
-#define     R200_YUV_TEMPERATURE_MASK                 (1  << 21)
-#define     R200_WRAPEN_S                             (1  << 22)
-#define     R200_CLAMP_S_WRAP                         (0  << 23)
-#define     R200_CLAMP_S_MIRROR                       (1  << 23)
-#define     R200_CLAMP_S_CLAMP_LAST                   (2  << 23)
-#define     R200_CLAMP_S_MIRROR_CLAMP_LAST            (3  << 23)
-#define     R200_CLAMP_S_CLAMP_BORDER                 (4  << 23)
-#define     R200_CLAMP_S_MIRROR_CLAMP_BORDER          (5  << 23)
-#define     R200_CLAMP_S_CLAMP_GL                     (6  << 23)
-#define     R200_CLAMP_S_MIRROR_CLAMP_GL              (7  << 23)
-#define     R200_CLAMP_S_MASK                         (7  << 23)
-#define     R200_WRAPEN_T                             (1  << 26)
-#define     R200_CLAMP_T_WRAP                         (0  << 27)
-#define     R200_CLAMP_T_MIRROR                       (1  << 27)
-#define     R200_CLAMP_T_CLAMP_LAST                   (2  << 27)
-#define     R200_CLAMP_T_MIRROR_CLAMP_LAST            (3  << 27)
-#define     R200_CLAMP_T_CLAMP_BORDER                 (4  << 27)
-#define     R200_CLAMP_T_MIRROR_CLAMP_BORDER          (5  << 27)
-#define     R200_CLAMP_T_CLAMP_GL                     (6  << 27)
-#define     R200_CLAMP_T_MIRROR_CLAMP_GL              (7  << 27)
-#define     R200_CLAMP_T_MASK                         (7  << 27)
-#define     R200_KILL_LT_ZERO                         (1  << 30)
-#define     R200_BORDER_MODE_OGL                      (0  << 31)
-#define     R200_BORDER_MODE_D3D                      (1  << 31)
-#define R200_PP_TXFORMAT_0                0x2c04
-#define     R200_TXFORMAT_I8                 (0  <<  0)
-#define     R200_TXFORMAT_AI88               (1  <<  0)
-#define     R200_TXFORMAT_RGB332             (2  <<  0)
-#define     R200_TXFORMAT_ARGB1555           (3  <<  0)
-#define     R200_TXFORMAT_RGB565             (4  <<  0)
-#define     R200_TXFORMAT_ARGB4444           (5  <<  0)
-#define     R200_TXFORMAT_ARGB8888           (6  <<  0)
-#define     R200_TXFORMAT_RGBA8888           (7  <<  0)
-#define     R200_TXFORMAT_Y8                 (8  <<  0)
-#define     R200_TXFORMAT_AVYU4444           (9  <<  0)
-#define     R200_TXFORMAT_VYUY422            (10  <<  0)
-#define     R200_TXFORMAT_YVYU422            (11  <<  0)
-#define     R200_TXFORMAT_DXT1               (12  <<  0)
-#define     R200_TXFORMAT_DXT23              (14  <<  0)
-#define     R200_TXFORMAT_DXT45              (15  <<  0)
-#define     R200_TXFORMAT_FORMAT_MASK        (31 <<  0)
-#define     R200_TXFORMAT_FORMAT_SHIFT       0
-#define     R200_TXFORMAT_ALPHA_IN_MAP       (1  <<  6)
-#define     R200_TXFORMAT_NON_POWER2         (1  <<  7)
-#define     R200_TXFORMAT_WIDTH_MASK         (15 <<  8)
-#define     R200_TXFORMAT_WIDTH_SHIFT        8
-#define     R200_TXFORMAT_HEIGHT_MASK        (15 << 12)
-#define     R200_TXFORMAT_HEIGHT_SHIFT       12
-#define     R200_TXFORMAT_F5_WIDTH_MASK      (15 << 16)	/* cube face 5 */
-#define     R200_TXFORMAT_F5_WIDTH_SHIFT     16
-#define     R200_TXFORMAT_F5_HEIGHT_MASK     (15 << 20)
-#define     R200_TXFORMAT_F5_HEIGHT_SHIFT    20
-#define     R200_TXFORMAT_ST_ROUTE_STQ0      (0  << 24)
-#define     R200_TXFORMAT_ST_ROUTE_STQ1      (1  << 24)
-#define     R200_TXFORMAT_ST_ROUTE_STQ2      (2  << 24)
-#define     R200_TXFORMAT_ST_ROUTE_STQ3      (3  << 24)
-#define     R200_TXFORMAT_ST_ROUTE_STQ4      (4  << 24)
-#define     R200_TXFORMAT_ST_ROUTE_STQ5      (5  << 24)
-#define     R200_TXFORMAT_ST_ROUTE_MASK      (7  << 24)
-#define     R200_TXFORMAT_ST_ROUTE_SHIFT     24
-#define     R200_TXFORMAT_ALPHA_MASK_ENABLE  (1  << 28)
-#define     R200_TXFORMAT_CHROMA_KEY_ENABLE  (1  << 29)
-#define     R200_TXFORMAT_CUBIC_MAP_ENABLE   (1  << 30)
-#define R200_PP_TXFORMAT_X_0              0x2c08
-#define     R200_DEPTH_LOG2_MASK                      (0xf << 0)
-#define     R200_DEPTH_LOG2_SHIFT                     0
-#define     R200_VOLUME_FILTER_SHIFT                  4
-#define     R200_VOLUME_FILTER_MASK                   (1 << 4)
-#define     R200_VOLUME_FILTER_NEAREST                (0 << 4)
-#define     R200_VOLUME_FILTER_LINEAR                 (1 << 4)
-#define     R200_WRAPEN_Q                             (1  << 8)
-#define     R200_CLAMP_Q_WRAP                         (0  << 9)
-#define     R200_CLAMP_Q_MIRROR                       (1  << 9)
-#define     R200_CLAMP_Q_CLAMP_LAST                   (2  << 9)
-#define     R200_CLAMP_Q_MIRROR_CLAMP_LAST            (3  << 9)
-#define     R200_CLAMP_Q_CLAMP_BORDER                 (4  << 9)
-#define     R200_CLAMP_Q_MIRROR_CLAMP_BORDER          (5  << 9)
-#define     R200_CLAMP_Q_CLAMP_GL                     (6  << 9)
-#define     R200_CLAMP_Q_MIRROR_CLAMP_GL              (7  << 9)
-#define     R200_CLAMP_Q_MASK                         (7  << 9)
-#define     R200_MIN_MIP_LEVEL_MASK                   (0xff << 12)
-#define     R200_MIN_MIP_LEVEL_SHIFT                  12
-#define     R200_TEXCOORD_NONPROJ                     (0  << 16)
-#define     R200_TEXCOORD_CUBIC_ENV                   (1  << 16)
-#define     R200_TEXCOORD_VOLUME                      (2  << 16)
-#define     R200_TEXCOORD_PROJ                        (3  << 16)
-#define     R200_TEXCOORD_DEPTH                       (4  << 16)
-#define     R200_TEXCOORD_1D_PROJ                     (5  << 16)
-#define     R200_TEXCOORD_1D                          (6  << 16)
-#define     R200_TEXCOORD_ZERO                        (7  << 16)
-#define     R200_TEXCOORD_MASK                        (7  << 16)
-#define     R200_LOD_BIAS_MASK                        (0xfff80000)
-#define     R200_LOD_BIAS_SHIFT                       19
-#define R200_PP_TXSIZE_0                  0x2c0c	/* NPOT only */
-#define R200_PP_TXPITCH_0                 0x2c10	/* NPOT only */
-#define R200_PP_BORDER_COLOR_0            0x2c14
-#define R200_PP_CUBIC_FACES_0             0x2c18
-#define     R200_FACE_WIDTH_1_SHIFT                   0
-#define     R200_FACE_HEIGHT_1_SHIFT                  4
-#define     R200_FACE_WIDTH_1_MASK                   (0xf << 0)
-#define     R200_FACE_HEIGHT_1_MASK                  (0xf << 4)
-#define     R200_FACE_WIDTH_2_SHIFT                   8
-#define     R200_FACE_HEIGHT_2_SHIFT                 12
-#define     R200_FACE_WIDTH_2_MASK                   (0xf << 8)
-#define     R200_FACE_HEIGHT_2_MASK                  (0xf << 12)
-#define     R200_FACE_WIDTH_3_SHIFT                  16
-#define     R200_FACE_HEIGHT_3_SHIFT                 20
-#define     R200_FACE_WIDTH_3_MASK                   (0xf << 16)
-#define     R200_FACE_HEIGHT_3_MASK                  (0xf << 20)
-#define     R200_FACE_WIDTH_4_SHIFT                  24
-#define     R200_FACE_HEIGHT_4_SHIFT                 28
-#define     R200_FACE_WIDTH_4_MASK                   (0xf << 24)
-#define     R200_FACE_HEIGHT_4_MASK                  (0xf << 28)
-#define R200_PP_TXFILTER_1                0x2c20
-#define R200_PP_TXFORMAT_1                0x2c24
-#define R200_PP_TXFORMAT_X_1              0x2c28
-#define R200_PP_TXSIZE_1                  0x2c2c
-#define R200_PP_TXPITCH_1                 0x2c30
-#define R200_PP_BORDER_COLOR_1            0x2c34
-#define R200_PP_CUBIC_FACES_1             0x2c38
-#define R200_PP_TXFILTER_2                0x2c40
-#define R200_PP_TXFORMAT_2                0x2c44
-#define R200_PP_TXSIZE_2                  0x2c4c
-#define R200_PP_TXFORMAT_X_2              0x2c48
-#define R200_PP_TXPITCH_2                 0x2c50
-#define R200_PP_BORDER_COLOR_2            0x2c54
-#define R200_PP_CUBIC_FACES_2             0x2c58
-#define R200_PP_TXFILTER_3                0x2c60
-#define R200_PP_TXFORMAT_3                0x2c64
-#define R200_PP_TXSIZE_3                  0x2c6c
-#define R200_PP_TXFORMAT_X_3              0x2c68
-#define R200_PP_TXPITCH_3                 0x2c70
-#define R200_PP_BORDER_COLOR_3            0x2c74
-#define R200_PP_CUBIC_FACES_3             0x2c78
-#define R200_PP_TXFILTER_4                0x2c80
-#define R200_PP_TXFORMAT_4                0x2c84
-#define R200_PP_TXSIZE_4                  0x2c8c
-#define R200_PP_TXFORMAT_X_4              0x2c88
-#define R200_PP_TXPITCH_4                 0x2c90
-#define R200_PP_BORDER_COLOR_4            0x2c94
-#define R200_PP_CUBIC_FACES_4             0x2c98
-#define R200_PP_TXFILTER_5                0x2ca0
-#define R200_PP_TXFORMAT_5                0x2ca4
-#define R200_PP_TXSIZE_5                  0x2cac
-#define R200_PP_TXFORMAT_X_5              0x2ca8
-#define R200_PP_TXPITCH_5                 0x2cb0
-#define R200_PP_BORDER_COLOR_5            0x2cb4
-#define R200_PP_CUBIC_FACES_5             0x2cb8
-/* gap */
-#define R200_PP_CNTL_X             0x2cc4
-/* gap */
-#define R200_PP_TXOFFSET_0                0x2d00
-#define     R200_TXO_ENDIAN_NO_SWAP     (0 << 0)
-#define     R200_TXO_ENDIAN_BYTE_SWAP   (1 << 0)
-#define     R200_TXO_ENDIAN_WORD_SWAP   (2 << 0)
-#define     R200_TXO_ENDIAN_HALFDW_SWAP (3 << 0)
-#define     R200_TXO_OFFSET_MASK        0xffffffe0
-#define     R200_TXO_OFFSET_SHIFT       5
-#define R200_PP_CUBIC_OFFSET_F1_0         0x2d04
-#define R200_PP_CUBIC_OFFSET_F2_0         0x2d08
-#define R200_PP_CUBIC_OFFSET_F3_0         0x2d0c
-#define R200_PP_CUBIC_OFFSET_F4_0         0x2d10
-#define R200_PP_CUBIC_OFFSET_F5_0         0x2d14
-#define R200_PP_TXOFFSET_1                0x2d18
-#define R200_PP_CUBIC_OFFSET_F1_1         0x2d1c
-#define R200_PP_CUBIC_OFFSET_F2_1         0x2d20
-#define R200_PP_CUBIC_OFFSET_F3_1         0x2d24
-#define R200_PP_CUBIC_OFFSET_F4_1         0x2d28
-#define R200_PP_CUBIC_OFFSET_F5_1         0x2d2c
-#define R200_PP_TXOFFSET_2                0x2d30
-#define R200_PP_CUBIC_OFFSET_F1_2         0x2d34
-#define R200_PP_CUBIC_OFFSET_F2_2         0x2d38
-#define R200_PP_CUBIC_OFFSET_F3_2         0x2d3c
-#define R200_PP_CUBIC_OFFSET_F4_2         0x2d40
-#define R200_PP_CUBIC_OFFSET_F5_2         0x2d44
-#define R200_PP_TXOFFSET_3                0x2d48
-#define R200_PP_CUBIC_OFFSET_F1_3         0x2d4c
-#define R200_PP_CUBIC_OFFSET_F2_3         0x2d50
-#define R200_PP_CUBIC_OFFSET_F3_3         0x2d54
-#define R200_PP_CUBIC_OFFSET_F4_3         0x2d58
-#define R200_PP_CUBIC_OFFSET_F5_3         0x2d5c
-#define R200_PP_TXOFFSET_4                0x2d60
-#define R200_PP_CUBIC_OFFSET_F1_4         0x2d64
-#define R200_PP_CUBIC_OFFSET_F2_4         0x2d68
-#define R200_PP_CUBIC_OFFSET_F3_4         0x2d6c
-#define R200_PP_CUBIC_OFFSET_F4_4         0x2d70
-#define R200_PP_CUBIC_OFFSET_F5_4         0x2d74
-#define R200_PP_TXOFFSET_5                0x2d78
-#define R200_PP_CUBIC_OFFSET_F1_5         0x2d7c
-#define R200_PP_CUBIC_OFFSET_F2_5         0x2d80
-#define R200_PP_CUBIC_OFFSET_F3_5         0x2d84
-#define R200_PP_CUBIC_OFFSET_F4_5         0x2d88
-#define R200_PP_CUBIC_OFFSET_F5_5         0x2d8c
-/* gap */
-#define R200_PP_TAM_DEBUG3                0x2d9c
-/* gap */
-#define R200_PP_TFACTOR_0                 0x2ee0
-#define R200_PP_TFACTOR_1                 0x2ee4
-#define R200_PP_TFACTOR_2                 0x2ee8
-#define R200_PP_TFACTOR_3                 0x2eec
-#define R200_PP_TFACTOR_4                 0x2ef0
-#define R200_PP_TFACTOR_5                 0x2ef4
-/* gap */
-#define R200_PP_TXCBLEND_0                0x2f00
-#define     R200_TXC_ARG_A_ZERO                (0)
-#define     R200_TXC_ARG_A_CURRENT_COLOR       (2)
-#define     R200_TXC_ARG_A_CURRENT_ALPHA       (3)
-#define     R200_TXC_ARG_A_DIFFUSE_COLOR       (4)
-#define     R200_TXC_ARG_A_DIFFUSE_ALPHA       (5)
-#define     R200_TXC_ARG_A_SPECULAR_COLOR      (6)
-#define     R200_TXC_ARG_A_SPECULAR_ALPHA      (7)
-#define     R200_TXC_ARG_A_TFACTOR_COLOR       (8)
-#define     R200_TXC_ARG_A_TFACTOR_ALPHA       (9)
-#define     R200_TXC_ARG_A_R0_COLOR            (10)
-#define     R200_TXC_ARG_A_R0_ALPHA            (11)
-#define     R200_TXC_ARG_A_R1_COLOR            (12)
-#define     R200_TXC_ARG_A_R1_ALPHA            (13)
-#define     R200_TXC_ARG_A_R2_COLOR            (14)
-#define     R200_TXC_ARG_A_R2_ALPHA            (15)
-#define     R200_TXC_ARG_A_R3_COLOR            (16)
-#define     R200_TXC_ARG_A_R3_ALPHA            (17)
-#define     R200_TXC_ARG_A_R4_COLOR            (18)
-#define     R200_TXC_ARG_A_R4_ALPHA            (19)
-#define     R200_TXC_ARG_A_R5_COLOR            (20)
-#define     R200_TXC_ARG_A_R5_ALPHA            (21)
-#define     R200_TXC_ARG_A_TFACTOR1_COLOR      (26)
-#define     R200_TXC_ARG_A_TFACTOR1_ALPHA      (27)
-#define     R200_TXC_ARG_A_MASK			(31 << 0)
-#define     R200_TXC_ARG_A_SHIFT			0
-#define     R200_TXC_ARG_B_ZERO                (0<<5)
-#define     R200_TXC_ARG_B_CURRENT_COLOR       (2<<5)
-#define     R200_TXC_ARG_B_CURRENT_ALPHA       (3<<5)
-#define     R200_TXC_ARG_B_DIFFUSE_COLOR       (4<<5)
-#define     R200_TXC_ARG_B_DIFFUSE_ALPHA       (5<<5)
-#define     R200_TXC_ARG_B_SPECULAR_COLOR      (6<<5)
-#define     R200_TXC_ARG_B_SPECULAR_ALPHA      (7<<5)
-#define     R200_TXC_ARG_B_TFACTOR_COLOR       (8<<5)
-#define     R200_TXC_ARG_B_TFACTOR_ALPHA       (9<<5)
-#define     R200_TXC_ARG_B_R0_COLOR            (10<<5)
-#define     R200_TXC_ARG_B_R0_ALPHA            (11<<5)
-#define     R200_TXC_ARG_B_R1_COLOR            (12<<5)
-#define     R200_TXC_ARG_B_R1_ALPHA            (13<<5)
-#define     R200_TXC_ARG_B_R2_COLOR            (14<<5)
-#define     R200_TXC_ARG_B_R2_ALPHA            (15<<5)
-#define     R200_TXC_ARG_B_R3_COLOR            (16<<5)
-#define     R200_TXC_ARG_B_R3_ALPHA            (17<<5)
-#define     R200_TXC_ARG_B_R4_COLOR            (18<<5)
-#define     R200_TXC_ARG_B_R4_ALPHA            (19<<5)
-#define     R200_TXC_ARG_B_R5_COLOR            (20<<5)
-#define     R200_TXC_ARG_B_R5_ALPHA            (21<<5)
-#define     R200_TXC_ARG_B_TFACTOR1_COLOR      (26<<5)
-#define     R200_TXC_ARG_B_TFACTOR1_ALPHA      (27<<5)
-#define     R200_TXC_ARG_B_MASK			(31 << 5)
-#define     R200_TXC_ARG_B_SHIFT			5
-#define     R200_TXC_ARG_C_ZERO                (0<<10)
-#define     R200_TXC_ARG_C_CURRENT_COLOR       (2<<10)
-#define     R200_TXC_ARG_C_CURRENT_ALPHA       (3<<10)
-#define     R200_TXC_ARG_C_DIFFUSE_COLOR       (4<<10)
-#define     R200_TXC_ARG_C_DIFFUSE_ALPHA       (5<<10)
-#define     R200_TXC_ARG_C_SPECULAR_COLOR      (6<<10)
-#define     R200_TXC_ARG_C_SPECULAR_ALPHA      (7<<10)
-#define     R200_TXC_ARG_C_TFACTOR_COLOR       (8<<10)
-#define     R200_TXC_ARG_C_TFACTOR_ALPHA       (9<<10)
-#define     R200_TXC_ARG_C_R0_COLOR            (10<<10)
-#define     R200_TXC_ARG_C_R0_ALPHA            (11<<10)
-#define     R200_TXC_ARG_C_R1_COLOR            (12<<10)
-#define     R200_TXC_ARG_C_R1_ALPHA            (13<<10)
-#define     R200_TXC_ARG_C_R2_COLOR            (14<<10)
-#define     R200_TXC_ARG_C_R2_ALPHA            (15<<10)
-#define     R200_TXC_ARG_C_R3_COLOR            (16<<10)
-#define     R200_TXC_ARG_C_R3_ALPHA            (17<<10)
-#define     R200_TXC_ARG_C_R4_COLOR            (18<<10)
-#define     R200_TXC_ARG_C_R4_ALPHA            (19<<10)
-#define     R200_TXC_ARG_C_R5_COLOR            (20<<10)
-#define     R200_TXC_ARG_C_R5_ALPHA            (21<<10)
-#define     R200_TXC_ARG_C_TFACTOR1_COLOR      (26<<10)
-#define     R200_TXC_ARG_C_TFACTOR1_ALPHA      (27<<10)
-#define     R200_TXC_ARG_C_MASK			(31 << 10)
-#define     R200_TXC_ARG_C_SHIFT			10
-#define     R200_TXC_COMP_ARG_A                    (1 << 16)
-#define     R200_TXC_COMP_ARG_A_SHIFT              (16)
-#define     R200_TXC_BIAS_ARG_A                    (1 << 17)
-#define     R200_TXC_SCALE_ARG_A                   (1 << 18)
-#define     R200_TXC_NEG_ARG_A                     (1 << 19)
-#define     R200_TXC_COMP_ARG_B                    (1 << 20)
-#define     R200_TXC_COMP_ARG_B_SHIFT              (20)
-#define     R200_TXC_BIAS_ARG_B                    (1 << 21)
-#define     R200_TXC_SCALE_ARG_B                   (1 << 22)
-#define     R200_TXC_NEG_ARG_B                     (1 << 23)
-#define     R200_TXC_COMP_ARG_C                    (1 << 24)
-#define     R200_TXC_COMP_ARG_C_SHIFT              (24)
-#define     R200_TXC_BIAS_ARG_C                    (1 << 25)
-#define     R200_TXC_SCALE_ARG_C                   (1 << 26)
-#define     R200_TXC_NEG_ARG_C                     (1 << 27)
-#define     R200_TXC_OP_MADD                        (0 << 28)
-#define     R200_TXC_OP_CND0                       (2 << 28)
-#define     R200_TXC_OP_LERP                       (3 << 28)
-#define     R200_TXC_OP_DOT3                       (4 << 28)
-#define     R200_TXC_OP_DOT4                       (5 << 28)
-#define     R200_TXC_OP_CONDITIONAL                (6 << 28)
-#define     R200_TXC_OP_DOT2_ADD                   (7 << 28)
-#define     R200_TXC_OP_MASK                       (7 << 28)
-#define R200_PP_TXCBLEND2_0                0x2f04
-#define     R200_TXC_TFACTOR_SEL_SHIFT             0
-#define     R200_TXC_TFACTOR_SEL_MASK              0x7
-#define     R200_TXC_TFACTOR1_SEL_SHIFT            4
-#define     R200_TXC_TFACTOR1_SEL_MASK             (0x7 << 4)
-#define     R200_TXC_SCALE_SHIFT                   8
-#define     R200_TXC_SCALE_MASK                    (7 << 8)
-#define     R200_TXC_SCALE_1X                      (0 << 8)
-#define     R200_TXC_SCALE_2X                      (1 << 8)
-#define     R200_TXC_SCALE_4X                      (2 << 8)
-#define     R200_TXC_SCALE_8X                      (3 << 8)
-#define     R200_TXC_SCALE_INV2                    (5 << 8)
-#define     R200_TXC_SCALE_INV4                    (6 << 8)
-#define     R200_TXC_SCALE_INV8                    (7 << 8)
-#define     R200_TXC_CLAMP_SHIFT                   12
-#define     R200_TXC_CLAMP_MASK                    (3 << 12)
-#define     R200_TXC_CLAMP_WRAP                    (0 << 12)
-#define     R200_TXC_CLAMP_0_1                     (1 << 12)
-#define     R200_TXC_CLAMP_8_8                     (2 << 12)
-#define     R200_TXC_OUTPUT_REG_MASK               (7 << 16)
-#define     R200_TXC_OUTPUT_REG_NONE               (0 << 16)
-#define     R200_TXC_OUTPUT_REG_R0                 (1 << 16)
-#define     R200_TXC_OUTPUT_REG_R1                 (2 << 16)
-#define     R200_TXC_OUTPUT_REG_R2                 (3 << 16)
-#define     R200_TXC_OUTPUT_REG_R3                 (4 << 16)
-#define     R200_TXC_OUTPUT_REG_R4                 (5 << 16)
-#define     R200_TXC_OUTPUT_REG_R5                 (6 << 16)
-#define     R200_TXC_OUTPUT_MASK_MASK              (7 << 20)
-#define     R200_TXC_OUTPUT_MASK_RGB               (0 << 20)
-#define     R200_TXC_OUTPUT_MASK_RG                (1 << 20)
-#define     R200_TXC_OUTPUT_MASK_RB                (2 << 20)
-#define     R200_TXC_OUTPUT_MASK_R                 (3 << 20)
-#define     R200_TXC_OUTPUT_MASK_GB                (4 << 20)
-#define     R200_TXC_OUTPUT_MASK_G                 (5 << 20)
-#define     R200_TXC_OUTPUT_MASK_B                 (6 << 20)
-#define     R200_TXC_OUTPUT_MASK_NONE              (7 << 20)
-#define     R200_TXC_REPL_NORMAL                   0
-#define     R200_TXC_REPL_RED                      1
-#define     R200_TXC_REPL_GREEN                    2
-#define     R200_TXC_REPL_BLUE                     3
-#define     R200_TXC_REPL_ARG_A_SHIFT              26
-#define     R200_TXC_REPL_ARG_A_MASK               (3 << 26)
-#define     R200_TXC_REPL_ARG_B_SHIFT              28
-#define     R200_TXC_REPL_ARG_B_MASK               (3 << 28)
-#define     R200_TXC_REPL_ARG_C_SHIFT              30
-#define     R200_TXC_REPL_ARG_C_MASK               (3 << 30)
-#define R200_PP_TXABLEND_0                0x2f08
-#define     R200_TXA_ARG_A_ZERO              (0)
-#define     R200_TXA_ARG_A_CURRENT_ALPHA     (2)	/* guess */
-#define     R200_TXA_ARG_A_CURRENT_BLUE      (3)	/* guess */
-#define     R200_TXA_ARG_A_DIFFUSE_ALPHA     (4)
-#define     R200_TXA_ARG_A_DIFFUSE_BLUE      (5)
-#define     R200_TXA_ARG_A_SPECULAR_ALPHA    (6)
-#define     R200_TXA_ARG_A_SPECULAR_BLUE     (7)
-#define     R200_TXA_ARG_A_TFACTOR_ALPHA     (8)
-#define     R200_TXA_ARG_A_TFACTOR_BLUE      (9)
-#define     R200_TXA_ARG_A_R0_ALPHA          (10)
-#define     R200_TXA_ARG_A_R0_BLUE           (11)
-#define     R200_TXA_ARG_A_R1_ALPHA          (12)
-#define     R200_TXA_ARG_A_R1_BLUE           (13)
-#define     R200_TXA_ARG_A_R2_ALPHA          (14)
-#define     R200_TXA_ARG_A_R2_BLUE           (15)
-#define     R200_TXA_ARG_A_R3_ALPHA          (16)
-#define     R200_TXA_ARG_A_R3_BLUE           (17)
-#define     R200_TXA_ARG_A_R4_ALPHA          (18)
-#define     R200_TXA_ARG_A_R4_BLUE           (19)
-#define     R200_TXA_ARG_A_R5_ALPHA          (20)
-#define     R200_TXA_ARG_A_R5_BLUE           (21)
-#define     R200_TXA_ARG_A_TFACTOR1_ALPHA    (26)
-#define     R200_TXA_ARG_A_TFACTOR1_BLUE     (27)
-#define     R200_TXA_ARG_A_MASK			(31 << 0)
-#define     R200_TXA_ARG_A_SHIFT			0
-#define     R200_TXA_ARG_B_ZERO              (0<<5)
-#define     R200_TXA_ARG_B_CURRENT_ALPHA     (2<<5)	/* guess */
-#define     R200_TXA_ARG_B_CURRENT_BLUE      (3<<5)	/* guess */
-#define     R200_TXA_ARG_B_DIFFUSE_ALPHA     (4<<5)
-#define     R200_TXA_ARG_B_DIFFUSE_BLUE      (5<<5)
-#define     R200_TXA_ARG_B_SPECULAR_ALPHA    (6<<5)
-#define     R200_TXA_ARG_B_SPECULAR_BLUE     (7<<5)
-#define     R200_TXA_ARG_B_TFACTOR_ALPHA     (8<<5)
-#define     R200_TXA_ARG_B_TFACTOR_BLUE      (9<<5)
-#define     R200_TXA_ARG_B_R0_ALPHA          (10<<5)
-#define     R200_TXA_ARG_B_R0_BLUE           (11<<5)
-#define     R200_TXA_ARG_B_R1_ALPHA          (12<<5)
-#define     R200_TXA_ARG_B_R1_BLUE           (13<<5)
-#define     R200_TXA_ARG_B_R2_ALPHA          (14<<5)
-#define     R200_TXA_ARG_B_R2_BLUE           (15<<5)
-#define     R200_TXA_ARG_B_R3_ALPHA          (16<<5)
-#define     R200_TXA_ARG_B_R3_BLUE           (17<<5)
-#define     R200_TXA_ARG_B_R4_ALPHA          (18<<5)
-#define     R200_TXA_ARG_B_R4_BLUE           (19<<5)
-#define     R200_TXA_ARG_B_R5_ALPHA          (20<<5)
-#define     R200_TXA_ARG_B_R5_BLUE           (21<<5)
-#define     R200_TXA_ARG_B_TFACTOR1_ALPHA    (26<<5)
-#define     R200_TXA_ARG_B_TFACTOR1_BLUE     (27<<5)
-#define     R200_TXA_ARG_B_MASK			(31 << 5)
-#define     R200_TXA_ARG_B_SHIFT			5
-#define     R200_TXA_ARG_C_ZERO              (0<<10)
-#define     R200_TXA_ARG_C_CURRENT_ALPHA     (2<<10)	/* guess */
-#define     R200_TXA_ARG_C_CURRENT_BLUE      (3<<10)	/* guess */
-#define     R200_TXA_ARG_C_DIFFUSE_ALPHA     (4<<10)
-#define     R200_TXA_ARG_C_DIFFUSE_BLUE      (5<<10)
-#define     R200_TXA_ARG_C_SPECULAR_ALPHA    (6<<10)
-#define     R200_TXA_ARG_C_SPECULAR_BLUE     (7<<10)
-#define     R200_TXA_ARG_C_TFACTOR_ALPHA     (8<<10)
-#define     R200_TXA_ARG_C_TFACTOR_BLUE      (9<<10)
-#define     R200_TXA_ARG_C_R0_ALPHA          (10<<10)
-#define     R200_TXA_ARG_C_R0_BLUE           (11<<10)
-#define     R200_TXA_ARG_C_R1_ALPHA          (12<<10)
-#define     R200_TXA_ARG_C_R1_BLUE           (13<<10)
-#define     R200_TXA_ARG_C_R2_ALPHA          (14<<10)
-#define     R200_TXA_ARG_C_R2_BLUE           (15<<10)
-#define     R200_TXA_ARG_C_R3_ALPHA          (16<<10)
-#define     R200_TXA_ARG_C_R3_BLUE           (17<<10)
-#define     R200_TXA_ARG_C_R4_ALPHA          (18<<10)
-#define     R200_TXA_ARG_C_R4_BLUE           (19<<10)
-#define     R200_TXA_ARG_C_R5_ALPHA          (20<<10)
-#define     R200_TXA_ARG_C_R5_BLUE           (21<<10)
-#define     R200_TXA_ARG_C_TFACTOR1_ALPHA    (26<<10)
-#define     R200_TXA_ARG_C_TFACTOR1_BLUE     (27<<10)
-#define     R200_TXA_ARG_C_MASK			(31 << 10)
-#define     R200_TXA_ARG_C_SHIFT			10
-#define     R200_TXA_COMP_ARG_A                    (1 << 16)
-#define     R200_TXA_COMP_ARG_A_SHIFT              (16)
-#define     R200_TXA_BIAS_ARG_A                    (1 << 17)
-#define     R200_TXA_SCALE_ARG_A                   (1 << 18)
-#define     R200_TXA_NEG_ARG_A                     (1 << 19)
-#define     R200_TXA_COMP_ARG_B                    (1 << 20)
-#define     R200_TXA_COMP_ARG_B_SHIFT              (20)
-#define     R200_TXA_BIAS_ARG_B                    (1 << 21)
-#define     R200_TXA_SCALE_ARG_B                   (1 << 22)
-#define     R200_TXA_NEG_ARG_B                     (1 << 23)
-#define     R200_TXA_COMP_ARG_C                    (1 << 24)
-#define     R200_TXA_COMP_ARG_C_SHIFT              (24)
-#define     R200_TXA_BIAS_ARG_C                    (1 << 25)
-#define     R200_TXA_SCALE_ARG_C                   (1 << 26)
-#define     R200_TXA_NEG_ARG_C                     (1 << 27)
-#define     R200_TXA_OP_MADD                       (0 << 28)
-#define     R200_TXA_OP_CND0                       (2 << 28)
-#define     R200_TXA_OP_LERP                       (3 << 28)
-#define     R200_TXA_OP_CONDITIONAL                (6 << 28)
-#define     R200_TXA_OP_MASK                       (7 << 28)
-#define R200_PP_TXABLEND2_0                0x2f0c
-#define     R200_TXA_TFACTOR_SEL_SHIFT             0
-#define     R200_TXA_TFACTOR_SEL_MASK              0x7
-#define     R200_TXA_TFACTOR1_SEL_SHIFT            4
-#define     R200_TXA_TFACTOR1_SEL_MASK             (0x7 << 4)
-#define     R200_TXA_SCALE_SHIFT                   8
-#define     R200_TXA_SCALE_MASK                    (7 << 8)
-#define     R200_TXA_SCALE_1X                      (0 << 8)
-#define     R200_TXA_SCALE_2X                      (1 << 8)
-#define     R200_TXA_SCALE_4X                      (2 << 8)
-#define     R200_TXA_SCALE_8X                      (3 << 8)
-#define     R200_TXA_SCALE_INV2                    (5 << 8)
-#define     R200_TXA_SCALE_INV4                    (6 << 8)
-#define     R200_TXA_SCALE_INV8                    (7 << 8)
-#define     R200_TXA_CLAMP_SHIFT                   12
-#define     R200_TXA_CLAMP_MASK                    (3 << 12)
-#define     R200_TXA_CLAMP_WRAP                    (0 << 12)
-#define     R200_TXA_CLAMP_0_1                     (1 << 12)
-#define     R200_TXA_CLAMP_8_8                     (2 << 12)
-#define     R200_TXA_OUTPUT_REG_MASK               (7 << 16)
-#define     R200_TXA_OUTPUT_REG_NONE               (0 << 16)
-#define     R200_TXA_OUTPUT_REG_R0                 (1 << 16)
-#define     R200_TXA_OUTPUT_REG_R1                 (2 << 16)
-#define     R200_TXA_OUTPUT_REG_R2                 (3 << 16)
-#define     R200_TXA_OUTPUT_REG_R3                 (4 << 16)
-#define     R200_TXA_OUTPUT_REG_R4                 (5 << 16)
-#define     R200_TXA_OUTPUT_REG_R5                 (6 << 16)
-#define     R200_TXA_DOT_ALPHA                     (1 << 20)
-#define     R200_TXA_REPL_NORMAL                   0
-#define     R200_TXA_REPL_RED                      1
-#define     R200_TXA_REPL_GREEN                    2
-#define     R200_TXA_REPL_ARG_A_SHIFT              26
-#define     R200_TXA_REPL_ARG_A_MASK               (3 << 26)
-#define     R200_TXA_REPL_ARG_B_SHIFT              28
-#define     R200_TXA_REPL_ARG_B_MASK               (3 << 28)
-#define     R200_TXA_REPL_ARG_C_SHIFT              30
-#define     R200_TXA_REPL_ARG_C_MASK               (3 << 30)
-#define R200_PP_TXCBLEND_1                0x2f10
-#define R200_PP_TXCBLEND2_1               0x2f14
-#define R200_PP_TXABLEND_1                0x2f18
-#define R200_PP_TXABLEND2_1               0x2f1c
-#define R200_PP_TXCBLEND_2                0x2f20
-#define R200_PP_TXCBLEND2_2               0x2f24
-#define R200_PP_TXABLEND_2                0x2f28
-#define R200_PP_TXABLEND2_2               0x2f2c
-#define R200_PP_TXCBLEND_3                0x2f30
-#define R200_PP_TXCBLEND2_3               0x2f34
-#define R200_PP_TXABLEND_3                0x2f38
-#define R200_PP_TXABLEND2_3               0x2f3c
-#define R200_PP_TXCBLEND_4                0x2f40
-#define R200_PP_TXCBLEND2_4               0x2f44
-#define R200_PP_TXABLEND_4                0x2f48
-#define R200_PP_TXABLEND2_4               0x2f4c
-#define R200_PP_TXCBLEND_5                0x2f50
-#define R200_PP_TXCBLEND2_5               0x2f54
-#define R200_PP_TXABLEND_5                0x2f58
-#define R200_PP_TXABLEND2_5               0x2f5c
-#define R200_PP_TXCBLEND_6                0x2f60
-#define R200_PP_TXCBLEND2_6               0x2f64
-#define R200_PP_TXABLEND_6                0x2f68
-#define R200_PP_TXABLEND2_6               0x2f6c
-#define R200_PP_TXCBLEND_7                0x2f70
-#define R200_PP_TXCBLEND2_7               0x2f74
-#define R200_PP_TXABLEND_7                0x2f78
-#define R200_PP_TXABLEND2_7               0x2f7c
-/* gap */
-#define R200_RB3D_BLENDCOLOR               0x3218	/* ARGB 8888 */
-#define R200_RB3D_ABLENDCNTL               0x321C	/* see BLENDCTL */
-#define R200_RB3D_CBLENDCNTL               0x3220	/* see BLENDCTL */
-
-/*
- * Offsets in TCL vector state.  NOTE: Hardwiring matrix positions.
- * Multiple contexts could collaberate to eliminate state bouncing.
- */
-#define R200_VS_LIGHT_AMBIENT_ADDR          0x00000028
-#define R200_VS_LIGHT_DIFFUSE_ADDR          0x00000030
-#define R200_VS_LIGHT_SPECULAR_ADDR         0x00000038
-#define R200_VS_LIGHT_DIRPOS_ADDR           0x00000040
-#define R200_VS_LIGHT_HWVSPOT_ADDR          0x00000048
-#define R200_VS_LIGHT_ATTENUATION_ADDR      0x00000050
-#define R200_VS_SPOT_DUAL_CONE              0x00000058
-#define R200_VS_GLOBAL_AMBIENT_ADDR         0x0000005C
-#define R200_VS_FOG_PARAM_ADDR              0x0000005D
-#define R200_VS_EYE_VECTOR_ADDR             0x0000005E
-#define R200_VS_UCP_ADDR                    0x00000060
-#define R200_VS_PNT_SPRITE_VPORT_SCALE      0x00000068
-#define R200_VS_MATRIX_0_MV                 0x00000080
-#define R200_VS_MATRIX_1_INV_MV        	    0x00000084
-#define R200_VS_MATRIX_2_MVP        	    0x00000088
-#define R200_VS_MATRIX_3_TEX0        	    0x0000008C
-#define R200_VS_MATRIX_4_TEX1        	    0x00000090
-#define R200_VS_MATRIX_5_TEX2        	    0x00000094
-#define R200_VS_MATRIX_6_TEX3        	    0x00000098
-#define R200_VS_MATRIX_7_TEX4        	    0x0000009C
-#define R200_VS_MATRIX_8_TEX5        	    0x000000A0
-#define R200_VS_MAT_0_EMISS                 0x000000B0
-#define R200_VS_MAT_0_AMB                   0x000000B1
-#define R200_VS_MAT_0_DIF                   0x000000B2
-#define R200_VS_MAT_0_SPEC                  0x000000B3
-#define R200_VS_MAT_1_EMISS                 0x000000B4
-#define R200_VS_MAT_1_AMB                   0x000000B5
-#define R200_VS_MAT_1_DIF                   0x000000B6
-#define R200_VS_MAT_1_SPEC                  0x000000B7
-#define R200_VS_EYE2CLIP_MTX                0x000000B8
-#define R200_VS_PNT_SPRITE_ATT_CONST        0x000000BC
-#define R200_VS_PNT_SPRITE_EYE_IN_MODEL     0x000000BD
-#define R200_VS_PNT_SPRITE_CLAMP            0x000000BE
-#define R200_VS_MAX                         0x000001C0
-
-/*
- * Offsets in TCL scalar state
- */
-#define R200_SS_LIGHT_DCD_ADDR              0x00000000
-#define R200_SS_LIGHT_DCM_ADDR              0x00000008
-#define R200_SS_LIGHT_SPOT_EXPONENT_ADDR    0x00000010
-#define R200_SS_LIGHT_SPOT_CUTOFF_ADDR      0x00000018
-#define R200_SS_LIGHT_SPECULAR_THRESH_ADDR  0x00000020
-#define R200_SS_LIGHT_RANGE_CUTOFF_SQRD     0x00000028
-#define R200_SS_LIGHT_RANGE_ATT_CONST       0x00000030
-#define R200_SS_VERT_GUARD_CLIP_ADJ_ADDR    0x00000080
-#define R200_SS_VERT_GUARD_DISCARD_ADJ_ADDR 0x00000081
-#define R200_SS_HORZ_GUARD_CLIP_ADJ_ADDR    0x00000082
-#define R200_SS_HORZ_GUARD_DISCARD_ADJ_ADDR 0x00000083
-#define R200_SS_MAT_0_SHININESS             0x00000100
-#define R200_SS_MAT_1_SHININESS             0x00000101
-
-/*
- * Matrix indices
- */
-#define R200_MTX_MV                        0
-#define R200_MTX_IMV                       1
-#define R200_MTX_MVP                       2
-#define R200_MTX_TEX0                      3
-#define R200_MTX_TEX1                      4
-#define R200_MTX_TEX2                      5
-#define R200_MTX_TEX3                      6
-#define R200_MTX_TEX4                      7
-#define R200_MTX_TEX5                      8
-
-/* Color formats for 2d packets
- */
-#define R200_CP_COLOR_FORMAT_CI8	2
-#define R200_CP_COLOR_FORMAT_ARGB1555	3
-#define R200_CP_COLOR_FORMAT_RGB565	4
-#define R200_CP_COLOR_FORMAT_ARGB8888	6
-#define R200_CP_COLOR_FORMAT_RGB332	7
-#define R200_CP_COLOR_FORMAT_RGB8	9
-#define R200_CP_COLOR_FORMAT_ARGB4444	15
-
-/*
- * CP type-3 packets
- */
-#define R200_CP_CMD_NOP                 0xC0001000
-#define R200_CP_CMD_NEXT_CHAR           0xC0001900
-#define R200_CP_CMD_PLY_NEXTSCAN        0xC0001D00
-#define R200_CP_CMD_SET_SCISSORS        0xC0001E00
-#define R200_CP_CMD_LOAD_MICROCODE      0xC0002400
-#define R200_CP_CMD_WAIT_FOR_IDLE       0xC0002600
-#define R200_CP_CMD_3D_DRAW_VBUF        0xC0002800
-#define R200_CP_CMD_3D_DRAW_IMMD        0xC0002900
-#define R200_CP_CMD_3D_DRAW_INDX        0xC0002A00
-#define R200_CP_CMD_LOAD_PALETTE        0xC0002C00
-#define R200_CP_CMD_3D_LOAD_VBPNTR      0xC0002F00
-#define R200_CP_CMD_INDX_BUFFER         0xC0003300
-#define R200_CP_CMD_3D_DRAW_VBUF_2      0xC0003400
-#define R200_CP_CMD_3D_DRAW_IMMD_2      0xC0003500
-#define R200_CP_CMD_3D_DRAW_INDX_2      0xC0003600
-#define R200_CP_CMD_PAINT		0xC0009100
-#define R200_CP_CMD_BITBLT		0xC0009200
-#define R200_CP_CMD_SMALLTEXT		0xC0009300
-#define R200_CP_CMD_HOSTDATA_BLT	0xC0009400
-#define R200_CP_CMD_POLYLINE		0xC0009500
-#define R200_CP_CMD_POLYSCANLINES	0xC0009800
-#define R200_CP_CMD_PAINT_MULTI		0xC0009A00
-#define R200_CP_CMD_BITBLT_MULTI	0xC0009B00
-#define R200_CP_CMD_TRANS_BITBLT	0xC0009C00
-
-#endif
diff --git a/src/mesa/drivers/dri/r300/r200_state.h b/src/mesa/drivers/dri/r300/r200_state.h
deleted file mode 100644
index 3e1a9c8ba19..00000000000
--- a/src/mesa/drivers/dri/r300/r200_state.h
+++ /dev/null
@@ -1,58 +0,0 @@
-/*
-Copyright (C) The Weather Channel, Inc.  2002.  All Rights Reserved.
-
-The Weather Channel (TM) funded Tungsten Graphics to develop the
-initial release of the Radeon 8500 driver under the XFree86 license.
-This notice must be preserved.
-
-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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __R200_STATE_H__
-#define __R200_STATE_H__
-
-#ifdef GLX_DIRECT_RENDERING
-
-#include "r200_context.h"
-
-extern void r200InitState(r200ContextPtr rmesa);
-extern void r200InitStateFuncs(struct dd_function_table *functions);
-extern void r200InitTnlFuncs(GLcontext * ctx);
-
-extern void r200UpdateMaterial(GLcontext * ctx);
-
-extern void r200UpdateViewportOffset(GLcontext * ctx);
-extern void r200UpdateWindow(GLcontext * ctx);
-
-extern void r200ValidateState(GLcontext * ctx);
-
-extern void r200PrintDirty(r200ContextPtr rmesa, const char *msg);
-
-extern void r200LightingSpaceChange(GLcontext * ctx);
-
-#endif
-#endif
diff --git a/src/mesa/drivers/dri/r300/r300_context.c b/src/mesa/drivers/dri/r300/r300_context.c
index 54eb081d055..d10a9d87d31 100644
--- a/src/mesa/drivers/dri/r300/r300_context.c
+++ b/src/mesa/drivers/dri/r300/r300_context.c
@@ -44,7 +44,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -73,6 +73,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 int future_hw_tcl_on=1;
 int hw_tcl_on=1;
 
+#define need_GL_EXT_stencil_two_side
 #define need_GL_ARB_multisample
 #define need_GL_ARB_texture_compression
 #define need_GL_ARB_vertex_buffer_object
@@ -126,6 +127,10 @@ const struct dri_extension card_extensions[] = {
   {NULL,				NULL}
 };
 
+const struct dri_extension stencil_two_side[] = {
+  {"GL_EXT_stencil_two_side",		GL_EXT_stencil_two_side_functions},
+};
+
 extern struct tnl_pipeline_stage _r300_render_stage;
 extern const struct tnl_pipeline_stage _r300_tcl_stage;
 extern const struct tnl_pipeline_stage _r300_texrect_stage;
@@ -195,6 +200,8 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	 */
 	driParseConfigFiles(&r300->radeon.optionCache, &screen->optionCache,
 			    screen->driScreen->myNum, "r300");
+	r300->initialMaxAnisotropy = driQueryOptionf(&r300->radeon.optionCache,
+						     "def_max_anisotropy");
 
 	//r300->texmicrotile = GL_TRUE;
 	
@@ -287,7 +294,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 	/* Initialize the software rasterizer and helper modules.
 	 */
 	_swrast_CreateContext(ctx);
-	_ac_CreateContext(ctx);
+	_vbo_CreateContext(ctx);
 	_tnl_CreateContext(ctx);
 	_swsetup_CreateContext(ctx);
 	_swsetup_Wakeup(ctx);
@@ -300,7 +307,7 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 
 	/* Try and keep materials and vertices separate:
 	 */
-	_tnl_isolate_materials(ctx, GL_TRUE);
+/* 	_tnl_isolate_materials(ctx, GL_TRUE); */
 
 	/* Configure swrast and TNL to match hardware characteristics:
 	 */
@@ -331,6 +338,9 @@ GLboolean r300CreateContext(const __GLcontextModes * glVisual,
 
 	driInitExtensions(ctx, card_extensions, GL_TRUE);
 	
+	if (driQueryOptionb(&r300->radeon.optionCache, "disable_stencil_two_side") == 0)
+		driInitSingleExtension(ctx, stencil_two_side);
+	
 	if (r300->radeon.glCtx->Mesa_DXTn && !driQueryOptionb (&r300->radeon.optionCache, "disable_s3tc")) {
 	  _mesa_enable_extension( ctx, "GL_EXT_texture_compression_s3tc" );
 	  _mesa_enable_extension( ctx, "GL_S3_s3tc" );
@@ -478,7 +488,7 @@ void r300DestroyContext(__DRIcontextPrivate * driContextPriv)
 		_swsetup_DestroyContext(r300->radeon.glCtx);
 		_tnl_ProgramCacheDestroy(r300->radeon.glCtx);
 		_tnl_DestroyContext(r300->radeon.glCtx);
-		_ac_DestroyContext(r300->radeon.glCtx);
+		_vbo_DestroyContext(r300->radeon.glCtx);
 		_swrast_DestroyContext(r300->radeon.glCtx);
 		
 		if (r300->dma.current.buf) {
diff --git a/src/mesa/drivers/dri/r300/r300_context.h b/src/mesa/drivers/dri/r300/r300_context.h
index dd3ecbb2357..48b50bca653 100644
--- a/src/mesa/drivers/dri/r300/r300_context.h
+++ b/src/mesa/drivers/dri/r300/r300_context.h
@@ -48,7 +48,10 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_context.h"
 
 #define USER_BUFFERS
-#define RADEON_VTXFMT_A
+/* KW: Disable this code.  Driver should hook into vbo module
+ * directly, see i965 driver for example.
+ */
+/* #define RADEON_VTXFMT_A */
 #define HW_VBOS
 
 /* We don't handle 16 bits elts swapping yet */
@@ -726,6 +729,11 @@ struct r300_fragment_program {
 	GLboolean params_uptodate;
 
 	int max_temp_idx;
+
+	/* the index of the sin constant is stored here */
+	GLint const_sin[2];
+	
+	GLuint optimization;
 };
 
 #define R300_MAX_AOS_ARRAYS		16
@@ -754,7 +762,7 @@ struct radeon_vertex_buffer {
 	
 	struct dt AttribPtr[VERT_ATTRIB_MAX];
 	
-	struct tnl_prim  *Primitive;	              
+	const struct _mesa_prim  *Primitive;	              
 	GLuint      PrimitiveCount;
 	GLint LockFirst;
 	GLsizei LockCount;
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.c b/src/mesa/drivers/dri/r300/r300_fragprog.c
index 32c0128eaa2..8e45bd54030 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog.c
+++ b/src/mesa/drivers/dri/r300/r300_fragprog.c
@@ -33,7 +33,6 @@
 
 /*TODO'S
  *
- * - COS/SIN/SCS instructions
  * - Depth write, WPOS/FOGC inputs
  * - FogOption
  * - Verify results of opcodes for accuracy, I've only checked them
@@ -51,18 +50,110 @@
 #include "r300_fragprog.h"
 #include "r300_reg.h"
 
+/*
+ * Usefull macros and values
+ */
+#define ERROR(fmt, args...) do {			\
+		fprintf(stderr, "%s::%s(): " fmt "\n",	\
+			__FILE__, __func__, ##args);	\
+		rp->error = GL_TRUE;			\
+	} while(0)
+
 #define PFS_INVAL 0xFFFFFFFF
 #define COMPILE_STATE struct r300_pfs_compile_state *cs = rp->cs
 
-static void dump_program(struct r300_fragment_program *rp);
-static void emit_arith(struct r300_fragment_program *rp, int op,
-				pfs_reg_t dest, int mask,
-				pfs_reg_t src0, pfs_reg_t src1, pfs_reg_t src2,
-				int flags);
+#define SWIZZLE_XYZ		0
+#define SWIZZLE_XXX		1
+#define SWIZZLE_YYY		2
+#define SWIZZLE_ZZZ		3
+#define SWIZZLE_WWW		4
+#define SWIZZLE_YZX		5
+#define SWIZZLE_ZXY		6
+#define SWIZZLE_WZY		7
+#define SWIZZLE_111		8
+#define SWIZZLE_000		9
+#define SWIZZLE_HHH		10
+
+#define swizzle(r, x, y, z, w) do_swizzle(rp, r,		\
+					  ((SWIZZLE_##x<<0)|	\
+					   (SWIZZLE_##y<<3)|	\
+					   (SWIZZLE_##z<<6)|	\
+					   (SWIZZLE_##w<<9)),	\
+					  0)
+
+#define REG_TYPE_INPUT		0
+#define REG_TYPE_OUTPUT		1
+#define REG_TYPE_TEMP		2
+#define REG_TYPE_CONST		3
+
+#define REG_TYPE_SHIFT		0
+#define REG_INDEX_SHIFT		2
+#define REG_VSWZ_SHIFT		8
+#define REG_SSWZ_SHIFT		13
+#define REG_NEGV_SHIFT		18
+#define REG_NEGS_SHIFT		19
+#define REG_ABS_SHIFT		20
+#define REG_NO_USE_SHIFT	21
+#define REG_VALID_SHIFT		22
+
+#define REG_TYPE_MASK		(0x03 << REG_TYPE_SHIFT)
+#define REG_INDEX_MASK		(0x3F << REG_INDEX_SHIFT)
+#define REG_VSWZ_MASK		(0x1F << REG_VSWZ_SHIFT)
+#define REG_SSWZ_MASK		(0x1F << REG_SSWZ_SHIFT)
+#define REG_NEGV_MASK		(0x01 << REG_NEGV_SHIFT)
+#define REG_NEGS_MASK		(0x01 << REG_NEGS_SHIFT)
+#define REG_ABS_MASK		(0x01 << REG_ABS_SHIFT)
+#define REG_NO_USE_MASK		(0x01 << REG_NO_USE_SHIFT)
+#define REG_VALID_MASK		(0x01 << REG_VALID_SHIFT)
+
+#define REG(type, index, vswz, sswz, nouse, valid)			\
+	(((type << REG_TYPE_SHIFT) & REG_TYPE_MASK) |			\
+	 ((index << REG_INDEX_SHIFT) & REG_INDEX_MASK) |		\
+	 ((nouse << REG_NO_USE_SHIFT) & REG_NO_USE_MASK) |		\
+	 ((valid << REG_VALID_SHIFT) & REG_VALID_MASK) |		\
+	 ((vswz << REG_VSWZ_SHIFT) & REG_VSWZ_MASK) |			\
+	 ((sswz << REG_SSWZ_SHIFT) & REG_SSWZ_MASK))
+#define REG_GET_TYPE(reg)						\
+	((reg & REG_TYPE_MASK) >> REG_TYPE_SHIFT)
+#define REG_GET_INDEX(reg)						\
+	((reg & REG_INDEX_MASK) >> REG_INDEX_SHIFT)
+#define REG_GET_VSWZ(reg)						\
+	((reg & REG_VSWZ_MASK) >> REG_VSWZ_SHIFT)
+#define REG_GET_SSWZ(reg)						\
+	((reg & REG_SSWZ_MASK) >> REG_SSWZ_SHIFT)
+#define REG_GET_NO_USE(reg)						\
+	((reg & REG_NO_USE_MASK) >> REG_NO_USE_SHIFT)
+#define REG_GET_VALID(reg)						\
+	((reg & REG_VALID_MASK) >> REG_VALID_SHIFT)
+#define REG_SET_TYPE(reg, type)						\
+	reg = ((reg & ~REG_TYPE_MASK) |					\
+	       ((type << REG_TYPE_SHIFT) & REG_TYPE_MASK))
+#define REG_SET_INDEX(reg, index)					\
+	reg = ((reg & ~REG_INDEX_MASK) |				\
+	       ((index << REG_INDEX_SHIFT) & REG_INDEX_MASK))
+#define REG_SET_VSWZ(reg, vswz)						\
+	reg = ((reg & ~REG_VSWZ_MASK) |					\
+	       ((vswz << REG_VSWZ_SHIFT) & REG_VSWZ_MASK))
+#define REG_SET_SSWZ(reg, sswz)						\
+	reg = ((reg & ~REG_SSWZ_MASK) |					\
+	       ((sswz << REG_SSWZ_SHIFT) & REG_SSWZ_MASK))
+#define REG_SET_NO_USE(reg, nouse)					\
+	reg = ((reg & ~REG_NO_USE_MASK) |				\
+	       ((nouse << REG_NO_USE_SHIFT) & REG_NO_USE_MASK))
+#define REG_SET_VALID(reg, valid)					\
+	reg = ((reg & ~REG_VALID_MASK) |				\
+	       ((valid << REG_VALID_SHIFT) & REG_VALID_MASK))
+#define REG_ABS(reg)							\
+	reg = (reg | REG_ABS_MASK)
+#define REG_NEGV(reg)							\
+	reg = (reg | REG_NEGV_MASK)
+#define REG_NEGS(reg)							\
+	reg = (reg | REG_NEGS_MASK)
 
-/***************************************
- * begin: useful data structions for fragment program generation
- ***************************************/
+
+/*
+ * Datas structures for fragment program generation
+ */
 
 /* description of r300 native hw instructions */
 static const struct {
@@ -86,20 +177,23 @@ static const struct {
 	{ "CMPH", 3, R300_FPI0_OUTC_CMPH, PFS_INVAL },
 };
 
-#define MAKE_SWZ3(x, y, z) (MAKE_SWIZZLE4(SWIZZLE_##x, \
-					  SWIZZLE_##y, \
-					  SWIZZLE_##z, \
-					  SWIZZLE_ZERO))
-
-#define SLOT_VECTOR	(1<<0)
-#define SLOT_SCALAR (1<<3)
-#define SLOT_BOTH	(SLOT_VECTOR|SLOT_SCALAR)
 
 /* vector swizzles r300 can support natively, with a couple of
  * cases we handle specially
  *
- * pfs_reg_t.v_swz/pfs_reg_t.s_swz is an index into this table
- **/
+ * REG_VSWZ/REG_SSWZ is an index into this table
+ */
+#define SLOT_VECTOR	(1<<0)
+#define SLOT_SCALAR	(1<<3)
+#define SLOT_BOTH	(SLOT_VECTOR | SLOT_SCALAR)
+
+/* mapping from SWIZZLE_* to r300 native values for scalar insns */
+#define SWIZZLE_HALF 6
+
+#define MAKE_SWZ3(x, y, z) (MAKE_SWIZZLE4(SWIZZLE_##x, \
+					  SWIZZLE_##y, \
+					  SWIZZLE_##z, \
+					  SWIZZLE_ZERO))
 static const struct r300_pfs_swizzle {
 	GLuint hash;	/* swizzle value this matches */
 	GLuint base;	/* base value for hw swizzle */
@@ -117,42 +211,30 @@ static const struct r300_pfs_swizzle {
 	{ MAKE_SWZ3(W, Z, Y), R300_FPI0_ARGC_SRC0CA_WZY, 1, SLOT_BOTH },
 	{ MAKE_SWZ3(ONE, ONE, ONE), R300_FPI0_ARGC_ONE, 0, 0},
 	{ MAKE_SWZ3(ZERO, ZERO, ZERO), R300_FPI0_ARGC_ZERO, 0, 0},
-	{ PFS_INVAL, R300_FPI0_ARGC_HALF, 0, 0},
+	{ MAKE_SWZ3(HALF, HALF, HALF), R300_FPI0_ARGC_HALF, 0, 0},
 	{ PFS_INVAL, 0, 0, 0},
 };
-#define SWIZZLE_XYZ		0
-#define SWIZZLE_XXX		1
-#define SWIZZLE_YYY		2
-#define SWIZZLE_ZZZ		3
-#define SWIZZLE_WWW		4
-#define SWIZZLE_YZX		5
-#define SWIZZLE_ZXY		6
-#define SWIZZLE_WZY		7
-#define SWIZZLE_111		8
-#define SWIZZLE_000		9
-#define SWIZZLE_HHH		10
 
+/* used during matching of non-native swizzles */
 #define SWZ_X_MASK (7 << 0)
 #define SWZ_Y_MASK (7 << 3)
 #define SWZ_Z_MASK (7 << 6)
 #define SWZ_W_MASK (7 << 9)
-/* used during matching of non-native swizzles */
 static const struct {
-	GLuint hash;	/* used to mask matching swizzle components */
+	GLuint hash;		/* used to mask matching swizzle components */
 	int mask;		/* actual outmask */
 	int count;		/* count of components matched */
 } s_mask[] = {
-    { SWZ_X_MASK|SWZ_Y_MASK|SWZ_Z_MASK, 1|2|4, 3},
-    { SWZ_X_MASK|SWZ_Y_MASK, 1|2, 2},
-    { SWZ_X_MASK|SWZ_Z_MASK, 1|4, 2},
-    { SWZ_Y_MASK|SWZ_Z_MASK, 2|4, 2},
-    { SWZ_X_MASK, 1, 1},
-    { SWZ_Y_MASK, 2, 1},
-    { SWZ_Z_MASK, 4, 1},
-    { PFS_INVAL, PFS_INVAL, PFS_INVAL}
+	{ SWZ_X_MASK|SWZ_Y_MASK|SWZ_Z_MASK, 1|2|4, 3},
+	{ SWZ_X_MASK|SWZ_Y_MASK, 1|2, 2},
+	{ SWZ_X_MASK|SWZ_Z_MASK, 1|4, 2},
+	{ SWZ_Y_MASK|SWZ_Z_MASK, 2|4, 2},
+	{ SWZ_X_MASK, 1, 1},
+	{ SWZ_Y_MASK, 2, 1},
+	{ SWZ_Z_MASK, 4, 1},
+	{ PFS_INVAL, PFS_INVAL, PFS_INVAL}
 };
 
-/* mapping from SWIZZLE_* to r300 native values for scalar insns */
 static const struct {
 	int base;	/* hw value of swizzle */
 	int stride;	/* difference between SRC0/1/2 */
@@ -166,58 +248,51 @@ static const struct {
 	{ R300_FPI2_ARGA_ONE    , 0, 0 },
 	{ R300_FPI2_ARGA_HALF   , 0, 0 }
 };
-#define SWIZZLE_HALF 6
 
 /* boiler-plate reg, for convenience */
-static const pfs_reg_t undef = {
-	type: REG_TYPE_TEMP,
-	index: 0,
-	v_swz: SWIZZLE_XYZ,
-	s_swz: SWIZZLE_W,
-	negate_v: 0,
-	negate_s: 0,
-	absolute: 0,
-	no_use: GL_FALSE,
-	valid: GL_FALSE
-};
+static const GLuint undef = REG(REG_TYPE_TEMP,
+				0,
+				SWIZZLE_XYZ,
+				SWIZZLE_W,
+				GL_FALSE,
+				GL_FALSE);
 
 /* constant one source */
-static const pfs_reg_t pfs_one = {
-	type: REG_TYPE_CONST,
-	index: 0,
-	v_swz: SWIZZLE_111,
-	s_swz: SWIZZLE_ONE,
-	valid: GL_TRUE
-};
+static const GLuint pfs_one = REG(REG_TYPE_CONST,
+				  0,
+				  SWIZZLE_111,
+				  SWIZZLE_ONE,
+				  GL_FALSE,
+				  GL_TRUE);
 
 /* constant half source */
-static const pfs_reg_t pfs_half = {
-	type: REG_TYPE_CONST,
-	index: 0,
-	v_swz: SWIZZLE_HHH,
-	s_swz: SWIZZLE_HALF,
-	valid: GL_TRUE
-};
+static const GLuint pfs_half = REG(REG_TYPE_CONST,
+				   0,
+				   SWIZZLE_HHH,
+				   SWIZZLE_HALF,
+				   GL_FALSE,
+				   GL_TRUE);
 
 /* constant zero source */
-static const pfs_reg_t pfs_zero = {
-	type: REG_TYPE_CONST,
-	index: 0,
-	v_swz: SWIZZLE_000,
-	s_swz: SWIZZLE_ZERO,
-	valid: GL_TRUE
-};
-
-/***************************************
- * end: data structures
- ***************************************/
+static const GLuint pfs_zero = REG(REG_TYPE_CONST,
+				   0,
+				   SWIZZLE_000,
+				   SWIZZLE_ZERO,
+				   GL_FALSE,
+				   GL_TRUE);
 
-#define ERROR(fmt, args...) do { \
-		fprintf(stderr, "%s::%s(): " fmt "\n",\
-			__FILE__, __func__, ##args);  \
-		rp->error = GL_TRUE; \
-} while(0)
+/*
+ * Common functions prototypes
+ */
+static void dump_program(struct r300_fragment_program *rp);
+static void emit_arith(struct r300_fragment_program *rp, int op,
+				GLuint dest, int mask,
+				GLuint src0, GLuint src1, GLuint src2,
+				int flags);
 
+/*
+ * Helper functions prototypes
+ */
 static int get_hw_temp(struct r300_fragment_program *rp)
 {
 	COMPILE_STATE;
@@ -256,263 +331,360 @@ static void free_hw_temp(struct r300_fragment_program *rp, int idx)
 	cs->hwreg_in_use &= ~(1<<idx);
 }
 
-static pfs_reg_t get_temp_reg(struct r300_fragment_program *rp)
+static GLuint get_temp_reg(struct r300_fragment_program *rp)
 {
 	COMPILE_STATE;
-	pfs_reg_t r = undef;
+	GLuint r = undef;
+	GLuint index;
 
-	r.index = ffs(~cs->temp_in_use);
-	if (!r.index) {
+	index = ffs(~cs->temp_in_use);
+	if (!index) {
 		ERROR("Out of program temps\n");
 		return r;
 	}
-	cs->temp_in_use |= (1 << --r.index);
-	
-	cs->temps[r.index].refcount = 0xFFFFFFFF;
-	cs->temps[r.index].reg = -1;
-	r.valid = GL_TRUE;
+
+	cs->temp_in_use |= (1 << --index);
+	cs->temps[index].refcount = 0xFFFFFFFF;
+	cs->temps[index].reg = -1;
+
+	REG_SET_TYPE(r, REG_TYPE_TEMP);
+	REG_SET_INDEX(r, index);
+	REG_SET_VALID(r, GL_TRUE);
 	return r;
 }
 
-static pfs_reg_t get_temp_reg_tex(struct r300_fragment_program *rp)
+static GLuint get_temp_reg_tex(struct r300_fragment_program *rp)
 {
 	COMPILE_STATE;
-	pfs_reg_t r = undef;
+	GLuint r = undef;
+	GLuint index;
 
-	r.index = ffs(~cs->temp_in_use);
-	if (!r.index) {
+	index = ffs(~cs->temp_in_use);
+	if (!index) {
 		ERROR("Out of program temps\n");
 		return r;
 	}
-	cs->temp_in_use |= (1 << --r.index);
-	
-	cs->temps[r.index].refcount = 0xFFFFFFFF;
-	cs->temps[r.index].reg = get_hw_temp_tex(rp);
-	r.valid = GL_TRUE;
+
+	cs->temp_in_use |= (1 << --index);
+	cs->temps[index].refcount = 0xFFFFFFFF;
+	cs->temps[index].reg = get_hw_temp_tex(rp);
+
+	REG_SET_TYPE(r, REG_TYPE_TEMP);
+	REG_SET_INDEX(r, index);
+	REG_SET_VALID(r, GL_TRUE);
 	return r;
 }
 
-static void free_temp(struct r300_fragment_program *rp, pfs_reg_t r)
+static void free_temp(struct r300_fragment_program *rp, GLuint r)
 {
 	COMPILE_STATE;
-	if (!(cs->temp_in_use & (1<<r.index))) return;
+	GLuint index = REG_GET_INDEX(r);
+
+	if (!(cs->temp_in_use & (1 << index)))
+		return;
 	
-	if (r.type == REG_TYPE_TEMP) {
-		free_hw_temp(rp, cs->temps[r.index].reg);
-		cs->temps[r.index].reg = -1;
-		cs->temp_in_use &= ~(1<<r.index);
-	} else if (r.type == REG_TYPE_INPUT) {
-		free_hw_temp(rp, cs->inputs[r.index].reg);
-		cs->inputs[r.index].reg = -1;
+	if (REG_GET_TYPE(r) == REG_TYPE_TEMP) {
+		free_hw_temp(rp, cs->temps[index].reg);
+		cs->temps[index].reg = -1;
+		cs->temp_in_use &= ~(1 << index);
+	} else if (REG_GET_TYPE(r) == REG_TYPE_INPUT) {
+		free_hw_temp(rp, cs->inputs[index].reg);
+		cs->inputs[index].reg = -1;
 	}
 }
 
-static pfs_reg_t emit_param4fv(struct r300_fragment_program *rp,
-			       GLfloat *values)
+static GLuint emit_param4fv(struct r300_fragment_program *rp,
+			    GLfloat *values)
 {
-	pfs_reg_t r = undef;
-		r.type = REG_TYPE_CONST;
+	GLuint r = undef;
+	GLuint index;
 	int pidx;
 
 	pidx = rp->param_nr++;
-	r.index = rp->const_nr++;
-	if (pidx >= PFS_NUM_CONST_REGS || r.index >= PFS_NUM_CONST_REGS) {
+	index = rp->const_nr++;
+	if (pidx >= PFS_NUM_CONST_REGS || index >= PFS_NUM_CONST_REGS) {
 		ERROR("Out of const/param slots!\n");
 		return r;
 	}
-	
-	rp->param[pidx].idx = r.index;
+
+	rp->param[pidx].idx = index;
 	rp->param[pidx].values = values;
 	rp->params_uptodate = GL_FALSE;
 
-	r.valid = GL_TRUE;
+	REG_SET_TYPE(r, REG_TYPE_CONST);
+	REG_SET_INDEX(r, index);
+	REG_SET_VALID(r, GL_TRUE);
 	return r;
 }
 
-static pfs_reg_t emit_const4fv(struct r300_fragment_program *rp, GLfloat *cp)
+static GLuint emit_const4fv(struct r300_fragment_program *rp, GLfloat *cp)
 { 
-	pfs_reg_t r = undef;
-		r.type = REG_TYPE_CONST;
+	GLuint r = undef;
+	GLuint index;
 
-	r.index = rp->const_nr++;
-	if (r.index >= PFS_NUM_CONST_REGS) {
+	index = rp->const_nr++;
+	if (index >= PFS_NUM_CONST_REGS) {
 		ERROR("Out of hw constants!\n");
 		return r;
 	}
 
-	COPY_4V(rp->constant[r.index], cp);
-	r.valid = GL_TRUE;
+	COPY_4V(rp->constant[index], cp);
+
+	REG_SET_TYPE(r, REG_TYPE_CONST);
+	REG_SET_INDEX(r, index);
+	REG_SET_VALID(r, GL_TRUE);
 	return r;
 }
 
-static __inline pfs_reg_t negate(pfs_reg_t r)
+static inline GLuint negate(GLuint r)
 {
-	r.negate_v = 1;
-	r.negate_s = 1;
+	REG_NEGS(r);
+	REG_NEGV(r);
 	return r;
 }
 
 /* Hack, to prevent clobbering sources used multiple times when
  * emulating non-native instructions
  */
-static __inline pfs_reg_t keep(pfs_reg_t r)
+static inline GLuint keep(GLuint r)
 {
-	r.no_use = GL_TRUE;
+	REG_SET_NO_USE(r, GL_TRUE);
 	return r;
 }
 
-static __inline pfs_reg_t absolute(pfs_reg_t r)
+static inline GLuint absolute(GLuint r)
 {
-	r.absolute = 1;
+	REG_ABS(r);
 	return r;
 }
 
 static int swz_native(struct r300_fragment_program *rp,
-		      pfs_reg_t src, pfs_reg_t *r, GLuint arbneg)
+		      GLuint src,
+		      GLuint *r,
+		      GLuint arbneg)
 {
-	/* Native swizzle, nothing to see here */
-	src.negate_s = (arbneg >> 3) & 1;
+	/* Native swizzle, handle negation */
+	src = (src & ~REG_NEGS_MASK) |
+		(((arbneg >> 3) & 1) << REG_NEGS_SHIFT);
 
 	if ((arbneg & 0x7) == 0x0) {
-		src.negate_v = 0;
+		src = src & ~REG_NEGV_MASK;
 		*r = src;
 	} else if ((arbneg & 0x7) == 0x7) {
-		src.negate_v = 1;
+		src |= REG_NEGV_MASK;
 		*r = src;
 	} else {
-		if (!r->valid)
+		if (!REG_GET_VALID(*r))
 			*r = get_temp_reg(rp);
-		src.negate_v = 1;
-		emit_arith(rp, PFS_OP_MAD, *r, arbneg & 0x7,
-			   keep(src), pfs_one, pfs_zero, 0);
-		src.negate_v = 0;
-		emit_arith(rp, PFS_OP_MAD, *r,
+		src |= REG_NEGV_MASK;
+		emit_arith(rp,
+			   PFS_OP_MAD,
+			   *r,
+			   arbneg & 0x7,
+			   keep(src),
+			   pfs_one,
+			   pfs_zero,
+			   0);
+		src = src & ~REG_NEGV_MASK;
+		emit_arith(rp,
+			   PFS_OP_MAD,
+			   *r,
 			   (arbneg ^ 0x7) | WRITEMASK_W,
-			   src, pfs_one, pfs_zero, 0);
+			   src,
+			   pfs_one,
+			   pfs_zero,
+			   0);
 	}
 
 	return 3;
 }
 
-static int swz_emit_partial(struct r300_fragment_program *rp, pfs_reg_t src,
-			    pfs_reg_t *r, int mask, int mc, GLuint arbneg)
+static int swz_emit_partial(struct r300_fragment_program *rp,
+			    GLuint src,
+			    GLuint *r,
+			    int mask,
+			    int mc,
+			    GLuint arbneg)
 {
 	GLuint tmp;
 	GLuint wmask = 0;
 
-	if (!r->valid)
+	if (!REG_GET_VALID(*r))
 		*r = get_temp_reg(rp);
 
-	/* A partial match, src.v_swz/mask define what parts of the
-	 * desired swizzle we match */
+	/* A partial match, VSWZ/mask define what parts of the
+	 * desired swizzle we match
+	 */
 	if (mc + s_mask[mask].count == 3) {
 		wmask = WRITEMASK_W;
-		src.negate_s = (arbneg >> 3) & 1;
+		src |= ((arbneg >> 3) & 1) << REG_NEGS_SHIFT;
 	}
 
 	tmp = arbneg & s_mask[mask].mask;
 	if (tmp) {
 		tmp = tmp ^ s_mask[mask].mask;
 		if (tmp) {
-			src.negate_v = 1;
-			emit_arith(rp, PFS_OP_MAD, *r,
+			emit_arith(rp,
+				   PFS_OP_MAD,
+				   *r,
 				   arbneg & s_mask[mask].mask,
-				   keep(src), pfs_one, pfs_zero, 0);
-			src.negate_v = 0;
-			if (!wmask) src.no_use = GL_TRUE;
-			else        src.no_use = GL_FALSE;
-			emit_arith(rp, PFS_OP_MAD, *r, tmp | wmask,
-				   src, pfs_one, pfs_zero, 0);
+				   keep(src) | REG_NEGV_MASK,
+				   pfs_one,
+				   pfs_zero,
+				   0);
+			if (!wmask) {
+				REG_SET_NO_USE(src, GL_TRUE);
+			} else {
+				REG_SET_NO_USE(src, GL_FALSE);
+			}
+			emit_arith(rp,
+				   PFS_OP_MAD,
+				   *r,
+				   tmp | wmask,
+				   src,
+				   pfs_one,
+				   pfs_zero,
+				   0);
 		} else {
-			src.negate_v = 1;
-			if (!wmask) src.no_use = GL_TRUE;
-			else        src.no_use = GL_FALSE;
-			emit_arith(rp, PFS_OP_MAD, *r,
+			if (!wmask) {
+				REG_SET_NO_USE(src, GL_TRUE);
+			} else {
+				REG_SET_NO_USE(src, GL_FALSE);
+			}
+			emit_arith(rp,
+				   PFS_OP_MAD,
+				   *r,
 				   (arbneg & s_mask[mask].mask) | wmask,
-				   src, pfs_one, pfs_zero, 0);
-			src.negate_v = 0;
+				   src | REG_NEGV_MASK,
+				   pfs_one,
+				   pfs_zero,
+				   0);
 		}
 	} else {
-		if (!wmask) src.no_use = GL_TRUE;
-		else        src.no_use = GL_FALSE;
-		emit_arith(rp, PFS_OP_MAD, *r,
+		if (!wmask) {
+			REG_SET_NO_USE(src, GL_TRUE);
+		} else {
+			REG_SET_NO_USE(src, GL_FALSE);
+		}
+		emit_arith(rp, PFS_OP_MAD,
+			   *r,
 			   s_mask[mask].mask | wmask,
-			   src, pfs_one, pfs_zero, 0);
+			   src,
+			   pfs_one,
+			   pfs_zero,
+			   0);
 	}
 
 	return s_mask[mask].count;
 }
 
-#define swizzle(r, x, y, z, w) do_swizzle(rp, r, \
-					  ((SWIZZLE_##x<<0)|	\
-					   (SWIZZLE_##y<<3)|	\
-					   (SWIZZLE_##z<<6)|	\
-					   (SWIZZLE_##w<<9)),	\
-					  0)
-
-static pfs_reg_t do_swizzle(struct r300_fragment_program *rp,
-			    pfs_reg_t src, GLuint arbswz, GLuint arbneg)
+static GLuint do_swizzle(struct r300_fragment_program *rp,
+			 GLuint src,
+			 GLuint arbswz,
+			 GLuint arbneg)
 {
-	pfs_reg_t r = undef;
-	
+	GLuint r = undef;
+	GLuint vswz;
 	int c_mask = 0;
-	int v_matched = 0;
+	int v_match = 0;
 
 	/* If swizzling from something without an XYZW native swizzle,
 	 * emit result to a temp, and do new swizzle from the temp.
 	 */
-	if (src.v_swz != SWIZZLE_XYZ || src.s_swz != SWIZZLE_W) {
-		pfs_reg_t temp = get_temp_reg(rp);
-		emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_XYZW, src, pfs_one,
-			   pfs_zero, 0);
+#if 0
+	if (REG_GET_VSWZ(src) != SWIZZLE_XYZ ||
+	    REG_GET_SSWZ(src) != SWIZZLE_W) {
+		GLuint temp = get_temp_reg(rp);
+		emit_arith(rp,
+			   PFS_OP_MAD,
+			   temp,
+			   WRITEMASK_XYZW,
+			   src,
+			   pfs_one,
+			   pfs_zero,
+			   0);
 		src = temp;
 	}
-	src.s_swz = GET_SWZ(arbswz, 3);
+#endif
+
+	if (REG_GET_VSWZ(src) != SWIZZLE_XYZ ||
+	    REG_GET_SSWZ(src) != SWIZZLE_W) {
+	    GLuint vsrcswz = (v_swiz[REG_GET_VSWZ(src)].hash & (SWZ_X_MASK|SWZ_Y_MASK|SWZ_Z_MASK)) | REG_GET_SSWZ(src) << 9;
+	    GLint i;
+
+	    GLuint newswz = 0;
+	    GLuint offset;
+	    for(i=0; i < 4; ++i){
+		offset = GET_SWZ(arbswz, i);
+		
+		newswz |= (offset <= 3)?GET_SWZ(vsrcswz, offset) << i*3:offset << i*3;
+	    }
+
+	    arbswz = newswz & (SWZ_X_MASK|SWZ_Y_MASK|SWZ_Z_MASK);
+	    REG_SET_SSWZ(src, GET_SWZ(newswz, 3));
+	}
+	else
+	{
+	    /* set scalar swizzling */
+	    REG_SET_SSWZ(src, GET_SWZ(arbswz, 3));
 
+	}
 	do {
+		vswz = REG_GET_VSWZ(src);
 		do {
-#define CUR_HASH (v_swiz[src.v_swz].hash & s_mask[c_mask].hash)
-			if (CUR_HASH == (arbswz & s_mask[c_mask].hash)) {
-				if (s_mask[c_mask].count == 3)
-					v_matched += swz_native(rp, src, &r,
+			int chash;
+
+			REG_SET_VSWZ(src, vswz);
+			chash = v_swiz[REG_GET_VSWZ(src)].hash &
+				s_mask[c_mask].hash;
+
+			if (chash == (arbswz & s_mask[c_mask].hash)) {
+				if (s_mask[c_mask].count == 3) {
+					v_match += swz_native(rp,
+								src,
+								&r,
 								arbneg);
-				else
-					v_matched += swz_emit_partial(rp, src,
-								      &r,
-								      c_mask,
-								      v_matched,
-								      arbneg);
-
-				if (v_matched == 3)
+				} else {
+					v_match += swz_emit_partial(rp,
+								    src,
+								    &r,
+								    c_mask,
+								    v_match,
+								    arbneg);
+				}
+
+				if (v_match == 3)
 					return r;
 
 				/* Fill with something invalid.. all 0's was
 				 * wrong before, matched SWIZZLE_X.  So all
-				 * 1's will be okay for now */
+				 * 1's will be okay for now
+				 */
 				arbswz |= (PFS_INVAL & s_mask[c_mask].hash);
 			}
-		} while(v_swiz[++src.v_swz].hash != PFS_INVAL);
-		src.v_swz = SWIZZLE_XYZ;
+		} while(v_swiz[++vswz].hash != PFS_INVAL);
+		REG_SET_VSWZ(src, SWIZZLE_XYZ);
 	} while (s_mask[++c_mask].hash != PFS_INVAL);
 
 	ERROR("should NEVER get here\n");
 	return r;
 }
-				
-static pfs_reg_t t_src(struct r300_fragment_program *rp,
-		       struct prog_src_register fpsrc)
+
+static GLuint t_src(struct r300_fragment_program *rp,
+		    struct prog_src_register fpsrc)
 {
-	pfs_reg_t r = undef;
+	GLuint r = undef;
 
 	switch (fpsrc.File) {
 	case PROGRAM_TEMPORARY:
-		r.index = fpsrc.Index;
-		r.valid = GL_TRUE;
+		REG_SET_INDEX(r, fpsrc.Index);
+		REG_SET_VALID(r, GL_TRUE);
+		REG_SET_TYPE(r, REG_TYPE_TEMP);
 		break;
 	case PROGRAM_INPUT:
-		r.index = fpsrc.Index;
-		r.type = REG_TYPE_INPUT;
-		r.valid = GL_TRUE;
+		REG_SET_INDEX(r, fpsrc.Index);
+		REG_SET_VALID(r, GL_TRUE);
+		REG_SET_TYPE(r, REG_TYPE_INPUT);
 		break;
 	case PROGRAM_LOCAL_PARAM:
 		r = emit_param4fv(rp,
@@ -533,13 +705,13 @@ static pfs_reg_t t_src(struct r300_fragment_program *rp,
 	}
 
 	/* no point swizzling ONE/ZERO/HALF constants... */
-	if (r.v_swz < SWIZZLE_111 || r.s_swz < SWIZZLE_ZERO)
+	if (REG_GET_VSWZ(r) < SWIZZLE_111 || REG_GET_SSWZ(r) < SWIZZLE_ZERO)
 		r = do_swizzle(rp, r, fpsrc.Swizzle, fpsrc.NegateBase);
 	return r;
 }
 
-static pfs_reg_t t_scalar_src(struct r300_fragment_program *rp,
-			      struct prog_src_register fpsrc)
+static GLuint t_scalar_src(struct r300_fragment_program *rp,
+			   struct prog_src_register fpsrc)
 {
 	struct prog_src_register src = fpsrc;
 	int sc = GET_SWZ(fpsrc.Swizzle, 0); /* X */
@@ -549,22 +721,24 @@ static pfs_reg_t t_scalar_src(struct r300_fragment_program *rp,
 	return t_src(rp, src);
 }
 
-static pfs_reg_t t_dst(struct r300_fragment_program *rp,
-		       struct prog_dst_register dest) {
-	pfs_reg_t r = undef;
+static GLuint t_dst(struct r300_fragment_program *rp,
+		       struct prog_dst_register dest)
+{
+	GLuint r = undef;
 	
 	switch (dest.File) {
 	case PROGRAM_TEMPORARY:
-		r.index = dest.Index;
-		r.valid = GL_TRUE;
+		REG_SET_INDEX(r, dest.Index);
+		REG_SET_VALID(r, GL_TRUE);
+		REG_SET_TYPE(r, REG_TYPE_TEMP);
 		return r;
 	case PROGRAM_OUTPUT:
-		r.type = REG_TYPE_OUTPUT;
+		REG_SET_TYPE(r, REG_TYPE_OUTPUT);
 		switch (dest.Index) {
 		case FRAG_RESULT_COLR:
 		case FRAG_RESULT_DEPR:
-			r.index = dest.Index;
-			r.valid = GL_TRUE;
+			REG_SET_INDEX(r, dest.Index);
+			REG_SET_VALID(r, GL_TRUE);
 			return r;
 		default:
 			ERROR("Bad DstReg->Index 0x%x\n", dest.Index);
@@ -576,66 +750,77 @@ static pfs_reg_t t_dst(struct r300_fragment_program *rp,
 	}
 }
 
-static int t_hw_src(struct r300_fragment_program *rp, pfs_reg_t src,
+static int t_hw_src(struct r300_fragment_program *rp,
+		    GLuint src,
 		    GLboolean tex)
 {
 	COMPILE_STATE;
 	int idx;
+	int index = REG_GET_INDEX(src);
 
-	switch (src.type) {
+	switch(REG_GET_TYPE(src)) {
 	case REG_TYPE_TEMP:
 		/* NOTE: if reg==-1 here, a source is being read that
-		 * 	 hasn't been written to. Undefined results */
-		if (cs->temps[src.index].reg == -1)
-			cs->temps[src.index].reg = get_hw_temp(rp);
-		idx = cs->temps[src.index].reg;
+		 * 	 hasn't been written to. Undefined results
+		 */
+		if (cs->temps[index].reg == -1)
+			cs->temps[index].reg = get_hw_temp(rp);
+
+		idx = cs->temps[index].reg;
 
-		if (!src.no_use && (--cs->temps[src.index].refcount == 0))
+		if (!REG_GET_NO_USE(src) &&
+		    (--cs->temps[index].refcount == 0))
 			free_temp(rp, src);
 		break;
 	case REG_TYPE_INPUT:
-		idx = cs->inputs[src.index].reg;
+		idx = cs->inputs[index].reg;
 
-		if (!src.no_use && (--cs->inputs[src.index].refcount == 0))
-			free_hw_temp(rp, cs->inputs[src.index].reg);
+		if (!REG_GET_NO_USE(src) &&
+		    (--cs->inputs[index].refcount == 0))
+			free_hw_temp(rp, cs->inputs[index].reg);
 		break;
 	case REG_TYPE_CONST:
-		return (src.index | SRC_CONST);
+		return (index | SRC_CONST);
 	default:
 		ERROR("Invalid type for source reg\n");
 		return (0 | SRC_CONST);
 	}
 
-	if (!tex) cs->used_in_node |= (1 << idx);
+	if (!tex)
+		cs->used_in_node |= (1 << idx);
 
 	return idx;
 }
 
-static int t_hw_dst(struct r300_fragment_program *rp, pfs_reg_t dest,
+static int t_hw_dst(struct r300_fragment_program *rp,
+		    GLuint dest,
 		    GLboolean tex)
 {
 	COMPILE_STATE;
 	int idx;
-	assert(dest.valid);
+	GLuint index = REG_GET_INDEX(dest);
+	assert(REG_GET_VALID(dest));
 
-	switch (dest.type) {
+	switch(REG_GET_TYPE(dest)) {
 	case REG_TYPE_TEMP:
-		if (cs->temps[dest.index].reg == -1) {
-			if (!tex)
-				cs->temps[dest.index].reg = get_hw_temp(rp);
-			else
-				cs->temps[dest.index].reg = get_hw_temp_tex(rp);
+		if (cs->temps[REG_GET_INDEX(dest)].reg == -1) {
+			if (!tex) {
+				cs->temps[index].reg = get_hw_temp(rp);
+			} else {
+				cs->temps[index].reg = get_hw_temp_tex(rp);
+			}
 		}
-		idx = cs->temps[dest.index].reg;
+		idx = cs->temps[index].reg;
 
-		if (!dest.no_use && (--cs->temps[dest.index].refcount == 0))
+		if (!REG_GET_NO_USE(dest) &&
+		    (--cs->temps[index].refcount == 0))
 			free_temp(rp, dest);
 
 		cs->dest_in_node |= (1 << idx);
 		cs->used_in_node |= (1 << idx);
 		break;
 	case REG_TYPE_OUTPUT:
-		switch (dest.index) {
+		switch(index) {
 		case FRAG_RESULT_COLR:
 			rp->node[rp->cur_node].flags |= R300_PFS_NODE_OUTPUT_COLOR;
 			break;
@@ -643,17 +828,18 @@ static int t_hw_dst(struct r300_fragment_program *rp, pfs_reg_t dest,
 			rp->node[rp->cur_node].flags |= R300_PFS_NODE_OUTPUT_DEPTH;
 			break;
 		}
-		return dest.index;
+		return index;
 		break;
 	default:
-		ERROR("invalid dest reg type %d\n", dest.type);
+		ERROR("invalid dest reg type %d\n", REG_GET_TYPE(dest));
 		return 0;
 	}
 	
 	return idx;
 }
 
-static void emit_nop(struct r300_fragment_program *rp, GLuint mask,
+static void emit_nop(struct r300_fragment_program *rp,
+		     GLuint mask,
 		     GLboolean sync)
 {
 	COMPILE_STATE;
@@ -679,8 +865,8 @@ static void emit_tex(struct r300_fragment_program *rp,
 		     int opcode)
 {
 	COMPILE_STATE;
-	pfs_reg_t coord = t_src(rp, fpi->SrcReg[0]);
-	pfs_reg_t dest = undef, rdest = undef;
+	GLuint coord = t_src(rp, fpi->SrcReg[0]);
+	GLuint dest = undef, rdest = undef;
 	GLuint din = cs->dest_in_node, uin = cs->used_in_node;
 	int unit = fpi->TexSrcUnit;
 	int hwsrc, hwdest;
@@ -691,7 +877,7 @@ static void emit_tex(struct r300_fragment_program *rp,
 		dest = t_dst(rp, fpi->DstReg);
 
 		/* r300 doesn't seem to be able to do TEX->output reg */
-		if (dest.type == REG_TYPE_OUTPUT) {
+		if (REG_GET_TYPE(dest) == REG_TYPE_OUTPUT) {
 			rdest = dest;
 			dest = get_temp_reg_tex(rp);
 		}
@@ -703,7 +889,7 @@ static void emit_tex(struct r300_fragment_program *rp,
 		if (uin & (1 << hwdest)) {
 			free_hw_temp(rp, hwdest);
 			hwdest = get_hw_temp_tex(rp);
-			cs->temps[dest.index].reg = hwdest;
+			cs->temps[REG_GET_INDEX(dest)].reg = hwdest;
 		}
 	} else {
 		hwdest = 0;
@@ -713,8 +899,8 @@ static void emit_tex(struct r300_fragment_program *rp,
 	/* Indirection if source has been written in this node, or if the
 	 * dest has been read/written in this node
 	 */
-	if ((coord.type != REG_TYPE_CONST && (din & (1<<hwsrc))) ||
-					(uin & (1<<hwdest))) {
+	if ((REG_GET_TYPE(coord) != REG_TYPE_CONST &&
+	     (din & (1<<hwsrc))) || (uin & (1<<hwdest))) {
 			
 		/* Finish off current node */
 		cs->v_pos = cs->s_pos = MAX2(cs->v_pos, cs->s_pos);
@@ -754,13 +940,13 @@ static void emit_tex(struct r300_fragment_program *rp,
 		| (opcode << R300_FPITX_OPCODE_SHIFT);
 
 	cs->dest_in_node |= (1 << hwdest); 
-	if (coord.type != REG_TYPE_CONST)
+	if (REG_GET_TYPE(coord) != REG_TYPE_CONST)
 		cs->used_in_node |= (1 << hwsrc);
 
 	rp->node[rp->cur_node].tex_end++;
 
 	/* Copy from temp to output if needed */
-	if (rdest.valid) {
+	if (REG_GET_VALID(rdest)) {
 		emit_arith(rp, PFS_OP_MAD, rdest, WRITEMASK_XYZW, dest,
 			   pfs_one, pfs_zero, 0);
 		free_temp(rp, dest);
@@ -770,7 +956,9 @@ static void emit_tex(struct r300_fragment_program *rp,
 /* Add sources to FPI1/FPI3 lists.  If source is already on list,
  * reuse the index instead of wasting a source.
  */
-static int add_src(struct r300_fragment_program *rp, int reg, int pos,
+static int add_src(struct r300_fragment_program *rp,
+		   int reg,
+		   int pos,
 		   int srcmask)
 {
 	COMPILE_STATE;
@@ -819,9 +1007,12 @@ static int add_src(struct r300_fragment_program *rp, int reg, int pos,
  * It's not necessary to force the first case, but it makes disassembled
  * shaders easier to read.
  */
-static GLboolean force_same_slot(int vop, int sop,
-				 GLboolean emit_vop, GLboolean emit_sop,
-				 int argc, pfs_reg_t *src)
+static GLboolean force_same_slot(int vop,
+				 int sop,
+				 GLboolean emit_vop,
+				 GLboolean emit_sop,
+				 int argc,
+				 GLuint *src)
 {
 	int i;
 
@@ -833,20 +1024,24 @@ static GLboolean force_same_slot(int vop, int sop,
 
 	if (emit_vop) {
 		for (i=0;i<argc;i++)
-			if (src[i].v_swz == SWIZZLE_WZY)
+			if (REG_GET_VSWZ(src[i]) == SWIZZLE_WZY)
 				return GL_TRUE;
 	}
 
 	return GL_FALSE;
 }
 
-static void emit_arith(struct r300_fragment_program *rp, int op,
-		       pfs_reg_t dest, int mask,
-		       pfs_reg_t src0, pfs_reg_t src1, pfs_reg_t src2,
+static void emit_arith(struct r300_fragment_program *rp,
+		       int op,
+		       GLuint dest,
+		       int mask,
+		       GLuint src0,
+		       GLuint src1,
+		       GLuint src2,
 		       int flags)
 {
 	COMPILE_STATE;
-	pfs_reg_t src[3] = { src0, src1, src2 };
+	GLuint src[3] = { src0, src1, src2 };
 	int hwsrc[3], sswz[3], vswz[3];
 	int hwdest;
 	GLboolean emit_vop = GL_FALSE, emit_sop = GL_FALSE;
@@ -863,7 +1058,8 @@ static void emit_arith(struct r300_fragment_program *rp, int op,
 	if ((mask & WRITEMASK_W) || vop == R300_FPI0_OUTC_REPL_ALPHA)
 		emit_sop = GL_TRUE;
 
-	if (dest.type == REG_TYPE_OUTPUT && dest.index == FRAG_RESULT_DEPR)
+	if (REG_GET_TYPE(dest) == REG_TYPE_OUTPUT &&
+	    REG_GET_INDEX(dest) == FRAG_RESULT_DEPR)
 		emit_vop = GL_FALSE;
 					
 	if (force_same_slot(vop, sop, emit_vop, emit_sop, argc, src)) {
@@ -879,12 +1075,12 @@ static void emit_arith(struct r300_fragment_program *rp, int op,
 		 */
 		for (i=0;i<3;i++) {
 			if (emit_vop &&
-			    (v_swiz[src[i].v_swz].flags & SLOT_SCALAR)) {
+			    (v_swiz[REG_GET_VSWZ(src[i])].flags & SLOT_SCALAR)) {
 				vpos = spos = MAX2(vpos, spos);
 				break;
 			}
 			if (emit_sop &&
-			    (s_swiz[src[i].s_swz].flags & SLOT_VECTOR)) {
+			    (s_swiz[REG_GET_SSWZ(src[i])].flags & SLOT_VECTOR)) {
 				vpos = spos = MAX2(vpos, spos);
 				break;
 			}
@@ -908,20 +1104,22 @@ static void emit_arith(struct r300_fragment_program *rp, int op,
 
 		if (emit_vop && vop != R300_FPI0_OUTC_REPL_ALPHA) {
 			srcpos = add_src(rp, hwsrc[i], vpos,
-					 v_swiz[src[i].v_swz].flags);	
-			vswz[i] = (v_swiz[src[i].v_swz].base +
-				   (srcpos * v_swiz[src[i].v_swz].stride)) |
-				(src[i].negate_v ? ARG_NEG : 0) |
-				(src[i].absolute ? ARG_ABS : 0);
+					 v_swiz[REG_GET_VSWZ(src[i])].flags);
+			vswz[i] = (v_swiz[REG_GET_VSWZ(src[i])].base +
+				   (srcpos *
+				    v_swiz[REG_GET_VSWZ(src[i])].stride)) |
+				((src[i] & REG_NEGV_MASK) ? ARG_NEG : 0) |
+				((src[i] & REG_ABS_MASK) ? ARG_ABS : 0);
 		} else vswz[i] = R300_FPI0_ARGC_ZERO;
 		
 		if (emit_sop) {
 			srcpos = add_src(rp, hwsrc[i], spos,
-					 s_swiz[src[i].s_swz].flags);
-			sswz[i] = (s_swiz[src[i].s_swz].base +
-				   (srcpos * s_swiz[src[i].s_swz].stride)) |
-				(src[i].negate_s ? ARG_NEG : 0) |
-				(src[i].absolute ? ARG_ABS : 0);	
+					 s_swiz[REG_GET_SSWZ(src[i])].flags);
+			sswz[i] = (s_swiz[REG_GET_SSWZ(src[i])].base +
+				   (srcpos *
+				    s_swiz[REG_GET_SSWZ(src[i])].stride)) |
+				((src[i] & REG_NEGS_MASK) ? ARG_NEG : 0) |
+				((src[i] & REG_ABS_MASK) ? ARG_ABS : 0);
 		} else sswz[i] = R300_FPI2_ARGA_ZERO;
 	}
 	hwdest = t_hw_dst(rp, dest, GL_FALSE);
@@ -943,8 +1141,8 @@ static void emit_arith(struct r300_fragment_program *rp, int op,
 				(vswz[2] << R300_FPI0_ARG2C_SHIFT);
 
 		rp->alu.inst[vpos].inst1 |= hwdest << R300_FPI1_DSTC_SHIFT;
-		if (dest.type == REG_TYPE_OUTPUT) {
-			if (dest.index == FRAG_RESULT_COLR) {
+		if (REG_GET_TYPE(dest) == REG_TYPE_OUTPUT) {
+			if (REG_GET_INDEX(dest) == FRAG_RESULT_COLR) {
 				rp->alu.inst[vpos].inst1 |=
 					(mask & WRITEMASK_XYZ) << R300_FPI1_DSTC_OUTPUT_MASK_SHIFT;
 			} else assert(0);
@@ -968,11 +1166,11 @@ static void emit_arith(struct r300_fragment_program *rp, int op,
 				sswz[2] << R300_FPI2_ARG2A_SHIFT;
 
 		if (mask & WRITEMASK_W) {
-			if (dest.type == REG_TYPE_OUTPUT) {
-				if (dest.index == FRAG_RESULT_COLR) {
+			if (REG_GET_TYPE(dest) == REG_TYPE_OUTPUT) {
+				if (REG_GET_INDEX(dest) == FRAG_RESULT_COLR) {
 					rp->alu.inst[spos].inst3 |= 
 							(hwdest << R300_FPI3_DSTA_SHIFT) | R300_FPI3_DSTA_OUTPUT;
-				} else if (dest.index == FRAG_RESULT_DEPR) {
+				} else if (REG_GET_INDEX(dest) == FRAG_RESULT_DEPR) {
 					rp->alu.inst[spos].inst3 |= R300_FPI3_DSTA_DEPTH;
 				} else assert(0);
 			} else {
@@ -985,33 +1183,52 @@ static void emit_arith(struct r300_fragment_program *rp, int op,
 		rp->alu.inst[vpos].inst2 = NOP_INST2;
 
 	return;
-};
+}
 
 #if 0
-static pfs_reg_t get_attrib(struct r300_fragment_program *rp, GLuint attr)
+static GLuint get_attrib(struct r300_fragment_program *rp, GLuint attr)
 {
 	struct gl_fragment_program *mp = &rp->mesa_program;
-	pfs_reg_t r = undef;
+	GLuint r = undef;
 
 	if (!(mp->Base.InputsRead & (1<<attr))) {
 		ERROR("Attribute %d was not provided!\n", attr);
 		return undef;
 	}
 
-	r.type  = REG_TYPE_INPUT;
-	r.index = attr;
-	r.valid = GL_TRUE;
+	REG_SET_TYPE(r, REG_TYPE_INPUT);
+	REG_SET_INDEX(r, attr);
+	REG_SET_VALID(r, GL_TRUE);
 	return r;
 }
 #endif
 
+static void make_sin_const(struct r300_fragment_program *rp)
+{
+	if(rp->const_sin[0] == -1){
+	    GLfloat cnstv[4];
+
+	    cnstv[0] = 1.273239545; // 4/PI
+	    cnstv[1] =-0.405284735; // -4/(PI*PI)
+	    cnstv[2] = 3.141592654; // PI
+	    cnstv[3] = 0.2225;      // weight
+	    rp->const_sin[0] = emit_const4fv(rp, cnstv);
+
+	    cnstv[0] = 0.5;
+	    cnstv[1] = -1.5;
+	    cnstv[2] = 0.159154943; // 1/(2*PI)
+	    cnstv[3] = 6.283185307; // 2*PI
+	    rp->const_sin[1] = emit_const4fv(rp, cnstv);
+	}
+}
+
 static GLboolean parse_program(struct r300_fragment_program *rp)
 {	
 	struct gl_fragment_program *mp = &rp->mesa_program;
 	const struct prog_instruction *inst = mp->Base.Instructions;
 	struct prog_instruction *fpi;
-	pfs_reg_t src[3], dest, temp;
-	pfs_reg_t cnst;
+	GLuint src[3], dest, temp;
+	GLuint cnst;
 	int flags, mask = 0;
 	GLfloat cnstv[4] = {0.0, 0.0, 0.0, 0.0};
 
@@ -1058,62 +1275,71 @@ static GLboolean parse_program(struct r300_fragment_program *rp)
 			break;
 		case OPCODE_COS:
 			/*
-			 * cos using taylor serie:
-			 * cos(x) = 1 - x^2/2! + x^4/4! - x^6/6!
+			 * cos using a parabola (see SIN):
+			 * cos(x):
+			 *   x += PI/2
+			 *   x = (x/(2*PI))+0.5
+			 *   x = frac(x)
+			 *   x = (x*2*PI)-PI
+			 *   result = sin(x)
 			 */
 			temp = get_temp_reg(rp);
-			cnstv[0] = 0.5;
-			cnstv[1] = 0.041666667;
-			cnstv[2] = 0.001388889;
-			cnstv[4] = 0.0;
-			cnst = emit_const4fv(rp, cnstv);
+			make_sin_const(rp);
 			src[0] = t_scalar_src(rp, fpi->SrcReg[0]);
 
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_XYZ,
-				   src[0],
-				   src[0],
-				   pfs_zero,
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_Y | WRITEMASK_Z,
-				   temp, temp,
-				   pfs_zero,
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_Z,
-				   temp,
-				   swizzle(temp, X, X, X, W),
-				   pfs_zero,
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_XYZ,
-				   temp, cnst,
-				   pfs_zero,
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_X,
-				   pfs_one,
-				   pfs_one,
-				   negate(temp),
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_X,
-				   temp,
-				   pfs_one,
-				   swizzle(temp, Y, Y, Y, W),
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_X,
-				   temp,
-				   pfs_one,
-				   negate(swizzle(temp, Z, Z, Z, W)),
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, dest, mask,
+			/* add 0.5*PI and do range reduction */
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X,
+				   swizzle(rp->const_sin[0], Z, Z, Z, Z), //PI
+				   pfs_half,
+				   swizzle(keep(src[0]), X, X, X, X),
+				   0);
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X,
 				   swizzle(temp, X, X, X, X),
-				   pfs_one,
+				   swizzle(rp->const_sin[1], Z, Z, Z, Z),
+				   pfs_half,
+				   0);
+
+			emit_arith(rp, PFS_OP_FRC, temp, WRITEMASK_X,
+				   swizzle(temp, X, X, X, X),
+				   undef,
+				   undef,
+				   0);
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_Z,
+				   swizzle(temp, X, X, X, X),
+				   swizzle(rp->const_sin[1], W, W, W, W), //2*PI
+				   negate(swizzle(rp->const_sin[0], Z, Z, Z, Z)), //-PI
+				   0);
+
+			/* SIN */
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X | WRITEMASK_Y,
+				   swizzle(temp, Z, Z, Z, Z),
+				   rp->const_sin[0],
 				   pfs_zero,
+				   0);
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X,
+				   swizzle(temp, Y, Y, Y, Y),
+				   absolute(swizzle(temp, Z, Z, Z, Z)),
+				   swizzle(temp, X, X, X, X),
+				   0);
+			
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_Y,
+				   swizzle(temp, X, X, X, X),
+				   absolute(swizzle(temp, X, X, X, X)),
+				   negate(swizzle(temp, X, X, X, X)),
+				   0);
+
+
+	    		emit_arith(rp, PFS_OP_MAD, dest, mask,
+				   swizzle(temp, Y, Y, Y, Y),
+				   swizzle(rp->const_sin[0], W, W, W, W),
+				   swizzle(temp, X, X, X, X),
 				   flags);
+
 			free_temp(rp, temp);
 			break;
 		case OPCODE_DP3:
@@ -1167,7 +1393,7 @@ static GLboolean parse_program(struct r300_fragment_program *rp)
 			/* result.x = 1.0
 			 * result.w = src1.w */
 			if (mask & WRITEMASK_XW) {
-				src[1].v_swz = SWIZZLE_111; /* Cheat.. */
+				REG_SET_VSWZ(src[1], SWIZZLE_111); /*Cheat*/
 				emit_arith(rp, PFS_OP_MAD, dest,
 					   mask & WRITEMASK_XW,
 					   src[1], pfs_one, pfs_zero,
@@ -1222,7 +1448,7 @@ static GLboolean parse_program(struct r300_fragment_program *rp)
 			 * change the compare to (t.x + 0.5) > 0.5 we may
 			 * save one instruction by doing CMP -t.x 
 			 */
-			cnstv[0] = cnstv[1] = cnstv[2] = cnstv[4] = 0.50001;
+			cnstv[0] = cnstv[1] = cnstv[2] = cnstv[3] = 0.50001;
 			src[0] = t_src(rp, fpi->SrcReg[0]);
 			temp = get_temp_reg(rp);
 			cnst = emit_const4fv(rp, cnstv);
@@ -1353,7 +1579,93 @@ static GLboolean parse_program(struct r300_fragment_program *rp)
 				   flags);
 			break;
 		case OPCODE_SCS:
-			ERROR("SCS not implemented\n");
+			/*
+			 * cos using a parabola (see SIN):
+			 * cos(x):
+			 *   x += PI/2
+			 *   x = (x/(2*PI))+0.5
+			 *   x = frac(x)
+			 *   x = (x*2*PI)-PI
+			 *   result = sin(x)
+			 */
+			temp = get_temp_reg(rp);
+			make_sin_const(rp);
+			src[0] = t_scalar_src(rp, fpi->SrcReg[0]);
+
+			/* add 0.5*PI and do range reduction */
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X|WRITEMASK_Y,
+				   swizzle(rp->const_sin[0], Z, Z, Z, Z),
+				   rp->const_sin[1],
+				   swizzle(keep(src[0]), X, X, X, X),
+				   0);
+
+			emit_arith(rp, PFS_OP_CMP, temp, WRITEMASK_W,
+				   swizzle(rp->const_sin[0], Z, Z, Z, Z),
+				   negate(pfs_half),
+				   swizzle(keep(src[0]), X, X, X, X),
+				   0);
+
+			emit_arith(rp, PFS_OP_CMP, temp, WRITEMASK_Z,
+				   swizzle(temp, X, X, X, X),
+				   swizzle(temp, Y, Y, Y, Y),
+				   swizzle(temp, W, W, W, W),
+				   0);
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X | WRITEMASK_Y,
+			           swizzle(temp, Z, Z, Z, Z),
+				   rp->const_sin[0],
+			           pfs_zero,
+				   0);
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_W,
+				   swizzle(temp, Y, Y, Y, Y),
+				   absolute(swizzle(temp, Z, Z, Z, Z)),
+				   swizzle(temp, X, X, X, X),
+				   0);
+
+			if(mask & WRITEMASK_Y)
+			{
+			    emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X | WRITEMASK_Y,
+				       swizzle(keep(src[0]), X, X, X, X),
+				       rp->const_sin[0],
+				       pfs_zero,
+				       0);
+
+			    emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X,
+				       swizzle(temp, Y, Y, Y, Y),
+				       absolute(swizzle(keep(src[0]), X, X, X, X)),
+				       swizzle(temp, X, X, X, X),
+				       0);
+			}
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_Z,
+				   swizzle(temp, W, W, W, W),
+				   absolute(swizzle(temp, W, W, W, W)),
+				   negate(swizzle(temp, W, W, W, W)),
+				   0);
+
+			emit_arith(rp, PFS_OP_MAD, dest, WRITEMASK_X,
+				   swizzle(temp, Z, Z, Z, Z),
+				   swizzle(rp->const_sin[0], W, W, W, W),
+				   swizzle(temp, W, W, W, W),
+				   flags);
+
+			if(mask & WRITEMASK_Y)
+			{
+			    emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_W,
+				       swizzle(temp, X, X, X, X),
+				       absolute(swizzle(temp, X, X, X, X)),
+				       negate(swizzle(temp, X, X, X, X)),
+				       0);
+
+	    		    emit_arith(rp, PFS_OP_MAD, dest, WRITEMASK_Y,
+				       swizzle(temp, W, W, W, W),
+				       swizzle(rp->const_sin[0], W, W, W, W),
+				       swizzle(temp, X, X, X, X),
+				       flags);
+			}
+			free_temp(rp, temp);
 			break;
 		case OPCODE_SGE:
 			src[0] = t_src(rp, fpi->SrcReg[0]);
@@ -1372,68 +1684,63 @@ static GLboolean parse_program(struct r300_fragment_program *rp)
 			break;
 		case OPCODE_SIN:
 			/*
-			 * sin using taylor serie:
-			 * sin(x) = x - x^3/3! + x^5/5! - x^7/7!
+			 *  using a parabola:
+			 * sin(x) = 4/pi * x + -4/(pi*pi) * x * abs(x)
+			 * extra precision is obtained by weighting against
+			 * itself squared.
 			 */
+
 			temp = get_temp_reg(rp);
-			cnstv[0] = 0.333333333;
-			cnstv[1] = 0.008333333;
-			cnstv[2] = 0.000198413;
-			cnstv[4] = 0.0;
-			cnst = emit_const4fv(rp, cnstv);
+			make_sin_const(rp);
 			src[0] = t_scalar_src(rp, fpi->SrcReg[0]);
 
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_XYZ,
-				   src[0],
-				   src[0],
-				   pfs_zero,
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_Y | WRITEMASK_Z,
-				   temp, temp,
-				   pfs_zero,
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_Z,
-				   temp,
-				   swizzle(temp, X, X, X, W),
-				   pfs_zero,
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_XYZ,
-				   src[0],
-				   temp,
-				   pfs_zero,
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_XYZ,
-				   temp, cnst,
-				   pfs_zero,
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_X,
-				   src[0],
-				   pfs_one,
-				   negate(temp),
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_X,
-				   temp,
-				   pfs_one,
-				   swizzle(temp, Y, Y, Y, W),
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, temp,
-				   WRITEMASK_X,
-				   temp,
-				   pfs_one,
-				   negate(swizzle(temp, Z, Z, Z, W)),
-				   flags);
-			emit_arith(rp, PFS_OP_MAD, dest, mask,
+			/* do range reduction */
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X,
+				   swizzle(keep(src[0]), X, X, X, X),
+				   swizzle(rp->const_sin[1], Z, Z, Z, Z),
+				   pfs_half,
+				   0);
+
+			emit_arith(rp, PFS_OP_FRC, temp, WRITEMASK_X,
 				   swizzle(temp, X, X, X, X),
-				   pfs_one,
+				   undef,
+				   undef,
+				   0);
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_Z,
+				   swizzle(temp, X, X, X, X),
+				   swizzle(rp->const_sin[1], W, W, W, W), //2*PI
+				   negate(swizzle(rp->const_sin[0], Z, Z, Z, Z)), //PI
+				   0);
+
+			/* SIN */
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X | WRITEMASK_Y,
+				   swizzle(temp, Z, Z, Z, Z),
+				   rp->const_sin[0],
 				   pfs_zero,
+				   0);
+
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_X,
+				   swizzle(temp, Y, Y, Y, Y),
+				   absolute(swizzle(temp, Z, Z, Z, Z)),
+				   swizzle(temp, X, X, X, X),
+				   0);
+			
+			emit_arith(rp, PFS_OP_MAD, temp, WRITEMASK_Y,
+				   swizzle(temp, X, X, X, X),
+				   absolute(swizzle(temp, X, X, X, X)),
+				   negate(swizzle(temp, X, X, X, X)),
+				   0);
+
+
+	    		emit_arith(rp, PFS_OP_MAD, dest, mask,
+				   swizzle(temp, Y, Y, Y, Y),
+				   swizzle(rp->const_sin[0], W, W, W, W),
+				   swizzle(temp, X, X, X, X),
 				   flags);
+
 			free_temp(rp, temp);
 			break;
 		case OPCODE_SLT:
@@ -1505,7 +1812,7 @@ static GLboolean parse_program(struct r300_fragment_program *rp)
 /* - Init structures
  * - Determine what hwregs each input corresponds to
  */
-static void init_program(struct r300_fragment_program *rp)
+static void init_program(r300ContextPtr r300, struct r300_fragment_program *rp)
 {
 	struct r300_pfs_compile_state *cs = NULL;
 	struct gl_fragment_program *mp = &rp->mesa_program;	
@@ -1515,6 +1822,7 @@ static void init_program(struct r300_fragment_program *rp)
 	int i,j;
 
 	/* New compile, reset tracking data */
+	rp->optimization = driQueryOptioni(&r300->radeon.optionCache, "fp_optimization");
 	rp->translated = GL_FALSE;
 	rp->error      = GL_FALSE;
 	rp->cs = cs	   = &(R300_CONTEXT(rp->ctx)->state.pfs_compile);
@@ -1527,6 +1835,7 @@ static void init_program(struct r300_fragment_program *rp)
 	rp->max_temp_idx = 0;
 	rp->node[0].alu_end = -1;
 	rp->node[0].tex_end = -1;
+	rp->const_sin[0] = -1;
 	
 	_mesa_memset(cs, 0, sizeof(*rp->cs));
 	for (i=0;i<PFS_MAX_ALU_INST;i++) {
@@ -1640,13 +1949,13 @@ static void update_params(struct r300_fragment_program *rp)
 	rp->params_uptodate = GL_TRUE;
 }
 
-void r300_translate_fragment_shader(struct r300_fragment_program *rp)
+void r300_translate_fragment_shader(r300ContextPtr r300, struct r300_fragment_program *rp)
 {
 	struct r300_pfs_compile_state *cs = NULL;
 
 	if (!rp->translated) {
 		
-		init_program(rp);
+		init_program(r300, rp);
 		cs = rp->cs;
 
 		if (parse_program(rp) == GL_FALSE) {
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog.h b/src/mesa/drivers/dri/r300/r300_fragprog.h
index 4bbaa07e012..73986abc3cc 100644
--- a/src/mesa/drivers/dri/r300/r300_fragprog.h
+++ b/src/mesa/drivers/dri/r300/r300_fragprog.h
@@ -41,6 +41,7 @@
 #include "r300_context.h"
 #include "program_instruction.h"
 
+#if 0
 /* representation of a register for emit_arith/swizzle */
 typedef struct _pfs_reg_t {
 	enum {
@@ -58,7 +59,7 @@ typedef struct _pfs_reg_t {
 	GLboolean no_use:1;
 	GLboolean valid:1;
 } pfs_reg_t;
-
+#endif
 typedef struct r300_fragment_program_swizzle {
 	GLuint length;
 	GLuint src[4];
@@ -111,8 +112,11 @@ typedef struct r300_fragment_program_swizzle {
 		((0 | SRC_CONST) << R300_FPI3_SRC1A_SHIFT) | \
 		((0 | SRC_CONST) << R300_FPI3_SRC2A_SHIFT))
 
+#define DRI_CONF_FP_OPTIMIZATION_SPEED   0
+#define DRI_CONF_FP_OPTIMIZATION_QUALITY 1
+
 struct r300_fragment_program;
 
-extern void r300_translate_fragment_shader(struct r300_fragment_program *rp);
+extern void r300_translate_fragment_shader(r300ContextPtr r300, struct r300_fragment_program *rp);
 
 #endif
diff --git a/src/mesa/drivers/dri/r300/r300_fragprog_swz.c b/src/mesa/drivers/dri/r300/r300_fragprog_swz.c
deleted file mode 100644
index b29331d7bde..00000000000
--- a/src/mesa/drivers/dri/r300/r300_fragprog_swz.c
+++ /dev/null
@@ -1,1328 +0,0 @@
-/*
- * Copyright (C) 2005 Jerome Glisse.  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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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.
- *
- */
-#include "r300_fragprog.h"
-#include "r300_reg.h"
-
-
-#define I0_000	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_ZERO) |				\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG2C_SHIFT) )
-#define I0_111	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_ZERO) |				\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ONE  << R300_FPI0_ARG2C_SHIFT) )
-#define I0_XXX	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_SRC0C_XXX) |				\
-		  (R300_FPI0_ARGC_ONE  << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG2C_SHIFT) )
-#define I0_YYY	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_SRC0C_YYY) |				\
-		  (R300_FPI0_ARGC_ONE  << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG2C_SHIFT) )
-#define I0_ZZZ	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_SRC0C_ZZZ) |				\
-		  (R300_FPI0_ARGC_ONE  << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG2C_SHIFT) )
-#define I0_XYZ	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_SRC0C_XYZ) |				\
-		  (R300_FPI0_ARGC_ONE  << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG2C_SHIFT) )
-#define I0_YZX	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_SRC0C_YZX) |				\
-		  (R300_FPI0_ARGC_ONE  << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG2C_SHIFT) )
-#define I0_ZXY	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_SRC0C_ZXY) |				\
-		  (R300_FPI0_ARGC_ONE  << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG2C_SHIFT) )
-#define I0_WZY	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_SRC0CA_WZY) |				\
-		  (R300_FPI0_ARGC_ONE  << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG2C_SHIFT) )
-#define I0_WWW	( (R300_FPI0_OUTC_MAD) |				\
-		  (R300_FPI0_ARGC_SRC0A) |				\
-		  (R300_FPI0_ARGC_ONE  << R300_FPI0_ARG1C_SHIFT) |	\
-		  (R300_FPI0_ARGC_ZERO << R300_FPI0_ARG2C_SHIFT) )
-
-#define IEMPTY	0
-
-#define I1_XYZ	( R300_FPI1_SRC1C_CONST |	\
-		  R300_FPI1_SRC2C_CONST |	\
-		  R300_FPI1_DSTC_REG_X |	\
-		  R300_FPI1_DSTC_REG_Y |	\
-		  R300_FPI1_DSTC_REG_Z )
-#define I1_XY_	( R300_FPI1_SRC1C_CONST |	\
-		  R300_FPI1_SRC2C_CONST |	\
-		  R300_FPI1_DSTC_REG_X |	\
-		  R300_FPI1_DSTC_REG_Y )	
-#define I1_X_Z	( R300_FPI1_SRC1C_CONST |	\
-		  R300_FPI1_SRC2C_CONST |	\
-		  R300_FPI1_DSTC_REG_X |	\
-		  R300_FPI1_DSTC_REG_Z )	
-#define I1__YZ	( R300_FPI1_SRC1C_CONST |	\
-		  R300_FPI1_SRC2C_CONST |	\
-		  R300_FPI1_DSTC_REG_Y |	\
-		  R300_FPI1_DSTC_REG_Z )
-#define I1_X__	( R300_FPI1_SRC1C_CONST |	\
-		  R300_FPI1_SRC2C_CONST |	\
-		  R300_FPI1_DSTC_REG_X )
-#define I1__Y_	( R300_FPI1_SRC1C_CONST |	\
-		  R300_FPI1_SRC2C_CONST |	\
-		  R300_FPI1_DSTC_REG_Y )
-#define I1___Z	( R300_FPI1_SRC1C_CONST |	\
-		  R300_FPI1_SRC2C_CONST |	\
-		  R300_FPI1_DSTC_REG_Z )
-
-#define SEMPTY	{0,{0,0,0,0},{0,0,0,0,0,0,0,0}}
-
-struct r300_fragment_program_swizzle r300_swizzle [512] = {
-	/* XXX */
-	{1,{0,0,0,0},{ I0_XXX, I1_XYZ,
-		       0, 0, 0, 0, 0, 0 } },
-	/* YXX */
-	{2,{0,0,0,0},{ I0_YZX, I1_X_Z,
-		       I0_XXX, I1__Y_,
-		       0,0,
-		       0,0 } },
-	/* ZXX */
-	{2,{0,0,0,0},{ I0_ZZZ, I1_X__,
-		       I0_XXX, I1__YZ,
-		       0,0,
-		       0,0 } },
-	/* WXX */
-	{2,{0,0,0,0},{ I0_WZY, I1_X__,
-		       I0_XXX, I1__YZ,
-		       0,0,
-		       0,0} },
-	/* 0XX */
-	{2,{0,2,0,0},{ I0_XXX, I1__YZ,
-		       I0_000, I1_X__,
-		       0,0,
-		       0,0 } },
-	/* 1XX */
-	{2,{0,2,0,0},{ I0_XXX, I1__YZ,
-		       I0_111, I1_X__,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XYX */
-	{2,{0,0,0,0},{ I0_YYY, I1__Y_,
-		       I0_XXX, I1_X_Z,
-		       0,0,0,0}},
-	/* YYX */
-	{2,{0,0,0,0},{ I0_YYY, I1_XY_,
-		       I0_XXX, I1___Z,
-		       0,0,0,0}},
-	/* ZYX */
-	{3,{0,0,0,0},{ I0_ZZZ, I1_X__,
-		       I0_YYY, I1__Y_,
-		       I0_XXX, I1___Z,
-		       0,0}},
-	/* WYX */
-	{3,{0,0,0,0},{ I0_WZY, I1_X__,
-		       I0_YYY, I1__Y_,
-		       I0_XXX, I1___Z,
-		       0,0}},
-	/* 0YX */
-	{3,{0,0,2,0},{ I0_YYY, I1__Y_,
-		       I0_XXX, I1___Z,
-	               I0_000, I1_X__,
-		       0,0}},
-	/* 1YX */
-	{3,{0,0,2,0},{ I0_YYY, I1__Y_,
-		       I0_XXX, I1___Z,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* XZX */
-	{2,{0,0,0,0},{ I0_YZX, I1__YZ,
-		       I0_XXX, I1_X__,
-		       0,0,0,0}},
-	/* YZX */
-	{1,{0,0,0,0},{ I0_YZX, I1_XYZ,
-		       0, 0, 0, 0, 0, 0 } },
-	/* ZZX */
-	{2,{0,0,0,0},{ I0_YZX, I1__YZ,
-		       I0_ZZZ, I1_X__,0,0,0,0}},
-	/* WZX */
-	{2,{0,0,0,0},{ I0_WZY, I1_XY_,
-		       I0_XXX, I1___Z,0,0,0,0}},
-	/* 0ZX */
-	{2,{0,2,0,0},{ I0_YZX, I1__YZ,
-		       I0_000, I1_X__,
-	               0,0,0,0}},
-	/* 1ZX */
-	{2,{0,2,0,0},{ I0_YZX, I1__YZ,
-		       I0_111, I1_X__,
-	               0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XWX */
-	{2,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_XXX, I1_X_Z,
-		       0,0,0,0}},
-	/* YWX */
-	{2,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_YZX, I1_X_Z,
-		       0,0,0,0}},
-	/* ZWX */
-	{3,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_ZZZ, I1_X__,
-		       I0_XXX, I1___Z,
-		       0,0}},
-	/* WWX */
-	{2,{0,0,0,0},{ I0_WWW, I1_XY_,
-		       I0_YZX, I1___Z,
-		       0,0,0,0}},
-	/* 0WX */
-	{3,{0,0,2,0},{ I0_WWW, I1__Y_,
-		       I0_XXX, I1___Z,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1WX */
-	{3,{0,0,2,0},{ I0_WWW, I1__Y_,
-		       I0_XXX, I1___Z,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* X0X */
-	{2,{0,2,0,0},{ I0_XXX, I1_X_Z,
-		       I0_000, I1__Y_,
-		       0,0,0,0}},
-	/* Y0X */
-	{2,{0,2,0,0},{ I0_YZX, I1_X_Z,
-		       I0_000, I1__Y_,
-		       0,0,0,0}},
-	/* Z0X */
-	{3,{0,2,0,0},{ I0_XXX, I1___Z,
-		       I0_000, I1__Y_,
-		       I0_ZZZ, I1_X__,
-		       0,0}},
-	/* W0X */
-	{3,{0,0,2,0},{ I0_WZY, I1_XYZ,
-		       I0_XXX, I1___Z,		       
-		       I0_000, I1__Y_,
-		       0,0}},
-	/* 00X */
-	{2,{0,2,0,0},{ I0_XXX, I1___Z,
-		       I0_000, I1_XY_,
-		       0,0,0,0}},
-	/* 10X */
-	{3,{0,2,0,0},{ I0_XXX, I1___Z,
-		       I0_000, I1__Y_,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* X1X */
-	{2,{0,2,0,0},{ I0_XXX, I1_X_Z,
-		       I0_111, I1__Y_,
-		       0,0,0,0}},
-	/* Y1X */
-	{2,{0,2,0,0},{ I0_YZX, I1_X_Z,
-		       I0_111, I1__Y_,
-		       0,0,0,0}},
-	/* Z1X */
-	{3,{0,2,0,0},{ I0_XXX, I1___Z,
-		       I0_111, I1__Y_,
-		       I0_ZZZ, I1_X__,
-		       0,0}},
-	/* W1X */
-	{3,{0,0,2,0},{ I0_WZY, I1_XYZ,
-		       I0_XXX, I1___Z,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* 01X */
-	{3,{0,2,0,0},{ I0_XXX, I1___Z,
-		       I0_111, I1__Y_,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 11X */
-	{2,{0,2,0,0},{ I0_XXX, I1___Z,
-		       I0_111, I1_XY_,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	/* XXY */
-	{2,{0,0,0,0},{ I0_YYY, I1___Z,
-		       I0_XXX, I1_XY_,
-		       0,0,0,0}},
-	/* YXY */
-	{2,{0,0,0,0},{ I0_YYY, I1_X_Z,
-		       I0_XXX, I1__Y_,
-		       0,0,0,0}},
-	/* ZXY */
-	{1,{0,0,0,0},{ I0_ZXY, I1_XYZ,
-		       0, 0, 0, 0, 0, 0 } },
-	/* WXY */
-	{2,{0,0,0,0},{ I0_WZY, I1_X__,
-		       I0_ZXY, I1__YZ,
-		       0,0,0,0}},
-	/* 0XY */
-	{2,{0,0,0,0},{ I0_ZXY, I1__YZ,
-		       I0_000, I1_X__,
-		       0,0,0,0}},
-	/* 1XY */
-	{2,{0,0,0,0},{ I0_ZXY, I1__YZ,
-		       I0_111, I1_X__,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XYY */
-	{2,{0,0,0,0},{ I0_YYY, I1__YZ,
-		       I0_XXX, I1_X__,
-		       0,0,0,0}},
-	/* YYY */
-	{1,{0,0,0,0},{ I0_YYY, I1_XYZ,
-		       0, 0, 0, 0, 0, 0 } },
-	/* ZYY */
-	{2,{0,0,0,0},{ I0_YYY, I1__YZ,
-		       I0_ZZZ, I1_X__,
-		       0,0,0,0}},
-	/* WYY */
-	{2,{0,0,0,0},{ I0_WZY, I1_XYZ,
-		       I0_YYY, I1__YZ,
-		       0,0,0,0}},
-	/* 0YY */
-	{2,{0,0,0,0},{ I0_YYY, I1__YZ,
-		       I0_000, I1_X__,
-		       0,0,0,0}},
-	/* 1YY */
-	{2,{0,0,0,0},{ I0_YYY, I1__YZ,
-		       I0_111, I1_X__,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XZY */
-	{2,{0,0,0,0},{ I0_WZY, I1__YZ,
-		       I0_XXX, I1_X__,
-		       0,0,0,0}},
-	/* YZY */
-	{2,{0,0,0,0},{ I0_WZY, I1__YZ,
-		       I0_YYY, I1_X__,
-		       0,0,0,0}},
-	/* ZZY */
-	{2,{0,0,0,0},{ I0_WZY, I1__YZ,
-		       I0_ZZZ, I1_X__,
-		       0,0,0,0}},
-	/* WZY */
-	{1,{0,0,0,0},{ I0_WZY, I1_XYZ,
-		       0, 0, 0, 0, 0, 0 } },
-	/* 0ZY */
-	{2,{0,0,0,0},{ I0_WZY, I1__YZ,
-		       I0_000, I1_X__,
-		       0,0,0,0}},
-	/* 1ZY */
-	{2,{0,0,0,0},{ I0_WZY, I1__YZ,
-		       I0_111, I1_X__,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XWY */
-	{3,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_XXX, I1_X__,
-		       I0_YYY, I1___Z,
-		       0,0}},
-	/* YWY */
-	{2,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_YYY, I1_X_Z,
-		       0,0,0,0}},
-	/* ZWY */
-	{2,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_ZXY, I1_X_Z,
-		       0,0,0,0}},
-	/* WWY */
-	{2,{0,0,0,0},{ I0_WWW, I1_XY_,
-		       I0_ZXY, I1___Z,
-		       0,0,0,0}},
-	/* 0WY */
-	{3,{0,0,2,0},{ I0_WWW, I1__Y_,
-		       I0_ZXY, I1___Z,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1WY */
-	{3,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_ZXY, I1___Z,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* X0Y */
-	{3,{0,2,0,0},{ I0_XXX, I1_X__,
-		       I0_000, I1__Y_,
-		       I0_YYY, I1___Z,
-		       0,0}},
-	/* Y0Y */
-	{2,{0,2,0,0},{ I0_YYY, I1_X_Z,
-		       I0_000, I1__Y_,
-		       0,0,0,0}},
-	/* Z0Y */
-	{2,{0,2,0,0},{ I0_ZXY, I1_X_Z,
-		       I0_000, I1__Y_,
-		       0,0,0,0}},
-	/* W0Y */
-	{2,{0,2,0,0},{ I0_WZY, I1_X_Z,
-		       I0_000, I1__Y_,
-		       0,0,0,0}},
-	/* 00Y */
-	{2,{0,2,0,0},{ I0_YYY, I1___Z,
-		       I0_000, I1_XY_,
-		       0,0,0,0}},
-	/* 10Y */
-	{3,{0,2,0,0},{ I0_YYY, I1___Z,
-		       I0_000, I1__Y_,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* X1Y */
-	{3,{0,2,0,0},{ I0_XXX, I1_X__,
-		       I0_111, I1__Y_,
-		       I0_YYY, I1___Z,
-		       0,0}},
-	/* Y1Y */
-	{2,{0,2,0,0},{ I0_YYY, I1_X_Z,
-		       I0_111, I1__Y_,
-		       0,0,0,0}},
-	/* Z1Y */
-	{2,{0,2,0,0},{ I0_ZXY, I1_X_Z,
-		       I0_111, I1__Y_,
-		       0,0,0,0}},
-	/* W1Y */
-	{3,{0,2,0,0},{ I0_WZY, I1_X_Z,
-		       I0_111, I1__Y_,
-		       0,0,0,0}},
-	/* 01Y */
-	{3,{0,2,0,0},{ I0_YYY, I1___Z,
-		       I0_111, I1__Y_,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 11Y */
-	{2,{0,2,0,0},{ I0_YYY, I1___Z,
-		       I0_111, I1_XY_,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	/* XXZ */
-	{2,{0,0,0,0},{ I0_XXX, I1_XY_,
-		       I0_ZZZ, I1___Z,
-		       0,0,0,0}},
-	/* YXZ */
-	{3,{0,0,0,0},{ I0_XXX, I1__Y_,
-		       I0_YYY, I1_X__,
-		       I0_ZZZ, I1___Z,
-		       0,0}},
-	/* ZXZ */
-	{2,{0,0,0,0},{ I0_XXX, I1__Y_,
-		       I0_ZZZ, I1_X_Z,
-		       0,0,0,0}},
-	/* WXZ */
-	{3,{0,0,0,0},{ I0_WZY, I1_XYZ,
-		       I0_XXX, I1__Y_,
-		       I0_ZZZ, I1___Z,
-		       0,0}},
-	/* 0XZ */
-	{3,{0,0,2,0},{ I0_XXX, I1__Y_,
-		       I0_ZZZ, I1___Z,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1XZ */
-	{3,{0,0,2,0},{ I0_XXX, I1__Y_,
-		       I0_ZZZ, I1___Z,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* XYZ */
-	{1,{0,0,0,0},{ I0_XYZ, I1_XYZ,
-		       0, 0, 0, 0, 0, 0 } },
-	/* YYZ */
-	{2,{0,0,0,0},{ I0_ZZZ, I1___Z,
-		       I0_YYY, I1_XY_,
-		       0,0,0,0}},
-	/* ZYZ */
-	{2,{0,0,0,0},{ I0_ZZZ, I1_X_Z,
-		       I0_YYY, I1__Y_,
-		       0,0,0,0}},
-	/* WYZ */
-	{2,{0,0,0,0},{ I0_WZY, I1_XYZ,
-		       I0_XYZ, I1__YZ,
-		       0,0,0,0}},
-	/* 0YZ */
-	{2,{0,2,0,0},{ I0_XYZ, I1__YZ,
-		       I0_000, I1_X__,
-		       0,0,0,0}},
-	/* 1YZ */
-	{2,{0,2,0,0},{ I0_XYZ, I1__YZ,
-		       I0_111, I1_X__,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XZZ */
-	{2,{0,0,0,0},{ I0_ZZZ, I1__YZ,
-		       I0_XXX, I1_X__,
-		       0,0,0,0}},
-	/* YZZ */
-	{2,{0,0,0,0},{ I0_ZZZ, I1__YZ,
-		       I0_YYY, I1_X__,
-		       0,0,0,0}},
-	/* ZZZ */
-	{1,{0,0,0,0},{ I0_ZZZ, I1_XYZ,
-		       0, 0, 0, 0, 0, 0 } },
-	/* WZZ */
-	{2,{0,0,0,0},{ I0_WZY, I1_XYZ,
-		       I0_ZZZ, I1__YZ,
-		       0,0,0,0}},
-	/* 0ZZ */
-	{2,{0,2,0,0},{ I0_ZZZ, I1__YZ,
-		       I0_000, I1_X__,
-		       0,0,0,0}},
-	/* 1ZZ */
-	{2,{0,2,0,0},{ I0_ZZZ, I1__YZ,
-		       I0_111, I1_X__,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XWZ */
-	{2,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_XYZ, I1_X_Z,
-		       0,0,0,0}},
-	/* YWZ */
-	{3,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_YYY, I1_X__,
-		       I0_XYZ, I1___Z,
-		       0,0}},
-	/* ZWZ */
-	{2,{0,0,0,0},{ I0_WWW, I1__Y_,
-		       I0_ZZZ, I1_X_Z,
-		       0,0,0,0}},
-	/* WWZ */
-	{2,{0,0,0,0},{ I0_WWW, I1_XY_,
-		       I0_XYZ, I1___Z,
-		       0,0,0,0}},
-	/* 0WZ */
-	{3,{0,0,2,0},{ I0_WWW, I1__Y_,
-		       I0_XYZ, I1___Z,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1WZ */
-	{3,{0,0,2,0},{ I0_WWW, I1__Y_,
-		       I0_XYZ, I1___Z,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* X0Z */
-	{2,{0,2,0,0},{ I0_XYZ, I1_X_Z,
-		       I0_000, I1__Y_,
-		       0,0,0,0}},
-	/* Y0Z */
-	{3,{0,2,0,0},{ I0_ZZZ, I1___Z,
-		       I0_000, I1__Y_,
-		       I0_YYY, I1_X__,
-		       0,0}},
-	/* Z0Z */
-	{2,{0,2,0,0},{ I0_ZZZ, I1_X_Z,
-		       I0_000, I1__Y_,
-		       0,0,0,0}},
-	/* W0Z */
-	{3,{0,0,2,0},{ I0_WZY, I1_X_Z,
-		       I0_ZZZ, I1___Z,
-		       I0_000, I1__Y_,
-		       0,0}},
-	/* 00Z */
-	{2,{0,2,0,0},{ I0_ZZZ, I1___Z,
-		       I0_000, I1_XY_,
-		       0,0,0,0}},
-	/* 10Z */
-	{3,{0,2,2,0},{ I0_ZZZ, I1___Z,
-		       I0_000, I1__Y_,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* X1Z */
-	{2,{0,2,0,0},{ I0_XYZ, I1_X_Z,
-		       I0_111, I1__Y_,
-		       0,0,0,0}},
-	/* Y1Z */
-	{3,{0,2,0,0},{ I0_ZZZ, I1___Z,
-		       I0_111, I1__Y_,
-		       I0_YYY, I1_X__,
-		       0,0}},
-	/* Z1Z */
-	{2,{0,2,0,0},{ I0_ZZZ, I1_X_Z,
-		       I0_111, I1__Y_,
-		       0,0,0,0}},
-	/* W1Z */
-	{3,{0,0,2,0},{ I0_WZY, I1_XYZ,
-		       I0_ZZZ, I1___Z,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* 01Z */
-	{3,{0,2,2,0},{ I0_ZZZ, I1___Z,
-		       I0_111, I1__Y_,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 11Z */
-	{2,{0,2,0,0},{ I0_ZZZ, I1___Z,
-		       I0_111, I1_XY_,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	/* XXW */
-	{2,{0,0,0,0},{ I0_WWW, I1___Z,
-		       I0_XXX, I1_XY_,
-		       0,0,0,0}},
-	/* YXW */
-	{3,{0,0,0,0},{ I0_WWW, I1___Z,
-		       I0_XXX, I1__Y_,
-		       I0_YYY, I1_X__,
-		       0,0}},
-	/* ZXW */
-	{2,{0,0,0,0},{ I0_WWW, I1___Z,
-		       I0_ZXY, I1_XY_,
-		       0,0,0,0}},
-	/* WXW */
-	{2,{0,0,0,0},{ I0_WWW, I1_X_Z,
-		       I0_XXX, I1__Y_,
-		       0,0,0,0}},
-	/* 0XW */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_XXX, I1__Y_,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1XW */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_XXX, I1__Y_,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* XYW */
-	{2,{0,0,0,0},{ I0_WWW, I1___Z,
-		       I0_XYZ, I1_XY_,
-		       0,0,0,0}},
-	/* YYW */
-	{2,{0,0,0,0},{ I0_WWW, I1___Z,
-		       I0_YYY, I1_XY_,
-		       0,0}},
-	/* ZYW */
-	{3,{0,0,0,0},{ I0_WWW, I1___Z,
-		       I0_XYZ, I1__Y_,
-		       I0_ZZZ, I1_X__,
-		       0,0}},
-	/* WYW */
-	{2,{0,0,0,0},{ I0_WWW, I1_X_Z,
-		       I0_YYY, I1__Y_,
-		       0,0,0,0}},
-	/* 0YW */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_YYY, I1__Y_,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1YW */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_YYY, I1__Y_,
-		       I0_111, I1_X__,
-		       0,0}},
-
-	SEMPTY,SEMPTY,
-	/* XZW */
-	{3,{0,0,0,0},{ I0_WWW, I1___Z,
-		       I0_XYZ, I1_X__,
-		       I0_ZZZ, I1__Y_,
-		       0,0}},
-	/* YZW */
-	{2,{0,0,0,0},{ I0_WWW, I1___Z,
-		       I0_YZX, I1_XY_,
-		       0,0,0,0}},
-	/* ZZW */
-	{2,{0,0,0,0},{ I0_WWW, I1___Z,
-		       I0_ZZZ, I1_XY_,
-		       0,0,0,0}},
-	/* WZW */
-	{2,{0,0,0,0},{ I0_WWW, I1_X_Z,
-		       I0_ZZZ, I1__Y_,
-		       0,0,0,0}},
-	/* 0ZW */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_ZZZ, I1__Y_,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1ZW */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_ZZZ, I1__Y_,
-		       I0_111, I1_X__,
-		       0,0}},
-
-	SEMPTY,SEMPTY,
-	/* XWW */
-	{2,{0,0,0,0},{ I0_WWW, I1__YZ,
-		       I0_XYZ, I1_X__,
-		       0,0,0,0}},
-	/* YWW */
-	{2,{0,0,0,0},{ I0_WWW, I1__YZ,
-		       I0_YYY, I1_X__,
-		       0,0,0,0}},
-	/* ZWW */
-	{2,{0,0,0,0},{ I0_WWW, I1__YZ,
-		       I0_ZZZ, I1_X__,
-		       0,0,0,0}},
-	/* WWW */
-	{1,{0,0,0,0},{ I0_WWW, I1_XYZ,
-		       0,0,0,0,0,0}},
-	/* 0WW */
-	{2,{0,2,0,0},{ I0_WWW, I1__YZ,
-		       I0_000, I1_X__,
-		       0,0,0,0}},
-	/* 1WW */
-	{2,{0,2,0,0},{ I0_WWW, I1__YZ,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* X0W */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_XYZ, I1_X__,
-		       I0_000, I1__Y_,
-		       0,0}},
-	/* Y0W */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_YYY, I1_X__,
-		       I0_000, I1__Y_,
-		       0,0}},
-	/* Z0W */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_ZZZ, I1_X__,
-		       I0_000, I1__Y_,
-		       0,0}},
-	/* W0W */
-	{2,{0,2,0,0},{ I0_WWW, I1_X_Z,
-		       I0_000, I1__Y_,
-		       0,0,0,0}},
-	/* 00W */
-	{2,{0,2,0,0},{ I0_WWW, I1___Z,
-		       I0_000, I1_XY_,
-		       0,0,0,0}},
-	/* 10W */
-	{3,{0,2,2,0},{ I0_WWW, I1___Z,
-		       I0_111, I1_X__,
-		       I0_000, I1__Y_,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* X1W */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_XYZ, I1_X__,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* Y1W */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_YYY, I1_X__,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* Z1W */
-	{3,{0,0,2,0},{ I0_WWW, I1___Z,
-		       I0_ZZZ, I1_X__,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* W1W */
-	{2,{0,2,0,0},{ I0_WWW, I1_XYZ,
-		       I0_111, I1__Y_,
-		       0,0,0,0}},
-	/* 01W */
-	{3,{0,2,2,0},{ I0_WWW, I1___Z,
-		       I0_000, I1_X__,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* 11W */
-	{2,{0,2,0,0},{ I0_WWW, I1___Z,
-		       I0_111, I1_XY_,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	/* XX0 */
-	{2,{0,2,0,0},{ I0_XXX, I1_XY_,
-		       I0_000, I1___Z,
-		       0,0,0,0}},
-	/* YX0 */
-	{3,{0,0,2,0},{ I0_YYY, I1_X__,
-		       I0_XXX, I1__Y_,
-		       I0_000, I1___Z,
-		       0,0}},
-	/* ZX0 */
-	{2,{0,2,0,0},{ I0_ZXY, I1_XY_,
-		       I0_000, I1___Z,
-		       0,0,0,0}},
-	/* WX0 */
-	{3,{0,0,2,0},{ I0_WZY, I1_X__,
-		       I0_XXX, I1__Y_,
-		       I0_000, I1___Z,
-		       0,0}},
-	/* 0X0 */
-	{2,{0,2,0,0},{ I0_XXX, I1__Y_,
-		       I0_000, I1_X_Z,
-		       0,0,0,0}},
-	/* 1X0 */
-	{3,{0,2,2,0},{ I0_XXX, I1__Y_,
-		       I0_000, I1___Z,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* XY0 */
-	{2,{0,2,0,0},{ I0_XYZ, I1_XY_,
-		       I0_000, I1___Z,
-		       0,0,0,0}},
-	/* YY0 */
-	{2,{0,2,0,0},{ I0_YYY, I1_XY_,
-		       I0_000, I1___Z,
-		       0,0,0,0}},
-	/* ZY0 */
-	{3,{0,0,2,0},{ I0_YYY, I1__Y_,
-		       I0_ZZZ, I1_X__,
-		       I0_000, I1___Z,
-		       0,0}},
-	/* WY0 */
-	{3,{0,0,2,0},{ I0_WZY, I1_X__,
-		       I0_XYZ, I1__Y_,
-		       I0_000, I1___Z,
-		       0,0}},
-	/* 0Y0 */
-	{2,{0,2,0,0},{ I0_XYZ, I1__Y_,
-		       I0_000, I1_X_Z,
-		       0,0,0,0}},
-	/* 1Y0 */
-	{3,{0,2,2,0},{ I0_XYZ, I1__Y_,
-		       I0_000, I1___Z,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* XZ0 */
-	{3,{0,0,2,0},{ I0_ZZZ, I1__Y_,
-		       I0_XYZ, I1_X__,
-		       I0_000, I1___Z,
-		       0,0}},
-	/* YZ0 */
-	{2,{0,2,0,0},{ I0_YZX, I1_XY_,
-		       I0_000, I1___Z,
-		       0,0,0,0}},
-	/* ZZ0 */
-	{2,{0,2,0,0},{ I0_ZZZ, I1_XY_,
-		       I0_000, I1___Z,
-		       0,0,0,0}},
-	/* WZ0 */
-	{3,{0,0,2,0},{ I0_XYZ, I1_XYZ,
-		       I0_WZY, I1_XY_,
-		       I0_000, I1___Z,
-		       0,0}},
-	/* 0Z0 */
-	{2,{0,2,0,0},{ I0_ZZZ, I1__Y_,
-		       I0_000, I1_X_Z,
-		       0,0,0,0}},
-	/* 1Z0 */
-	{3,{0,2,2,0},{ I0_ZZZ, I1__Y_,
-		       I0_000, I1___Z,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* XW0 */
-	{3,{0,0,2,0},{ I0_WWW, I1__Y_,
-		       I0_XYZ, I1_X__,
-		       I0_000, I1___Z,
-		       0,0}},
-	/* YW0 */
-	{3,{0,2,0,0},{ I0_WWW, I1__Y_,
-		       I0_000, I1___Z,
-		       I0_YYY, I1_X__,
-		       0,0}},
-	/* ZW0 */
-	{3,{0,2,0,0},{ I0_WWW, I1__Y_,
-		       I0_000, I1___Z,
-		       I0_ZZZ, I1_X__,
-		       0,0}},
-	/* WW0 */
-	{2,{0,2,0,0},{ I0_WWW, I1_XY_,
-		       I0_000, I1___Z,
-		       0,0,0,0}},
-	/* 0W0 */
-	{2,{0,2,0,0},{ I0_WWW, I1__Y_,
-		       I0_000, I1_X_Z,
-		       0,0,0,0}},
-	/* 1W0 */
-	{3,{0,2,2,0},{ I0_WWW, I1__Y_,
-		       I0_000, I1___Z,
-		       I0_111, I1_X__,
-		       0,0}},
-	SEMPTY,SEMPTY,
-	/* X00 */
-	{2,{0,2,0,0},{ I0_XYZ, I1_X__,
-		       I0_000, I1__YZ,
-		       0,0,0,0}},
-	/* Y00 */
-	{2,{0,2,0,0},{ I0_YYY, I1_X__,
-		       I0_000, I1__YZ,
-		       0,0,0,0}},
-	/* Z00 */
-	{2,{0,2,0,0},{ I0_ZZZ, I1_X__,
-		       I0_000, I1__YZ,
-		       0,0,0,0}},
-	/* W00 */
-	{2,{2,0,0,0},{ I0_WZY, I1_X__,
-		       I0_000, I1__YZ,
-		       0,0,0,0}},
-	/* 000 */
-	{1,{2,0,0,0},{ I0_000, I1_XYZ,
-		       0, 0, 0, 0, 0, 0 } },
-	/* 100 */
-	{2,{2,2,0,0},{ I0_000, I1__YZ,
-		       I0_111, I1_X__,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* X10 */
-	{3,{0,2,2,0},{ I0_XYZ, I1_XYZ,
-		       I0_000, I1___Z,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* Y10 */
-	{3,{0,2,2,0},{ I0_YYY, I1_XYZ,
-		       I0_000, I1___Z,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* Z10 */
-	{3,{0,2,2,0},{ I0_ZZZ, I1_XYZ,
-		       I0_000, I1___Z,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* W10 */
-	{3,{0,2,2,0},{ I0_WZY, I1_XYZ,
-		       I0_000, I1___Z,
-		       I0_111, I1__Y_,
-		       0,0}},
-	/* 010 */
-	{2,{2,2,0,0},{ I0_000, I1_X_Z,
-		       I0_111, I1__Y_,
-		       0, 0, 0, 0 } },
-	/* 110 */
-	{2,{2,2,0,0},{ I0_000, I1___Z,
-		       I0_111, I1_XY_,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-
-
-
-	/* XX1 */
-	{2,{0,2,0,0},{ I0_XXX, I1_XY_,
-		       I0_111, I1___Z,
-		       0,0,0,0}},
-	/* YX1 */
-	{3,{0,0,2,0},{ I0_YYY, I1_X__,
-		       I0_XXX, I1__Y_,
-		       I0_111, I1___Z,
-		       0,0}},
-	/* ZX1 */
-	{2,{0,2,0,0},{ I0_ZXY, I1_XY_,
-		       I0_111, I1___Z,
-		       0,0,0,0}},
-	/* WX1 */
-	{3,{0,0,2,0},{ I0_WZY, I1_XYZ,
-		       I0_XXX, I1__Y_,
-		       I0_111, I1___Z,
-		       0,0}},
-	/* 0X1 */
-	{3,{0,2,2,0},{ I0_XXX, I1__Y_,
-		       I0_111, I1___Z,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1X1 */
-	{2,{0,2,0,0},{ I0_XXX, I1__Y_,
-		       I0_111, I1_X_Z,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XY1 */
-	{2,{0,2,0,0},{ I0_XYZ, I1_XY_,
-		       I0_111, I1___Z,
-		       0,0,0,0}},
-	/* YY1 */
-	{2,{0,2,0,0},{ I0_YYY, I1_XY_,
-		       I0_111, I1___Z,
-		       0,0,0,0}},
-	/* ZY1 */
-	{3,{0,0,2,0},{ I0_YYY, I1__Y_,
-		       I0_ZZZ, I1_X__,
-		       I0_111, I1___Z,
-		       0,0}},
-	/* WY1 */
-	{3,{0,0,2,0},{ I0_WZY, I1_XYZ,
-		       I0_XYZ, I1__Y_,
-		       I0_111, I1___Z,
-		       0,0}},
-	/* 0Y1 */
-	{3,{0,2,2,0},{ I0_XYZ, I1__Y_,
-		       I0_111, I1___Z,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1Y1 */
-	{2,{0,2,0,0},{ I0_XYZ, I1__Y_,
-		       I0_111, I1_X_Z,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XZ1 */
-	{3,{0,0,2,0},{ I0_ZZZ, I1__Y_,
-		       I0_XYZ, I1_X__,
-		       I0_111, I1___Z,
-		       0,0}},
-	/* YZ1 */
-	{2,{0,2,0,0},{ I0_YZX, I1_XY_,
-		       I0_111, I1___Z,
-		       0,0,0,0}},
-	/* ZZ1 */
-	{2,{0,2,0,0},{ I0_ZZZ, I1_XYZ,
-		       I0_111, I1___Z,
-		       0,0,0,0}},
-	/* WZ1 */
-	{2,{0,2,0,0},{ I0_WZY, I1_XY_,
-		       I0_111, I1___Z,
-		       0,0,0,0}},
-	/* 0Z1 */
-	{3,{0,2,2,0},{ I0_ZZZ, I1_XYZ,
-		       I0_111, I1___Z,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1Z1 */
-	{2,{0,2,0,0},{ I0_ZZZ, I1__Y_,
-		       I0_111, I1_X_Z,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* XW1 */
-	{3,{0,0,2,0},{ I0_WWW, I1__Y_,
-		       I0_XYZ, I1_X__,
-		       I0_111, I1___Z,
-		       0,0}},
-	/* YW1 */
-	{3,{0,2,0,0},{ I0_WWW, I1__Y_,
-		       I0_111, I1___Z,
-		       I0_YYY, I1_X__,
-		       0,0}},
-	/* ZW1 */
-	{3,{0,2,0,0},{ I0_WWW, I1__Y_,
-		       I0_111, I1___Z,
-		       I0_ZZZ, I1_X__,
-		       0,0}},
-	/* WW1 */
-	{2,{0,2,0,0},{ I0_WWW, I1_XY_,
-		       I0_111, I1___Z,
-		       0,0,0,0}},
-	/* 0W1 */
-	{3,{0,2,2,0},{ I0_WWW, I1__Y_,
-		       I0_111, I1___Z,
-		       I0_000, I1_X__,
-		       0,0}},
-	/* 1W1 */
-	{2,{0,2,0,0},{ I0_WWW, I1__Y_,
-		       I0_111, I1_X_Z,
-		       0,0,0,0}},
-	SEMPTY,SEMPTY,
-	/* X01 */
-	{3,{0,2,2,0},{ I0_XYZ, I1_X__,
-		       I0_111, I1___Z,
-		       I0_000, I1__Y_,
-		       0,0}},
-	/* Y01 */
-	{3,{0,2,2,0},{ I0_YYY, I1_X__,
-		       I0_111, I1___Z,
-		       I0_000, I1__Y_,
-		       0,0}},
-	/* Z01 */
-	{3,{0,2,2,0},{ I0_ZZZ, I1_X__,
-		       I0_111, I1___Z,
-		       I0_000, I1__Y_,
-		       0,0}},
-	/* W01 */
-	{3,{0,2,2,0},{ I0_WZY, I1_XYZ,
-		       I0_111, I1___Z,
-		       I0_000, I1__Y_,
-		       0,0}},
-	/* 001 */
-	{2,{2,2,0,0},{ I0_111, I1___Z,
-		       I0_000, I1_XY_,
-		       0,0,0,0}},
-	/* 101 */
-	{2,{2,2,0,0},{ I0_111, I1_X_Z,
-		       I0_000, I1__Y_,
-		       0, 0, 0, 0 } },
-	SEMPTY,SEMPTY,
-	/* X11 */
-	{2,{0,2,0,0},{ I0_XYZ, I1_X__,
-		       I0_111, I1__YZ,
-		       0,0,0,0}},
-	/* Y11 */
-	{2,{0,2,0,0},{ I0_YYY, I1_X__,
-		       I0_111, I1__YZ,
-		       0,0,0,0}},
-	/* Z11 */
-	{2,{0,2,0,0},{ I0_ZZZ, I1_X__,
-		       I0_111, I1__YZ,
-		       0,0,0,0}},
-	/* W11 */
-	{2,{0,2,0,0},{ I0_WZY, I1_XYZ,
-		       I0_111, I1__YZ,
-		       0,0,0,0}},
-	/* 011 */
-	{2,{2,2,0,0},{ I0_111, I1__YZ,
-		       I0_000, I1_X__,
-		       0,0,0,0}},
-	/* 111 */
-	{1,{2,0,0,0},{ I0_111, I1_XYZ,
-		       0, 0, 0, 0, 0, 0 } },
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,
-	SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY,SEMPTY
-};
-
-/******************************************************************************
-* Color source mask table
-******************************************************************************/
-
-#define S_111	R300_FPI0_ARGC_ONE
-#define S_000	R300_FPI0_ARGC_ZERO
-
-#define S0XXX	R300_FPI0_ARGC_SRC0C_XXX
-#define S0YYY	R300_FPI0_ARGC_SRC0C_YYY
-#define S0ZZZ	R300_FPI0_ARGC_SRC0C_ZZZ
-#define S0WWW	R300_FPI0_ARGC_SRC0A
-#define S0XYZ	R300_FPI0_ARGC_SRC0C_XYZ
-#define S0ZXY	R300_FPI0_ARGC_SRC0C_ZXY
-#define S0YZX	R300_FPI0_ARGC_SRC0C_YZX
-#define S0WZY	R300_FPI0_ARGC_SRC0CA_WZY
-#define S0WZY	R300_FPI0_ARGC_SRC0CA_WZY
-
-#define S1XXX	R300_FPI0_ARGC_SRC1C_XXX
-#define S1YYY	R300_FPI0_ARGC_SRC1C_YYY
-#define S1ZZZ	R300_FPI0_ARGC_SRC1C_ZZZ
-#define S1WWW	R300_FPI0_ARGC_SRC1A
-#define S1XYZ	R300_FPI0_ARGC_SRC1C_XYZ
-#define S1ZXY	R300_FPI0_ARGC_SRC1C_ZXY
-#define S1YZX	R300_FPI0_ARGC_SRC1C_YZX
-#define S1WZY	R300_FPI0_ARGC_SRC1CA_WZY
-
-#define S2XXX	R300_FPI0_ARGC_SRC2C_XXX
-#define S2YYY	R300_FPI0_ARGC_SRC2C_YYY
-#define S2ZZZ	R300_FPI0_ARGC_SRC2C_ZZZ
-#define S2WWW	R300_FPI0_ARGC_SRC2A
-#define S2XYZ	R300_FPI0_ARGC_SRC2C_XYZ
-#define S2ZXY	R300_FPI0_ARGC_SRC2C_ZXY
-#define S2YZX	R300_FPI0_ARGC_SRC2C_YZX
-#define S2WZY	R300_FPI0_ARGC_SRC2CA_WZY
-
-#define ntnat	32
-
-const GLuint r300_swz_srcc_mask[3][512] = {
-	{
-		S0XXX,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S0YZX,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S0ZXY,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,S0YYY,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,S0WZY,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S0XYZ,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S0ZZZ,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S0WWW,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,S_000,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,S_111,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat
-	},
-	{
-		S1XXX,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S1YZX,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S1ZXY,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,S1YYY,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,S1WZY,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S1XYZ,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S1ZZZ,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S1WWW,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,S_000,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,S_111,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat
-	},
-	{
-		S2XXX,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S2YZX,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S2ZXY,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,S2YYY,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,S2WZY,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S2XYZ,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S2ZZZ,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,S2WWW,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,S_000,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,S_111,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,ntnat,
-		ntnat,ntnat
-	}
-};
-
-/******************************************************************************
-* Alpha source mask table
-******************************************************************************/
-
-GLuint r300_swz_srca_mask[3][6] = {
-	{ R300_FPI2_ARGA_SRC0C_X,
-	  R300_FPI2_ARGA_SRC0C_Y,
-	  R300_FPI2_ARGA_SRC0C_Z,
-	  R300_FPI2_ARGA_SRC0A,
-	  R300_FPI2_ARGA_ZERO,
-	  R300_FPI2_ARGA_ONE },
-	{ R300_FPI2_ARGA_SRC1C_X,
-	  R300_FPI2_ARGA_SRC1C_Y,
-	  R300_FPI2_ARGA_SRC1C_Z,
-	  R300_FPI2_ARGA_SRC1A,
-	  R300_FPI2_ARGA_ZERO,
-	  R300_FPI2_ARGA_ONE },
-	{ R300_FPI2_ARGA_SRC2C_X,
-	  R300_FPI2_ARGA_SRC2C_Y,
-	  R300_FPI2_ARGA_SRC2C_Z,
-	  R300_FPI2_ARGA_SRC2A,
-	  R300_FPI2_ARGA_ZERO,
-	  R300_FPI2_ARGA_ONE },
-};
diff --git a/src/mesa/drivers/dri/r300/r300_render.c b/src/mesa/drivers/dri/r300/r300_render.c
index 03f168365dc..211c451f661 100644
--- a/src/mesa/drivers/dri/r300/r300_render.c
+++ b/src/mesa/drivers/dri/r300/r300_render.c
@@ -42,7 +42,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "api_arrayelt.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_vp_build.h"
 
@@ -352,7 +352,7 @@ GLboolean r300_run_vb_render(GLcontext *ctx,
 	r300EmitState(rmesa);
 	
 	for(i=0; i < VB->PrimitiveCount; i++){
-		GLuint prim = VB->Primitive[i].mode;
+		GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
 		GLuint start = VB->Primitive[i].start;
 		GLuint length = VB->Primitive[i].count;
 		
@@ -385,8 +385,18 @@ GLboolean r300_run_vb_render(GLcontext *ctx,
 int r300Fallback(GLcontext *ctx)
 {
 	r300ContextPtr r300 = R300_CONTEXT(ctx);
+	struct r300_fragment_program *rp =
+		(struct r300_fragment_program *)
+		(char *)ctx->FragmentProgram._Current;
 	int i;
 
+	if (rp) {
+		if (!rp->translated)
+			r300_translate_fragment_shader(r300, rp);
+
+		FALLBACK_IF(!rp->translated);
+	}
+
 	/* We do not do SELECT or FEEDBACK (yet ?)
 	 * Is it worth doing them ?
 	 */
@@ -406,6 +416,10 @@ int r300Fallback(GLcontext *ctx)
 	 */
 	FALLBACK_IF(ctx->Fog.Enabled);
 #endif
+	FALLBACK_IF(ctx->Stencil._TestTwoSide &&
+		    (ctx->Stencil.Ref[0] != ctx->Stencil.Ref[1] ||
+		     ctx->Stencil.ValueMask[0] != ctx->Stencil.ValueMask[1] ||
+		     ctx->Stencil.WriteMask[0] != ctx->Stencil.WriteMask[1]));
 
 	if(!r300->disable_lowimpact_fallback){
 		/* GL_POLYGON_OFFSET_POINT */
diff --git a/src/mesa/drivers/dri/r300/r300_state.c b/src/mesa/drivers/dri/r300/r300_state.c
index 72255066d59..906dfceb482 100644
--- a/src/mesa/drivers/dri/r300/r300_state.c
+++ b/src/mesa/drivers/dri/r300/r300_state.c
@@ -46,7 +46,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "api_arrayelt.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "texformat.h"
 
@@ -509,7 +509,6 @@ static void r300Enable(GLcontext* ctx, GLenum cap, GLboolean state)
 		if (r300->state.stencil.hw_stencil) {
 			R300_STATECHANGE(r300, zs);
 			if (state) {
-				WARN_ONCE("TODO - double side stencil !\n");
 				r300->hw.zs.cmd[R300_ZS_CNTL_0] |=
 				    R300_RB3D_STENCIL_ENABLE;
 			} else {
@@ -863,9 +862,12 @@ static void r300StencilFuncSeparate(GLcontext * ctx, GLenum face,
 						(R300_RB3D_ZS2_STENCIL_MASK << R300_RB3D_ZS2_STENCIL_MASK_SHIFT));
 	
 	flag = translate_func(ctx->Stencil.Function[0]);
-
-	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |= (flag << R300_RB3D_ZS1_FRONT_FUNC_SHIFT)
-					  | (flag << R300_RB3D_ZS1_BACK_FUNC_SHIFT);
+	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |= (flag << R300_RB3D_ZS1_FRONT_FUNC_SHIFT);
+	
+	if (ctx->Stencil._TestTwoSide)
+		flag = translate_func(ctx->Stencil.Function[1]);
+	
+	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |= (flag << R300_RB3D_ZS1_BACK_FUNC_SHIFT);
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_2] |= refmask;
 }
 
@@ -894,10 +896,19 @@ static void r300StencilOpSeparate(GLcontext * ctx, GLenum face, GLenum fail,
 	rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |=
 		 (translate_stencil_op(ctx->Stencil.FailFunc[0]) << R300_RB3D_ZS1_FRONT_FAIL_OP_SHIFT)
 		|(translate_stencil_op(ctx->Stencil.ZFailFunc[0]) << R300_RB3D_ZS1_FRONT_ZFAIL_OP_SHIFT)
-		|(translate_stencil_op(ctx->Stencil.ZPassFunc[0]) << R300_RB3D_ZS1_FRONT_ZPASS_OP_SHIFT)
-		|(translate_stencil_op(ctx->Stencil.FailFunc[0]) << R300_RB3D_ZS1_BACK_FAIL_OP_SHIFT)
-		|(translate_stencil_op(ctx->Stencil.ZFailFunc[0]) << R300_RB3D_ZS1_BACK_ZFAIL_OP_SHIFT)
-		|(translate_stencil_op(ctx->Stencil.ZPassFunc[0]) << R300_RB3D_ZS1_BACK_ZPASS_OP_SHIFT);
+		|(translate_stencil_op(ctx->Stencil.ZPassFunc[0]) << R300_RB3D_ZS1_FRONT_ZPASS_OP_SHIFT);
+	
+	if (ctx->Stencil._TestTwoSide) {
+		rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |=
+			 (translate_stencil_op(ctx->Stencil.FailFunc[1]) << R300_RB3D_ZS1_BACK_FAIL_OP_SHIFT)
+			|(translate_stencil_op(ctx->Stencil.ZFailFunc[1]) << R300_RB3D_ZS1_BACK_ZFAIL_OP_SHIFT)
+			|(translate_stencil_op(ctx->Stencil.ZPassFunc[1]) << R300_RB3D_ZS1_BACK_ZPASS_OP_SHIFT);
+	} else {
+		rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |=
+			 (translate_stencil_op(ctx->Stencil.FailFunc[0]) << R300_RB3D_ZS1_BACK_FAIL_OP_SHIFT)
+			|(translate_stencil_op(ctx->Stencil.ZFailFunc[0]) << R300_RB3D_ZS1_BACK_ZFAIL_OP_SHIFT)
+			|(translate_stencil_op(ctx->Stencil.ZPassFunc[0]) << R300_RB3D_ZS1_BACK_ZPASS_OP_SHIFT);
+	}
 }
 
 static void r300ClearStencil(GLcontext * ctx, GLint s)
@@ -1809,10 +1820,10 @@ void r300SetupPixelShader(r300ContextPtr rmesa)
 	if (!rp)	/* should only happenen once, just after context is created */
 		return;
 	
-	r300_translate_fragment_shader(rp);
+	r300_translate_fragment_shader(rmesa, rp);
 	if (!rp->translated) {
 		fprintf(stderr, "%s: No valid fragment shader, exiting\n", __func__);
-		exit(-1);
+		return;
 	}
 	
 #define OUTPUT_FIELD(st, reg, field)  \
@@ -1874,7 +1885,7 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
 	
 	_swrast_InvalidateState(ctx, new_state);
 	_swsetup_InvalidateState(ctx, new_state);
-	_ac_InvalidateState(ctx, new_state);
+	_vbo_InvalidateState(ctx, new_state);
 	_tnl_InvalidateState(ctx, new_state);
 	_ae_invalidate_state(ctx, new_state);
 
diff --git a/src/mesa/drivers/dri/r300/r300_vertexprog.c b/src/mesa/drivers/dri/r300/r300_vertexprog.c
index 2492a4a3a0a..2ff92e13289 100644
--- a/src/mesa/drivers/dri/r300/r300_vertexprog.c
+++ b/src/mesa/drivers/dri/r300/r300_vertexprog.c
@@ -960,26 +960,23 @@ static void position_invariant(struct gl_program *prog)
 
 static void insert_wpos(struct r300_vertex_program *vp,
 		       struct gl_program *prog,
-		       GLint pos)
+		       GLuint temp_index)
 {
 
 	GLint tokens[6] = { STATE_INTERNAL, STATE_R300_WINDOW_DIMENSION, 0, 0, 0, 0 };
 	struct prog_instruction *vpi;
 	struct prog_instruction *vpi_insert;
-	GLuint temp_index;
 	GLuint window_index;
 	int i = 0;
 	
 	vpi = malloc((prog->NumInstructions + 5) * sizeof(struct prog_instruction));
-	memcpy(vpi, prog->Instructions, (pos+1) * sizeof(struct prog_instruction));
+	/* all but END */
+	memcpy(vpi, prog->Instructions, (prog->NumInstructions - 1) * sizeof(struct prog_instruction));
+	/* END */
+	memcpy(&vpi[prog->NumInstructions + 4], &prog->Instructions[prog->NumInstructions - 1],
+		sizeof(struct prog_instruction));
 	
-	vpi_insert = &vpi[pos];
-
-	/* make a copy before outputting VERT_RESULT_HPOS */
-	vpi_insert->DstReg.File = vpi_insert->SrcReg[2].File;
-	vpi_insert->DstReg.Index = temp_index = vpi_insert->SrcReg[2].Index;
-	
-	vpi_insert++;
+	vpi_insert = &vpi[prog->NumInstructions - 1];
 	memset(vpi_insert, 0, 5 * sizeof(struct prog_instruction));
 
 	vpi_insert[i].Opcode = OPCODE_MOV;
@@ -1062,8 +1059,6 @@ static void insert_wpos(struct r300_vertex_program *vp,
 	vpi_insert[i].SrcReg[1].Swizzle = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_Y, SWIZZLE_ONE, SWIZZLE_ONE);
 	i++;
 
-	memcpy(&vpi_insert[i], &prog->Instructions[pos+1], (prog->NumInstructions-(pos+1)) * sizeof(struct prog_instruction));
-
 	free(prog->Instructions);
 
 	prog->Instructions = vpi;
@@ -1078,16 +1073,18 @@ static void pos_as_texcoord(struct r300_vertex_program *vp,
 			    struct gl_program *prog)
 {
 	struct prog_instruction *vpi;
-	int pos = 0;
-	
-	for(vpi = prog->Instructions; vpi->Opcode != OPCODE_END; vpi++, pos++){
+	GLuint tempregi = prog->NumTemporaries;
+	/* should do something else if no temps left... */
+	prog->NumTemporaries++;
+
+	for(vpi = prog->Instructions; vpi->Opcode != OPCODE_END; vpi++){
 		if( vpi->DstReg.File == PROGRAM_OUTPUT &&
 		    vpi->DstReg.Index == VERT_RESULT_HPOS ){
-			insert_wpos(vp, prog, pos);
-			break;
+			vpi->DstReg.File = PROGRAM_TEMPORARY;
+			vpi->DstReg.Index = tempregi;
 		}
 	}
-
+	insert_wpos(vp, prog, tempregi);
 }
 
 static struct r300_vertex_program *build_program(struct r300_vertex_program_key *wanted_key,
@@ -1101,8 +1098,9 @@ static struct r300_vertex_program *build_program(struct r300_vertex_program_key
 
 	vp->wpos_idx = wpos_idx;
 
-	if(mesa_vp->IsPositionInvariant)
+	if(mesa_vp->IsPositionInvariant) {
 		position_invariant(&mesa_vp->Base);
+	}
 
 	if(wpos_idx > -1)
 		pos_as_texcoord(vp, &mesa_vp->Base);
@@ -1158,6 +1156,10 @@ void r300_select_vertex_shader(r300ContextPtr r300)
 			wanted_key.OutputsWritten |= 1 << (VERT_RESULT_TEX0 + i);
 
 	wanted_key.InputsRead = vpc->mesa_program.Base.InputsRead;
+	if(vpc->mesa_program.IsPositionInvariant) {
+		/* we wan't position don't we ? */
+		wanted_key.InputsRead |= (1 << VERT_ATTRIB_POS);
+	}
 
 	for (vp = vpc->progs; vp; vp = vp->next)
 		if (_mesa_memcmp(&vp->key, &wanted_key, sizeof(wanted_key)) == 0) {
@@ -1170,6 +1172,5 @@ void r300_select_vertex_shader(r300ContextPtr r300)
 	vp = build_program(&wanted_key, &vpc->mesa_program, wpos_idx);
 	vp->next = vpc->progs;
 	vpc->progs = vp;
-
 	r300->selected_vp = vp;
 }
diff --git a/src/mesa/drivers/dri/r300/radeon_state.c b/src/mesa/drivers/dri/r300/radeon_state.c
index 0ceaf7ebb7d..ddadf83a003 100644
--- a/src/mesa/drivers/dri/r300/radeon_state.c
+++ b/src/mesa/drivers/dri/r300/radeon_state.c
@@ -41,7 +41,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "light.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
diff --git a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
index 72c03c53ad9..0625e5bc571 100644
--- a/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
+++ b/src/mesa/drivers/dri/r300/radeon_vtxfmt_a.c
@@ -46,6 +46,8 @@
 #include "state.h"
 #include "image.h"
 
+#include "vbo/vbo_context.h"
+
 #define CONV_VB(a, b) rvb->AttribPtr[(a)].size = vb->b->size, \
 			rvb->AttribPtr[(a)].type = GL_FLOAT, \
 			rvb->AttribPtr[(a)].stride = vb->b->stride, \
@@ -129,15 +131,7 @@ static int setup_arrays(r300ContextPtr rmesa, GLint start)
 			CONV(i, VertexAttrib[i]);
 	
 	for (i=0; i < VERT_ATTRIB_MAX; i++) {
-		if (enabled & (1 << i)) {
-			rmesa->state.VB.AttribPtr[i].data += rmesa->state.VB.AttribPtr[i].stride * start;
-		} else {
-			def.data = ctx->Current.Attrib[i];
-			memcpy(&rmesa->state.VB.AttribPtr[i], &def, sizeof(struct dt));
-		}
-		
-		/*if(rmesa->state.VB.AttribPtr[i].data == ctx->Current.Attrib[i])
-			fprintf(stderr, "%d is default coord\n", i);*/
+	   rmesa->state.VB.AttribPtr[i].data += rmesa->state.VB.AttribPtr[i].stride * start;
 	}
 	
 	for(i=0; i < VERT_ATTRIB_MAX; i++){
@@ -177,177 +171,18 @@ static int setup_arrays(r300ContextPtr rmesa, GLint start)
 
 void radeon_init_vtxfmt_a(r300ContextPtr rmesa);
 
-static void radeonDrawElements( GLenum mode, GLsizei count, GLenum type, const GLvoid *c_indices )
-{
-	GET_CURRENT_CONTEXT(ctx);
-	r300ContextPtr rmesa = R300_CONTEXT(ctx);
-	int elt_size;
-	int i;
-	unsigned int min = ~0, max = 0;
-	struct tnl_prim prim;
-	static void *ptr = NULL;
-	struct r300_dma_region rvb;
-	const GLvoid *indices = c_indices;
-	
-	if (count > 65535) {
-		WARN_ONCE("Too many verts!\n");
-		goto fallback;
-	}
-	
-	if (ctx->Array.ElementArrayBufferObj->Name) {
-		/* use indices in the buffer object */
-		if (!ctx->Array.ElementArrayBufferObj->Data) {
-			_mesa_warning(ctx, "DrawRangeElements with empty vertex elements buffer!");
-			return;
-		}
-		/* actual address is the sum of pointers */
-		indices = (GLvoid *)
-		ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data, (const GLubyte *) c_indices);
-	}
-	
-	if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
-		return;
-	
-	FLUSH_CURRENT( ctx, 0 );
-	
-	memset(&rvb, 0, sizeof(rvb));
-	switch (type) {
-	case GL_UNSIGNED_BYTE:
-		for (i=0; i < count; i++) {
-			if(((unsigned char *)indices)[i] < min)
-				min = ((unsigned char *)indices)[i];
-			if(((unsigned char *)indices)[i] > max)
-				max = ((unsigned char *)indices)[i];
-		}
-		
-#ifdef FORCE_32BITS_ELTS
-		elt_size = 4;
-#else
-		elt_size = 2;
-#endif		
-		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
-		rvb.aos_offset = GET_START(&rvb);
-		ptr = rvb.address + rvb.start;
-			
-#ifdef FORCE_32BITS_ELTS
-		for (i=0; i < count; i++)
-			((unsigned int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
-#else
-		for (i=0; i < count; i++)
-			((unsigned short int *)ptr)[i] = ((unsigned char *)indices)[i] - min;
-#endif
-	break;
-		
-	case GL_UNSIGNED_SHORT:
-		for (i=0; i < count; i++) {
-			if(((unsigned short int *)indices)[i] < min)
-				min = ((unsigned short int *)indices)[i];
-			if(((unsigned short int *)indices)[i] > max)
-				max = ((unsigned short int *)indices)[i];
-		}
-		
-#ifdef FORCE_32BITS_ELTS
-		elt_size = 4;
-#else
-		elt_size = 2;
-#endif
-		
-		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
-		rvb.aos_offset = GET_START(&rvb);
-		ptr = rvb.address + rvb.start;
-		
-#ifdef FORCE_32BITS_ELTS
-		for (i=0; i < count; i++)
-			((unsigned int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
-#else
-		for (i=0; i < count; i++)
-			((unsigned short int *)ptr)[i] = ((unsigned short int *)indices)[i] - min;
-#endif
-	break;
-	
-	case GL_UNSIGNED_INT:
-		for (i=0; i < count; i++) {
-			if(((unsigned int *)indices)[i] < min)
-				min = ((unsigned int *)indices)[i];
-			if(((unsigned int *)indices)[i] > max)
-				max = ((unsigned int *)indices)[i];
-		}
-		
-#ifdef FORCE_32BITS_ELTS
-		elt_size = 4;
-#else
-		if (max - min <= 65535)
-			elt_size = 2;
-		else 
-			elt_size = 4;
-#endif
-		r300AllocDmaRegion(rmesa, &rvb, count * elt_size, elt_size);
-		rvb.aos_offset = GET_START(&rvb);
-		ptr = rvb.address + rvb.start;
-		
-		
-		if (elt_size == 2)
-			for (i=0; i < count; i++)
-				((unsigned short int *)ptr)[i] = ((unsigned int *)indices)[i] - min;
-		else
-			for (i=0; i < count; i++)
-				((unsigned int *)ptr)[i] = ((unsigned int *)indices)[i] - min;
-	break;
-	
-	default:
-		WARN_ONCE("Unknown elt type!\n");
-	goto fallback;
-	}
-	
-	if (ctx->NewState) 
-		_mesa_update_state( ctx );
-	
-	r300UpdateShaders(rmesa);
-	
-	if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL) {
-		r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-		goto fallback;
-	}
-	
-	rmesa->state.VB.Count = max - min + 1;
-	
-	r300UpdateShaderStates(rmesa);
-	
-	rmesa->state.VB.Primitive = &prim;
-	rmesa->state.VB.PrimitiveCount = 1;
-	
-	prim.mode = mode | PRIM_BEGIN | PRIM_END;
-	if (rmesa->state.VB.LockCount)
-		prim.start = min - rmesa->state.VB.LockFirst;
-	else
-		prim.start = 0;
-	prim.count = count;
-	
-	rmesa->state.VB.Elts = ptr;
-	rmesa->state.VB.elt_size = elt_size;
-	
-	if (r300_run_vb_render(ctx, NULL)) {
-		r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-		goto fallback;
-	}
-	
-	if(rvb.buf)
-		radeon_mm_use(rmesa, rvb.buf->id);
-	
-	r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-	return;
-	
-	fallback:
-	_tnl_array_init(ctx);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
-	CALL_DrawElements(GET_DISPATCH(), (mode, count, type, c_indices));
-	radeon_init_vtxfmt_a(rmesa);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
-}
 
-static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei count, GLenum type, const GLvoid *c_indices)
+static void radeonDrawRangeElements(GLcontext *ctx,
+				    GLenum mode,
+				    GLuint min,
+				    GLuint max,
+				    GLsizei count, 
+				    GLenum type,
+				    const GLvoid *c_indices)
 {
-	GET_CURRENT_CONTEXT(ctx);
+#if 1
+	return GL_FALSE;
+#else
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct tnl_prim prim;
 	int elt_size;
@@ -371,26 +206,23 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 				indices += i * _mesa_sizeof_type(type);
 				count -= i;
 			}
-			return ;
+			return GL_TRUE;
 		}
 		WARN_ONCE("Too many verts!\n");
-		goto fallback;
+		return GL_FALSE;
 	}
 	
 	if (ctx->Array.ElementArrayBufferObj->Name) {
 		/* use indices in the buffer object */
 		if (!ctx->Array.ElementArrayBufferObj->Data) {
 			_mesa_warning(ctx, "DrawRangeElements with empty vertex elements buffer!");
-			return;
+			return GL_TRUE;
 		}
 		/* actual address is the sum of pointers */
 		indices = (GLvoid *)
 		ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data, (const GLubyte *) c_indices);
 	}
 	
-	if (!_mesa_validate_DrawRangeElements( ctx, mode, min, max, count, type, indices ))
-		return;
-	
 	FLUSH_CURRENT( ctx, 0 );
 #ifdef OPTIMIZE_ELTS
 	min = 0;
@@ -465,7 +297,7 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 	
 	default:
 		WARN_ONCE("Unknown elt type!\n");
-	goto fallback;
+		return GL_FALSE;
 	}
 	
 	/* XXX: setup_arrays before state update? */
@@ -477,7 +309,7 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 
 	if (setup_arrays(rmesa, min) >= R300_FALLBACK_TCL) {
 		r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-		goto fallback;
+		return GL_FALSE;
 	}
 
 	rmesa->state.VB.Count = max - min + 1;
@@ -501,37 +333,34 @@ static void radeonDrawRangeElements(GLenum mode, GLuint min, GLuint max, GLsizei
 	
 	if (r300_run_vb_render(ctx, NULL)) {
 		r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-		goto fallback;
+		return GL_FALSE;
 	}
 	
 	if(rvb.buf)
 		radeon_mm_use(rmesa, rvb.buf->id);
 	
 	r300ReleaseDmaRegion(rmesa, &rvb, __FUNCTION__);
-	return ;
-	
-	fallback:
-	_tnl_array_init(ctx);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
-	CALL_DrawRangeElements(GET_DISPATCH(), (mode, min, max, count, type, c_indices));
-	radeon_init_vtxfmt_a(rmesa);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
+	return GL_TRUE;
+#endif
 }
 
-static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
+static GLboolean radeonDrawArrays( GLcontext *ctx,
+				   GLenum mode, GLint start, GLsizei count )
 {
+#if 1
+	return GL_FALSE;
+#else
 	GET_CURRENT_CONTEXT(ctx);
 	r300ContextPtr rmesa = R300_CONTEXT(ctx);
 	struct tnl_prim prim;
 	
 	if (count > 65535) {
+	   /* TODO: split into multiple draws.
+	    */
 		WARN_ONCE("Too many verts!\n");
-		goto fallback;
+		return GL_FALSE;
 	}
 	
-	if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
-		return;
-	
 	FLUSH_CURRENT( ctx, 0 );
 	
 	if (ctx->NewState) 
@@ -542,7 +371,7 @@ static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
 	r300UpdateShaders(rmesa);
 
 	if (setup_arrays(rmesa, start) >= R300_FALLBACK_TCL)
-		goto fallback;
+		return GL_FALSE;
 
 	rmesa->state.VB.Count = count;
 
@@ -564,31 +393,70 @@ static void radeonDrawArrays( GLenum mode, GLint start, GLsizei count )
 	rmesa->state.VB.elt_max = 0;
 	
 	if (r300_run_vb_render(ctx, NULL))
-		goto fallback;
+	   return GL_FALSE;
 
-	return ;
-	
-	fallback:
-	_tnl_array_init(ctx);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
-	CALL_DrawArrays(GET_DISPATCH(), (mode, start, count));
-	radeon_init_vtxfmt_a(rmesa);
-	_mesa_install_exec_vtxfmt( ctx, &TNL_CONTEXT(ctx)->exec_vtxfmt );
+	return GL_TRUE;
+#endif
 }
 
+static void radeon_draw_prims( GLcontext *ctx,
+			       const struct gl_client_array *arrays[],
+			       const struct _mesa_prim *prim,
+			       GLuint nr_prims,
+			       const struct _mesa_index_buffer *ib,
+			       GLuint min_index,
+			       GLuint max_index)
+{
+	if (ib == NULL) {
+		for (i = 0; i < nr_prims; i++) {
+			if (!radeonDrawArrays(ctx,
+					      prim->mode,
+					      prim->start,
+					      prim->count)) {
+				/* Fallback
+				 */
+				_tnl_draw_prims(ctx,
+						arrays,
+						prim + i,
+						nr_prims - i,
+						ib, 
+						min_index,
+						max_index);
+				return;
+			}
+		}
+	} else {
+		for (i = 0; i < nr_prims; i++) {
+			if (!radeonDrawRangeElements(ctx,
+						     prim->mode,
+						     min_index,
+						     max_index,
+						     prim->count,
+						     ib->types,
+						     ib->ptr)) {
+				/* Fallback
+				 */
+				_tnl_draw_prims(ctx,
+						arrays,
+						prim + i,
+						nr_prims - i,
+						ib, 
+						min_index,
+						max_index);
+				return;
+			}
+		}
+	}
+}
+	   
 void radeon_init_vtxfmt_a(r300ContextPtr rmesa)
 {
 	GLcontext *ctx;
-	GLvertexformat *vfmt;
-	
-	ctx = rmesa->radeon.glCtx; 
-	vfmt = (GLvertexformat *)ctx->TnlModule.Current;
-   
-	vfmt->DrawElements = radeonDrawElements;
-	vfmt->DrawArrays = radeonDrawArrays;
-	vfmt->DrawRangeElements = radeonDrawRangeElements;
+	struct vbo_context *vbo = vbo_context(ctx);
 	
+	vbo->draw_prims = radeon_draw_prims;
 }
+
 #endif
 
 #ifdef HW_VBOS
diff --git a/src/mesa/drivers/dri/radeon/Makefile b/src/mesa/drivers/dri/radeon/Makefile
index 6e46620b397..f223b2d9228 100644
--- a/src/mesa/drivers/dri/radeon/Makefile
+++ b/src/mesa/drivers/dri/radeon/Makefile
@@ -22,11 +22,7 @@ DRIVER_SOURCES = \
 	radeon_swtcl.c \
 	radeon_span.c \
 	radeon_maos.c \
-	radeon_sanity.c \
-	radeon_vtxfmt.c \
-	radeon_vtxfmt_c.c \
-	radeon_vtxfmt_sse.c \
-	radeon_vtxfmt_x86.c 
+	radeon_sanity.c 
 
 C_SOURCES = \
 	$(COMMON_SOURCES) \
@@ -34,8 +30,7 @@ C_SOURCES = \
 
 DRIVER_DEFINES = -DRADEON_COMMON=0
 
-X86_SOURCES = \
-	radeon_vtxtmp_x86.S 
+X86_SOURCES = 
 
 include ../Makefile.template
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.c b/src/mesa/drivers/dri/radeon/radeon_context.c
index 8b92e3fe65f..6bc2c4aa5c5 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.c
+++ b/src/mesa/drivers/dri/radeon/radeon_context.c
@@ -46,7 +46,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -60,7 +60,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_tex.h"
 #include "radeon_swtcl.h"
 #include "radeon_tcl.h"
-#include "radeon_vtxfmt.h"
 #include "radeon_maos.h"
 
 #define need_GL_ARB_multisample
@@ -361,7 +360,7 @@ radeonCreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
    _ae_create_context( ctx );
@@ -370,13 +369,10 @@ radeonCreateContext( const __GLcontextModes *glVisual,
     */
    _tnl_destroy_pipeline( ctx );
    _tnl_install_pipeline( ctx, radeon_pipeline );
-   ctx->Driver.FlushVertices = radeonFlushVertices;
 
    /* Try and keep materials and vertices separate:
     */
-   _tnl_isolate_materials( ctx, GL_TRUE );
-
-/*     _mesa_allow_light_in_model( ctx, GL_FALSE ); */
+/*    _tnl_isolate_materials( ctx, GL_TRUE ); */
 
    /* Configure swrast and T&L to match hardware characteristics:
     */
@@ -450,10 +446,7 @@ radeonCreateContext( const __GLcontextModes *glVisual,
    }
 
    if (rmesa->radeonScreen->chip_flags & RADEON_CHIPSET_TCL) {
-      if (tcl_mode >= DRI_CONF_TCL_VTXFMT)
-	 radeonVtxfmtInit( ctx, tcl_mode >= DRI_CONF_TCL_CODEGEN );
-
-      _tnl_need_dlist_norm_lengths( ctx, GL_FALSE );
+/*       _tnl_need_dlist_norm_lengths( ctx, GL_FALSE ); */
    }
    return GL_TRUE;
 }
@@ -484,7 +477,7 @@ void radeonDestroyContext( __DRIcontextPrivate *driContextPriv )
       release_texture_heaps = (rmesa->glCtx->Shared->RefCount == 1);
       _swsetup_DestroyContext( rmesa->glCtx );
       _tnl_DestroyContext( rmesa->glCtx );
-      _ac_DestroyContext( rmesa->glCtx );
+      _vbo_DestroyContext( rmesa->glCtx );
       _swrast_DestroyContext( rmesa->glCtx );
 
       radeonDestroySwtcl( rmesa->glCtx );
@@ -494,12 +487,6 @@ void radeonDestroyContext( __DRIcontextPrivate *driContextPriv )
 	 radeonFlushCmdBuf( rmesa, __FUNCTION__ );
       }
 
-      if (!(rmesa->TclFallback & RADEON_TCL_FALLBACK_TCL_DISABLE)) {
-	 int tcl_mode = driQueryOptioni(&rmesa->optionCache, "tcl_mode");
-	 if (tcl_mode >= DRI_CONF_TCL_VTXFMT)
-	    radeonVtxfmtDestroy( rmesa->glCtx );
-      }
-
       _mesa_vector4f_free( &rmesa->tcl.ObjClean );
 
       if (rmesa->state.scissor.pClipRects) {
@@ -621,9 +608,6 @@ radeonMakeCurrent( __DRIcontextPrivate *driContextPriv,
 			  (GLframebuffer *) driDrawPriv->driverPrivate,
 			  (GLframebuffer *) driReadPriv->driverPrivate );
 
-      if (newCtx->vb.enabled)
-	 radeonVtxfmtMakeCurrent( newCtx->glCtx );
-
    } else {
       if (RADEON_DEBUG & DEBUG_DRI)
 	 fprintf(stderr, "%s ctx is null\n", __FUNCTION__);
diff --git a/src/mesa/drivers/dri/radeon/radeon_context.h b/src/mesa/drivers/dri/radeon/radeon_context.h
index 0a7c3b2f544..02cea2f4e37 100644
--- a/src/mesa/drivers/dri/radeon/radeon_context.h
+++ b/src/mesa/drivers/dri/radeon/radeon_context.h
@@ -529,7 +529,6 @@ struct radeon_store {
  */
 struct radeon_tcl_info {
    GLuint vertex_format;
-   GLint last_offset;
    GLuint hw_primitive;
 
    /* Temporary for cases where incoming vertex data is incompatible
@@ -600,75 +599,6 @@ struct radeon_ioctl {
 #define RADEON_MAX_PRIMS 64
 
 
-/* Want to keep a cache of these around.  Each is parameterized by
- * only a single value which has only a small range.  Only expect a
- * few, so just rescan the list each time?
- */
-struct dynfn {
-   struct dynfn *next, *prev;
-   int key;
-   char *code;
-};
-
-struct dfn_lists {
-   struct dynfn Vertex2f;
-   struct dynfn Vertex2fv;
-   struct dynfn Vertex3f;
-   struct dynfn Vertex3fv;
-   struct dynfn Color4ub;
-   struct dynfn Color4ubv;
-   struct dynfn Color3ub;
-   struct dynfn Color3ubv;
-   struct dynfn Color4f;
-   struct dynfn Color4fv;
-   struct dynfn Color3f;
-   struct dynfn Color3fv;
-   struct dynfn SecondaryColor3ubEXT;
-   struct dynfn SecondaryColor3ubvEXT;
-   struct dynfn SecondaryColor3fEXT;
-   struct dynfn SecondaryColor3fvEXT;
-   struct dynfn Normal3f;
-   struct dynfn Normal3fv;
-   struct dynfn TexCoord2f;
-   struct dynfn TexCoord2fv;
-   struct dynfn TexCoord1f;
-   struct dynfn TexCoord1fv;
-   struct dynfn MultiTexCoord2fARB;
-   struct dynfn MultiTexCoord2fvARB;
-   struct dynfn MultiTexCoord1fARB;
-   struct dynfn MultiTexCoord1fvARB;
-};
-
-struct dfn_generators {
-   struct dynfn *(*Vertex2f)( GLcontext *, int );
-   struct dynfn *(*Vertex2fv)( GLcontext *, int );
-   struct dynfn *(*Vertex3f)( GLcontext *, int );
-   struct dynfn *(*Vertex3fv)( GLcontext *, int );
-   struct dynfn *(*Color4ub)( GLcontext *, int );
-   struct dynfn *(*Color4ubv)( GLcontext *, int );
-   struct dynfn *(*Color3ub)( GLcontext *, int );
-   struct dynfn *(*Color3ubv)( GLcontext *, int );
-   struct dynfn *(*Color4f)( GLcontext *, int );
-   struct dynfn *(*Color4fv)( GLcontext *, int );
-   struct dynfn *(*Color3f)( GLcontext *, int );
-   struct dynfn *(*Color3fv)( GLcontext *, int );
-   struct dynfn *(*SecondaryColor3ubEXT)( GLcontext *, int );
-   struct dynfn *(*SecondaryColor3ubvEXT)( GLcontext *, int );
-   struct dynfn *(*SecondaryColor3fEXT)( GLcontext *, int );
-   struct dynfn *(*SecondaryColor3fvEXT)( GLcontext *, int );
-   struct dynfn *(*Normal3f)( GLcontext *, int );
-   struct dynfn *(*Normal3fv)( GLcontext *, int );
-   struct dynfn *(*TexCoord2f)( GLcontext *, int );
-   struct dynfn *(*TexCoord2fv)( GLcontext *, int );
-   struct dynfn *(*TexCoord1f)( GLcontext *, int );
-   struct dynfn *(*TexCoord1fv)( GLcontext *, int );
-   struct dynfn *(*MultiTexCoord2fARB)( GLcontext *, int );
-   struct dynfn *(*MultiTexCoord2fvARB)( GLcontext *, int );
-   struct dynfn *(*MultiTexCoord1fARB)( GLcontext *, int );
-   struct dynfn *(*MultiTexCoord1fvARB)( GLcontext *, int );
-};
-
-
 
 struct radeon_prim {
    GLuint start;
@@ -685,43 +615,6 @@ struct radeon_prim {
  */
 #define RADEON_MAX_VERTEX_SIZE 20
 
-struct radeon_vbinfo {
-   GLint counter, initial_counter;
-   GLint *dmaptr;
-   void (*notify)( void );
-   GLint vertex_size;
-
-   union { float f; int i; radeon_color_t color; } vertex[RADEON_MAX_VERTEX_SIZE];
-
-   GLfloat *normalptr;
-   GLfloat *floatcolorptr;
-   radeon_color_t *colorptr;
-   GLfloat *floatspecptr;
-   radeon_color_t *specptr;
-   GLfloat *texcoordptr[4];	/* 3 (TMU) + 1 for radeon_vtxfmt_c.c when GL_TEXTURE3 */
-
-   GLenum *prim;		/* &ctx->Driver.CurrentExecPrimitive */
-   GLuint primflags;
-   GLboolean enabled;		/* *_NO_VTXFMT / *_NO_TCL env vars */
-   GLboolean installed;
-   GLboolean fell_back;
-   GLboolean recheck;
-   GLint nrverts;
-   GLuint vertex_format;
-
-   GLuint installed_vertex_format;
-   GLuint installed_color_3f_sz;
-
-   struct radeon_prim primlist[RADEON_MAX_PRIMS];
-   int nrprims;
-
-   struct dfn_lists dfn_cache;
-   struct dfn_generators codegen;
-   GLvertexformat vtxfmt;
-};
-
-
-
 
 struct radeon_context {
    GLcontext *glCtx;			/* Mesa context */
@@ -808,10 +701,6 @@ struct radeon_context {
     */
    struct radeon_swtcl_info swtcl;
 
-   /* radeon_vtxfmt.c
-    */
-   struct radeon_vbinfo vb;
-
    /* Mirrors of some DRI state
     */
    struct radeon_dri_mirror dri;
diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.c b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
index 737490fcde0..4c64bc201ae 100644
--- a/src/mesa/drivers/dri/radeon/radeon_ioctl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.c
@@ -1087,10 +1087,6 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
       }
    }
 
-   /* Flip top to bottom */
-   cx += dPriv->x;
-   cy  = dPriv->y + dPriv->h - cy - ch;
-
    LOCK_HARDWARE( rmesa );
 
    /* compute region after locking: */
@@ -1099,6 +1095,10 @@ static void radeonClear( GLcontext *ctx, GLbitfield mask )
    cw = ctx->DrawBuffer->_Xmax - cx;
    ch = ctx->DrawBuffer->_Ymax - cy;
 
+   /* Flip top to bottom */
+   cx += dPriv->x;
+   cy  = dPriv->y + dPriv->h - cy - ch;
+
    /* Throttle the number of clear ioctls we do.
     */
    while ( 1 ) {
diff --git a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
index 91a60bb9f15..65dbecf7a6c 100644
--- a/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
+++ b/src/mesa/drivers/dri/radeon/radeon_maos_verts.c
@@ -37,7 +37,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "imports.h"
 #include "mtypes.h"
 
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "math/m_translate.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
diff --git a/src/mesa/drivers/dri/radeon/radeon_screen.c b/src/mesa/drivers/dri/radeon/radeon_screen.c
index 279357ab590..fc5aa114621 100644
--- a/src/mesa/drivers/dri/radeon/radeon_screen.c
+++ b/src/mesa/drivers/dri/radeon/radeon_screen.c
@@ -55,6 +55,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r200_span.h"
 #elif RADEON_COMMON && defined(RADEON_COMMON_FOR_R300)
 #include "r300_context.h"
+#include "r300_fragprog.h"
 #include "radeon_span.h"
 #endif
 
@@ -163,6 +164,18 @@ DRI_CONF_OPT_BEGIN(disable_lowimpact_fallback,bool,def) \
         DRI_CONF_DESC(en,"Disable Low-impact fallback") \
 DRI_CONF_OPT_END
 
+#define DRI_CONF_DISABLE_DOUBLE_SIDE_STENCIL(def) \
+DRI_CONF_OPT_BEGIN(disable_stencil_two_side,bool,def) \
+        DRI_CONF_DESC(en,"Disable GL_EXT_stencil_two_side") \
+DRI_CONF_OPT_END
+
+#define DRI_CONF_FP_OPTIMIZATION(def) \
+DRI_CONF_OPT_BEGIN_V(fp_optimization,enum,def,"0:1") \
+	DRI_CONF_DESC_BEGIN(en,"Fragment Program optimization") \
+                DRI_CONF_ENUM(0,"Optimize for Speed") \
+                DRI_CONF_ENUM(1,"Optimize for Quality") \
+        DRI_CONF_DESC_END \
+DRI_CONF_OPT_END
 
 const char __driConfigOptions[] =
 DRI_CONF_BEGIN
@@ -174,6 +187,7 @@ DRI_CONF_BEGIN
 		DRI_CONF_MAX_TEXTURE_COORD_UNITS(8, 2, 8)
 		DRI_CONF_COMMAND_BUFFER_SIZE(8, 8, 32)
 		DRI_CONF_DISABLE_FALLBACK(false)
+		DRI_CONF_DISABLE_DOUBLE_SIDE_STENCIL(false)
 	DRI_CONF_SECTION_END
 	DRI_CONF_SECTION_QUALITY
 		DRI_CONF_TEXTURE_DEPTH(DRI_CONF_TEXTURE_DEPTH_FB)
@@ -184,12 +198,13 @@ DRI_CONF_BEGIN
 		DRI_CONF_COLOR_REDUCTION(DRI_CONF_COLOR_REDUCTION_DITHER)
 		DRI_CONF_ROUND_MODE(DRI_CONF_ROUND_TRUNC)
 		DRI_CONF_DITHER_MODE(DRI_CONF_DITHER_XERRORDIFF)
+		DRI_CONF_FP_OPTIMIZATION(DRI_CONF_FP_OPTIMIZATION_SPEED)
 	DRI_CONF_SECTION_END
 	DRI_CONF_SECTION_DEBUG
 		DRI_CONF_NO_RAST(false)
 	DRI_CONF_SECTION_END
 DRI_CONF_END;
-static const GLuint __driNConfigOptions = 16;
+static const GLuint __driNConfigOptions = 18;
 
 #ifndef RADEON_DEBUG
 int RADEON_DEBUG = 0;
diff --git a/src/mesa/drivers/dri/radeon/radeon_state.c b/src/mesa/drivers/dri/radeon/radeon_state.c
index 216d7dca174..e19202fa443 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state.c
@@ -42,7 +42,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "context.h"
 #include "framebuffer.h"
 
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -53,7 +53,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "radeon_tcl.h"
 #include "radeon_tex.h"
 #include "radeon_swtcl.h"
-#include "radeon_vtxfmt.h"
 #include "drirenderbuffer.h"
 
 static void radeonUpdateSpecular( GLcontext *ctx );
@@ -2303,11 +2302,10 @@ static void radeonInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    _ae_invalidate_state( ctx, new_state );
    RADEON_CONTEXT(ctx)->NewGLState |= new_state;
-   radeonVtxfmtInvalidate( ctx );
 }
 
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_state_init.c b/src/mesa/drivers/dri/radeon/radeon_state_init.c
index 5682d95ae10..5fc34f09331 100644
--- a/src/mesa/drivers/dri/radeon/radeon_state_init.c
+++ b/src/mesa/drivers/dri/radeon/radeon_state_init.c
@@ -33,7 +33,7 @@
 #include "api_arrayelt.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -44,7 +44,6 @@
 #include "radeon_tcl.h"
 #include "radeon_tex.h"
 #include "radeon_swtcl.h"
-#include "radeon_vtxfmt.h"
 
 #include "xmlpool.h"
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.c b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
index e36a710d33a..7ce1fa67cf6 100644
--- a/src/mesa/drivers/dri/radeon/radeon_swtcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.c
@@ -45,7 +45,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
-#include "tnl/t_vtx_api.h"	/* for _tnl_FlushVertices */
 
 #include "radeon_context.h"
 #include "radeon_ioctl.h"
@@ -848,14 +847,6 @@ void radeonFallback( GLcontext *ctx, GLuint bit, GLboolean mode )
 }
 
 
-void radeonFlushVertices( GLcontext *ctx, GLuint flags )
-{
-   _tnl_FlushVertices( ctx, flags );
-
-   if (flags & FLUSH_STORED_VERTICES)
-      RADEON_NEWPRIM( RADEON_CONTEXT( ctx ) );
-}
-
 /**********************************************************************/
 /*                            Initialization.                         */
 /**********************************************************************/
diff --git a/src/mesa/drivers/dri/radeon/radeon_swtcl.h b/src/mesa/drivers/dri/radeon/radeon_swtcl.h
index f95a52c0b8e..64f9019513d 100644
--- a/src/mesa/drivers/dri/radeon/radeon_swtcl.h
+++ b/src/mesa/drivers/dri/radeon/radeon_swtcl.h
@@ -43,7 +43,6 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 extern void radeonInitSwtcl( GLcontext *ctx );
 extern void radeonDestroySwtcl( GLcontext *ctx );
 
-extern void radeonFlushVertices( GLcontext *ctx, GLuint flags );
 extern void radeonChooseRenderState( GLcontext *ctx );
 extern void radeonChooseVertexState( GLcontext *ctx );
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_tcl.c b/src/mesa/drivers/dri/radeon/radeon_tcl.c
index ffd49b525b8..5ad044c2628 100644
--- a/src/mesa/drivers/dri/radeon/radeon_tcl.c
+++ b/src/mesa/drivers/dri/radeon/radeon_tcl.c
@@ -39,7 +39,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "mtypes.h"
 #include "enums.h"
 
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c b/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c
deleted file mode 100644
index a5a9eb144b5..00000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt.c
+++ /dev/null
@@ -1,1086 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt.c,v 1.6 2003/05/06 23:52:08 daenzer Exp $ */
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-#include "glheader.h"
-#include "imports.h"
-#include "api_noop.h"
-#include "api_arrayelt.h"
-#include "context.h"
-#include "mtypes.h"
-#include "enums.h"
-#include "glapi.h"
-#include "colormac.h"
-#include "light.h"
-#include "state.h"
-#include "vtxfmt.h"
-
-#include "tnl/tnl.h"
-#include "tnl/t_context.h"
-#include "tnl/t_array_api.h"
-#include "tnl/t_save_api.h"
-
-#include "radeon_context.h"
-#include "radeon_state.h"
-#include "radeon_ioctl.h"
-#include "radeon_tex.h"
-#include "radeon_tcl.h"
-#include "radeon_swtcl.h"
-#include "radeon_vtxfmt.h"
-
-#define VERT_ATTRIB_TEX(u)	(VERT_ATTRIB_TEX0 + (u))
-
-#include "dispatch.h"
-
-static void radeonVtxfmtFlushVertices( GLcontext *, GLuint );
-
-static void count_func( const char *name,  struct dynfn *l )
-{
-   int i = 0;
-   struct dynfn *f;
-   foreach (f, l) i++;
-   if (i) fprintf(stderr, "%s: %d\n", name, i );
-}
-
-static void count_funcs( radeonContextPtr rmesa )
-{
-   count_func( "Vertex2f", &rmesa->vb.dfn_cache.Vertex2f );
-   count_func( "Vertex2fv", &rmesa->vb.dfn_cache.Vertex2fv );
-   count_func( "Vertex3f", &rmesa->vb.dfn_cache.Vertex3f );
-   count_func( "Vertex3fv", &rmesa->vb.dfn_cache.Vertex3fv );
-   count_func( "Color4ub", &rmesa->vb.dfn_cache.Color4ub );
-   count_func( "Color4ubv", &rmesa->vb.dfn_cache.Color4ubv );
-   count_func( "Color3ub", &rmesa->vb.dfn_cache.Color3ub );
-   count_func( "Color3ubv", &rmesa->vb.dfn_cache.Color3ubv );
-   count_func( "Color4f", &rmesa->vb.dfn_cache.Color4f );
-   count_func( "Color4fv", &rmesa->vb.dfn_cache.Color4fv );
-   count_func( "Color3f", &rmesa->vb.dfn_cache.Color3f );
-   count_func( "Color3fv", &rmesa->vb.dfn_cache.Color3fv );
-   count_func( "SecondaryColor3f", &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   count_func( "SecondaryColor3fv", &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   count_func( "SecondaryColor3ub", &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   count_func( "SecondaryColor3ubv", &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   count_func( "Normal3f", &rmesa->vb.dfn_cache.Normal3f );
-   count_func( "Normal3fv", &rmesa->vb.dfn_cache.Normal3fv );
-   count_func( "TexCoord2f", &rmesa->vb.dfn_cache.TexCoord2f );
-   count_func( "TexCoord2fv", &rmesa->vb.dfn_cache.TexCoord2fv );
-   count_func( "TexCoord1f", &rmesa->vb.dfn_cache.TexCoord1f );
-   count_func( "TexCoord1fv", &rmesa->vb.dfn_cache.TexCoord1fv );
-   count_func( "MultiTexCoord2fARB", &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   count_func( "MultiTexCoord2fvARB", &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   count_func( "MultiTexCoord1fARB", &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   count_func( "MultiTexCoord1fvARB", &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-}
-
-
-void radeon_copy_to_current( GLcontext *ctx ) 
-{
-   GLuint unit;
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   assert(ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT);
-
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_N0) {
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][0] = rmesa->vb.normalptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][1] = rmesa->vb.normalptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_NORMAL][2] = rmesa->vb.normalptr[2];
-   }
-
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_PKCOLOR) {
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->red );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->green );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->blue );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT( rmesa->vb.colorptr->alpha );
-   } 
-   
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_FPCOLOR) {
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] = rmesa->vb.floatcolorptr[0];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] = rmesa->vb.floatcolorptr[1];
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] = rmesa->vb.floatcolorptr[2];
-   }
-
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_FPALPHA)
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = rmesa->vb.floatcolorptr[3];
-      
-   if (rmesa->vb.vertex_format & RADEON_CP_VC_FRMT_PKSPEC) {
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][0] = UBYTE_TO_FLOAT( rmesa->vb.specptr->red );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][1] = UBYTE_TO_FLOAT( rmesa->vb.specptr->green );
-      ctx->Current.Attrib[VERT_ATTRIB_COLOR1][2] = UBYTE_TO_FLOAT( rmesa->vb.specptr->blue );
-   } 
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (rmesa->vb.vertex_format & RADEON_ST_BIT(unit)) {
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][0] = rmesa->vb.texcoordptr[unit][0];
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][1] = rmesa->vb.texcoordptr[unit][1];
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][2] = 0.0F;
-	 ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][3] = 1.0F;
-      }
-   }
-
-   ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-}
-
-static GLboolean discreet_gl_prim[GL_POLYGON+1] = {
-   1,				/* 0 points */
-   1,				/* 1 lines */
-   0,				/* 2 line_strip */
-   0,				/* 3 line_loop */
-   1,				/* 4 tris */
-   0,				/* 5 tri_fan */
-   0,				/* 6 tri_strip */
-   1,				/* 7 quads */
-   0,				/* 8 quadstrip */
-   0,				/* 9 poly */
-};
-
-static void flush_prims( radeonContextPtr rmesa )
-{
-   int i,j;
-   struct radeon_dma_region tmp = rmesa->dma.current;
-   
-   tmp.buf->refcount++;
-   tmp.aos_size = rmesa->vb.vertex_size;
-   tmp.aos_stride = rmesa->vb.vertex_size;
-   tmp.aos_start = GET_START(&tmp);
-
-   rmesa->dma.current.ptr = rmesa->dma.current.start += 
-      (rmesa->vb.initial_counter - rmesa->vb.counter) * rmesa->vb.vertex_size * 4; 
-
-   rmesa->tcl.vertex_format = rmesa->vb.vertex_format;
-   rmesa->tcl.aos_components[0] = &tmp;
-   rmesa->tcl.nr_aos_components = 1;
-   rmesa->dma.flush = NULL;
-
-   /* Optimize the primitive list:
-    */
-   if (rmesa->vb.nrprims > 1) {
-      for (j = 0, i = 1 ; i < rmesa->vb.nrprims; i++) {
-	 int pj = rmesa->vb.primlist[j].prim & 0xf;
-	 int pi = rmesa->vb.primlist[i].prim & 0xf;
-      
-	 if (pj == pi && discreet_gl_prim[pj] &&
-	     rmesa->vb.primlist[i].start == rmesa->vb.primlist[j].end) {
-	    rmesa->vb.primlist[j].end = rmesa->vb.primlist[i].end;
-	 }
-	 else {
-	    j++;
-	    if (j != i) rmesa->vb.primlist[j] = rmesa->vb.primlist[i];
-	 }
-      }
-      rmesa->vb.nrprims = j+1;
-   }
-
-   for (i = 0 ; i < rmesa->vb.nrprims; i++) {
-      if (RADEON_DEBUG & DEBUG_PRIMS)
-	 fprintf(stderr, "vtxfmt prim %d: %s %d..%d\n", i,
-		 _mesa_lookup_enum_by_nr( rmesa->vb.primlist[i].prim & 
-					  PRIM_MODE_MASK ),
-		 rmesa->vb.primlist[i].start,
-		 rmesa->vb.primlist[i].end);
-
-      radeonEmitPrimitive( rmesa->glCtx,
-			   rmesa->vb.primlist[i].start,
-			   rmesa->vb.primlist[i].end,
-			   rmesa->vb.primlist[i].prim );
-   }
-
-   rmesa->vb.nrprims = 0;
-   radeonReleaseDmaRegion( rmesa, &tmp, __FUNCTION__ );
-}
-
-
-static void start_prim( radeonContextPtr rmesa, GLuint mode )
-{
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   rmesa->vb.primlist[rmesa->vb.nrprims].start = rmesa->vb.initial_counter - rmesa->vb.counter;
-   rmesa->vb.primlist[rmesa->vb.nrprims].prim = mode;
-}
-
-static void note_last_prim( radeonContextPtr rmesa, GLuint flags )
-{
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d\n", __FUNCTION__, rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      rmesa->vb.primlist[rmesa->vb.nrprims].prim |= flags;
-      rmesa->vb.primlist[rmesa->vb.nrprims].end = rmesa->vb.initial_counter - rmesa->vb.counter;
-
-      if (++(rmesa->vb.nrprims) == RADEON_MAX_PRIMS)
-	 flush_prims( rmesa );
-   }
-}
-
-
-static void copy_vertex( radeonContextPtr rmesa, GLuint n, GLfloat *dst )
-{
-   GLuint i;
-   GLfloat *src = (GLfloat *)(rmesa->dma.current.address + 
-			      rmesa->dma.current.ptr + 
-			      (rmesa->vb.primlist[rmesa->vb.nrprims].start + n) * 
-			      rmesa->vb.vertex_size * 4);
-
-   if (RADEON_DEBUG & DEBUG_VFMT) 
-      fprintf(stderr, "copy_vertex %d\n", rmesa->vb.primlist[rmesa->vb.nrprims].start + n);
-
-   for (i = 0 ; i < rmesa->vb.vertex_size; i++) {
-      dst[i] = src[i];
-   }
-}
-
-/* NOTE: This actually reads the copied vertices back from uncached
- * memory.  Could also use the counter/notify mechanism to populate
- * tmp on the fly as vertices are generated.  
- */
-static GLuint copy_dma_verts( radeonContextPtr rmesa, GLfloat (*tmp)[RADEON_MAX_VERTEX_SIZE] )
-{
-   GLuint ovf, i;
-   GLuint nr = (rmesa->vb.initial_counter - rmesa->vb.counter) - rmesa->vb.primlist[rmesa->vb.nrprims].start;
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s %d verts\n", __FUNCTION__, nr);
-
-   switch( rmesa->vb.prim[0] )
-   {
-   case GL_POINTS:
-      return 0;
-   case GL_LINES:
-      ovf = nr&1;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_TRIANGLES:
-      ovf = nr%3;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_QUADS:
-      ovf = nr&3;
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_LINE_STRIP:
-      if (nr == 0) 
-	 return 0;
-      copy_vertex( rmesa, nr-1, tmp[0] );
-      return 1;
-   case GL_LINE_LOOP:
-   case GL_TRIANGLE_FAN:
-   case GL_POLYGON:
-      if (nr == 0) 
-	 return 0;
-      else if (nr == 1) {
-	 copy_vertex( rmesa, 0, tmp[0] );
-	 return 1;
-      } else {
-	 copy_vertex( rmesa, 0, tmp[0] );
-	 copy_vertex( rmesa, nr-1, tmp[1] );
-	 return 2;
-      }
-   case GL_TRIANGLE_STRIP:
-      ovf = MIN2(nr, 2);
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   case GL_QUAD_STRIP:
-      switch (nr) {
-      case 0: ovf = 0; break;
-      case 1: ovf = 1; break;
-      default: ovf = 2 + (nr&1); break;
-      }
-      for (i = 0 ; i < ovf ; i++)
-	 copy_vertex( rmesa, nr-ovf+i, tmp[i] );
-      return i;
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
-static void VFMT_FALLBACK_OUTSIDE_BEGIN_END( const char *caller )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_FALLBACKS))
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-   if (ctx->Driver.NeedFlush) 
-      radeonVtxfmtFlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   if (ctx->NewState)
-      _mesa_update_state( ctx ); /* clear state so fell_back sticks */
-
-   _tnl_wakeup_exec( ctx );
-   ctx->Driver.FlushVertices = radeonFlushVertices;
-   ctx->Driver.NewList =_tnl_NewList;
-
-   assert( rmesa->dma.flush == 0 );
-   rmesa->vb.fell_back = GL_TRUE;
-   rmesa->vb.installed = GL_FALSE;
-}
-
-
-static void VFMT_FALLBACK( const char *caller )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat tmp[3][RADEON_MAX_VERTEX_SIZE];
-   GLuint i, prim;
-   GLuint ind = rmesa->vb.vertex_format;
-   GLuint nrverts;
-   GLfloat alpha = 1.0;
-   GLuint unit;
-
-   if (RADEON_DEBUG & (DEBUG_FALLBACKS|DEBUG_VFMT))
-      fprintf(stderr, "%s from %s\n", __FUNCTION__, caller);
-
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) {
-      VFMT_FALLBACK_OUTSIDE_BEGIN_END( __FUNCTION__ );
-      return;
-   }
-
-   /* Copy vertices out of dma:
-    */
-   nrverts = copy_dma_verts( rmesa, tmp );
-
-   /* Finish the prim at this point:
-    */
-   note_last_prim( rmesa, 0 );
-   flush_prims( rmesa );
-
-   /* Update ctx->Driver.CurrentExecPrimitive and swap in swtnl. 
-    */
-   prim = rmesa->vb.prim[0];
-   ctx->Driver.CurrentExecPrimitive = GL_POLYGON+1;
-   _tnl_wakeup_exec( ctx );
-   ctx->Driver.FlushVertices = radeonFlushVertices;
-
-   assert(rmesa->dma.flush == 0);
-   rmesa->vb.fell_back = GL_TRUE;
-   rmesa->vb.installed = GL_FALSE;
-   CALL_Begin(GET_DISPATCH(), (prim));
-   
-   if (rmesa->vb.installed_color_3f_sz == 4)
-      alpha = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3];
-
-   /* Replay saved vertices
-    */
-   for (i = 0 ; i < nrverts; i++) {
-      GLuint offset = 3;
-      if (ind & RADEON_CP_VC_FRMT_N0) {
-	 CALL_Normal3fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset += 3;
-      }
-
-      if (ind & RADEON_CP_VC_FRMT_PKCOLOR) {
-	 radeon_color_t *col = (radeon_color_t *)&tmp[i][offset];
-	 CALL_Color4ub(GET_DISPATCH(), (col->red, col->green, col->blue, col->alpha));
-	 offset++;
-      }
-      else if (ind & RADEON_CP_VC_FRMT_FPALPHA) {
-	 CALL_Color4fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset+=4;
-      } 
-      else if (ind & RADEON_CP_VC_FRMT_FPCOLOR) {
-	 CALL_Color3fv(GET_DISPATCH(), (&tmp[i][offset]));
-	 offset+=3;
-      }
-
-      if (ind & RADEON_CP_VC_FRMT_PKSPEC) {
-	 radeon_color_t *spec = (radeon_color_t *)&tmp[i][offset];
-	 CALL_SecondaryColor3ubEXT(GET_DISPATCH(), (spec->red, spec->green, spec->blue));
-	 offset++;
-      }
-
-      for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-	 if (ind & RADEON_ST_BIT(unit)) {
-	    CALL_MultiTexCoord2fvARB(GET_DISPATCH(), ((GL_TEXTURE0 + unit), &tmp[i][offset]));
-	    offset += 2;
-	 }
-      }
-      CALL_Vertex3fv(GET_DISPATCH(), (&tmp[i][0]));
-   }
-
-   /* Replay current vertex
-    */
-   if (ind & RADEON_CP_VC_FRMT_N0) 
-      CALL_Normal3fv(GET_DISPATCH(), (rmesa->vb.normalptr));
-
-   if (ind & RADEON_CP_VC_FRMT_PKCOLOR)
-      CALL_Color4ub(GET_DISPATCH(), (rmesa->vb.colorptr->red, rmesa->vb.colorptr->green,
-      				     rmesa->vb.colorptr->blue, rmesa->vb.colorptr->alpha));
-   else if (ind & RADEON_CP_VC_FRMT_FPALPHA)
-      CALL_Color4fv(GET_DISPATCH(), (rmesa->vb.floatcolorptr));
-   else if (ind & RADEON_CP_VC_FRMT_FPCOLOR) {
-      if (rmesa->vb.installed_color_3f_sz == 4 && alpha != 1.0)
-	 CALL_Color4f(GET_DISPATCH(), (rmesa->vb.floatcolorptr[0],
-				       rmesa->vb.floatcolorptr[1],
-				       rmesa->vb.floatcolorptr[2],
-				       alpha));
-      else
-	 CALL_Color3fv(GET_DISPATCH(), (rmesa->vb.floatcolorptr));
-   }
-
-   if (ind & RADEON_CP_VC_FRMT_PKSPEC) 
-       CALL_SecondaryColor3ubEXT(GET_DISPATCH(), (rmesa->vb.specptr->red,
-       						  rmesa->vb.specptr->green,
-						  rmesa->vb.specptr->blue));
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (ind & RADEON_ST_BIT(unit)) {
-	 CALL_MultiTexCoord2fvARB(GET_DISPATCH(), ((GL_TEXTURE0 + unit),
-	 			  rmesa->vb.texcoordptr[unit]));
-      }
-   }
-}
-
-static void radeonNewList( GLcontext *ctx, GLuint list, GLenum mode )
-{
-   VFMT_FALLBACK( __FUNCTION__ );
-   _tnl_NewList( ctx, list, mode );
-   return;
-}
-
-
-static void wrap_buffer( void )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat tmp[3][RADEON_MAX_VERTEX_SIZE];
-   GLuint i, nrverts;
-
-   if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_PRIMS))
-      fprintf(stderr, "%s %d\n", __FUNCTION__, rmesa->vb.initial_counter - rmesa->vb.counter);
-
-   /* Don't deal with parity.
-    */
-   if ((((rmesa->vb.initial_counter - rmesa->vb.counter) -  
-	 rmesa->vb.primlist[rmesa->vb.nrprims].start) & 1)) {
-      rmesa->vb.counter++;
-      rmesa->vb.initial_counter++;
-      return;
-   }
-
-   /* Copy vertices out of dma:
-    */
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) 
-      nrverts = 0;
-   else {
-      nrverts = copy_dma_verts( rmesa, tmp );
-
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%d vertices to copy\n", nrverts);
-   
-      /* Finish the prim at this point:
-       */
-      note_last_prim( rmesa, 0 );
-   }
-
-   /* Fire any buffered primitives
-    */
-   flush_prims( rmesa );
-
-   /* Get new buffer
-    */
-   radeonRefillCurrentDmaRegion( rmesa );
-
-   /* Reset counter, dmaptr
-    */
-   rmesa->vb.dmaptr = (int *)(rmesa->dma.current.ptr + rmesa->dma.current.address);
-   rmesa->vb.counter = (rmesa->dma.current.end - rmesa->dma.current.ptr) / 
-      (rmesa->vb.vertex_size * 4);
-   rmesa->vb.counter--;
-   rmesa->vb.initial_counter = rmesa->vb.counter;
-   rmesa->vb.notify = wrap_buffer;
-
-   rmesa->dma.flush = flush_prims;
-
-   /* Restart wrapped primitive:
-    */
-   if (rmesa->vb.prim[0] != GL_POLYGON+1)
-      start_prim( rmesa, rmesa->vb.prim[0] );
-
-   /* Reemit saved vertices
-    */
-   for (i = 0 ; i < nrverts; i++) {
-      if (RADEON_DEBUG & DEBUG_VERTS) {
-	 int j;
-	 fprintf(stderr, "re-emit vertex %d to %p\n", i, (void *)rmesa->vb.dmaptr);
-	 if (RADEON_DEBUG & DEBUG_VERBOSE)
-	    for (j = 0 ; j < rmesa->vb.vertex_size; j++) 
-	       fprintf(stderr, "\t%08x/%f\n", *(int*)&tmp[i][j], tmp[i][j]);
-      }
-
-      memcpy( rmesa->vb.dmaptr, tmp[i], rmesa->vb.vertex_size * 4 );
-      rmesa->vb.dmaptr += rmesa->vb.vertex_size;
-      rmesa->vb.counter--;
-   }
-}
-
-
-
-static GLboolean check_vtx_fmt( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLuint ind = RADEON_CP_VC_FRMT_Z;
-   GLuint unit;
-
-   if (rmesa->TclFallback || rmesa->vb.fell_back || ctx->CompileFlag ||
-      (ctx->Fog.Enabled && (ctx->Fog.FogCoordinateSource == GL_FOG_COORD)))
-      return GL_FALSE;
-
-   if (ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) 
-      ctx->Driver.FlushVertices( ctx, FLUSH_UPDATE_CURRENT );
-   
-   /* Make all this event-driven:
-    */
-   if (ctx->Light.Enabled) {
-      ind |= RADEON_CP_VC_FRMT_N0;
-
-      /* TODO: make this data driven: If we receive only ubytes, send
-       * color as ubytes.  Also check if converting (with free
-       * checking for overflow) is cheaper than sending floats
-       * directly.
-       */
-      if (ctx->Light.ColorMaterialEnabled) {
-	 ind |= (RADEON_CP_VC_FRMT_FPCOLOR |
-		 RADEON_CP_VC_FRMT_FPALPHA);
-      }
-      else
-	 ind |= RADEON_CP_VC_FRMT_PKCOLOR; /* for alpha? */
-   }
-   else {
-      /* TODO: make this data driven?
-       */
-      ind |= RADEON_CP_VC_FRMT_PKCOLOR;
-	 
-      if (ctx->_TriangleCaps & DD_SEPARATE_SPECULAR) {
-	 ind |= RADEON_CP_VC_FRMT_PKSPEC;
-      }
-   }
-
-   if ( ctx->Fog.FogCoordinateSource == GL_FOG_COORD ) {
-      ind |= RADEON_CP_VC_FRMT_PKSPEC;
-   }
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (ctx->Texture.Unit[unit]._ReallyEnabled) {
-	 if (ctx->Texture.Unit[unit].TexGenEnabled) {
-	    if (rmesa->TexGenNeedNormals[unit]) {
-	       ind |= RADEON_CP_VC_FRMT_N0;
-	    }
-	 } else {
-	    if (ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][3] != 1.0) {
-	       if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_FALLBACKS))
-		  fprintf(stderr, "%s: q%u\n", __FUNCTION__, unit);
-	       return GL_FALSE;
-	    }
-	    ind |= RADEON_ST_BIT(unit);
-	 }
-      }
-   }
-
-   if (RADEON_DEBUG & (DEBUG_VFMT|DEBUG_STATE))
-      fprintf(stderr, "%s: format: 0x%x\n", __FUNCTION__, ind );
-
-   RADEON_NEWPRIM(rmesa);
-   rmesa->vb.vertex_format = ind;
-   rmesa->vb.vertex_size = 3;
-   rmesa->vb.prim = &ctx->Driver.CurrentExecPrimitive;
-
-   rmesa->vb.normalptr = ctx->Current.Attrib[VERT_ATTRIB_NORMAL];
-   rmesa->vb.colorptr = NULL;
-   rmesa->vb.floatcolorptr = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
-   rmesa->vb.specptr = NULL;
-   rmesa->vb.floatspecptr = ctx->Current.Attrib[VERT_ATTRIB_COLOR1];
-   rmesa->vb.texcoordptr[0] = ctx->Current.Attrib[VERT_ATTRIB_TEX0];
-   rmesa->vb.texcoordptr[1] = ctx->Current.Attrib[VERT_ATTRIB_TEX1];
-   rmesa->vb.texcoordptr[2] = ctx->Current.Attrib[VERT_ATTRIB_TEX2];
-   rmesa->vb.texcoordptr[3] = ctx->Current.Attrib[VERT_ATTRIB_TEX0]; /* dummy */
-
-   /* Run through and initialize the vertex components in the order
-    * the hardware understands:
-    */
-   if (ind & RADEON_CP_VC_FRMT_N0) {
-      rmesa->vb.normalptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 3;
-      rmesa->vb.normalptr[0] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][0];
-      rmesa->vb.normalptr[1] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][1];
-      rmesa->vb.normalptr[2] = ctx->Current.Attrib[VERT_ATTRIB_NORMAL][2];
-   }
-
-   if (ind & RADEON_CP_VC_FRMT_PKCOLOR) {
-      rmesa->vb.colorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].color;
-      rmesa->vb.vertex_size += 1;
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->red,   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->green, ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->blue,  ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.colorptr->alpha, ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] );
-   }
-
-   if (ind & RADEON_CP_VC_FRMT_FPCOLOR) {
-      assert(!(ind & RADEON_CP_VC_FRMT_PKCOLOR));
-      rmesa->vb.floatcolorptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-      rmesa->vb.vertex_size += 3;
-      rmesa->vb.floatcolorptr[0] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][0];
-      rmesa->vb.floatcolorptr[1] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][1];
-      rmesa->vb.floatcolorptr[2] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][2];
-
-      if (ind & RADEON_CP_VC_FRMT_FPALPHA) {
-	 rmesa->vb.vertex_size += 1;
-	 rmesa->vb.floatcolorptr[3] = ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3];
-      }
-   }
-   
-   if (ind & RADEON_CP_VC_FRMT_PKSPEC) {
-      rmesa->vb.specptr = &rmesa->vb.vertex[rmesa->vb.vertex_size].color;
-      rmesa->vb.vertex_size += 1;
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->red,   ctx->Current.Attrib[VERT_ATTRIB_COLOR1][0] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->green, ctx->Current.Attrib[VERT_ATTRIB_COLOR1][1] );
-      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->blue,  ctx->Current.Attrib[VERT_ATTRIB_COLOR1][2] );
-      /* fog ??? */
-/*      UNCLAMPED_FLOAT_TO_CHAN( rmesa->vb.specptr->alpha,
-			       radeonComputeFogFactor(ctx->Current.Attrib[VERT_ATTRIB_FOG][0]) ); */
-   }
-
-   for (unit = 0 ; unit < ctx->Const.MaxTextureUnits; unit++) {
-      if (ind & RADEON_ST_BIT(unit)) {
-	 rmesa->vb.texcoordptr[unit] = &rmesa->vb.vertex[rmesa->vb.vertex_size].f;
-	 rmesa->vb.vertex_size += 2;
-	 rmesa->vb.texcoordptr[unit][0] = ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][0];
-	 rmesa->vb.texcoordptr[unit][1] = ctx->Current.Attrib[VERT_ATTRIB_TEX(unit)][1];
-      }
-   }
-
-   if (rmesa->vb.installed_vertex_format != rmesa->vb.vertex_format) {
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "reinstall on vertex_format change\n");
-      _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-      rmesa->vb.installed_vertex_format = rmesa->vb.vertex_format;
-   }
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s -- success\n", __FUNCTION__);
-   
-   return GL_TRUE;
-}
-
-void radeonVtxfmtInvalidate( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   rmesa->vb.recheck = GL_TRUE;
-   rmesa->vb.fell_back = GL_FALSE;
-}
-
-
-static void radeonVtxfmtValidate( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (ctx->Driver.NeedFlush)
-      ctx->Driver.FlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   rmesa->vb.recheck = GL_FALSE;
-
-   if (check_vtx_fmt( ctx )) {
-      if (!rmesa->vb.installed) {
-	 if (RADEON_DEBUG & DEBUG_VFMT)
-	    fprintf(stderr, "reinstall (new install)\n");
-
-	 _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-	 ctx->Driver.FlushVertices = radeonVtxfmtFlushVertices;
-	 ctx->Driver.NewList = radeonNewList;
-	 rmesa->vb.installed = GL_TRUE;
-      }
-      else if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: already installed", __FUNCTION__);
-   } 
-   else {
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: failed\n", __FUNCTION__);
-
-      if (rmesa->vb.installed) {
-	 if (rmesa->dma.flush)
-	    rmesa->dma.flush( rmesa );
-	 _tnl_wakeup_exec( ctx );
-	 ctx->Driver.FlushVertices = radeonFlushVertices;
-	 ctx->Driver.NewList =_tnl_NewList;
-	 rmesa->vb.installed = GL_FALSE;
-      }
-   }      
-}
-
-
-
-/* Materials:
- */
-static void radeon_Materialfv( GLenum face, GLenum pname, 
-			       const GLfloat *params )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      VFMT_FALLBACK( __FUNCTION__ );
-      CALL_Materialfv(GET_DISPATCH(), (face, pname, params));
-      return;
-   }
-   _mesa_noop_Materialfv( face, pname, params );
-   radeonUpdateMaterial( ctx );
-}
-
-
-/* Begin/End
- */
-static void radeon_Begin( GLenum mode )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s( %s )\n", __FUNCTION__,
-	      _mesa_lookup_enum_by_nr( mode ));
-
-   if (mode > GL_POLYGON) {
-      _mesa_error( ctx, GL_INVALID_ENUM, "glBegin" );
-      return;
-   }
-
-   if (rmesa->vb.prim[0] != GL_POLYGON+1) {
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glBegin" );
-      return;
-   }
-   
-   if (ctx->NewState) 
-      _mesa_update_state( ctx );
-
-   if (rmesa->NewGLState)
-      radeonValidateState( ctx );
-
-   if (rmesa->vb.recheck) 
-      radeonVtxfmtValidate( ctx );
-
-   if (!rmesa->vb.installed) {
-      CALL_Begin(GET_DISPATCH(), (mode));
-      return;
-   }
-
-
-   if (rmesa->dma.flush && rmesa->vb.counter < 12) {
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s: flush almost-empty buffers\n", __FUNCTION__);
-      flush_prims( rmesa );
-   }
-
-   /* Need to arrange to save vertices here?  Or always copy from dma (yuk)?
-    */
-   if (!rmesa->dma.flush) {
-/* FIXME: what are these constants? */
-      if (rmesa->dma.current.ptr + 12*rmesa->vb.vertex_size*4 > 
-	  rmesa->dma.current.end) {
-	 RADEON_NEWPRIM( rmesa );
-	 radeonRefillCurrentDmaRegion( rmesa );
-      }
-
-      rmesa->vb.dmaptr = (int *)(rmesa->dma.current.address + rmesa->dma.current.ptr);
-      rmesa->vb.counter = (rmesa->dma.current.end - rmesa->dma.current.ptr) / 
-	 (rmesa->vb.vertex_size * 4);
-      rmesa->vb.counter--;
-      rmesa->vb.initial_counter = rmesa->vb.counter;
-      rmesa->vb.notify = wrap_buffer;
-      rmesa->dma.flush = flush_prims;
-      ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   }
-   
-   
-   rmesa->vb.prim[0] = mode;
-   start_prim( rmesa, mode | PRIM_BEGIN );
-}
-
-
-
-static void radeon_End( void )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   if (rmesa->vb.prim[0] == GL_POLYGON+1) {
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glEnd" );
-      return;
-   }
-	  
-   note_last_prim( rmesa, PRIM_END );
-   rmesa->vb.prim[0] = GL_POLYGON+1;
-}
-
-
-/* Fallback on difficult entrypoints:
- */
-#define PRE_LOOPBACK( FUNC )			\
-do {						\
-   if (RADEON_DEBUG & DEBUG_VFMT) 		\
-      fprintf(stderr, "%s\n", __FUNCTION__);	\
-   VFMT_FALLBACK( __FUNCTION__ );		\
-} while (0)
-#define TAG(x) radeon_fallback_##x
-#include "vtxfmt_tmp.h"
-
-
-
-static GLboolean radeonNotifyBegin( GLcontext *ctx, GLenum p )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-   
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(!rmesa->vb.installed);
-
-   if (ctx->NewState) 
-      _mesa_update_state( ctx );
-
-   if (rmesa->NewGLState)
-      radeonValidateState( ctx );
-
-   if (ctx->Driver.NeedFlush)
-      ctx->Driver.FlushVertices( ctx, ctx->Driver.NeedFlush );
-
-   if (rmesa->vb.recheck) 
-      radeonVtxfmtValidate( ctx );
-
-   if (!rmesa->vb.installed) {
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "%s -- failed\n", __FUNCTION__);
-      return GL_FALSE;
-   }
-
-   radeon_Begin( p );
-   return GL_TRUE;
-}
-
-static void radeonVtxfmtFlushVertices( GLcontext *ctx, GLuint flags )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   if (RADEON_DEBUG & DEBUG_VFMT)
-      fprintf(stderr, "%s\n", __FUNCTION__);
-
-   assert(rmesa->vb.installed);
-
-   if (flags & FLUSH_UPDATE_CURRENT) {
-      radeon_copy_to_current( ctx );
-      if (RADEON_DEBUG & DEBUG_VFMT)
-	 fprintf(stderr, "reinstall on update_current\n");
-      _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );
-      ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-   }
-
-   if (flags & FLUSH_STORED_VERTICES) {
-      radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-      assert (rmesa->dma.flush == 0 ||
-	      rmesa->dma.flush == flush_prims);
-      if (rmesa->dma.flush == flush_prims)
-	 flush_prims( RADEON_CONTEXT( ctx ) );
-      ctx->Driver.NeedFlush &= ~FLUSH_STORED_VERTICES;
-   }
-}
-
-
-
-/* At this point, don't expect very many versions of each function to
- * be generated, so not concerned about freeing them?
- */
-
-
-void radeonVtxfmtInit( GLcontext *ctx, GLboolean useCodegen )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-   GLvertexformat *vfmt = &(rmesa->vb.vtxfmt);
-
-   /* start by initializing to no-op functions */
-   _mesa_noop_vtxfmt_init(vfmt);
-
-   /* Hook in chooser functions for codegen, etc:
-    */
-   radeonVtxfmtInitChoosers( vfmt );
-
-   /* Handled fully in supported states, but no codegen:
-    */
-   vfmt->Materialfv = radeon_Materialfv;
-   vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = radeon_Begin;
-   vfmt->End = radeon_End;
-
-   /* Fallback for performance reasons:  (Fix with cva/elt path here and
-    * dmatmp2.h style primitive-merging)
-    *
-    * These should call NotifyBegin(), as should _tnl_EvalMesh, to allow
-    * a driver-hook.
-    */
-   vfmt->DrawArrays = radeon_fallback_DrawArrays;
-   vfmt->DrawElements = radeon_fallback_DrawElements;
-   vfmt->DrawRangeElements = radeon_fallback_DrawRangeElements; 
-
-   /* Active but unsupported -- fallback if we receive these:
-    */
-   vfmt->CallList = radeon_fallback_CallList;
-   vfmt->CallLists = radeon_fallback_CallLists;
-   vfmt->EvalCoord1f = radeon_fallback_EvalCoord1f;
-   vfmt->EvalCoord1fv = radeon_fallback_EvalCoord1fv;
-   vfmt->EvalCoord2f = radeon_fallback_EvalCoord2f;
-   vfmt->EvalCoord2fv = radeon_fallback_EvalCoord2fv;
-   vfmt->EvalMesh1 = radeon_fallback_EvalMesh1;
-   vfmt->EvalMesh2 = radeon_fallback_EvalMesh2;
-   vfmt->EvalPoint1 = radeon_fallback_EvalPoint1;
-   vfmt->EvalPoint2 = radeon_fallback_EvalPoint2;
-   vfmt->TexCoord3f = radeon_fallback_TexCoord3f;
-   vfmt->TexCoord3fv = radeon_fallback_TexCoord3fv;
-   vfmt->TexCoord4f = radeon_fallback_TexCoord4f;
-   vfmt->TexCoord4fv = radeon_fallback_TexCoord4fv;
-   vfmt->MultiTexCoord3fARB = radeon_fallback_MultiTexCoord3fARB;
-   vfmt->MultiTexCoord3fvARB = radeon_fallback_MultiTexCoord3fvARB;
-   vfmt->MultiTexCoord4fARB = radeon_fallback_MultiTexCoord4fARB;
-   vfmt->MultiTexCoord4fvARB = radeon_fallback_MultiTexCoord4fvARB;
-   vfmt->Vertex4f = radeon_fallback_Vertex4f;
-   vfmt->Vertex4fv = radeon_fallback_Vertex4fv;
-   vfmt->VertexAttrib1fNV  = radeon_fallback_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = radeon_fallback_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV  = radeon_fallback_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = radeon_fallback_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV  = radeon_fallback_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = radeon_fallback_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV  = radeon_fallback_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = radeon_fallback_VertexAttrib4fvNV;
-   vfmt->FogCoordfEXT = radeon_fallback_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = radeon_fallback_FogCoordfvEXT;
-
-   (void)radeon_fallback_vtxfmt;
-
-   TNL_CONTEXT(ctx)->Driver.NotifyBegin = radeonNotifyBegin;
-
-   rmesa->vb.enabled = 1;
-   rmesa->vb.prim = &ctx->Driver.CurrentExecPrimitive;
-   rmesa->vb.primflags = 0;
-
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex2f );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex2fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Vertex3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4ub );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4ubv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3ub );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3ubv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4f );
-   make_empty_list( &rmesa->vb.dfn_cache.Color4fv );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Color3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   make_empty_list( &rmesa->vb.dfn_cache.Normal3f );
-   make_empty_list( &rmesa->vb.dfn_cache.Normal3fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord2f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord2fv );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord1f );
-   make_empty_list( &rmesa->vb.dfn_cache.TexCoord1fv );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   make_empty_list( &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-
-   radeonInitCodegen( &rmesa->vb.codegen, useCodegen );
-}
-
-static void free_funcs( struct dynfn *l )
-{
-   struct dynfn *f, *tmp;
-   foreach_s (f, tmp, l) {
-      remove_from_list( f );
-      _mesa_exec_free( f->code );
-      _mesa_free( f );
-   }
-}
-
-
-
-void radeonVtxfmtMakeCurrent( GLcontext *ctx )
-{
-}
-
-
-void radeonVtxfmtDestroy( GLcontext *ctx )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT( ctx );
-
-   count_funcs( rmesa );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex2f );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex2fv );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex3f );
-   free_funcs( &rmesa->vb.dfn_cache.Vertex3fv );
-   free_funcs( &rmesa->vb.dfn_cache.Color4ub );
-   free_funcs( &rmesa->vb.dfn_cache.Color4ubv );
-   free_funcs( &rmesa->vb.dfn_cache.Color3ub );
-   free_funcs( &rmesa->vb.dfn_cache.Color3ubv );
-   free_funcs( &rmesa->vb.dfn_cache.Color4f );
-   free_funcs( &rmesa->vb.dfn_cache.Color4fv );
-   free_funcs( &rmesa->vb.dfn_cache.Color3f );
-   free_funcs( &rmesa->vb.dfn_cache.Color3fv );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3ubEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3ubvEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3fEXT );
-   free_funcs( &rmesa->vb.dfn_cache.SecondaryColor3fvEXT );
-   free_funcs( &rmesa->vb.dfn_cache.Normal3f );
-   free_funcs( &rmesa->vb.dfn_cache.Normal3fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord2f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord2fv );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord1f );
-   free_funcs( &rmesa->vb.dfn_cache.TexCoord1fv );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord1fARB );
-   free_funcs( &rmesa->vb.dfn_cache.MultiTexCoord1fvARB );
-}
-
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h b/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h
deleted file mode 100644
index a656e49e445..00000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt.h
+++ /dev/null
@@ -1,120 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt.h,v 1.3 2002/12/21 17:02:16 dawes Exp $ */
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#ifndef __RADEON_VTXFMT_H__
-#define __RADEON_VTXFMT_H__
-
-#include "radeon_context.h"
-
-
-extern void radeonVtxfmtUpdate( GLcontext *ctx );
-extern void radeonVtxfmtInit( GLcontext *ctx, GLboolean useCodegen );
-extern void radeonVtxfmtInvalidate( GLcontext *ctx );
-extern void radeonVtxfmtDestroy( GLcontext *ctx );
-extern void radeonVtxfmtInitChoosers( GLvertexformat *vfmt );
-
-extern void radeonVtxfmtMakeCurrent( GLcontext *ctx );
-extern void radeonVtxfmtUnbindContext( GLcontext *ctx );
-
-extern void radeon_copy_to_current( GLcontext *ctx );
-
-#define DFN( FUNC, CACHE)				\
-do {							\
-   char *start = (char *)&FUNC;				\
-   char *end = (char *)&FUNC##_end;			\
-   insert_at_head( &CACHE, dfn );			\
-   dfn->key = key;					\
-   dfn->code = _mesa_exec_malloc( end - start );	\
-   _mesa_memcpy(dfn->code, start, end - start);		\
-}							\
-while ( 0 )
-
-#define FIXUP( CODE, OFFSET, CHECKVAL, NEWVAL )	\
-do {						\
-   int *icode = (int *)(CODE+OFFSET);		\
-   assert (*icode == CHECKVAL);			\
-   *icode = (int)NEWVAL;			\
-} while (0)
-
-
-/* Useful for figuring out the offsets:
- */
-#define FIXUP2( CODE, OFFSET, CHECKVAL, NEWVAL )		\
-do {								\
-   while (*(int *)(CODE+OFFSET) != CHECKVAL) OFFSET++;		\
-   fprintf(stderr, "%s/%d CVAL %x OFFSET %d VAL %x\n", __FUNCTION__,	\
-	   __LINE__, CHECKVAL, OFFSET, (int)(NEWVAL));			\
-   *(int *)(CODE+OFFSET) = (int)(NEWVAL);				\
-   OFFSET += 4;							\
-} while (0)
-
-/* 
- */
-void radeonInitCodegen( struct dfn_generators *gen, GLboolean useCodegen );
-void radeonInitX86Codegen( struct dfn_generators *gen );
-void radeonInitSSECodegen( struct dfn_generators *gen );
-
-
-
-/* Defined in radeon_vtxfmt_x86.c
- */
-struct dynfn *radeon_makeX86Vertex2f( GLcontext *, int );
-struct dynfn *radeon_makeX86Vertex2fv( GLcontext *, int );
-struct dynfn *radeon_makeX86Vertex3f( GLcontext *, int );
-struct dynfn *radeon_makeX86Vertex3fv( GLcontext *, int );
-struct dynfn *radeon_makeX86Color4ub( GLcontext *, int );
-struct dynfn *radeon_makeX86Color4ubv( GLcontext *, int );
-struct dynfn *radeon_makeX86Color3ub( GLcontext *, int );
-struct dynfn *radeon_makeX86Color3ubv( GLcontext *, int );
-struct dynfn *radeon_makeX86Color4f( GLcontext *, int );
-struct dynfn *radeon_makeX86Color4fv( GLcontext *, int );
-struct dynfn *radeon_makeX86Color3f( GLcontext *, int );
-struct dynfn *radeon_makeX86Color3fv( GLcontext *, int );
-struct dynfn *radeon_makeX86SecondaryColor3ubEXT( GLcontext *, int );
-struct dynfn *radeon_makeX86SecondaryColor3ubvEXT( GLcontext *, int );
-struct dynfn *radeon_makeX86SecondaryColor3fEXT( GLcontext *, int );
-struct dynfn *radeon_makeX86SecondaryColor3fvEXT( GLcontext *, int );
-struct dynfn *radeon_makeX86Normal3f( GLcontext *, int );
-struct dynfn *radeon_makeX86Normal3fv( GLcontext *, int );
-struct dynfn *radeon_makeX86TexCoord2f( GLcontext *, int );
-struct dynfn *radeon_makeX86TexCoord2fv( GLcontext *, int );
-struct dynfn *radeon_makeX86TexCoord1f( GLcontext *, int );
-struct dynfn *radeon_makeX86TexCoord1fv( GLcontext *, int );
-struct dynfn *radeon_makeX86MultiTexCoord2fARB( GLcontext *, int );
-struct dynfn *radeon_makeX86MultiTexCoord2fvARB( GLcontext *, int );
-struct dynfn *radeon_makeX86MultiTexCoord1fARB( GLcontext *, int );
-struct dynfn *radeon_makeX86MultiTexCoord1fvARB( GLcontext *, int );
-
-#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c b/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c
deleted file mode 100644
index aac029aa0e4..00000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_c.c
+++ /dev/null
@@ -1,924 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt_c.c,v 1.2 2002/12/16 16:18:59 dawes Exp $ */
-/**************************************************************************
-
-Copyright 2002 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-#include "glheader.h"
-#include "mtypes.h"
-#include "colormac.h"
-#include "simple_list.h"
-#include "api_noop.h"
-#include "vtxfmt.h"
-
-#include "radeon_vtxfmt.h"
-
-#include "dispatch.h"
-
-/* Fallback versions of all the entrypoints for situations where
- * codegen isn't available.  This is still a lot faster than the
- * vb/pipeline implementation in Mesa.
- */
-static void radeon_Vertex3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&x;
-   *rmesa->vb.dmaptr++ = *(int *)&y;
-   *rmesa->vb.dmaptr++ = *(int *)&z;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void radeon_Vertex3fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&v[0];
-   *rmesa->vb.dmaptr++ = *(int *)&v[1];
-   *rmesa->vb.dmaptr++ = *(int *)&v[2];
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void radeon_Vertex2f( GLfloat x, GLfloat y )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&x;
-   *rmesa->vb.dmaptr++ = *(int *)&y;
-   *rmesa->vb.dmaptr++ = 0;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = *(int *)&rmesa->vb.vertex[i];
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-static void radeon_Vertex2fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   int i;
-
-   *rmesa->vb.dmaptr++ = *(int *)&v[0];
-   *rmesa->vb.dmaptr++ = *(int *)&v[1];
-   *rmesa->vb.dmaptr++ = 0;
-
-   for (i = 3; i < rmesa->vb.vertex_size; i++)
-      *rmesa->vb.dmaptr++ = rmesa->vb.vertex[i].i;
-   
-   if (--rmesa->vb.counter == 0)
-      rmesa->vb.notify();
-}
-
-
-#if 0
-/* Color for ubyte (packed) color formats:
- */
-static void radeon_Color3ub_ub( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= 0xff;
-}
-
-static void radeon_Color3ubv_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= v[0];
-   dest->green	= v[1];
-   dest->blue	= v[2];
-   dest->alpha	= 0xff;
-}
-
-static void radeon_Color4ub_ub( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= a;
-}
-
-static void radeon_Color4ubv_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   *(GLuint *)rmesa->vb.colorptr = LE32_TO_CPU(*(GLuint *)v);
-}
-#endif /* 0 */
-
-static void radeon_Color3f_ub( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   dest->alpha = 255;
-}
-
-static void radeon_Color3fv_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   dest->alpha = 255;
-}
-
-static void radeon_Color4f_ub( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,   r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, a );
-}
-
-static void radeon_Color4fv_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.colorptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->alpha, v[3] );
-}
-
-
-/* Color for float color+alpha formats:
- */
-#if 0
-static void radeon_Color3ub_4f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = 1.0;
-}
-
-static void radeon_Color3ubv_4f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = 1.0;
-}
-
-static void radeon_Color4ub_4f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = UBYTE_TO_FLOAT(a);
-}
-
-static void radeon_Color4ubv_4f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = UBYTE_TO_FLOAT(v[3]);
-}
-#endif /* 0 */
-
-
-static void radeon_Color3f_4f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = 1.0;		
-}
-
-static void radeon_Color3fv_4f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = 1.0;
-}
-
-static void radeon_Color4f_4f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = a;
-}
-
-static void radeon_Color4fv_4f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = v[3];
-}
-
-
-/* Color for float color formats:
- */
-#if 0
-static void radeon_Color3ub_3f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-}
-
-static void radeon_Color3ubv_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-}
-
-static void radeon_Color4ub_3f( GLubyte r, GLubyte g, GLubyte b, GLubyte a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(a);
-}
-
-static void radeon_Color4ubv_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = UBYTE_TO_FLOAT(v[3]);
-}
-#endif /* 0 */
-
-
-static void radeon_Color3f_3f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-}
-
-static void radeon_Color3fv_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-}
-
-static void radeon_Color4f_3f( GLfloat r, GLfloat g, GLfloat b, GLfloat a )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = a;
-}
-
-static void radeon_Color4fv_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatcolorptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = v[3]; 
-}
-
-
-/* Secondary Color:
- */
-#if 0
-static void radeon_SecondaryColor3ubEXT_ub( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.specptr;
-   dest->red	= r;
-   dest->green	= g;
-   dest->blue	= b;
-   dest->alpha	= 0xff;
-}
-
-static void radeon_SecondaryColor3ubvEXT_ub( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.specptr;
-   dest->red	= v[0];
-   dest->green	= v[1];
-   dest->blue	= v[2];
-   dest->alpha	= 0xff;
-}
-#endif /* 0 */
-
-static void radeon_SecondaryColor3fEXT_ub( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.specptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  r );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, g );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  b );
-   dest->alpha = 255;
-}
-
-static void radeon_SecondaryColor3fvEXT_ub( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   radeon_color_t *dest = rmesa->vb.specptr;
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->red,	  v[0] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->green, v[1] );
-   UNCLAMPED_FLOAT_TO_UBYTE( dest->blue,  v[2] );
-   dest->alpha = 255;
-}
-
-#if 0
-static void radeon_SecondaryColor3ubEXT_3f( GLubyte r, GLubyte g, GLubyte b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = UBYTE_TO_FLOAT(r);
-   dest[1] = UBYTE_TO_FLOAT(g);
-   dest[2] = UBYTE_TO_FLOAT(b);
-   dest[3] = 1.0;
-}
-
-static void radeon_SecondaryColor3ubvEXT_3f( const GLubyte *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = UBYTE_TO_FLOAT(v[0]);
-   dest[1] = UBYTE_TO_FLOAT(v[1]);
-   dest[2] = UBYTE_TO_FLOAT(v[2]);
-   dest[3] = 1.0;
-}
-#endif /* 0 */
-
-static void radeon_SecondaryColor3fEXT_3f( GLfloat r, GLfloat g, GLfloat b )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = r;
-   dest[1] = g;
-   dest[2] = b;
-   dest[3] = 1.0;
-}
-
-static void radeon_SecondaryColor3fvEXT_3f( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.floatspecptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-   dest[3] = 1.0;
-}
-
-
-/* Normal
- */
-static void radeon_Normal3f( GLfloat n0, GLfloat n1, GLfloat n2 )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.normalptr;
-   dest[0] = n0;
-   dest[1] = n1;
-   dest[2] = n2;
-}
-
-static void radeon_Normal3fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.normalptr;
-   dest[0] = v[0];
-   dest[1] = v[1];
-   dest[2] = v[2];
-}
-
-
-/* TexCoord
- */
-static void radeon_TexCoord1f( GLfloat s )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = s;
-   dest[1] = 0;
-}
-
-static void radeon_TexCoord1fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = v[0];
-   dest[1] = 0;
-}
-
-static void radeon_TexCoord2f( GLfloat s, GLfloat t )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = s;
-   dest[1] = t;
-}
-
-static void radeon_TexCoord2fv( const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[0];
-   dest[0] = v[0];
-   dest[1] = v[1];
-}
-
-
-/* MultiTexcoord
- * 
- * Technically speaking, these functions should subtract GL_TEXTURE0 from
- * \c target before masking and using it.  The value of GL_TEXTURE0 is 0x84C0,
- * which has the low-order 5 bits 0.  For all possible valid values of 
- * \c target.  Subtracting GL_TEXTURE0 has the net effect of masking \c target
- * with 0x1F.  Masking with 0x1F and then masking with 0x01 is redundant, so
- * the subtraction has been omitted.
- */
-
-static void radeon_MultiTexCoord1fARB( GLenum target, GLfloat s  )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 3];
-   dest[0] = s;
-   dest[1] = 0;
-}
-
-static void radeon_MultiTexCoord1fvARB( GLenum target, const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 3];
-   dest[0] = v[0];
-   dest[1] = 0;
-}
-
-static void radeon_MultiTexCoord2fARB( GLenum target, GLfloat s, GLfloat t )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 3];
-   dest[0] = s;
-   dest[1] = t;
-}
-
-static void radeon_MultiTexCoord2fvARB( GLenum target, const GLfloat *v )
-{
-   GET_CURRENT_CONTEXT(ctx);
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   GLfloat *dest = rmesa->vb.texcoordptr[target & 3];
-   dest[0] = v[0];
-   dest[1] = v[1];
-}
-
-static struct dynfn *lookup( struct dynfn *l, int key )
-{
-   struct dynfn *f;
-
-   foreach( f, l ) {
-      if (f->key == key) 
-	 return f;
-   }
-
-   return NULL;
-}
-
-/* Can't use the loopback template for this:
- */
-
-#define CHOOSE(FN, FNTYPE, MASK, ACTIVE, ARGS1, ARGS2 )			\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);			\
-   int key = rmesa->vb.vertex_format & (MASK|ACTIVE);			\
-   struct dynfn *dfn;							\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (dfn == 0)							\
-      dfn = rmesa->vb.codegen.FN( ctx, key );				\
-   else if (RADEON_DEBUG & DEBUG_CODEGEN)				\
-      fprintf(stderr, "%s -- cached codegen\n", __FUNCTION__ );		\
-									\
-   if (dfn)								\
-      SET_ ## FN (ctx->Exec, (FNTYPE)(dfn->code));			\
-   else {								\
-      if (RADEON_DEBUG & DEBUG_CODEGEN)					\
-	 fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, radeon_##FN);				\
-   }									\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-/* For the _3f case, only allow one color function to be hooked in at
- * a time.  Eventually, use a similar mechanism to allow selecting the
- * color component of the vertex format based on client behaviour.  
- *
- * Note:  Perform these actions even if there is a codegen or cached 
- * codegen version of the chosen function.
- */
-#define CHOOSE_COLOR(FN, FNTYPE, NR, MASK, ACTIVE, ARGS1, ARGS2 )	\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx); \
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);			\
-   int key = rmesa->vb.vertex_format & (MASK|ACTIVE);			\
-   struct dynfn *dfn;							\
-									\
-   if (rmesa->vb.vertex_format & ACTIVE_PKCOLOR) {			\
-      SET_ ## FN (ctx->Exec, radeon_##FN##_ub);				\
-   }									\
-   else if ((rmesa->vb.vertex_format &					\
-            (ACTIVE_FPCOLOR|ACTIVE_FPALPHA)) == ACTIVE_FPCOLOR) {	\
-									\
-      if (rmesa->vb.installed_color_3f_sz != NR) {			\
-         rmesa->vb.installed_color_3f_sz = NR;				\
-         if (NR == 3) ctx->Current.Attrib[VERT_ATTRIB_COLOR0][3] = 1.0;	\
-         if (ctx->Driver.NeedFlush & FLUSH_UPDATE_CURRENT) {		\
-            radeon_copy_to_current( ctx );				\
-            _mesa_install_exec_vtxfmt( ctx, &rmesa->vb.vtxfmt );	\
-            CALL_ ## FN (ctx->Exec, ARGS2);				\
-            return;							\
-         }								\
-      }									\
-									\
-      SET_ ## FN (ctx->Exec, radeon_##FN##_3f);				\
-   }									\
-   else {								\
-      SET_ ## FN (ctx->Exec, radeon_##FN##_4f);				\
-   }									\
-									\
-									\
-   dfn = lookup( &rmesa->vb.dfn_cache.FN, key );			\
-   if (!dfn) dfn = rmesa->vb.codegen.FN( ctx, key );			\
-									\
-   if (dfn) {								\
-      if (RADEON_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- codegen version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, (FNTYPE)dfn->code);			\
-   }									\
-   else if (RADEON_DEBUG & DEBUG_CODEGEN)				\
-         fprintf(stderr, "%s -- 'c' version\n", __FUNCTION__ );		\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-/* Right now there are both _ub and _3f versions of the secondary color
- * functions.  Currently, we only set-up the hardware to use the _ub versions.
- * The _3f versions are needed for the cases where secondary color isn't used
- * in the vertex format, but it still needs to be stored in the context
- * state vector.
- */
-#define CHOOSE_SECONDARY_COLOR(FN, FNTYPE, MASK, ACTIVE, ARGS1, ARGS2 )	\
-static void choose_##FN ARGS1						\
-{									\
-   GET_CURRENT_CONTEXT(ctx);						\
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);			\
-   int key = rmesa->vb.vertex_format & (MASK|ACTIVE);			\
-   struct dynfn *dfn = lookup( &rmesa->vb.dfn_cache.FN, key );		\
-									\
-   if (dfn == 0)							\
-      dfn = rmesa->vb.codegen.FN( ctx, key );				\
-   else  if (RADEON_DEBUG & DEBUG_CODEGEN)				\
-      fprintf(stderr, "%s -- cached version\n", __FUNCTION__ );		\
-									\
-   if (dfn)								\
-      SET_ ## FN (ctx->Exec, (FNTYPE)(dfn->code));			\
-   else {								\
-      if (RADEON_DEBUG & DEBUG_CODEGEN)					\
-         fprintf(stderr, "%s -- generic version\n", __FUNCTION__ );	\
-      SET_ ## FN (ctx->Exec, ((rmesa->vb.vertex_format & ACTIVE_PKSPEC) != 0)	\
-	  ? radeon_##FN##_ub : radeon_##FN##_3f);			\
-   }									\
-									\
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;			\
-   CALL_ ## FN (ctx->Exec, ARGS2);					\
-}
-
-
-
-
-
-/* Shorthands
- */
-#define ACTIVE_XYZW (RADEON_CP_VC_FRMT_W0|RADEON_CP_VC_FRMT_Z)
-#define ACTIVE_NORM RADEON_CP_VC_FRMT_N0
-
-#define ACTIVE_PKCOLOR RADEON_CP_VC_FRMT_PKCOLOR
-#define ACTIVE_FPCOLOR RADEON_CP_VC_FRMT_FPCOLOR
-#define ACTIVE_FPALPHA RADEON_CP_VC_FRMT_FPALPHA
-#define ACTIVE_COLOR (ACTIVE_FPCOLOR|ACTIVE_PKCOLOR)
-
-#define ACTIVE_PKSPEC RADEON_CP_VC_FRMT_PKSPEC
-#define ACTIVE_FPSPEC RADEON_CP_VC_FRMT_FPSPEC
-#define ACTIVE_SPEC   (ACTIVE_FPSPEC|ACTIVE_PKSPEC)
-
-#define ACTIVE_ST0 RADEON_CP_VC_FRMT_ST0
-#define ACTIVE_ST1 RADEON_CP_VC_FRMT_ST1
-#define ACTIVE_ST2 RADEON_CP_VC_FRMT_ST2
-#define ACTIVE_ST_ALL (RADEON_CP_VC_FRMT_ST1|RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST2)
-
-/* Each codegen function should be able to be fully specified by a
- * subsetted version of rmesa->vb.vertex_format.
- */
-#define MASK_NORM    (ACTIVE_XYZW)
-#define MASK_COLOR   (MASK_NORM|ACTIVE_NORM)
-#define MASK_SPEC    (MASK_COLOR|ACTIVE_COLOR)
-#define MASK_ST0     (MASK_SPEC|ACTIVE_SPEC)
-#define MASK_ST1     (MASK_ST0|ACTIVE_ST0)
-#define MASK_ST2     (MASK_ST1|ACTIVE_ST1)
-#define MASK_ST_ALL  (MASK_ST2|ACTIVE_ST2)
-#define MASK_VERTEX  (MASK_ST_ALL|ACTIVE_FPALPHA) 
-
-
-typedef void (*p4f)( GLfloat, GLfloat, GLfloat, GLfloat );
-typedef void (*p3f)( GLfloat, GLfloat, GLfloat );
-typedef void (*p2f)( GLfloat, GLfloat );
-typedef void (*p1f)( GLfloat );
-typedef void (*pe2f)( GLenum, GLfloat, GLfloat );
-typedef void (*pe1f)( GLenum, GLfloat );
-typedef void (*p4ub)( GLubyte, GLubyte, GLubyte, GLubyte );
-typedef void (*p3ub)( GLubyte, GLubyte, GLubyte );
-typedef void (*pfv)( const GLfloat * );
-typedef void (*pefv)( GLenum, const GLfloat * );
-typedef void (*pubv)( const GLubyte * );
-
-
-CHOOSE(Normal3f, p3f, MASK_NORM, ACTIVE_NORM, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(Normal3fv, pfv, MASK_NORM, ACTIVE_NORM, 
-       (const GLfloat *v), (v))
-
-#if 0
-CHOOSE_COLOR(Color4ub, p4ub, 4, MASK_COLOR, ACTIVE_COLOR,
-	(GLubyte a,GLubyte b, GLubyte c, GLubyte d), (a,b,c,d))
-CHOOSE_COLOR(Color4ubv, pubv, 4, MASK_COLOR, ACTIVE_COLOR, 
-	(const GLubyte *v), (v))
-CHOOSE_COLOR(Color3ub, p3ub, 3, MASK_COLOR, ACTIVE_COLOR, 
-	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
-CHOOSE_COLOR(Color3ubv, pubv, 3, MASK_COLOR, ACTIVE_COLOR, 
-	(const GLubyte *v), (v))
-#endif
-
-CHOOSE_COLOR(Color4f, p4f, 4, MASK_COLOR, ACTIVE_COLOR, 
-	(GLfloat a,GLfloat b, GLfloat c, GLfloat d), (a,b,c,d))
-CHOOSE_COLOR(Color4fv, pfv, 4, MASK_COLOR, ACTIVE_COLOR, 
-	(const GLfloat *v), (v))
-CHOOSE_COLOR(Color3f, p3f, 3, MASK_COLOR, ACTIVE_COLOR,
-	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
-CHOOSE_COLOR(Color3fv, pfv, 3, MASK_COLOR, ACTIVE_COLOR,
-	(const GLfloat *v), (v))
-
-
-#if 0
-CHOOSE_SECONDARY_COLOR(SecondaryColor3ubEXT, p3ub, MASK_SPEC, ACTIVE_SPEC,
-	(GLubyte a,GLubyte b, GLubyte c), (a,b,c))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3ubvEXT, pubv, MASK_SPEC, ACTIVE_SPEC,
-	(const GLubyte *v), (v))
-#endif
-CHOOSE_SECONDARY_COLOR(SecondaryColor3fEXT, p3f, MASK_SPEC, ACTIVE_SPEC,
-	(GLfloat a,GLfloat b, GLfloat c), (a,b,c))
-CHOOSE_SECONDARY_COLOR(SecondaryColor3fvEXT, pfv, MASK_SPEC, ACTIVE_SPEC,
-	(const GLfloat *v), (v))
-
-CHOOSE(TexCoord2f, p2f, MASK_ST0, ACTIVE_ST0, 
-       (GLfloat a,GLfloat b), (a,b))
-CHOOSE(TexCoord2fv, pfv, MASK_ST0, ACTIVE_ST0, 
-       (const GLfloat *v), (v))
-CHOOSE(TexCoord1f, p1f, MASK_ST0, ACTIVE_ST0, 
-       (GLfloat a), (a))
-CHOOSE(TexCoord1fv, pfv, MASK_ST0, ACTIVE_ST0, 
-       (const GLfloat *v), (v))
-
-CHOOSE(MultiTexCoord2fARB, pe2f, MASK_ST_ALL, ACTIVE_ST_ALL,
-	 (GLenum u,GLfloat a,GLfloat b), (u,a,b))
-CHOOSE(MultiTexCoord2fvARB, pefv, MASK_ST_ALL, ACTIVE_ST_ALL,
-	(GLenum u,const GLfloat *v), (u,v))
-CHOOSE(MultiTexCoord1fARB, pe1f, MASK_ST_ALL, ACTIVE_ST_ALL,
-	 (GLenum u,GLfloat a), (u,a))
-CHOOSE(MultiTexCoord1fvARB, pefv, MASK_ST_ALL, ACTIVE_ST_ALL,
-	(GLenum u,const GLfloat *v), (u,v))
-
-CHOOSE(Vertex3f, p3f, MASK_VERTEX, MASK_VERTEX, 
-       (GLfloat a,GLfloat b,GLfloat c), (a,b,c))
-CHOOSE(Vertex3fv, pfv, MASK_VERTEX, MASK_VERTEX, 
-       (const GLfloat *v), (v))
-CHOOSE(Vertex2f, p2f, MASK_VERTEX, MASK_VERTEX, 
-       (GLfloat a,GLfloat b), (a,b))
-CHOOSE(Vertex2fv, pfv, MASK_VERTEX, MASK_VERTEX, 
-       (const GLfloat *v), (v))
-
-
-
-
-
-void radeonVtxfmtInitChoosers( GLvertexformat *vfmt )
-{
-   vfmt->Color3f = choose_Color3f;
-   vfmt->Color3fv = choose_Color3fv;
-   vfmt->Color4f = choose_Color4f;
-   vfmt->Color4fv = choose_Color4fv;
-   vfmt->SecondaryColor3fEXT = choose_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = choose_SecondaryColor3fvEXT;
-   vfmt->MultiTexCoord1fARB = choose_MultiTexCoord1fARB;
-   vfmt->MultiTexCoord1fvARB = choose_MultiTexCoord1fvARB;
-   vfmt->MultiTexCoord2fARB = choose_MultiTexCoord2fARB;
-   vfmt->MultiTexCoord2fvARB = choose_MultiTexCoord2fvARB;
-   vfmt->Normal3f = choose_Normal3f;
-   vfmt->Normal3fv = choose_Normal3fv;
-   vfmt->TexCoord1f = choose_TexCoord1f;
-   vfmt->TexCoord1fv = choose_TexCoord1fv;
-   vfmt->TexCoord2f = choose_TexCoord2f;
-   vfmt->TexCoord2fv = choose_TexCoord2fv;
-   vfmt->Vertex2f = choose_Vertex2f;
-   vfmt->Vertex2fv = choose_Vertex2fv;
-   vfmt->Vertex3f = choose_Vertex3f;
-   vfmt->Vertex3fv = choose_Vertex3fv;
-
-#if 0
-   vfmt->Color3ub = choose_Color3ub;
-   vfmt->Color3ubv = choose_Color3ubv;
-   vfmt->Color4ub = choose_Color4ub;
-   vfmt->Color4ubv = choose_Color4ubv;
-   vfmt->SecondaryColor3ubEXT = choose_SecondaryColor3ubEXT;
-   vfmt->SecondaryColor3ubvEXT = choose_SecondaryColor3ubvEXT;
-#endif
-}
-
-
-static struct dynfn *codegen_noop( GLcontext *ctx, int key )
-{
-   (void) ctx; (void) key;
-   return NULL;
-}
-
-void radeonInitCodegen( struct dfn_generators *gen, GLboolean useCodegen )
-{
-   gen->Vertex3f = codegen_noop;
-   gen->Vertex3fv = codegen_noop;
-   gen->Color4ub = codegen_noop;
-   gen->Color4ubv = codegen_noop;
-   gen->Normal3f = codegen_noop;
-   gen->Normal3fv = codegen_noop;
-   gen->TexCoord2f = codegen_noop;
-   gen->TexCoord2fv = codegen_noop;
-   gen->MultiTexCoord2fARB = codegen_noop;
-   gen->MultiTexCoord2fvARB = codegen_noop;
-   gen->Vertex2f = codegen_noop;
-   gen->Vertex2fv = codegen_noop;
-   gen->Color3ub = codegen_noop;
-   gen->Color3ubv = codegen_noop;
-   gen->Color4f = codegen_noop;
-   gen->Color4fv = codegen_noop;
-   gen->Color3f = codegen_noop;
-   gen->Color3fv = codegen_noop;
-   gen->SecondaryColor3fEXT = codegen_noop;
-   gen->SecondaryColor3fvEXT = codegen_noop;
-   gen->SecondaryColor3ubEXT = codegen_noop;
-   gen->SecondaryColor3ubvEXT = codegen_noop;
-   gen->TexCoord1f = codegen_noop;
-   gen->TexCoord1fv = codegen_noop;
-   gen->MultiTexCoord1fARB = codegen_noop;
-   gen->MultiTexCoord1fvARB = codegen_noop;
-
-   if (useCodegen) {
-#if defined(USE_X86_ASM)
-      radeonInitX86Codegen( gen );
-#endif
-
-#if defined(USE_SSE_ASM)
-      radeonInitSSECodegen( gen );
-#endif
-   }
-}
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c b/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c
deleted file mode 100644
index 0f0fc9e0653..00000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_sse.c
+++ /dev/null
@@ -1,236 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt_sse.c,v 1.1 2002/10/30 12:51:58 alanh Exp $ */
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "simple_list.h" 
-#include "radeon_vtxfmt.h"
-
-#if defined(USE_SSE_ASM)
-#include "x86/common_x86_asm.h"
-
-#define EXTERN( FUNC )		\
-extern const char *FUNC;	\
-extern const char *FUNC##_end
-
-EXTERN( _sse_Attribute2fv );
-EXTERN( _sse_Attribute2f );
-EXTERN( _sse_Attribute3fv );
-EXTERN( _sse_Attribute3f );
-EXTERN( _sse_MultiTexCoord2fv );
-EXTERN( _sse_MultiTexCoord2f );
-EXTERN( _sse_MultiTexCoord2fv_2 );
-EXTERN( _sse_MultiTexCoord2f_2 );
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.
- */
-
-static struct dynfn *radeon_makeSSEAttribute2fv( struct dynfn * cache, int key,
-					       const char * name, void * dest)
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _sse_Attribute2fv, (*cache) );
-   FIXUP(dfn->code, 10, 0x0, (int)dest);
-   return dfn;
-}
-
-static struct dynfn *radeon_makeSSEAttribute2f( struct dynfn * cache, int key,
-					      const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _sse_Attribute2f, (*cache) );
-   FIXUP(dfn->code, 8, 0x0, (int)dest); 
-   return dfn;
-}
-
-static struct dynfn *radeon_makeSSEAttribute3fv( struct dynfn * cache, int key,
-					       const char * name, void * dest)
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _sse_Attribute3fv, (*cache) );
-   FIXUP(dfn->code, 13, 0x0, (int)dest);
-   FIXUP(dfn->code, 18, 0x8, 8+(int)dest);
-   return dfn;
-}
-
-static struct dynfn *radeon_makeSSEAttribute3f( struct dynfn * cache, int key,
-					      const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _sse_Attribute3f, (*cache) );
-   FIXUP(dfn->code, 12, 0x0, (int)dest); 
-   FIXUP(dfn->code, 17, 0x8, 8+(int)dest); 
-   return dfn;
-}
-
-static struct dynfn * radeon_makeSSENormal3fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeSSEAttribute3fv( & rmesa->vb.dfn_cache.Normal3fv, key,
-				      __FUNCTION__, rmesa->vb.normalptr );
-}
-
-static struct dynfn *radeon_makeSSENormal3f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeSSEAttribute3f( & rmesa->vb.dfn_cache.Normal3f, key,
-				     __FUNCTION__, rmesa->vb.normalptr );
-}
-
-static struct dynfn *radeon_makeSSEColor3fv( GLcontext *ctx, int key )
-{
-   if (key & (RADEON_CP_VC_FRMT_PKCOLOR|RADEON_CP_VC_FRMT_FPALPHA))
-      return NULL;
-   else
-   {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      return radeon_makeSSEAttribute3fv( & rmesa->vb.dfn_cache.Color3fv, key,
-					 __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-static struct dynfn *radeon_makeSSEColor3f( GLcontext *ctx, int key )
-{
-   if (key & (RADEON_CP_VC_FRMT_PKCOLOR|RADEON_CP_VC_FRMT_FPALPHA))
-      return NULL;
-   else
-   {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      return radeon_makeSSEAttribute3f( & rmesa->vb.dfn_cache.Color3f, key,
-					__FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-static struct dynfn *radeon_makeSSETexCoord2fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeSSEAttribute2fv( & rmesa->vb.dfn_cache.TexCoord2fv, key,
-				       __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-static struct dynfn *radeon_makeSSETexCoord2f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeSSEAttribute2f( & rmesa->vb.dfn_cache.TexCoord2f, key,
-				     __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-#if 0 /* Temporarily disabled - probably needs adjustments for more than 2 tex units -rs */
-static struct dynfn *radeon_makeSSEMultiTexCoord2fv( GLcontext *ctx, int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if ((key & (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) ==
-      (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) {
-      DFN ( _sse_MultiTexCoord2fv, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 18, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-   } else {
-      DFN ( _sse_MultiTexCoord2fv_2, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 14, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-
-static struct dynfn *radeon_makeSSEMultiTexCoord2f( GLcontext *ctx, int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if ((key & (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) ==
-      (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) {
-      DFN ( _sse_MultiTexCoord2f, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 16, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);	
-   } else {
-      DFN ( _sse_MultiTexCoord2f_2, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 15, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-#endif
-
-void radeonInitSSECodegen( struct dfn_generators *gen )
-{
-   if ( cpu_has_xmm ) {
-      gen->Normal3fv = (void *) radeon_makeSSENormal3fv;
-      gen->Normal3f = (void *) radeon_makeSSENormal3f;
-      gen->Color3fv = (void *) radeon_makeSSEColor3fv;
-      gen->Color3f = (void *) radeon_makeSSEColor3f;
-      gen->TexCoord2fv = (void *) radeon_makeSSETexCoord2fv;
-      gen->TexCoord2f = (void *) radeon_makeSSETexCoord2f;
-#if 0 /* Temporarily disabled - probably needs adjustments for more than 2 tex units -rs */
-      gen->MultiTexCoord2fvARB = (void *) radeon_makeSSEMultiTexCoord2fv;
-      gen->MultiTexCoord2fARB = (void *) radeon_makeSSEMultiTexCoord2f;
-#endif
-   }
-}
-
-#else 
-
-void radeonInitSSECodegen( struct dfn_generators *gen )
-{
-   (void) gen;
-}
-
-#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c b/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c
deleted file mode 100644
index 529e79065e9..00000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxfmt_x86.c
+++ /dev/null
@@ -1,440 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxfmt_x86.c,v 1.2 2002/12/21 17:02:16 dawes Exp $ */
-/**************************************************************************
-
-Copyright 2000, 2001 ATI Technologies Inc., Ontario, Canada, and
-                     Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"), 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 COPYRIGHT OWNER(S) AND/OR ITS SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "imports.h"
-#include "simple_list.h" 
-#include "radeon_vtxfmt.h"
-
-#if defined(USE_X86_ASM)
-
-#define EXTERN( FUNC )		\
-extern const char *FUNC;	\
-extern const char *FUNC##_end
-
-EXTERN ( _x86_Attribute2fv );
-EXTERN ( _x86_Attribute2f );
-EXTERN ( _x86_Attribute3fv );
-EXTERN ( _x86_Attribute3f );
-EXTERN ( _x86_Vertex3fv_6 );
-EXTERN ( _x86_Vertex3fv_8 );
-EXTERN ( _x86_Vertex3fv );
-EXTERN ( _x86_Vertex3f_4 );
-EXTERN ( _x86_Vertex3f_6 );
-EXTERN ( _x86_Vertex3f );
-EXTERN ( _x86_Color4ubv_ub );
-EXTERN ( _x86_Color4ubv_4f );
-EXTERN ( _x86_Color4ub_ub );
-EXTERN ( _x86_MultiTexCoord2fv );
-EXTERN ( _x86_MultiTexCoord2fv_2 );
-EXTERN ( _x86_MultiTexCoord2f );
-EXTERN ( _x86_MultiTexCoord2f_2 );
-
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.  Generic x86 versions.
- */
-
-struct dynfn *radeon_makeX86Vertex3f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x %d\n", __FUNCTION__, key, rmesa->vb.vertex_size );
-
-   switch (rmesa->vb.vertex_size) {
-   case 4: {
-
-      DFN ( _x86_Vertex3f_4, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 2, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 25, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 36, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 46, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 51, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 60, 0x0, (int)&rmesa->vb.notify);
-      break;
-   }
-   case 6: {
-
-      DFN ( _x86_Vertex3f_6, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 3, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 28, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 34, 0x0, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 40, 0x0, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 57, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 63, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 70, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 79, 0x0, (int)&rmesa->vb.notify);
-      break;
-   }
-   default: {
-
-      DFN ( _x86_Vertex3f, rmesa->vb.dfn_cache.Vertex3f );
-      FIXUP(dfn->code, 3, 0x0, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 9, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 37, 0x0, rmesa->vb.vertex_size-3);
-      FIXUP(dfn->code, 44, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 50, 0x0, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 56, 0x0, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x0, (int)&rmesa->vb.notify);
-   break;
-   }
-   }
-
-   return dfn;
-}
-
-
-
-struct dynfn *radeon_makeX86Vertex3fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x %d\n", __FUNCTION__, key, rmesa->vb.vertex_size );
-
-   switch (rmesa->vb.vertex_size) {
-   case 6: {
-
-      DFN ( _x86_Vertex3fv_6, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 1, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 27, 0x0000001c, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 33, 0x00000020, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 45, 0x00000024, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 56, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 61, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 76, 0x00000008, (int)&rmesa->vb.notify);
-      break;
-   }
-   
-
-   case 8: {
-
-      DFN ( _x86_Vertex3fv_8, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 1, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 27, 0x0000001c, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 33, 0x00000020, (int)&rmesa->vb.vertex[4]);
-      FIXUP(dfn->code, 45, 0x0000001c, (int)&rmesa->vb.vertex[5]);
-      FIXUP(dfn->code, 51, 0x00000020, (int)&rmesa->vb.vertex[6]);
-      FIXUP(dfn->code, 63, 0x00000024, (int)&rmesa->vb.vertex[7]);
-      FIXUP(dfn->code, 74, 0x00000000, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 79, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 85, 0x00000004, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 94, 0x00000008, (int)&rmesa->vb.notify);
-      break;
-   }
-   
-
-
-   default: {
-
-      DFN ( _x86_Vertex3fv, rmesa->vb.dfn_cache.Vertex3fv );
-      FIXUP(dfn->code, 8, 0x01010101, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 32, 0x00000006, rmesa->vb.vertex_size-3);
-      FIXUP(dfn->code, 37, 0x00000058, (int)&rmesa->vb.vertex[3]);
-      FIXUP(dfn->code, 45, 0x01010101, (int)&rmesa->vb.dmaptr);
-      FIXUP(dfn->code, 50, 0x02020202, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 58, 0x02020202, (int)&rmesa->vb.counter);
-      FIXUP(dfn->code, 67, 0x0, (int)&rmesa->vb.notify);
-   break;
-   }
-   }
-
-   return dfn;
-}
-
-static struct dynfn *
-radeon_makeX86Attribute2fv( struct dynfn * cache, int key,
-			    const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _x86_Attribute2fv, (*cache) );
-   FIXUP(dfn->code, 11, 0x0, (int)dest); 
-   FIXUP(dfn->code, 16, 0x4, 4+(int)dest); 
-
-   return dfn;
-}
-
-static struct dynfn *
-radeon_makeX86Attribute2f( struct dynfn * cache, int key,
-			   const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _x86_Attribute2f, (*cache) );
-   FIXUP(dfn->code, 1, 0x0, (int)dest); 
-
-   return dfn;
-}
-
-
-static struct dynfn *
-radeon_makeX86Attribute3fv( struct dynfn * cache, int key,
-			    const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _x86_Attribute3fv, (*cache) );
-   FIXUP(dfn->code, 14, 0x0, (int)dest); 
-   FIXUP(dfn->code, 20, 0x4, 4+(int)dest); 
-   FIXUP(dfn->code, 25, 0x8, 8+(int)dest);
-
-   return dfn;
-}
-
-static struct dynfn *
-radeon_makeX86Attribute3f( struct dynfn * cache, int key,
-			 const char * name, void * dest )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", name, key );
-
-   DFN ( _x86_Attribute3f, (*cache) );
-   FIXUP(dfn->code, 14, 0x0, (int)dest); 
-   FIXUP(dfn->code, 20, 0x4, 4+(int)dest); 
-   FIXUP(dfn->code, 25, 0x8, 8+(int)dest);
-
-   return dfn;
-}
-
-struct dynfn *radeon_makeX86Normal3fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeX86Attribute3fv( & rmesa->vb.dfn_cache.Normal3fv, key,
-				      __FUNCTION__, rmesa->vb.normalptr );
-}
-
-struct dynfn *radeon_makeX86Normal3f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeX86Attribute3f( & rmesa->vb.dfn_cache.Normal3f, key,
-				     __FUNCTION__, rmesa->vb.normalptr );
-}
-
-struct dynfn *radeon_makeX86Color4ubv( GLcontext *ctx, int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if (key & RADEON_CP_VC_FRMT_PKCOLOR) {
-      DFN ( _x86_Color4ubv_ub, rmesa->vb.dfn_cache.Color4ubv);
-      FIXUP(dfn->code, 5, 0x12345678, (int)rmesa->vb.colorptr); 
-      return dfn;
-   } 
-   else {
-
-      DFN ( _x86_Color4ubv_4f, rmesa->vb.dfn_cache.Color4ubv);
-      FIXUP(dfn->code, 2, 0x00000000, (int)_mesa_ubyte_to_float_color_tab); 
-      FIXUP(dfn->code, 27, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr); 
-      FIXUP(dfn->code, 33, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+4); 
-      FIXUP(dfn->code, 55, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+8); 
-      FIXUP(dfn->code, 61, 0xdeadbeaf, (int)rmesa->vb.floatcolorptr+12); 
-      return dfn;
-   }
-}
-
-struct dynfn *radeon_makeX86Color4ub( GLcontext *ctx, int key )
-{
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if (key & RADEON_CP_VC_FRMT_PKCOLOR) {
-      struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      DFN ( _x86_Color4ub_ub, rmesa->vb.dfn_cache.Color4ub );
-      FIXUP(dfn->code, 18, 0x0, (int)rmesa->vb.colorptr); 
-      FIXUP(dfn->code, 24, 0x0, (int)rmesa->vb.colorptr+1); 
-      FIXUP(dfn->code, 30, 0x0, (int)rmesa->vb.colorptr+2); 
-      FIXUP(dfn->code, 36, 0x0, (int)rmesa->vb.colorptr+3); 
-      return dfn;
-   }
-   else
-      return NULL;
-}
-
-
-struct dynfn *radeon_makeX86Color3fv( GLcontext *ctx, int key )
-{
-   if (key & (RADEON_CP_VC_FRMT_PKCOLOR|RADEON_CP_VC_FRMT_FPALPHA))
-      return NULL;
-   else
-   {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      return radeon_makeX86Attribute3fv( & rmesa->vb.dfn_cache.Color3fv, key,
-					 __FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-struct dynfn *radeon_makeX86Color3f( GLcontext *ctx, int key )
-{
-   if (key & (RADEON_CP_VC_FRMT_PKCOLOR|RADEON_CP_VC_FRMT_FPALPHA))
-      return NULL;
-   else
-   {
-      radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-      return radeon_makeX86Attribute3f( & rmesa->vb.dfn_cache.Color3f, key,
-					__FUNCTION__, rmesa->vb.floatcolorptr );
-   }
-}
-
-
-
-struct dynfn *radeon_makeX86TexCoord2fv( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeX86Attribute2fv( & rmesa->vb.dfn_cache.TexCoord2fv, key,
-				      __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-struct dynfn *radeon_makeX86TexCoord2f( GLcontext *ctx, int key )
-{
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   return radeon_makeX86Attribute2f( & rmesa->vb.dfn_cache.TexCoord2f, key,
-				     __FUNCTION__, rmesa->vb.texcoordptr[0] );
-}
-
-#if 0 /* Temporarily disabled - probably needs adjustments for more than 2 tex units -rs */
-struct dynfn *radeon_makeX86MultiTexCoord2fvARB( GLcontext *ctx, int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if ((key & (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) ==
-      (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) {
-      DFN ( _x86_MultiTexCoord2fv, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 21, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);
-      FIXUP(dfn->code, 27, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]+4);
-   } else {
-      DFN ( _x86_MultiTexCoord2fv_2, rmesa->vb.dfn_cache.MultiTexCoord2fvARB );
-      FIXUP(dfn->code, 14, 0x0, (int)rmesa->vb.texcoordptr);
-   }
-   return dfn;
-}
-
-struct dynfn *radeon_makeX86MultiTexCoord2fARB( GLcontext *ctx, 
-						int key )
-{
-   struct dynfn *dfn = MALLOC_STRUCT( dynfn );
-   radeonContextPtr rmesa = RADEON_CONTEXT(ctx);
-
-   if (RADEON_DEBUG & DEBUG_CODEGEN)
-      fprintf(stderr, "%s 0x%08x\n", __FUNCTION__, key );
-
-   if ((key & (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) ==
-       (RADEON_CP_VC_FRMT_ST0|RADEON_CP_VC_FRMT_ST1)) {
-      DFN ( _x86_MultiTexCoord2f, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 20, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]);
-      FIXUP(dfn->code, 26, 0xdeadbeef, (int)rmesa->vb.texcoordptr[0]+4); 
-   }
-   else {
-      /* Note: this might get generated multiple times, even though the
-       * actual emitted code is the same.
-       */
-      DFN ( _x86_MultiTexCoord2f_2, rmesa->vb.dfn_cache.MultiTexCoord2fARB );
-      FIXUP(dfn->code, 18, 0x0, (int)rmesa->vb.texcoordptr); 
-   }      
-   return dfn;
-}
-#endif
-
-void radeonInitX86Codegen( struct dfn_generators *gen )
-{
-   gen->Vertex3f = radeon_makeX86Vertex3f;
-   gen->Vertex3fv = radeon_makeX86Vertex3fv;
-   gen->Color4ub = radeon_makeX86Color4ub; /* PKCOLOR only */
-   gen->Color4ubv = radeon_makeX86Color4ubv; /* PKCOLOR only */
-   gen->Normal3f = radeon_makeX86Normal3f;
-   gen->Normal3fv = radeon_makeX86Normal3fv;
-   gen->TexCoord2f = radeon_makeX86TexCoord2f;
-   gen->TexCoord2fv = radeon_makeX86TexCoord2fv;
-#if 0 /* Temporarily disabled - probably needs adjustments for more than 2 tex units -rs */
-   gen->MultiTexCoord2fARB = radeon_makeX86MultiTexCoord2fARB;
-   gen->MultiTexCoord2fvARB = radeon_makeX86MultiTexCoord2fvARB;
-#endif
-   gen->Color3f = radeon_makeX86Color3f;
-   gen->Color3fv = radeon_makeX86Color3fv;
-
-   /* Not done:
-    */
-/*     gen->Vertex2f = radeon_makeX86Vertex2f; */
-/*     gen->Vertex2fv = radeon_makeX86Vertex2fv; */
-/*     gen->Color3ub = radeon_makeX86Color3ub; */
-/*     gen->Color3ubv = radeon_makeX86Color3ubv; */
-/*     gen->Color4f = radeon_makeX86Color4f; */
-/*     gen->Color4fv = radeon_makeX86Color4fv; */
-/*     gen->TexCoord1f = radeon_makeX86TexCoord1f; */
-/*     gen->TexCoord1fv = radeon_makeX86TexCoord1fv; */
-/*     gen->MultiTexCoord1fARB = radeon_makeX86MultiTexCoord1fARB; */
-/*     gen->MultiTexCoord1fvARB = radeon_makeX86MultiTexCoord1fvARB; */
-}
-
-
-#else 
-
-void radeonInitX86Codegen( struct dfn_generators *gen )
-{
-   (void) gen;
-}
-
-#endif
diff --git a/src/mesa/drivers/dri/radeon/radeon_vtxtmp_x86.S b/src/mesa/drivers/dri/radeon/radeon_vtxtmp_x86.S
deleted file mode 100644
index 1b433491aa8..00000000000
--- a/src/mesa/drivers/dri/radeon/radeon_vtxtmp_x86.S
+++ /dev/null
@@ -1,498 +0,0 @@
-/* $XFree86: xc/lib/GL/mesa/src/drv/radeon/radeon_vtxtmp_x86.S,v 1.1 2002/10/30 12:51:58 alanh Exp $ */
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS 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.
-
-**************************************************************************/
-
-#define GLOBL( x )	\
-.globl x;		\
-x:
-
-.data
-.align 4
-
-/*
-	vertex 3f vertex size 4
-*/
-	
-GLOBL ( _x86_Vertex3f_4 )
-	movl	(0), %ecx
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	%eax, (%ecx)
-	movl	%edx, 4(%ecx)
-	movl	12(%esp), %eax
-	movl	(0), %edx
-	movl	%eax, 8(%ecx)
-	movl	%edx, 12(%ecx)
-	movl	(0), %eax
-	addl	$16, %ecx
-	dec 	%eax
-	movl	%ecx, (0)
-	movl	%eax, (0)
-	je	.1 
-	ret
-.1:	jmp	*0
-	
-GLOBL ( _x86_Vertex3f_4_end )
-
-/*
-	vertex 3f vertex size 6
-*/
-GLOBL ( _x86_Vertex3f_6 )
-	push	%edi
-	movl	(0), %edi
-	movl	8(%esp), %eax
-	movl	12(%esp), %edx
-	movl	16(%esp), %ecx
-	movl	%eax, (%edi)
-	movl	%edx, 4(%edi)
-	movl	%ecx, 8(%edi)
-	movl	(0), %eax
-	movl	(0), %edx
-	movl	(0), %ecx
-	movl	%eax, 12(%edi)
-	movl	%edx, 16(%edi)
-	movl	%ecx, 20(%edi)
-	addl	$24, %edi
-	movl	(0), %eax
-	movl	%edi, (0)
-	dec 	%eax
-	pop 	%edi
-	movl	%eax, (0)
-	je	.2
-	ret
-.2:	jmp	*0
-GLOBL ( _x86_Vertex3f_6_end )
-/*
-	vertex 3f generic size
-*/
-GLOBL ( _x86_Vertex3f )
-	push	%edi
-	push	%esi
-	movl	$0, %esi
-	movl	(0), %edi
-	movl	12(%esp), %eax
-	movl	16(%esp), %edx
-	movl	20(%esp), %ecx
-	movl	%eax, (%edi)
-	movl	%edx, 4(%edi)
-	movl	%ecx, 8(%edi)
-	addl	$12, %edi
-	movl	$0, %ecx
-	repz
-	movsl %ds:(%esi), %es:(%edi)
-	movl	(0), %eax
-	movl	%edi, (0)
-	dec 	%eax
-	movl	%eax, (0)
-	pop 	%esi
-	pop 	%edi
-	je  	.3
-	ret
-.3:	jmp	*0
-
-GLOBL ( _x86_Vertex3f_end )
-
-/*
-	Vertex 3fv vertex size 6
-*/
-GLOBL ( _x86_Vertex3fv_6 )
-	movl	(0), %eax
-	movl	4(%esp), %ecx
-	movl	(%ecx), %edx
-	movl	%edx, (%eax)
-	movl	4(%ecx), %edx
-	movl	8(%ecx), %ecx
-	movl	%edx, 4(%eax)
-	movl	%ecx, 8(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 12(%eax)
-	movl	%ecx, 16(%eax)
-	movl	(36), %edx
-	movl	%edx, 20(%eax)
-	addl	$24, %eax
-	movl	%eax, 0
-	movl	4, %eax
-	dec 	%eax
-	movl	%eax, 4
-	je	.4
-	ret
-.4:	jmp    *8
-	
-GLOBL ( _x86_Vertex3fv_6_end )
-
-/*
-	Vertex 3fv vertex size 8
-*/
-GLOBL ( _x86_Vertex3fv_8 )
-	movl	(0), %eax
-	movl	4(%esp), %ecx
-	movl	(%ecx), %edx
-	movl	%edx ,(%eax)
-	movl	4(%ecx) ,%edx
-	movl	8(%ecx) ,%ecx
-	movl	%edx, 4(%eax)
-	movl	%ecx, 8(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 12(%eax)
-	movl	%ecx, 16(%eax)
-	movl	(28), %edx
-	movl	(32), %ecx
-	movl	%edx, 20(%eax)
-	movl	%ecx, 24(%eax)
-	movl	(36), %edx
-	movl	%edx, 28(%eax)
-	addl	$32, %eax
-	movl	%eax, (0)
-	movl	4, %eax
-	dec	%eax
-	movl    %eax, (4)
-	je	.5
-	ret
-.5:	jmp    *8
-	
-GLOBL ( _x86_Vertex3fv_8_end )
-
-/*
-	Vertex 3fv generic vertex size
-*/
-GLOBL ( _x86_Vertex3fv )
-	movl	4(%esp), %edx
-	push	%edi
-	push	%esi
-	movl	(0x1010101), %edi
-	movl	(%edx), %eax
-	movl	4(%edx), %ecx
-	movl	8(%edx), %esi
-	movl	%eax, (%edi)
-	movl	%ecx, 4(%edi)
-	movl	%esi, 8(%edi)
-	addl	$12, %edi
-	movl	$6, %ecx
-	movl	$0x58, %esi
-	repz
-	movsl %ds:(%esi), %es:(%edi)
-	movl	%edi, (0x1010101)
-	movl	(0x2020202), %eax
-	pop	%esi
-	pop	%edi
-	dec	%eax
-	movl	%eax, (0x2020202)
-	je	.6
-	ret
-.6:	jmp    *0
-GLOBL ( _x86_Vertex3fv_end )
-
-
-/**
- * Generic handler for 2 float format data.  This can be used for
- * TexCoord2f and possibly other functions.
- */
-
-GLOBL ( _x86_Attribute2f )
-	movl	$0x0, %edx
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	movl	%eax, (%edx)
-	movl	%ecx, 4(%edx)
-	ret
-GLOBL ( _x86_Attribute2f_end )
-
-
-/**
- * Generic handler for 2 float vector format data.  This can be used for
- * TexCoord2fv and possibly other functions.
- */
-
-GLOBL( _x86_Attribute2fv)
-	movl 4(%esp), %eax      /* load 'v' off stack */
-	movl (%eax), %ecx       /* load v[0] */
-	movl 4(%eax), %eax      /* load v[1] */
-	movl %ecx, 0      	/* store v[0] to current vertex */
-	movl %eax, 4      	/* store v[1] to current vertex */
-	ret
-GLOBL ( _x86_Attribute2fv_end )
-
-
-/**
- * Generic handler for 3 float format data.  This can be used for
- * Normal3f, Color3f (when the color target is also float), or
- * TexCoord3f.
- */
-
-GLOBL ( _x86_Attribute3f )
-	movl	4(%esp), %ecx
-	movl	8(%esp), %edx
-	movl	12(%esp), %eax
-	movl	%ecx, 0
-	movl	%edx, 4
-	movl	%eax, 8
-	ret
-GLOBL ( _x86_Attribute3f_end )
-
-/**
- * Generic handler for 3 float vector format data.  This can be used for
- * Normal3f, Color3f (when the color target is also float), or
- * TexCoord3f.
- */
-
-GLOBL( _x86_Attribute3fv)
-	movl 4(%esp), %eax      /* load 'v' off stack */
-	movl (%eax), %ecx       /* load v[0] */
-	movl 4(%eax), %edx      /* load v[1] */
-	movl 8(%eax), %eax      /* load v[2] */
-	movl %ecx, 0      	/* store v[0] to current vertex */
-	movl %edx, 4      	/* store v[1] to current vertex */
-	movl %eax, 8      	/* store v[2] to current vertex */
-	ret
-GLOBL ( _x86_Attribute3fv_end )
-
-
-/*
-	Color 4ubv_ub
-*/
-GLOBL ( _x86_Color4ubv_ub )
-	movl 4(%esp), %eax
-	movl $0x12345678, %edx
-	movl (%eax), %eax
-	movl %eax, (%edx)
-	ret
-GLOBL ( _x86_Color4ubv_ub_end )
-
-/*
-	Color 4ubv 4f
-*/
-GLOBL ( _x86_Color4ubv_4f )
-	push	%ebx
-	movl	$0, %edx
-	xor	%eax, %eax
-	xor	%ecx, %ecx
-	movl	8(%esp), %ebx
-	movl	(%ebx), %ebx
-	mov	%bl, %al
-	mov	%bh, %cl
-	movl	(%edx,%eax,4),%eax
-	movl	(%edx,%ecx,4),%ecx
-	movl	%eax, (0xdeadbeaf)
-	movl	%ecx, (0xdeadbeaf)
-	xor	%eax, %eax
-	xor	%ecx, %ecx
-	shr	$16, %ebx
-	mov	%bl, %al
-	mov	%bh, %cl
-	movl	(%edx,%eax,4), %eax
-	movl	(%edx,%ecx,4), %ecx
-	movl	%eax, (0xdeadbeaf)
-	movl	%ecx, (0xdeadbeaf)
-	pop	%ebx
-	ret
-GLOBL ( _x86_Color4ubv_4f_end )
-
-/*
-
-	Color4ub_ub
-*/
-GLOBL( _x86_Color4ub_ub )
-	push	%ebx
-	movl	8(%esp), %eax
-	movl	12(%esp), %edx
-	movl	16(%esp), %ecx
-	movl	20(%esp), %ebx
-	mov	%al, (0)
-	mov	%dl, (0)
-	mov	%cl, (0)
-	mov	%bl, (0)
-	pop	%ebx
-	ret
-GLOBL( _x86_Color4ub_ub_end )
-
-
-/*
-	MultiTexCoord2fv st0/st1
-*/
-GLOBL( _x86_MultiTexCoord2fv )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$3, %eax
-	movl	(%ecx), %edx
-	shl	$3, %eax
-	movl	4(%ecx), %ecx
-	movl	%edx, 0xdeadbeef(%eax)
-	movl	%ecx, 0xdeadbeef(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2fv_end )
-
-/*
-	MultiTexCoord2fv
-*/
-
-GLOBL( _x86_MultiTexCoord2fv_2 )
-	movl	4(%esp,1), %eax
-	movl	8(%esp,1), %ecx
-	and	$3, %eax
-	movl	0(,%eax,4), %edx
-	movl	(%ecx), %eax
-	movl	%eax, (%edx)
-	movl	4(%ecx), %eax
-	movl	%eax, 4(%edx)
-	ret
-GLOBL( _x86_MultiTexCoord2fv_2_end )
-
-/*
-	MultiTexCoord2f st0/st1
-*/
-GLOBL( _x86_MultiTexCoord2f )
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	12(%esp), %ecx
-	and	$3, %eax
-	shl	$3, %eax
-	movl	%edx, 0xdeadbeef(%eax)
-	movl	%ecx, 0xdeadbeef(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2f_end )
-
-/*
-	MultiTexCoord2f
-*/
-GLOBL( _x86_MultiTexCoord2f_2 )
-	movl	4(%esp), %eax
-	movl	8(%esp), %edx
-	movl	12(%esp,1), %ecx
-	and	$3,%eax
-	movl	0(,%eax,4), %eax
-	movl	%edx, (%eax)
-	movl	%ecx, 4(%eax)
-	ret
-GLOBL( _x86_MultiTexCoord2f_2_end )
-
-#if defined(USE_SSE_ASM)
-/**
- * This can be used as a template for either Color3fv (when the color
- * target is also a 3f) or Normal3fv.
- */
-
-GLOBL( _sse_Attribute3fv )
-	movl	4(%esp), %eax
-	movlps	(%eax), %xmm0
-	movl	8(%eax), %eax
-	movlps	%xmm0, 0
-	movl	%eax, 8
-	ret
-GLOBL( _sse_Attribute3fv_end )
-
-/**
- * This can be used as a template for either Color3f (when the color
- * target is also a 3f) or Normal3f.
- */
-
-GLOBL( _sse_Attribute3f )
-	movlps	4(%esp), %xmm0
-	movl	12(%esp), %eax
-	movlps	%xmm0, 0
-	movl	%eax, 8
-	ret
-GLOBL( _sse_Attribute3f_end )
-
-
-/**
- * Generic handler for 2 float vector format data.  This can be used for
- * TexCoord2fv and possibly other functions.
- */
-
-GLOBL( _sse_Attribute2fv )
-	movl	4(%esp), %eax
-	movlps	(%eax), %xmm0
-	movlps	%xmm0, 0
-	ret
-GLOBL( _sse_Attribute2fv_end )
-
-
-/**
- * Generic handler for 2 float format data.  This can be used for
- * TexCoord2f and possibly other functions.
- */
-
-GLOBL( _sse_Attribute2f )
-	movlps	4(%esp), %xmm0
-	movlps	%xmm0, 0
-	ret
-GLOBL( _sse_Attribute2f_end )
-
-/*
-	MultiTexCoord2fv st0/st1
-*/
-GLOBL( _sse_MultiTexCoord2fv )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$3, %eax
-	movlps	(%ecx), %xmm0
-	movlps	%xmm0, 0xdeadbeef(,%eax,8)
-	ret
-GLOBL( _sse_MultiTexCoord2fv_end )
-
-/*
-	MultiTexCoord2fv
-*/
-GLOBL( _sse_MultiTexCoord2fv_2 )
-	movl	4(%esp), %eax
-	movl	8(%esp), %ecx
-	and	$3, %eax
-	movl	0(,%eax,4), %edx
-	movlps	(%ecx), %xmm0
-	movlps	%xmm0, (%edx)
-	ret
-GLOBL( _sse_MultiTexCoord2fv_2_end )
-
-/*
-	MultiTexCoord2f st0/st1
-*/
-GLOBL( _sse_MultiTexCoord2f )
-	movl	4(%esp), %eax
-	and	$3, %eax
-	movlps	8(%esp), %xmm0
-	movlps	%xmm0, 0xdeadbeef(,%eax,8)
-	ret
-GLOBL( _sse_MultiTexCoord2f_end )
-
-/*
-	MultiTexCoord2f
-*/
-GLOBL( _sse_MultiTexCoord2f_2 )
-	movl	4(%esp), %eax
-	movlps	8(%esp), %xmm0
-	and	$3,%eax
-	movl	0(,%eax,4), %eax
-	movlps	%xmm0, (%eax)
-	ret
-GLOBL( _sse_MultiTexCoord2f_2_end )
-#endif
-
-#if defined (__ELF__) && defined (__linux__)
-	.section .note.GNU-stack,"",%progbits
-#endif
diff --git a/src/mesa/drivers/dri/s3v/s3v_context.c b/src/mesa/drivers/dri/s3v/s3v_context.c
index 7b0aa0daee0..2d2f704ad77 100644
--- a/src/mesa/drivers/dri/s3v/s3v_context.c
+++ b/src/mesa/drivers/dri/s3v/s3v_context.c
@@ -6,7 +6,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -139,7 +139,7 @@ GLboolean s3vCreateContext(const __GLcontextModes *glVisual,
 	/* Initialize the software rasterizer and helper modules.
 	 */
 	_swrast_CreateContext( ctx );
-	_ac_CreateContext( ctx );
+	_vbo_CreateContext( ctx );
 	_tnl_CreateContext( ctx );
 	_swsetup_CreateContext( ctx );
 
diff --git a/src/mesa/drivers/dri/s3v/s3v_state.c b/src/mesa/drivers/dri/s3v/s3v_state.c
index fcea901c1c9..b86b618c117 100644
--- a/src/mesa/drivers/dri/s3v/s3v_state.c
+++ b/src/mesa/drivers/dri/s3v/s3v_state.c
@@ -10,7 +10,7 @@
 #include "colormac.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 
 /* #define DEBUG(str) printf str */
@@ -825,7 +825,7 @@ static void s3vDDUpdateState( GLcontext *ctx, GLuint new_state )
 {
 	_swrast_InvalidateState( ctx, new_state );
 	_swsetup_InvalidateState( ctx, new_state );
-	_ac_InvalidateState( ctx, new_state );
+	_vbo_InvalidateState( ctx, new_state );
 	_tnl_InvalidateState( ctx, new_state );
 	S3V_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/s3v/s3v_xmesa.c b/src/mesa/drivers/dri/s3v/s3v_xmesa.c
index 533424cdeab..c451f7452bf 100644
--- a/src/mesa/drivers/dri/s3v/s3v_xmesa.c
+++ b/src/mesa/drivers/dri/s3v/s3v_xmesa.c
@@ -13,7 +13,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 /* #define DEBUG(str) printf str */
 
@@ -38,7 +38,7 @@ s3vDestroyContext(__DRIcontextPrivate *driContextPriv)
     if (vmesa) {
       _swsetup_DestroyContext( vmesa->glCtx );
       _tnl_DestroyContext( vmesa->glCtx );
-      _ac_DestroyContext( vmesa->glCtx );
+      _vbo_DestroyContext( vmesa->glCtx );
       _swrast_DestroyContext( vmesa->glCtx );
 
       s3vFreeVB( vmesa->glCtx );
diff --git a/src/mesa/drivers/dri/savage/savage_xmesa.c b/src/mesa/drivers/dri/savage/savage_xmesa.c
index f51e321a4a6..ad79b9235c1 100644
--- a/src/mesa/drivers/dri/savage/savage_xmesa.c
+++ b/src/mesa/drivers/dri/savage/savage_xmesa.c
@@ -40,7 +40,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 
@@ -509,7 +509,7 @@ savageCreateContext( const __GLcontextModes *mesaVis,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    
    _swsetup_CreateContext( ctx );
@@ -599,7 +599,7 @@ savageDestroyContext(__DRIcontextPrivate *driContextPriv)
 
       _swsetup_DestroyContext(imesa->glCtx );
       _tnl_DestroyContext( imesa->glCtx );
-      _ac_DestroyContext( imesa->glCtx );
+      _vbo_DestroyContext( imesa->glCtx );
       _swrast_DestroyContext( imesa->glCtx );
 
       /* free the Mesa context */
diff --git a/src/mesa/drivers/dri/savage/savagestate.c b/src/mesa/drivers/dri/savage/savagestate.c
index 2f20dc23b6e..741a9dda4f7 100644
--- a/src/mesa/drivers/dri/savage/savagestate.c
+++ b/src/mesa/drivers/dri/savage/savagestate.c
@@ -41,7 +41,7 @@
 #include "savage_bci.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1673,7 +1673,7 @@ static void savageDDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    SAVAGE_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/sis/sis6326_state.c b/src/mesa/drivers/dri/sis/sis6326_state.c
index 6bc2c6de3bf..08402fb3e2a 100644
--- a/src/mesa/drivers/dri/sis/sis6326_state.c
+++ b/src/mesa/drivers/dri/sis/sis6326_state.c
@@ -37,7 +37,7 @@
 #include "enums.h"
 #include "colormac.h"
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -645,7 +645,7 @@ sis6326DDInvalidateState( GLcontext *ctx, GLuint new_state )
 
 	_swrast_InvalidateState( ctx, new_state );
 	_swsetup_InvalidateState( ctx, new_state );
-	_ac_InvalidateState( ctx, new_state );
+	_vbo_InvalidateState( ctx, new_state );
 	_tnl_InvalidateState( ctx, new_state );
 	smesa->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/sis/sis_context.c b/src/mesa/drivers/dri/sis/sis_context.c
index a300a080ec0..89b81da347a 100644
--- a/src/mesa/drivers/dri/sis/sis_context.c
+++ b/src/mesa/drivers/dri/sis/sis_context.c
@@ -53,7 +53,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -306,7 +306,7 @@ sisCreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -353,7 +353,7 @@ sisDestroyContext ( __DRIcontextPrivate *driContextPriv )
    if ( smesa != NULL ) {
       _swsetup_DestroyContext( smesa->glCtx );
       _tnl_DestroyContext( smesa->glCtx );
-      _ac_DestroyContext( smesa->glCtx );
+      _vbo_DestroyContext( smesa->glCtx );
       _swrast_DestroyContext( smesa->glCtx );
 
       if (smesa->using_agp)
diff --git a/src/mesa/drivers/dri/sis/sis_state.c b/src/mesa/drivers/dri/sis/sis_state.c
index 2a10a5fbf62..33a2f089b82 100644
--- a/src/mesa/drivers/dri/sis/sis_state.c
+++ b/src/mesa/drivers/dri/sis/sis_state.c
@@ -42,7 +42,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "enums.h"
 #include "colormac.h"
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -707,7 +707,7 @@ sisDDInvalidateState( GLcontext *ctx, GLuint new_state )
 
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    smesa->NewGLState |= new_state;
 }
diff --git a/src/mesa/drivers/dri/tdfx/tdfx_context.c b/src/mesa/drivers/dri/tdfx/tdfx_context.c
index bd9dade89b5..a9163f49a8f 100644
--- a/src/mesa/drivers/dri/tdfx/tdfx_context.c
+++ b/src/mesa/drivers/dri/tdfx/tdfx_context.c
@@ -49,7 +49,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -328,7 +328,7 @@ GLboolean tdfxCreateContext( const __GLcontextModes *mesaVis,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -609,7 +609,7 @@ tdfxDestroyContext( __DRIcontextPrivate *driContextPriv )
 
       _swsetup_DestroyContext( fxMesa->glCtx );
       _tnl_DestroyContext( fxMesa->glCtx );
-      _ac_DestroyContext( fxMesa->glCtx );
+      _vbo_DestroyContext( fxMesa->glCtx );
       _swrast_DestroyContext( fxMesa->glCtx );
 
       tdfxFreeVB( fxMesa->glCtx );
diff --git a/src/mesa/drivers/dri/tdfx/tdfx_state.c b/src/mesa/drivers/dri/tdfx/tdfx_state.c
index 59e6549e5ef..42cb5dfaa38 100644
--- a/src/mesa/drivers/dri/tdfx/tdfx_state.c
+++ b/src/mesa/drivers/dri/tdfx/tdfx_state.c
@@ -46,7 +46,7 @@
 #include "teximage.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
 #include "swrast_setup/swrast_setup.h"
@@ -1234,7 +1234,7 @@ static void tdfxDDInvalidateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    TDFX_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/trident/trident_context.c b/src/mesa/drivers/dri/trident/trident_context.c
index d3806586475..dbbd1ac0c69 100644
--- a/src/mesa/drivers/dri/trident/trident_context.c
+++ b/src/mesa/drivers/dri/trident/trident_context.c
@@ -30,7 +30,7 @@
 
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/tnl.h"
 #include "tnl/t_pipeline.h"
@@ -143,7 +143,7 @@ tridentCreateContext( const __GLcontextModes *glVisual,
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext( ctx );
-   _ac_CreateContext( ctx );
+   _vbo_CreateContext( ctx );
    _tnl_CreateContext( ctx );
    _swsetup_CreateContext( ctx );
 
@@ -183,7 +183,7 @@ tridentDestroyContext(__DRIcontextPrivate *driContextPriv)
     if (tmesa) {
       _swsetup_DestroyContext( tmesa->glCtx );
       _tnl_DestroyContext( tmesa->glCtx );
-      _ac_DestroyContext( tmesa->glCtx );
+      _vbo_DestroyContext( tmesa->glCtx );
       _swrast_DestroyContext( tmesa->glCtx );
 
       /* free the Mesa context */
diff --git a/src/mesa/drivers/dri/trident/trident_state.c b/src/mesa/drivers/dri/trident/trident_state.c
index 037cce5f862..5303bd422ed 100644
--- a/src/mesa/drivers/dri/trident/trident_state.c
+++ b/src/mesa/drivers/dri/trident/trident_state.c
@@ -26,7 +26,7 @@
  */
 #include "trident_context.h"
 #include "trident_lock.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
@@ -468,7 +468,7 @@ tridentDDUpdateState( GLcontext *ctx, GLuint new_state )
 {
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
    TRIDENT_CONTEXT(ctx)->new_gl_state |= new_state;
 }
diff --git a/src/mesa/drivers/dri/unichrome/server/via_dri.c b/src/mesa/drivers/dri/unichrome/server/via_dri.c
index 2fe61e40390..6944bd66f9c 100644
--- a/src/mesa/drivers/dri/unichrome/server/via_dri.c
+++ b/src/mesa/drivers/dri/unichrome/server/via_dri.c
@@ -1179,6 +1179,7 @@ static int viaInitFBDev(DRIDriverContext *ctx)
         pVia->Chipset = VIA_KM400;
         break;
     case PCI_CHIP_VT3204:
+    case PCI_CHIP_VT3344:
         pVia->Chipset = VIA_K8M800;
         break;
     case PCI_CHIP_VT3259:
diff --git a/src/mesa/drivers/dri/unichrome/server/via_regs.h b/src/mesa/drivers/dri/unichrome/server/via_regs.h
index 5c5eecbe831..87e1e9daa92 100644
--- a/src/mesa/drivers/dri/unichrome/server/via_regs.h
+++ b/src/mesa/drivers/dri/unichrome/server/via_regs.h
@@ -44,6 +44,7 @@
 #define PCI_CHIP_VT7205         0x7205
 #define PCI_CHIP_VT3204         0x3108
 #define PCI_CHIP_VT3259         0x3118
+#define PCI_CHIP_VT3344         0x3344
 
 
 #define BIOS_BSIZE              1024
diff --git a/src/mesa/drivers/dri/unichrome/via_context.c b/src/mesa/drivers/dri/unichrome/via_context.c
index 49c24656551..bc5a414df60 100644
--- a/src/mesa/drivers/dri/unichrome/via_context.c
+++ b/src/mesa/drivers/dri/unichrome/via_context.c
@@ -42,7 +42,7 @@
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 #include "tnl/t_pipeline.h"
 
@@ -583,7 +583,7 @@ viaCreateContext(const __GLcontextModes *visual,
     /* Initialize the software rasterizer and helper modules.
      */
     _swrast_CreateContext(ctx);
-    _ac_CreateContext(ctx);
+    _vbo_CreateContext(ctx);
     _tnl_CreateContext(ctx);
     _swsetup_CreateContext(ctx);
 
@@ -708,7 +708,7 @@ viaDestroyContext(__DRIcontextPrivate *driContextPriv)
 	
 	_swsetup_DestroyContext(vmesa->glCtx);
         _tnl_DestroyContext(vmesa->glCtx);
-        _ac_DestroyContext(vmesa->glCtx);
+        _vbo_DestroyContext(vmesa->glCtx);
         _swrast_DestroyContext(vmesa->glCtx);
         /* free the Mesa context */
 	_mesa_destroy_context(vmesa->glCtx);
diff --git a/src/mesa/drivers/dri/unichrome/via_state.c b/src/mesa/drivers/dri/unichrome/via_state.c
index 2c9cfca522e..30b9dc289a7 100644
--- a/src/mesa/drivers/dri/unichrome/via_state.c
+++ b/src/mesa/drivers/dri/unichrome/via_state.c
@@ -40,7 +40,7 @@
 #include "via_3d_reg.h"
 
 #include "swrast/swrast.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "swrast_setup/swrast_setup.h"
 
@@ -1507,7 +1507,7 @@ static void viaInvalidateState(GLcontext *ctx, GLuint newState)
 
     _swrast_InvalidateState(ctx, newState);
     _swsetup_InvalidateState(ctx, newState);
-    _ac_InvalidateState(ctx, newState);
+    _vbo_InvalidateState(ctx, newState);
     _tnl_InvalidateState(ctx, newState);
 }
 
diff --git a/src/mesa/drivers/fbdev/glfbdev.c b/src/mesa/drivers/fbdev/glfbdev.c
index bd9d3ac0ad9..f9d57dbdd81 100644
--- a/src/mesa/drivers/fbdev/glfbdev.c
+++ b/src/mesa/drivers/fbdev/glfbdev.c
@@ -54,7 +54,7 @@
 #include "texformat.h"
 #include "teximage.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
@@ -143,7 +143,7 @@ update_state( GLcontext *ctx, GLuint new_state )
    /* not much to do here - pass it on */
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
 }
 
@@ -776,7 +776,7 @@ glFBDevCreateContext( const GLFBDevVisualPtr visual, GLFBDevContextPtr share )
    /* Create module contexts */
    glctx = (GLcontext *) &ctx->glcontext;
    _swrast_CreateContext( glctx );
-   _ac_CreateContext( glctx );
+   _vbo_CreateContext( glctx );
    _tnl_CreateContext( glctx );
    _swsetup_CreateContext( glctx );
    _swsetup_Wakeup( glctx );
diff --git a/src/mesa/drivers/ggi/ggimesa.c b/src/mesa/drivers/ggi/ggimesa.c
index 47d7f2bdb99..85592d98dab 100644
--- a/src/mesa/drivers/ggi/ggimesa.c
+++ b/src/mesa/drivers/ggi/ggimesa.c
@@ -39,7 +39,7 @@
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "teximage.h"
 #include "texformat.h"
 #include "texstore.h"
@@ -563,7 +563,7 @@ ggi_mesa_context_t ggiMesaCreateContext(ggi_visual_t vis)
         _mesa_enable_sw_extensions(ctx->gl_ctx);
 	
 	_swrast_CreateContext(ctx->gl_ctx);
-	_ac_CreateContext(ctx->gl_ctx);
+	_vbo_CreateContext(ctx->gl_ctx);
 	_tnl_CreateContext(ctx->gl_ctx);
 	_swsetup_CreateContext(ctx->gl_ctx);
 	
diff --git a/src/mesa/drivers/glide/fxdd.c b/src/mesa/drivers/glide/fxdd.c
index ef60602c832..43d3cd94051 100644
--- a/src/mesa/drivers/glide/fxdd.c
+++ b/src/mesa/drivers/glide/fxdd.c
@@ -55,7 +55,7 @@
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 
 
 
@@ -1783,7 +1783,7 @@ fxDDInitFxMesaContext(fxMesaContext fxMesa)
    /* Initialize the software rasterizer and helper modules.
     */
    _swrast_CreateContext(ctx);
-   _ac_CreateContext(ctx);
+   _vbo_CreateContext(ctx);
    _tnl_CreateContext(ctx);
    _swsetup_CreateContext(ctx);
 
@@ -1826,7 +1826,7 @@ fxDDDestroyFxMesaContext(fxMesaContext fxMesa)
 {
    _swsetup_DestroyContext(fxMesa->glCtx);
    _tnl_DestroyContext(fxMesa->glCtx);
-   _ac_DestroyContext(fxMesa->glCtx);
+   _vbo_DestroyContext(fxMesa->glCtx);
    _swrast_DestroyContext(fxMesa->glCtx);
 
    if (fxMesa->state)
@@ -2101,7 +2101,7 @@ fxDDUpdateDDPointers(GLcontext * ctx, GLuint new_state)
    }
 
    _swrast_InvalidateState(ctx, new_state);
-   _ac_InvalidateState(ctx, new_state);
+   _vbo_InvalidateState(ctx, new_state);
    _tnl_InvalidateState(ctx, new_state);
    _swsetup_InvalidateState(ctx, new_state);
 
diff --git a/src/mesa/drivers/osmesa/osmesa.c b/src/mesa/drivers/osmesa/osmesa.c
index 0b027eddec5..c4fc8821e50 100644
--- a/src/mesa/drivers/osmesa/osmesa.c
+++ b/src/mesa/drivers/osmesa/osmesa.c
@@ -41,7 +41,6 @@
 #include "imports.h"
 #include "mtypes.h"
 #include "renderbuffer.h"
-#include "array_cache/acache.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast/s_context.h"
@@ -51,6 +50,7 @@
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
 #include "drivers/common/driverfuncs.h"
+#include "vbo/vbo.h"
 
 
 
@@ -110,8 +110,8 @@ osmesa_update_state( GLcontext *ctx, GLuint new_state )
    /* easy - just propogate */
    _swrast_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
 }
 
 
@@ -1261,7 +1261,7 @@ OSMesaCreateContextExt( GLenum format, GLint depthBits, GLint stencilBits,
          TNLcontext *tnl;
 
 	 if (!_swrast_CreateContext( ctx ) ||
-             !_ac_CreateContext( ctx ) ||
+             !_vbo_CreateContext( ctx ) ||
              !_tnl_CreateContext( ctx ) ||
              !_swsetup_CreateContext( ctx )) {
             _mesa_destroy_visual(osmesa->gl_visual);
@@ -1299,7 +1299,7 @@ OSMesaDestroyContext( OSMesaContext osmesa )
    if (osmesa) {
       _swsetup_DestroyContext( &osmesa->mesa );
       _tnl_DestroyContext( &osmesa->mesa );
-      _ac_DestroyContext( &osmesa->mesa );
+      _vbo_DestroyContext( &osmesa->mesa );
       _swrast_DestroyContext( &osmesa->mesa );
 
       _mesa_destroy_visual( osmesa->gl_visual );
diff --git a/src/mesa/drivers/svga/svgamesa15.c b/src/mesa/drivers/svga/svgamesa15.c
index a4c0901e38f..ae5104d0c01 100644
--- a/src/mesa/drivers/svga/svgamesa15.c
+++ b/src/mesa/drivers/svga/svgamesa15.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa15.c,v 1.12 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa15.c,v 1.11.36.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/svga/svgamesa16.c b/src/mesa/drivers/svga/svgamesa16.c
index 999c6331c9c..a59937bfb4b 100644
--- a/src/mesa/drivers/svga/svgamesa16.c
+++ b/src/mesa/drivers/svga/svgamesa16.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa16.c,v 1.12 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa16.c,v 1.11.36.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/svga/svgamesa24.c b/src/mesa/drivers/svga/svgamesa24.c
index 93c84167ebb..dd15bf38db2 100644
--- a/src/mesa/drivers/svga/svgamesa24.c
+++ b/src/mesa/drivers/svga/svgamesa24.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa24.c,v 1.13 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa24.c,v 1.12.36.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/svga/svgamesa32.c b/src/mesa/drivers/svga/svgamesa32.c
index 5a3a6776ce7..4da18795d8c 100644
--- a/src/mesa/drivers/svga/svgamesa32.c
+++ b/src/mesa/drivers/svga/svgamesa32.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa32.c,v 1.13 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa32.c,v 1.12.36.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/svga/svgamesa8.c b/src/mesa/drivers/svga/svgamesa8.c
index e3f5eaebf6a..4264fcd9599 100644
--- a/src/mesa/drivers/svga/svgamesa8.c
+++ b/src/mesa/drivers/svga/svgamesa8.c
@@ -1,4 +1,4 @@
-/* $Id: svgamesa8.c,v 1.10 2006/11/01 19:35:23 brianp Exp $ */
+/* $Id: svgamesa8.c,v 1.9.10.1 2006/11/02 12:02:17 alanh Exp $ */
 
 /*
  * Mesa 3-D graphics library
diff --git a/src/mesa/drivers/windows/gdi/mesa.def b/src/mesa/drivers/windows/gdi/mesa.def
index f0b275730ca..8d654281215 100644
--- a/src/mesa/drivers/windows/gdi/mesa.def
+++ b/src/mesa/drivers/windows/gdi/mesa.def
@@ -859,9 +859,9 @@ EXPORTS
 	wglGetExtensionsStringARB
 ;
 ; Mesa internals - mostly for OSMESA
-	_ac_CreateContext
-	_ac_DestroyContext
-	_ac_InvalidateState
+	_vbo_CreateContext
+	_vbo_DestroyContext
+	_vbo_InvalidateState
 	_glapi_check_multithread
 	_glapi_get_context
 	_glapi_get_proc_address
diff --git a/src/mesa/drivers/windows/gdi/wmesa.c b/src/mesa/drivers/windows/gdi/wmesa.c
index 8c3229ab14c..078b590af87 100644
--- a/src/mesa/drivers/windows/gdi/wmesa.c
+++ b/src/mesa/drivers/windows/gdi/wmesa.c
@@ -11,7 +11,7 @@
 #include "framebuffer.h"
 #include "renderbuffer.h"
 #include "drivers/common/driverfuncs.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "tnl/tnl.h"
@@ -1125,7 +1125,7 @@ static void wmesa_update_state(GLcontext *ctx, GLuint new_state)
 {
     _swrast_InvalidateState(ctx, new_state);
     _swsetup_InvalidateState(ctx, new_state);
-    _ac_InvalidateState(ctx, new_state);
+    _vbo_InvalidateState(ctx, new_state);
     _tnl_InvalidateState(ctx, new_state);
 
     /* TODO - This code is not complete yet because I 
@@ -1242,7 +1242,7 @@ WMesaContext WMesaCreateContext(HDC hDC,
   
     /* Initialize the software rasterizer and helper modules. */
     if (!_swrast_CreateContext(ctx) ||
-        !_ac_CreateContext(ctx) ||
+        !_vbo_CreateContext(ctx) ||
         !_tnl_CreateContext(ctx) ||
 	!_swsetup_CreateContext(ctx)) {
 	_mesa_free_context_data(ctx);
@@ -1285,7 +1285,7 @@ void WMesaDestroyContext( WMesaContext pwc )
     
     _swsetup_DestroyContext(ctx);
     _tnl_DestroyContext(ctx);
-    _ac_DestroyContext(ctx);
+    _vbo_DestroyContext(ctx);
     _swrast_DestroyContext(ctx);
     
     _mesa_free_context_data(ctx);
diff --git a/src/mesa/drivers/windows/gldirect/dglcontext.c b/src/mesa/drivers/windows/gldirect/dglcontext.c
index 5dafe14518d..4ad7a76e67f 100644
--- a/src/mesa/drivers/windows/gldirect/dglcontext.c
+++ b/src/mesa/drivers/windows/gldirect/dglcontext.c
@@ -1453,7 +1453,7 @@ SkipPrimaryCreate:
 #ifdef _USE_GLD3_WGL
 	// Init Mesa internals
 	_swrast_CreateContext( lpCtx->glCtx );
-	_ac_CreateContext( lpCtx->glCtx );
+	_vbo_CreateContext( lpCtx->glCtx );
 	_tnl_CreateContext( lpCtx->glCtx );
 	_swsetup_CreateContext( lpCtx->glCtx );
 
diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_driver_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_driver_dx7.c
index cad299540d7..d5fa6428006 100644
--- a/src/mesa/drivers/windows/gldirect/dx7/gld_driver_dx7.c
+++ b/src/mesa/drivers/windows/gldirect/dx7/gld_driver_dx7.c
@@ -52,7 +52,7 @@
 #include "texformat.h"
 #include "teximage.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast_setup/ss_context.h"
 #include "tnl/tnl.h"
@@ -868,7 +868,7 @@ void gld_update_state_DX7(
 		return;
 
 	_swsetup_InvalidateState( ctx, new_state );
-	_ac_InvalidateState( ctx, new_state );
+	_vbo_InvalidateState( ctx, new_state );
 	_tnl_InvalidateState( ctx, new_state );
 
 	// SetupIndex will be used in the pipelines for choosing setup function
diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_ext_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_ext_dx7.c
index 6be41a80dd3..ba60980bbe8 100644
--- a/src/mesa/drivers/windows/gldirect/dx7/gld_ext_dx7.c
+++ b/src/mesa/drivers/windows/gldirect/dx7/gld_ext_dx7.c
@@ -58,7 +58,7 @@
 #include "mtypes.h"
 #include "texformat.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast_setup/ss_context.h"
 #include "tnl/tnl.h"
diff --git a/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c b/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c
index 5da25003c08..c99ba0bba55 100644
--- a/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c
+++ b/src/mesa/drivers/windows/gldirect/dx7/gld_primitive_dx7.c
@@ -54,7 +54,7 @@
 #include "mtypes.h"
 #include "texformat.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast_setup/ss_context.h"
diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_driver_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_driver_dx8.c
index 0a6d9f8555c..7afa9190cdf 100644
--- a/src/mesa/drivers/windows/gldirect/dx8/gld_driver_dx8.c
+++ b/src/mesa/drivers/windows/gldirect/dx8/gld_driver_dx8.c
@@ -52,7 +52,7 @@
 #include "texformat.h"
 #include "teximage.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast_setup/ss_context.h"
 #include "tnl/tnl.h"
@@ -850,7 +850,7 @@ void gld_update_state_DX8(
 		return;
 
 	_swsetup_InvalidateState( ctx, new_state );
-	_ac_InvalidateState( ctx, new_state );
+	_vbo_InvalidateState( ctx, new_state );
 	_tnl_InvalidateState( ctx, new_state );
 
 	// SetupIndex will be used in the pipelines for choosing setup function
diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_ext_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_ext_dx8.c
index 108f12a9d16..b51bba9b3ca 100644
--- a/src/mesa/drivers/windows/gldirect/dx8/gld_ext_dx8.c
+++ b/src/mesa/drivers/windows/gldirect/dx8/gld_ext_dx8.c
@@ -58,7 +58,7 @@
 #include "mtypes.h"
 #include "texformat.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast_setup/ss_context.h"
 #include "tnl/tnl.h"
diff --git a/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c b/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c
index 700b5200862..a5b5462f039 100644
--- a/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c
+++ b/src/mesa/drivers/windows/gldirect/dx8/gld_primitive_dx8.c
@@ -54,7 +54,7 @@
 #include "mtypes.h"
 #include "texformat.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast_setup/ss_context.h"
diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_driver_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_driver_dx9.c
index 1b01cb1f7f3..c191564d6eb 100644
--- a/src/mesa/drivers/windows/gldirect/dx9/gld_driver_dx9.c
+++ b/src/mesa/drivers/windows/gldirect/dx9/gld_driver_dx9.c
@@ -52,7 +52,7 @@
 #include "texformat.h"
 #include "teximage.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast_setup/ss_context.h"
 #include "tnl/tnl.h"
@@ -876,7 +876,7 @@ void gld_update_state_DX9(
 		return;
 
 	_swsetup_InvalidateState( ctx, new_state );
-	_ac_InvalidateState( ctx, new_state );
+	_vbo_InvalidateState( ctx, new_state );
 	_tnl_InvalidateState( ctx, new_state );
 
 	// SetupIndex will be used in the pipelines for choosing setup function
diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_ext_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_ext_dx9.c
index 745c987602b..e8c73a6ff80 100644
--- a/src/mesa/drivers/windows/gldirect/dx9/gld_ext_dx9.c
+++ b/src/mesa/drivers/windows/gldirect/dx9/gld_ext_dx9.c
@@ -58,7 +58,7 @@
 #include "mtypes.h"
 #include "texformat.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast_setup/ss_context.h"
 #include "tnl/tnl.h"
diff --git a/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c b/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c
index 65fd821276e..403a9d5f86e 100644
--- a/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c
+++ b/src/mesa/drivers/windows/gldirect/dx9/gld_primitive_dx9.c
@@ -54,7 +54,7 @@
 #include "mtypes.h"
 #include "texformat.h"
 #include "texstore.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast_setup/ss_context.h"
diff --git a/src/mesa/drivers/windows/gldirect/mesasw/gld_wgl_mesasw.c b/src/mesa/drivers/windows/gldirect/mesasw/gld_wgl_mesasw.c
index 0f11b4fe513..b590dc795a0 100644
--- a/src/mesa/drivers/windows/gldirect/mesasw/gld_wgl_mesasw.c
+++ b/src/mesa/drivers/windows/gldirect/mesasw/gld_wgl_mesasw.c
@@ -54,7 +54,7 @@
 #include "texformat.h"
 #include "texstore.h"
 #include "teximage.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
 #include "swrast/s_context.h"
@@ -1321,7 +1321,7 @@ static void wmesa_update_state(
 {
     _swrast_InvalidateState( ctx, new_state );
 	_swsetup_InvalidateState( ctx, new_state );
-	_ac_InvalidateState( ctx, new_state );
+	_vbo_InvalidateState( ctx, new_state );
 	_tnl_InvalidateState( ctx, new_state );
 }
 
@@ -1631,7 +1631,7 @@ BOOL gldInitialiseMesa_MesaSW(
 	_mesa_enable_1_3_extensions(ctx);
 	
 //	_swrast_CreateContext( ctx );
-//	_ac_CreateContext( ctx );
+//	_vbo_CreateContext( ctx );
 //	_tnl_CreateContext( ctx );
 //	_swsetup_CreateContext( ctx );
 	
diff --git a/src/mesa/drivers/windows/icd/mesa.def b/src/mesa/drivers/windows/icd/mesa.def
index 1826192d307..465b380a0cb 100644
--- a/src/mesa/drivers/windows/icd/mesa.def
+++ b/src/mesa/drivers/windows/icd/mesa.def
@@ -24,9 +24,9 @@ EXPORTS
 
 ;
 ; Mesa internals - mostly for OSMESA
-	_ac_CreateContext
-	_ac_DestroyContext
-	_ac_InvalidateState
+	_vbo_CreateContext
+	_vbo_DestroyContext
+	_vbo_InvalidateState
 	_glapi_get_context
 	_glapi_get_proc_address
 	_mesa_buffer_data
diff --git a/src/mesa/drivers/x11/glxapi.c b/src/mesa/drivers/x11/glxapi.c
index 7945f1c5b06..973f3940451 100644
--- a/src/mesa/drivers/x11/glxapi.c
+++ b/src/mesa/drivers/x11/glxapi.c
@@ -241,7 +241,7 @@ glXGetCurrentContext(void)
 #if defined(GLX_USE_TLS)
    return CurrentContext;
 #elif defined(THREADS)
-   return _glthread_GetTSD(&ContextTSD);
+   return (GLXContext) _glthread_GetTSD(&ContextTSD);
 #else
    return CurrentContext;
 #endif
diff --git a/src/mesa/drivers/x11/xm_api.c b/src/mesa/drivers/x11/xm_api.c
index b50c4757afc..b0ef422de2e 100644
--- a/src/mesa/drivers/x11/xm_api.c
+++ b/src/mesa/drivers/x11/xm_api.c
@@ -74,7 +74,7 @@
 #include "renderbuffer.h"
 #include "swrast/swrast.h"
 #include "swrast_setup/swrast_setup.h"
-#include "array_cache/acache.h"
+#include "vbo/vbo.h"
 #include "tnl/tnl.h"
 #include "tnl/t_context.h"
 #include "tnl/t_pipeline.h"
@@ -1545,6 +1545,13 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
     _mesa_enable_extension(mesaCtx, "GL_EXT_timer_query");
 #endif
 
+#ifdef XFree86Server
+   /* If we're running in the X server, do bounds checking to prevent
+    * segfaults and server crashes!
+    */
+   mesaCtx->Const.CheckArrayBounds = GL_TRUE;
+#endif
+
    /* finish up xmesa context initializations */
    c->swapbytes = CHECK_BYTE_ORDER(v) ? GL_FALSE : GL_TRUE;
    c->xm_visual = v;
@@ -1555,7 +1562,7 @@ XMesaContext XMesaCreateContext( XMesaVisual v, XMesaContext share_list )
    /* Initialize the software rasterizer and helper modules.
     */
    if (!_swrast_CreateContext( mesaCtx ) ||
-       !_ac_CreateContext( mesaCtx ) ||
+       !_vbo_CreateContext( mesaCtx ) ||
        !_tnl_CreateContext( mesaCtx ) ||
        !_swsetup_CreateContext( mesaCtx )) {
       _mesa_free_context_data(&c->mesa);
@@ -1587,7 +1594,7 @@ void XMesaDestroyContext( XMesaContext c )
    _swsetup_DestroyContext( mesaCtx );
    _swrast_DestroyContext( mesaCtx );
    _tnl_DestroyContext( mesaCtx );
-   _ac_DestroyContext( mesaCtx );
+   _vbo_DestroyContext( mesaCtx );
    _mesa_free_context_data( mesaCtx );
    _mesa_free( c );
 }
diff --git a/src/mesa/drivers/x11/xm_dd.c b/src/mesa/drivers/x11/xm_dd.c
index d272630d18a..dbac3b8cdb6 100644
--- a/src/mesa/drivers/x11/xm_dd.c
+++ b/src/mesa/drivers/x11/xm_dd.c
@@ -47,7 +47,6 @@
 #include "texstore.h"
 #include "texformat.h"
 #include "xmesaP.h"
-#include "array_cache/acache.h"
 #include "swrast/swrast.h"
 #include "swrast/s_context.h"
 #include "swrast_setup/swrast_setup.h"
@@ -829,8 +828,8 @@ xmesa_update_state( GLcontext *ctx, GLbitfield new_state )
     * modules.  The X11 driver has no internal GL-dependent state.
     */
    _swrast_InvalidateState( ctx, new_state );
-   _ac_InvalidateState( ctx, new_state );
    _tnl_InvalidateState( ctx, new_state );
+   _vbo_InvalidateState( ctx, new_state );
    _swsetup_InvalidateState( ctx, new_state );
 
    if (ctx->DrawBuffer->Name != 0)
diff --git a/src/mesa/glapi/Makefile b/src/mesa/glapi/Makefile
index 2f8da126b8e..c52c0903478 100644
--- a/src/mesa/glapi/Makefile
+++ b/src/mesa/glapi/Makefile
@@ -26,6 +26,8 @@ GLX_DIR   = $(XORG_BASE)/GL/glx
 SERVER_OUTPUTS = $(GLX_DIR)/indirect_dispatch.c \
 	$(GLX_DIR)/indirect_dispatch_swap.c \
 	$(GLX_DIR)/indirect_dispatch.h \
+	$(GLX_DIR)/indirect_reqsize.c \
+	$(GLX_DIR)/indirect_reqsize.h \
 	$(GLX_DIR)/indirect_size_get.c \
 	$(GLX_DIR)/indirect_size_get.h \
 	$(GLX_DIR)/indirect_table.c
@@ -37,7 +39,7 @@ API_XML = gl_API.xml \
 COMMON = gl_XML.py glX_XML.py license.py $(API_XML) typeexpr.py
 COMMON_GLX = $(COMMON) glX_API.xml glX_XML.py glX_proto_common.py
 
-INDENT_FLAGS = -i4 -nut -br -brs -npcs -ce
+INDENT_FLAGS = -i4 -nut -br -brs -npcs -ce -T GLubyte -T GLbyte -T Bool
 
 all: $(OUTPUTS)
 
@@ -104,6 +106,12 @@ $(GLX_DIR)/indirect_size_get.h: $(COMMON_GLX) glX_proto_size.py
 $(GLX_DIR)/indirect_size_get.c: $(COMMON_GLX) glX_proto_size.py
 	$(PYTHON2) $(PYTHON_FLAGS) glX_proto_size.py -m size_c | indent $(INDENT_FLAGS) > $@
 
+$(GLX_DIR)/indirect_reqsize.h: $(COMMON_GLX) glX_proto_size.py
+	$(PYTHON2) $(PYTHON_FLAGS) glX_proto_size.py -m reqsize_h --only-get -h '_INDIRECT_SIZE_GET_H_' | indent $(INDENT_FLAGS) -l200 > $@
+
+$(GLX_DIR)/indirect_reqsize.c: $(COMMON_GLX) glX_proto_size.py
+	$(PYTHON2) $(PYTHON_FLAGS) glX_proto_size.py -m reqsize_c | indent $(INDENT_FLAGS) > $@
+
 $(GLX_DIR)/indirect_table.c: $(COMMON_GLX) glX_server_table.py glX_API.xml
 	$(PYTHON2) $(PYTHON_FLAGS) glX_server_table.py -f gl_and_glX_API.xml > $@
 
diff --git a/src/mesa/glapi/glX_proto_recv.py b/src/mesa/glapi/glX_proto_recv.py
index 527f6f10eef..20f75575cf1 100644
--- a/src/mesa/glapi/glX_proto_recv.py
+++ b/src/mesa/glapi/glX_proto_recv.py
@@ -81,31 +81,12 @@ class PrintGlxDispatchFunctions(glX_proto_common.glx_print_proto):
 		print '#include <GL/gl.h>'
 		print '#include <GL/glxproto.h>'
 
-
-		# FIXME: Since this block will require changes as other
-		# FIXME: platforms are added, it should probably be in a
-		# FIXME: header file that is not generated by a script.
-
-		if self.do_swap:
-			print '#ifdef __linux__'
-			print '#include <byteswap.h>'
-			print '#elif defined(__OpenBSD__)'
-			print '#include <sys/endian.h>'
-			print '#define bswap_16 __swap16'
-			print '#define bswap_32 __swap32'
-			print '#define bswap_64 __swap64'
-			print '#else'
-			print '#include <sys/endian.h>'
-			print '#define bswap_16 bswap16'
-			print '#define bswap_32 bswap32'
-			print '#define bswap_64 bswap64'
-			print '#endif'
-
 		print '#include <inttypes.h>'
 		print '#include "indirect_size.h"'
 		print '#include "indirect_size_get.h"'
 		print '#include "indirect_dispatch.h"'
 		print '#include "glxserver.h"'
+		print '#include "glxbyteorder.h"'
 		print '#include "indirect_util.h"'
 		print '#include "singlesize.h"'
 		print '#include "glapitable.h"'
diff --git a/src/mesa/glapi/glX_proto_size.py b/src/mesa/glapi/glX_proto_size.py
index 18bfa1b71bc..2b9a6433622 100644
--- a/src/mesa/glapi/glX_proto_size.py
+++ b/src/mesa/glapi/glX_proto_size.py
@@ -447,18 +447,10 @@ class PrintGlxReqSize_c(PrintGlxReqSize_common):
 		print ''
 		print '#include <GL/gl.h>'
 		print '#include "glxserver.h"'
+		print '#include "glxbyteorder.h"'
 		print '#include "indirect_size.h"'
 		print '#include "indirect_reqsize.h"'
 		print ''
-		print '#if defined(linux)'
-		print '#  include <byteswap.h>'
-		print '#  define SWAP_32(v)  do { (v) = bswap_32(v); } while(0)'
-		print '#else'
-		print '#  include <X11/misc.h>'
-		print '#  define SWAP_32(v)  do { char tmp; swapl(&v, tmp); } while(0)'
-		print '#endif'
-		
-		print ''
 		print '#define __GLX_PAD(x)  (((x) + 3) & ~3)'
 		print ''
 		print '#if defined(__CYGWIN__) || defined(__MINGW32__)'
@@ -538,7 +530,7 @@ class PrintGlxReqSize_c(PrintGlxReqSize_common):
 		if fixup:
 			print '    if (swap) {'
 			for name in fixup:
-				print '        SWAP_32( %s );' % (name)
+				print '        %s = bswap_32(%s);' % (name, name)
 			print '    }'
 
 		return
diff --git a/src/mesa/glapi/gl_API.xml b/src/mesa/glapi/gl_API.xml
index 905c4a03a67..bd31e8871b0 100644
--- a/src/mesa/glapi/gl_API.xml
+++ b/src/mesa/glapi/gl_API.xml
@@ -5318,7 +5318,7 @@
         <glx ignore="true"/>
     </function>
 
-    <function name="GetVertexAttribPointerv" alias="GetVertexAttribPointervARB">
+    <function name="GetVertexAttribPointerv" alias="GetVertexAttribPointervNV">
         <param name="index" type="GLuint"/>
         <param name="pname" type="GLenum"/>
         <param name="pointer" type="GLvoid **" output="true"/>
@@ -7209,7 +7209,7 @@
     <function name="GetVertexAttribPointervARB" alias="GetVertexAttribPointervNV">
         <param name="index" type="GLuint"/>
         <param name="pname" type="GLenum"/>
-        <param name="params" type="GLvoid **" output="true"/>
+        <param name="pointer" type="GLvoid **" output="true"/>
     </function>
 </category>
 
@@ -11234,7 +11234,7 @@
     <function name="GetVertexAttribPointervNV" offset="assign">
         <param name="index" type="GLuint"/>
         <param name="pname" type="GLenum"/>
-        <param name="params" type="GLvoid **" output="true"/>
+        <param name="pointer" type="GLvoid **" output="true"/>
         <glx handcode="true"/>
     </function>
 
diff --git a/src/mesa/glapi/glapi.c b/src/mesa/glapi/glapi.c
index d1ccc5fba69..348f12b3e74 100644
--- a/src/mesa/glapi/glapi.c
+++ b/src/mesa/glapi/glapi.c
@@ -1003,7 +1003,6 @@ _glapi_check_table(const struct _glapi_table *table)
       GLuint offset = (secondaryColor3fFunc - (char *) table) / sizeof(void *);
       assert(secondaryColor3fOffset == _gloffset_SecondaryColor3fEXT);
       assert(secondaryColor3fOffset == offset);
-      assert(_glapi_get_proc_address("glSecondaryColor3fEXT") == (_glapi_proc) &glSecondaryColor3fEXT);
    }
    {
       GLuint pointParameterivOffset = _glapi_get_proc_offset("glPointParameterivNV");
@@ -1011,7 +1010,6 @@ _glapi_check_table(const struct _glapi_table *table)
       GLuint offset = (pointParameterivFunc - (char *) table) / sizeof(void *);
       assert(pointParameterivOffset == _gloffset_PointParameterivNV);
       assert(pointParameterivOffset == offset);
-      assert(_glapi_get_proc_address("glPointParameterivNV") == (_glapi_proc) &glPointParameterivNV);
    }
    {
       GLuint setFenceOffset = _glapi_get_proc_offset("glSetFenceNV");
diff --git a/src/mesa/glapi/glapitable.h b/src/mesa/glapi/glapitable.h
index 026af32af8f..55548adcd3b 100644
--- a/src/mesa/glapi/glapitable.h
+++ b/src/mesa/glapi/glapitable.h
@@ -704,7 +704,7 @@ struct _glapi_table
    void (GLAPIENTRYP GetProgramStringNV)(GLuint id, GLenum pname, GLubyte * program); /* 664 */
    void (GLAPIENTRYP GetProgramivNV)(GLuint id, GLenum pname, GLint * params); /* 665 */
    void (GLAPIENTRYP GetTrackMatrixivNV)(GLenum target, GLuint address, GLenum pname, GLint * params); /* 666 */
-   void (GLAPIENTRYP GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** params); /* 667 */
+   void (GLAPIENTRYP GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** pointer); /* 667 */
    void (GLAPIENTRYP GetVertexAttribdvNV)(GLuint index, GLenum pname, GLdouble * params); /* 668 */
    void (GLAPIENTRYP GetVertexAttribfvNV)(GLuint index, GLenum pname, GLfloat * params); /* 669 */
    void (GLAPIENTRYP GetVertexAttribivNV)(GLuint index, GLenum pname, GLint * params); /* 670 */
diff --git a/src/mesa/glapi/glapitemp.h b/src/mesa/glapi/glapitemp.h
index 340a4f6a040..62407968ccc 100644
--- a/src/mesa/glapi/glapitemp.h
+++ b/src/mesa/glapi/glapitemp.h
@@ -4865,14 +4865,19 @@ KEYWORD1 void KEYWORD2 NAME(GetTrackMatrixivNV)(GLenum target, GLuint address, G
    DISPATCH(GetTrackMatrixivNV, (target, address, pname, params), (F, "glGetTrackMatrixivNV(0x%x, %d, 0x%x, %p);\n", target, address, pname, (const void *) params));
 }
 
-KEYWORD1 void KEYWORD2 NAME(GetVertexAttribPointervARB)(GLuint index, GLenum pname, GLvoid ** params)
+KEYWORD1 void KEYWORD2 NAME(GetVertexAttribPointerv)(GLuint index, GLenum pname, GLvoid ** pointer)
 {
-   DISPATCH(GetVertexAttribPointervNV, (index, pname, params), (F, "glGetVertexAttribPointervARB(%d, 0x%x, %p);\n", index, pname, (const void *) params));
+   DISPATCH(GetVertexAttribPointervNV, (index, pname, pointer), (F, "glGetVertexAttribPointerv(%d, 0x%x, %p);\n", index, pname, (const void *) pointer));
 }
 
-KEYWORD1 void KEYWORD2 NAME(GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** params)
+KEYWORD1 void KEYWORD2 NAME(GetVertexAttribPointervARB)(GLuint index, GLenum pname, GLvoid ** pointer)
 {
-   DISPATCH(GetVertexAttribPointervNV, (index, pname, params), (F, "glGetVertexAttribPointervNV(%d, 0x%x, %p);\n", index, pname, (const void *) params));
+   DISPATCH(GetVertexAttribPointervNV, (index, pname, pointer), (F, "glGetVertexAttribPointervARB(%d, 0x%x, %p);\n", index, pname, (const void *) pointer));
+}
+
+KEYWORD1 void KEYWORD2 NAME(GetVertexAttribPointervNV)(GLuint index, GLenum pname, GLvoid ** pointer)
+{
+   DISPATCH(GetVertexAttribPointervNV, (index, pname, pointer), (F, "glGetVertexAttribPointervNV(%d, 0x%x, %p);\n", index, pname, (const void *) pointer));
 }
 
 KEYWORD1 void KEYWORD2 NAME(GetVertexAttribdvNV)(GLuint index, GLenum pname, GLdouble * params)
@@ -6577,6 +6582,7 @@ static _glapi_proc UNUSED_TABLE_NAME[] = {
    TABLE_ENTRY(BindProgramARB),
    TABLE_ENTRY(DeleteProgramsARB),
    TABLE_ENTRY(GenProgramsARB),
+   TABLE_ENTRY(GetVertexAttribPointerv),
    TABLE_ENTRY(GetVertexAttribPointervARB),
    TABLE_ENTRY(IsProgramARB),
    TABLE_ENTRY(PointParameteri),
diff --git a/src/mesa/glapi/glprocs.h b/src/mesa/glapi/glprocs.h
index 905c65862d1..190d9ed149b 100644
--- a/src/mesa/glapi/glprocs.h
+++ b/src/mesa/glapi/glprocs.h
@@ -1086,6 +1086,7 @@ static const char gl_string_table[] =
     "glBindProgramARB\0"
     "glDeleteProgramsARB\0"
     "glGenProgramsARB\0"
+    "glGetVertexAttribPointerv\0"
     "glGetVertexAttribPointervARB\0"
     "glIsProgramARB\0"
     "glPointParameteri\0"
@@ -2229,11 +2230,12 @@ static const glprocs_table_t static_functions[] = {
     NAME_FUNC_OFFSET(18432, glDeleteProgramsNV, glDeleteProgramsNV, NULL, _gloffset_DeleteProgramsNV),
     NAME_FUNC_OFFSET(18452, glGenProgramsNV, glGenProgramsNV, NULL, _gloffset_GenProgramsNV),
     NAME_FUNC_OFFSET(18469, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
-    NAME_FUNC_OFFSET(18498, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
-    NAME_FUNC_OFFSET(18513, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
-    NAME_FUNC_OFFSET(18531, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
-    NAME_FUNC_OFFSET(18550, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT),
-    NAME_FUNC_OFFSET(18574, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(18495, glGetVertexAttribPointervNV, glGetVertexAttribPointervNV, NULL, _gloffset_GetVertexAttribPointervNV),
+    NAME_FUNC_OFFSET(18524, glIsProgramNV, glIsProgramNV, NULL, _gloffset_IsProgramNV),
+    NAME_FUNC_OFFSET(18539, glPointParameteriNV, glPointParameteriNV, NULL, _gloffset_PointParameteriNV),
+    NAME_FUNC_OFFSET(18557, glPointParameterivNV, glPointParameterivNV, NULL, _gloffset_PointParameterivNV),
+    NAME_FUNC_OFFSET(18576, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT),
+    NAME_FUNC_OFFSET(18600, gl_dispatch_stub_749, gl_dispatch_stub_749, NULL, _gloffset_BlendEquationSeparateEXT),
     NAME_FUNC_OFFSET(-1, NULL, NULL, NULL, 0)
 };
 
diff --git a/src/mesa/main/api_noop.c b/src/mesa/main/api_noop.c
index 33d44e43296..0c1a35361f1 100644
--- a/src/mesa/main/api_noop.c
+++ b/src/mesa/main/api_noop.c
@@ -45,7 +45,7 @@
 static void GLAPIENTRY _mesa_noop_EdgeFlag( GLboolean b )
 {
    GET_CURRENT_CONTEXT(ctx);
-   ctx->Current.EdgeFlag = b;
+   ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] = (GLfloat)b;
 }
 
 static void GLAPIENTRY _mesa_noop_Indexf( GLfloat f )
diff --git a/src/mesa/main/arrayobj.c b/src/mesa/main/arrayobj.c
index 852b9aaee98..d601ee461e6 100644
--- a/src/mesa/main/arrayobj.c
+++ b/src/mesa/main/arrayobj.c
@@ -114,40 +114,34 @@ _mesa_initialize_array_object( GLcontext *ctx,
    obj->Vertex.StrideB = 0;
    obj->Vertex.Ptr = NULL;
    obj->Vertex.Enabled = GL_FALSE;
-   obj->Vertex.Flags = CA_CLIENT_DATA;
    obj->Normal.Type = GL_FLOAT;
    obj->Normal.Stride = 0;
    obj->Normal.StrideB = 0;
    obj->Normal.Ptr = NULL;
    obj->Normal.Enabled = GL_FALSE;
-   obj->Normal.Flags = CA_CLIENT_DATA;
    obj->Color.Size = 4;
    obj->Color.Type = GL_FLOAT;
    obj->Color.Stride = 0;
    obj->Color.StrideB = 0;
    obj->Color.Ptr = NULL;
    obj->Color.Enabled = GL_FALSE;
-   obj->Color.Flags = CA_CLIENT_DATA;
    obj->SecondaryColor.Size = 4;
    obj->SecondaryColor.Type = GL_FLOAT;
    obj->SecondaryColor.Stride = 0;
    obj->SecondaryColor.StrideB = 0;
    obj->SecondaryColor.Ptr = NULL;
    obj->SecondaryColor.Enabled = GL_FALSE;
-   obj->SecondaryColor.Flags = CA_CLIENT_DATA;
    obj->FogCoord.Size = 1;
    obj->FogCoord.Type = GL_FLOAT;
    obj->FogCoord.Stride = 0;
    obj->FogCoord.StrideB = 0;
    obj->FogCoord.Ptr = NULL;
    obj->FogCoord.Enabled = GL_FALSE;
-   obj->FogCoord.Flags = CA_CLIENT_DATA;
    obj->Index.Type = GL_FLOAT;
    obj->Index.Stride = 0;
    obj->Index.StrideB = 0;
    obj->Index.Ptr = NULL;
    obj->Index.Enabled = GL_FALSE;
-   obj->Index.Flags = CA_CLIENT_DATA;
    for (i = 0; i < MAX_TEXTURE_UNITS; i++) {
       obj->TexCoord[i].Size = 4;
       obj->TexCoord[i].Type = GL_FLOAT;
@@ -155,13 +149,11 @@ _mesa_initialize_array_object( GLcontext *ctx,
       obj->TexCoord[i].StrideB = 0;
       obj->TexCoord[i].Ptr = NULL;
       obj->TexCoord[i].Enabled = GL_FALSE;
-      obj->TexCoord[i].Flags = CA_CLIENT_DATA;
    }
    obj->EdgeFlag.Stride = 0;
    obj->EdgeFlag.StrideB = 0;
    obj->EdgeFlag.Ptr = NULL;
    obj->EdgeFlag.Enabled = GL_FALSE;
-   obj->EdgeFlag.Flags = CA_CLIENT_DATA;
    for (i = 0; i < VERT_ATTRIB_MAX; i++) {
       obj->VertexAttrib[i].Size = 4;
       obj->VertexAttrib[i].Type = GL_FLOAT;
@@ -170,7 +162,6 @@ _mesa_initialize_array_object( GLcontext *ctx,
       obj->VertexAttrib[i].Ptr = NULL;
       obj->VertexAttrib[i].Enabled = GL_FALSE;
       obj->VertexAttrib[i].Normalized = GL_FALSE;
-      obj->VertexAttrib[i].Flags = CA_CLIENT_DATA;
    }
 
 #if FEATURE_ARB_vertex_buffer_object
diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c
index 9993a0021bc..2b1a35f3de3 100644
--- a/src/mesa/main/attrib.c
+++ b/src/mesa/main/attrib.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.2
+ * Version:  6.5.3
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  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"),
@@ -1285,6 +1285,12 @@ _mesa_PushClientAttrib(GLbitfield mask)
       attr = MALLOC_STRUCT( gl_array_attrib );
       obj = MALLOC_STRUCT( gl_array_object );
 
+#if FEATURE_ARB_vertex_buffer_object
+      /* increment ref counts since we're copying pointers to these objects */
+      ctx->Array.ArrayBufferObj->RefCount++;
+      ctx->Array.ElementArrayBufferObj->RefCount++;
+#endif
+
       MEMCPY( attr, &ctx->Array, sizeof(struct gl_array_attrib) );
       MEMCPY( obj, ctx->Array.ArrayObj, sizeof(struct gl_array_object) );
 
@@ -1359,6 +1365,13 @@ _mesa_PopClientAttrib(void)
 
 	    _mesa_BindVertexArrayAPPLE( data->ArrayObj->Name );
 	    
+#if FEATURE_ARB_vertex_buffer_object
+            _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB,
+                                data->ArrayBufferObj->Name);
+            _mesa_BindBufferARB(GL_ELEMENT_ARRAY_BUFFER_ARB,
+                                data->ElementArrayBufferObj->Name);
+#endif
+
 	    MEMCPY( ctx->Array.ArrayObj, data->ArrayObj,
 		    sizeof( struct gl_array_object ) );
 
diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c
index 3f9f7985465..009055a6ab6 100644
--- a/src/mesa/main/bufferobj.c
+++ b/src/mesa/main/bufferobj.c
@@ -119,7 +119,7 @@ buffer_object_subdata_range_good( GLcontext * ctx, GLenum target,
       _mesa_error(ctx, GL_INVALID_OPERATION, "%s", caller);
       return NULL;
    }
-   if ((GLuint) (offset + size) > bufObj->Size) {
+   if (offset + size > bufObj->Size) {
       _mesa_error(ctx, GL_INVALID_VALUE,
 		  "%s(size + offset > buffer size)", caller);
       return NULL;
@@ -297,7 +297,7 @@ _mesa_buffer_subdata( GLcontext *ctx, GLenum target, GLintptrARB offset,
    (void) ctx; (void) target;
 
    /* this should have been caught in _mesa_BufferSubData() */
-   ASSERT((GLuint) (size + offset) <= bufObj->Size);
+   ASSERT(size + offset <= bufObj->Size);
 
    if (bufObj->Data) {
       _mesa_memcpy( (GLubyte *) bufObj->Data + offset, data, size );
diff --git a/src/mesa/main/colortab.c b/src/mesa/main/colortab.c
index e9349516eb8..9fb0baf4a7c 100644
--- a/src/mesa/main/colortab.c
+++ b/src/mesa/main/colortab.c
@@ -472,8 +472,8 @@ _mesa_ColorTable( GLenum target, GLenum internalFormat,
       _mesa_free_colortable_data(table);
 
       if (width > 0) {
-         table->TableF = _mesa_malloc(comps * width * sizeof(GLfloat));
-         table->TableUB = _mesa_malloc(comps * width * sizeof(GLubyte));
+         table->TableF = (GLfloat *) _mesa_malloc(comps * width * sizeof(GLfloat));
+         table->TableUB = (GLubyte *) _mesa_malloc(comps * width * sizeof(GLubyte));
 
 	 if (!table->TableF || !table->TableUB) {
 	    _mesa_error(ctx, GL_OUT_OF_MEMORY, "glColorTable");
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index eb5a0318dea..9b043488068 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -292,13 +292,8 @@ _mesa_forceCurrent(__GLcontext *gc)
 GLboolean
 _mesa_notifyResize(__GLcontext *gc)
 {
-   GLint x, y;
-   GLuint width, height;
-   __GLdrawablePrivate *d = gc->imports.getDrawablePrivate(gc);
-   if (!d || !d->getDrawableSize)
-      return GL_FALSE;
-   d->getDrawableSize( d, &x, &y, &width, &height );
    /* update viewport, resize software buffers, etc. */
+   (void) gc;
    return GL_TRUE;
 }
 
@@ -990,9 +985,8 @@ _mesa_init_current(GLcontext *ctx)
    ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_NORMAL], 0.0, 0.0, 1.0, 1.0 );
    ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_COLOR0], 1.0, 1.0, 1.0, 1.0 );
    ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_COLOR1], 0.0, 0.0, 0.0, 1.0 );
-   ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_FOG], 0.0, 0.0, 0.0, 0.0 );
-   ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX][0] = 1.0;
-   ctx->Current.EdgeFlag = GL_TRUE;
+   ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX], 1.0, 0.0, 0.0, 1.0 );
+   ASSIGN_4V( ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG], 1.0, 0.0, 0.0, 1.0 );
 }
 
 
@@ -1093,14 +1087,8 @@ _mesa_init_constants(GLcontext *ctx)
    ctx->Const.MaxProgramMatrices = MAX_PROGRAM_MATRICES;
    ctx->Const.MaxProgramMatrixStackDepth = MAX_PROGRAM_MATRIX_STACK_DEPTH;
 
-   /* If we're running in the X server, do bounds checking to prevent
-    * segfaults and server crashes!
-    */
-#if defined(XFree86Server)
-   ctx->Const.CheckArrayBounds = GL_TRUE;
-#else
+   /* CheckArrayBounds is overriden by drivers/x11 for X server */
    ctx->Const.CheckArrayBounds = GL_FALSE;
-#endif
 
    /* GL_ARB_draw_buffers */
    ctx->Const.MaxDrawBuffers = MAX_DRAW_BUFFERS;
diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c
index dca6ede6cb3..49bc2933bf5 100644
--- a/src/mesa/main/dlist.c
+++ b/src/mesa/main/dlist.c
@@ -343,8 +343,6 @@ typedef enum
    OPCODE_ATTR_3F_ARB,
    OPCODE_ATTR_4F_ARB,
    OPCODE_MATERIAL,
-   OPCODE_INDEX,
-   OPCODE_EDGEFLAG,
    OPCODE_BEGIN,
    OPCODE_END,
    OPCODE_RECTF,
@@ -4476,7 +4474,7 @@ save_ProgramLocalParameters4fvEXT(GLenum target, GLuint index, GLsizei count,
    ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx);
 
    if (count > 0) {
-      unsigned i;
+      GLint i;
       const GLfloat * p = params;
 
       for (i = 0 ; i < count ; i++) {
@@ -4710,7 +4708,7 @@ save_ProgramEnvParameters4fvEXT(GLenum target, GLuint index, GLsizei count,
    ASSERT_OUTSIDE_SAVE_BEGIN_END_AND_FLUSH(ctx);
 
    if (count > 0) {
-      unsigned i;
+      GLint i;
       const GLfloat * p = params;
 
       for (i = 0 ; i < count ; i++) {
@@ -5110,45 +5108,19 @@ save_EvalPoint2(GLint x, GLint y)
 static void GLAPIENTRY
 save_Indexf(GLfloat x)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   Node *n;
-   SAVE_FLUSH_VERTICES(ctx);
-   n = ALLOC_INSTRUCTION(ctx, OPCODE_INDEX, 1);
-   if (n) {
-      n[1].f = x;
-   }
-
-   ctx->ListState.ActiveIndex = 1;
-   ctx->ListState.CurrentIndex = x;
-
-   if (ctx->ExecuteFlag) {
-      CALL_Indexf(ctx->Exec, (x));
-   }
+   save_Attr1fNV(VERT_ATTRIB_COLOR_INDEX, x);
 }
 
 static void GLAPIENTRY
 save_Indexfv(const GLfloat * v)
 {
-   save_Indexf(v[0]);
+   save_Attr1fNV(VERT_ATTRIB_COLOR_INDEX, v[0]);
 }
 
 static void GLAPIENTRY
 save_EdgeFlag(GLboolean x)
 {
-   GET_CURRENT_CONTEXT(ctx);
-   Node *n;
-   SAVE_FLUSH_VERTICES(ctx);
-   n = ALLOC_INSTRUCTION(ctx, OPCODE_EDGEFLAG, 1);
-   if (n) {
-      n[1].b = x;
-   }
-
-   ctx->ListState.ActiveEdgeFlag = 1;
-   ctx->ListState.CurrentEdgeFlag = x;
-
-   if (ctx->ExecuteFlag) {
-      CALL_EdgeFlag(ctx->Exec, (x));
-   }
+   save_Attr1fNV(VERT_ATTRIB_EDGEFLAG, x ? 1.0 : 0.0);
 }
 
 static void GLAPIENTRY
@@ -6602,12 +6574,6 @@ execute_list(GLcontext *ctx, GLuint list)
                CALL_Materialfv(ctx->Exec, (n[1].e, n[2].e, f));
             }
             break;
-         case OPCODE_INDEX:
-            CALL_Indexf(ctx->Exec, (n[1].f));
-            break;
-         case OPCODE_EDGEFLAG:
-            CALL_EdgeFlag(ctx->Exec, (n[1].b));
-            break;
          case OPCODE_BEGIN:
             CALL_Begin(ctx->Exec, (n[1].e));
             break;
@@ -6793,9 +6759,6 @@ _mesa_NewList(GLuint list, GLenum mode)
    for (i = 0; i < MAT_ATTRIB_MAX; i++)
       ctx->ListState.ActiveMaterialSize[i] = 0;
 
-   ctx->ListState.ActiveIndex = 0;
-   ctx->ListState.ActiveEdgeFlag = 0;
-
    ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
    ctx->Driver.NewList(ctx, list, mode);
 
@@ -8421,12 +8384,6 @@ print_list(GLcontext *ctx, GLuint list)
             _mesa_printf("MATERIAL %x %x: %f %f %f %f\n",
                          n[1].i, n[2].i, n[3].f, n[4].f, n[5].f, n[6].f);
             break;
-         case OPCODE_INDEX:
-            _mesa_printf("INDEX: %f\n", n[1].f);
-            break;
-         case OPCODE_EDGEFLAG:
-            _mesa_printf("EDGEFLAG: %d\n", n[1].i);
-            break;
          case OPCODE_BEGIN:
             _mesa_printf("BEGIN %x\n", n[1].i);
             break;
diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c
index 91268b596d2..0d54c29949b 100644
--- a/src/mesa/main/enable.c
+++ b/src/mesa/main/enable.c
@@ -49,11 +49,14 @@
    }
 
 
+/**
+ * Helper to enable/disable client-side state.
+ */
 static void
-client_state( GLcontext *ctx, GLenum cap, GLboolean state )
+client_state(GLcontext *ctx, GLenum cap, GLboolean state)
 {
    GLuint flag;
-   GLuint *var;
+   GLboolean *var;
 
    switch (cap) {
       case GL_VERTEX_ARRAY:
@@ -134,17 +137,14 @@ client_state( GLcontext *ctx, GLenum cap, GLboolean state )
       ctx->Array.ArrayObj->_Enabled &= ~flag;
 
    if (ctx->Driver.Enable) {
-      (*ctx->Driver.Enable)( ctx, cap, state );
+      ctx->Driver.Enable( ctx, cap, state );
    }
 }
 
 
 /**
  * Enable GL capability.
- *
- * \param cap capability.
- *
- * \sa glEnable().
+ * \param cap  state to enable/disable.
  *
  * Get's the current context, assures that we're outside glBegin()/glEnd() and
  * calls client_state().
@@ -160,10 +160,7 @@ _mesa_EnableClientState( GLenum cap )
 
 /**
  * Disable GL capability.
- *
- * \param cap capability.
- *
- * \sa glDisable().
+ * \param cap  state to enable/disable.
  *
  * Get's the current context, assures that we're outside glBegin()/glEnd() and
  * calls client_state().
@@ -195,10 +192,10 @@ _mesa_DisableClientState( GLenum cap )
 
 
 /**
- * Perform glEnable() and glDisable() calls.
+ * Helper function to enable or disable state.
  *
  * \param ctx GL context.
- * \param cap capability.
+ * \param cap  the state to enable/disable
  * \param state whether to enable or disable the specified capability.
  *
  * Updates the current context and flushes the vertices as needed. For
@@ -206,7 +203,8 @@ _mesa_DisableClientState( GLenum cap )
  * are effectivly present before updating. Notifies the driver via
  * dd_function_table::Enable.
  */
-void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
+void
+_mesa_set_enable(GLcontext *ctx, GLenum cap, GLboolean state)
 {
    if (MESA_VERBOSE & VERBOSE_API)
       _mesa_debug(ctx, "%s %s (newstate is %x)\n",
@@ -285,7 +283,6 @@ void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
          FLUSH_VERTICES(ctx, _NEW_POLYGON);
          ctx->Polygon.CullFlag = state;
          break;
-
       case GL_CULL_VERTEX_EXT:
          CHECK_EXTENSION(EXT_cull_vertex, cap);
          if (ctx->Transform.CullVertexFlag == state)
@@ -293,13 +290,12 @@ void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
          FLUSH_VERTICES(ctx, _NEW_TRANSFORM);
          ctx->Transform.CullVertexFlag = state;
          break;
-
       case GL_DEPTH_TEST:
          if (state && ctx->DrawBuffer->Visual.depthBits == 0) {
             _mesa_warning(ctx,"glEnable(GL_DEPTH_TEST) but no depth buffer");
             return;
          }
-         if (ctx->Depth.Test==state)
+         if (ctx->Depth.Test == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_DEPTH);
          ctx->Depth.Test = state;
@@ -308,13 +304,13 @@ void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
          if (ctx->NoDither) {
             state = GL_FALSE; /* MESA_NO_DITHER env var */
          }
-         if (ctx->Color.DitherFlag==state)
+         if (ctx->Color.DitherFlag == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_COLOR);
          ctx->Color.DitherFlag = state;
          break;
       case GL_FOG:
-         if (ctx->Fog.Enabled==state)
+         if (ctx->Fog.Enabled == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_FOG);
          ctx->Fog.Enabled = state;
@@ -351,26 +347,18 @@ void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
             return;
          FLUSH_VERTICES(ctx, _NEW_LIGHT);
          ctx->Light.Enabled = state;
-
-         if (ctx->Light.Enabled && ctx->Light.Model.TwoSide)
-   	   ctx->_TriangleCaps |= DD_TRI_LIGHT_TWOSIDE;
-         else
- 	   ctx->_TriangleCaps &= ~DD_TRI_LIGHT_TWOSIDE;
- 
          break;
       case GL_LINE_SMOOTH:
          if (ctx->Line.SmoothFlag == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_LINE);
          ctx->Line.SmoothFlag = state;
-         ctx->_TriangleCaps ^= DD_LINE_SMOOTH;
          break;
       case GL_LINE_STIPPLE:
          if (ctx->Line.StippleFlag == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_LINE);
          ctx->Line.StippleFlag = state;
-         ctx->_TriangleCaps ^= DD_LINE_STIPPLE;
          break;
       case GL_INDEX_LOGIC_OP:
          if (ctx->Color.IndexLogicOpEnabled == state)
@@ -505,41 +493,38 @@ void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
          ctx->Transform.Normalize = state;
          break;
       case GL_POINT_SMOOTH:
-         if (ctx->Point.SmoothFlag==state)
+         if (ctx->Point.SmoothFlag == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_POINT);
          ctx->Point.SmoothFlag = state;
-         ctx->_TriangleCaps ^= DD_POINT_SMOOTH;
          break;
       case GL_POLYGON_SMOOTH:
-         if (ctx->Polygon.SmoothFlag==state)
+         if (ctx->Polygon.SmoothFlag == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_POLYGON);
          ctx->Polygon.SmoothFlag = state;
-         ctx->_TriangleCaps ^= DD_TRI_SMOOTH;
          break;
       case GL_POLYGON_STIPPLE:
-         if (ctx->Polygon.StippleFlag==state)
+         if (ctx->Polygon.StippleFlag == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_POLYGON);
          ctx->Polygon.StippleFlag = state;
-         ctx->_TriangleCaps ^= DD_TRI_STIPPLE;
          break;
       case GL_POLYGON_OFFSET_POINT:
-         if (ctx->Polygon.OffsetPoint==state)
+         if (ctx->Polygon.OffsetPoint == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_POLYGON);
          ctx->Polygon.OffsetPoint = state;
          break;
       case GL_POLYGON_OFFSET_LINE:
-         if (ctx->Polygon.OffsetLine==state)
+         if (ctx->Polygon.OffsetLine == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_POLYGON);
          ctx->Polygon.OffsetLine = state;
          break;
       case GL_POLYGON_OFFSET_FILL:
          /*case GL_POLYGON_OFFSET_EXT:*/
-         if (ctx->Polygon.OffsetFill==state)
+         if (ctx->Polygon.OffsetFill == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_POLYGON);
          ctx->Polygon.OffsetFill = state;
@@ -551,7 +536,7 @@ void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
          ctx->Transform.RescaleNormals = state;
          break;
       case GL_SCISSOR_TEST:
-         if (ctx->Scissor.Enabled==state)
+         if (ctx->Scissor.Enabled == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_SCISSOR);
          ctx->Scissor.Enabled = state;
@@ -568,7 +553,7 @@ void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
                           "glEnable(GL_STENCIL_TEST) but no stencil buffer");
             return;
          }
-         if (ctx->Stencil.Enabled==state)
+         if (ctx->Stencil.Enabled == state)
             return;
          FLUSH_VERTICES(ctx, _NEW_STENCIL);
          ctx->Stencil.Enabled = state;
@@ -916,11 +901,6 @@ void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
             return;
          FLUSH_VERTICES(ctx, _NEW_STENCIL);
          ctx->Stencil.TestTwoSide = state;
-         if (state) {
-            ctx->_TriangleCaps |= DD_TRI_TWOSTENCIL;
-         } else {
-            ctx->_TriangleCaps &= ~DD_TRI_TWOSTENCIL;
-         }
          break;
 
 #if FEATURE_ARB_fragment_program
@@ -973,20 +953,14 @@ void _mesa_set_enable( GLcontext *ctx, GLenum cap, GLboolean state )
    }
 
    if (ctx->Driver.Enable) {
-      (*ctx->Driver.Enable)( ctx, cap, state );
+      ctx->Driver.Enable( ctx, cap, state );
    }
 }
 
 
 /**
- * Enable GL capability.
- *
- * \param cap capability.
- *
- * \sa glEnable().
- *
- * Get's the current context, assures that we're outside glBegin()/glEnd() and
- * calls _mesa_set_enable().
+ * Enable GL capability.  Called by glEnable()
+ * \param cap  state to enable.
  */
 void GLAPIENTRY
 _mesa_Enable( GLenum cap )
@@ -999,14 +973,8 @@ _mesa_Enable( GLenum cap )
 
 
 /**
- * Disable GL capability.
- *
- * \param cap capability.
- *
- * \sa glDisable().
- *
- * Get's the current context, assures that we're outside glBegin()/glEnd() and
- * calls _mesa_set_enable().
+ * Disable GL capability.  Called by glDisable()
+ * \param cap  state to disable.
  */
 void GLAPIENTRY
 _mesa_Disable( GLenum cap )
@@ -1032,10 +1000,11 @@ _mesa_Disable( GLenum cap )
       return GL_FALSE;						\
    }
 
+
 /**
- * Test whether a capability is enabled.
+ * Return simple enable/disable state.
  *
- * \param cap capability.
+ * \param cap  state variable to query.
  *
  * Returns the state of the specified capability from the current GL context.
  * For the capabilities associated with extensions verifies that those
diff --git a/src/mesa/main/enums.c b/src/mesa/main/enums.c
index fc56809e978..8c1b785aab3 100644
--- a/src/mesa/main/enums.c
+++ b/src/mesa/main/enums.c
@@ -3517,7 +3517,6 @@ static const enum_elt all_enums[1737] =
 
 static const unsigned reduced_enums[1277] =
 {
-        30, /* GL_ALL_CLIENT_ATTRIB_BITS */
        435, /* GL_FALSE */
        643, /* GL_LINES */
        645, /* GL_LINE_LOOP */
@@ -4794,6 +4793,7 @@ static const unsigned reduced_enums[1277] =
       1314, /* GL_SCISSOR_BIT */
         29, /* GL_ALL_ATTRIB_BITS */
        938, /* GL_MULTISAMPLE_BIT */
+        30, /* GL_ALL_CLIENT_ATTRIB_BITS */
 };
 
 #define Elements(x) sizeof(x)/sizeof(*x)
diff --git a/src/mesa/main/extensions.c b/src/mesa/main/extensions.c
index 7845ea018ef..b61cb821f7d 100644
--- a/src/mesa/main/extensions.c
+++ b/src/mesa/main/extensions.c
@@ -136,6 +136,7 @@ static const struct {
    { OFF, "GL_ATI_texture_env_combine3",       F(ATI_texture_env_combine3)},
    { OFF, "GL_ATI_texture_mirror_once",        F(ATI_texture_mirror_once)},
    { OFF, "GL_ATI_fragment_shader",            F(ATI_fragment_shader)},
+   { OFF, "GL_ATI_separate_stencil",           F(ATI_separate_stencil)},
    { OFF, "GL_IBM_multimode_draw_arrays",      F(IBM_multimode_draw_arrays) },
    { ON,  "GL_IBM_rasterpos_clip",             F(IBM_rasterpos_clip) },
    { OFF, "GL_IBM_texture_mirrored_repeat",    F(ARB_texture_mirrored_repeat)},
@@ -223,6 +224,7 @@ _mesa_enable_sw_extensions(GLcontext *ctx)
 #endif
    ctx->Extensions.ATI_texture_env_combine3 = GL_TRUE;
    ctx->Extensions.ATI_texture_mirror_once = GL_TRUE;
+   ctx->Extensions.ATI_separate_stencil = GL_TRUE;
    ctx->Extensions.EXT_blend_color = GL_TRUE;
    ctx->Extensions.EXT_blend_equation_separate = GL_TRUE;
    ctx->Extensions.EXT_blend_func_separate = GL_TRUE;
diff --git a/src/mesa/main/get.c b/src/mesa/main/get.c
index 335443ef4f2..4cd07cb7b11 100644
--- a/src/mesa/main/get.c
+++ b/src/mesa/main/get.c
@@ -323,7 +323,7 @@ _mesa_GetBooleanv( GLenum pname, GLboolean *params )
       case GL_EDGE_FLAG:
          {
          FLUSH_CURRENT(ctx, 0);
-         params[0] = ctx->Current.EdgeFlag;
+         params[0] = (ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] == 1.0);
          }
          break;
       case GL_FEEDBACK_BUFFER_SIZE:
@@ -2150,7 +2150,7 @@ _mesa_GetFloatv( GLenum pname, GLfloat *params )
       case GL_EDGE_FLAG:
          {
          FLUSH_CURRENT(ctx, 0);
-         params[0] = BOOLEAN_TO_FLOAT(ctx->Current.EdgeFlag);
+         params[0] = BOOLEAN_TO_FLOAT((ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] == 1.0));
          }
          break;
       case GL_FEEDBACK_BUFFER_SIZE:
@@ -3977,7 +3977,7 @@ _mesa_GetIntegerv( GLenum pname, GLint *params )
       case GL_EDGE_FLAG:
          {
          FLUSH_CURRENT(ctx, 0);
-         params[0] = BOOLEAN_TO_INT(ctx->Current.EdgeFlag);
+         params[0] = BOOLEAN_TO_INT((ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] == 1.0));
          }
          break;
       case GL_FEEDBACK_BUFFER_SIZE:
diff --git a/src/mesa/main/get_gen.py b/src/mesa/main/get_gen.py
index 5a9ca0720e3..2f07cc51926 100644
--- a/src/mesa/main/get_gen.py
+++ b/src/mesa/main/get_gen.py
@@ -190,7 +190,7 @@ StateVars = [
 	( "GL_DOUBLEBUFFER", GLboolean,
 	  ["ctx->DrawBuffer->Visual.doubleBufferMode"], "", None ),
 	( "GL_DRAW_BUFFER", GLenum, ["ctx->DrawBuffer->ColorDrawBuffer[0]"], "", None ),
-	( "GL_EDGE_FLAG", GLboolean, ["ctx->Current.EdgeFlag"],
+	( "GL_EDGE_FLAG", GLboolean, ["(ctx->Current.Attrib[VERT_ATTRIB_EDGEFLAG][0] == 1.0)"],
 	  "FLUSH_CURRENT(ctx, 0);", None ),
 	( "GL_FEEDBACK_BUFFER_SIZE", GLint, ["ctx->Feedback.BufferSize"], "", None ),
 	( "GL_FEEDBACK_BUFFER_TYPE", GLenum, ["ctx->Feedback.Type"], "", None ),
diff --git a/src/mesa/main/glheader.h b/src/mesa/main/glheader.h
index e37499e4be7..5abea137d74 100644
--- a/src/mesa/main/glheader.h
+++ b/src/mesa/main/glheader.h
@@ -5,8 +5,8 @@
  * This is the top-most include file of the Mesa sources.
  * It includes gl.h and all system headers which are needed.
  * Other Mesa source files should \e not directly include any system
- * headers.  This allows Mesa to be integrated into XFree86 and
- * allows system-dependent hacks/workarounds to be collected in one place.
+ * headers.  This allows system-dependent hacks/workarounds to be
+ * collected in one place.
  *
  * \note Actually, a lot of system-dependent stuff is now in imports.[ch].
  *
@@ -46,18 +46,15 @@
 #ifndef GLHEADER_H
 #define GLHEADER_H
 
+/* This allows Mesa to be integrated into XFree86 */
 #ifdef HAVE_DIX_CONFIG_H
 #include "dix-config.h"
 #endif
 
-#if defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
-#include "xf86_ansic.h"
-#else
 #include <assert.h>
 #include <ctype.h>
-/* If we can use Compaq's Fast Math Library on Alpha */
 #if defined(__alpha__) && defined(CCPML)
-#include <cpml.h>
+#include <cpml.h> /* use Compaq's Fast Math Library on Alpha */
 #else
 #include <math.h>
 #endif
@@ -68,7 +65,6 @@
 #if defined(__linux__) && defined(__i386__)
 #include <fpu_control.h>
 #endif
-#endif
 #include <float.h>
 #include <stdarg.h>
 
diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c
index 6ff4089f0ea..fc8e1f0f574 100644
--- a/src/mesa/main/image.c
+++ b/src/mesa/main/image.c
@@ -1331,6 +1331,7 @@ _mesa_pack_rgba_span_float(GLcontext *ctx, GLuint n, GLfloat rgba[][4],
                      dst[i*4+2] = FLOAT_TO_BYTE(rgba[i][RCOMP]);
                      dst[i*4+3] = FLOAT_TO_BYTE(rgba[i][ACOMP]);
                   }
+		  break;
                case GL_ABGR_EXT:
                   for (i=0;i<n;i++) {
                      dst[i*4+0] = FLOAT_TO_BYTE(rgba[i][ACOMP]);
@@ -1479,6 +1480,7 @@ _mesa_pack_rgba_span_float(GLcontext *ctx, GLuint n, GLfloat rgba[][4],
                      dst[i*4+2] = FLOAT_TO_SHORT(rgba[i][RCOMP]);
                      dst[i*4+3] = FLOAT_TO_SHORT(rgba[i][ACOMP]);
                   }
+		  break;
                case GL_ABGR_EXT:
                   for (i=0;i<n;i++) {
                      dst[i*4+0] = FLOAT_TO_SHORT(rgba[i][ACOMP]);
@@ -1815,7 +1817,7 @@ _mesa_pack_rgba_span_float(GLcontext *ctx, GLuint n, GLfloat rgba[][4],
             for (i=0;i<n;i++) {
                dst[i] = (((GLint) (rgba[i][RCOMP] * 7.0F))     )
                       | (((GLint) (rgba[i][GCOMP] * 7.0F)) << 3)
-                      | (((GLint) (rgba[i][BCOMP] * 3.0F)) << 5);
+                      | (((GLint) (rgba[i][BCOMP] * 3.0F)) << 6);
             }
          }
          break;
@@ -1861,9 +1863,9 @@ _mesa_pack_rgba_span_float(GLcontext *ctx, GLuint n, GLfloat rgba[][4],
          else if (dstFormat == GL_ABGR_EXT) {
             GLushort *dst = (GLushort *) dstAddr;
             for (i=0;i<n;i++) {
-               dst[i] = (((GLint) (rgba[i][ACOMP] * 15.0F)) <<  4)
+               dst[i] = (((GLint) (rgba[i][ACOMP] * 15.0F)) << 12)
                       | (((GLint) (rgba[i][BCOMP] * 15.0F)) <<  8)
-                      | (((GLint) (rgba[i][GCOMP] * 15.0F)) << 12)
+                      | (((GLint) (rgba[i][GCOMP] * 15.0F)) <<  4)
                       | (((GLint) (rgba[i][RCOMP] * 15.0F))      );
             }
          }
diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c
index 9c7ebf9287d..e2d44fa07c4 100644
--- a/src/mesa/main/imports.c
+++ b/src/mesa/main/imports.c
@@ -24,9 +24,6 @@
  * - scanf
  * - qsort
  * - rand and RAND_MAX
- *
- * \note When compiled into a XFree86 module these functions wrap around
- * XFree86 own wrappers.
  */
 
 /*
@@ -71,50 +68,29 @@ extern int vsnprintf(char *str, size_t count, const char *fmt, va_list arg);
 #endif
 #endif
 
-/* If we don't actually want to use the libcwrapper junk (even though we're
- * building an Xorg server module), then just undef IN_MODULE to signal that to
- * the following code.  It's left around for now to allow compiling of newish
- * Mesa with older servers, but this whole mess should go away at some point.
- */
-#ifdef NO_LIBCWRAPPER
-#undef IN_MODULE
-#endif
-
 /**********************************************************************/
 /** \name Memory */
 /*@{*/
 
-/** Wrapper around either malloc() or xf86malloc() */
+/** Wrapper around malloc() */
 void *
 _mesa_malloc(size_t bytes)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86malloc(bytes);
-#else
    return malloc(bytes);
-#endif
 }
 
-/** Wrapper around either calloc() or xf86calloc() */
+/** Wrapper around calloc() */
 void *
 _mesa_calloc(size_t bytes)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86calloc(1, bytes);
-#else
    return calloc(1, bytes);
-#endif
 }
 
-/** Wrapper around either free() or xf86free() */
+/** Wrapper around free() */
 void
 _mesa_free(void *ptr)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   xf86free(ptr);
-#else
    free(ptr);
-#endif
 }
 
 /**
@@ -131,7 +107,7 @@ _mesa_free(void *ptr)
 void *
 _mesa_align_malloc(size_t bytes, unsigned long alignment)
 {
-#if defined(HAVE_POSIX_MEMALIGN) && !(defined(XFree86LOADER) && defined(IN_MODULE))
+#if defined(HAVE_POSIX_MEMALIGN)
    void *mem;
 
    (void) posix_memalign(& mem, alignment, bytes);
@@ -157,7 +133,7 @@ _mesa_align_malloc(size_t bytes, unsigned long alignment)
 #endif
 
    return (void *) buf;
-#endif /* defined(HAVE_POSIX_MEMALIGN) && !(defined(XFree86LOADER) && defined(IN_MODULE)) */
+#endif /* defined(HAVE_POSIX_MEMALIGN) */
 }
 
 /**
@@ -167,7 +143,7 @@ _mesa_align_malloc(size_t bytes, unsigned long alignment)
 void *
 _mesa_align_calloc(size_t bytes, unsigned long alignment)
 {
-#if defined(HAVE_POSIX_MEMALIGN) && !(defined(XFree86LOADER) && defined(IN_MODULE))
+#if defined(HAVE_POSIX_MEMALIGN)
    void *mem;
    
    mem = _mesa_align_malloc(bytes, alignment);
@@ -197,7 +173,7 @@ _mesa_align_calloc(size_t bytes, unsigned long alignment)
 #endif
 
    return (void *)buf;
-#endif /* defined(HAVE_POSIX_MEMALIGN) && !(defined(XFree86LOADER) && defined(IN_MODULE)) */
+#endif /* defined(HAVE_POSIX_MEMALIGN) */
 }
 
 /**
@@ -210,13 +186,13 @@ _mesa_align_calloc(size_t bytes, unsigned long alignment)
 void
 _mesa_align_free(void *ptr)
 {
-#if defined(HAVE_POSIX_MEMALIGN) && !(defined(XFree86LOADER) && defined(IN_MODULE))
+#if defined(HAVE_POSIX_MEMALIGN)
    free(ptr);
 #else
    void **cubbyHole = (void **) ((char *) ptr - sizeof(void *));
    void *realAddr = *cubbyHole;
    _mesa_free(realAddr);
-#endif /* defined(HAVE_POSIX_MEMALIGN) && !(defined(XFree86LOADER) && defined(IN_MODULE)) */
+#endif /* defined(HAVE_POSIX_MEMALIGN) */
 }
 
 /**
@@ -255,22 +231,18 @@ _mesa_realloc(void *oldBuffer, size_t oldSize, size_t newSize)
 void *
 _mesa_memcpy(void *dest, const void *src, size_t n)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86memcpy(dest, src, n);
-#elif defined(SUNOS4)
+#if defined(SUNOS4)
    return memcpy((char *) dest, (char *) src, (int) n);
 #else
    return memcpy(dest, src, n);
 #endif
 }
 
-/** Wrapper around either memset() or xf86memset() */
+/** Wrapper around memset() */
 void
 _mesa_memset( void *dst, int val, size_t n )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   xf86memset( dst, val, n );
-#elif defined(SUNOS4)
+#if defined(SUNOS4)
    memset( (char *) dst, (int) val, (int) n );
 #else
    memset(dst, val, n);
@@ -290,26 +262,22 @@ _mesa_memset16( unsigned short *dst, unsigned short val, size_t n )
       *dst++ = val;
 }
 
-/** Wrapper around either memcpy() or xf86memcpy() or bzero() */
+/** Wrapper around either memcpy() or bzero() */
 void
 _mesa_bzero( void *dst, size_t n )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   xf86memset( dst, 0, n );
-#elif defined(__FreeBSD__)
+#if defined(__FreeBSD__)
    bzero( dst, n );
 #else
    memset( dst, 0, n );
 #endif
 }
 
-/** Wrapper around either memcmp() or xf86memcmp() */
+/** Wrapper around memcmp() */
 int
 _mesa_memcmp( const void *s1, const void *s2, size_t n )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86memcmp( s1, s2, n );
-#elif defined(SUNOS4)
+#if defined(SUNOS4)
    return memcmp( (char *) s1, (char *) s2, (int) n );
 #else
    return memcmp(s1, s2, n);
@@ -323,70 +291,46 @@ _mesa_memcmp( const void *s1, const void *s2, size_t n )
 /** \name Math */
 /*@{*/
 
-/** Wrapper around either sin() or xf86sin() */
+/** Wrapper around sin() */
 double
 _mesa_sin(double a)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86sin(a);
-#else
    return sin(a);
-#endif
 }
 
-/** Single precision wrapper around either sin() or xf86sin() */
+/** Single precision wrapper around sin() */
 float
 _mesa_sinf(float a)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return (float) xf86sin((double) a);
-#else
    return (float) sin((double) a);
-#endif
 }
 
-/** Wrapper around either cos() or xf86cos() */
+/** Wrapper around cos() */
 double
 _mesa_cos(double a)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86cos(a);
-#else
    return cos(a);
-#endif
 }
 
-/** Single precision wrapper around either asin() or xf86asin() */
+/** Single precision wrapper around asin() */
 float
 _mesa_asinf(float x)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return (float) xf86asin((double) x);
-#else
    return (float) asin((double) x);
-#endif
 }
 
-/** Single precision wrapper around either atan() or xf86atan() */
+/** Single precision wrapper around atan() */
 float
 _mesa_atanf(float x)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return (float) xf86atan((double) x);
-#else
    return (float) atan((double) x);
-#endif
 }
 
-/** Wrapper around either sqrt() or xf86sqrt() */
+/** Wrapper around sqrt() */
 double
 _mesa_sqrtd(double x)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86sqrt(x);
-#else
    return sqrt(x);
-#endif
 }
 
 
@@ -584,25 +528,17 @@ _mesa_inv_sqrtf(float n)
 
         return x3 * r3;
 #endif
-#elif defined(XFree86LOADER) && defined(IN_MODULE)
-        return 1.0F / xf86sqrt(n);
 #else
         return (float) (1.0 / sqrt(n));
 #endif
 }
 
 
-/**
- * Wrapper around either pow() or xf86pow().
- */
+/** Wrapper around pow() */
 double
 _mesa_pow(double x, double y)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86pow(x, y);
-#else
    return pow(x, y);
-#endif
 }
 
 
@@ -633,8 +569,6 @@ _mesa_ffs(int i)
       }
    }
    return bit;
-#elif defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86ffs(i);
 #else
    return ffs(i);
 #endif
@@ -642,6 +576,35 @@ _mesa_ffs(int i)
 
 
 /**
+ * Find position of first bit set in given value.
+ * XXX Warning: this function can only be used on 64-bit systems!
+ * \return  position of least-significant bit set, starting at 1, return zero
+ *          if no bits set.
+ */
+int
+_mesa_ffsll(long long val)
+{
+#ifdef ffsll
+   return ffsll(val);
+#else
+   int bit;
+
+   assert(sizeof(val) == 8);
+
+   bit = ffs(val);
+   if (bit != 0)
+      return bit;
+
+   bit = ffs(val >> 32);
+   if (bit != 0)
+      return 32 + bit;
+
+   return 0;
+#endif
+}
+
+
+/**
  * Return number of bits set in given GLuint.
  */
 unsigned int
@@ -807,11 +770,7 @@ void *
 _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size, 
                int (*compar)(const void *, const void *) )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86bsearch(key, base, nmemb, size, compar);
-#else
    return bsearch(key, base, nmemb, size, compar);
-#endif
 }
 
 /*@}*/
@@ -827,9 +786,7 @@ _mesa_bsearch( const void *key, const void *base, size_t nmemb, size_t size,
 char *
 _mesa_getenv( const char *var )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86getenv(var);
-#elif defined(_XBOX)
+#if defined(_XBOX)
    return NULL;
 #else
    return getenv(var);
@@ -843,81 +800,53 @@ _mesa_getenv( const char *var )
 /** \name String */
 /*@{*/
 
-/** Wrapper around either strstr() or xf86strstr() */
+/** Wrapper around strstr() */
 char *
 _mesa_strstr( const char *haystack, const char *needle )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86strstr(haystack, needle);
-#else
    return strstr(haystack, needle);
-#endif
 }
 
-/** Wrapper around either strncat() or xf86strncat() */
+/** Wrapper around strncat() */
 char *
 _mesa_strncat( char *dest, const char *src, size_t n )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86strncat(dest, src, n);
-#else
    return strncat(dest, src, n);
-#endif
 }
 
-/** Wrapper around either strcpy() or xf86strcpy() */
+/** Wrapper around strcpy() */
 char *
 _mesa_strcpy( char *dest, const char *src )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86strcpy(dest, src);
-#else
    return strcpy(dest, src);
-#endif
 }
 
-/** Wrapper around either strncpy() or xf86strncpy() */
+/** Wrapper around strncpy() */
 char *
 _mesa_strncpy( char *dest, const char *src, size_t n )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86strncpy(dest, src, n);
-#else
    return strncpy(dest, src, n);
-#endif
 }
 
-/** Wrapper around either strlen() or xf86strlen() */
+/** Wrapper around strlen() */
 size_t
 _mesa_strlen( const char *s )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86strlen(s);
-#else
    return strlen(s);
-#endif
 }
 
-/** Wrapper around either strcmp() or xf86strcmp() */
+/** Wrapper around strcmp() */
 int
 _mesa_strcmp( const char *s1, const char *s2 )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86strcmp(s1, s2);
-#else
    return strcmp(s1, s2);
-#endif
 }
 
-/** Wrapper around either strncmp() or xf86strncmp() */
+/** Wrapper around strncmp() */
 int
 _mesa_strncmp( const char *s1, const char *s2, size_t n )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86strncmp(s1, s2, n);
-#else
    return strncmp(s1, s2, n);
-#endif
 }
 
 /**
@@ -939,26 +868,18 @@ _mesa_strdup( const char *s )
    }
 }
 
-/** Wrapper around either atoi() or xf86atoi() */
+/** Wrapper around atoi() */
 int
 _mesa_atoi(const char *s)
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86atoi(s);
-#else
    return atoi(s);
-#endif
 }
 
-/** Wrapper around either strtod() or xf86strtod() */
+/** Wrapper around strtod() */
 double
 _mesa_strtod( const char *s, char **end )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86strtod(s, end);
-#else
    return strtod(s, end);
-#endif
 }
 
 /*@}*/
@@ -968,24 +889,19 @@ _mesa_strtod( const char *s, char **end )
 /** \name I/O */
 /*@{*/
 
-/** Wrapper around either vsprintf() or xf86vsprintf() */
+/** Wrapper around vsprintf() */
 int
 _mesa_sprintf( char *str, const char *fmt, ... )
 {
    int r;
    va_list args;
    va_start( args, fmt );  
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   r = xf86vsprintf( str, fmt, args );
-#else
    r = vsprintf( str, fmt, args );
-#endif
    va_end( args );
    return r;
 }
 
-/** Wrapper around either printf() or xf86printf(), using vsprintf() for
- * the formatting. */
+/** Wrapper around printf(), using vsprintf() for the formatting. */
 void
 _mesa_printf( const char *fmtString, ... )
 {
@@ -994,22 +910,14 @@ _mesa_printf( const char *fmtString, ... )
    va_start( args, fmtString );  
    vsnprintf(s, MAXSTRING, fmtString, args);
    va_end( args );
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   xf86printf("%s", s);
-#else
    fprintf(stderr,"%s", s);
-#endif
 }
 
-/** Wrapper around either vsprintf() or xf86vsprintf() */
+/** Wrapper around vsprintf() */
 int
 _mesa_vsprintf( char *str, const char *fmt, va_list args )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86vsprintf( str, fmt, args );
-#else
    return vsprintf( str, fmt, args );
-#endif
 }
 
 /*@}*/
@@ -1027,7 +935,7 @@ _mesa_vsprintf( char *str, const char *fmt, va_list args )
  * 
  * If debugging is enabled (either at compile-time via the DEBUG macro, or
  * run-time via the MESA_DEBUG environment variable), prints the warning to
- * stderr, either via fprintf() or xf86printf().
+ * stderr via fprintf().
  */
 void
 _mesa_warning( GLcontext *ctx, const char *fmtString, ... )
@@ -1045,11 +953,7 @@ _mesa_warning( GLcontext *ctx, const char *fmtString, ... )
    debug = _mesa_getenv("MESA_DEBUG") ? GL_TRUE : GL_FALSE;
 #endif
    if (debug) {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-      xf86fprintf(stderr, "Mesa warning: %s\n", str);
-#else
       fprintf(stderr, "Mesa warning: %s\n", str);
-#endif
    }
 }
 
@@ -1060,7 +964,7 @@ _mesa_warning( GLcontext *ctx, const char *fmtString, ... )
  * \param ctx GL context.
  * \param s problem description string.
  *
- * Prints the message to stderr, either via fprintf() or xf86fprintf().
+ * Prints the message to stderr via fprintf().
  */
 void
 _mesa_problem( const GLcontext *ctx, const char *fmtString, ... )
@@ -1073,13 +977,8 @@ _mesa_problem( const GLcontext *ctx, const char *fmtString, ... )
    vsnprintf( str, MAXSTRING, fmtString, args );
    va_end( args );
 
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   xf86fprintf(stderr, "Mesa %s implementation error: %s\n", MESA_VERSION_STRING, str);
-   xf86fprintf(stderr, "Please report at bugzilla.freedesktop.org\n");
-#else
    fprintf(stderr, "Mesa %s implementation error: %s\n", MESA_VERSION_STRING, str);
    fprintf(stderr, "Please report at bugzilla.freedesktop.org\n");
-#endif
 }
 
 /**
@@ -1169,7 +1068,7 @@ _mesa_error( GLcontext *ctx, GLenum error, const char *fmtString, ... )
  * \param ctx GL context.
  * \param fmtString printf() alike format string.
  * 
- * Prints the message to stderr, either via fprintf() or xf86printf().
+ * Prints the message to stderr via fprintf().
  */
 void
 _mesa_debug( const GLcontext *ctx, const char *fmtString, ... )
@@ -1180,11 +1079,7 @@ _mesa_debug( const GLcontext *ctx, const char *fmtString, ... )
    va_start(args, fmtString);
    vsnprintf(s, MAXSTRING, fmtString, args);
    va_end(args);
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   xf86fprintf(stderr, "Mesa: %s", s);
-#else
    fprintf(stderr, "Mesa: %s", s);
-#endif
 #endif /* DEBUG */
    (void) ctx;
    (void) fmtString;
@@ -1203,11 +1098,7 @@ _mesa_debug( const GLcontext *ctx, const char *fmtString, ... )
 void
 _mesa_exit( int status )
 {
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   xf86exit(status);
-#else
    exit(status);
-#endif
 }
 
 /*@}*/
@@ -1233,16 +1124,12 @@ default_calloc(__GLcontext *gc, size_t numElem, size_t elemSize)
    return _mesa_calloc(numElem * elemSize);
 }
 
-/** Wrapper around either realloc() or xf86realloc() */
+/** Wrapper around realloc() */
 static void *
 default_realloc(__GLcontext *gc, void *oldAddr, size_t newSize)
 {
    (void) gc;
-#if defined(XFree86LOADER) && defined(IN_MODULE)
-   return xf86realloc(oldAddr, newSize);
-#else
    return realloc(oldAddr, newSize);
-#endif
 }
 
 /** Wrapper around _mesa_free() */
@@ -1326,16 +1213,6 @@ default_fprintf(__GLcontext *gc, void *stream, const char *fmt, ...)
    return r;
 }
 
-/**
- * \todo this really is driver-specific and can't be here 
- */
-static __GLdrawablePrivate *
-default_GetDrawablePrivate(__GLcontext *gc)
-{
-   (void) gc;
-   return NULL;
-}
-
 /*@}*/
 
 
@@ -1372,6 +1249,7 @@ _mesa_init_default_imports(__GLimports *imports, void *driverCtx)
    imports->fopen = default_fopen;
    imports->fclose = default_fclose;
    imports->fprintf = default_fprintf;
-   imports->getDrawablePrivate = default_GetDrawablePrivate;
+   imports->getDrawablePrivate = NULL; /* driver-specific */
+   imports->getReadablePrivate = NULL; /* driver-specific */
    imports->other = driverCtx;
 }
diff --git a/src/mesa/main/imports.h b/src/mesa/main/imports.h
index dad2767e72d..0633b3b8bf7 100644
--- a/src/mesa/main/imports.h
+++ b/src/mesa/main/imports.h
@@ -138,13 +138,16 @@ typedef union { GLfloat f; GLint i; } fi_type;
 #define M_E (2.7182818284590452354)
 #endif
 
-#ifndef FLT_MAX_EXP
-#define FLT_MAX_EXP 128
+#ifndef ONE_DIV_LN2
+#define ONE_DIV_LN2 (1.442695040888963456)
+#endif
+
+#ifndef ONE_DIV_SQRT_LN2
+#define ONE_DIV_SQRT_LN2 (1.201122408786449815)
 #endif
 
-/* XXX this is a bit of a hack needed for compilation within XFree86 */
-#ifndef FLT_MIN
-#define FLT_MIN (1.0e-37)
+#ifndef FLT_MAX_EXP
+#define FLT_MAX_EXP 128
 #endif
 
 /* Degrees to radians conversion: */
@@ -173,8 +176,6 @@ typedef union { GLfloat f; GLint i; } fi_type;
  ***/
 #if 0 /* _mesa_sqrtf() not accurate enough - temporarily disabled */
 #  define SQRTF(X)  _mesa_sqrtf(X)
-#elif defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
-#  define SQRTF(X)  (float) xf86sqrt((float) (X))
 #else
 #  define SQRTF(X)  (float) sqrt((float) (X))
 #endif
@@ -221,8 +222,6 @@ static INLINE GLfloat LOG2(GLfloat val)
    num.f = ((-1.0f/3) * num.f + 2) * num.f - 2.0f/3;
    return num.f + log_2;
 }
-#elif defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
-#define LOG2(x) ((GLfloat) (xf86log(x) * 1.442695))
 #else
 /*
  * NOTE: log_base_2(x) = log(x) / log(2)
@@ -293,15 +292,7 @@ static INLINE int GET_FLOAT_BITS( float x )
  *** LDEXPF: multiply value by an integral power of two
  *** FREXPF: extract mantissa and exponent from value
  ***/
-#if defined(XFree86LOADER) && defined(IN_MODULE) && !defined(NO_LIBCWRAPPER)
-#define CEILF(x)   ((GLfloat) xf86ceil(x))
-#define FLOORF(x)  ((GLfloat) xf86floor(x))
-#define FABSF(x)   ((GLfloat) xf86fabs(x))
-#define LOGF(x)    ((GLfloat) xf86log(x))
-#define EXPF(x)    ((GLfloat) xf86exp(x))
-#define LDEXPF(x,y)   ((GLfloat) xf86ldexp(x,y))
-#define FREXPF(x,y)   ((GLfloat) xf86frexp(x,y))
-#elif defined(__gnu_linux__)
+#if defined(__gnu_linux__)
 /* C99 functions */
 #define CEILF(x)   ceilf(x)
 #define FLOORF(x)  floorf(x)
@@ -705,6 +696,9 @@ _mesa_pow(double x, double y);
 extern int
 _mesa_ffs(int i);
 
+extern int
+_mesa_ffsll(long long i);
+
 extern unsigned int
 _mesa_bitcount(unsigned int n);
 
diff --git a/src/mesa/main/light.c b/src/mesa/main/light.c
index 984f7b2abc4..92d8a0ae0d4 100644
--- a/src/mesa/main/light.c
+++ b/src/mesa/main/light.c
@@ -1,6 +1,6 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  6.5.3
  *
  * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
  *
@@ -44,7 +44,7 @@ _mesa_ShadeModel( GLenum mode )
       _mesa_debug(ctx, "glShadeModel %s\n", _mesa_lookup_enum_by_nr(mode));
 
    if (mode != GL_FLAT && mode != GL_SMOOTH) {
-      _mesa_error( ctx, GL_INVALID_ENUM, "glShadeModel" );
+      _mesa_error(ctx, GL_INVALID_ENUM, "glShadeModel");
       return;
    }
 
@@ -53,9 +53,8 @@ _mesa_ShadeModel( GLenum mode )
 
    FLUSH_VERTICES(ctx, _NEW_LIGHT);
    ctx->Light.ShadeModel = mode;
-   ctx->_TriangleCaps ^= DD_FLATSHADE;
    if (ctx->Driver.ShadeModel)
-      (*ctx->Driver.ShadeModel)( ctx, mode );
+      ctx->Driver.ShadeModel( ctx, mode );
 }
 
 
@@ -442,11 +441,6 @@ _mesa_LightModelfv( GLenum pname, const GLfloat *params )
 	    return;
 	 FLUSH_VERTICES(ctx, _NEW_LIGHT);
 	 ctx->Light.Model.TwoSide = newbool;
-
-	 if (ctx->Light.Enabled && ctx->Light.Model.TwoSide)
-	    ctx->_TriangleCaps |= DD_TRI_LIGHT_TWOSIDE;
-	 else
-	    ctx->_TriangleCaps &= ~DD_TRI_LIGHT_TWOSIDE;
          break;
       case GL_LIGHT_MODEL_COLOR_CONTROL:
          if (params[0] == (GLfloat) GL_SINGLE_COLOR)
@@ -728,7 +722,7 @@ _mesa_ColorMaterial( GLenum face, GLenum mode )
    }
 
    if (ctx->Driver.ColorMaterial)
-      (*ctx->Driver.ColorMaterial)( ctx, face, mode );
+      ctx->Driver.ColorMaterial( ctx, face, mode );
 }
 
 
diff --git a/src/mesa/main/light.h b/src/mesa/main/light.h
index f47fe58a839..b97e17b5be4 100644
--- a/src/mesa/main/light.h
+++ b/src/mesa/main/light.h
@@ -92,7 +92,8 @@ do {									\
    struct gl_shine_tab *_tab = table;					\
    float f = (dp * (SHINE_TABLE_SIZE-1));				\
    int k = (int) f;							\
-   if (k > SHINE_TABLE_SIZE-2) 						\
+   if (k < 0 /* gcc may cast an overflow float value to negative int value*/ \
+	|| k > SHINE_TABLE_SIZE-2)					\
       result = (GLfloat) _mesa_pow( dp, _tab->shininess );		\
    else									\
       result = _tab->tab[k] + (f-k)*(_tab->tab[k+1]-_tab->tab[k]);	\
diff --git a/src/mesa/main/lines.c b/src/mesa/main/lines.c
index c30d9ac109e..dc7195d4ebf 100644
--- a/src/mesa/main/lines.c
+++ b/src/mesa/main/lines.c
@@ -1,13 +1,8 @@
-/**
- * \file lines.c
- * Line operations.
- */
-
 /*
  * Mesa 3-D graphics library
- * Version:  5.1
+ * Version:  6.5.3
  *
- * Copyright (C) 1999-2003  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2006  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"),
@@ -43,12 +38,6 @@
  * \param width line width in pixels.
  *
  * \sa glLineWidth().
- *
- * Verifies the parameter and updates gl_line_attrib::Width. On a change,
- * flushes the vertices, updates the clamped line width and marks the
- * DD_LINE_WIDTH flag in __GLcontextRec::_TriangleCaps for the drivers if the
- * width is different from one. Notifies the driver via the
- * dd_function_table::LineWidth callback.
  */
 void GLAPIENTRY
 _mesa_LineWidth( GLfloat width )
@@ -70,14 +59,8 @@ _mesa_LineWidth( GLfloat width )
 			    ctx->Const.MinLineWidth,
 			    ctx->Const.MaxLineWidth);
 
-
-   if (width != 1.0)
-      ctx->_TriangleCaps |= DD_LINE_WIDTH;
-   else
-      ctx->_TriangleCaps &= ~DD_LINE_WIDTH;
-
    if (ctx->Driver.LineWidth)
-      (*ctx->Driver.LineWidth)(ctx, width);
+      ctx->Driver.LineWidth(ctx, width);
 }
 
 
diff --git a/src/mesa/main/matrix.c b/src/mesa/main/matrix.c
index 7339b0ce409..b2aa83e1890 100644
--- a/src/mesa/main/matrix.c
+++ b/src/mesa/main/matrix.c
@@ -427,7 +427,7 @@ _mesa_Scalef( GLfloat x, GLfloat y, GLfloat z )
 
 
 /**
- * Multiply the current matrix with a general scaling matrix.
+ * Multiply the current matrix with a translation matrix.
  *
  * \param x translation vector x coordinate.
  * \param y translation vector y coordinate.
diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c
index f154bd46513..cc1fb97eedd 100644
--- a/src/mesa/main/mipmap.c
+++ b/src/mesa/main/mipmap.c
@@ -928,6 +928,9 @@ _mesa_generate_mipmap(GLcontext *ctx, GLenum target,
          return;
       }
 
+      if (dstImage->ImageOffsets)
+         _mesa_free(dstImage->ImageOffsets);
+
       /* Free old image data */
       if (dstImage->Data)
          ctx->Driver.FreeTexImageData(ctx, dstImage);
diff --git a/src/mesa/main/mtypes.h b/src/mesa/main/mtypes.h
index 76b36d60f1b..ce58154a0af 100644
--- a/src/mesa/main/mtypes.h
+++ b/src/mesa/main/mtypes.h
@@ -149,7 +149,7 @@ enum
    VERT_ATTRIB_COLOR1 = 4,
    VERT_ATTRIB_FOG = 5,
    VERT_ATTRIB_COLOR_INDEX = 6,
-   VERT_ATTRIB_SEVEN = 7,
+   VERT_ATTRIB_EDGEFLAG = 7,
    VERT_ATTRIB_TEX0 = 8,
    VERT_ATTRIB_TEX1 = 9,
    VERT_ATTRIB_TEX2 = 10,
@@ -189,7 +189,7 @@ enum
 #define VERT_BIT_COLOR1      (1 << VERT_ATTRIB_COLOR1)
 #define VERT_BIT_FOG         (1 << VERT_ATTRIB_FOG)
 #define VERT_BIT_COLOR_INDEX (1 << VERT_ATTRIB_COLOR_INDEX)
-#define VERT_BIT_SEVEN       (1 << VERT_ATTRIB_SEVEN)
+#define VERT_BIT_EDGEFLAG    (1 << VERT_ATTRIB_EDGEFLAG)
 #define VERT_BIT_TEX0        (1 << VERT_ATTRIB_TEX0)
 #define VERT_BIT_TEX1        (1 << VERT_ATTRIB_TEX1)
 #define VERT_BIT_TEX2        (1 << VERT_ATTRIB_TEX2)
@@ -613,11 +613,11 @@ struct gl_current_attrib
    /**
     * \name Current vertex attributes.
     * \note Values are valid only after FLUSH_VERTICES has been called.
+    * \note Index and Edgeflag current values are stored as floats in the 
+    * SIX and SEVEN attribute slots.
     */
    /*@{*/
    GLfloat Attrib[VERT_ATTRIB_MAX][4];	/**< Position, color, texcoords, etc */
-   GLfloat Index;			/**< Current color index */
-   GLboolean EdgeFlag;			/**< Current edge flag */
    /*@}*/
 
    /**
@@ -1637,8 +1637,6 @@ struct gl_pixelstore_attrib
 };
 
 
-#define CA_CLIENT_DATA     0x1	/**< Data not allocated by mesa */
-
 
 /**
  * Client vertex array attributes
@@ -1650,14 +1648,12 @@ struct gl_client_array
    GLsizei Stride;		/**< user-specified stride */
    GLsizei StrideB;		/**< actual stride in bytes */
    const GLubyte *Ptr;          /**< Points to array data */
-   GLbitfield Enabled;		/**< one of the _NEW_ARRAY_ bits */
+   GLboolean Enabled;		/**< Enabled flag is a boolean */
    GLboolean Normalized;        /**< GL_ARB_vertex_program */
 
    /**< GL_ARB_vertex_buffer_object */
    struct gl_buffer_object *BufferObj;
    GLuint _MaxElement;
-
-   GLbitfield Flags;
 };
 
 
@@ -1678,8 +1674,8 @@ struct gl_array_object
    struct gl_client_array SecondaryColor;
    struct gl_client_array FogCoord;
    struct gl_client_array Index;
-   struct gl_client_array TexCoord[MAX_TEXTURE_COORD_UNITS];
    struct gl_client_array EdgeFlag;
+   struct gl_client_array TexCoord[MAX_TEXTURE_COORD_UNITS];
    /*@}*/
 
    /** Generic arrays for vertex programs/shaders */
@@ -2547,6 +2543,7 @@ struct gl_extensions
    GLboolean ATI_texture_mirror_once;
    GLboolean ATI_texture_env_combine3;
    GLboolean ATI_fragment_shader;
+   GLboolean ATI_separate_stencil;
    GLboolean IBM_rasterpos_clip;
    GLboolean IBM_multimode_draw_arrays;
    GLboolean MESA_pack_invert;
@@ -2680,7 +2677,7 @@ struct matrix_stack
 #define _NEW_ARRAY_COLOR1           VERT_BIT_COLOR1
 #define _NEW_ARRAY_FOGCOORD         VERT_BIT_FOG
 #define _NEW_ARRAY_INDEX            VERT_BIT_COLOR_INDEX
-#define _NEW_ARRAY_EDGEFLAG         VERT_BIT_SEVEN
+#define _NEW_ARRAY_EDGEFLAG         VERT_BIT_EDGEFLAG
 #define _NEW_ARRAY_TEXCOORD_0       VERT_BIT_TEX0
 #define _NEW_ARRAY_TEXCOORD_1       VERT_BIT_TEX1
 #define _NEW_ARRAY_TEXCOORD_2       VERT_BIT_TEX2
@@ -2689,7 +2686,7 @@ struct matrix_stack
 #define _NEW_ARRAY_TEXCOORD_5       VERT_BIT_TEX5
 #define _NEW_ARRAY_TEXCOORD_6       VERT_BIT_TEX6
 #define _NEW_ARRAY_TEXCOORD_7       VERT_BIT_TEX7
-#define _NEW_ARRAY_ATTRIB_0         0x10000  /* start at bit 16 */
+#define _NEW_ARRAY_ATTRIB_0         VERT_BIT_GENERIC0  /* start at bit 16 */
 #define _NEW_ARRAY_ALL              0xffffffff
 
 
diff --git a/src/mesa/main/points.c b/src/mesa/main/points.c
index aa36fb62877..9caa9ab3ab2 100644
--- a/src/mesa/main/points.c
+++ b/src/mesa/main/points.c
@@ -5,9 +5,9 @@
 
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  6.5.1
  *
- * Copyright (C) 1999-2005  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2006  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"),
@@ -57,6 +57,13 @@ _mesa_PointSize( GLfloat size )
 
    FLUSH_VERTICES(ctx, _NEW_POINT);
    ctx->Point.Size = size;
+   ctx->Point._Size = CLAMP(ctx->Point.Size,
+			    ctx->Point.MinSize,
+			    ctx->Point.MaxSize);
+
+   ctx->Point._Attenuated = (ctx->Point.Params[0] != 1.0 ||
+                             ctx->Point.Params[1] != 0.0 ||
+                             ctx->Point.Params[2] != 0.0);
 
    if (ctx->Driver.PointSize)
       ctx->Driver.PointSize(ctx, size);
@@ -232,36 +239,6 @@ _mesa_PointParameterfvEXT( GLenum pname, const GLfloat *params)
 
 
 /**
- * Update derived point-related state.
- */
-void
-_mesa_update_point(GLcontext *ctx)
-{
-   /* clamp to user-specified limits now, clamp to ctx->Const.Min/Max
-    * limits during rasterization.
-    */
-   ctx->Point._Size = CLAMP(ctx->Point.Size,
-			    ctx->Point.MinSize,
-			    ctx->Point.MaxSize);
-
-   if (ctx->Point._Size == 1.0F)
-      ctx->_TriangleCaps &= ~DD_POINT_SIZE;
-   else
-      ctx->_TriangleCaps |= DD_POINT_SIZE;
-
-   ctx->Point._Attenuated = (ctx->Point.Params[0] != 1.0 ||
-                             ctx->Point.Params[1] != 0.0 ||
-                             ctx->Point.Params[2] != 0.0);
-
-   if (ctx->Point._Attenuated)
-      ctx->_TriangleCaps |= DD_POINT_ATTEN;
-   else
-      ctx->_TriangleCaps &= ~DD_POINT_ATTEN;
-}
-
-
-
-/**
  * Initialize the context point state.
  *
  * \param ctx GL context.
diff --git a/src/mesa/main/points.h b/src/mesa/main/points.h
index 56acd9ee574..951ff677db4 100644
--- a/src/mesa/main/points.h
+++ b/src/mesa/main/points.h
@@ -50,9 +50,6 @@ _mesa_PointParameterfEXT( GLenum pname, GLfloat param );
 extern void GLAPIENTRY
 _mesa_PointParameterfvEXT( GLenum pname, const GLfloat *params );
 
-extern void
-_mesa_update_point(GLcontext *ctx);
-
 extern void 
 _mesa_init_point( GLcontext * ctx );
 
diff --git a/src/mesa/main/polygon.c b/src/mesa/main/polygon.c
index b771408cf36..fd02e5a652d 100644
--- a/src/mesa/main/polygon.c
+++ b/src/mesa/main/polygon.c
@@ -5,9 +5,9 @@
 
 /*
  * Mesa 3-D graphics library
- * Version:  6.3
+ * Version:  6.5.1
  *
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2006  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"),
@@ -167,13 +167,8 @@ _mesa_PolygonMode( GLenum face, GLenum mode )
       return;
    }
 
-   ctx->_TriangleCaps &= ~DD_TRI_UNFILLED;
-   if (ctx->Polygon.FrontMode!=GL_FILL || ctx->Polygon.BackMode!=GL_FILL)
-      ctx->_TriangleCaps |= DD_TRI_UNFILLED;
-
-   if (ctx->Driver.PolygonMode) {
-      (*ctx->Driver.PolygonMode)( ctx, face, mode );
-   }
+   if (ctx->Driver.PolygonMode)
+      ctx->Driver.PolygonMode(ctx, face, mode);
 }
 
 #if _HAVE_FULL_GL
@@ -321,32 +316,6 @@ _mesa_PolygonOffsetEXT( GLfloat factor, GLfloat bias )
 
 
 /**********************************************************************/
-/** \name State Management */
-/*@{*/
-
-/*
- * Check polygon state and set DD_TRI_CULL_FRONT_BACK and/or DD_TRI_OFFSET
- * in ctx->_TriangleCaps if needed.
- */
-void _mesa_update_polygon( GLcontext *ctx )
-{
-   ctx->_TriangleCaps &= ~(DD_TRI_CULL_FRONT_BACK | DD_TRI_OFFSET);
-
-   if (ctx->Polygon.CullFlag && ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK)
-      ctx->_TriangleCaps |= DD_TRI_CULL_FRONT_BACK;
-
-   /* Any Polygon offsets enabled? */
-   if (ctx->Polygon.OffsetPoint ||
-       ctx->Polygon.OffsetLine ||
-       ctx->Polygon.OffsetFill) {
-      ctx->_TriangleCaps |= DD_TRI_OFFSET;
-   }
-}
-
-/*@}*/
-
-
-/**********************************************************************/
 /** \name Initialization */
 /*@{*/
 
diff --git a/src/mesa/main/polygon.h b/src/mesa/main/polygon.h
index 2550ed16877..78e8394d053 100644
--- a/src/mesa/main/polygon.h
+++ b/src/mesa/main/polygon.h
@@ -5,9 +5,9 @@
 
 /*
  * Mesa 3-D graphics library
- * Version:  6.3
+ * Version:  6.5.1
  *
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2006  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"),
@@ -60,9 +60,6 @@ _mesa_PolygonStipple( const GLubyte *mask );
 extern void GLAPIENTRY
 _mesa_GetPolygonStipple( GLubyte *mask );
 
-extern void  
-_mesa_update_polygon( GLcontext *ctx );
-
 extern void 
 _mesa_init_polygon( GLcontext * ctx );
 
diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c
index 947d3136d5b..6ed7ae6c7dd 100644
--- a/src/mesa/main/state.c
+++ b/src/mesa/main/state.c
@@ -821,16 +821,6 @@ _mesa_init_exec_table(struct _glapi_table *exec)
 /*@{*/
 
 
-static void
-update_separate_specular( GLcontext *ctx )
-{
-   if (NEED_SECONDARY_COLOR(ctx))
-      ctx->_TriangleCaps |= DD_SEPARATE_SPECULAR;
-   else
-      ctx->_TriangleCaps &= ~DD_SEPARATE_SPECULAR;
-}
-
-
 /**
  * Update state dependent on vertex arrays.
  */
@@ -909,8 +899,8 @@ update_arrays( GLcontext *ctx )
 
    /* 7 */
    if (ctx->VertexProgram._Enabled
-       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_SEVEN].Enabled) {
-      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_SEVEN]._MaxElement);
+       && ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG].Enabled) {
+      min = MIN2(min, ctx->Array.ArrayObj->VertexAttrib[VERT_ATTRIB_EDGEFLAG]._MaxElement);
    }
 
    /* 8..15 */
@@ -1047,6 +1037,79 @@ update_color(GLcontext *ctx)
 
 
 /**
+ * Update the ctx->_TriangleCaps bitfield.
+ * XXX that bitfield should really go away someday!
+ * This function must be called after other update_*() functions since
+ * there are dependencies on some other derived values.
+ */
+static void
+update_tricaps(GLcontext *ctx, GLbitfield new_state)
+{
+   ctx->_TriangleCaps = 0;
+
+   /*
+    * Points
+    */
+   if (new_state & _NEW_POINT) {
+      if (ctx->Point.SmoothFlag)
+         ctx->_TriangleCaps |= DD_POINT_SMOOTH;
+      if (ctx->Point._Size != 1.0F)
+         ctx->_TriangleCaps |= DD_POINT_SIZE;
+      if (ctx->Point._Attenuated)
+         ctx->_TriangleCaps |= DD_POINT_ATTEN;
+   }
+
+   /*
+    * Lines
+    */
+   if (new_state & _NEW_LINE) {
+      if (ctx->Line.SmoothFlag)
+         ctx->_TriangleCaps |= DD_LINE_SMOOTH;
+      if (ctx->Line.StippleFlag)
+         ctx->_TriangleCaps |= DD_LINE_STIPPLE;
+      if (ctx->Line._Width != 1.0)
+         ctx->_TriangleCaps |= DD_LINE_WIDTH;
+   }
+
+   /*
+    * Polygons
+    */
+   if (new_state & _NEW_POLYGON) {
+      if (ctx->Polygon.SmoothFlag)
+         ctx->_TriangleCaps |= DD_TRI_SMOOTH;
+      if (ctx->Polygon.StippleFlag)
+         ctx->_TriangleCaps |= DD_TRI_STIPPLE;
+      if (ctx->Polygon.FrontMode != GL_FILL
+          || ctx->Polygon.BackMode != GL_FILL)
+         ctx->_TriangleCaps |= DD_TRI_UNFILLED;
+      if (ctx->Polygon.CullFlag
+          && ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK)
+         ctx->_TriangleCaps |= DD_TRI_CULL_FRONT_BACK;
+      if (ctx->Polygon.OffsetPoint ||
+          ctx->Polygon.OffsetLine ||
+          ctx->Polygon.OffsetFill)
+         ctx->_TriangleCaps |= DD_TRI_OFFSET;
+   }
+
+   /*
+    * Lighting and shading
+    */
+   if (ctx->Light.Enabled && ctx->Light.Model.TwoSide)
+      ctx->_TriangleCaps |= DD_TRI_LIGHT_TWOSIDE;
+   if (ctx->Light.ShadeModel == GL_FLAT)
+      ctx->_TriangleCaps |= DD_FLATSHADE;
+   if (NEED_SECONDARY_COLOR(ctx))
+      ctx->_TriangleCaps |= DD_SEPARATE_SPECULAR;
+
+   /*
+    * Stencil
+    */
+   if (ctx->Stencil._TestTwoSide)
+      ctx->_TriangleCaps |= DD_TRI_TWOSTENCIL;
+}
+
+
+/**
  * Compute derived GL state.
  * If __GLcontextRec::NewState is non-zero then this function \b must
  * be called before rendering anything.
@@ -1055,7 +1118,7 @@ update_color(GLcontext *ctx)
  * management necessary.
  * 
  * \sa _mesa_update_modelview_project(), _mesa_update_texture(),
- * _mesa_update_buffer_bounds(), _mesa_update_polygon(),
+ * _mesa_update_buffer_bounds(),
  * _mesa_update_lighting() and _mesa_update_tnl_spaces().
  */
 void
@@ -1081,12 +1144,6 @@ _mesa_update_state_locked( GLcontext *ctx )
    if (new_state & (_NEW_SCISSOR | _NEW_BUFFERS | _NEW_VIEWPORT))
       _mesa_update_draw_buffer_bounds( ctx );
 
-   if (new_state & _NEW_POINT)
-      _mesa_update_point( ctx );
-
-   if (new_state & _NEW_POLYGON)
-      _mesa_update_polygon( ctx );
-
    if (new_state & _NEW_LIGHT)
       _mesa_update_lighting( ctx );
 
@@ -1096,9 +1153,6 @@ _mesa_update_state_locked( GLcontext *ctx )
    if (new_state & _IMAGE_NEW_TRANSFER_STATE)
       _mesa_update_pixel( ctx, new_state );
 
-   if (new_state & _DD_NEW_SEPARATE_SPECULAR)
-      update_separate_specular( ctx );
-
    if (new_state & (_NEW_ARRAY | _NEW_PROGRAM))
       update_arrays( ctx );
 
@@ -1108,6 +1162,10 @@ _mesa_update_state_locked( GLcontext *ctx )
    if (new_state & _NEW_COLOR)
       update_color( ctx );
 
+   if (new_state & (_NEW_POINT | _NEW_LINE | _NEW_POLYGON | _NEW_LIGHT
+                    | _NEW_STENCIL | _DD_NEW_SEPARATE_SPECULAR))
+      update_tricaps( ctx, new_state );
+
    if (ctx->FragmentProgram._MaintainTexEnvProgram) {
       if (new_state & (_NEW_TEXTURE | _DD_NEW_SEPARATE_SPECULAR | _NEW_FOG))
 	 _mesa_UpdateTexEnvProgram(ctx);
diff --git a/src/mesa/main/stencil.c b/src/mesa/main/stencil.c
index ad71a81f6e0..e61eb0030c0 100644
--- a/src/mesa/main/stencil.c
+++ b/src/mesa/main/stencil.c
@@ -119,23 +119,7 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask )
 
    ref = CLAMP( ref, 0, stencilMax );
 
-   if (ctx->Extensions.EXT_stencil_two_side) {
-      /* only set active face state */
-      const GLint face = ctx->Stencil.ActiveFace;
-      if (ctx->Stencil.Function[face] == func &&
-          ctx->Stencil.ValueMask[face] == mask &&
-          ctx->Stencil.Ref[face] == ref)
-         return;
-      FLUSH_VERTICES(ctx, _NEW_STENCIL);
-      ctx->Stencil.Function[face] = func;
-      ctx->Stencil.Ref[face] = ref;
-      ctx->Stencil.ValueMask[face] = mask;
-      if (ctx->Driver.StencilFuncSeparate) {
-         ctx->Driver.StencilFuncSeparate(ctx, face ? GL_BACK : GL_FRONT,
-                                         func, ref, mask);
-      }
-   }
-   else {
+   if (ctx->Extensions.ATI_separate_stencil) {
       /* set both front and back state */
       if (ctx->Stencil.Function[0] == func &&
           ctx->Stencil.Function[1] == func &&
@@ -153,6 +137,22 @@ _mesa_StencilFunc( GLenum func, GLint ref, GLuint mask )
                                          func, ref, mask);
       }
    }
+   else {
+      /* only set active face state */
+      const GLint face = ctx->Stencil.ActiveFace;
+      if (ctx->Stencil.Function[face] == func &&
+          ctx->Stencil.ValueMask[face] == mask &&
+          ctx->Stencil.Ref[face] == ref)
+         return;
+      FLUSH_VERTICES(ctx, _NEW_STENCIL);
+      ctx->Stencil.Function[face] = func;
+      ctx->Stencil.Ref[face] = ref;
+      ctx->Stencil.ValueMask[face] = mask;
+      if (ctx->Driver.StencilFuncSeparate) {
+         ctx->Driver.StencilFuncSeparate(ctx, face ? GL_BACK : GL_FRONT,
+                                         func, ref, mask);
+      }
+   }
 }
 
 
@@ -173,26 +173,26 @@ _mesa_StencilMask( GLuint mask )
    GET_CURRENT_CONTEXT(ctx);
    ASSERT_OUTSIDE_BEGIN_END(ctx);
 
-   if (ctx->Extensions.EXT_stencil_two_side) {
-      /* only set active face state */
-      const GLint face = ctx->Stencil.ActiveFace;
-      if (ctx->Stencil.WriteMask[face] == mask)
+   if (ctx->Extensions.ATI_separate_stencil) {
+      /* set both front and back state */
+      if (ctx->Stencil.WriteMask[0] == mask &&
+          ctx->Stencil.WriteMask[1] == mask)
          return;
       FLUSH_VERTICES(ctx, _NEW_STENCIL);
-      ctx->Stencil.WriteMask[face] = mask;
+      ctx->Stencil.WriteMask[0] = ctx->Stencil.WriteMask[1] = mask;
       if (ctx->Driver.StencilMaskSeparate) {
-         ctx->Driver.StencilMaskSeparate(ctx, face ? GL_BACK : GL_FRONT, mask);
+         ctx->Driver.StencilMaskSeparate(ctx, GL_FRONT_AND_BACK, mask);
       }
    }
    else {
-      /* set both front and back state */
-      if (ctx->Stencil.WriteMask[0] == mask &&
-          ctx->Stencil.WriteMask[1] == mask)
+      /* only set active face state */
+      const GLint face = ctx->Stencil.ActiveFace;
+      if (ctx->Stencil.WriteMask[face] == mask)
          return;
       FLUSH_VERTICES(ctx, _NEW_STENCIL);
-      ctx->Stencil.WriteMask[0] = ctx->Stencil.WriteMask[1] = mask;
+      ctx->Stencil.WriteMask[face] = mask;
       if (ctx->Driver.StencilMaskSeparate) {
-         ctx->Driver.StencilMaskSeparate(ctx, GL_FRONT_AND_BACK, mask);
+         ctx->Driver.StencilMaskSeparate(ctx, face ? GL_BACK : GL_FRONT, mask);
       }
    }
 }
@@ -273,23 +273,7 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
          return;
    }
 
-   if (ctx->Extensions.EXT_stencil_two_side) {
-      /* only set active face state */
-      const GLint face = ctx->Stencil.ActiveFace;
-      if (ctx->Stencil.ZFailFunc[face] == zfail &&
-          ctx->Stencil.ZPassFunc[face] == zpass &&
-          ctx->Stencil.FailFunc[face] == fail)
-         return;
-      FLUSH_VERTICES(ctx, _NEW_STENCIL);
-      ctx->Stencil.ZFailFunc[face] = zfail;
-      ctx->Stencil.ZPassFunc[face] = zpass;
-      ctx->Stencil.FailFunc[face] = fail;
-      if (ctx->Driver.StencilOpSeparate) {
-         ctx->Driver.StencilOpSeparate(ctx, face ? GL_BACK : GL_FRONT,
-                                       fail, zfail, zpass);
-      }
-   }
-   else {
+   if (ctx->Extensions.ATI_separate_stencil) {
       /* set both front and back state */
       if (ctx->Stencil.ZFailFunc[0] == zfail &&
           ctx->Stencil.ZFailFunc[1] == zfail &&
@@ -307,6 +291,22 @@ _mesa_StencilOp(GLenum fail, GLenum zfail, GLenum zpass)
                                        fail, zfail, zpass);
       }
    }
+   else {
+      /* only set active face state */
+      const GLint face = ctx->Stencil.ActiveFace;
+      if (ctx->Stencil.ZFailFunc[face] == zfail &&
+          ctx->Stencil.ZPassFunc[face] == zpass &&
+          ctx->Stencil.FailFunc[face] == fail)
+         return;
+      FLUSH_VERTICES(ctx, _NEW_STENCIL);
+      ctx->Stencil.ZFailFunc[face] = zfail;
+      ctx->Stencil.ZPassFunc[face] = zpass;
+      ctx->Stencil.FailFunc[face] = fail;
+      if (ctx->Driver.StencilOpSeparate) {
+         ctx->Driver.StencilOpSeparate(ctx, face ? GL_BACK : GL_FRONT,
+                                       fail, zfail, zpass);
+      }
+   }
 }
 
 
diff --git a/src/mesa/main/texcompress_s3tc.c b/src/mesa/main/texcompress_s3tc.c
index 99b703de4a5..c823967b7a0 100644
--- a/src/mesa/main/texcompress_s3tc.c
+++ b/src/mesa/main/texcompress_s3tc.c
@@ -91,7 +91,7 @@ _mesa_dlopen(const char *libname, int flags)
    return dlopen(libname, flags);
 #endif
 #else
-   return (GenericFunc) NULL;
+   return NULL;
 #endif /* USE_EXTERNAL_DXTN_LIB */
 }
 
diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c
index 1d27cd3f7c6..3cfbfa5eb54 100644
--- a/src/mesa/main/texobj.c
+++ b/src/mesa/main/texobj.c
@@ -699,7 +699,7 @@ _mesa_DeleteTextures( GLsizei n, const GLuint *textures)
             = _mesa_lookup_texture(ctx, textures[i]);
 
          if (delObj) {
-	    GLboolean delete;
+	    GLboolean deleted;
 
 	    _mesa_lock_texture(ctx, delObj);
 
@@ -728,14 +728,14 @@ _mesa_DeleteTextures( GLsizei n, const GLuint *textures)
              * XXX all RefCount accesses should be protected by a mutex.
              */
             delObj->RefCount--;
-	    delete = (delObj->RefCount == 0);
+	    deleted = (delObj->RefCount == 0);
 	    _mesa_unlock_texture(ctx, delObj);
 
 	    /* We know that refcount went to zero above, so this is
 	     * the only pointer left to delObj, so we don't have to
 	     * worry about locking any more:
 	     */
-            if (delete) {
+            if (deleted) {
                ASSERT(delObj->Name != 0); /* Never delete default tex objs */
                ASSERT(ctx->Driver.DeleteTexture);
                (*ctx->Driver.DeleteTexture)(ctx, delObj);
diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c
index 89563842c25..87f8fa7a0d4 100644
--- a/src/mesa/main/texstore.c
+++ b/src/mesa/main/texstore.c
@@ -808,7 +808,8 @@ _mesa_swizzle_ubyte_image(GLcontext *ctx,
 
 /*    _mesa_printf("map %d %d %d %d\n", map[0], map[1], map[2], map[3]);  */
 
-   if (srcRowStride == srcWidth * srcComponents &&
+   if (srcRowStride == dstRowStride &&
+       srcRowStride == srcWidth * srcComponents &&
        dimensions < 3) {
       /* 1 and 2D images only */
       GLubyte *dstImage = (GLubyte *) dstAddr
diff --git a/src/mesa/main/version.h b/src/mesa/main/version.h
index e6196050b9e..e4a74cb0d41 100644
--- a/src/mesa/main/version.h
+++ b/src/mesa/main/version.h
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.2
+ * Version:  6.5.3
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  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"),
@@ -30,8 +30,8 @@
 /* Mesa version */
 #define MESA_MAJOR 6
 #define MESA_MINOR 5
-#define MESA_PATCH 2
-#define MESA_VERSION_STRING "6.5.2"
+#define MESA_PATCH 3
+#define MESA_VERSION_STRING "6.5.3"
 
 /* To make version comparison easy */
 #define MESA_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))
@@ -39,10 +39,10 @@
 
 
 /* OpenGL API version */
-#define OPENGL_MAJOR 1
-#define OPENGL_MINOR 5
+#define OPENGL_MAJOR 2
+#define OPENGL_MINOR 0
 #define OPENGL_PATCH 0
-#define OPENGL_VERSION_STRING "1.5"
+#define OPENGL_VERSION_STRING "2.0"
 
 /* To make version comparison easy */
 #define OPENGL_VERSION(a,b,c) (((a) << 16) + ((b) << 8) + (c))
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c
index 0ac35a57682..480fdf512ca 100644
--- a/src/mesa/shader/arbprogparse.c
+++ b/src/mesa/shader/arbprogparse.c
@@ -3720,7 +3720,7 @@ parse_instructions(GLcontext * ctx, const GLubyte * inst,
 
 
 /* XXX temporary */
-__extension__ static char core_grammar_text[] =
+LONGSTRING static char core_grammar_text[] =
 #include "grammar_syn.h"
 ;
 
@@ -4118,7 +4118,7 @@ _mesa_parse_arb_vertex_program(GLcontext *ctx, GLenum target,
    program->Base.Parameters = ap.Base.Parameters; 
 
 #if DEBUG_VP
-   _mesa_printf("____________Vertex program %u __________\n", program->Base.ID);
+   _mesa_printf("____________Vertex program %u __________\n", program->Base.Id);
    _mesa_print_program(&program->Base);
 #endif
 }
diff --git a/src/mesa/shader/programopt.c b/src/mesa/shader/programopt.c
index ca11a4e547c..05a05cd3697 100644
--- a/src/mesa/shader/programopt.c
+++ b/src/mesa/shader/programopt.c
@@ -125,19 +125,15 @@ _mesa_insert_mvp_code(GLcontext *ctx, struct gl_vertex_program *vprog)
 void
 _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog)
 {
-   static const GLint fogParamsState[]
-      = { STATE_FOG, STATE_FOG_PARAMS, 0, 0, 0 };
-   static const GLint fogColorState[]
-      = { STATE_FOG, STATE_FOG_COLOR, 0, 0, 0 };
+   static const GLint fogPStateOpt[] = { STATE_INTERNAL,
+					 STATE_FOG_PARAMS_OPTIMIZED, 0, 0, 0 };
+   static const GLint fogColorState[] = { STATE_FOG, STATE_FOG_COLOR, 0, 0, 0};
    struct prog_instruction *newInst, *inst;
    const GLuint origLen = fprog->Base.NumInstructions;
-   const GLuint newLen = origLen + 6;
+   const GLuint newLen = origLen + 5;
    GLuint i;
-   GLint fogParamsRef, fogColorRef; /* state references */
+   GLint fogPRefOpt, fogColorRef; /* state references */
    GLuint colorTemp, fogFactorTemp; /* temporary registerss */
-   GLfloat fogVals[4];
-   GLuint fogConsts;                /* constant values for EXP, EXP2 mode */
-   GLuint swizzle;
 
    if (fprog->FogOption == GL_NONE) {
       _mesa_problem(ctx, "_mesa_append_fog_code() called for fragment program"
@@ -157,9 +153,9 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog)
    _mesa_memcpy(newInst, fprog->Base.Instructions,
                 origLen * sizeof(struct prog_instruction));
 
-   /* PARAM fogParamsRef = state.fog.params; */
-   fogParamsRef
-      = _mesa_add_state_reference(fprog->Base.Parameters, fogParamsState);
+   /* PARAM fogParamsRefOpt = internal optimized fog params; */
+   fogPRefOpt
+      = _mesa_add_state_reference(fprog->Base.Parameters, fogPStateOpt);
    /* PARAM fogColorRef = state.fog.color; */
    fogColorRef
       = _mesa_add_state_reference(fprog->Base.Parameters, fogColorState);
@@ -169,15 +165,6 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog)
    /* TEMP fogFactorTemp; */
    fogFactorTemp = fprog->Base.NumTemporaries++;
 
-   /* PARAM fogVals = { 1/ln(2), 1/sqrt(ln(2), 0, 0 }; */
-   fogVals[0] = 1.0 / log(2.0);
-   fogVals[1] = 1.0 / SQRTF(log(2.0));
-   fogVals[2] = 0.0;
-   fogVals[3] = 0.0;
-   fogConsts = _mesa_add_unnamed_constant(fprog->Base.Parameters,
-                                          fogVals, 4, &swizzle);
-   ASSERT(swizzle == SWIZZLE_NOOP);
-
    /* Scan program to find where result.color is written */
    inst = newInst;
    for (i = 0; i < fprog->Base.NumInstructions; i++) {
@@ -195,43 +182,39 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog)
    }
    assert(inst->Opcode == OPCODE_END); /* we'll overwrite this inst */
 
-   _mesa_init_instructions(inst, 6);
+   _mesa_init_instructions(inst, 5);
 
    /* emit instructions to compute fog blending factor */
    if (fprog->FogOption == GL_LINEAR) {
-      /* SUB fogFactorTemp.x, fogParamsRef.z, fragment.fogcoord.x; */
-      inst->Opcode = OPCODE_SUB;
-      inst->DstReg.File = PROGRAM_TEMPORARY;
-      inst->DstReg.Index = fogFactorTemp;
-      inst->DstReg.WriteMask = WRITEMASK_X;
-      inst->SrcReg[0].File = PROGRAM_STATE_VAR;
-      inst->SrcReg[0].Index = fogParamsRef;
-      inst->SrcReg[0].Swizzle = SWIZZLE_Z;
-      inst->SrcReg[1].File = PROGRAM_INPUT;
-      inst->SrcReg[1].Index = FRAG_ATTRIB_FOGC;
-      inst++;
-      /* MUL fogFactorTemp.x, fogFactorTemp, fogParamsRef.w; */
-      inst->Opcode = OPCODE_MUL;
+      /* MAD fogFactorTemp.x, fragment.fogcoord.x, fogPRefOpt.x, fogPRefOpt.y; */
+      inst->Opcode = OPCODE_MAD;
       inst->DstReg.File = PROGRAM_TEMPORARY;
       inst->DstReg.Index = fogFactorTemp;
       inst->DstReg.WriteMask = WRITEMASK_X;
-      inst->SrcReg[0].File = PROGRAM_TEMPORARY;
-      inst->SrcReg[0].Index = fogFactorTemp;
+      inst->SrcReg[0].File = PROGRAM_INPUT;
+      inst->SrcReg[0].Index = FRAG_ATTRIB_FOGC;
+      inst->SrcReg[0].Swizzle = SWIZZLE_X;
       inst->SrcReg[1].File = PROGRAM_STATE_VAR;
-      inst->SrcReg[1].Index = fogParamsRef;
-      inst->SrcReg[1].Swizzle = SWIZZLE_W;
+      inst->SrcReg[1].Index = fogPRefOpt;
+      inst->SrcReg[1].Swizzle = SWIZZLE_X;
+      inst->SrcReg[2].File = PROGRAM_STATE_VAR;
+      inst->SrcReg[2].Index = fogPRefOpt;
+      inst->SrcReg[2].Swizzle = SWIZZLE_Y;
       inst++;
    }
    else {
       ASSERT(fprog->FogOption == GL_EXP || fprog->FogOption == GL_EXP2);
-      /* MUL fogFactorTemp.x, fogParamsRef.x, fragment.fogcoord; */
+      /* fogPRefOpt.z = d/ln(2), fogPRefOpt.w = d/sqrt(ln(2) */
+      /* EXP: MUL fogFactorTemp.x, fogPRefOpt.z, fragment.fogcoord.x; */
+      /* EXP2: MUL fogFactorTemp.x, fogPRefOpt.w, fragment.fogcoord.x; */
       inst->Opcode = OPCODE_MUL;
       inst->DstReg.File = PROGRAM_TEMPORARY;
       inst->DstReg.Index = fogFactorTemp;
       inst->DstReg.WriteMask = WRITEMASK_X;
       inst->SrcReg[0].File = PROGRAM_STATE_VAR;
-      inst->SrcReg[0].Index = fogParamsRef;
-      inst->SrcReg[0].Swizzle = SWIZZLE_X; /* X=density */
+      inst->SrcReg[0].Index = fogPRefOpt;
+      inst->SrcReg[0].Swizzle
+         = (fprog->FogOption == GL_EXP) ? SWIZZLE_Z : SWIZZLE_W;
       inst->SrcReg[1].File = PROGRAM_INPUT;
       inst->SrcReg[1].Index = FRAG_ATTRIB_FOGC;
       inst->SrcReg[1].Swizzle = SWIZZLE_X;
@@ -244,23 +227,12 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog)
          inst->DstReg.WriteMask = WRITEMASK_X;
          inst->SrcReg[0].File = PROGRAM_TEMPORARY;
          inst->SrcReg[0].Index = fogFactorTemp;
+         inst->SrcReg[0].Swizzle = SWIZZLE_X;
          inst->SrcReg[1].File = PROGRAM_TEMPORARY;
          inst->SrcReg[1].Index = fogFactorTemp;
+         inst->SrcReg[1].Swizzle = SWIZZLE_X;
          inst++;
       }
-      /* EXP:  MUL fogFactorTemp.x, fogFactorTemp.x, {1/ln(2)}; */
-      /* EXP2: MUL fogFactorTemp.x, fogFactorTemp.x, {1/sqrt(ln(2))}; */
-      inst->Opcode = OPCODE_MUL;
-      inst->DstReg.File = PROGRAM_TEMPORARY;
-      inst->DstReg.Index = fogFactorTemp;
-      inst->DstReg.WriteMask = WRITEMASK_X;
-      inst->SrcReg[0].File = PROGRAM_TEMPORARY;
-      inst->SrcReg[0].Index = fogFactorTemp;
-      inst->SrcReg[1].File = PROGRAM_CONSTANT;
-      inst->SrcReg[1].Index = fogConsts;
-      inst->SrcReg[1].Swizzle
-         = (fprog->FogOption == GL_EXP) ? SWIZZLE_X : SWIZZLE_Y;
-      inst++;
       /* EX2_SAT fogFactorTemp.x, -fogFactorTemp.x; */
       inst->Opcode = OPCODE_EX2;
       inst->DstReg.File = PROGRAM_TEMPORARY;
@@ -269,6 +241,7 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog)
       inst->SrcReg[0].File = PROGRAM_TEMPORARY;
       inst->SrcReg[0].Index = fogFactorTemp;
       inst->SrcReg[0].NegateBase = GL_TRUE;
+      inst->SrcReg[0].Swizzle = SWIZZLE_X;
       inst->SaturateMode = SATURATE_ZERO_ONE;
       inst++;
    }
@@ -283,8 +256,10 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog)
       = MAKE_SWIZZLE4(SWIZZLE_X, SWIZZLE_X, SWIZZLE_X, SWIZZLE_X);
    inst->SrcReg[1].File = PROGRAM_TEMPORARY;
    inst->SrcReg[1].Index = colorTemp;
+   inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
    inst->SrcReg[2].File = PROGRAM_STATE_VAR;
    inst->SrcReg[2].Index = fogColorRef;
+   inst->SrcReg[2].Swizzle = SWIZZLE_NOOP;
    inst++;
    /* MOV result.color.w, colorTemp.x;  # copy alpha */
    inst->Opcode = OPCODE_MOV;
@@ -293,6 +268,7 @@ _mesa_append_fog_code(GLcontext *ctx, struct gl_fragment_program *fprog)
    inst->DstReg.WriteMask = WRITEMASK_W;
    inst->SrcReg[0].File = PROGRAM_TEMPORARY;
    inst->SrcReg[0].Index = colorTemp;
+   inst->SrcReg[0].Swizzle = SWIZZLE_NOOP;
    inst++;
    /* END; */
    inst->Opcode = OPCODE_END;
diff --git a/src/mesa/sources b/src/mesa/sources
index 55c282cad90..a15c5c479fe 100644
--- a/src/mesa/sources
+++ b/src/mesa/sources
@@ -79,10 +79,6 @@ MATH_SOURCES = \
 	math/m_vector.c \
 	math/m_xform.c
 
-ARRAY_CACHE_SOURCES = \
-	array_cache/ac_context.c \
-	array_cache/ac_import.c
-
 SWRAST_SOURCES = \
 	swrast/s_aaline.c \
 	swrast/s_aatriangle.c \
@@ -119,13 +115,9 @@ SWRAST_SETUP_SOURCES = \
 	swrast_setup/ss_triangle.c 
 
 TNL_SOURCES = \
-	tnl/t_array_api.c \
-	tnl/t_array_import.c \
 	tnl/t_context.c \
 	tnl/t_pipeline.c \
-	tnl/t_save_api.c \
-	tnl/t_save_loopback.c \
-	tnl/t_save_playback.c \
+	tnl/t_draw.c \
 	tnl/t_vb_arbprogram.c \
 	tnl/t_vb_arbprogram_sse.c \
 	tnl/t_vb_program.c \
@@ -141,12 +133,24 @@ TNL_SOURCES = \
 	tnl/t_vp_build.c \
 	tnl/t_vertex.c \
 	tnl/t_vertex_sse.c \
-	tnl/t_vertex_generic.c \
-	tnl/t_vtx_api.c \
-	tnl/t_vtx_generic.c \
-	tnl/t_vtx_x86.c \
-	tnl/t_vtx_eval.c \
-	tnl/t_vtx_exec.c 
+	tnl/t_vertex_generic.c 
+
+VBO_SOURCES = \
+	vbo/vbo_context.c \
+	vbo/vbo_exec.c \
+	vbo/vbo_exec_api.c \
+	vbo/vbo_exec_array.c \
+	vbo/vbo_exec_draw.c \
+	vbo/vbo_exec_eval.c \
+	vbo/vbo_rebase.c \
+	vbo/vbo_split.c \
+	vbo/vbo_split_copy.c \
+	vbo/vbo_split_inplace.c \
+	vbo/vbo_save.c \
+	vbo/vbo_save_api.c \
+	vbo/vbo_save_draw.c \
+	vbo/vbo_save_loopback.c 
+
 
 SHADER_SOURCES = \
 	shader/arbprogparse.c \
@@ -213,8 +217,7 @@ X86_SOURCES =			\
 	x86/sse_xform3.S	\
 	x86/sse_xform4.S	\
 	x86/sse_normal.S	\
-	x86/read_rgba_span_x86.S \
-	tnl/t_vtx_x86_gcc.S
+	x86/read_rgba_span_x86.S
 
 X86_API =			\
 	x86/glapi_x86.S
@@ -288,7 +291,7 @@ ALL_SOURCES = \
 SOLO_SOURCES = \
 	$(MAIN_SOURCES)		\
 	$(MATH_SOURCES)		\
-	$(ARRAY_CACHE_SOURCES)	\
+	$(VBO_SOURCES)		\
 	$(TNL_SOURCES)		\
 	$(SHADER_SOURCES)	\
 	$(SWRAST_SOURCES)	\
diff --git a/src/mesa/sparc/glapi_sparc.S b/src/mesa/sparc/glapi_sparc.S
index c2a8ca6154d..86c9f30e143 100644
--- a/src/mesa/sparc/glapi_sparc.S
+++ b/src/mesa/sparc/glapi_sparc.S
@@ -1855,6 +1855,7 @@ _mesa_sparc_glapi_end:
 	.globl glBindProgramARB ; .type glBindProgramARB,#function ; glBindProgramARB = glBindProgramNV
 	.globl glDeleteProgramsARB ; .type glDeleteProgramsARB,#function ; glDeleteProgramsARB = glDeleteProgramsNV
 	.globl glGenProgramsARB ; .type glGenProgramsARB,#function ; glGenProgramsARB = glGenProgramsNV
+	.globl glGetVertexAttribPointerv ; .type glGetVertexAttribPointerv,#function ; glGetVertexAttribPointerv = glGetVertexAttribPointervNV
 	.globl glGetVertexAttribPointervARB ; .type glGetVertexAttribPointervARB,#function ; glGetVertexAttribPointervARB = glGetVertexAttribPointervNV
 	.globl glIsProgramARB ; .type glIsProgramARB,#function ; glIsProgramARB = glIsProgramNV
 	.globl glPointParameteri ; .type glPointParameteri,#function ; glPointParameteri = glPointParameteriNV
diff --git a/src/mesa/swrast/s_atifragshader.c b/src/mesa/swrast/s_atifragshader.c
index 1ac6528a91a..947054faa30 100644
--- a/src/mesa/swrast/s_atifragshader.c
+++ b/src/mesa/swrast/s_atifragshader.c
@@ -326,7 +326,8 @@ execute_shader(GLcontext *ctx, const struct ati_fragment_shader *shader,
    struct atifs_instruction *inst;
    struct atifs_setupinst *texinst;
    GLint optype;
-   GLint i, j, pass;
+   GLuint i;
+   GLint j, pass;
    GLint dstreg;
    GLfloat src[2][3][4];
    GLfloat zeros[4] = { 0.0, 0.0, 0.0, 0.0 };
@@ -349,7 +350,7 @@ execute_shader(GLcontext *ctx, const struct ati_fragment_shader *shader,
 
 	 /* setup the source registers for color and alpha ops */
 	 for (optype = 0; optype < 2; optype++) {
-	    for (i = 0; i < inst->ArgCount[optype]; i++) {
+ 	    for (i = 0; i < inst->ArgCount[optype]; i++) {
 	       GLint index = inst->SrcReg[optype][i].Index;
 
 	       if (index >= GL_REG_0_ATI && index <= GL_REG_5_ATI)
diff --git a/src/mesa/swrast/s_buffers.c b/src/mesa/swrast/s_buffers.c
index 284ea2b5178..35f2dd64909 100644
--- a/src/mesa/swrast/s_buffers.c
+++ b/src/mesa/swrast/s_buffers.c
@@ -80,7 +80,10 @@ clear_rgba_buffer_with_masking(GLcontext *ctx, struct gl_renderbuffer *rb)
    else {
       ASSERT(span.array->ChanType == GL_FLOAT);
       for (i = 0; i < width; i++) {
-         COPY_4V(span.array->rgba[i], ctx->Color.ClearColor);
+         CLAMPED_FLOAT_TO_CHAN(span.array->rgba[i][0], ctx->Color.ClearColor[0]);
+         CLAMPED_FLOAT_TO_CHAN(span.array->rgba[i][1], ctx->Color.ClearColor[1]);
+         CLAMPED_FLOAT_TO_CHAN(span.array->rgba[i][2], ctx->Color.ClearColor[2]);
+         CLAMPED_FLOAT_TO_CHAN(span.array->rgba[i][3], ctx->Color.ClearColor[3]);
       }
    }
 
diff --git a/src/mesa/swrast/s_context.c b/src/mesa/swrast/s_context.c
index d8a5520f6b1..5e174f81d5f 100644
--- a/src/mesa/swrast/s_context.c
+++ b/src/mesa/swrast/s_context.c
@@ -395,7 +395,7 @@ _swrast_validate_texture_images(GLcontext *ctx)
             GLuint numFaces = (texObj->Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
             GLuint face;
             for (face = 0; face < numFaces; face++) {
-               GLuint lvl;
+               GLint lvl;
                for (lvl = texObj->BaseLevel; lvl <= texObj->_MaxLevel; lvl++) {
                   struct gl_texture_image *texImg = texObj->Image[face][lvl];
                   if (texImg && !texImg->Data) {
@@ -433,7 +433,7 @@ _swrast_eject_texture_images(GLcontext *ctx)
             GLuint numFaces = (texObj->Target == GL_TEXTURE_CUBE_MAP) ? 6 : 1;
             GLuint face;
             for (face = 0; face < numFaces; face++) {
-               GLuint lvl;
+               GLint lvl;
                for (lvl = texObj->BaseLevel; lvl <= texObj->_MaxLevel; lvl++) {
                   struct gl_texture_image *texImg = texObj->Image[face][lvl];
                   if (texImg && texImg->Data) {
diff --git a/src/mesa/swrast/s_copypix.c b/src/mesa/swrast/s_copypix.c
index a9bc6433de5..afab7c4d37a 100644
--- a/src/mesa/swrast/s_copypix.c
+++ b/src/mesa/swrast/s_copypix.c
@@ -860,8 +860,8 @@ fast_copy_pixels(GLcontext *ctx,
    }
 
    /* clipping not supported */
-   if (srcX < 0 || srcX + width > srcFb->Width ||
-       srcY < 0 || srcY + height > srcFb->Height ||
+   if (srcX < 0 || srcX + width > (GLint) srcFb->Width ||
+       srcY < 0 || srcY + height > (GLint) srcFb->Height ||
        dstX < dstFb->_Xmin || dstX + width > dstFb->_Xmax ||
        dstY < dstFb->_Ymin || dstY + height > dstFb->_Ymax) {
       return GL_FALSE;
diff --git a/src/mesa/swrast/s_depth.c b/src/mesa/swrast/s_depth.c
index cb584b5de8f..408174c990f 100644
--- a/src/mesa/swrast/s_depth.c
+++ b/src/mesa/swrast/s_depth.c
@@ -1289,7 +1289,7 @@ _swrast_read_depth_span_uint( GLcontext *ctx, struct gl_renderbuffer *rb,
       GLint dx = -x;
       GLint i;
       for (i = 0; i < dx; i++)
-         depth[i] = 0.0;
+         depth[i] = 0;
       x = 0;
       n -= dx;
       depth += dx;
@@ -1298,7 +1298,7 @@ _swrast_read_depth_span_uint( GLcontext *ctx, struct gl_renderbuffer *rb,
       GLint dx = x + n - (GLint) rb->Width;
       GLint i;
       for (i = 0; i < dx; i++)
-         depth[n - i - 1] = 0.0;
+         depth[n - i - 1] = 0;
       n -= dx;
    }
    if (n <= 0) {
diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c
index dd2b57c7d36..595fd5c5cde 100644
--- a/src/mesa/swrast/s_drawpix.c
+++ b/src/mesa/swrast/s_drawpix.c
@@ -110,8 +110,9 @@ fast_draw_rgba_pixels(GLcontext *ctx, GLint x, GLint y,
     */
 
    if (format == GL_RGBA && type == rbType) {
-      const GLubyte *src = _mesa_image_address2d(&unpack, pixels, width,
-                                                 height, format, type, 0, 0);
+      const GLubyte *src
+         = (const GLubyte *) _mesa_image_address2d(&unpack, pixels, width,
+                                                   height, format, type, 0, 0);
       const GLint srcStride = _mesa_image_row_stride(&unpack, width,
                                                      format, type);
       if (simpleZoom) {
@@ -139,8 +140,9 @@ fast_draw_rgba_pixels(GLcontext *ctx, GLint x, GLint y,
    }
 
    if (format == GL_RGB && type == rbType) {
-      const GLubyte *src = _mesa_image_address2d(&unpack, pixels, width,
-                                                 height, format, type, 0, 0);
+      const GLubyte *src
+         = (const GLubyte *) _mesa_image_address2d(&unpack, pixels, width,
+                                                   height, format, type, 0, 0);
       const GLint srcStride = _mesa_image_row_stride(&unpack, width,
                                                      format, type);
       if (simpleZoom) {
@@ -640,8 +642,10 @@ draw_rgba_pixels( GLcontext *ctx, GLint x, GLint y,
       /* if the span is wider than MAX_WIDTH we have to do it in chunks */
       while (skipPixels < width) {
          const GLint spanWidth = MIN2(width - skipPixels, MAX_WIDTH);
-         const GLubyte *source = _mesa_image_address2d(unpack, pixels,
-                               width, height, format, type, 0, skipPixels);
+         const GLubyte *source
+            = (const GLubyte *) _mesa_image_address2d(unpack, pixels,
+                                                      width, height, format,
+                                                      type, 0, skipPixels);
          GLint row;
 
          for (row = 0; row < height; row++) {
diff --git a/src/mesa/swrast/s_linetemp.h b/src/mesa/swrast/s_linetemp.h
index d2f37bf391c..7b66291eb5e 100644
--- a/src/mesa/swrast/s_linetemp.h
+++ b/src/mesa/swrast/s_linetemp.h
@@ -82,14 +82,15 @@ NAME( GLcontext *ctx, const SWvertex *vert0, const SWvertex *vert1 )
    GLint numPixels;
    GLint xstep, ystep;
 #if defined(DEPTH_TYPE)
-   const GLint depthBits = ctx->Visual.depthBits;
+   const GLint depthBits = ctx->DrawBuffer->Visual.depthBits;
    const GLint fixedToDepthShift = depthBits <= 16 ? FIXED_SHIFT : 0;
    struct gl_renderbuffer *zrb = ctx->DrawBuffer->Attachment[BUFFER_DEPTH].Renderbuffer;
 #define FixedToDepth(F)  ((F) >> fixedToDepthShift)
    GLint zPtrXstep, zPtrYstep;
    DEPTH_TYPE *zPtr;
 #elif defined(INTERP_Z)
-   const GLint depthBits = ctx->Visual.depthBits;
+   const GLint depthBits = ctx->DrawBuffer->Visual.depthBits;
+/*ctx->Visual.depthBits;*/
 #endif
 #ifdef PIXEL_ADDRESS
    PIXEL_TYPE *pixelPtr;
@@ -272,7 +273,7 @@ NAME( GLcontext *ctx, const SWvertex *vert0, const SWvertex *vert1 )
       }
       else {
          /* don't use fixed point */
-         span.z = (GLint) vert0->win[2];
+         span.z = (GLuint) vert0->win[2];
          span.zStep = (GLint) ((vert1->win[2] - vert0->win[2]) / numPixels);
       }
    }
diff --git a/src/mesa/swrast/s_pointtemp.h b/src/mesa/swrast/s_pointtemp.h
index eca0b9d38d8..cc6f999b91b 100644
--- a/src/mesa/swrast/s_pointtemp.h
+++ b/src/mesa/swrast/s_pointtemp.h
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.2
+ * Version:  6.5.3
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  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"),
@@ -195,7 +195,7 @@ NAME ( GLcontext *ctx, const SWvertex *vert )
    {{
       GLint x, y;
       const GLfloat radius = 0.5F * size;
-      const GLint z = (GLint) (vert->win[2] + 0.5F);
+      const GLuint z = (GLuint) (vert->win[2] + 0.5F);
       GLuint count;
 #if FLAGS & SMOOTH
       const GLfloat rmin = radius - 0.7071F;  /* 0.7071 = sqrt(2)/2 */
diff --git a/src/mesa/swrast/s_readpix.c b/src/mesa/swrast/s_readpix.c
index 75134ab4623..a7b25a46339 100644
--- a/src/mesa/swrast/s_readpix.c
+++ b/src/mesa/swrast/s_readpix.c
@@ -94,8 +94,8 @@ read_depth_pixels( GLcontext *ctx,
    /* clipping should have been done already */
    ASSERT(x >= 0);
    ASSERT(y >= 0);
-   ASSERT(x + width <= rb->Width);
-   ASSERT(y + height <= rb->Height);
+   ASSERT(x + width <= (GLint) rb->Width);
+   ASSERT(y + height <= (GLint) rb->Height);
    /* width should never be > MAX_WIDTH since we did clipping earlier */
    ASSERT(width <= MAX_WIDTH);
 
@@ -210,8 +210,8 @@ fast_read_rgba_pixels( GLcontext *ctx,
    ASSERT(rb->_BaseFormat == GL_RGBA || rb->_BaseFormat == GL_RGB);
 
    /* clipping should have already been done */
-   ASSERT(x + width <= rb->Width);
-   ASSERT(y + height <= rb->Height);
+   ASSERT(x + width <= (GLint) rb->Width);
+   ASSERT(y + height <= (GLint) rb->Height);
 
    /* check for things we can't handle here */
    if (transferOps ||
@@ -223,8 +223,9 @@ fast_read_rgba_pixels( GLcontext *ctx,
    if (format == GL_RGBA && rb->DataType == type) {
       const GLint dstStride = _mesa_image_row_stride(packing, width,
                                                      format, type);
-      GLubyte *dest = _mesa_image_address2d(packing, pixels, width, height,
-                                            format, type, 0, 0);
+      GLubyte *dest
+         = (GLubyte *) _mesa_image_address2d(packing, pixels, width, height,
+                                             format, type, 0, 0);
       GLint row;
       ASSERT(rb->GetRow);
       for (row = 0; row < height; row++) {
@@ -239,8 +240,9 @@ fast_read_rgba_pixels( GLcontext *ctx,
        type == GL_UNSIGNED_BYTE) {
       const GLint dstStride = _mesa_image_row_stride(packing, width,
                                                      format, type);
-      GLubyte *dest = _mesa_image_address2d(packing, pixels, width, height,
-                                            format, type, 0, 0);
+      GLubyte *dest
+         = (GLubyte *) _mesa_image_address2d(packing, pixels, width, height,
+                                             format, type, 0, 0);
       GLint row;
       ASSERT(rb->GetRow);
       for (row = 0; row < height; row++) {
@@ -396,8 +398,9 @@ read_rgba_pixels( GLcontext *ctx,
          = _mesa_image_row_stride(packing, width, format, type);
       GLfloat (*rgba)[4] = swrast->SpanArrays->attribs[FRAG_ATTRIB_COL0];
       GLint row;
-      GLubyte *dst = _mesa_image_address2d(packing, pixels, width, height,
-                                           format, type, 0, 0);
+      GLubyte *dst
+         = (GLubyte *) _mesa_image_address2d(packing, pixels, width, height,
+                                             format, type, 0, 0);
 
       for (row = 0; row < height; row++, y++) {
 
diff --git a/src/mesa/swrast/s_stencil.c b/src/mesa/swrast/s_stencil.c
index dbab6b3c20b..a8aa1d4b6dc 100644
--- a/src/mesa/swrast/s_stencil.c
+++ b/src/mesa/swrast/s_stencil.c
@@ -1062,7 +1062,8 @@ void
 _swrast_read_stencil_span(GLcontext *ctx, struct gl_renderbuffer *rb,
                           GLint n, GLint x, GLint y, GLstencil stencil[])
 {
-   if (y < 0 || y >= rb->Height || x + n <= 0 || x >= rb->Width) {
+   if (y < 0 || y >= (GLint) rb->Height ||
+       x + n <= 0 || x >= (GLint) rb->Width) {
       /* span is completely outside framebuffer */
       return; /* undefined values OK */
    }
@@ -1073,7 +1074,7 @@ _swrast_read_stencil_span(GLcontext *ctx, struct gl_renderbuffer *rb,
       n -= dx;
       stencil += dx;
    }
-   if (x + n > rb->Width) {
+   if (x + n > (GLint) rb->Width) {
       GLint dx = x + n - rb->Width;
       n -= dx;
    }
@@ -1103,7 +1104,8 @@ _swrast_write_stencil_span(GLcontext *ctx, GLint n, GLint x, GLint y,
    const GLuint stencilMax = (1 << fb->Visual.stencilBits) - 1;
    const GLuint stencilMask = ctx->Stencil.WriteMask[0];
 
-   if (y < 0 || y >= rb->Height || x + n <= 0 || x >= rb->Width) {
+   if (y < 0 || y >= (GLint) rb->Height ||
+       x + n <= 0 || x >= (GLint) rb->Width) {
       /* span is completely outside framebuffer */
       return; /* undefined values OK */
    }
@@ -1113,7 +1115,7 @@ _swrast_write_stencil_span(GLcontext *ctx, GLint n, GLint x, GLint y,
       n -= dx;
       stencil += dx;
    }
-   if (x + n > rb->Width) {
+   if (x + n > (GLint) rb->Width) {
       GLint dx = x + n - rb->Width;
       n -= dx;
    }
@@ -1191,7 +1193,7 @@ _swrast_clear_stencil_buffer( GLcontext *ctx, struct gl_renderbuffer *rb )
       }
       else {
          /* no bit masking */
-         if (width == rb->Width && rb->DataType == GL_UNSIGNED_BYTE) {
+         if (width == (GLint) rb->Width && rb->DataType == GL_UNSIGNED_BYTE) {
             /* optimized case */
             /* Note: bottom-to-top raster assumed! */
             GLubyte *stencil = (GLubyte *) rb->GetPointer(ctx, rb, x, y);
diff --git a/src/mesa/swrast/s_texfilter.c b/src/mesa/swrast/s_texfilter.c
index f55e23422d4..b2c5574d95d 100644
--- a/src/mesa/swrast/s_texfilter.c
+++ b/src/mesa/swrast/s_texfilter.c
@@ -1196,7 +1196,9 @@ sample_linear_2d( GLcontext *ctx,
    GLuint i;
    struct gl_texture_image *image = tObj->Image[0][tObj->BaseLevel];
    (void) lambda;
-   if (tObj->WrapS == GL_REPEAT && tObj->WrapT == GL_REPEAT) {
+   if (tObj->WrapS == GL_REPEAT &&
+       tObj->WrapT == GL_REPEAT &&
+       image->_IsPowerOfTwo) {
       for (i=0;i<n;i++) {
          sample_2d_linear_repeat(ctx, tObj, image, texcoords[i], rgba[i]);
       }
diff --git a/src/mesa/swrast/s_zoom.c b/src/mesa/swrast/s_zoom.c
index a4e8d9a36fe..9bec0bb1a1c 100644
--- a/src/mesa/swrast/s_zoom.c
+++ b/src/mesa/swrast/s_zoom.c
@@ -210,7 +210,7 @@ zoom_span( GLcontext *ctx, GLint imgX, GLint imgY, const SWspan *span,
          for (i = 0; i < zoomedWidth; i++) {
             GLint j = unzoom_x(ctx->Pixel.ZoomX, imgX, x0 + i) - span->x;
             ASSERT(j >= 0);
-            ASSERT(j < span->end);
+            ASSERT(j < (GLint) span->end);
             COPY_4UBV(zoomed.array->color.sz1.rgba[i], rgba[j]);
          }
       }
@@ -220,7 +220,7 @@ zoom_span( GLcontext *ctx, GLint imgX, GLint imgY, const SWspan *span,
          for (i = 0; i < zoomedWidth; i++) {
             GLint j = unzoom_x(ctx->Pixel.ZoomX, imgX, x0 + i) - span->x;
             ASSERT(j >= 0);
-            ASSERT(j < span->end);
+            ASSERT(j < (GLint) span->end);
             COPY_4V(zoomed.array->color.sz2.rgba[i], rgba[j]);
          }
       }
@@ -242,7 +242,7 @@ zoom_span( GLcontext *ctx, GLint imgX, GLint imgY, const SWspan *span,
          for (i = 0; i < zoomedWidth; i++) {
             GLint j = unzoom_x(ctx->Pixel.ZoomX, imgX, x0 + i) - span->x;
             ASSERT(j >= 0);
-            ASSERT(j < span->end);
+            ASSERT(j < (GLint) span->end);
             zoomed.array->color.sz1.rgba[i][0] = rgb[j][0];
             zoomed.array->color.sz1.rgba[i][1] = rgb[j][1];
             zoomed.array->color.sz1.rgba[i][2] = rgb[j][2];
@@ -255,7 +255,7 @@ zoom_span( GLcontext *ctx, GLint imgX, GLint imgY, const SWspan *span,
          for (i = 0; i < zoomedWidth; i++) {
             GLint j = unzoom_x(ctx->Pixel.ZoomX, imgX, x0 + i) - span->x;
             ASSERT(j >= 0);
-            ASSERT(j < span->end);
+            ASSERT(j < (GLint) span->end);
             zoomed.array->color.sz2.rgba[i][0] = rgb[j][0];
             zoomed.array->color.sz2.rgba[i][1] = rgb[j][1];
             zoomed.array->color.sz2.rgba[i][2] = rgb[j][2];
@@ -282,7 +282,7 @@ zoom_span( GLcontext *ctx, GLint imgX, GLint imgY, const SWspan *span,
       for (i = 0; i < zoomedWidth; i++) {
          GLint j = unzoom_x(ctx->Pixel.ZoomX, imgX, x0 + i) - span->x;
          ASSERT(j >= 0);
-         ASSERT(j < span->end);
+         ASSERT(j < (GLint) span->end);
          zoomed.array->index[i] = indexes[j];
       }
    }
@@ -292,7 +292,7 @@ zoom_span( GLcontext *ctx, GLint imgX, GLint imgY, const SWspan *span,
       for (i = 0; i < zoomedWidth; i++) {
          GLint j = unzoom_x(ctx->Pixel.ZoomX, imgX, x0 + i) - span->x;
          ASSERT(j >= 0);
-         ASSERT(j < span->end);
+         ASSERT(j < (GLint) span->end);
          zoomed.array->z[i] = zValues[j];
       }
       /* Now, fall into either the RGB or COLOR_INDEX path below */
diff --git a/src/mesa/tnl/sources b/src/mesa/tnl/sources
index e01f55dbafb..a0888be11d6 100644
--- a/src/mesa/tnl/sources
+++ b/src/mesa/tnl/sources
@@ -1,12 +1,7 @@
 # List of source files in this directory used for X.org xserver build
 MESA_TNL_SOURCES = \
-t_array_api.c \
-t_array_import.c \
 t_context.c \
 t_pipeline.c \
-t_save_api.c \
-t_save_loopback.c \
-t_save_playback.c \
 t_vb_arbprogram.c \
 t_vb_arbprogram_sse.c \
 t_vb_arbshader.c \
@@ -23,24 +18,17 @@ t_vb_vertex.c \
 t_vertex.c \
 t_vertex_generic.c \
 t_vertex_sse.c \
-t_vp_build.c \
-t_vtx_api.c \
-t_vtx_eval.c \
-t_vtx_exec.c \
-t_vtx_generic.c \
-t_vtx_x86.c
+t_vp_build.c 
 
 MESA_TNL_HEADERS = \
 t_array_api.h \
 t_array_import.h \
 t_context.h \
 t_pipeline.h \
-t_save_api.h \
 t_vb_arbprogram.h \
 t_vb_cliptmp.h \
 t_vb_lighttmp.h \
 t_vb_rendertmp.h \
 t_vertex.h \
 t_vp_build.h \
-t_vtx_api.h \
 tnl.h
diff --git a/src/mesa/tnl/t_array_api.c b/src/mesa/tnl/t_array_api.c
deleted file mode 100644
index 6826cf7e24c..00000000000
--- a/src/mesa/tnl/t_array_api.c
+++ /dev/null
@@ -1,432 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.2
- *
- * Copyright (C) 1999-2006  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"),
- * 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 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
- * BRIAN PAUL 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.
- */
-
-/**
- * \file t_array_api.c
- * \brief Vertex array API functions (glDrawArrays, etc)
- * \author Keith Whitwell
- */
-
-#include "glheader.h"
-#include "api_validate.h"
-#include "context.h"
-#include "imports.h"
-#include "macros.h"
-#include "mtypes.h"
-#include "state.h"
-
-#include "array_cache/acache.h"
-
-#include "t_array_api.h"
-#include "t_array_import.h"
-#include "t_save_api.h"
-#include "t_context.h"
-#include "t_pipeline.h"
-#include "dispatch.h"
-
-static void fallback_drawarrays( GLcontext *ctx, GLenum mode, GLint start,
-				 GLsizei count )
-{
-   GLint i;
-
-   assert(!ctx->CompileFlag);
-   assert(ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END);
-
-   CALL_Begin(GET_DISPATCH(), (mode));
-   for (i = 0; i < count; i++) 
-       CALL_ArrayElement(GET_DISPATCH(), ( start + i ));
-   CALL_End(GET_DISPATCH(), ());
-}
-
-
-static void fallback_drawelements( GLcontext *ctx, GLenum mode, GLsizei count,
-				   const GLuint *indices)
-{
-   GLint i;
-
-   assert(!ctx->CompileFlag);
-   assert(ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END);
-
-   /* Here, indices will already reflect the buffer object if active */
-
-   CALL_Begin(GET_DISPATCH(), (mode));
-   for (i = 0 ; i < count ; i++) {
-      CALL_ArrayElement(GET_DISPATCH(), ( indices[i] ));
-   }
-   CALL_End(GET_DISPATCH(), ());
-}
-
-
-/* Note this function no longer takes a 'start' value, the range is
- * assumed to start at zero.  The old trick of subtracting 'start'
- * from each index won't work if the indices are not in writeable
- * memory.
- */
-static void _tnl_draw_range_elements( GLcontext *ctx, GLenum mode,
-				      GLuint max_index,
-				      GLsizei index_count, GLuint *indices )
-
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct tnl_prim prim;
-   FLUSH_CURRENT( ctx, 0 );
-   
-   _tnl_vb_bind_arrays( ctx, 0, max_index );
-
-   tnl->vb.Primitive = &prim;
-   tnl->vb.Primitive[0].mode = mode | PRIM_BEGIN | PRIM_END;
-   tnl->vb.Primitive[0].start = 0;
-   tnl->vb.Primitive[0].count = index_count;
-   tnl->vb.PrimitiveCount = 1;
-
-   tnl->vb.Elts = (GLuint *)indices;
-
-   tnl->Driver.RunPipeline( ctx );
-}
-
-
-
-/**
- * Called via the GL API dispatcher.
- */
-void GLAPIENTRY
-_tnl_DrawArrays(GLenum mode, GLint start, GLsizei count)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   /* It's tempting to get rid of this threshold value because we take
-    * very different paths if 'count' is less than or greater than 'thresh'.
-    * I've found/fixed at least one bug which only occured for particular
-    * array sizes.  Also, several conformance tests use very short arrays
-    * which means the long-array path doesn't get tested.  -Brian
-    */
-   GLuint thresh = (ctx->Driver.NeedFlush & FLUSH_STORED_VERTICES) ? 30 : 10;
-   
-   if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(NULL, "_tnl_DrawArrays %d %d\n", start, count); 
-   
-   /* Check arguments, etc.
-    */
-   if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
-      return;
-
-   assert(!ctx->CompileFlag);
-
-   if (!ctx->Array.LockCount && (GLuint) count < thresh) {
-      /* Small primitives: attempt to share a vb (at the expense of
-       * using the immediate interface).
-      */
-      fallback_drawarrays( ctx, mode, start, count );
-   } 
-   else if (start >= (GLint) ctx->Array.LockFirst &&
-	    start + count <= (GLint)(ctx->Array.LockFirst + ctx->Array.LockCount)) {
-      
-      struct tnl_prim prim;
-
-      /* Locked primitives which can fit in a single vertex buffer:
-       */
-      FLUSH_CURRENT( ctx, 0 );
-
-      /* Locked drawarrays.  Reuse any previously transformed data.
-       */
-      _tnl_vb_bind_arrays( ctx, ctx->Array.LockFirst, 
-			   ctx->Array.LockFirst + ctx->Array.LockCount );
-
-      tnl->vb.Primitive = &prim;
-      tnl->vb.Primitive[0].mode = mode | PRIM_BEGIN | PRIM_END;
-      tnl->vb.Primitive[0].start = start;
-      tnl->vb.Primitive[0].count = count;
-      tnl->vb.PrimitiveCount = 1;
-
-      tnl->Driver.RunPipeline( ctx );
-   } 
-   else {
-      int bufsz = 256;		/* Use a small buffer for cache goodness */
-      int j, nr;
-      int minimum, modulo, skip;
-
-      /* Large primitives requiring decomposition to multiple vertex
-       * buffers:
-       */
-      switch (mode) {
-      case GL_POINTS:
-	 minimum = 0;
-	 modulo = 1;
-	 skip = 0;
-	 break;
-      case GL_LINES:
-	 minimum = 1;
-	 modulo = 2;
-	 skip = 1;
-	 break;
-      case GL_LINE_STRIP:
-	 minimum = 1;
-	 modulo = 1;
-	 skip = 0;
-	 break;
-      case GL_TRIANGLES:
-	 minimum = 2;
-	 modulo = 3;
-	 skip = 2;
-	 break;
-      case GL_TRIANGLE_STRIP:
-	 minimum = 2;
-	 modulo = 1;
-	 skip = 0;
-	 break;
-      case GL_QUADS:
-	 minimum = 3;
-	 modulo = 4;
-	 skip = 3;
-	 break;
-      case GL_QUAD_STRIP:
-	 minimum = 3;
-	 modulo = 2;
-	 skip = 0;
-	 break;
-      case GL_LINE_LOOP:
-      case GL_TRIANGLE_FAN:
-      case GL_POLYGON:
-      default:
-	 /* Primitives requiring a copied vertex (fan-like primitives)
-	  * must use the slow path if they cannot fit in a single
-	  * vertex buffer.  
-	  */
-	 if (count <= (GLint) ctx->Const.MaxArrayLockSize) {
-	    bufsz = ctx->Const.MaxArrayLockSize;
-	    minimum = 0;
-	    modulo = 1;
-	    skip = 0;
-	 }
-	 else {
-	    fallback_drawarrays( ctx, mode, start, count );
-	    return;
-	 }
-      }
-
-      FLUSH_CURRENT( ctx, 0 );
-
-      bufsz -= bufsz % modulo;
-      bufsz -= minimum;
-      count += start;
-
-      for (j = start + minimum ; j < count ; j += nr + skip ) {
-
-	 struct tnl_prim prim;
-
-	 nr = MIN2( bufsz, count - j );
-
-         /* XXX is the last parameter a count or index into the array??? */
-	 _tnl_vb_bind_arrays( ctx, j - minimum, j + nr );
-
-	 tnl->vb.Primitive = &prim;
-	 tnl->vb.Primitive[0].mode = mode;
-
-	 if (j == start + minimum)
-	    tnl->vb.Primitive[0].mode |= PRIM_BEGIN;
-
-	 if (j + nr + skip >= count)
-	    tnl->vb.Primitive[0].mode |= PRIM_END;
-
-	 tnl->vb.Primitive[0].start = 0;
-	 tnl->vb.Primitive[0].count = nr + minimum;
-	 tnl->vb.PrimitiveCount = 1;
-
-	 tnl->Driver.RunPipeline( ctx );
-      }
-   }
-}
-
-
-/**
- * Called via the GL API dispatcher.
- */
-void GLAPIENTRY
-_tnl_DrawRangeElements(GLenum mode,
-		       GLuint start, GLuint end,
-		       GLsizei count, GLenum type, const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLuint *ui_indices;
-
-   if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(NULL, "_tnl_DrawRangeElements %d %d %d\n", start, end, count); 
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      /* use indices in the buffer object */
-      if (!ctx->Array.ElementArrayBufferObj->Data) {
-         _mesa_warning(ctx,
-                       "DrawRangeElements with empty vertex elements buffer!");
-         return;
-      }
-      /* actual address is the sum of pointers */
-      indices = (const GLvoid *)
-         ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data,
-                      (const GLubyte *) indices);
-   }
-
-   /* Check arguments, etc.
-    */
-   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count,
-                                          type, indices ))
-      return;
-
-   ui_indices = (GLuint *)_ac_import_elements( ctx, GL_UNSIGNED_INT,
-					       count, type, indices );
-
-#ifdef DEBUG
-   /* check that array indices really fall inside [start, end] range */
-   {
-      GLuint i;
-      for (i = 0; i < count; i++) {
-         if (ui_indices[i] < start || ui_indices[i] > end) {
-            _mesa_warning(ctx, "Invalid array index in "
-                          "glDrawRangeElements(index=%u)", ui_indices[i]);
-         }
-      }
-   }
-#endif
-
-   assert(!ctx->CompileFlag);
-
-   if (ctx->Array.LockCount) {
-      /* Are the arrays already locked?  If so we currently have to look
-       * at the whole locked range.
-       */
-
-      if (start == 0 && ctx->Array.LockFirst == 0 && 
-	  end < (ctx->Array.LockFirst + ctx->Array.LockCount))
-	 _tnl_draw_range_elements( ctx, mode,
-				   ctx->Array.LockCount,
-				   count, ui_indices );
-      else {
-	 fallback_drawelements( ctx, mode, count, ui_indices );
-      }
-   }
-   else if (start == 0 && end < ctx->Const.MaxArrayLockSize) {
-      /* The arrays aren't locked but we can still fit them inside a
-       * single vertexbuffer.
-       */
-      _tnl_draw_range_elements( ctx, mode, end + 1, count, ui_indices );
-   }
-   else {
-      /* Range is too big to optimize:
-       */
-      fallback_drawelements( ctx, mode, count, ui_indices );
-   }
-}
-
-
-
-/**
- * Called via the GL API dispatcher.
- */
-void GLAPIENTRY
-_tnl_DrawElements(GLenum mode, GLsizei count, GLenum type,
-		  const GLvoid *indices)
-{
-   GET_CURRENT_CONTEXT(ctx);
-   GLuint *ui_indices;
-
-   if (MESA_VERBOSE & VERBOSE_API)
-      _mesa_debug(NULL, "_tnl_DrawElements %d\n", count); 
-
-   /* Check arguments, etc. */
-   if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
-      return;
-
-   if (ctx->Array.ElementArrayBufferObj->Name) {
-      /* actual address is the sum of pointers */
-      indices = (const GLvoid *)
-         ADD_POINTERS(ctx->Array.ElementArrayBufferObj->Data,
-                      (const GLubyte *) indices);
-   }
-
-   ui_indices = (GLuint *)_ac_import_elements( ctx, GL_UNSIGNED_INT,
-					       count, type, indices );
-
-   assert(!ctx->CompileFlag);
-
-   if (ctx->Array.LockCount) {
-      if (ctx->Array.LockFirst == 0)
-	 _tnl_draw_range_elements( ctx, mode,
-				   ctx->Array.LockCount,
-				   count, ui_indices );
-      else
-	 fallback_drawelements( ctx, mode, count, ui_indices );
-   }
-   else {
-      /* Scan the index list and see if we can use the locked path anyway.
-       */
-      GLuint max_elt = 0;
-      GLint i;
-
-      for (i = 0 ; i < count ; i++)
-	 if (ui_indices[i] > max_elt)
-            max_elt = ui_indices[i];
-
-      if (max_elt < ctx->Const.MaxArrayLockSize && /* can we use it? */
-	  max_elt < (GLuint) count) 	           /* do we want to use it? */
-	 _tnl_draw_range_elements( ctx, mode, max_elt+1, count, ui_indices );
-      else
-	 fallback_drawelements( ctx, mode, count, ui_indices );
-   }
-}
-
-
-/**
- * Initialize context's vertex array fields.  Called during T 'n L context
- * creation.
- */
-void _tnl_array_init( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct tnl_vertex_arrays *tmp = &tnl->array_inputs;
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-   GLuint i;
-
-   vfmt->DrawArrays = _tnl_DrawArrays;
-   vfmt->DrawElements = _tnl_DrawElements;
-   vfmt->DrawRangeElements = _tnl_DrawRangeElements;
-
-   /* Setup vector pointers that will be used to bind arrays to VB's.
-    */
-   _mesa_vector4f_init( &tmp->Obj, 0, NULL);
-   _mesa_vector4f_init( &tmp->Normal, 0, NULL);   
-   _mesa_vector4f_init( &tmp->FogCoord, 0, NULL);
-   _mesa_vector4f_init( &tmp->Index, 0, NULL);
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++)
-      _mesa_vector4f_init( &tmp->TexCoord[i], 0, NULL);
-}
-
-
-/**
- * Destroy the context's vertex array stuff.
- * Called during T 'n L context destruction.
- */
-void _tnl_array_destroy( GLcontext *ctx )
-{
-   (void) ctx;
-}
diff --git a/src/mesa/tnl/t_array_api.h b/src/mesa/tnl/t_array_api.h
deleted file mode 100644
index 61d1f696b00..00000000000
--- a/src/mesa/tnl/t_array_api.h
+++ /dev/null
@@ -1,46 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  3.5
- *
- * Copyright (C) 1999-2001  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"),
- * 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 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
- * BRIAN PAUL 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.
- */
-
-#ifndef _T_VARRAY_H
-#define _T_VARRAY_H
-
-#include "mtypes.h"
-#include "t_context.h"
-
-
-extern void GLAPIENTRY _tnl_DrawArrays(GLenum mode, GLint first, GLsizei count);
-
-extern void GLAPIENTRY _tnl_DrawElements(GLenum mode, GLsizei count, GLenum type,
-			      const GLvoid *indices);
-
-extern void GLAPIENTRY _tnl_DrawRangeElements(GLenum mode, GLuint start,
-				   GLuint end, GLsizei count, GLenum type,
-				   const GLvoid *indices);
-
-
-extern void _tnl_array_init( GLcontext *ctx );
-extern void _tnl_array_destroy( GLcontext *ctx );
-
-#endif
diff --git a/src/mesa/tnl/t_array_import.c b/src/mesa/tnl/t_array_import.c
deleted file mode 100644
index 13c5689ceb9..00000000000
--- a/src/mesa/tnl/t_array_import.c
+++ /dev/null
@@ -1,376 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.1
- *
- * Copyright (C) 1999-2006  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"),
- * 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 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
- * BRIAN PAUL 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:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "imports.h"
-#include "state.h"
-#include "mtypes.h"
-
-#include "array_cache/acache.h"
-
-#include "t_array_import.h"
-#include "t_context.h"
-
-
-/**
- * XXX writable and stride are always false in these functions...
- */
-static void _tnl_import_vertex( GLcontext *ctx,
-				GLboolean writable,
-				GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_vertex(ctx,
-			   GL_FLOAT,
-			   stride ? 4*sizeof(GLfloat) : 0,
-			   0,
-			   writable,
-			   &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Obj.data = (GLfloat (*)[4]) data;
-   inputs->Obj.start = (GLfloat *) data;
-   inputs->Obj.stride = tmp->StrideB;
-   inputs->Obj.size = tmp->Size;
-}
-
-static void _tnl_import_normal( GLcontext *ctx,
-				GLboolean writable,
-				GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_normal(ctx, GL_FLOAT,
-			   stride ? 3*sizeof(GLfloat) : 0, writable,
-			   &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Normal.data = (GLfloat (*)[4]) data;
-   inputs->Normal.start = (GLfloat *) data;
-   inputs->Normal.stride = tmp->StrideB;
-   inputs->Normal.size = 3;
-}
-
-
-static void _tnl_import_color( GLcontext *ctx,
-			       GLboolean writable,
-			       GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_color(ctx,
-			  GL_FLOAT,
-			  stride ? 4*sizeof(GLfloat) : 0,
-			  4,
-			  writable,
-			  &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Color.data = (GLfloat (*)[4]) data;
-   inputs->Color.start = (GLfloat *) data;
-   inputs->Color.stride = tmp->StrideB;
-   inputs->Color.size = tmp->Size;
-}
-
-
-static void _tnl_import_secondarycolor( GLcontext *ctx,
-					GLboolean writable,
-					GLboolean stride )
-{
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   const GLubyte *data;
-
-   tmp = _ac_import_secondarycolor(ctx, 
-				   GL_FLOAT,
-				   stride ? 4*sizeof(GLfloat) : 0,
-				   4,
-				   writable,
-				   &is_writable);
-
-   data = tmp->Ptr;
-   inputs->SecondaryColor.data = (GLfloat (*)[4]) data;
-   inputs->SecondaryColor.start = (GLfloat *) data;
-   inputs->SecondaryColor.stride = tmp->StrideB;
-   inputs->SecondaryColor.size = tmp->Size;
-}
-
-static void _tnl_import_fogcoord( GLcontext *ctx,
-				  GLboolean writable,
-				  GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   tmp = _ac_import_fogcoord(ctx, GL_FLOAT,
-			     stride ? sizeof(GLfloat) : 0, writable,
-			     &is_writable);
-
-   data = tmp->Ptr;
-   inputs->FogCoord.data = (GLfloat (*)[4]) data;
-   inputs->FogCoord.start = (GLfloat *) data;
-   inputs->FogCoord.stride = tmp->StrideB;
-}
-
-static void _tnl_import_index( GLcontext *ctx,
-			       GLboolean writable,
-			       GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   tmp = _ac_import_index(ctx, GL_FLOAT,
-			  stride ? sizeof(GLfloat) : 0, writable,
-			  &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Index.data = (GLfloat (*)[4]) data;
-   inputs->Index.start = (GLfloat *) data;
-   inputs->Index.stride = tmp->StrideB;
-}
-
-
-static void _tnl_import_texcoord( GLcontext *ctx,
-				  GLuint unit,
-				  GLboolean writable,
-				  GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   tmp = _ac_import_texcoord(ctx, unit, GL_FLOAT,
-			     stride ? 4 * sizeof(GLfloat) : 0,
-			     0,
-			     writable,
-			     &is_writable);
-
-   data = tmp->Ptr;
-   inputs->TexCoord[unit].data = (GLfloat (*)[4]) data;
-   inputs->TexCoord[unit].start = (GLfloat *) data;
-   inputs->TexCoord[unit].stride = tmp->StrideB;
-   inputs->TexCoord[unit].size = tmp->Size;
-}
-
-
-static void _tnl_import_edgeflag( GLcontext *ctx,
-				  GLboolean writable,
-				  GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-   (void) writable; (void) stride;
-
-   tmp = _ac_import_edgeflag(ctx, GL_UNSIGNED_BYTE,
-			     sizeof(GLubyte),
-			     0,
-			     &is_writable);
-
-   data = tmp->Ptr;
-   inputs->EdgeFlag = (GLubyte *) data;
-}
-
-
-
-static void _tnl_import_attrib( GLcontext *ctx,
-                                GLuint index,
-                                GLboolean writable,
-                                GLboolean stride )
-{
-   struct tnl_vertex_arrays *inputs = &TNL_CONTEXT(ctx)->array_inputs;
-   struct gl_client_array *tmp;
-   GLboolean is_writable = 0;
-   const GLubyte *data;
-
-   ASSERT(index < MAX_VERTEX_PROGRAM_ATTRIBS);
-
-   tmp = _ac_import_attrib(ctx, index, GL_FLOAT,
-                           stride ? 4 * sizeof(GLfloat) : 0,
-                           4,  /* want GLfloat[4] */
-                           writable,
-                           &is_writable);
-
-   data = tmp->Ptr;
-   inputs->Attribs[index].data = (GLfloat (*)[4]) data;
-   inputs->Attribs[index].start = (GLfloat *) data;
-   inputs->Attribs[index].stride = tmp->StrideB;
-   inputs->Attribs[index].size = tmp->Size;
-}
-
-
-static void _tnl_constant_attrib( TNLcontext *tnl,
-				  struct tnl_vertex_arrays *tmp,
-				  GLuint i )
-{
-   tmp->Attribs[i].count = 1;
-   tmp->Attribs[i].data = (GLfloat (*)[4]) tnl->vtx.current[i];
-   tmp->Attribs[i].start = tnl->vtx.current[i];
-   tmp->Attribs[i].size = 4; 
-   tmp->Attribs[i].stride = 0;
-   tnl->vb.AttribPtr[i] = &tmp->Attribs[i];
-}
-
-
-
-void _tnl_vb_bind_arrays( GLcontext *ctx, GLint start, GLint end)
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   struct tnl_vertex_arrays *tmp = &tnl->array_inputs;
-   const struct gl_vertex_program *program
-      = ctx->VertexProgram._Enabled ? ctx->VertexProgram.Current : NULL;
-   GLuint i, index;
-
-   VB->Count = end - start;
-   VB->Elts = NULL;
-
-   _ac_import_range( ctx, start, end );
-
-   /* Note that the generic attribute arrays are treated differently
-    * depending on whether an NV or ARB vertex program is enabled
-    * (corresponding to aliasing vs. non-aliasing behaviour).
-    * Generic array 0 always aliases vertex position.
-    */
-   for (index = 0; index < VERT_ATTRIB_MAX; index++) {
-      if (ctx->VertexProgram._Enabled
-          && (program->IsNVProgram || index == 0)
-          && ctx->Array.ArrayObj->VertexAttrib[index].Enabled) {
-         /* Use generic attribute array.  If an NV vertex program is active,
-          * the generic arrays override the conventional attributes.
-          * Otherwise, if an ARB vertex program is active, we'll import the
-          * generic attributes without aliasing over conventional attribs
-          * (see below).
-          */
-         _tnl_import_attrib( ctx, index, GL_FALSE, GL_TRUE );
-         VB->AttribPtr[index] = &tmp->Attribs[index];
-      }
-      /* use conventional arrays... */
-      else if (index == VERT_ATTRIB_POS) {
-	 _tnl_import_vertex( ctx, GL_FALSE, GL_FALSE );
-	 tmp->Obj.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_POS] = &tmp->Obj;
-      }
-      else if (index == VERT_ATTRIB_NORMAL) {
-	 _tnl_import_normal( ctx, GL_FALSE, GL_FALSE );
-	 tmp->Normal.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_NORMAL] = &tmp->Normal;
-      }
-      else if (index == VERT_ATTRIB_COLOR0) {
-	 _tnl_import_color( ctx, GL_FALSE, GL_FALSE );
-	 tmp->Color.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_COLOR0] = &tmp->Color;
-      }
-      else if (index == VERT_ATTRIB_COLOR1) {
-	 _tnl_import_secondarycolor( ctx, GL_FALSE, GL_FALSE );
-	 tmp->SecondaryColor.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_COLOR1] = &tmp->SecondaryColor;
-      }
-      else if (index == VERT_ATTRIB_FOG) {
-	 _tnl_import_fogcoord( ctx, GL_FALSE, GL_FALSE );
-	 tmp->FogCoord.count = VB->Count;
-	 VB->AttribPtr[_TNL_ATTRIB_FOG] = &tmp->FogCoord;
-      }
-      else if (index == VERT_ATTRIB_COLOR_INDEX) {
-         _tnl_import_index( ctx, GL_FALSE, GL_FALSE );
-         tmp->Index.count = VB->Count;
-         VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX] = &tmp->Index;
-      }
-      else if (index >= VERT_ATTRIB_TEX0 && index <= VERT_ATTRIB_TEX7) {
-	 i = index - VERT_ATTRIB_TEX0;
-	 _tnl_import_texcoord( ctx, i, GL_FALSE, GL_FALSE );
-	 tmp->TexCoord[i].count = VB->Count;
-	 VB->AttribPtr[index] = &tmp->TexCoord[i];
-      }
-      else if (index >= VERT_ATTRIB_GENERIC1 &&
-               index <= VERT_ATTRIB_GENERIC15) {
-         const GLuint arrayIndex = index - VERT_ATTRIB_GENERIC0;
-         if (program && !program->IsNVProgram &&
-	     ctx->Array.ArrayObj->VertexAttrib[arrayIndex].Enabled) {
-            /* GL_ARB_vertex_program: bind a generic attribute array */
-            _tnl_import_attrib(ctx, arrayIndex, GL_FALSE, GL_TRUE);
-            VB->AttribPtr[index] = &tmp->Attribs[arrayIndex];
-         }
-	 else {
-            _tnl_constant_attrib(tnl, tmp, index);
-         }
-      }
-      else {
-	 _tnl_constant_attrib(tnl, tmp, index);
-      }
-      assert(VB->AttribPtr[index]);
-      assert(VB->AttribPtr[index]->size);
-   }
-
-   /* odd-ball vertex attributes */
-   {
-      _tnl_import_edgeflag( ctx, GL_TRUE, sizeof(GLboolean) );
-      VB->EdgeFlag = (GLboolean *) tmp->EdgeFlag;
-   }
-
-   /* These are constant & could be precalculated:
-    */
-   for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
-      _tnl_constant_attrib(tnl, tmp, i);
-   }
-
-
-   /* Legacy pointers -- remove one day.
-    */
-   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
-   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
-   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
-   VB->ColorPtr[1] = NULL;
-   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
-   VB->IndexPtr[1] = NULL;
-   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
-   VB->SecondaryColorPtr[1] = NULL;
-   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
-   }
-}
diff --git a/src/mesa/tnl/t_array_import.h b/src/mesa/tnl/t_array_import.h
deleted file mode 100644
index 39b77641d53..00000000000
--- a/src/mesa/tnl/t_array_import.h
+++ /dev/null
@@ -1,36 +0,0 @@
-
-/*
- * Mesa 3-D graphics library
- * Version:  3.5
- *
- * Copyright (C) 1999-2001  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"),
- * 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 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
- * BRIAN PAUL 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.
- */
-
-#ifndef _T_ARRAY_IMPORT_H
-#define _T_ARRAY_IMPORT_H
-
-#include "mtypes.h"
-#include "t_context.h"
-
-extern void _tnl_vb_bind_arrays( GLcontext *ctx, GLint start, GLint end );
-
-extern void _tnl_array_import_init( GLcontext *ctx );
-
-#endif
diff --git a/src/mesa/tnl/t_context.c b/src/mesa/tnl/t_context.c
index 3887f18213a..3b2f91acbaf 100644
--- a/src/mesa/tnl/t_context.c
+++ b/src/mesa/tnl/t_context.c
@@ -26,38 +26,19 @@
  */
 
 
-#include "api_arrayelt.h"
 #include "glheader.h"
 #include "imports.h"
 #include "context.h"
 #include "macros.h"
 #include "mtypes.h"
-#include "dlist.h"
 #include "light.h"
-#include "vtxfmt.h"
 
 #include "tnl.h"
-#include "t_array_api.h"
 #include "t_context.h"
 #include "t_pipeline.h"
-#include "t_save_api.h"
 #include "t_vp_build.h"
-#include "t_vtx_api.h"
-
-
-
-static void
-install_driver_callbacks( GLcontext *ctx )
-{
-   ctx->Driver.NewList = _tnl_NewList;
-   ctx->Driver.EndList = _tnl_EndList;
-   ctx->Driver.FlushVertices = _tnl_FlushVertices;
-   ctx->Driver.SaveFlushVertices = _tnl_SaveFlushVertices;
-   ctx->Driver.BeginCallList = _tnl_BeginCallList;
-   ctx->Driver.EndCallList = _tnl_EndCallList;
-}
-
 
+#include "vbo/vbo.h"
 
 GLboolean
 _tnl_CreateContext( GLcontext *ctx )
@@ -72,20 +53,13 @@ _tnl_CreateContext( GLcontext *ctx )
       return GL_FALSE;
    }
 
-   if (_mesa_getenv("MESA_CODEGEN"))
-      tnl->AllowCodegen = GL_TRUE;
-
    /* Initialize the VB.
     */
    tnl->vb.Size = ctx->Const.MaxArrayLockSize + MAX_CLIPPED_VERTICES;
 
 
-   /* Initialize tnl state and tnl->vtxfmt.
+   /* Initialize tnl state.
     */
-   _tnl_save_init( ctx );
-   _tnl_array_init( ctx );
-   _tnl_vtx_init( ctx );
-
    if (ctx->VertexProgram._MaintainTnlProgram) {
       _tnl_ProgramCacheInit( ctx );
       _tnl_install_pipeline( ctx, _tnl_vp_pipeline );
@@ -93,34 +67,18 @@ _tnl_CreateContext( GLcontext *ctx )
       _tnl_install_pipeline( ctx, _tnl_default_pipeline );
    }
 
-   /* Initialize the arrayelt helper
-    */
-   if (!_ae_create_context( ctx ))
-      return GL_FALSE;
-
-
    tnl->NeedNdcCoords = GL_TRUE;
-   tnl->LoopbackDListCassettes = GL_FALSE;
-   tnl->CalcDListNormalLengths = GL_TRUE;
    tnl->AllowVertexFog = GL_TRUE;
    tnl->AllowPixelFog = GL_TRUE;
 
-   /* Hook our functions into exec and compile dispatch tables.
-    */
-   _mesa_install_exec_vtxfmt( ctx, &tnl->exec_vtxfmt );
-
-
    /* Set a few default values in the driver struct.
     */
-   install_driver_callbacks(ctx);
-   ctx->Driver.NeedFlush = 0;
-   ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-   ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
-
    tnl->Driver.Render.PrimTabElts = _tnl_render_tab_elts;
    tnl->Driver.Render.PrimTabVerts = _tnl_render_tab_verts;
    tnl->Driver.NotifyMaterialChange = _mesa_validate_all_lighting_tables;
 
+   tnl->nr_blocks = 0;
+
    return GL_TRUE;
 }
 
@@ -130,11 +88,7 @@ _tnl_DestroyContext( GLcontext *ctx )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
 
-   _tnl_array_destroy( ctx );
-   _tnl_vtx_destroy( ctx );
-   _tnl_save_destroy( ctx );
    _tnl_destroy_pipeline( ctx );
-   _ae_destroy_context( ctx );
 
    if (ctx->VertexProgram._MaintainTnlProgram)
       _tnl_ProgramCacheDestroy( ctx );
@@ -155,10 +109,7 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state )
          || !tnl->AllowPixelFog;
    }
 
-   _ae_invalidate_state(ctx, new_state);
-
    tnl->pipeline.new_state |= new_state;
-   tnl->vtx.eval.new_state |= new_state;
 
    /* Calculate tnl->render_inputs:
     */
@@ -210,36 +161,21 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state )
 
 
 void
-_tnl_wakeup_exec( GLcontext *ctx )
+_tnl_wakeup( GLcontext *ctx )
 {
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   install_driver_callbacks(ctx);
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-
-   /* Hook our functions into exec and compile dispatch tables.
-    */
-   _mesa_install_exec_vtxfmt( ctx, &tnl->exec_vtxfmt );
-
    /* Assume we haven't been getting state updates either:
     */
    _tnl_InvalidateState( ctx, ~0 );
 
+#if 0
    if (ctx->Light.ColorMaterialEnabled) {
       _mesa_update_color_material( ctx, 
 				   ctx->Current.Attrib[VERT_ATTRIB_COLOR0] );
    }
+#endif
 }
 
 
-void
-_tnl_wakeup_save_exec( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   _tnl_wakeup_exec( ctx );
-   _mesa_install_save_vtxfmt( ctx, &tnl->save_vtxfmt );
-}
 
 
 /**
@@ -255,27 +191,6 @@ _tnl_need_projected_coords( GLcontext *ctx, GLboolean mode )
 }
 
 void
-_tnl_need_dlist_loopback( GLcontext *ctx, GLboolean mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->LoopbackDListCassettes = mode;
-}
-
-void
-_tnl_need_dlist_norm_lengths( GLcontext *ctx, GLboolean mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->CalcDListNormalLengths = mode;
-}
-
-void
-_tnl_isolate_materials( GLcontext *ctx, GLboolean mode )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   tnl->IsolateMaterials = mode;
-}
-
-void
 _tnl_allow_vertex_fog( GLcontext *ctx, GLboolean value )
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
diff --git a/src/mesa/tnl/t_context.h b/src/mesa/tnl/t_context.h
index ea2a6bed546..733b3a5fe1a 100644
--- a/src/mesa/tnl/t_context.h
+++ b/src/mesa/tnl/t_context.h
@@ -43,29 +43,6 @@
  * stages to the vertex_buffer TNLcontext::vb, where the vertex data
  * is stored. The last stage in the pipeline is the rasterizer.
  *
- * The initial vertex_buffer data may either come from an ::immediate
- * structure or client vertex_arrays or display lists:
- *
- *
- * - The ::immediate structure records all the GL commands issued between
- * glBegin and glEnd.  \n
- * The structure accumulates data, until it is either full or it is
- * flushed (usually by a state change). Before starting then the pipeline,
- * the collected vertex data in ::immediate has to be pushed into
- * TNLcontext::vb.
- * This happens in ::_tnl_vb_bind_immediate. The pipeline is then run by
- * calling tnl_device_driver::RunPipeline = ::_tnl_run_pipeline, which
- * is stored in TNLcontext::Driver.   \n
- * An ::immediate does (for performance reasons) usually not finish with a
- * glEnd, and hence it also does not need to start with a glBegin.
- * This means that the last vertices of one ::immediate may need to be
- * saved for the next one.
- *
- *
- * - NOT SURE ABOUT THIS: The vertex_arrays structure is used to handle
- * glDrawArrays etc.  \n
- * Here, the data of the vertex_arrays is copied by ::_tnl_vb_bind_arrays
- * into TNLcontext::vb, so that the pipeline can be started.
  */
 
 
@@ -79,6 +56,7 @@
 #include "math/m_vector.h"
 #include "math/m_xform.h"
 
+#include "vbo/vbo.h"
 
 #define MAX_PIPELINE_STAGES     30
 
@@ -106,7 +84,7 @@ enum {
 	_TNL_ATTRIB_COLOR1 = 4,
 	_TNL_ATTRIB_FOG = 5,
 	_TNL_ATTRIB_COLOR_INDEX = 6,
-	_TNL_ATTRIB_SEVEN = 7,
+	_TNL_ATTRIB_EDGEFLAG = 7,
 	_TNL_ATTRIB_TEX0 = 8,
 	_TNL_ATTRIB_TEX1 = 9,
 	_TNL_ATTRIB_TEX2 = 10,
@@ -115,7 +93,8 @@ enum {
 	_TNL_ATTRIB_TEX5 = 13,
 	_TNL_ATTRIB_TEX6 = 14,
 	_TNL_ATTRIB_TEX7 = 15,
-	_TNL_ATTRIB_GENERIC0 = 16,
+
+	_TNL_ATTRIB_GENERIC0 = 16, /* doesn't really exist! */
 	_TNL_ATTRIB_GENERIC1 = 17,
 	_TNL_ATTRIB_GENERIC2 = 18,
 	_TNL_ATTRIB_GENERIC3 = 19,
@@ -131,21 +110,36 @@ enum {
 	_TNL_ATTRIB_GENERIC13 = 29,
 	_TNL_ATTRIB_GENERIC14 = 30,
 	_TNL_ATTRIB_GENERIC15 = 31,
-	_TNL_ATTRIB_MAT_FRONT_AMBIENT = 32,
-	_TNL_ATTRIB_MAT_BACK_AMBIENT = 33,
-	_TNL_ATTRIB_MAT_FRONT_DIFFUSE = 34,
-	_TNL_ATTRIB_MAT_BACK_DIFFUSE = 35,
-	_TNL_ATTRIB_MAT_FRONT_SPECULAR = 36,
-	_TNL_ATTRIB_MAT_BACK_SPECULAR = 37,
-	_TNL_ATTRIB_MAT_FRONT_EMISSION = 38,
-	_TNL_ATTRIB_MAT_BACK_EMISSION = 39,
-	_TNL_ATTRIB_MAT_FRONT_SHININESS = 40,
-	_TNL_ATTRIB_MAT_BACK_SHININESS = 41,
-	_TNL_ATTRIB_MAT_FRONT_INDEXES = 42,
-	_TNL_ATTRIB_MAT_BACK_INDEXES = 43,
-	_TNL_ATTRIB_EDGEFLAG = 44,
-	_TNL_ATTRIB_POINTSIZE = 45,
-	_TNL_ATTRIB_MAX = 46
+
+	/* These alias with the generics, but they are not active
+	 * concurrently, so it's not a problem.  The TNL module
+	 * doesn't have to do anything about this as this is how they
+	 * are passed into the _draw_prims callback.
+	 *
+	 * When we generate fixed-function replacement programs (in
+	 * t_vp_build.c currently), they refer to the appropriate
+	 * generic attribute in order to pick up per-vertex material
+	 * data.
+	 */
+	_TNL_ATTRIB_MAT_FRONT_AMBIENT = 16,
+	_TNL_ATTRIB_MAT_BACK_AMBIENT = 17,
+	_TNL_ATTRIB_MAT_FRONT_DIFFUSE = 18,
+	_TNL_ATTRIB_MAT_BACK_DIFFUSE = 19,
+	_TNL_ATTRIB_MAT_FRONT_SPECULAR = 20,
+	_TNL_ATTRIB_MAT_BACK_SPECULAR = 21,
+	_TNL_ATTRIB_MAT_FRONT_EMISSION = 22,
+	_TNL_ATTRIB_MAT_BACK_EMISSION = 23,
+	_TNL_ATTRIB_MAT_FRONT_SHININESS = 24,
+	_TNL_ATTRIB_MAT_BACK_SHININESS = 25,
+	_TNL_ATTRIB_MAT_FRONT_INDEXES = 26,
+	_TNL_ATTRIB_MAT_BACK_INDEXES = 27,
+
+	/* This is really a VERT_RESULT, not an attrib.  Need to fix
+	 * tnl to understand the difference.
+	 */
+	_TNL_ATTRIB_POINTSIZE = 16,
+
+	_TNL_ATTRIB_MAX = 32
 } ;
 
 #define _TNL_ATTRIB_TEX(u)       (_TNL_ATTRIB_TEX0 + (u))
@@ -166,8 +160,8 @@ enum {
 #define _TNL_FIRST_GENERIC _TNL_ATTRIB_GENERIC0
 #define _TNL_LAST_GENERIC  _TNL_ATTRIB_GENERIC15
 
-#define _TNL_FIRST_MAT       _TNL_ATTRIB_MAT_FRONT_AMBIENT
-#define _TNL_LAST_MAT        _TNL_ATTRIB_MAT_BACK_INDEXES
+#define _TNL_FIRST_MAT       _TNL_ATTRIB_MAT_FRONT_AMBIENT /* GENERIC0 */
+#define _TNL_LAST_MAT        _TNL_ATTRIB_MAT_BACK_INDEXES  /* GENERIC11 */
 
 /* Number of available generic attributes */
 #define _TNL_NUM_GENERIC 16
@@ -175,221 +169,22 @@ enum {
 /* Number of attributes used for evaluators */
 #define _TNL_NUM_EVAL 16
 
+
 #define PRIM_BEGIN     0x10
 #define PRIM_END       0x20
-#define PRIM_WEAK      0x40
 #define PRIM_MODE_MASK 0x0f
 
-/* 
- */
-struct tnl_prim {
-   GLuint mode;
-   GLuint start;
-   GLuint count;
-};
-
-
-
-struct tnl_eval1_map {
-   struct gl_1d_map *map;
-   GLuint sz;
-};
-
-struct tnl_eval2_map {
-   struct gl_2d_map *map;
-   GLuint sz;
-};
-
-struct tnl_eval {
-   GLuint new_state;
-   struct tnl_eval1_map map1[_TNL_NUM_EVAL];
-   struct tnl_eval2_map map2[_TNL_NUM_EVAL];
-};
-
-
-#define TNL_MAX_PRIM 16
-#define TNL_MAX_COPIED_VERTS 3
-
-struct tnl_copied_vtx {
-   GLfloat buffer[_TNL_ATTRIB_MAX * 4 * TNL_MAX_COPIED_VERTS];
-   GLuint nr;
-};
-
-#define VERT_BUFFER_SIZE 2048	/* 8kbytes */
-
-
-typedef void (*tnl_attrfv_func)( const GLfloat * );
-
-struct _tnl_dynfn {
-   struct _tnl_dynfn *next, *prev;
-   GLuint key;
-   char *code;
-};
-
-struct _tnl_dynfn_lists {
-   struct _tnl_dynfn Vertex[4];
-   struct _tnl_dynfn Attribute[4];
-};
-
-struct _tnl_dynfn_generators {
-   struct _tnl_dynfn *(*Vertex[4])( GLcontext *ctx, int key );
-   struct _tnl_dynfn *(*Attribute[4])( GLcontext *ctx, int key );
-};
-
-#define _TNL_MAX_ATTR_CODEGEN 32
-
-
-/**
- * The assembly of vertices in immediate mode is separated from
- * display list compilation.  This allows a simpler immediate mode
- * treatment and a display list compiler better suited to
- * hardware-acceleration.
- */
-struct tnl_vtx {
-   GLfloat buffer[VERT_BUFFER_SIZE];
-   GLubyte attrsz[_TNL_ATTRIB_MAX];
-   GLubyte active_sz[_TNL_ATTRIB_MAX];
-   GLuint vertex_size;
-   struct tnl_prim prim[TNL_MAX_PRIM];
-   GLuint prim_count;
-   GLfloat *vbptr;		      /* cursor, points into buffer */
-   GLfloat vertex[_TNL_ATTRIB_MAX*4]; /* current vertex */
-   GLfloat *attrptr[_TNL_ATTRIB_MAX]; /* points into vertex */
-   GLfloat *current[_TNL_ATTRIB_MAX]; /* points into ctx->Current, etc */
-   GLfloat CurrentFloatEdgeFlag;
-   GLuint counter, initial_counter;
-   struct tnl_copied_vtx copied;
-
-   /** Note extra space for error handler: */
-   tnl_attrfv_func tabfv[_TNL_ATTRIB_ERROR+1][4];
-
-   struct _tnl_dynfn_lists cache;
-   struct _tnl_dynfn_generators gen;
-
-   struct tnl_eval eval;
-   GLboolean *edgeflag_tmp;
-   GLboolean have_materials;
-};
-
-
-
-
-/* For display lists, this structure holds a run of vertices of the
- * same format, and a strictly well-formed set of begin/end pairs,
- * starting on the first vertex and ending at the last.  Vertex
- * copying on buffer breaks is precomputed according to these
- * primitives, though there are situations where the copying will need
- * correction at execute-time, perhaps by replaying the list as
- * immediate mode commands.
- *
- * On executing this list, the 'current' values may be updated with
- * the values of the final vertex, and often no fixup of the start of
- * the vertex list is required.
- *
- * Eval and other commands that don't fit into these vertex lists are
- * compiled using the fallback opcode mechanism provided by dlist.c.
- */
-struct tnl_vertex_list {
-   GLubyte attrsz[_TNL_ATTRIB_MAX];
-   GLuint vertex_size;
-
-   GLfloat *buffer;
-   GLuint count;
-   GLuint wrap_count;		/* number of copied vertices at start */
-   GLboolean have_materials;	/* bit of a hack - quick check for materials */
-   GLboolean dangling_attr_ref;	/* current attr implicitly referenced 
-				   outside the list */
-
-   GLfloat *normal_lengths;
-   struct tnl_prim *prim;
-   GLuint prim_count;
-
-   struct tnl_vertex_store *vertex_store;
-   struct tnl_primitive_store *prim_store;
-};
-
-/* These buffers should be a reasonable size to support upload to
- * hardware?  Maybe drivers should stitch them back together, or
- * specify a desired size?
- */
-#define SAVE_BUFFER_SIZE (16*1024)
-#define SAVE_PRIM_SIZE   128
-
-/* Storage to be shared among several vertex_lists.
- */
-struct tnl_vertex_store {
-   GLfloat buffer[SAVE_BUFFER_SIZE];
-   GLuint used;
-   GLuint refcount;
-};
-
-struct tnl_primitive_store {
-   struct tnl_prim buffer[SAVE_PRIM_SIZE];
-   GLuint used;
-   GLuint refcount;
-};
-
-
-struct tnl_save {
-   GLubyte attrsz[_TNL_ATTRIB_MAX];
-   GLuint vertex_size;
-
-   GLfloat *buffer;
-   GLuint count;
-   GLuint wrap_count;
-   GLuint replay_flags;
-
-   struct tnl_prim *prim;
-   GLuint prim_count, prim_max;
-
-   struct tnl_vertex_store *vertex_store;
-   struct tnl_primitive_store *prim_store;
-
-   GLfloat *vbptr;		   /* cursor, points into buffer */
-   GLfloat vertex[_TNL_ATTRIB_MAX*4];	   /* current values */
-   GLfloat *attrptr[_TNL_ATTRIB_MAX];
-   GLuint counter, initial_counter;
-   GLboolean dangling_attr_ref;
-   GLboolean have_materials;
-
-   GLuint opcode_vertex_list;
-
-   struct tnl_copied_vtx copied;
-   
-   GLfloat CurrentFloatEdgeFlag;
-
-   GLfloat *current[_TNL_ATTRIB_MAX]; /* points into ctx->ListState */
-   GLubyte *currentsz[_TNL_ATTRIB_MAX];
+static INLINE GLuint _tnl_translate_prim( const struct _mesa_prim *prim )
+{
+   GLuint flag;
+   flag = prim->mode;
+   if (prim->begin) flag |= PRIM_BEGIN;
+   if (prim->end) flag |= PRIM_END;
+   return flag;
+}
 
-   void (*tabfv[_TNL_ATTRIB_MAX][4])( const GLfloat * );
-};
 
 
-/**
- * A collection of vertex arrays.
- */
-struct tnl_vertex_arrays
-{
-   /* Conventional vertex attribute arrays */
-   GLvector4f  Obj;
-   GLvector4f  Normal;
-   GLvector4f  Color;
-   GLvector4f  SecondaryColor;
-   GLvector4f  FogCoord;
-   GLvector4f  TexCoord[MAX_TEXTURE_COORD_UNITS];
-   GLvector4f  Index;
-
-   GLubyte     *EdgeFlag;
-   GLuint      *Elt;
-
-   /* These attributes don't alias with the conventional attributes.
-    * The GL_NV_vertex_program extension defines 16 extra sets of vertex
-    * arrays which have precedent over the conventional arrays when enabled.
-    */
-   /* XXX I think the array size is wronge (47 vs. 16) */
-   GLvector4f  Attribs[_TNL_ATTRIB_MAX];
-};
-
 
 /**
  * Contains the current state of a running pipeline.
@@ -424,11 +219,10 @@ struct vertex_buffer
    GLvector4f  *FogCoordPtr;	                /* _TNL_BIT_FOG */
    GLvector4f  *VaryingPtr[MAX_VARYING];
 
-   struct tnl_prim  *Primitive;	              
+   const struct _mesa_prim  *Primitive;	              
    GLuint      PrimitiveCount;	      
 
    /* Inputs to the vertex program stage */
-   /* XXX This array may be too large (47 vs. 16) */
    GLvector4f *AttribPtr[_TNL_ATTRIB_MAX];      /* GL_NV_vertex_program */
 };
 
@@ -626,12 +420,6 @@ struct tnl_device_driver
     * arrays.
     */
 
-   GLboolean (*NotifyBegin)(GLcontext *ctx, GLenum p);
-   /* Allow drivers to hook in optimized begin/end engines.
-    * Return value:  GL_TRUE - driver handled the begin
-    *                GL_FALSE - driver didn't handle the begin
-    */
-
    /***
     *** Rendering -- These functions called only from t_vb_render.c
     ***/
@@ -737,26 +525,11 @@ typedef struct
     */
    struct tnl_device_driver Driver;
 
-   /* Execute:
-    */
-   struct tnl_vtx vtx;
-   
-   /* Compile:
-    */
-   struct tnl_save save;
-
    /* Pipeline
     */
    struct tnl_pipeline pipeline;
    struct vertex_buffer vb;
 
-   /* GLvectors for binding to vb:
-    */
-   struct tnl_vertex_arrays vtx_inputs;
-   struct tnl_vertex_arrays save_inputs;
-   struct tnl_vertex_arrays current;
-   struct tnl_vertex_arrays array_inputs;
-
    /* Clipspace/ndc/window vertex managment:
     */
    struct tnl_clipspace clipspace;
@@ -764,26 +537,21 @@ typedef struct
    /* Probably need a better configuration mechanism:
     */
    GLboolean NeedNdcCoords;
-   GLboolean LoopbackDListCassettes;
-   GLboolean CalcDListNormalLengths;
-   GLboolean IsolateMaterials;
    GLboolean AllowVertexFog;
    GLboolean AllowPixelFog;
-   GLboolean AllowCodegen;
-
    GLboolean _DoVertexFog;  /* eval fog function at each vertex? */
 
-   /* If True, it means we started a glBegin/End primtive with an invalid
-    * vertex/fragment program or incomplete framebuffer.  In that case,
-    * discard any buffered vertex data.
-    */
-   GLboolean DiscardPrimitive;
-
    DECLARE_RENDERINPUTS(render_inputs_bitset);
 
-   GLvertexformat exec_vtxfmt;
-   GLvertexformat save_vtxfmt;
+   GLvector4f tmp_inputs[VERT_ATTRIB_MAX];
 
+   /* Temp storage for t_draw.c: 
+    */
+   GLubyte *block[VERT_ATTRIB_MAX];
+   GLuint nr_blocks;
+
+   /* Cache of fixed-function-replacing vertex programs:
+    */
    struct tnl_cache *vp_cache;
 
 } TNLcontext;
diff --git a/src/mesa/tnl/t_draw.c b/src/mesa/tnl/t_draw.c
new file mode 100644
index 00000000000..c97cf5f7b21
--- /dev/null
+++ b/src/mesa/tnl/t_draw.c
@@ -0,0 +1,406 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  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"),
+ * 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 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
+ * BRIAN PAUL 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:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "glheader.h"
+#include "context.h"
+#include "imports.h"
+#include "state.h"
+#include "mtypes.h"
+#include "macros.h"
+#include "enums.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+#include "t_vp_build.h"
+#include "t_vertex.h"
+#include "tnl.h"
+
+
+
+static GLubyte *get_space(GLcontext *ctx, GLuint bytes)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLubyte *space = _mesa_malloc(bytes);
+   
+   tnl->block[tnl->nr_blocks++] = space;
+   return space;
+}
+
+
+static void free_space(GLcontext *ctx)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLuint i;
+   for (i = 0; i < tnl->nr_blocks; i++)
+      _mesa_free(tnl->block[i]);
+   tnl->nr_blocks = 0;
+}
+
+
+/* Convert the incoming array to GLfloats.  Understands the
+ * array->Normalized flag and selects the correct conversion method.
+ */
+#define CONVERT( TYPE, MACRO ) do {		\
+   GLuint i, j;					\
+   if (input->Normalized) {			\
+      for (i = 0; i < count; i++) {		\
+	 const TYPE *in = (TYPE *)ptr;		\
+	 for (j = 0; j < sz; j++) {		\
+	    *fptr++ = MACRO(*in);		\
+	    in++;				\
+	 }					\
+	 ptr += input->StrideB;			\
+      }						\
+   } else {					\
+      for (i = 0; i < count; i++) {		\
+	 const TYPE *in = (TYPE *)ptr;		\
+	 for (j = 0; j < sz; j++) {		\
+	    *fptr++ = (GLfloat)(*in);		\
+	    in++;				\
+	 }					\
+	 ptr += input->StrideB;			\
+      }						\
+   }						\
+} while (0)
+
+
+
+/* Adjust pointer to point at first requested element, convert to
+ * floating point, populate VB->AttribPtr[].
+ */
+static void _tnl_import_array( GLcontext *ctx,
+			       GLuint attrib,
+			       GLuint count,
+			       const struct gl_client_array *input,
+			       const GLubyte *ptr )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint stride = input->StrideB;
+
+   if (input->Type != GL_FLOAT) {
+      const GLuint sz = input->Size;
+      GLubyte *buf = get_space(ctx, count * sz * sizeof(GLfloat));
+      GLfloat *fptr = (GLfloat *)buf;
+
+      switch (input->Type) {
+      case GL_BYTE: 
+	 CONVERT(GLbyte, BYTE_TO_FLOAT); 
+	 break;
+      case GL_UNSIGNED_BYTE: 
+	 CONVERT(GLubyte, UBYTE_TO_FLOAT); 
+	 break;
+      case GL_SHORT: 
+	 CONVERT(GLshort, SHORT_TO_FLOAT); 
+	 break;
+      case GL_UNSIGNED_SHORT: 
+	 CONVERT(GLushort, USHORT_TO_FLOAT); 
+	 break;
+      case GL_INT: 
+	 CONVERT(GLint, INT_TO_FLOAT); 
+	 break;
+      case GL_UNSIGNED_INT: 
+	 CONVERT(GLuint, UINT_TO_FLOAT); 
+	 break;
+      case GL_DOUBLE: 
+	 CONVERT(GLdouble, (GLfloat)); 
+	 break;
+      default:
+	 assert(0);
+	 break;
+      }
+
+      ptr = buf;
+      stride = sz * sizeof(GLfloat);
+   }
+
+   VB->AttribPtr[attrib] = &tnl->tmp_inputs[attrib];
+   VB->AttribPtr[attrib]->data = (GLfloat (*)[4])ptr;
+   VB->AttribPtr[attrib]->start = (GLfloat *)ptr;
+   VB->AttribPtr[attrib]->count = count;
+   VB->AttribPtr[attrib]->stride = stride;
+   VB->AttribPtr[attrib]->size = input->Size;
+
+   /* This should die, but so should the whole GLvector4f concept: 
+    */
+   VB->AttribPtr[attrib]->flags = (((1<<input->Size)-1) | 
+				   VEC_NOT_WRITEABLE |
+				   (stride == 4*sizeof(GLfloat) ? 0 : VEC_BAD_STRIDE));
+   
+   VB->AttribPtr[attrib]->storage = NULL;
+}
+
+#define CLIPVERTS  ((6 + MAX_CLIP_PLANES) * 2)
+
+
+static GLboolean *_tnl_import_edgeflag( GLcontext *ctx,
+					const GLvector4f *input,
+					GLuint count)
+{
+   const GLubyte *ptr = (const GLubyte *)input->data;
+   const GLuint stride = input->stride;
+   GLboolean *space = (GLboolean *)get_space(ctx, count + CLIPVERTS);
+   GLboolean *bptr = space;
+   GLuint i;
+
+   for (i = 0; i < count; i++) {
+      *bptr++ = ((GLfloat *)ptr)[0] == 1.0;
+      ptr += stride;
+   }
+
+   return space;
+}
+
+
+static void bind_inputs( GLcontext *ctx, 
+			 const struct gl_client_array *inputs[],
+			 GLint count,
+			 struct gl_buffer_object **bo,
+			 GLuint *nr_bo )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint i;
+
+   /* Map all the VBOs
+    */
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      const void *ptr;
+
+      if (inputs[i]->BufferObj->Name) { 
+	 if (!inputs[i]->BufferObj->Pointer) {
+	    bo[*nr_bo] = inputs[i]->BufferObj;
+	    (*nr_bo)++;
+	    ctx->Driver.MapBuffer(ctx, 
+				  GL_ARRAY_BUFFER,
+				  GL_READ_ONLY_ARB,
+				  inputs[i]->BufferObj);
+	    
+	    assert(inputs[i]->BufferObj->Pointer);
+	 }
+	 
+	 ptr = ADD_POINTERS(inputs[i]->BufferObj->Pointer,
+			    inputs[i]->Ptr);
+      }
+      else
+	 ptr = inputs[i]->Ptr;
+
+      /* Just make sure the array is floating point, otherwise convert to
+       * temporary storage.  
+       *
+       * XXX: remove the GLvector4f type at some stage and just use
+       * client arrays.
+       */
+      _tnl_import_array(ctx, i, count, inputs[i], ptr);
+   }
+
+   /* We process only the vertices between min & max index:
+    */
+   VB->Count = count;
+
+
+   /* Legacy pointers -- remove one day.
+    */
+   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
+   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
+   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
+   VB->ColorPtr[1] = NULL;
+   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
+   VB->IndexPtr[1] = NULL;
+   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
+   VB->SecondaryColorPtr[1] = NULL;
+   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
+
+   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
+   }
+
+   /* Clipping and drawing code still requires this to be a packed
+    * array of ubytes which can be written into.  TODO: Fix and
+    * remove.
+    */
+   if (ctx->Polygon.FrontMode != GL_FILL ||
+       ctx->Polygon.BackMode != GL_FILL)
+   {
+      VB->EdgeFlag = _tnl_import_edgeflag( ctx, 
+					   VB->AttribPtr[_TNL_ATTRIB_EDGEFLAG],
+					   VB->Count );
+   }
+
+}
+
+
+/* Translate indices to GLuints and store in VB->Elts.
+ */
+static void bind_indices( GLcontext *ctx,
+			  const struct _mesa_index_buffer *ib,
+			  struct gl_buffer_object **bo,
+			  GLuint *nr_bo)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLuint i;
+   void *ptr;
+
+   if (!ib) {
+      VB->Elts = NULL;
+      return;
+   }
+
+   if (ib->obj->Name && !ib->obj->Pointer) {
+      bo[*nr_bo] = ib->obj;
+      (*nr_bo)++;
+      ctx->Driver.MapBuffer(ctx, 
+			    GL_ELEMENT_ARRAY_BUFFER,
+			    GL_READ_ONLY_ARB,
+			    ib->obj);
+
+      assert(ib->obj->Pointer);
+   }
+
+   ptr = ADD_POINTERS(ib->obj->Pointer, ib->ptr);
+
+   if (ib->type == GL_UNSIGNED_INT) {
+      VB->Elts = (GLuint *) ptr;
+   }
+   else {
+      GLuint *elts = (GLuint *)get_space(ctx, ib->count * sizeof(GLuint));
+      VB->Elts = elts;
+
+      if (ib->type == GL_UNSIGNED_SHORT) {
+	 const GLushort *in = (GLushort *)ptr;
+	 for (i = 0; i < ib->count; i++) 
+	    *elts++ = (GLuint)(*in++);
+      }
+      else {
+	 const GLubyte *in = (GLubyte *)ptr;
+	 for (i = 0; i < ib->count; i++) 
+	    *elts++ = (GLuint)(*in++);
+      }
+   }
+}
+
+static void bind_prims( GLcontext *ctx,
+			const struct _mesa_prim *prim,
+			GLuint nr_prims )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+
+   VB->Primitive = prim;
+   VB->PrimitiveCount = nr_prims;
+}
+
+static void unmap_vbos( GLcontext *ctx,
+			struct gl_buffer_object **bo,
+			GLuint nr_bo )
+{
+   GLuint i;
+   for (i = 0; i < nr_bo; i++) { 
+      ctx->Driver.UnmapBuffer(ctx, 
+			      0, /* target -- I don't see why this would be needed */
+			      bo[i]);
+   }
+}
+
+
+
+/* This is the main entrypoint into the slimmed-down software tnl
+ * module.  In a regular swtnl driver, this can be plugged straight
+ * into the vbo->Driver.DrawPrims() callback.
+ */
+void _tnl_draw_prims( GLcontext *ctx,
+		      const struct gl_client_array *arrays[],
+		      const struct _mesa_prim *prim,
+		      GLuint nr_prims,
+		      const struct _mesa_index_buffer *ib,
+		      GLuint min_index,
+		      GLuint max_index)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   const GLuint TEST_SPLIT = 0;
+   const GLint max = TEST_SPLIT ? 8 : tnl->vb.Size - MAX_CLIPPED_VERTICES;
+
+   if (0)
+   {
+      GLuint i;
+      _mesa_printf("%s %d..%d\n", __FUNCTION__, min_index, max_index);
+      for (i = 0; i < nr_prims; i++)
+	 _mesa_printf("prim %d: %s start %d count %d\n", i, 
+		      _mesa_lookup_enum_by_nr(prim[i].mode),
+		      prim[i].start,
+		      prim[i].count);
+   }
+
+   if (min_index) {
+      /* We always translate away calls with min_index != 0. 
+       */
+      vbo_rebase_prims( ctx, arrays, prim, nr_prims, ib, 
+			min_index, max_index,
+			_tnl_draw_prims );
+      return;
+   }
+   else if (max_index >= max) {
+      /* The software TNL pipeline has a fixed amount of storage for
+       * vertices and it is necessary to split incoming drawing commands
+       * if they exceed that limit.
+       */
+      struct split_limits limits;
+      limits.max_verts = max;
+      limits.max_vb_size = ~0;
+      limits.max_indices = ~0;
+
+      /* This will split the buffers one way or another and
+       * recursively call back into this function.
+       */
+      vbo_split_prims( ctx, arrays, prim, nr_prims, ib, 
+		       0, max_index,
+		       _tnl_draw_prims,
+		       &limits );
+   }
+   else {
+      /* May need to map a vertex buffer object for every attribute plus
+       * one for the index buffer.
+       */
+      struct gl_buffer_object *bo[VERT_ATTRIB_MAX + 1];
+      GLuint nr_bo = 0;
+
+      /* Binding inputs may imply mapping some vertex buffer objects.
+       * They will need to be unmapped below.
+       */
+      bind_inputs(ctx, arrays, max_index+1, bo, &nr_bo);
+      bind_indices(ctx, ib, bo, &nr_bo);
+      bind_prims(ctx, prim, nr_prims );
+
+      TNL_CONTEXT(ctx)->Driver.RunPipeline(ctx);
+
+      unmap_vbos(ctx, bo, nr_bo);
+      free_space(ctx);
+   }
+}
+
diff --git a/src/mesa/tnl/t_save_loopback.c b/src/mesa/tnl/t_save_loopback.c
deleted file mode 100644
index b96bb4ec552..00000000000
--- a/src/mesa/tnl/t_save_loopback.c
+++ /dev/null
@@ -1,330 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.3
- *
- * Copyright (C) 1999-2004  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"),
- * 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 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
- * BRIAN PAUL 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.
- */
-
-/* Author:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "enums.h"
-#include "glapi.h"
-#include "imports.h"
-#include "macros.h"
-#include "mtypes.h"
-#include "t_context.h"
-#include "t_save_api.h"
-#include "dispatch.h"
-
-/* If someone compiles a display list like:
- *      glBegin(Triangles)
- *      glVertex() 
- *      ... lots of vertices ...
- *      glEnd()
- *
- * or: 
- *      glDrawArrays(...)
- * 
- * and then tries to execute it like this:
- *
- *      glBegin(Lines)
- *      glCallList()
- *      glEnd()
- *
- * it will wind up in here, as the vertex copying used when wrapping
- * buffers in list compilation (Triangles) won't be right for how the
- * list is being executed (as Lines). 
- *
- * This could be avoided by not compiling as vertex_lists until after
- * the first glEnd() has been seen.  However, that would miss an
- * important category of display lists, for the sake of a degenerate
- * usage. 
- *
- * Further, replaying degenerately-called lists in this fashion is
- * probably still faster than the replay using opcodes.
- */
-
-typedef void (*attr_func)( GLcontext *ctx, GLint target, const GLfloat * );
-
-
-/* Wrapper functions in case glVertexAttrib*fvNV doesn't exist */
-static void VertexAttrib1fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvNV(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvNV(ctx->Exec, (target, v));
-}
-
-static attr_func vert_attrfunc[4] = {
-   VertexAttrib1fvNV,
-   VertexAttrib2fvNV,
-   VertexAttrib3fvNV,
-   VertexAttrib4fvNV
-};
-
-
-static void VertexAttrib1fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib1fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib2fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib2fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib3fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib3fvARB(ctx->Exec, (target, v));
-}
-
-static void VertexAttrib4fvARB(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   CALL_VertexAttrib4fvARB(ctx->Exec, (target, v));
-}
-
-static attr_func vert_attrfunc_arb[4] = {
-   VertexAttrib1fvARB,
-   VertexAttrib2fvARB,
-   VertexAttrib3fvARB,
-   VertexAttrib4fvARB
-};
-
-
-
-
-
-
-
-static void mat_attr1fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case _TNL_ATTRIB_MAT_FRONT_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SHININESS, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_SHININESS:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SHININESS, v ));
-      break;
-   }
-}
-
-
-static void mat_attr3fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case _TNL_ATTRIB_MAT_FRONT_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_COLOR_INDEXES, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_INDEXES:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_COLOR_INDEXES, v ));
-      break;
-   }
-}
-
-
-static void mat_attr4fv( GLcontext *ctx, GLint target, const GLfloat *v )
-{
-   switch (target) {
-   case _TNL_ATTRIB_MAT_FRONT_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_EMISSION, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_EMISSION:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_EMISSION, v ));
-      break;
-   case _TNL_ATTRIB_MAT_FRONT_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_AMBIENT, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_AMBIENT:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_AMBIENT, v ));
-      break;
-   case _TNL_ATTRIB_MAT_FRONT_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_DIFFUSE, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_DIFFUSE:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_DIFFUSE, v ));
-      break;
-   case _TNL_ATTRIB_MAT_FRONT_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_FRONT, GL_SPECULAR, v ));
-      break;
-   case _TNL_ATTRIB_MAT_BACK_SPECULAR:
-      CALL_Materialfv(ctx->Exec, ( GL_BACK, GL_SPECULAR, v ));
-      break;
-   }
-}
-
-
-static attr_func mat_attrfunc[4] = {
-   mat_attr1fv,
-   NULL,
-   mat_attr3fv,
-   mat_attr4fv
-};
-
-
-static void edgeflag_attr1fv(GLcontext *ctx, GLint target, const GLfloat *v)
-{
-   (void) target;
-   CALL_EdgeFlag(ctx->Exec, ((GLboolean)(v[0] == 1.0)));
-}
-
-struct loopback_attr {
-   GLint target;
-   GLint sz;
-   attr_func func;
-};
-
-/* Don't emit ends and begins on wrapped primitives.  Don't replay
- * wrapped vertices.  If we get here, it's probably because the the
- * precalculated wrapping is wrong.
- */
-static void loopback_prim( GLcontext *ctx,
-			   const struct tnl_vertex_list *list, GLuint i,
-			   const struct loopback_attr *la, GLuint nr )
-{
-   struct tnl_prim *prim = &list->prim[i];
-   GLint begin = prim->start;
-   GLint end = begin + prim->count;
-   GLfloat *data;
-   GLint j;
-   GLuint k;
-
-   if (prim->mode & PRIM_BEGIN) {
-      CALL_Begin(GET_DISPATCH(), ( prim->mode & PRIM_MODE_MASK ));
-   }
-   else {
-      assert(i == 0);
-      assert(begin == 0);
-      begin += list->wrap_count;
-   }
-
-   data = list->buffer + begin * list->vertex_size;
-
-   for (j = begin ; j < end ; j++) {
-      GLfloat *tmp = data + la[0].sz;
-
-      for (k = 1 ; k < nr ; k++) {
-	 la[k].func( ctx, la[k].target, tmp );
-	 tmp += la[k].sz;
-      }
-	 
-      /* Fire the vertex
-       */
-      la[0].func( ctx, VERT_ATTRIB_POS, data );
-      data = tmp;
-   }
-
-   if (prim->mode & PRIM_END) {
-      CALL_End(GET_DISPATCH(), ());
-   }
-   else {
-      assert (i == list->prim_count-1);
-   }
-}
-
-/* Primitives generated by DrawArrays/DrawElements/Rectf may be
- * caught here.  If there is no primitive in progress, execute them
- * normally, otherwise need to track and discard the generated
- * primitives.
- */
-static void loopback_weak_prim( GLcontext *ctx,
-				const struct tnl_vertex_list *list, GLuint i,
-				const struct loopback_attr *la, GLuint nr )
-{
-   if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END) 
-      loopback_prim( ctx, list, i, la, nr );
-   else {
-      struct tnl_prim *prim = &list->prim[i];
-
-      /* Use the prim_weak flag to ensure that if this primitive
-       * wraps, we don't mistake future vertex_lists for part of the
-       * surrounding primitive.
-       *
-       * While this flag is set, we are simply disposing of data
-       * generated by an operation now known to be a noop.
-       */
-      if (prim->mode & PRIM_BEGIN)
-	 ctx->Driver.CurrentExecPrimitive |= PRIM_WEAK;
-      if (prim->mode & PRIM_END)
-	 ctx->Driver.CurrentExecPrimitive &= ~PRIM_WEAK;
-   }
-}
-
-
-
-void _tnl_loopback_vertex_list( GLcontext *ctx,
-                                const struct tnl_vertex_list *list )
-{
-   struct loopback_attr la[_TNL_ATTRIB_MAX];
-   GLuint i, nr = 0;
-
-   /* conventional + generic attributes */
-   for (i = 0 ; i <= _TNL_ATTRIB_GENERIC15 ; i++) {
-      if (list->attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = list->attrsz[i];
-	 la[nr].func = vert_attrfunc[list->attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   /* material attributes */
-   for (i = _TNL_ATTRIB_MAT_FRONT_AMBIENT ; 
-	i <= _TNL_ATTRIB_MAT_BACK_INDEXES ; 
-	i++) {
-      if (list->attrsz[i]) {
-	 la[nr].target = i;
-	 la[nr].sz = list->attrsz[i];
-	 la[nr].func = mat_attrfunc[list->attrsz[i]-1];
-	 nr++;
-      }
-   }
-
-   /* special-case: edgeflag */
-   if (list->attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      la[nr].target = _TNL_ATTRIB_EDGEFLAG;
-      la[nr].sz = list->attrsz[_TNL_ATTRIB_EDGEFLAG];
-      la[nr].func = edgeflag_attr1fv;
-      nr++;
-   }
-
-   for (i = 0 ; i < list->prim_count ; i++) {
-      if (list->prim[i].mode & PRIM_WEAK)
-	 loopback_weak_prim( ctx, list, i, la, nr );
-      else
-	 loopback_prim( ctx, list, i, la, nr );
-   }
-}
diff --git a/src/mesa/tnl/t_save_playback.c b/src/mesa/tnl/t_save_playback.c
deleted file mode 100644
index 426e94372ec..00000000000
--- a/src/mesa/tnl/t_save_playback.c
+++ /dev/null
@@ -1,216 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.1
- *
- * Copyright (C) 1999-2004  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"),
- * 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 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
- * BRIAN PAUL 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.
- */
-
-/* Author:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "imports.h"
-#include "mtypes.h"
-#include "macros.h"
-#include "light.h"
-#include "state.h"
-#include "t_pipeline.h"
-#include "t_save_api.h"
-#include "t_vtx_api.h"
-
-static INLINE GLint get_size( const GLfloat *f )
-{
-   if (f[3] != 1.0) return 4;
-   if (f[2] != 0.0) return 3;
-   return 2;
-}
-
-
-/* Some nasty stuff still hanging on here.  
- *
- * TODO - remove VB->ColorPtr, etc and just use the AttrPtr's.
- */
-static void _tnl_bind_vertex_list( GLcontext *ctx,
-                                   const struct tnl_vertex_list *node )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   struct tnl_vertex_arrays *tmp = &tnl->save_inputs;
-   GLfloat *data = node->buffer;
-   GLuint attr, i;
-
-   /* Setup constant data in the VB.
-    */
-   VB->Count = node->count;
-   VB->Primitive = node->prim;
-   VB->PrimitiveCount = node->prim_count;
-   VB->Elts = NULL;
-   VB->NormalLengthPtr = node->normal_lengths;
-
-   for (attr = 0; attr <= _TNL_ATTRIB_EDGEFLAG; attr++) {
-      if (node->attrsz[attr]) {
-	 tmp->Attribs[attr].count = node->count;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) data;
-	 tmp->Attribs[attr].start = data;
-	 tmp->Attribs[attr].size = node->attrsz[attr];
-	 tmp->Attribs[attr].stride = node->vertex_size * sizeof(GLfloat);
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-	 data += node->attrsz[attr];
-      }
-      else {
-	 tmp->Attribs[attr].count = 1;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) tnl->vtx.current[attr];
-	 tmp->Attribs[attr].start = tnl->vtx.current[attr];
-	 tmp->Attribs[attr].size = get_size( tnl->vtx.current[attr] );
-	 tmp->Attribs[attr].stride = 0;
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-      }
-   }
-
-   
-   /* Copy edgeflag to a contiguous array
-    */
-   if (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL) {
-      if (node->attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-	 VB->EdgeFlag = _tnl_translate_edgeflag( ctx, data, 
-						 node->count,
-						 node->vertex_size );
-	 data++;
-      }
-      else 
-	 VB->EdgeFlag = _tnl_import_current_edgeflag( ctx, node->count );
-   }
-
-   /* Legacy pointers -- remove one day.
-    */
-   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
-   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
-   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
-   VB->ColorPtr[1] = NULL;
-   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
-   VB->IndexPtr[1] = NULL;
-   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
-   VB->SecondaryColorPtr[1] = NULL;
-   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
-   }
-}
-
-static void _playback_copy_to_current( GLcontext *ctx,
-				       const struct tnl_vertex_list *node )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   const GLfloat *data;
-   GLuint i;
-
-   if (node->count)
-      data = node->buffer + (node->count-1) * node->vertex_size;
-   else
-      data = node->buffer;
-
-   for (i = _TNL_ATTRIB_POS+1 ; i <= _TNL_ATTRIB_EDGEFLAG ; i++) {
-      if (node->attrsz[i]) {
-	 COPY_CLEAN_4V(tnl->vtx.current[i], node->attrsz[i], data);
-	 data += node->attrsz[i];
-      }
-   }
-
-   /* Edgeflag requires special treatment:
-    */
-   if (node->attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = (data[0] == 1.0);
-   }
-
-   /* Colormaterial -- this kindof sucks.
-    */
-   if (ctx->Light.ColorMaterialEnabled) {
-      _mesa_update_color_material(ctx, ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
-   }
-
-   if (node->have_materials) {
-      tnl->Driver.NotifyMaterialChange( ctx );
-      ctx->NewState |= _NEW_LIGHT;
-   }
-
-   /* CurrentExecPrimitive
-    */
-   if (node->prim_count) {
-      GLenum mode = node->prim[node->prim_count - 1].mode;
-      if (mode & PRIM_END)
-	 ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-      else
-	 ctx->Driver.CurrentExecPrimitive = (mode & PRIM_MODE_MASK);
-   }
-}
-
-
-/**
- * Execute the buffer and save copied verts.
- */
-void _tnl_playback_vertex_list( GLcontext *ctx, void *data )
-{
-   const struct tnl_vertex_list *node = (const struct tnl_vertex_list *) data;
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   FLUSH_CURRENT(ctx, 0);
-
-   if (node->prim_count > 0 && node->count > 0) {
-
-      if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END &&
-	  (node->prim[0].mode & PRIM_BEGIN)) {
-
-	 /* Degenerate case: list is called inside begin/end pair and
-	  * includes operations such as glBegin or glDrawArrays.
-	  */
-	 _mesa_error( ctx, GL_INVALID_OPERATION, "displaylist recursive begin");
-	 _tnl_loopback_vertex_list( ctx, node );
-	 return;
-      }
-      else if (tnl->save.replay_flags) {
-	 /* Various degnerate cases: translate into immediate mode
-	  * calls rather than trying to execute in place.
-	  */
-	 _tnl_loopback_vertex_list( ctx, node );
-	 return;
-      }
-      
-      if (ctx->NewState)
-	 _mesa_update_state( ctx );
-
-      if ((ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) ||
-          (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glBegin (invalid vertex/fragment program)");
-         return;
-      }
-
-      _tnl_bind_vertex_list( ctx, node );
-
-      tnl->Driver.RunPipeline( ctx );
-   }
-
-   /* Copy to current?
-    */
-   _playback_copy_to_current( ctx, node );
-}
diff --git a/src/mesa/tnl/t_vb_arbprogram.c b/src/mesa/tnl/t_vb_arbprogram.c
index 0a443b3e017..97ff4c7f102 100644
--- a/src/mesa/tnl/t_vb_arbprogram.c
+++ b/src/mesa/tnl/t_vb_arbprogram.c
@@ -1096,7 +1096,7 @@ static void compile_vertex_program( struct gl_vertex_program *program,
 { 
    struct compilation cp;
    struct tnl_compiled_program *p = CALLOC_STRUCT(tnl_compiled_program);
-   GLuint i;
+   GLint i;
 
 #if 1
    if (!program->IsNVProgram && program->IsPositionInvariant) {
@@ -1293,8 +1293,7 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
    m->nr_inputs = m->nr_outputs = 0;
 
    for (i = 0; i < VERT_ATTRIB_MAX; i++) {
-      if (program->Base.InputsRead & (1<<i) ||
-	  (i == VERT_ATTRIB_POS && program->IsPositionInvariant)) {
+      if (program->Base.InputsRead & (1<<i)) {
 	 GLuint j = m->nr_inputs++;
 	 m->input[j].idx = i;
 	 m->input[j].data = (GLfloat *)m->VB->AttribPtr[i]->data;
@@ -1302,16 +1301,15 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
 	 m->input[j].size = m->VB->AttribPtr[i]->size;
 	 ASSIGN_4V(m->File[0][REG_IN0 + i], 0, 0, 0, 1);
       }
-   }     
+   }
 
    for (i = 0; i < VERT_RESULT_MAX; i++) {
-      if (program->Base.OutputsWritten & (1 << i) ||
-	  (i == VERT_RESULT_HPOS && program->IsPositionInvariant)) {
+      if (program->Base.OutputsWritten & (1 << i)) {
 	 GLuint j = m->nr_outputs++;
 	 m->output[j].idx = i;
 	 m->output[j].data = (GLfloat *)m->attribs[i].data;
       }
-   }     
+   }
 
 
    /* Run the actual program:
@@ -1325,7 +1323,13 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
 	 case 2: m->File[0][idx][1] = m->input[j].data[1];
 	 case 1: m->File[0][idx][0] = m->input[j].data[0];
 	 }
-
+#if 0
+         printf("  attr %d/%d: %g %g %g %g\n", j, idx-REG_IN0,
+                m->input[j].data[0],
+                m->input[j].data[1],
+                m->input[j].data[2],
+                m->input[j].data[3]);
+#endif
 	 STRIDE_F(m->input[j].data, m->input[j].stride);
       }
 
@@ -1334,22 +1338,13 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
 	 call_func( p, m );
       }
       else {
+         GLint j;
 	 for (j = 0; j < p->nr_instructions; j++) {
 	    union instruction inst = p->instructions[j];	 
 	    opcode_func[inst.alu.opcode]( m, inst );
 	 }
       }
 
-      /* If the program is position invariant, multiply the input position
-       * by the MVP matrix and store in the vertex position result register.
-       */
-#if 0
-      if (program->IsPositionInvariant) {
-	 TRANSFORM_POINT( m->File[0][REG_OUT0+0], 
-			  ctx->_ModelProjectMatrix.m, 
-			  m->File[0][REG_IN0+0]);
-      }
-#endif
       for (j = 0; j < m->nr_outputs; j++) {
 	 GLuint idx = REG_OUT0 + m->output[j].idx;
 	 m->output[j].data[0] = m->File[0][idx][0];
@@ -1371,15 +1366,39 @@ run_arb_vertex_program(GLcontext *ctx, struct tnl_pipeline_stage *stage)
     * TODO: 2) Integrate t_vertex.c so that we just go straight ahead
     * and build machine vertices here.
     */
-   VB->ClipPtr = &m->attribs[VERT_RESULT_HPOS];
-   VB->ClipPtr->count = VB->Count;
 
    /* XXX There seems to be confusion between using the VERT_ATTRIB_*
     * values vs _TNL_ATTRIB_* tokens here:
     */
    outputs = program->Base.OutputsWritten;
-   if (program->IsPositionInvariant) 
-      outputs |= (1<<VERT_RESULT_HPOS);
+
+   if (program->IsPositionInvariant) {
+      /* We need the exact same transform as in the fixed function path here
+         to guarantee invariance, depending on compiler optimization flags results
+         could be different otherwise */
+      VB->ClipPtr = TransformRaw( &m->attribs[0],
+				  &ctx->_ModelProjectMatrix,
+				  m->VB->AttribPtr[0] );
+
+      /* Drivers expect this to be clean to element 4...
+       */
+      switch (VB->ClipPtr->size) {
+      case 1:
+	 /* impossible */
+      case 2:
+	 _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 2 );
+	 /* fall-through */
+      case 3:
+	 _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 3 );
+	 /* fall-through */
+      case 4:
+	 break;
+      }
+   }
+   else {
+      VB->ClipPtr = &m->attribs[VERT_RESULT_HPOS];
+      VB->ClipPtr->count = VB->Count;
+   }
 
    if (outputs & (1<<VERT_RESULT_COL0)) {
       VB->ColorPtr[0] =
diff --git a/src/mesa/tnl/t_vb_render.c b/src/mesa/tnl/t_vb_render.c
index 163fa312579..c38f0745e1b 100644
--- a/src/mesa/tnl/t_vb_render.c
+++ b/src/mesa/tnl/t_vb_render.c
@@ -305,7 +305,7 @@ static GLboolean run_render( GLcontext *ctx,
 
       for (i = 0 ; i < VB->PrimitiveCount ; i++)
       {
-	 GLuint prim = VB->Primitive[i].mode;
+	 GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
 	 GLuint start = VB->Primitive[i].start;
 	 GLuint length = VB->Primitive[i].count;
 
diff --git a/src/mesa/tnl/t_vb_rendertmp.h b/src/mesa/tnl/t_vb_rendertmp.h
index 90319eddfc5..2b5f4e93b22 100644
--- a/src/mesa/tnl/t_vb_rendertmp.h
+++ b/src/mesa/tnl/t_vb_rendertmp.h
@@ -45,8 +45,8 @@
 #endif
 
 #ifndef TEST_PRIM_END
-#define TEST_PRIM_END(flags) (flags & PRIM_END)
-#define TEST_PRIM_BEGIN(flags) (flags & PRIM_BEGIN)
+#define TEST_PRIM_END(prim) (flags & PRIM_END)
+#define TEST_PRIM_BEGIN(prim) (flags & PRIM_BEGIN)
 #endif
 
 #ifndef ELT
diff --git a/src/mesa/tnl/t_vp_build.c b/src/mesa/tnl/t_vp_build.c
index 8b75dff96a8..283782588aa 100644
--- a/src/mesa/tnl/t_vp_build.c
+++ b/src/mesa/tnl/t_vp_build.c
@@ -812,14 +812,13 @@ static struct ureg calculate_light_attenuation( struct tnl_program *p,
    /* Calculate spot attenuation:
     */
    if (!p->state->unit[i].light_spotcutoff_is_180) {
-      struct ureg spot_dir = register_param3(p, STATE_LIGHT, i,
-					     STATE_SPOT_DIRECTION);
+      struct ureg spot_dir_norm = register_param3(p, STATE_INTERNAL,
+						  STATE_SPOT_DIR_NORMALIZED, i);
       struct ureg spot = get_temp(p);
       struct ureg slt = get_temp(p);
-	       
-      emit_normalize_vec3( p, spot, spot_dir ); /* XXX: precompute! */
-      emit_op2(p, OPCODE_DP3, spot, 0, negate(VPpli), spot);
-      emit_op2(p, OPCODE_SLT, slt, 0, swizzle1(spot_dir,W), spot);
+
+      emit_op2(p, OPCODE_DP3, spot, 0, 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);
 
@@ -1109,29 +1108,27 @@ static void build_fog( struct tnl_program *p )
    }
 
    if (p->state->tnl_do_vertex_fog) {
-      struct ureg params = register_param2(p, STATE_FOG, STATE_FOG_PARAMS);
+      struct ureg params = register_param2(p, STATE_INTERNAL,
+					   STATE_FOG_PARAMS_OPTIMIZED);
       struct ureg tmp = get_temp(p);
 
       switch (p->state->fog_mode) {
       case FOG_LINEAR: {
 	 struct ureg id = get_identity_param(p);
-	 emit_op2(p, OPCODE_SUB, tmp, 0, swizzle1(params,Z), input); 
-	 emit_op2(p, OPCODE_MUL, tmp, 0, tmp, swizzle1(params,W)); 
+	 emit_op3(p, OPCODE_MAD, tmp, 0, 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_op1(p, OPCODE_ABS, tmp, 0, input); 
-	 emit_op2(p, OPCODE_MUL, tmp, 0, tmp, swizzle1(params,X)); 
-	 emit_op2(p, OPCODE_POW, fog, WRITEMASK_X, 
-		  register_const1f(p, M_E), negate(tmp)); 
+	 emit_op2(p, OPCODE_MUL, tmp, 0, tmp, swizzle1(params,Z));
+	 emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
 	 break;
       case FOG_EXP2:
-	 emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,X)); 
+	 emit_op2(p, OPCODE_MUL, tmp, 0, input, swizzle1(params,W));
 	 emit_op2(p, OPCODE_MUL, tmp, 0, tmp, tmp); 
-	 emit_op2(p, OPCODE_POW, fog, WRITEMASK_X, 
-		  register_const1f(p, M_E), negate(tmp)); 
+	 emit_op1(p, OPCODE_EX2, fog, WRITEMASK_X, negate(tmp));
 	 break;
       }
       
@@ -1323,8 +1320,6 @@ static void build_texture_transform( struct tnl_program *p )
 }
 
 
-/* Seems like it could be tighter:
- */
 static void build_pointsize( struct tnl_program *p )
 {
    struct ureg eye = get_eye_position(p);
@@ -1333,20 +1328,25 @@ static void build_pointsize( struct tnl_program *p )
    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, 0, swizzle1(get_identity_param(p), W));
-   emit_op2(p, OPCODE_MUL, ut, WRITEMASK_YZ, ut, negate(swizzle1(eye, Z)));
-   emit_op2(p, OPCODE_MUL, ut, WRITEMASK_Z, ut, negate(swizzle1(eye, Z)));
-
-
-   /* p1 +  p2 * dist + p3 * dist * dist, 0 */
-   emit_op2(p, OPCODE_DP3, ut, 0, ut, state_attenuation);
-
-   /* 1 / factor */
-   emit_op1(p, OPCODE_RCP, ut, 0, ut ); 
-
-   /* out = pointSize / factor */
-   emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size); 
+   /* p1 + dist * (p2 + dist * p3); */
+   emit_op3(p, OPCODE_MAD, ut, 0, negate(swizzle1(eye, Z)),
+		swizzle1(state_attenuation, Z), swizzle1(state_attenuation, Y));
+   emit_op3(p, OPCODE_MAD, ut, 0, negate(swizzle1(eye, Z)),
+		ut, swizzle1(state_attenuation, X));
+
+   /* 1 / sqrt(factor) */
+   emit_op1(p, OPCODE_RSQ, ut, 0, ut );
+
+#if 1
+   /* out = pointSize / sqrt(factor) */
+   emit_op2(p, OPCODE_MUL, out, WRITEMASK_X, ut, state_size);
+#else
+   /* not sure, might make sense to do clamping here,
+      but it's not done in t_vb_points neither */
+   emit_op2(p, OPCODE_MUL, ut, 0, ut, state_size);
+   emit_op2(p, OPCODE_MAX, ut, 0, ut, swizzle1(state_size, Y));
+   emit_op2(p, OPCODE_MIN, out, WRITEMASK_X, ut, swizzle1(state_size, Z));
+#endif
 
    release_temp(p, ut);
 }
diff --git a/src/mesa/tnl/t_vtx_api.c b/src/mesa/tnl/t_vtx_api.c
deleted file mode 100644
index 77eec8b3b44..00000000000
--- a/src/mesa/tnl/t_vtx_api.c
+++ /dev/null
@@ -1,1046 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
-TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "vtxfmt.h"
-#include "dlist.h"
-#include "state.h"
-#include "light.h"
-#include "api_arrayelt.h"
-#include "api_noop.h"
-#include "t_vtx_api.h"
-#include "simple_list.h"
-
-#include "dispatch.h"
-
-static void reset_attrfv( TNLcontext *tnl );
-
-/** Note extra space for error index: */
-static tnl_attrfv_func choose[_TNL_ATTRIB_ERROR+1][4];
-static tnl_attrfv_func generic_attr_func[_TNL_MAX_ATTR_CODEGEN][4];
-
-
-/* Close off the last primitive, execute the buffer, restart the
- * primitive.  
- */
-static void _tnl_wrap_buffers( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   
-
-   if (tnl->vtx.prim_count == 0) {
-      tnl->vtx.copied.nr = 0;
-      tnl->vtx.counter = tnl->vtx.initial_counter;
-      tnl->vtx.vbptr = tnl->vtx.buffer;
-   }
-   else {
-      GLuint last_prim = tnl->vtx.prim[tnl->vtx.prim_count-1].mode;
-      GLuint last_count;
-
-      if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-	 GLint i = tnl->vtx.prim_count - 1;
-	 assert(i >= 0);
-	 tnl->vtx.prim[i].count = ((tnl->vtx.initial_counter - 
-				    tnl->vtx.counter) - 
-				   tnl->vtx.prim[i].start);
-      }
-
-      last_count = tnl->vtx.prim[tnl->vtx.prim_count-1].count;
-
-      /* Execute the buffer and save copied vertices.
-       */
-      if (tnl->vtx.counter != tnl->vtx.initial_counter)
-	 _tnl_flush_vtx( ctx );
-      else {
-	 tnl->vtx.prim_count = 0;
-	 tnl->vtx.copied.nr = 0;
-      }
-
-      /* Emit a glBegin to start the new list.
-       */
-      assert(tnl->vtx.prim_count == 0);
-
-      if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-	 tnl->vtx.prim[0].mode = ctx->Driver.CurrentExecPrimitive;
-	 tnl->vtx.prim[0].start = 0;
-	 tnl->vtx.prim[0].count = 0;
-	 tnl->vtx.prim_count++;
-      
-	 if (tnl->vtx.copied.nr == last_count)
-	    tnl->vtx.prim[0].mode |= last_prim & PRIM_BEGIN;
-      }
-   }
-}
-
-
-/* Deal with buffer wrapping where provoked by the vertex buffer
- * filling up, as opposed to upgrade_vertex().
- *
- * Make it GLAPIENTRY, so we can tail from the codegen'ed Vertex*fv
- */
-void GLAPIENTRY _tnl_wrap_filled_vertex( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLfloat *data = tnl->vtx.copied.buffer;
-   GLuint i;
-
-   /* Run pipeline on current vertices, copy wrapped vertices
-    * to tnl->copied.
-    */
-   _tnl_wrap_buffers( ctx );
-   
-   /* Copy stored stored vertices to start of new list. 
-    */
-   assert(tnl->vtx.counter > tnl->vtx.copied.nr);
-
-   for (i = 0 ; i < tnl->vtx.copied.nr ; i++) {
-      _mesa_memcpy( tnl->vtx.vbptr, data, 
-		    tnl->vtx.vertex_size * sizeof(GLfloat));
-      tnl->vtx.vbptr += tnl->vtx.vertex_size;
-      data += tnl->vtx.vertex_size;
-      tnl->vtx.counter--;
-   }
-
-   tnl->vtx.copied.nr = 0;
-}
-
-
-/*
- * Copy the active vertex's values to the ctx->Current fields.
- */
-static void _tnl_copy_to_current( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint i;
-
-   for (i = _TNL_ATTRIB_POS+1 ; i < _TNL_ATTRIB_EDGEFLAG ; i++) {
-      if (tnl->vtx.attrsz[i]) {
-         /* Note: the tnl->vtx.current[i] pointers points to
-          * the ctx->Current fields.  The first 16 or so, anyway.
-          */
-	 COPY_CLEAN_4V(tnl->vtx.current[i], 
-                       tnl->vtx.attrsz[i], 
-                       tnl->vtx.attrptr[i]);
-      }
-   }
-
-   /* Edgeflag requires additional treatment:
-    */
-   if (tnl->vtx.attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = 
-	 (tnl->vtx.CurrentFloatEdgeFlag == 1.0);
-   }
-
-   /* Colormaterial -- this kindof sucks.
-    */
-   if (ctx->Light.ColorMaterialEnabled) {
-      _mesa_update_color_material(ctx, 
-				  ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
-   }
-
-   if (tnl->vtx.have_materials) {
-      tnl->Driver.NotifyMaterialChange( ctx );
-      ctx->NewState |= _NEW_LIGHT;
-   }
-         
-   ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
-}
-
-
-static void _tnl_copy_from_current( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLint i;
-
-   /* Edgeflag requires additional treatment:
-    */
-   tnl->vtx.CurrentFloatEdgeFlag = (GLfloat) ctx->Current.EdgeFlag;
-   
-   for (i = _TNL_ATTRIB_POS+1 ; i < _TNL_ATTRIB_MAX ; i++) 
-      switch (tnl->vtx.attrsz[i]) {
-      case 4: tnl->vtx.attrptr[i][3] = tnl->vtx.current[i][3];
-      case 3: tnl->vtx.attrptr[i][2] = tnl->vtx.current[i][2];
-      case 2: tnl->vtx.attrptr[i][1] = tnl->vtx.current[i][1];
-      case 1: tnl->vtx.attrptr[i][0] = tnl->vtx.current[i][0];
-	 break;
-      }
-
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-}
-
-
-/* Flush existing data, set new attrib size, replay copied vertices.
- */ 
-static void _tnl_wrap_upgrade_vertex( GLcontext *ctx, 
-				      GLuint attr,
-				      GLuint newsz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint oldsz;
-   GLuint i;
-   GLfloat *tmp;
-   GLint lastcount = tnl->vtx.initial_counter - tnl->vtx.counter;
-
-   /* Run pipeline on current vertices, copy wrapped vertices
-    * to tnl->vtx.copied.
-    */
-   _tnl_wrap_buffers( ctx );
-
-
-   /* Do a COPY_TO_CURRENT to ensure back-copying works for the case
-    * when the attribute already exists in the vertex and is having
-    * its size increased.  
-    */
-   _tnl_copy_to_current( ctx );
-
-
-   /* Heuristic: Attempt to isolate attributes received outside
-    * begin/end so that they don't bloat the vertices.
-    */
-   if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END &&
-       tnl->vtx.attrsz[attr] == 0 && 
-       lastcount > 8 &&
-       tnl->vtx.vertex_size) {
-      reset_attrfv( tnl );
-   }
-
-   /* Fix up sizes:
-    */
-   oldsz = tnl->vtx.attrsz[attr];
-   tnl->vtx.attrsz[attr] = newsz;
-
-   tnl->vtx.vertex_size += newsz - oldsz;
-   tnl->vtx.counter = MIN2( VERT_BUFFER_SIZE / tnl->vtx.vertex_size,
-			    ctx->Const.MaxArrayLockSize );
-   tnl->vtx.initial_counter = tnl->vtx.counter;
-   tnl->vtx.vbptr = tnl->vtx.buffer;
-
-
-   /* Recalculate all the attrptr[] values
-    */
-   for (i = 0, tmp = tnl->vtx.vertex ; i < _TNL_ATTRIB_MAX ; i++) {
-      if (tnl->vtx.attrsz[i]) {
-	 tnl->vtx.attrptr[i] = tmp;
-	 tmp += tnl->vtx.attrsz[i];
-      }
-      else 
-	 tnl->vtx.attrptr[i] = NULL; /* will not be dereferenced */
-   }
-
-   /* Copy from current to repopulate the vertex with correct values.
-    */
-   _tnl_copy_from_current( ctx );
-
-   /* Replay stored vertices to translate them
-    * to new format here.
-    *
-    * -- No need to replay - just copy piecewise
-    */
-   if (tnl->vtx.copied.nr)
-   {
-      const GLfloat *data = tnl->vtx.copied.buffer;
-      GLfloat *dest = tnl->vtx.buffer;
-      GLuint j;
-
-      for (i = 0 ; i < tnl->vtx.copied.nr ; i++) {
-	 for (j = 0 ; j < _TNL_ATTRIB_MAX ; j++) {
-	    if (tnl->vtx.attrsz[j]) {
-	       if (j == attr) {
-		  if (oldsz) {
-		     COPY_CLEAN_4V( dest, oldsz, data );
-		     data += oldsz;
-		     dest += newsz;
-		  } else {
-		     COPY_SZ_4V( dest, newsz, tnl->vtx.current[j] );
-		     dest += newsz;
-		  }
-	       }
-	       else {
-		  GLuint sz = tnl->vtx.attrsz[j];
-		  COPY_SZ_4V( dest, sz, data );
-		  dest += sz;
-		  data += sz;
-	       }
-	    }
-	 }
-      }
-
-      tnl->vtx.vbptr = dest;
-      tnl->vtx.counter -= tnl->vtx.copied.nr;
-      tnl->vtx.copied.nr = 0;
-   }
-
-   /* For codegen - attrptr's may have changed, so need to redo
-    * codegen.  Might be a reasonable place to try & detect attributes
-    * in the vertex which aren't being submitted any more.
-    */
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++) 
-      if (tnl->vtx.attrsz[i]) {
-	 GLuint j = tnl->vtx.attrsz[i] - 1;
-
-	 if (i < _TNL_MAX_ATTR_CODEGEN)
-	    tnl->vtx.tabfv[i][j] = choose[i][j];
-      }
-
-}
-
-
-static void _tnl_fixup_vertex( GLcontext *ctx, GLuint attr, GLuint sz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   static const GLfloat id[4] = { 0, 0, 0, 1 };
-   int i;
-
-   if (0) 
-      _mesa_printf("%s attr %d sz %d -> %d\n",
-		   __FUNCTION__, attr, tnl->vtx.attrsz[attr], sz);
-
-   if (tnl->vtx.attrsz[attr] < sz) {
-      /* New size is larger.  Need to flush existing vertices and get
-       * an enlarged vertex format.
-       */
-      _tnl_wrap_upgrade_vertex( ctx, attr, sz );
-   }
-   else if (sz < tnl->vtx.active_sz[attr]) {
-      /* New size is smaller - just need to fill in some
-       * zeros.  Don't need to flush or wrap.
-       */
-      for (i = sz ; i <= tnl->vtx.attrsz[attr] ; i++)
-	 tnl->vtx.attrptr[attr][i-1] = id[i-1];
-   }
-
-   tnl->vtx.active_sz[attr] = sz;
-
-   /* Does setting NeedFlush belong here?  Necessitates resetting
-    * vtxfmt on each flush (otherwise flags won't get reset
-    * afterwards).
-    */
-   if (attr == 0) 
-      ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;
-   else 
-      ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-}
-
-#ifdef USE_X86_ASM
-
-static struct _tnl_dynfn *lookup( struct _tnl_dynfn *l, GLuint key )
-{
-   struct _tnl_dynfn *f;
-
-   foreach( f, l ) {
-      if (f->key == key) 
-	 return f;
-   }
-
-   return NULL;
-}
-
-
-static tnl_attrfv_func do_codegen( GLcontext *ctx, GLuint attr, GLuint sz )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   struct _tnl_dynfn *dfn = NULL;
-
-   if (attr == 0) {
-      GLuint key = tnl->vtx.vertex_size;
-
-      dfn = lookup( &tnl->vtx.cache.Vertex[sz-1], key );
-
-      if (!dfn)
-	 dfn = tnl->vtx.gen.Vertex[sz-1]( ctx, key );
-   }
-   else {
-      GLuint key = (GLuint) tnl->vtx.attrptr[attr];
-
-      dfn = lookup( &tnl->vtx.cache.Attribute[sz-1], key );
-
-      if (!dfn)
-	 dfn = tnl->vtx.gen.Attribute[sz-1]( ctx, key );
-   }
-
-   if (dfn) 
-      return *(tnl_attrfv_func *) &dfn->code;
-   else
-      return NULL;
-}
-
-#endif /* USE_X86_ASM */
-
-/* Helper function for 'CHOOSE' macro.  Do what's necessary when an
- * entrypoint is called for the first time.
- */
-
-static tnl_attrfv_func do_choose( GLuint attr, GLuint sz )
-{ 
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint oldsz = tnl->vtx.active_sz[attr];
-
-   assert(attr < _TNL_MAX_ATTR_CODEGEN);
-
-   if (oldsz != sz) {
-      /* Reset any active pointers for this attribute 
-       */
-      if (oldsz)
-	 tnl->vtx.tabfv[attr][oldsz-1] = choose[attr][oldsz-1];
-   
-      _tnl_fixup_vertex( ctx, attr, sz );
- 
-   }
-
-
-   /* Try to use codegen:
-    */
-#ifdef USE_X86_ASM
-   if (tnl->AllowCodegen)
-      tnl->vtx.tabfv[attr][sz-1] = do_codegen( ctx, attr, sz );
-   else
-#endif
-      tnl->vtx.tabfv[attr][sz-1] = NULL;
-
-   /* Else use generic version:
-    */
-   if (!tnl->vtx.tabfv[attr][sz-1])
-      tnl->vtx.tabfv[attr][sz-1] = generic_attr_func[attr][sz-1];
-
-   ASSERT(tnl->vtx.tabfv[attr][sz-1]);
-   return tnl->vtx.tabfv[attr][sz-1];
-}
-
-
-
-#define CHOOSE( ATTR, N )				\
-static void choose_##ATTR##_##N( const GLfloat *v )	\
-{							\
-   tnl_attrfv_func f = do_choose(ATTR, N);		\
-   ASSERT(f); \
-   f( v );						\
-}
-
-#define CHOOSERS( ATTRIB )	\
-   CHOOSE( ATTRIB, 1 )		\
-   CHOOSE( ATTRIB, 2 )		\
-   CHOOSE( ATTRIB, 3 )		\
-   CHOOSE( ATTRIB, 4 )		\
-
-
-#define INIT_CHOOSERS(ATTR)			\
-   ASSERT(ATTR <= _TNL_ATTRIB_ERROR);\
-   choose[ATTR][0] = choose_##ATTR##_1;		\
-   choose[ATTR][1] = choose_##ATTR##_2;		\
-   choose[ATTR][2] = choose_##ATTR##_3;		\
-   choose[ATTR][3] = choose_##ATTR##_4;
-
-/* conventional attributes */
-CHOOSERS( 0 )
-CHOOSERS( 1 )
-CHOOSERS( 2 )
-CHOOSERS( 3 )
-CHOOSERS( 4 )
-CHOOSERS( 5 )
-CHOOSERS( 6 )
-CHOOSERS( 7 )
-CHOOSERS( 8 )
-CHOOSERS( 9 )
-CHOOSERS( 10 )
-CHOOSERS( 11 )
-CHOOSERS( 12 )
-CHOOSERS( 13 )
-CHOOSERS( 14 )
-CHOOSERS( 15 )
-
-/* generic attributes */
-CHOOSERS( 16 )
-CHOOSERS( 17 )
-CHOOSERS( 18 )
-CHOOSERS( 19 )
-CHOOSERS( 20 )
-CHOOSERS( 21 )
-CHOOSERS( 22 )
-CHOOSERS( 23 )
-CHOOSERS( 24 )
-CHOOSERS( 25 )
-CHOOSERS( 26 )
-CHOOSERS( 27 )
-CHOOSERS( 28 )
-CHOOSERS( 29 )
-CHOOSERS( 30 )
-CHOOSERS( 31 )
-
-
-/**
- * This function will get called when glVertexAttribNV/ARB() is called
- * with an invalid index parameter.
- */
-static void
-error_attrib(const GLfloat *unused)
-{
-   GET_CURRENT_CONTEXT( ctx );
-   (void) unused;
-   _mesa_error( ctx, GL_INVALID_VALUE, "glVertexAttrib(index)" );
-}   
-
-
-
-/**
- * Reset all the per-vertex functions pointers to point to the default
- * "chooser" functions.
- */
-static void
-reset_attrfv(TNLcontext *tnl)
-{   
-   GLuint i;
-
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++) 
-      if (tnl->vtx.attrsz[i]) {
-	 GLint j = tnl->vtx.attrsz[i] - 1;
-	 tnl->vtx.attrsz[i] = 0;
-	 tnl->vtx.active_sz[i] = 0;
-
-	 if (i < _TNL_MAX_ATTR_CODEGEN) {
-            while (j >= 0) {
-	       tnl->vtx.tabfv[i][j] = choose[i][j];
-               j--;
-            }
-         }
-      }
-
-   tnl->vtx.vertex_size = 0;
-   tnl->vtx.have_materials = 0;
-}
-      
-
-
-/**
- * Materials:  
- * 
- * These are treated as per-vertex attributes, at indices above where
- * the NV_vertex_program leaves off.  There are a lot of good things
- * about treating materials this way.  
- *
- * However: I don't want to double the number of generated functions
- * just to cope with this, so I unroll the 'C' varients of CHOOSE and
- * ATTRF into this function, and dispense with codegen and
- * second-level dispatch.
- *
- * There is no aliasing of material attributes with other entrypoints.
- */
-#define OTHER_ATTR( A, N, params )			\
-do {							\
-   if (tnl->vtx.active_sz[A] != N) {			\
-      _tnl_fixup_vertex( ctx, A, N );			\
-   }							\
-							\
-   {							\
-      GLfloat *dest = tnl->vtx.attrptr[A];		\
-      if (N>0) dest[0] = (params)[0];			\
-      if (N>1) dest[1] = (params)[1];			\
-      if (N>2) dest[2] = (params)[2];			\
-      if (N>3) dest[3] = (params)[3];			\
-   }							\
-} while (0)
-
-
-#define MAT( ATTR, N, face, params )			\
-do {							\
-   if (face != GL_BACK)					\
-      OTHER_ATTR( ATTR, N, params ); /* front */	\
-   if (face != GL_FRONT)				\
-      OTHER_ATTR( ATTR + 1, N, params ); /* back */	\
-} while (0)
-
-
-/**
- * Called by glMaterialfv().
- * Colormaterial is dealt with later on.
- */
-static void GLAPIENTRY
-_tnl_Materialfv( GLenum face, GLenum pname, const GLfloat *params )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   switch (face) {
-   case GL_FRONT:
-   case GL_BACK:
-   case GL_FRONT_AND_BACK:
-      break;
-   default:
-      _mesa_error( ctx, GL_INVALID_ENUM, "glMaterialfv" );
-      return;
-   }
-
-   switch (pname) {
-   case GL_EMISSION:
-      MAT( _TNL_ATTRIB_MAT_FRONT_EMISSION, 4, face, params );
-      break;
-   case GL_AMBIENT:
-      MAT( _TNL_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      break;
-   case GL_DIFFUSE:
-      MAT( _TNL_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
-      break;
-   case GL_SPECULAR:
-      MAT( _TNL_ATTRIB_MAT_FRONT_SPECULAR, 4, face, params );
-      break;
-   case GL_SHININESS:
-      MAT( _TNL_ATTRIB_MAT_FRONT_SHININESS, 1, face, params );
-      break;
-   case GL_COLOR_INDEXES:
-      MAT( _TNL_ATTRIB_MAT_FRONT_INDEXES, 3, face, params );
-      break;
-   case GL_AMBIENT_AND_DIFFUSE:
-      MAT( _TNL_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      MAT( _TNL_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
-      break;
-   default:
-      _mesa_error( ctx, GL_INVALID_ENUM, "glMaterialfv" );
-      return;
-   }
-
-   tnl->vtx.have_materials = GL_TRUE;
-}
-
-
-static void GLAPIENTRY _tnl_EdgeFlag( GLboolean b )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLfloat f = (GLfloat)b;
-
-   OTHER_ATTR( _TNL_ATTRIB_EDGEFLAG, 1, &f );
-}
-
-
-/* Eval
- */
-static void GLAPIENTRY _tnl_EvalCoord1f( GLfloat u )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   /* TODO: use a CHOOSE() function for this: */
-   {
-      GLint i;
-      if (tnl->vtx.eval.new_state) 
-	 _tnl_update_eval( ctx );
-
-      for (i = 0; i < _TNL_NUM_EVAL; i++) {
-	 if (tnl->vtx.eval.map1[i].map) 
-	    if (tnl->vtx.active_sz[i] != tnl->vtx.eval.map1[i].sz)
-	       _tnl_fixup_vertex( ctx, i, tnl->vtx.eval.map1[i].sz );
-      }
-   }
-
-
-   _mesa_memcpy( tnl->vtx.copied.buffer, tnl->vtx.vertex, 
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-
-   _tnl_do_EvalCoord1f( ctx, u );
-
-   _mesa_memcpy( tnl->vtx.vertex, tnl->vtx.copied.buffer,
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-}
-
-static void GLAPIENTRY _tnl_EvalCoord2f( GLfloat u, GLfloat v )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   /* TODO: use a CHOOSE() function for this: */
-   {
-      GLint i;
-      if (tnl->vtx.eval.new_state) 
-	 _tnl_update_eval( ctx );
-
-      for (i = 0; i < _TNL_NUM_EVAL; i++) {
-	 if (tnl->vtx.eval.map2[i].map) 
-	    if (tnl->vtx.active_sz[i] != tnl->vtx.eval.map2[i].sz)
-	       _tnl_fixup_vertex( ctx, i, tnl->vtx.eval.map2[i].sz );
-      }
-
-      if (ctx->Eval.AutoNormal) 
-	 if (tnl->vtx.active_sz[_TNL_ATTRIB_NORMAL] != 3)
-	    _tnl_fixup_vertex( ctx, _TNL_ATTRIB_NORMAL, 3 );
-   }
-
-   _mesa_memcpy( tnl->vtx.copied.buffer, tnl->vtx.vertex, 
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-
-   _tnl_do_EvalCoord2f( ctx, u, v );
-
-   _mesa_memcpy( tnl->vtx.vertex, tnl->vtx.copied.buffer, 
-                 tnl->vtx.vertex_size * sizeof(GLfloat));
-}
-
-static void GLAPIENTRY _tnl_EvalCoord1fv( const GLfloat *u )
-{
-   _tnl_EvalCoord1f( u[0] );
-}
-
-static void GLAPIENTRY _tnl_EvalCoord2fv( const GLfloat *u )
-{
-   _tnl_EvalCoord2f( u[0], u[1] );
-}
-
-static void GLAPIENTRY _tnl_EvalPoint1( GLint i )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   GLfloat du = ((ctx->Eval.MapGrid1u2 - ctx->Eval.MapGrid1u1) /
-		 (GLfloat) ctx->Eval.MapGrid1un);
-   GLfloat u = i * du + ctx->Eval.MapGrid1u1;
-
-   _tnl_EvalCoord1f( u );
-}
-
-
-static void GLAPIENTRY _tnl_EvalPoint2( GLint i, GLint j )
-{
-   GET_CURRENT_CONTEXT( ctx );
-   GLfloat du = ((ctx->Eval.MapGrid2u2 - ctx->Eval.MapGrid2u1) / 
-		 (GLfloat) ctx->Eval.MapGrid2un);
-   GLfloat dv = ((ctx->Eval.MapGrid2v2 - ctx->Eval.MapGrid2v1) / 
-		 (GLfloat) ctx->Eval.MapGrid2vn);
-   GLfloat u = i * du + ctx->Eval.MapGrid2u1;
-   GLfloat v = j * dv + ctx->Eval.MapGrid2v1;
-
-   _tnl_EvalCoord2f( u, v );
-}
-
-
-/**
- * Called from glBegin.
- * ctx->Driver.CurrentExecPrimitive will be set to <mode>.
- */
-static void GLAPIENTRY _tnl_Begin( GLenum mode )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-
-   if (mode > GL_POLYGON) {
-      _mesa_error(ctx, GL_INVALID_ENUM, "glBegin(mode)");
-      return;
-   }
-
-   if (ctx->Driver.CurrentExecPrimitive == PRIM_OUTSIDE_BEGIN_END) {
-      /* we're not inside a glBegin/End pair */
-      TNLcontext *tnl = TNL_CONTEXT(ctx); 
-      GLuint i;
-
-      if (ctx->NewState) {
-	 _mesa_update_state( ctx );
-
-         if ((ctx->VertexProgram.Enabled && !ctx->VertexProgram._Enabled) ||
-            (ctx->FragmentProgram.Enabled && !ctx->FragmentProgram._Enabled)) {
-            _mesa_error(ctx, GL_INVALID_OPERATION,
-                        "glBegin (invalid vertex/fragment program)");
-            tnl->DiscardPrimitive = GL_TRUE;
-            return;
-         }
-
-         if (ctx->DrawBuffer->_Status != GL_FRAMEBUFFER_COMPLETE_EXT) {
-            _mesa_error(ctx, GL_INVALID_FRAMEBUFFER_OPERATION_EXT,
-                        "glBegin(incomplete framebuffer)");
-            tnl->DiscardPrimitive = GL_TRUE;
-            return;
-         }
-
-         tnl->DiscardPrimitive = GL_FALSE;
-
-	 if (!(tnl->Driver.NotifyBegin && 
-	       tnl->Driver.NotifyBegin( ctx, mode )))
-	     CALL_Begin(ctx->Exec, (mode));
-	 return;
-      }
-
-      /* Heuristic: attempt to isolate attributes occuring outside
-       * begin/end pairs.
-       */
-      if (tnl->vtx.vertex_size && !tnl->vtx.attrsz[0]) 
-	 _tnl_FlushVertices( ctx, ~0 );
-
-      i = tnl->vtx.prim_count++;
-      tnl->vtx.prim[i].mode = mode | PRIM_BEGIN;
-      tnl->vtx.prim[i].start = tnl->vtx.initial_counter - tnl->vtx.counter;
-      tnl->vtx.prim[i].count = 0;
-
-      ctx->Driver.CurrentExecPrimitive = mode;
-   }
-   else {
-      /* already inside glBegin/End */
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glBegin" );
-   }      
-}
-
-
-/**
- * Called from glEnd.
- */
-static void GLAPIENTRY _tnl_End( void )
-{
-   GET_CURRENT_CONTEXT( ctx ); 
-
-   if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-      /* closing an open glBegin primitive */
-      TNLcontext *tnl = TNL_CONTEXT(ctx); 
-      int idx = tnl->vtx.initial_counter - tnl->vtx.counter;
-      int i = tnl->vtx.prim_count - 1;
-
-      tnl->vtx.prim[i].mode |= PRIM_END; 
-      tnl->vtx.prim[i].count = idx - tnl->vtx.prim[i].start;
-
-      ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-
-      /* Two choices which effect the way vertex attributes are
-       * carried over (or not) between adjacent primitives.
-       */
-#if 0
-      if (tnl->vtx.prim_count == TNL_MAX_PRIM) 
-	 _tnl_FlushVertices( ctx, ~0 );
-#else
-      if (tnl->vtx.prim_count == TNL_MAX_PRIM)
-	 _tnl_flush_vtx( ctx );	
-#endif
-
-   }
-   else {
-      /* glBegin hasn't been called! */
-      _mesa_error( ctx, GL_INVALID_OPERATION, "glEnd" );
-   }
-}
-
-
-/**
- * XXX why aren't all members initialized here??
- */
-static void _tnl_exec_vtxfmt_init( GLcontext *ctx )
-{
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-
-   vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = _tnl_Begin;
-   vfmt->CallList = _mesa_CallList;
-   vfmt->CallLists = _mesa_CallLists;
-   vfmt->EdgeFlag = _tnl_EdgeFlag;
-   vfmt->End = _tnl_End;
-   vfmt->EvalCoord1f = _tnl_EvalCoord1f;
-   vfmt->EvalCoord1fv = _tnl_EvalCoord1fv;
-   vfmt->EvalCoord2f = _tnl_EvalCoord2f;
-   vfmt->EvalCoord2fv = _tnl_EvalCoord2fv;
-   vfmt->EvalPoint1 = _tnl_EvalPoint1;
-   vfmt->EvalPoint2 = _tnl_EvalPoint2;
-   vfmt->Materialfv = _tnl_Materialfv;
-
-   vfmt->Rectf = _mesa_noop_Rectf;
-   vfmt->EvalMesh1 = _mesa_noop_EvalMesh1;
-   vfmt->EvalMesh2 = _mesa_noop_EvalMesh2;
-}
-
-
-
-void _tnl_FlushVertices( GLcontext *ctx, GLuint flags )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   (void) flags;
-
-   if (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END) {
-      /* still inside a glBegin/End pair.  How'd we get here??? */
-      return;
-   }
-
-   if (tnl->DiscardPrimitive) {
-      /* discard any primitives */
-      tnl->vtx.prim_count = 0;
-      tnl->vtx.counter = tnl->vtx.initial_counter;
-      tnl->vtx.vbptr = tnl->vtx.buffer;
-   }
-
-   if (tnl->vtx.counter != tnl->vtx.initial_counter) {
-      _tnl_flush_vtx( ctx );
-   }
-
-   if (tnl->vtx.vertex_size) {
-      _tnl_copy_to_current( ctx );
-      reset_attrfv( tnl );
-   }
-
-   ctx->Driver.NeedFlush = 0;
-}
-
-
-/**
- * Init the tnl->vtx->current[] pointers to point to the corresponding
- * fields in ctx->Current attribute group.
- */
-static void _tnl_current_init( GLcontext *ctx ) 
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLint i;
-
-   /* setup the pointers for the typical (32) vertex attributes */
-   for (i = 0; i < VERT_ATTRIB_MAX; i++) 
-      tnl->vtx.current[i] = ctx->Current.Attrib[i];
-
-   /* setup pointers for the 12 material attributes */
-   for (i = 0; i < MAT_ATTRIB_MAX; i++)
-      tnl->vtx.current[_TNL_ATTRIB_MAT_FRONT_AMBIENT + i] = 
-	 ctx->Light.Material.Attrib[i];
-
-   /* special case */
-   tnl->vtx.current[_TNL_ATTRIB_EDGEFLAG] = &tnl->vtx.CurrentFloatEdgeFlag;
-}
-
-static struct _tnl_dynfn *no_codegen( GLcontext *ctx, int key )
-{
-   (void) ctx; (void) key;
-   return NULL;
-}
-
-void _tnl_vtx_init( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   struct tnl_vertex_arrays *tmp = &tnl->vtx_inputs;
-   GLuint i;
-   static int firsttime = 1;
-   
-   if (firsttime) {
-      firsttime = 0;
-
-      /* conventional attributes */
-      INIT_CHOOSERS( 0 );
-      INIT_CHOOSERS( 1 );
-      INIT_CHOOSERS( 2 );
-      INIT_CHOOSERS( 3 );
-      INIT_CHOOSERS( 4 );
-      INIT_CHOOSERS( 5 );
-      INIT_CHOOSERS( 6 );
-      INIT_CHOOSERS( 7 );
-      INIT_CHOOSERS( 8 );
-      INIT_CHOOSERS( 9 );
-      INIT_CHOOSERS( 10 );
-      INIT_CHOOSERS( 11 );
-      INIT_CHOOSERS( 12 );
-      INIT_CHOOSERS( 13 );
-      INIT_CHOOSERS( 14 );
-      INIT_CHOOSERS( 15 );
-
-      /* generic attributes */
-      INIT_CHOOSERS( 16 );
-      INIT_CHOOSERS( 17 );
-      INIT_CHOOSERS( 18 );
-      INIT_CHOOSERS( 19 );
-      INIT_CHOOSERS( 20 );
-      INIT_CHOOSERS( 21 );
-      INIT_CHOOSERS( 22 );
-      INIT_CHOOSERS( 23 );
-      INIT_CHOOSERS( 24 );
-      INIT_CHOOSERS( 25 );
-      INIT_CHOOSERS( 26 );
-      INIT_CHOOSERS( 27 );
-      INIT_CHOOSERS( 28 );
-      INIT_CHOOSERS( 29 );
-      INIT_CHOOSERS( 30 );
-      INIT_CHOOSERS( 31 );
-
-      choose[_TNL_ATTRIB_ERROR][0] = error_attrib;
-      choose[_TNL_ATTRIB_ERROR][1] = error_attrib;
-      choose[_TNL_ATTRIB_ERROR][2] = error_attrib;
-      choose[_TNL_ATTRIB_ERROR][3] = error_attrib;
-
-#ifdef USE_X86_ASM
-      if (tnl->AllowCodegen) {
-         _tnl_x86choosers(choose, do_choose); /* x86 INIT_CHOOSERS */
-      }
-#endif
-
-      _tnl_generic_attr_table_init( generic_attr_func );
-   }
-
-   for (i = 0; i < _TNL_ATTRIB_EDGEFLAG; i++)
-      _mesa_vector4f_init( &tmp->Attribs[i], 0, NULL);
-
-   for (i = 0; i < 4; i++) {
-      make_empty_list( &tnl->vtx.cache.Vertex[i] );
-      make_empty_list( &tnl->vtx.cache.Attribute[i] );
-      tnl->vtx.gen.Vertex[i] = no_codegen;
-      tnl->vtx.gen.Attribute[i] = no_codegen;
-   }
-
-#ifdef USE_X86_ASM
-   _tnl_InitX86Codegen( &tnl->vtx.gen );
-#endif
-
-   _tnl_current_init( ctx );
-   _tnl_exec_vtxfmt_init( ctx );
-   _tnl_generic_exec_vtxfmt_init( ctx );
-#ifdef USE_X86_ASM
-   if (tnl->AllowCodegen) {
-      _tnl_x86_exec_vtxfmt_init( ctx ); /* x86 DISPATCH_ATTRFV */
-   }
-#endif
-
-   _mesa_install_exec_vtxfmt( ctx, &tnl->exec_vtxfmt );
-
-   _mesa_memcpy( tnl->vtx.tabfv, choose, sizeof(choose) );
-
-   for (i = 0 ; i < _TNL_ATTRIB_MAX ; i++) {
-      tnl->vtx.attrsz[i] = 0;
-      tnl->vtx.active_sz[i] = 0;
-   }
-
-   tnl->vtx.vertex_size = 0;
-   tnl->vtx.have_materials = 0;
-}
-
-static void free_funcs( struct _tnl_dynfn *l )
-{
-   struct _tnl_dynfn *f, *tmp;
-   foreach_s (f, tmp, l) {
-      remove_from_list( f );
-      ALIGN_FREE( f->code );
-      FREE( f );
-   }
-}
-
-
-void _tnl_vtx_destroy( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 
-   GLuint i;
-
-   for (i = 0; i < 4; i++) {
-      free_funcs( &tnl->vtx.cache.Vertex[i] );
-      free_funcs( &tnl->vtx.cache.Attribute[i] ); 
-   }
-}
-
diff --git a/src/mesa/tnl/t_vtx_api.h b/src/mesa/tnl/t_vtx_api.h
deleted file mode 100644
index 5311f40a852..00000000000
--- a/src/mesa/tnl/t_vtx_api.h
+++ /dev/null
@@ -1,86 +0,0 @@
-/**************************************************************************
-
-Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
-TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *
- */
-
-#ifndef __T_VTX_API_H__
-#define __T_VTX_API_H__
-
-#include "t_context.h"
-
-
-/* t_vtx_api.c:
- */
-extern void _tnl_vtx_init( GLcontext *ctx );
-extern void _tnl_vtx_destroy( GLcontext *ctx );
-
-extern void _tnl_FlushVertices( GLcontext *ctx, GLuint flags );
-extern void _tnl_flush_vtx( GLcontext *ctx );
-
-extern void GLAPIENTRY _tnl_wrap_filled_vertex( GLcontext *ctx );
-
-/* t_vtx_exec.c:
- */
-
-extern void _tnl_do_EvalCoord2f( GLcontext* ctx, GLfloat u, GLfloat v );
-extern void _tnl_do_EvalCoord1f(GLcontext* ctx, GLfloat u);
-extern void _tnl_update_eval( GLcontext *ctx );
-
-extern GLboolean *_tnl_translate_edgeflag( GLcontext *ctx, 
-					   const GLfloat *data, 
-					   GLuint count,
-					   GLuint stride );
-
-extern GLboolean *_tnl_import_current_edgeflag( GLcontext *ctx,
-						GLuint count );
-
-
-
-/* t_vtx_generic.c:
- */
-extern void _tnl_generic_exec_vtxfmt_init( GLcontext *ctx );
-
-extern void _tnl_generic_attr_table_init( tnl_attrfv_func (*tab)[4] );
-
-/* t_vtx_x86.c:
- */
-extern void _tnl_InitX86Codegen( struct _tnl_dynfn_generators *gen );
-
-extern void _tnl_x86_exec_vtxfmt_init( GLcontext *ctx );
-
-extern void _tnl_x86choosers( tnl_attrfv_func (*choose)[4],
-			      tnl_attrfv_func (*do_choose)( GLuint attr,
-							GLuint sz ));
-
-
-
-
-#endif
diff --git a/src/mesa/tnl/t_vtx_eval.c b/src/mesa/tnl/t_vtx_eval.c
deleted file mode 100644
index bb621ecbe39..00000000000
--- a/src/mesa/tnl/t_vtx_eval.c
+++ /dev/null
@@ -1,264 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  6.5.1
- *
- * Copyright (C) 1999-2006  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"),
- * 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 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
- * BRIAN PAUL 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:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "api_eval.h"
-#include "context.h"
-#include "macros.h"
-#include "math/m_eval.h"
-#include "t_vtx_api.h"
-#include "dispatch.h"
-
-
-static void clear_active_eval1( TNLcontext *tnl, GLuint attr ) 
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   tnl->vtx.eval.map1[attr].map = NULL;
-}
-
-static void clear_active_eval2( TNLcontext *tnl, GLuint attr ) 
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   tnl->vtx.eval.map2[attr].map = NULL;
-}
-
-static void set_active_eval1( TNLcontext *tnl, GLuint attr, GLuint dim, 
-			      struct gl_1d_map *map )
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   if (!tnl->vtx.eval.map1[attr].map) {
-      tnl->vtx.eval.map1[attr].map = map;
-      tnl->vtx.eval.map1[attr].sz = dim;
-   }
-} 
-
-static void set_active_eval2( TNLcontext *tnl, GLuint attr, GLuint dim, 
-			      struct gl_2d_map *map )
-{
-   ASSERT(attr < _TNL_NUM_EVAL);
-   if (!tnl->vtx.eval.map2[attr].map) {
-      tnl->vtx.eval.map2[attr].map = map;
-      tnl->vtx.eval.map2[attr].sz = dim;
-   }
-} 
-
-void _tnl_update_eval( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint attr;
-
-   /* Vertex program maps have priority over conventional attribs */
-
-   for (attr = 0; attr < _TNL_NUM_EVAL; attr++) {
-      clear_active_eval1( tnl, attr );
-      clear_active_eval2( tnl, attr );
-   }
-
-   if (ctx->Eval.Map1Color4) 
-      set_active_eval1( tnl, VERT_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map1Color4 );
-      
-   if (ctx->Eval.Map2Color4) 
-      set_active_eval2( tnl, VERT_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map2Color4 );
-
-   if (ctx->Eval.Map1TextureCoord4) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 4, &ctx->EvalMap.Map1Texture4 );
-   else if (ctx->Eval.Map1TextureCoord3) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 3, &ctx->EvalMap.Map1Texture3 );
-   else if (ctx->Eval.Map1TextureCoord2) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 2, &ctx->EvalMap.Map1Texture2 );
-   else if (ctx->Eval.Map1TextureCoord1) 
-      set_active_eval1( tnl, VERT_ATTRIB_TEX0, 1, &ctx->EvalMap.Map1Texture1 );
-
-   if (ctx->Eval.Map2TextureCoord4) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 4, &ctx->EvalMap.Map2Texture4 );
-   else if (ctx->Eval.Map2TextureCoord3) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 3, &ctx->EvalMap.Map2Texture3 );
-   else if (ctx->Eval.Map2TextureCoord2) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 2, &ctx->EvalMap.Map2Texture2 );
-   else if (ctx->Eval.Map2TextureCoord1) 
-      set_active_eval2( tnl, VERT_ATTRIB_TEX0, 1, &ctx->EvalMap.Map2Texture1 );
-
-   if (ctx->Eval.Map1Normal) 
-      set_active_eval1( tnl, VERT_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map1Normal );
-
-   if (ctx->Eval.Map2Normal) 
-      set_active_eval2( tnl, VERT_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map2Normal );
-
-   if (ctx->Eval.Map1Vertex4) 
-      set_active_eval1( tnl, VERT_ATTRIB_POS, 4, &ctx->EvalMap.Map1Vertex4 );
-   else if (ctx->Eval.Map1Vertex3) 
-      set_active_eval1( tnl, VERT_ATTRIB_POS, 3, &ctx->EvalMap.Map1Vertex3 );
-
-   if (ctx->Eval.Map2Vertex4) 
-      set_active_eval2( tnl, VERT_ATTRIB_POS, 4, &ctx->EvalMap.Map2Vertex4 );
-   else if (ctx->Eval.Map2Vertex3) 
-      set_active_eval2( tnl, VERT_ATTRIB_POS, 3, &ctx->EvalMap.Map2Vertex3 );
-
-   /* Evaluators with generic attributes is only supported for NV vertex
-    * programs, not ARB vertex programs.  16 evaluator maps are supported.
-    * We do this after the conventional attributes since the spec says that
-    * these generic maps have higher priority.
-    */
-   if (ctx->VertexProgram._Current &&
-       ctx->VertexProgram._Current->IsNVProgram) {
-      for (attr = 0; attr < 16; attr++) {
-	 if (ctx->Eval.Map1Attrib[attr]) 
-	    set_active_eval1( tnl, attr, 4, &ctx->EvalMap.Map1Attrib[attr] );
-
-	 if (ctx->Eval.Map2Attrib[attr]) 
-	    set_active_eval2( tnl, attr, 4, &ctx->EvalMap.Map2Attrib[attr] );
-      }
-   }
-
-   tnl->vtx.eval.new_state = 0;
-}
-
-
-
-void _tnl_do_EvalCoord1f(GLcontext* ctx, GLfloat u)
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint attr;
-
-   for (attr = 1; attr < _TNL_NUM_EVAL; attr++) {
-      struct gl_1d_map *map = tnl->vtx.eval.map1[attr].map;
-      if (map) {
-	 GLfloat uu = (u - map->u1) * map->du;
-	 GLfloat data[4];
-
-	 ASSIGN_4V(data, 0, 0, 0, 1);
-
-	 _math_horner_bezier_curve(map->Points, data, uu, 
-				   tnl->vtx.eval.map1[attr].sz, 
-				   map->Order);
-
-	 COPY_SZ_4V( tnl->vtx.attrptr[attr],
-		     tnl->vtx.attrsz[attr],
-		     data );
-      }
-   }
-
-   /** Vertex -- EvalCoord1f is a noop if this map not enabled:
-    **/
-   if (tnl->vtx.eval.map1[0].map) {
-      struct gl_1d_map *map = tnl->vtx.eval.map1[0].map;
-      GLfloat uu = (u - map->u1) * map->du;
-      GLfloat vertex[4];
-
-      ASSIGN_4V(vertex, 0, 0, 0, 1);
-
-      _math_horner_bezier_curve(map->Points, vertex, uu, 
-				tnl->vtx.eval.map1[0].sz, 
-				map->Order);
-
-      if (tnl->vtx.eval.map1[0].sz == 4) 
-	 CALL_Vertex4fv(GET_DISPATCH(), ( vertex ));
-      else
-	 CALL_Vertex3fv(GET_DISPATCH(), ( vertex )); 
-   }
-}
-
-
-
-void _tnl_do_EvalCoord2f( GLcontext* ctx, GLfloat u, GLfloat v )
-{   
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint attr;
-
-   for (attr = 1; attr < _TNL_NUM_EVAL; attr++) {
-      struct gl_2d_map *map = tnl->vtx.eval.map2[attr].map;
-      if (map) {
-	 GLfloat uu = (u - map->u1) * map->du;
-	 GLfloat vv = (v - map->v1) * map->dv;
-	 GLfloat data[4];
-
-	 ASSIGN_4V(data, 0, 0, 0, 1);
-
-	 _math_horner_bezier_surf(map->Points, 
-				  data, 
-				  uu, vv, 
-				  tnl->vtx.eval.map2[attr].sz, 
-				  map->Uorder, map->Vorder);
-
-	 COPY_SZ_4V( tnl->vtx.attrptr[attr],
-		     tnl->vtx.attrsz[attr],
-		     data );
-      }
-   }
-
-   /** Vertex -- EvalCoord2f is a noop if this map not enabled:
-    **/
-   if (tnl->vtx.eval.map2[0].map) {
-      struct gl_2d_map *map = tnl->vtx.eval.map2[0].map;
-      GLfloat uu = (u - map->u1) * map->du;
-      GLfloat vv = (v - map->v1) * map->dv;
-      GLfloat vertex[4];
-
-      ASSIGN_4V(vertex, 0, 0, 0, 1);
-
-      if (ctx->Eval.AutoNormal) {
-	 GLfloat normal[4];
-         GLfloat du[4], dv[4];
-
-         _math_de_casteljau_surf(map->Points, vertex, du, dv, uu, vv, 
-				 tnl->vtx.eval.map2[0].sz,
-				 map->Uorder, map->Vorder);
-
-	 if (tnl->vtx.eval.map2[0].sz == 4) {
-	    du[0] = du[0]*vertex[3] - du[3]*vertex[0];
-	    du[1] = du[1]*vertex[3] - du[3]*vertex[1];
-	    du[2] = du[2]*vertex[3] - du[3]*vertex[2];
-	 
-	    dv[0] = dv[0]*vertex[3] - dv[3]*vertex[0];
-	    dv[1] = dv[1]*vertex[3] - dv[3]*vertex[1];
-	    dv[2] = dv[2]*vertex[3] - dv[3]*vertex[2];
-	 }
-
-
-         CROSS3(normal, du, dv);
-         NORMALIZE_3FV(normal);
-	 normal[3] = 1.0;
-
-	 COPY_SZ_4V( tnl->vtx.attrptr[_TNL_ATTRIB_NORMAL],
-		     tnl->vtx.attrsz[_TNL_ATTRIB_NORMAL],
-		     normal );
-
-      }
-      else {
-         _math_horner_bezier_surf(map->Points, vertex, uu, vv, 
-				  tnl->vtx.eval.map2[0].sz,
-				  map->Uorder, map->Vorder);
-      }
-
-      if (tnl->vtx.attrsz[0] == 4) 
-	 CALL_Vertex4fv(GET_DISPATCH(), ( vertex ));
-      else
-	 CALL_Vertex3fv(GET_DISPATCH(), ( vertex )); 
-   }
-}
-
-
diff --git a/src/mesa/tnl/t_vtx_exec.c b/src/mesa/tnl/t_vtx_exec.c
deleted file mode 100644
index 900c4ab6cc1..00000000000
--- a/src/mesa/tnl/t_vtx_exec.c
+++ /dev/null
@@ -1,293 +0,0 @@
-/*
- * Mesa 3-D graphics library
- * Version:  5.1
- *
- * Copyright (C) 1999-2003  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"),
- * 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 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
- * BRIAN PAUL 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:
- *    Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "api_eval.h"
-#include "context.h"
-#include "enums.h"
-#include "state.h"
-#include "macros.h"
-#include "math/m_eval.h"
-#include "t_vtx_api.h"
-#include "t_pipeline.h"
-
-
-static void _tnl_print_vtx( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint count = tnl->vtx.initial_counter - tnl->vtx.counter;
-   GLuint i;
-
-   _mesa_debug(ctx, "_tnl_print_vtx: %u vertices %d primitives, %d vertsize\n",
-               count,
-	       tnl->vtx.prim_count,
-	       tnl->vtx.vertex_size);
-
-   for (i = 0 ; i < tnl->vtx.prim_count ; i++) {
-      struct tnl_prim *prim = &tnl->vtx.prim[i];
-      _mesa_debug(NULL, "   prim %d: %s %d..%d %s %s\n",
-		  i, 
-		  _mesa_lookup_enum_by_nr(prim->mode & PRIM_MODE_MASK),
-		  prim->start, 
-		  prim->start + prim->count,
-		  (prim->mode & PRIM_BEGIN) ? "BEGIN" : "(wrap)",
-		  (prim->mode & PRIM_END) ? "END" : "(wrap)");
-   }
-}
-
-GLboolean *_tnl_translate_edgeflag( GLcontext *ctx, const GLfloat *data,
-				    GLuint count, GLuint stride )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLboolean *ef = tnl->vtx.edgeflag_tmp;
-   GLuint i;
-
-   if (!ef) 
-      ef = tnl->vtx.edgeflag_tmp = (GLboolean *) MALLOC( tnl->vb.Size );
-   
-   for (i = 0 ; i < count ; i++, data += stride)
-      ef[i] = (data[0] == 1.0);
-
-   return ef;
-}
-
-
-GLboolean *_tnl_import_current_edgeflag( GLcontext *ctx,
-					 GLuint count )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLboolean *ef = tnl->vtx.edgeflag_tmp;
-   GLboolean tmp = ctx->Current.EdgeFlag;
-   GLuint i;
-
-   if (!ef) 
-      ef = tnl->vtx.edgeflag_tmp = (GLboolean *) MALLOC( tnl->vb.Size );
-
-   for (i = 0 ; i < count ; i++)
-      ef[i] = tmp;
-
-   return ef;
-}
-
-static INLINE GLint get_size( const GLfloat *f )
-{
-   if (f[3] != 1.0) return 4;
-   if (f[2] != 0.0) return 3;
-   return 2;
-}
-
-/* Some nasty stuff still hanging on here.  
- *
- * TODO - remove VB->NormalPtr, etc and just use the AttrPtr's.
- */
-static void _tnl_vb_bind_vtx( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   struct vertex_buffer *VB = &tnl->vb;
-   struct tnl_vertex_arrays *tmp = &tnl->vtx_inputs;
-   GLfloat *data = tnl->vtx.buffer;
-   GLuint count = tnl->vtx.initial_counter - tnl->vtx.counter;
-   GLuint attr, i;
-
-#undef DEBUG_VTX
-
-#ifdef DEBUG_VTX
-   fprintf(stderr, "_tnl_vb_bind_vtx(): %d verts %d vertsize\n",
-		  count, tnl->vtx.vertex_size);
-#endif
-
-
-   /* Setup constant data in the VB.
-    */
-   VB->Count = count;
-   VB->Primitive = tnl->vtx.prim;
-   VB->PrimitiveCount = tnl->vtx.prim_count;
-   VB->Elts = NULL;
-   VB->NormalLengthPtr = NULL;
-
-   for (attr = 0; attr <= _TNL_ATTRIB_EDGEFLAG ; attr++) {
-      if (tnl->vtx.attrsz[attr]) {
-	 tmp->Attribs[attr].count = count;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) data;
-	 tmp->Attribs[attr].start = data;
-	 tmp->Attribs[attr].size = tnl->vtx.attrsz[attr];
-	 tmp->Attribs[attr].stride = tnl->vtx.vertex_size * sizeof(GLfloat);
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-	 data += tnl->vtx.attrsz[attr];
-      }
-      else {
-/* 	 VB->AttribPtr[attr] = &tnl->current.Attribs[attr]; */
-
-
-	 tmp->Attribs[attr].count = 1;
-	 tmp->Attribs[attr].data = (GLfloat (*)[4]) tnl->vtx.current[attr];
-	 tmp->Attribs[attr].start = tnl->vtx.current[attr];
-	 tmp->Attribs[attr].size = get_size( tnl->vtx.current[attr] );
-	 tmp->Attribs[attr].stride = 0;
-	 VB->AttribPtr[attr] = &tmp->Attribs[attr];
-      }
-   }
-
-   
-   /* Copy and translate EdgeFlag to a contiguous array of GLbooleans
-    */
-   if (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL) {
-      if (tnl->vtx.attrsz[_TNL_ATTRIB_EDGEFLAG]) {
-	 VB->EdgeFlag = _tnl_translate_edgeflag( ctx, data, count,
-						 tnl->vtx.vertex_size );
-	 data++;
-      }
-      else 
-	 VB->EdgeFlag = _tnl_import_current_edgeflag( ctx, count );
-   }
-
-   /* Legacy pointers -- remove one day.
-    */
-   VB->ObjPtr = VB->AttribPtr[_TNL_ATTRIB_POS];
-   VB->NormalPtr = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
-   VB->ColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR0];
-   VB->ColorPtr[1] = NULL;
-   VB->SecondaryColorPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR1];
-   VB->SecondaryColorPtr[1] = NULL;
-   VB->IndexPtr[0] = VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX];
-   VB->IndexPtr[1] = NULL;
-   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
-
-   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
-      VB->TexCoordPtr[i] = VB->AttribPtr[_TNL_ATTRIB_TEX0 + i];
-   }
-}
-
-
-/*
- * NOTE: Need to have calculated primitives by this point -- do it on the fly.
- * NOTE: Old 'parity' issue is gone.
- */
-static GLuint _tnl_copy_vertices( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT( ctx );
-   GLuint nr = tnl->vtx.prim[tnl->vtx.prim_count-1].count;
-   GLuint ovf, i;
-   GLuint sz = tnl->vtx.vertex_size;
-   GLfloat *dst = tnl->vtx.copied.buffer;
-   GLfloat *src = (tnl->vtx.buffer + 
-		   tnl->vtx.prim[tnl->vtx.prim_count-1].start * 
-		   tnl->vtx.vertex_size);
-
-
-   switch( ctx->Driver.CurrentExecPrimitive )
-   {
-   case GL_POINTS:
-      return 0;
-   case GL_LINES:
-      ovf = nr&1;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_TRIANGLES:
-      ovf = nr%3;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_QUADS:
-      ovf = nr&3;
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case GL_LINE_STRIP:
-      if (nr == 0) 
-	 return 0;
-      else {
-	 _mesa_memcpy( dst, src+(nr-1)*sz, sz * sizeof(GLfloat) );
-	 return 1;
-      }
-   case GL_LINE_LOOP:
-   case GL_TRIANGLE_FAN:
-   case GL_POLYGON:
-      if (nr == 0) 
-	 return 0;
-      else if (nr == 1) {
-	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
-	 return 1;
-      } else {
-	 _mesa_memcpy( dst, src+0, sz * sizeof(GLfloat) );
-	 _mesa_memcpy( dst+sz, src+(nr-1)*sz, sz * sizeof(GLfloat) );
-	 return 2;
-      }
-   case GL_TRIANGLE_STRIP:
-      /* no parity issue, but need to make sure the tri is not drawn twice */
-      if (nr & 1) {
-	 tnl->vtx.prim[tnl->vtx.prim_count-1].count--;
-      }
-      /* fallthrough */
-   case GL_QUAD_STRIP:
-      switch (nr) {
-      case 0: ovf = 0; break;
-      case 1: ovf = 1; break;
-      default: ovf = 2 + (nr&1); break;
-      }
-      for (i = 0 ; i < ovf ; i++)
-	 _mesa_memcpy( dst+i*sz, src+(nr-ovf+i)*sz, sz * sizeof(GLfloat) );
-      return i;
-   case PRIM_OUTSIDE_BEGIN_END:
-      return 0;
-   default:
-      assert(0);
-      return 0;
-   }
-}
-
-
-/**
- * Execute the buffer and save copied verts.
- */
-void _tnl_flush_vtx( GLcontext *ctx )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   GLuint vertex_count = tnl->vtx.initial_counter - tnl->vtx.counter;
-
-   if (0)
-      _tnl_print_vtx( ctx );
-
-   if (tnl->vtx.prim_count && vertex_count) {
-
-      tnl->vtx.copied.nr = _tnl_copy_vertices( ctx ); 
-
-      if (tnl->vtx.copied.nr != vertex_count) {
-	 if (ctx->NewState)
-	    _mesa_update_state( ctx );
-      
-	 _tnl_vb_bind_vtx( ctx );
-
-	 tnl->Driver.RunPipeline( ctx );
-      }
-   }
-
-   tnl->vtx.prim_count = 0;
-   tnl->vtx.counter = tnl->vtx.initial_counter;
-   tnl->vtx.vbptr = tnl->vtx.buffer;
-}
diff --git a/src/mesa/tnl/t_vtx_generic.c b/src/mesa/tnl/t_vtx_generic.c
deleted file mode 100644
index 28de445ed41..00000000000
--- a/src/mesa/tnl/t_vtx_generic.c
+++ /dev/null
@@ -1,615 +0,0 @@
-/**************************************************************************
-
-Copyright 2004 Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- */
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "vtxfmt.h"
-#include "dlist.h"
-#include "state.h"
-#include "light.h"
-#include "api_arrayelt.h"
-#include "api_noop.h"
-#include "t_vtx_api.h"
-
-
-/* Versions of all the entrypoints for situations where codegen isn't
- * available.  
- *
- * Note: Only one size for each attribute may be active at once.
- * Eg. if Color3f is installed/active, then Color4f may not be, even
- * if the vertex actually contains 4 color coordinates.  This is
- * because the 3f version won't otherwise set color[3] to 1.0 -- this
- * is the job of the chooser function when switching between Color4f
- * and Color3f.
- */
-#define ATTRFV( ATTR, N )				\
-static void attrib_##ATTR##_##N( const GLfloat *v )	\
-{							\
-   GET_CURRENT_CONTEXT( ctx );				\
-   TNLcontext *tnl = TNL_CONTEXT(ctx);			\
-							\
-   if ((ATTR) == 0) {					\
-      GLuint i;						\
-							\
-      if (N>0) tnl->vtx.vbptr[0] = v[0];		\
-      if (N>1) tnl->vtx.vbptr[1] = v[1];		\
-      if (N>2) tnl->vtx.vbptr[2] = v[2];		\
-      if (N>3) tnl->vtx.vbptr[3] = v[3];		\
-							\
-      for (i = N; i < tnl->vtx.vertex_size; i++)	\
-	 tnl->vtx.vbptr[i] = tnl->vtx.vertex[i];	\
-							\
-      tnl->vtx.vbptr += tnl->vtx.vertex_size;		\
-							\
-      if (--tnl->vtx.counter == 0)			\
-	 _tnl_wrap_filled_vertex( ctx );		\
-   }							\
-   else {						\
-      GLfloat *dest = tnl->vtx.attrptr[ATTR];		\
-      if (N>0) dest[0] = v[0];				\
-      if (N>1) dest[1] = v[1];				\
-      if (N>2) dest[2] = v[2];				\
-      if (N>3) dest[3] = v[3];				\
-   }							\
-}
-
-#define INIT(TAB, ATTR)						\
-   TAB[ATTR][0] = attrib_##ATTR##_1;				\
-   TAB[ATTR][1] = attrib_##ATTR##_2;				\
-   TAB[ATTR][2] = attrib_##ATTR##_3;				\
-   TAB[ATTR][3] = attrib_##ATTR##_4;
-
-
-#define ATTRS( ATTRIB )				\
-   ATTRFV( ATTRIB, 1 )				\
-   ATTRFV( ATTRIB, 2 )				\
-   ATTRFV( ATTRIB, 3 )				\
-   ATTRFV( ATTRIB, 4 )			
-
-/* conventional attribs */
-ATTRS( 0 )
-ATTRS( 1 )
-ATTRS( 2 )
-ATTRS( 3 )
-ATTRS( 4 )
-ATTRS( 5 )
-ATTRS( 6 )
-ATTRS( 7 )
-ATTRS( 8 )
-ATTRS( 9 )
-ATTRS( 10 )
-ATTRS( 11 )
-ATTRS( 12 )
-ATTRS( 13 )
-ATTRS( 14 )
-ATTRS( 15 )
-
-/* generic attribs */
-ATTRS( 16 )
-ATTRS( 17 )
-ATTRS( 18 )
-ATTRS( 19 )
-ATTRS( 20 )
-ATTRS( 21 )
-ATTRS( 22 )
-ATTRS( 23 )
-ATTRS( 24 )
-ATTRS( 25 )
-ATTRS( 26 )
-ATTRS( 27 )
-ATTRS( 28 )
-ATTRS( 29 )
-ATTRS( 30 )
-ATTRS( 31 )
-
-
-void _tnl_generic_attr_table_init( tnl_attrfv_func (*tab)[4] )
-{
-   /* conventional attribs */
-   INIT( tab, 0 );
-   INIT( tab, 1 );
-   INIT( tab, 2 );
-   INIT( tab, 3 );
-   INIT( tab, 4 );
-   INIT( tab, 5 );
-   INIT( tab, 6 );
-   INIT( tab, 7 );
-   INIT( tab, 8 );
-   INIT( tab, 9 );
-   INIT( tab, 10 );
-   INIT( tab, 11 );
-   INIT( tab, 12 );
-   INIT( tab, 13 );
-   INIT( tab, 14 );
-   INIT( tab, 15 );
-
-   /* generic attribs */
-   INIT( tab, 16 );
-   INIT( tab, 17 );
-   INIT( tab, 18 );
-   INIT( tab, 19 );
-   INIT( tab, 20 );
-   INIT( tab, 21 );
-   INIT( tab, 22 );
-   INIT( tab, 23 );
-   INIT( tab, 24 );
-   INIT( tab, 25 );
-   INIT( tab, 26 );
-   INIT( tab, 27 );
-   INIT( tab, 28 );
-   INIT( tab, 29 );
-   INIT( tab, 30 );
-   INIT( tab, 31 );
-}
-
-/* These can be made efficient with codegen.  Further, by adding more
- * logic to do_choose(), the double-dispatch for legacy entrypoints
- * like glVertex3f() can be removed.
- */
-#define DISPATCH_ATTRFV( ATTR, COUNT, P )	\
-do {						\
-   GET_CURRENT_CONTEXT( ctx ); 			\
-   TNLcontext *tnl = TNL_CONTEXT(ctx); 		\
-   tnl->vtx.tabfv[ATTR][COUNT-1]( P );		\
-} while (0)
-
-#define DISPATCH_ATTR1FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 1, V )
-#define DISPATCH_ATTR2FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 2, V )
-#define DISPATCH_ATTR3FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 3, V )
-#define DISPATCH_ATTR4FV( ATTR, V ) DISPATCH_ATTRFV( ATTR, 4, V )
-
-#define DISPATCH_ATTR1F( ATTR, S ) DISPATCH_ATTRFV( ATTR, 1, &(S) )
-
-#if defined(USE_X86_ASM) && 0 /* will break register calling convention */
-/* Naughty cheat:
- */
-#define DISPATCH_ATTR2F( ATTR, S,T ) DISPATCH_ATTRFV( ATTR, 2, &(S) )
-#define DISPATCH_ATTR3F( ATTR, S,T,R ) DISPATCH_ATTRFV( ATTR, 3, &(S) )
-#define DISPATCH_ATTR4F( ATTR, S,T,R,Q ) DISPATCH_ATTRFV( ATTR, 4, &(S) )
-#else
-/* Safe:
- */
-#define DISPATCH_ATTR2F( ATTR, S,T ) 		\
-do { 						\
-   GLfloat v[2]; 				\
-   v[0] = S; v[1] = T;				\
-   DISPATCH_ATTR2FV( ATTR, v );			\
-} while (0)
-#define DISPATCH_ATTR3F( ATTR, S,T,R ) 		\
-do { 						\
-   GLfloat v[3]; 				\
-   v[0] = S; v[1] = T; v[2] = R;		\
-   DISPATCH_ATTR3FV( ATTR, v );			\
-} while (0)
-#define DISPATCH_ATTR4F( ATTR, S,T,R,Q )	\
-do { 						\
-   GLfloat v[4]; 				\
-   v[0] = S; v[1] = T; v[2] = R; v[3] = Q;	\
-   DISPATCH_ATTR4FV( ATTR, v );			\
-} while (0)
-#endif
-
-
-static void GLAPIENTRY _tnl_Vertex2f( GLfloat x, GLfloat y )
-{
-   DISPATCH_ATTR2F( _TNL_ATTRIB_POS, x, y );
-}
-
-static void GLAPIENTRY _tnl_Vertex2fv( const GLfloat *v )
-{
-   DISPATCH_ATTR2FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _tnl_Vertex3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_POS, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_Vertex3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _tnl_Vertex4f( GLfloat x, GLfloat y, GLfloat z, 
-				      GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_POS, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_Vertex4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_POS, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord1f( GLfloat x )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_TEX0, x );
-}
-
-static void GLAPIENTRY _tnl_TexCoord1fv( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord2f( GLfloat x, GLfloat y )
-{
-   DISPATCH_ATTR2F( _TNL_ATTRIB_TEX0, x, y );
-}
-
-static void GLAPIENTRY _tnl_TexCoord2fv( const GLfloat *v )
-{
-   DISPATCH_ATTR2FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_TEX0, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_TexCoord3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_TexCoord4f( GLfloat x, GLfloat y, GLfloat z,
-					GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_TEX0, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_TexCoord4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_TEX0, v );
-}
-
-static void GLAPIENTRY _tnl_Normal3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_NORMAL, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_Normal3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_NORMAL, v );
-}
-
-static void GLAPIENTRY _tnl_FogCoordfEXT( GLfloat x )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_FOG, x );
-}
-
-static void GLAPIENTRY _tnl_FogCoordfvEXT( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_FOG, v );
-}
-
-static void GLAPIENTRY _tnl_Color3f( GLfloat x, GLfloat y, GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_COLOR0, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_Color3fv( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_COLOR0, v );
-}
-
-static void GLAPIENTRY _tnl_Color4f( GLfloat x, GLfloat y, GLfloat z, 
-				     GLfloat w )
-{
-   DISPATCH_ATTR4F( _TNL_ATTRIB_COLOR0, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_Color4fv( const GLfloat *v )
-{
-   DISPATCH_ATTR4FV( _TNL_ATTRIB_COLOR0, v );
-}
-
-static void GLAPIENTRY _tnl_Indexf( GLfloat v )
-{
-   DISPATCH_ATTR1F( _TNL_ATTRIB_COLOR_INDEX, v );
-}
-
-static void GLAPIENTRY _tnl_Indexfv( const GLfloat *v )
-{
-   DISPATCH_ATTR1FV( _TNL_ATTRIB_COLOR_INDEX, v );
-}
-
-static void GLAPIENTRY _tnl_SecondaryColor3fEXT( GLfloat x, GLfloat y, 
-						 GLfloat z )
-{
-   DISPATCH_ATTR3F( _TNL_ATTRIB_COLOR1, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_SecondaryColor3fvEXT( const GLfloat *v )
-{
-   DISPATCH_ATTR3FV( _TNL_ATTRIB_COLOR1, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord1f( GLenum target, GLfloat x  )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR1F( attr, x );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord1fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR1FV( attr, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord2f( GLenum target, GLfloat x, GLfloat y )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR2F( attr, x, y );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord2fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR2FV( attr, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord3f( GLenum target, GLfloat x, 
-					     GLfloat y, GLfloat z)
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR3F( attr, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord3fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR3FV( attr, v );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord4f( GLenum target, GLfloat x, 
-					     GLfloat y, GLfloat z,
-					     GLfloat w )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR4F( attr, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_MultiTexCoord4fv( GLenum target, const GLfloat *v )
-{
-   GLuint attr = (target & 0x7) + _TNL_ATTRIB_TEX0;
-   DISPATCH_ATTR4FV( attr, v );
-}
-
-
-/**
- * GL_NV_vertex_program Vertex Attributes
- * Note that these attributes DO alias the conventional attributes.
- * Also, calling glVertexAttribNV(0, xxx) is equivalent to glVertex(xxx).
- */
-
-static void GLAPIENTRY _tnl_VertexAttrib1fNV( GLuint index, GLfloat x )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR1F( index, x );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib1fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR1FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fNV(GLuint index, GLfloat x, GLfloat y)
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR2F( index, x, y );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR2FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fNV( GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR3F( index, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR3FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fNV( GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z,
-					      GLfloat w )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR4F( index, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fvNV( GLuint index, const GLfloat *v )
-{
-   if (index >= MAX_VERTEX_PROGRAM_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   DISPATCH_ATTR4FV( index, v );
-}
-
-
-
-/**
- * GL_ARB_vertex_program Vertex Attributes
- * Note that these attributes do NOT alias the conventional attributes.
- * Also, calling glVertexAttribARB(0, xxx) is equivalent to glVertex(xxx).
- */
-
-static void GLAPIENTRY _tnl_VertexAttrib1fARB( GLuint index, GLfloat x )
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR1F( index, x );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib1fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR1FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fARB( GLuint index, GLfloat x, 
-					      GLfloat y )
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR2F( index, x, y );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib2fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR2FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fARB(GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR3F( index, x, y, z );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib3fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR3FV( index, v );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fARB(GLuint index, GLfloat x,
-					      GLfloat y, GLfloat z, GLfloat w)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR4F( index, x, y, z, w );
-}
-
-static void GLAPIENTRY _tnl_VertexAttrib4fvARB(GLuint index, const GLfloat *v)
-{
-   if (index >= MAX_VERTEX_ATTRIBS)
-      index = _TNL_ATTRIB_ERROR;
-   else if (index > 0)
-      index += VERT_ATTRIB_GENERIC0;
-   DISPATCH_ATTR4FV( index, v );
-}
-
-
-/* Install the generic versions of the 2nd level dispatch
- * functions.  Some of these have a codegen alternative.
- */
-void _tnl_generic_exec_vtxfmt_init( GLcontext *ctx )
-{
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-
-   vfmt->Color3f = _tnl_Color3f;
-   vfmt->Color3fv = _tnl_Color3fv;
-   vfmt->Color4f = _tnl_Color4f;
-   vfmt->Color4fv = _tnl_Color4fv;
-   vfmt->Indexf = _tnl_Indexf;
-   vfmt->Indexfv = _tnl_Indexfv;
-   vfmt->FogCoordfEXT = _tnl_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = _tnl_FogCoordfvEXT;
-   vfmt->MultiTexCoord1fARB = _tnl_MultiTexCoord1f;
-   vfmt->MultiTexCoord1fvARB = _tnl_MultiTexCoord1fv;
-   vfmt->MultiTexCoord2fARB = _tnl_MultiTexCoord2f;
-   vfmt->MultiTexCoord2fvARB = _tnl_MultiTexCoord2fv;
-   vfmt->MultiTexCoord3fARB = _tnl_MultiTexCoord3f;
-   vfmt->MultiTexCoord3fvARB = _tnl_MultiTexCoord3fv;
-   vfmt->MultiTexCoord4fARB = _tnl_MultiTexCoord4f;
-   vfmt->MultiTexCoord4fvARB = _tnl_MultiTexCoord4fv;
-   vfmt->Normal3f = _tnl_Normal3f;
-   vfmt->Normal3fv = _tnl_Normal3fv;
-   vfmt->SecondaryColor3fEXT = _tnl_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = _tnl_SecondaryColor3fvEXT;
-   vfmt->TexCoord1f = _tnl_TexCoord1f;
-   vfmt->TexCoord1fv = _tnl_TexCoord1fv;
-   vfmt->TexCoord2f = _tnl_TexCoord2f;
-   vfmt->TexCoord2fv = _tnl_TexCoord2fv;
-   vfmt->TexCoord3f = _tnl_TexCoord3f;
-   vfmt->TexCoord3fv = _tnl_TexCoord3fv;
-   vfmt->TexCoord4f = _tnl_TexCoord4f;
-   vfmt->TexCoord4fv = _tnl_TexCoord4fv;
-   vfmt->Vertex2f = _tnl_Vertex2f;
-   vfmt->Vertex2fv = _tnl_Vertex2fv;
-   vfmt->Vertex3f = _tnl_Vertex3f;
-   vfmt->Vertex3fv = _tnl_Vertex3fv;
-   vfmt->Vertex4f = _tnl_Vertex4f;
-   vfmt->Vertex4fv = _tnl_Vertex4fv;
-   vfmt->VertexAttrib1fNV = _tnl_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = _tnl_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV = _tnl_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = _tnl_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV = _tnl_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = _tnl_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV = _tnl_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = _tnl_VertexAttrib4fvNV;
-   vfmt->VertexAttrib1fARB = _tnl_VertexAttrib1fARB;
-   vfmt->VertexAttrib1fvARB = _tnl_VertexAttrib1fvARB;
-   vfmt->VertexAttrib2fARB = _tnl_VertexAttrib2fARB;
-   vfmt->VertexAttrib2fvARB = _tnl_VertexAttrib2fvARB;
-   vfmt->VertexAttrib3fARB = _tnl_VertexAttrib3fARB;
-   vfmt->VertexAttrib3fvARB = _tnl_VertexAttrib3fvARB;
-   vfmt->VertexAttrib4fARB = _tnl_VertexAttrib4fARB;
-   vfmt->VertexAttrib4fvARB = _tnl_VertexAttrib4fvARB;
-}
diff --git a/src/mesa/tnl/t_vtx_x86.c b/src/mesa/tnl/t_vtx_x86.c
deleted file mode 100644
index 99a05509327..00000000000
--- a/src/mesa/tnl/t_vtx_x86.c
+++ /dev/null
@@ -1,396 +0,0 @@
-/**************************************************************************
-
-Copyright 2004 Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *   Daniel Borca <dborca@yahoo.com>
- */
-
-
-#include "glheader.h"
-#include "context.h"
-#include "macros.h"
-#include "vtxfmt.h"
-#include "dlist.h"
-#include "state.h"
-#include "light.h"
-#include "api_arrayelt.h"
-#include "api_noop.h"
-#include "t_vtx_api.h"
-#include "simple_list.h"
-
-#if defined(USE_X86_ASM) && !defined(HAVE_NONSTANDARD_GLAPIENTRY)
-
-#define EXTERN( FUNC )		\
-extern const char FUNC[];	\
-extern const char FUNC##_end[]
-
-EXTERN( _tnl_x86_Attribute1fv );
-EXTERN( _tnl_x86_Attribute2fv );
-EXTERN( _tnl_x86_Attribute3fv );
-EXTERN( _tnl_x86_Attribute4fv );
-EXTERN( _tnl_x86_Vertex1fv );
-EXTERN( _tnl_x86_Vertex2fv );
-EXTERN( _tnl_x86_Vertex3fv );
-EXTERN( _tnl_x86_Vertex4fv );
-
-EXTERN( _tnl_x86_dispatch_attrf1 );
-EXTERN( _tnl_x86_dispatch_attrf2 );
-EXTERN( _tnl_x86_dispatch_attrf3 );
-EXTERN( _tnl_x86_dispatch_attrf4 );
-EXTERN( _tnl_x86_dispatch_attrfv );
-EXTERN( _tnl_x86_dispatch_multitexcoordf1 );
-EXTERN( _tnl_x86_dispatch_multitexcoordf2 );
-EXTERN( _tnl_x86_dispatch_multitexcoordf3 );
-EXTERN( _tnl_x86_dispatch_multitexcoordf4 );
-EXTERN( _tnl_x86_dispatch_multitexcoordfv );
-EXTERN( _tnl_x86_dispatch_vertexattribf1 );
-EXTERN( _tnl_x86_dispatch_vertexattribf2 );
-EXTERN( _tnl_x86_dispatch_vertexattribf3 );
-EXTERN( _tnl_x86_dispatch_vertexattribf4 );
-EXTERN( _tnl_x86_dispatch_vertexattribfv );
-
-EXTERN( _tnl_x86_choose_fv );
-
-
-#define DONT_KNOW_OFFSETS 1
-
-
-#define DFN( FUNC, CACHE, KEY )				\
-   struct _tnl_dynfn *dfn = MALLOC_STRUCT( _tnl_dynfn );\
-   const char *start = FUNC;				\
-   const char *end = FUNC##_end;			\
-   int offset = 0;               			\
-   insert_at_head( &CACHE, dfn );			\
-   dfn->key = KEY;					\
-   dfn->code = ALIGN_MALLOC( end - start, 16 );		\
-   _mesa_memcpy (dfn->code, start, end - start)
-
-
-
-#undef DEBUG_VTX
-
-#ifdef DEBUG_VTX
-#define FIXUP_PRINTF( offset, NEWVAL ) \
-      fprintf(stderr, "%s/%d: offset %d, new value: 0x%x\n", __FILE__, __LINE__, offset, (int)(NEWVAL))
-#define FIXUPREL_PRINTF( offset, NEWVAL, CODE ) \
-      fprintf(stderr, "%s/%d: offset %d, new value: 0x%x\n", __FILE__, __LINE__, offset, (int)(NEWVAL) - ((int)(CODE)+offset) - 4)
-#else
-#define FIXUP_PRINTF( offset, NEWVAL )
-#define FIXUPREL_PRINTF( offset, NEWVAL, CODE )
-#endif
-
-#define FIXUP( CODE, KNOWN_OFFSET, CHECKVAL, NEWVAL )	\
-do {							\
-   GLint subst = 0x10101010 + CHECKVAL;			\
-							\
-   if (DONT_KNOW_OFFSETS) {				\
-      while (*(int *)(CODE+offset) != subst) offset++;	\
-      *(int *)(CODE+offset) = (int)(NEWVAL);		\
-      FIXUP_PRINTF(offset, NEWVAL);			\
-      offset += 4;					\
-   }							\
-   else {						\
-      int *icode = (int *)(CODE+KNOWN_OFFSET);		\
-      assert (*icode == subst);				\
-      *icode = (int)NEWVAL;				\
-   }							\
-} while (0)
-
-
-
-#define FIXUPREL( CODE, KNOWN_OFFSET, CHECKVAL, NEWVAL )\
-do {							\
-   GLint subst = 0x10101010 + CHECKVAL;			\
-							\
-   if (DONT_KNOW_OFFSETS) {				\
-      while (*(int *)(CODE+offset) != subst) offset++;	\
-      *(int *)(CODE+offset) = (int)(NEWVAL) - ((int)(CODE)+offset) - 4; \
-      FIXUPREL_PRINTF(offset, NEWVAL, CODE);		\
-      offset += 4;					\
-   }							\
-   else {						\
-      int *icode = (int *)(CODE+KNOWN_OFFSET);		\
-      assert (*icode == subst);				\
-      *icode = (int)(NEWVAL) - (int)(icode) - 4;	\
-   }							\
-} while (0)
-
-
-
-
-/* Build specialized versions of the immediate calls on the fly for
- * the current state.  Generic x86 versions.
- */
-
-static struct _tnl_dynfn *makeX86Vertex1fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Vertex1fv, tnl->vtx.cache.Vertex[1-1], vertex_size );
-
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 1, vertex_size - 1);
-   FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[1]);
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 4, (int)ctx);
-   FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Vertex2fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Vertex2fv, tnl->vtx.cache.Vertex[2-1], vertex_size );
-
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 1, vertex_size - 2);
-   FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[2]);
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 4, (int)ctx);
-   FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Vertex3fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-
-   switch (vertex_size) {
-      default: {
-         DFN ( _tnl_x86_Vertex3fv, tnl->vtx.cache.Vertex[3-1], vertex_size );
-
-         FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-         FIXUP(dfn->code, 0, 1, vertex_size - 3);
-         FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[3]);
-         FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-         FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-         FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-         FIXUP(dfn->code, 0, 4, (int)ctx);
-         FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-         return dfn;
-      }
-   }
-}
-
-static struct _tnl_dynfn *makeX86Vertex4fv( GLcontext *ctx, int vertex_size )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Vertex4fv, tnl->vtx.cache.Vertex[4-1], vertex_size );
-
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 1, vertex_size - 4);
-   FIXUP(dfn->code, 0, 2, (int)&tnl->vtx.vertex[4]);
-   FIXUP(dfn->code, 0, 0, (int)&tnl->vtx.vbptr);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 3, (int)&tnl->vtx.counter);
-   FIXUP(dfn->code, 0, 4, (int)ctx);
-   FIXUPREL(dfn->code, 0, 5, (int)&_tnl_wrap_filled_vertex);
-
-   return dfn;
-}
-
-
-static struct _tnl_dynfn *makeX86Attribute1fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute1fv, tnl->vtx.cache.Attribute[1-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Attribute2fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute2fv, tnl->vtx.cache.Attribute[2-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-   FIXUP(dfn->code, 0, 1, 4+dest); 
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Attribute3fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute3fv, tnl->vtx.cache.Attribute[3-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-   FIXUP(dfn->code, 0, 1, 4+dest); 
-   FIXUP(dfn->code, 0, 2, 8+dest);
-
-   return dfn;
-}
-
-static struct _tnl_dynfn *makeX86Attribute4fv( GLcontext *ctx, int dest )
-{
-   TNLcontext *tnl = TNL_CONTEXT(ctx);
-   DFN ( _tnl_x86_Attribute4fv, tnl->vtx.cache.Attribute[4-1], dest );
-
-   FIXUP(dfn->code, 0, 0, dest); 
-   FIXUP(dfn->code, 0, 1, 4+dest); 
-   FIXUP(dfn->code, 0, 2, 8+dest);
-   FIXUP(dfn->code, 0, 3, 12+dest);
-
-   return dfn;
-}
-
-
-void _tnl_InitX86Codegen( struct _tnl_dynfn_generators *gen )
-{
-   gen->Vertex[0] = makeX86Vertex1fv;
-   gen->Vertex[1] = makeX86Vertex2fv;
-   gen->Vertex[2] = makeX86Vertex3fv;
-   gen->Vertex[3] = makeX86Vertex4fv;
-   gen->Attribute[0] = makeX86Attribute1fv;
-   gen->Attribute[1] = makeX86Attribute2fv;
-   gen->Attribute[2] = makeX86Attribute3fv;
-   gen->Attribute[3] = makeX86Attribute4fv;
-}
-
-
-#define MKDISP(FUNC, SIZE, ATTR, WARP)					\
-do {									\
-   char *code;								\
-   const char *start = WARP;						\
-   const char *end = WARP##_end;					\
-   int offset = 0;							\
-   code = ALIGN_MALLOC( end - start, 16 );				\
-   _mesa_memcpy (code, start, end - start);					\
-   FIXUP(code, 0, 0, (int)&(TNL_CONTEXT(ctx)->vtx.tabfv[ATTR][SIZE-1]));\
-   *(void **)&vfmt->FUNC = code;					\
-} while (0)
-
-
-/* Install the codegen'ed versions of the 2nd level dispatch
- * functions.  We should keep a list and free them in the end...
- */
-void _tnl_x86_exec_vtxfmt_init( GLcontext *ctx )
-{
-   GLvertexformat *vfmt = &(TNL_CONTEXT(ctx)->exec_vtxfmt);
-
-   MKDISP(Color3f,             3, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrf3);
-   MKDISP(Color3fv,            3, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrfv);
-   MKDISP(Color4f,             4, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrf4);
-   MKDISP(Color4fv,            4, _TNL_ATTRIB_COLOR0, _tnl_x86_dispatch_attrfv);
-   MKDISP(FogCoordfEXT,        1, _TNL_ATTRIB_FOG,    _tnl_x86_dispatch_attrf1);
-   MKDISP(FogCoordfvEXT,       1, _TNL_ATTRIB_FOG,    _tnl_x86_dispatch_attrfv);
-   MKDISP(Normal3f,            3, _TNL_ATTRIB_NORMAL, _tnl_x86_dispatch_attrf3);
-   MKDISP(Normal3fv,           3, _TNL_ATTRIB_NORMAL, _tnl_x86_dispatch_attrfv);
-   MKDISP(SecondaryColor3fEXT, 3, _TNL_ATTRIB_COLOR1, _tnl_x86_dispatch_attrf3);
-   MKDISP(SecondaryColor3fvEXT,3, _TNL_ATTRIB_COLOR1, _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord1f,          1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf1);
-   MKDISP(TexCoord1fv,         1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord2f,          2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf2);
-   MKDISP(TexCoord2fv,         2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord3f,          3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf3);
-   MKDISP(TexCoord3fv,         3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(TexCoord4f,          4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrf4);
-   MKDISP(TexCoord4fv,         4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_attrfv);
-   MKDISP(Vertex2f,            2, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrf2);
-   MKDISP(Vertex2fv,           2, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrfv);
-   MKDISP(Vertex3f,            3, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrf3);
-   MKDISP(Vertex3fv,           3, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrfv);
-   MKDISP(Vertex4f,            4, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrf4);
-   MKDISP(Vertex4fv,           4, _TNL_ATTRIB_POS,    _tnl_x86_dispatch_attrfv);
-
-   MKDISP(MultiTexCoord1fARB,  1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf1);
-   MKDISP(MultiTexCoord1fvARB, 1, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-   MKDISP(MultiTexCoord2fARB,  2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf2);
-   MKDISP(MultiTexCoord2fvARB, 2, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-   MKDISP(MultiTexCoord3fARB,  3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf3);
-   MKDISP(MultiTexCoord3fvARB, 3, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-   MKDISP(MultiTexCoord4fARB,  4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordf4);
-   MKDISP(MultiTexCoord4fvARB, 4, _TNL_ATTRIB_TEX0,   _tnl_x86_dispatch_multitexcoordfv);
-
-   MKDISP(VertexAttrib1fNV,    1, 0,                  _tnl_x86_dispatch_vertexattribf1);
-   MKDISP(VertexAttrib1fvNV,   1, 0,                  _tnl_x86_dispatch_vertexattribfv);
-   MKDISP(VertexAttrib2fNV,    2, 0,                  _tnl_x86_dispatch_vertexattribf2);
-   MKDISP(VertexAttrib2fvNV,   2, 0,                  _tnl_x86_dispatch_vertexattribfv);
-   MKDISP(VertexAttrib3fNV,    3, 0,                  _tnl_x86_dispatch_vertexattribf3);
-   MKDISP(VertexAttrib3fvNV,   3, 0,                  _tnl_x86_dispatch_vertexattribfv);
-   MKDISP(VertexAttrib4fNV,    4, 0,                  _tnl_x86_dispatch_vertexattribf4);
-   MKDISP(VertexAttrib4fvNV,   4, 0,                  _tnl_x86_dispatch_vertexattribfv);
-}
-
-
-/* Install the codegen'ed choosers.
- * We should keep a list and free them in the end...
- */
-void _tnl_x86choosers( tnl_attrfv_func (*choose)[4],
-		       tnl_attrfv_func (*do_choose)( GLuint attr,
-						 GLuint sz ))
-{
-   int attr, size;
-
-   for (attr = 0; attr < _TNL_MAX_ATTR_CODEGEN; attr++) {
-      for (size = 0; size < 4; size++) {
-         char *code;
-         const char *start = _tnl_x86_choose_fv;
-         const char *end = _tnl_x86_choose_fv_end;
-         int offset = 0;
-         code = ALIGN_MALLOC( end - start, 16 );
-         _mesa_memcpy (code, start, end - start);
-         FIXUP(code, 0, 0, attr);
-         FIXUP(code, 0, 1, size + 1);
-         FIXUPREL(code, 0, 2, do_choose);
-         choose[attr][size] = (tnl_attrfv_func)code;
-      }
-   }
-}
-
-#else 
-
-void _tnl_InitX86Codegen( struct _tnl_dynfn_generators *gen )
-{
-   (void) gen;
-}
-
-
-void _tnl_x86_exec_vtxfmt_init( GLcontext *ctx )
-{
-   (void) ctx;
-}
-
-
-void _tnl_x86choosers( tnl_attrfv_func (*choose)[4],
-		       tnl_attrfv_func (*do_choose)( GLuint attr,
-						 GLuint sz ))
-{
-   (void) choose;
-   (void) do_choose;
-}
-
-#endif
diff --git a/src/mesa/tnl/t_vtx_x86_gcc.S b/src/mesa/tnl/t_vtx_x86_gcc.S
deleted file mode 100644
index 93753be759b..00000000000
--- a/src/mesa/tnl/t_vtx_x86_gcc.S
+++ /dev/null
@@ -1,561 +0,0 @@
-/**************************************************************************
-
-Copyright 2004 Tungsten Graphics Inc., Cedar Park, Texas.
-
-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"),
-to deal in the Software without restriction, including without limitation
-on the rights to use, copy, modify, merge, publish, distribute, sub
-license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
-ATI, TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS 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:
- *   Keith Whitwell <keith@tungstengraphics.com>
- *   Daniel Borca <dborca@yahoo.com>
- */
-
-#if defined (__DJGPP__) || defined (__MINGW32__) || defined (__CYGWIN__)
-#define GLOBL( x )	\
-.globl _##x;		\
-_##x:
-#else  /* !defined (__DJGPP__) && !defined (__MINGW32__) && !defined (__CYGWIN__) */
-#define GLOBL( x )	\
-.globl x;		\
-x:
-#endif /* !defined (__DJGPP__) && !defined (__MINGW32__) && !defined (__CYGWIN__) */
-
-
-#if !defined (STDCALL_API)
-#define RETCLEAN( x )	ret
-#else
-#define RETCLEAN( x )	ret $x
-#endif
-
-
-#define _JMP(x)		\
-.byte 0xe9;		\
-.long x
-
-#define _CALL(x)	\
-.byte 0xe8;		\
-.long x
-
-
-/* Someone who knew a lot about this sort of thing would use this
- * macro to note current offsets, etc in a special region of the
- * object file & just make everything work out neat.  I don't know
- * enough to do that...
- */
-
-#define SUBST( x ) (0x10101010 + x)
-
-
-.data
-
-
-/* [dBorca] TODO
- * Unfold functions for each vertex size?
- * Build super-specialized SSE versions?
- *
- * There is a trick in Vertex*fv: under certain conditions,
- * we tail to _tnl_wrap_filled_vertex(ctx). This means that
- * if Vertex*fv is STDCALL, then _tnl_wrap_filled_vertex must
- * be STDCALL as well, because (GLcontext *) and (GLfloat *)
- * have the same size.
- */
-.align 4
-GLOBL ( _tnl_x86_Vertex1fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* 0x0 --> tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	addl	$4, %edi	/* tnl->vtx.vbptr += 1 */
-	movl	$SUBST(1), %ecx	/* 0x1 --> (tnl->vtx.vertex_size - 1) */
-	movl	$SUBST(2), %esi	/* 0x2 --> (tnl->vtx.vertex + 1) */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* 0x0 --> tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* 0x3 --> counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* 0x3 --> counter */
-	je	.0		/* if (counter == 0) goto .0 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.0:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex1fv_end )
-
-.align 4
-GLOBL ( _tnl_x86_Vertex2fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* load tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	4(%ecx), %eax	/* load v[1] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	movl	%eax, 4(%edi)	/* tnl->vtx.vbptr[1] = v[1] */
-	addl	$8, %edi	/* tnl->vtx.vbptr += 2 */
-	movl	$SUBST(1), %ecx	/* vertex_size - 2 */
-	movl	$SUBST(2), %esi	/* tnl->vtx.vertex + 2 */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* save tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* load counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* save counter */
-	je	.1		/* if (counter == 0) goto .1 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.1:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex2fv_end )
-
-.align 4
-GLOBL ( _tnl_x86_Vertex3fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* load tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	4(%ecx), %eax	/* load v[1] */
-	movl	8(%ecx), %esi	/* load v[2] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	movl	%eax, 4(%edi)	/* tnl->vtx.vbptr[1] = v[1] */
-	movl	%esi, 8(%edi)	/* tnl->vtx.vbptr[2] = v[2] */
-	addl	$12, %edi	/* tnl->vtx.vbptr += 3 */
-	movl	$SUBST(1), %ecx	/* vertex_size - 3 */
-	movl	$SUBST(2), %esi	/* tnl->vtx.vertex + 3 */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* save tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* load counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* save counter */
-	je	.2		/* if (counter == 0) goto .2 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.2:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex3fv_end )
-
-.align 4
-GLOBL ( _tnl_x86_Vertex4fv )
-	movl	4(%esp), %ecx
-	push	%edi
-	push	%esi
-	movl	SUBST(0), %edi	/* load tnl->vtx.vbptr */
-	movl	(%ecx), %edx	/* load v[0] */
-	movl	4(%ecx), %eax	/* load v[1] */
-	movl	8(%ecx), %esi	/* load v[2] */
-	movl	12(%ecx), %ecx	/* load v[3] */
-	movl	%edx, (%edi)	/* tnl->vtx.vbptr[0] = v[0] */
-	movl	%eax, 4(%edi)	/* tnl->vtx.vbptr[1] = v[1] */
-	movl	%esi, 8(%edi)	/* tnl->vtx.vbptr[2] = v[2] */
-	movl	%ecx, 12(%edi)	/* tnl->vtx.vbptr[3] = v[3] */
-	addl	$16, %edi	/* tnl->vtx.vbptr += 4 */
-	movl	$SUBST(1), %ecx	/* vertex_size - 4 */
-	movl	$SUBST(2), %esi	/* tnl->vtx.vertex + 4 */
-	repz
-	movsl	%ds:(%esi), %es:(%edi)
-	movl	%edi, SUBST(0)	/* save tnl->vtx.vbptr */
-	movl	SUBST(3), %edx	/* load counter */
-	pop	%esi
-	pop	%edi
-	dec	%edx		/* counter-- */
-	movl	%edx, SUBST(3)	/* save counter */
-	je	.3		/* if (counter == 0) goto .3 */
-	RETCLEAN(4)		/* return */
-	.balign	16
-.3:
-	movl	$SUBST(4), %eax	/* load ctx */
-	movl	%eax, 4(%esp)	/* push ctx */
-	_JMP	(SUBST(5))	/* jmp _tnl_wrap_filled_vertex */
-GLOBL ( _tnl_x86_Vertex4fv_end )
-
-
-/**
- * Generic handlers for vector format data. 
- */
-GLOBL( _tnl_x86_Attribute1fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute1fv_end )
-
-GLOBL( _tnl_x86_Attribute2fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	4(%ecx), %edx	/* load v[1] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	movl	%edx, SUBST(1)	/* store v[1] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute2fv_end )
-
-GLOBL( _tnl_x86_Attribute3fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	4(%ecx), %edx	/* load v[1] */
-	movl	8(%ecx), %ecx	/* load v[2] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	movl	%edx, SUBST(1)	/* store v[1] to current vertex */
-	movl	%ecx, SUBST(2)	/* store v[2] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute3fv_end )
-
-GLOBL( _tnl_x86_Attribute4fv )
-	movl	4(%esp), %ecx
-	movl	(%ecx), %eax	/* load v[0] */
-	movl	4(%ecx), %edx	/* load v[1] */
-	movl	%eax, SUBST(0)	/* store v[0] to current vertex */
-	movl	%edx, SUBST(1)	/* store v[1] to current vertex */
-	movl	8(%ecx), %eax	/* load v[2] */
-	movl	12(%ecx), %edx	/* load v[3] */
-	movl	%eax, SUBST(2)	/* store v[2] to current vertex */
-	movl	%edx, SUBST(3)	/* store v[3] to current vertex */
-	RETCLEAN(4)
-GLOBL ( _tnl_x86_Attribute4fv_end )
-
-
-/* Choosers:
- *
- * Must generate all of these ahead of first usage.  Generate at
- * compile-time?
- */
-GLOBL( _tnl_x86_choose_fv )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	$SUBST(0), (%esp)	/* arg 0 - attrib */
-	movl	$SUBST(1), 4(%esp)	/* arg 1 - N */
-	_CALL	(SUBST(2))		/* call do_choose */
-	add	$12, %esp		/* tear down stack frame */
-	jmp	*%eax			/* jump to new func */
-GLOBL ( _tnl_x86_choose_fv_end )
-
-
-/* FIRST LEVEL FUNCTIONS -- these are plugged directly into GL dispatch.
- *
- * In the 1st level dispatch functions, switch to a different
- * calling convention -- (const GLfloat *v) in %ecx.
- *
- * As with regular (x86) dispatch, don't create a new stack frame -
- * just let the 'ret' in the dispatched function return straight
- * back to the original caller.
- *
- * Vertex/Normal/Color, etc: the address of the function pointer
- * is known at codegen time.
- */
-
-/* Unfortunately, have to play with the stack in the non-fv case:
- */
-#if !defined (STDCALL_API)
-GLOBL( _tnl_x86_dispatch_attrf1 )
-GLOBL( _tnl_x86_dispatch_attrf2 )
-GLOBL( _tnl_x86_dispatch_attrf3 )
-GLOBL( _tnl_x86_dispatch_attrf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$12, %esp	/* tear down frame */
-	ret			/* return */
-GLOBL( _tnl_x86_dispatch_attrf4_end )
-GLOBL( _tnl_x86_dispatch_attrf3_end )
-GLOBL( _tnl_x86_dispatch_attrf2_end )
-GLOBL( _tnl_x86_dispatch_attrf1_end )
-
-#else  /* defined(STDCALL_API) */
-
-GLOBL( _tnl_x86_dispatch_attrf1 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$4		/* return */
-GLOBL( _tnl_x86_dispatch_attrf1_end )
-
-GLOBL( _tnl_x86_dispatch_attrf2 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_attrf2_end )
-
-GLOBL( _tnl_x86_dispatch_attrf3 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$12		/* return */
-GLOBL( _tnl_x86_dispatch_attrf3_end )
-
-GLOBL( _tnl_x86_dispatch_attrf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	leal	16(%esp), %edx	/* address of first float on stack */
-	movl	%edx, (%esp)	/* save as 'v' */
-	call	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$16		/* return */
-GLOBL( _tnl_x86_dispatch_attrf4_end )
-#endif /* defined(STDCALL_API) */
-
-/* The fv case is simpler:
- */
-GLOBL( _tnl_x86_dispatch_attrfv )
-	jmp	*SUBST(0)	/* 0x0 --> tabfv[attr][n] */
-GLOBL( _tnl_x86_dispatch_attrfv_end )
-
-
-/* MultiTexcoord: the address of the function pointer must be
- * calculated, but can use the index argument slot to hold 'v', and
- * avoid setting up a new stack frame.
- *
- * [dBorca]
- * right, this would be the preferred approach, but gcc does not
- * clean up the stack after each function call when optimizing (-fdefer-pop);
- * can it make assumptions about what's already on the stack?  I dunno,
- * but in this case, we can't mess with the caller's stack frame, and
- * we must use a model like `_x86_dispatch_attrfv' above.  Caveat emptor!
- */
-
-/* Also, will only need a maximum of four of each of these per context:
- */
-#if !defined (STDCALL_API)
-GLOBL( _tnl_x86_dispatch_multitexcoordf1 )
-GLOBL( _tnl_x86_dispatch_multitexcoordf2 )
-GLOBL( _tnl_x86_dispatch_multitexcoordf3 )
-GLOBL( _tnl_x86_dispatch_multitexcoordf4 )
-	movl	4(%esp), %ecx
-	leal	8(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, 4(%esp)
-	sall	$4, %ecx
-	jmp	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-GLOBL( _tnl_x86_dispatch_multitexcoordf4_end )
-GLOBL( _tnl_x86_dispatch_multitexcoordf3_end )
-GLOBL( _tnl_x86_dispatch_multitexcoordf2_end )
-GLOBL( _tnl_x86_dispatch_multitexcoordf1_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordfv )
-	movl	4(%esp), %ecx
-	movl	8(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, 4(%esp)
-	sall	$4, %ecx
-	jmp	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-GLOBL( _tnl_x86_dispatch_multitexcoordfv_end )
-
-#else  /* defined (STDCALL_API) */
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf1 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf1_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf2 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$12		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf2_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf3 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$16		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf3_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	leal	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$20		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordf4_end )
-
-GLOBL( _tnl_x86_dispatch_multitexcoordfv )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %ecx
-	movl	20(%esp), %edx
-	andl	$7, %ecx
-	movl	%edx, (%esp)
-	sall	$4, %ecx
-	call	*SUBST(0)(%ecx)	/* 0x0 - tabfv[tex0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_multitexcoordfv_end )
-#endif /* defined (STDCALL_API) */
-
-
-/* VertexAttrib: the address of the function pointer must be
- * calculated.
- */
-#if !defined (STDCALL_API)
-GLOBL( _tnl_x86_dispatch_vertexattribf1 )
-GLOBL( _tnl_x86_dispatch_vertexattribf2 )
-GLOBL( _tnl_x86_dispatch_vertexattribf3 )
-GLOBL( _tnl_x86_dispatch_vertexattribf4 )
-	movl	4(%esp), %eax
-	cmpl	$16, %eax
-	jb	.8		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.8:
-	leal	8(%esp), %ecx	/* calculate 'v' */
-	movl	%ecx, 4(%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	jmp	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-GLOBL( _tnl_x86_dispatch_vertexattribf4_end )
-GLOBL( _tnl_x86_dispatch_vertexattribf3_end )
-GLOBL( _tnl_x86_dispatch_vertexattribf2_end )
-GLOBL( _tnl_x86_dispatch_vertexattribf1_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribfv )
-	movl	4(%esp), %eax
-	cmpl	$16, %eax
-	jb	.9		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.9:
-	movl	8(%esp), %ecx	/* load 'v' */
-	movl	%ecx, 4(%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	jmp	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-GLOBL( _tnl_x86_dispatch_vertexattribfv_end )
-
-#else  /* defined (STDCALL_API) */
-
-GLOBL( _tnl_x86_dispatch_vertexattribf1 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.81		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.81:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf1_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribf2 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.82		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.82:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$12		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf2_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribf3 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.83		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.83:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$16		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf3_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribf4 )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.84		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.84:
-	leal	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$20		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribf4_end )
-
-GLOBL( _tnl_x86_dispatch_vertexattribfv )
-	subl	$12, %esp	/* gcc does 16 byte alignment of stack frames? */
-	movl	16(%esp), %eax
-	cmpl	$16, %eax
-	jb	.9		/* "cmovge" is not supported on all CPUs */
-	movl	$16, %eax
-.9:
-	movl	20(%esp), %ecx	/* load 'v' */
-	movl	%ecx, (%esp)	/* save in 1st arg slot */
-	sall	$4, %eax
-	call	*SUBST(0)(%eax)	/* 0x0 - tabfv[0][n] */
-	addl	$8, %esp	/* tear down frame (4 shaved off by the callee) */
-	ret	$8		/* return */
-GLOBL( _tnl_x86_dispatch_vertexattribfv_end )
-#endif /* defined (STDCALL_API) */
-
-#if defined (__ELF__) && defined (__linux__)
-	.section .note.GNU-stack,"",%progbits
-#endif
diff --git a/src/mesa/tnl/tnl.h b/src/mesa/tnl/tnl.h
index dab422429a1..17cb30adc76 100644
--- a/src/mesa/tnl/tnl.h
+++ b/src/mesa/tnl/tnl.h
@@ -50,30 +50,14 @@ _tnl_InvalidateState( GLcontext *ctx, GLuint new_state );
  * dispatch and/or driver callbacks.
  */
 
-/* Restore just the ctx->Exec table:
- */
-extern void
-_tnl_wakeup_exec( GLcontext *ctx );
-
-/* Restore both ctx->Exec and ctx->Save:
- */
 extern void
-_tnl_wakeup_save_exec( GLcontext *ctx );
+_tnl_wakeup( GLcontext *ctx );
 
 /* Driver configuration options:
  */
 extern void
 _tnl_need_projected_coords( GLcontext *ctx, GLboolean flag );
 
-extern void
-_tnl_need_dlist_loopback( GLcontext *ctx, GLboolean flag );
-
-extern void
-_tnl_need_dlist_norm_lengths( GLcontext *ctx, GLboolean flag );
-
-extern void
-_tnl_isolate_materials( GLcontext *ctx, GLboolean flag );
-
 
 /* Control whether T&L does per-vertex fog
  */
@@ -86,4 +70,16 @@ _tnl_allow_pixel_fog( GLcontext *ctx, GLboolean value );
 extern void
 _tnl_program_string(GLcontext *ctx, GLenum target, struct gl_program *program);
 
+struct _mesa_prim;
+struct _mesa_index_buffer;
+
+void
+_tnl_draw_prims( GLcontext *ctx,
+		 const struct gl_client_array *arrays[],
+		 const struct _mesa_prim *prim,
+		 GLuint nr_prims,
+		 const struct _mesa_index_buffer *ib,
+		 GLuint min_index,
+		 GLuint max_index);
+
 #endif
diff --git a/src/mesa/vbo/vbo.h b/src/mesa/vbo/vbo.h
new file mode 100644
index 00000000000..874a5f9e0e2
--- /dev/null
+++ b/src/mesa/vbo/vbo.h
@@ -0,0 +1,117 @@
+/*
+ * mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  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"),
+ * 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 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
+ * BRIAN PAUL 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.
+ */
+
+/**
+ * \file vbo_context.h
+ * \brief VBO builder module datatypes and definitions.
+ * \author Keith Whitwell
+ */
+
+
+#ifndef _VBO_H
+#define _VBO_H
+
+#include "mtypes.h"
+
+struct _mesa_prim {
+   GLuint mode:8;
+   GLuint indexed:1;
+   GLuint begin:1;
+   GLuint end:1;
+   GLuint weak:1;
+   GLuint pad:20;
+
+   GLuint start;
+   GLuint count;
+};
+
+/* Would like to call this a "vbo_index_buffer", but this would be
+ * confusing as the indices are not neccessarily yet in a non-null
+ * buffer object.
+ */
+struct _mesa_index_buffer {
+   GLuint count;
+   GLenum type;
+   struct gl_buffer_object *obj;
+   const void *ptr;
+};
+
+
+
+GLboolean _vbo_CreateContext( GLcontext *ctx );
+void _vbo_DestroyContext( GLcontext *ctx );
+void _vbo_InvalidateState( GLcontext *ctx, GLuint new_state );
+
+
+typedef void (*vbo_draw_func)( GLcontext *ctx,
+			       const struct gl_client_array **arrays,
+			       const struct _mesa_prim *prims,
+			       GLuint nr_prims,
+			       const struct _mesa_index_buffer *ib,
+			       GLuint min_index,
+			       GLuint max_index );
+
+
+
+
+/* Utility function to cope with various constraints on tnl modules or
+ * hardware.  This can be used to split an incoming set of arrays and
+ * primitives against the following constraints:
+ *    - Maximum number of indices in index buffer.
+ *    - Maximum number of vertices referenced by index buffer.
+ *    - Maximum hardware vertex buffer size.
+ */
+struct split_limits {
+   GLuint max_verts;
+   GLuint max_indices;
+   GLuint max_vb_size;		/* bytes */
+};
+
+
+void vbo_split_prims( GLcontext *ctx,
+		      const struct gl_client_array *arrays[],
+		      const struct _mesa_prim *prim,
+		      GLuint nr_prims,
+		      const struct _mesa_index_buffer *ib,
+		      GLuint min_index,
+		      GLuint max_index,
+		      vbo_draw_func draw,
+		      const struct split_limits *limits );
+
+
+/* Helpers for dealing translating away non-zero min_index.
+ */
+GLboolean vbo_all_varyings_in_vbos( const struct gl_client_array *arrays[] );
+
+void vbo_rebase_prims( GLcontext *ctx,
+		       const struct gl_client_array *arrays[],
+		       const struct _mesa_prim *prim,
+		       GLuint nr_prims,
+		       const struct _mesa_index_buffer *ib,
+		       GLuint min_index,
+		       GLuint max_index,
+		       vbo_draw_func draw );
+
+
+#endif
diff --git a/src/mesa/drivers/dri/i965/brw_attrib.h b/src/mesa/vbo/vbo_attrib.h
index 12659bd1cf0..0ae928f2af4 100644
--- a/src/mesa/drivers/dri/i965/brw_attrib.h
+++ b/src/mesa/vbo/vbo_attrib.h
@@ -29,8 +29,8 @@
   *   Keith Whitwell <keith@tungstengraphics.com>
   */
 
-#ifndef BRW_ATTRIB_H
-#define BRW_ATTRIB_H
+#ifndef VBO_ATTRIB_H
+#define VBO_ATTRIB_H
 
 
 /*
@@ -44,71 +44,64 @@
  * program instruction formats and register layouts.
  */
 enum {
-	BRW_ATTRIB_POS = 0,
-	BRW_ATTRIB_WEIGHT = 1,
-	BRW_ATTRIB_NORMAL = 2,
-	BRW_ATTRIB_COLOR0 = 3,
-	BRW_ATTRIB_COLOR1 = 4,
-	BRW_ATTRIB_FOG = 5,
-	BRW_ATTRIB_INDEX = 6,        
-	BRW_ATTRIB_EDGEFLAG = 7,     
-	BRW_ATTRIB_TEX0 = 8,
-	BRW_ATTRIB_TEX1 = 9,
-	BRW_ATTRIB_TEX2 = 10,
-	BRW_ATTRIB_TEX3 = 11,
-	BRW_ATTRIB_TEX4 = 12,
-	BRW_ATTRIB_TEX5 = 13,
-	BRW_ATTRIB_TEX6 = 14,
-	BRW_ATTRIB_TEX7 = 15,
+	VBO_ATTRIB_POS = 0,
+	VBO_ATTRIB_WEIGHT = 1,
+	VBO_ATTRIB_NORMAL = 2,
+	VBO_ATTRIB_COLOR0 = 3,
+	VBO_ATTRIB_COLOR1 = 4,
+	VBO_ATTRIB_FOG = 5,
+	VBO_ATTRIB_INDEX = 6,        
+	VBO_ATTRIB_EDGEFLAG = 7,     
+	VBO_ATTRIB_TEX0 = 8,
+	VBO_ATTRIB_TEX1 = 9,
+	VBO_ATTRIB_TEX2 = 10,
+	VBO_ATTRIB_TEX3 = 11,
+	VBO_ATTRIB_TEX4 = 12,
+	VBO_ATTRIB_TEX5 = 13,
+	VBO_ATTRIB_TEX6 = 14,
+	VBO_ATTRIB_TEX7 = 15,
 
-	BRW_ATTRIB_GENERIC0 = 16, /* Not used? */
-	BRW_ATTRIB_GENERIC1 = 17,
-	BRW_ATTRIB_GENERIC2 = 18,
-	BRW_ATTRIB_GENERIC3 = 19,
-	BRW_ATTRIB_GENERIC4 = 20,
-	BRW_ATTRIB_GENERIC5 = 21,
-	BRW_ATTRIB_GENERIC6 = 22,
-	BRW_ATTRIB_GENERIC7 = 23,
-	BRW_ATTRIB_GENERIC8 = 24,
-	BRW_ATTRIB_GENERIC9 = 25,
-	BRW_ATTRIB_GENERIC10 = 26,
-	BRW_ATTRIB_GENERIC11 = 27,
-	BRW_ATTRIB_GENERIC12 = 28,
-	BRW_ATTRIB_GENERIC13 = 29,
-	BRW_ATTRIB_GENERIC14 = 30,
-	BRW_ATTRIB_GENERIC15 = 31,
+	VBO_ATTRIB_GENERIC0 = 16, /* Not used? */
+	VBO_ATTRIB_GENERIC1 = 17,
+	VBO_ATTRIB_GENERIC2 = 18,
+	VBO_ATTRIB_GENERIC3 = 19,
+	VBO_ATTRIB_GENERIC4 = 20,
+	VBO_ATTRIB_GENERIC5 = 21,
+	VBO_ATTRIB_GENERIC6 = 22,
+	VBO_ATTRIB_GENERIC7 = 23,
+	VBO_ATTRIB_GENERIC8 = 24,
+	VBO_ATTRIB_GENERIC9 = 25,
+	VBO_ATTRIB_GENERIC10 = 26,
+	VBO_ATTRIB_GENERIC11 = 27,
+	VBO_ATTRIB_GENERIC12 = 28,
+	VBO_ATTRIB_GENERIC13 = 29,
+	VBO_ATTRIB_GENERIC14 = 30,
+	VBO_ATTRIB_GENERIC15 = 31,
 
-	BRW_ATTRIB_MAT_FRONT_AMBIENT = 32, 
-	BRW_ATTRIB_MAT_BACK_AMBIENT = 33,
-	BRW_ATTRIB_MAT_FRONT_DIFFUSE = 34,
-	BRW_ATTRIB_MAT_BACK_DIFFUSE = 35,
-	BRW_ATTRIB_MAT_FRONT_SPECULAR = 36,
-	BRW_ATTRIB_MAT_BACK_SPECULAR = 37,
-	BRW_ATTRIB_MAT_FRONT_EMISSION = 38,
-	BRW_ATTRIB_MAT_BACK_EMISSION = 39,
-	BRW_ATTRIB_MAT_FRONT_SHININESS = 40,
-	BRW_ATTRIB_MAT_BACK_SHININESS = 41,
-	BRW_ATTRIB_MAT_FRONT_INDEXES = 42,
-	BRW_ATTRIB_MAT_BACK_INDEXES = 43, 
+	/* XXX: in the vertex program InputsRead flag, we alias
+	 * materials and generics and use knowledge about the program
+	 * (whether it is a fixed-function emulation) to
+	 * differentiate.  Here we must keep them apart instead.
+	 */
+	VBO_ATTRIB_MAT_FRONT_AMBIENT = 32, 
+	VBO_ATTRIB_MAT_BACK_AMBIENT = 33,
+	VBO_ATTRIB_MAT_FRONT_DIFFUSE = 34,
+	VBO_ATTRIB_MAT_BACK_DIFFUSE = 35,
+	VBO_ATTRIB_MAT_FRONT_SPECULAR = 36,
+	VBO_ATTRIB_MAT_BACK_SPECULAR = 37,
+	VBO_ATTRIB_MAT_FRONT_EMISSION = 38,
+	VBO_ATTRIB_MAT_BACK_EMISSION = 39,
+	VBO_ATTRIB_MAT_FRONT_SHININESS = 40,
+	VBO_ATTRIB_MAT_BACK_SHININESS = 41,
+	VBO_ATTRIB_MAT_FRONT_INDEXES = 42,
+	VBO_ATTRIB_MAT_BACK_INDEXES = 43, 
 
-	BRW_ATTRIB_MAX = 44
-} ;
+	VBO_ATTRIB_MAX = 44
+};
 
-#define BRW_ATTRIB_FIRST_MATERIAL BRW_ATTRIB_MAT_FRONT_AMBIENT
-#define BRW_ATTRIB_LAST_MATERIAL BRW_ATTRIB_MAT_BACK_INDEXES
-
-#define BRW_MAX_COPIED_VERTS 3
-
-
-static inline GLuint64EXT brw_translate_inputs( GLboolean vp_enabled,
-						GLuint mesa_inputs )
-{
-   GLuint64EXT inputs = mesa_inputs;
-   if (vp_enabled)
-      return inputs;
-   else 
-      return (inputs & 0xffff) | ((inputs & 0xffff0000) << 16);
-}
+#define VBO_ATTRIB_FIRST_MATERIAL VBO_ATTRIB_MAT_FRONT_AMBIENT
+#define VBO_ATTRIB_LAST_MATERIAL VBO_ATTRIB_MAT_BACK_INDEXES
 
+#define VBO_MAX_COPIED_VERTS 3
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_attrib_tmp.h b/src/mesa/vbo/vbo_attrib_tmp.h
index 3089bd6cac2..ff11c7d59a7 100644
--- a/src/mesa/drivers/dri/i965/brw_attrib_tmp.h
+++ b/src/mesa/vbo/vbo_attrib_tmp.h
@@ -40,192 +40,192 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 static void GLAPIENTRY TAG(Vertex2f)( GLfloat x, GLfloat y )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR2F( BRW_ATTRIB_POS, x, y );
+   ATTR2F( VBO_ATTRIB_POS, x, y );
 }
 
 static void GLAPIENTRY TAG(Vertex2fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR2FV( BRW_ATTRIB_POS, v );
+   ATTR2FV( VBO_ATTRIB_POS, v );
 }
 
 static void GLAPIENTRY TAG(Vertex3f)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_POS, x, y, z );
+   ATTR3F( VBO_ATTRIB_POS, x, y, z );
 }
 
 static void GLAPIENTRY TAG(Vertex3fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_POS, v );
+   ATTR3FV( VBO_ATTRIB_POS, v );
 }
 
 static void GLAPIENTRY TAG(Vertex4f)( GLfloat x, GLfloat y, GLfloat z, GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4F( BRW_ATTRIB_POS, x, y, z, w );
+   ATTR4F( VBO_ATTRIB_POS, x, y, z, w );
 }
 
 static void GLAPIENTRY TAG(Vertex4fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4FV( BRW_ATTRIB_POS, v );
+   ATTR4FV( VBO_ATTRIB_POS, v );
 }
 
 static void GLAPIENTRY TAG(TexCoord1f)( GLfloat x )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1F( BRW_ATTRIB_TEX0, x );
+   ATTR1F( VBO_ATTRIB_TEX0, x );
 }
 
 static void GLAPIENTRY TAG(TexCoord1fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1FV( BRW_ATTRIB_TEX0, v );
+   ATTR1FV( VBO_ATTRIB_TEX0, v );
 }
 
 static void GLAPIENTRY TAG(TexCoord2f)( GLfloat x, GLfloat y )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR2F( BRW_ATTRIB_TEX0, x, y );
+   ATTR2F( VBO_ATTRIB_TEX0, x, y );
 }
 
 static void GLAPIENTRY TAG(TexCoord2fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR2FV( BRW_ATTRIB_TEX0, v );
+   ATTR2FV( VBO_ATTRIB_TEX0, v );
 }
 
 static void GLAPIENTRY TAG(TexCoord3f)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_TEX0, x, y, z );
+   ATTR3F( VBO_ATTRIB_TEX0, x, y, z );
 }
 
 static void GLAPIENTRY TAG(TexCoord3fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_TEX0, v );
+   ATTR3FV( VBO_ATTRIB_TEX0, v );
 }
 
 static void GLAPIENTRY TAG(TexCoord4f)( GLfloat x, GLfloat y, GLfloat z, GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4F( BRW_ATTRIB_TEX0, x, y, z, w );
+   ATTR4F( VBO_ATTRIB_TEX0, x, y, z, w );
 }
 
 static void GLAPIENTRY TAG(TexCoord4fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4FV( BRW_ATTRIB_TEX0, v );
+   ATTR4FV( VBO_ATTRIB_TEX0, v );
 }
 
 static void GLAPIENTRY TAG(Normal3f)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_NORMAL, x, y, z );
+   ATTR3F( VBO_ATTRIB_NORMAL, x, y, z );
 }
 
 static void GLAPIENTRY TAG(Normal3fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_NORMAL, v );
+   ATTR3FV( VBO_ATTRIB_NORMAL, v );
 }
 
 static void GLAPIENTRY TAG(FogCoordfEXT)( GLfloat x )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1F( BRW_ATTRIB_FOG, x );
+   ATTR1F( VBO_ATTRIB_FOG, x );
 }
 
 static void GLAPIENTRY TAG(FogCoordfvEXT)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1FV( BRW_ATTRIB_FOG, v );
+   ATTR1FV( VBO_ATTRIB_FOG, v );
 }
 
 static void GLAPIENTRY TAG(Color3f)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_COLOR0, x, y, z );
+   ATTR3F( VBO_ATTRIB_COLOR0, x, y, z );
 }
 
 static void GLAPIENTRY TAG(Color3fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_COLOR0, v );
+   ATTR3FV( VBO_ATTRIB_COLOR0, v );
 }
 
 static void GLAPIENTRY TAG(Color4f)( GLfloat x, GLfloat y, GLfloat z, GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4F( BRW_ATTRIB_COLOR0, x, y, z, w );
+   ATTR4F( VBO_ATTRIB_COLOR0, x, y, z, w );
 }
 
 static void GLAPIENTRY TAG(Color4fv)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR4FV( BRW_ATTRIB_COLOR0, v );
+   ATTR4FV( VBO_ATTRIB_COLOR0, v );
 }
 
 static void GLAPIENTRY TAG(SecondaryColor3fEXT)( GLfloat x, GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3F( BRW_ATTRIB_COLOR1, x, y, z );
+   ATTR3F( VBO_ATTRIB_COLOR1, x, y, z );
 }
 
 static void GLAPIENTRY TAG(SecondaryColor3fvEXT)( const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR3FV( BRW_ATTRIB_COLOR1, v );
+   ATTR3FV( VBO_ATTRIB_COLOR1, v );
 }
 
 
 static void GLAPIENTRY TAG(EdgeFlag)( GLboolean b )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1F( BRW_ATTRIB_EDGEFLAG, (GLfloat)b );
+   ATTR1F( VBO_ATTRIB_EDGEFLAG, (GLfloat)b );
 }
 
 static void GLAPIENTRY TAG(Indexf)( GLfloat f )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1F( BRW_ATTRIB_INDEX, f );
+   ATTR1F( VBO_ATTRIB_INDEX, f );
 }
 
 static void GLAPIENTRY TAG(Indexfv)( const GLfloat *f )
 {
    GET_CURRENT_CONTEXT( ctx );
-   ATTR1FV( BRW_ATTRIB_INDEX, f );
+   ATTR1FV( VBO_ATTRIB_INDEX, f );
 }
 
 
 static void GLAPIENTRY TAG(MultiTexCoord1f)( GLenum target, GLfloat x  )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR1F( attr, x );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord1fv)( GLenum target, const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR1FV( attr, v );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord2f)( GLenum target, GLfloat x, GLfloat y )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR2F( attr, x, y );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord2fv)( GLenum target, const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR2FV( attr, v );
 }
 
@@ -233,14 +233,14 @@ static void GLAPIENTRY TAG(MultiTexCoord3f)( GLenum target, GLfloat x, GLfloat y
 				    GLfloat z)
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR3F( attr, x, y, z );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord3fv)( GLenum target, const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR3FV( attr, v );
 }
 
@@ -248,14 +248,14 @@ static void GLAPIENTRY TAG(MultiTexCoord4f)( GLenum target, GLfloat x, GLfloat y
 				    GLfloat z, GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR4F( attr, x, y, z, w );
 }
 
 static void GLAPIENTRY TAG(MultiTexCoord4fv)( GLenum target, const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   GLuint attr = (target & 0x7) + BRW_ATTRIB_TEX0;
+   GLuint attr = (target & 0x7) + VBO_ATTRIB_TEX0;
    ATTR4FV( attr, v );
 }
 
@@ -266,7 +266,7 @@ static void GLAPIENTRY TAG(VertexAttrib1fARB)( GLuint index, GLfloat x )
    if (index == 0) 
       ATTR1F(0, x);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR1F(BRW_ATTRIB_GENERIC0 + index, x);
+      ATTR1F(VBO_ATTRIB_GENERIC0 + index, x);
    else
       ERROR();
 }
@@ -278,7 +278,7 @@ static void GLAPIENTRY TAG(VertexAttrib1fvARB)( GLuint index,
    if (index == 0) 
       ATTR1FV(0, v);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR1FV(BRW_ATTRIB_GENERIC0 + index, v);
+      ATTR1FV(VBO_ATTRIB_GENERIC0 + index, v);
    else
       ERROR();
 }
@@ -290,7 +290,7 @@ static void GLAPIENTRY TAG(VertexAttrib2fARB)( GLuint index, GLfloat x,
    if (index == 0) 
       ATTR2F(0, x, y);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR2F(BRW_ATTRIB_GENERIC0 + index, x, y);
+      ATTR2F(VBO_ATTRIB_GENERIC0 + index, x, y);
    else
       ERROR();
 }
@@ -302,7 +302,7 @@ static void GLAPIENTRY TAG(VertexAttrib2fvARB)( GLuint index,
    if (index == 0) 
       ATTR2FV(0, v);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR2FV(BRW_ATTRIB_GENERIC0 + index, v);
+      ATTR2FV(VBO_ATTRIB_GENERIC0 + index, v);
    else
       ERROR();
 }
@@ -314,7 +314,7 @@ static void GLAPIENTRY TAG(VertexAttrib3fARB)( GLuint index, GLfloat x,
    if (index == 0) 
       ATTR3F(0, x, y, z);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR3F(BRW_ATTRIB_GENERIC0 + index, x, y, z);
+      ATTR3F(VBO_ATTRIB_GENERIC0 + index, x, y, z);
    else
       ERROR();
 }
@@ -326,7 +326,7 @@ static void GLAPIENTRY TAG(VertexAttrib3fvARB)( GLuint index,
    if (index == 0) 
       ATTR3FV(0, v);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR3FV(BRW_ATTRIB_GENERIC0 + index, v);
+      ATTR3FV(VBO_ATTRIB_GENERIC0 + index, v);
    else
       ERROR();
 }
@@ -339,7 +339,7 @@ static void GLAPIENTRY TAG(VertexAttrib4fARB)( GLuint index, GLfloat x,
    if (index == 0) 
       ATTR4F(0, x, y, z, w);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR4F(BRW_ATTRIB_GENERIC0 + index, x, y, z, w);
+      ATTR4F(VBO_ATTRIB_GENERIC0 + index, x, y, z, w);
    else
       ERROR();
 }
@@ -351,20 +351,21 @@ static void GLAPIENTRY TAG(VertexAttrib4fvARB)( GLuint index,
    if (index == 0) 
       ATTR4FV(0, v);
    else if (index < MAX_VERTEX_ATTRIBS)
-      ATTR4FV(BRW_ATTRIB_GENERIC0 + index, v);
+      ATTR4FV(VBO_ATTRIB_GENERIC0 + index, v);
    else
       ERROR();
 }
 
 
-/* Although we don't export NV_vertex_program, these entrypoints are
+/* In addition to supporting NV_vertex_program, these entrypoints are
  * used by the display list and other code specifically because of
- * their property of aliasing with other attributes.
+ * their property of aliasing with other attributes.  (See
+ * vbo_save_loopback.c)
  */
 static void GLAPIENTRY TAG(VertexAttrib1fNV)( GLuint index, GLfloat x )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX)
+   if (index < VBO_ATTRIB_MAX)
       ATTR1F(index, x);
 }
 
@@ -372,7 +373,7 @@ static void GLAPIENTRY TAG(VertexAttrib1fvNV)( GLuint index,
 					       const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR1FV(index, v);
 }
 
@@ -380,7 +381,7 @@ static void GLAPIENTRY TAG(VertexAttrib2fNV)( GLuint index, GLfloat x,
 					      GLfloat y )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR2F(index, x, y);
 }
 
@@ -388,7 +389,7 @@ static void GLAPIENTRY TAG(VertexAttrib2fvNV)( GLuint index,
 					       const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR2FV(index, v);
 }
 
@@ -396,7 +397,7 @@ static void GLAPIENTRY TAG(VertexAttrib3fNV)( GLuint index, GLfloat x,
 					      GLfloat y, GLfloat z )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR3F(index, x, y, z);
 }
 
@@ -404,7 +405,7 @@ static void GLAPIENTRY TAG(VertexAttrib3fvNV)( GLuint index,
 					       const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR3FV(index, v);
 }
 
@@ -413,7 +414,7 @@ static void GLAPIENTRY TAG(VertexAttrib4fNV)( GLuint index, GLfloat x,
 					      GLfloat w )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR4F(index, x, y, z, w);
 }
 
@@ -421,7 +422,7 @@ static void GLAPIENTRY TAG(VertexAttrib4fvNV)( GLuint index,
 					       const GLfloat *v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   if (index < BRW_ATTRIB_MAX) 
+   if (index < VBO_ATTRIB_MAX) 
       ATTR4FV(index, v);
 }
 
@@ -443,26 +444,26 @@ static void GLAPIENTRY TAG(Materialfv)( GLenum face, GLenum pname,
    GET_CURRENT_CONTEXT( ctx );
    switch (pname) {
    case GL_EMISSION:
-      MAT( BRW_ATTRIB_MAT_FRONT_EMISSION, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_EMISSION, 4, face, params );
       break;
    case GL_AMBIENT:
-      MAT( BRW_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
       break;
    case GL_DIFFUSE:
-      MAT( BRW_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
       break;
    case GL_SPECULAR:
-      MAT( BRW_ATTRIB_MAT_FRONT_SPECULAR, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_SPECULAR, 4, face, params );
       break;
    case GL_SHININESS:
-      MAT( BRW_ATTRIB_MAT_FRONT_SHININESS, 1, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_SHININESS, 1, face, params );
       break;
    case GL_COLOR_INDEXES:
-      MAT( BRW_ATTRIB_MAT_FRONT_INDEXES, 3, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_INDEXES, 3, face, params );
       break;
    case GL_AMBIENT_AND_DIFFUSE:
-      MAT( BRW_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
-      MAT( BRW_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_AMBIENT, 4, face, params );
+      MAT( VBO_ATTRIB_MAT_FRONT_DIFFUSE, 4, face, params );
       break;
    default:
       ERROR();
diff --git a/src/mesa/vbo/vbo_context.c b/src/mesa/vbo/vbo_context.c
new file mode 100644
index 00000000000..65998e7ff75
--- /dev/null
+++ b/src/mesa/vbo/vbo_context.c
@@ -0,0 +1,246 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.3
+ *
+ * Copyright (C) 1999-2005  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"),
+ * 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 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
+ * BRIAN PAUL 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:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+#include "mtypes.h"
+#include "vbo.h"
+#include "vbo_context.h"
+#include "imports.h"
+#include "api_arrayelt.h"
+
+/* Reach out and grab this to use as the default:
+ */
+extern void _tnl_draw_prims( GLcontext *ctx,
+			     const struct gl_client_array *arrays[],
+			     const struct _mesa_prim *prims,
+			     GLuint nr_prims,
+			     const struct _mesa_index_buffer *ib,
+			     GLuint min_index,
+			     GLuint max_index );
+
+
+
+#define NR_LEGACY_ATTRIBS 16
+#define NR_GENERIC_ATTRIBS 16
+#define NR_MAT_ATTRIBS 12
+
+static GLuint check_size( const GLfloat *attr )
+{
+   if (attr[3] != 1.0) return 4;
+   if (attr[2] != 0.0) return 3;
+   if (attr[1] != 0.0) return 2;
+   return 1;		
+}
+
+static void init_legacy_currval(GLcontext *ctx)
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct gl_client_array *arrays = vbo->legacy_currval;
+   GLuint i;
+
+   memset(arrays, 0, sizeof(*arrays) * NR_LEGACY_ATTRIBS);
+
+   /* Set up a constant (StrideB == 0) array for each current
+    * attribute:
+    */
+   for (i = 0; i < NR_LEGACY_ATTRIBS; i++) {
+      struct gl_client_array *cl = &arrays[i];
+
+      /* Size will have to be determined at runtime:
+       */
+      cl->Size = check_size(ctx->Current.Attrib[i]);
+      cl->Stride = 0;
+      cl->StrideB = 0;
+      cl->Enabled = 1;
+      cl->Type = GL_FLOAT;
+      cl->Ptr = (const void *)ctx->Current.Attrib[i];
+      cl->BufferObj = ctx->Array.NullBufferObj;
+   }
+}
+
+
+static void init_generic_currval(GLcontext *ctx)
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct gl_client_array *arrays = vbo->generic_currval;
+   GLuint i;
+
+   memset(arrays, 0, sizeof(*arrays) * NR_GENERIC_ATTRIBS);
+
+   for (i = 0; i < NR_GENERIC_ATTRIBS; i++) {
+      struct gl_client_array *cl = &arrays[i];
+
+      /* This will have to be determined at runtime:
+       */
+      cl->Size = 1;
+      cl->Type = GL_FLOAT;
+      cl->Ptr = (const void *)ctx->Current.Attrib[VERT_ATTRIB_GENERIC0 + i];
+      cl->Stride = 0;
+      cl->StrideB = 0;
+      cl->Enabled = 1;
+      cl->BufferObj = ctx->Array.NullBufferObj;
+   }
+}
+
+
+static void init_mat_currval(GLcontext *ctx)
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct gl_client_array *arrays = vbo->mat_currval;
+   GLuint i;
+
+   memset(arrays, 0, sizeof(*arrays) * NR_MAT_ATTRIBS);
+
+   /* Set up a constant (StrideB == 0) array for each current
+    * attribute:
+    */
+   for (i = 0; i < NR_MAT_ATTRIBS; i++) {
+      struct gl_client_array *cl = &arrays[i];
+
+      /* Size is fixed for the material attributes, for others will
+       * be determined at runtime:
+       */
+      switch (i - VERT_ATTRIB_GENERIC0) {
+      case MAT_ATTRIB_FRONT_SHININESS:
+      case MAT_ATTRIB_BACK_SHININESS:
+	 cl->Size = 1;
+	 break;
+      case MAT_ATTRIB_FRONT_INDEXES:
+      case MAT_ATTRIB_BACK_INDEXES:
+	 cl->Size = 3;
+	 break;
+      default:
+	 cl->Size = 4;
+	 break;
+      }
+
+      if (i < MAT_ATTRIB_MAX)
+	 cl->Ptr = (const void *)ctx->Light.Material.Attrib[i];
+      else 
+	 cl->Ptr = (const void *)ctx->Current.Attrib[VERT_ATTRIB_GENERIC0 + i];
+
+      cl->Type = GL_FLOAT;
+      cl->Stride = 0;
+      cl->StrideB = 0;
+      cl->Enabled = 1;
+      cl->BufferObj = ctx->Array.NullBufferObj;
+   }
+}
+
+#if 0
+
+static void vbo_exec_current_init( struct vbo_exec_context *exec ) 
+{
+   GLcontext *ctx = exec->ctx;
+   GLint i;
+
+   /* setup the pointers for the typical 16 vertex attributes */
+   for (i = 0; i < VBO_ATTRIB_FIRST_MATERIAL; i++) 
+      exec->vtx.current[i] = ctx->Current.Attrib[i];
+
+   /* setup pointers for the 12 material attributes */
+   for (i = 0; i < MAT_ATTRIB_MAX; i++)
+      exec->vtx.current[VBO_ATTRIB_FIRST_MATERIAL + i] = 
+	 ctx->Light.Material.Attrib[i];
+}
+#endif
+
+GLboolean _vbo_CreateContext( GLcontext *ctx )
+{
+   struct vbo_context *vbo = CALLOC_STRUCT(vbo_context);
+
+   ctx->swtnl_im = (void *)vbo;
+
+   /* Initialize the arrayelt helper
+    */
+   if (!ctx->aelt_context &&
+       !_ae_create_context( ctx )) {
+      return GL_FALSE;
+   }
+
+   /* TODO: remove these pointers.
+    */
+   vbo->legacy_currval = &vbo->currval[VBO_ATTRIB_POS];
+   vbo->generic_currval = &vbo->currval[VBO_ATTRIB_GENERIC0];
+   vbo->mat_currval = &vbo->currval[VBO_ATTRIB_MAT_FRONT_AMBIENT];
+
+   init_legacy_currval( ctx );
+   init_generic_currval( ctx );
+   init_mat_currval( ctx );
+
+   /* Build mappings from VERT_ATTRIB -> VBO_ATTRIB depending on type
+    * of vertex program active.
+    */
+   {
+      GLuint i;
+
+      /* When no vertex program, pull in the material attributes in
+       * the 16..32 generic range.
+       */
+      for (i = 0; i < 16; i++) 
+	 vbo->map_vp_none[i] = i;
+      for (i = 0; i < 12; i++) 
+	 vbo->map_vp_none[16+i] = VBO_ATTRIB_MAT_FRONT_AMBIENT + i;
+      for (i = 0; i < 4; i++)
+	 vbo->map_vp_none[28+i] = i;	
+      
+      for (i = 0; i < VERT_ATTRIB_MAX; i++)
+	 vbo->map_vp_arb[i] = i;
+   }
+
+
+   /* By default: 
+    */
+   vbo->draw_prims = _tnl_draw_prims;
+
+   /* Hook our functions into exec and compile dispatch tables.  These
+    * will pretty much be permanently installed, which means that the
+    * vtxfmt mechanism can be removed now.
+    */
+   vbo_exec_init( ctx );
+   vbo_save_init( ctx );
+
+   
+   return GL_TRUE;
+}
+
+void _vbo_InvalidateState( GLcontext *ctx, GLuint new_state )
+{
+   _ae_invalidate_state(ctx, new_state);
+   vbo_exec_invalidate_state(ctx, new_state);
+}
+
+
+void _vbo_DestroyContext( GLcontext *ctx )
+{
+   if (ctx->aelt_context) {
+      _ae_destroy_context( ctx );
+      ctx->aelt_context = NULL;
+   }
+
+   FREE(vbo_context(ctx));
+   ctx->swtnl_im = NULL;
+}
diff --git a/src/mesa/vbo/vbo_context.h b/src/mesa/vbo/vbo_context.h
new file mode 100644
index 00000000000..0dc1019b39f
--- /dev/null
+++ b/src/mesa/vbo/vbo_context.h
@@ -0,0 +1,108 @@
+/*
+ * mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  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"),
+ * 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 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
+ * BRIAN PAUL 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.
+ */
+
+/**
+ * \file vbo_context.h
+ * \brief VBO builder module datatypes and definitions.
+ * \author Keith Whitwell
+ */
+
+
+/**
+ * \mainpage The VBO builder module
+ *
+ * This module hooks into the GL dispatch table and catches all vertex
+ * building and drawing commands, such as glVertex3f, glBegin and
+ * glDrawArrays.  The module stores all incoming vertex data as arrays
+ * in GL vertex buffer objects (VBOs), and translates all drawing
+ * commands into calls to a driver supplied DrawPrimitives() callback.
+ *
+ * The module captures both immediate mode and display list drawing,
+ * and manages the allocation, reference counting and deallocation of
+ * vertex buffer objects itself.
+ * 
+ * The DrawPrimitives() callback can be either implemented by the
+ * driver itself or hooked to the tnl module's _tnl_draw_primitives()
+ * function for hardware without tnl capablilties or during fallbacks.
+ */
+
+
+#ifndef _VBO_CONTEXT_H
+#define _VBO_CONTEXT_H
+
+#include "vbo.h"
+#include "vbo_attrib.h"
+#include "vbo_save.h"
+#include "vbo_exec.h"
+
+
+struct vbo_context {
+   struct gl_client_array currval[VBO_ATTRIB_MAX];
+   
+   /* These point into the above.  TODO: remove. 
+    */
+   struct gl_client_array *legacy_currval;
+   struct gl_client_array *generic_currval;
+   struct gl_client_array *mat_currval;
+
+   GLuint map_vp_none[32];
+   GLuint map_vp_arb[32];
+
+   GLfloat *current[VBO_ATTRIB_MAX]; /* points into ctx->Current, ctx->Light.Material */
+   GLfloat CurrentFloatEdgeFlag;
+
+
+   struct vbo_exec_context exec;
+   struct vbo_save_context save;
+
+   /* Callback into the driver.  This must always succeed, the driver
+    * is responsible for initiating any fallback actions required:
+    */
+   vbo_draw_func draw_prims;
+};
+
+
+static INLINE struct vbo_context *vbo_context(GLcontext *ctx) 
+{
+   return (struct vbo_context *)(ctx->swtnl_im);
+}
+
+enum {
+   VP_NONE = 1,
+   VP_NV,
+   VP_ARB
+};
+
+static INLINE GLuint get_program_mode( GLcontext *ctx )
+{
+   if (!ctx->VertexProgram._Enabled)
+      return VP_NONE;
+   else if (ctx->VertexProgram.Current->IsNVProgram)
+      return VP_NV;
+   else
+      return VP_ARB;
+}
+
+
+#endif
diff --git a/src/mesa/drivers/dri/i965/brw_exec.c b/src/mesa/vbo/vbo_exec.c
index fc06c3c3617..7d958732479 100644
--- a/src/mesa/drivers/dri/i965/brw_exec.c
+++ b/src/mesa/vbo/vbo_exec.c
@@ -35,19 +35,13 @@
 #include "dlist.h"
 #include "vtxfmt.h"
 
-#include "brw_exec.h"
+#include "vbo_context.h"
 
-
-void brw_exec_init( GLcontext *ctx )
+void vbo_exec_init( GLcontext *ctx )
 {
-   struct brw_exec_context *exec = CALLOC_STRUCT(brw_exec_context);
-
-   if (ctx->swtnl_im == NULL) {
-      ctx->swtnl_im = CALLOC_STRUCT(brw_exec_save);
-   }
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    exec->ctx = ctx;
-   IMM_CONTEXT(ctx)->exec = exec;
 
    /* Initialize the arrayelt helper
     */
@@ -55,48 +49,41 @@ void brw_exec_init( GLcontext *ctx )
        !_ae_create_context( ctx )) 
       return;
 
-   brw_exec_vtx_init( exec );
-   brw_exec_array_init( exec );
+   vbo_exec_vtx_init( exec );
+   vbo_exec_array_init( exec );
+
+   /* Hook our functions into exec and compile dispatch tables.
+    */
+   _mesa_install_exec_vtxfmt( ctx, &exec->vtxfmt );
 
    ctx->Driver.NeedFlush = 0;
    ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
-   ctx->Driver.FlushVertices = brw_exec_FlushVertices;
+   ctx->Driver.FlushVertices = vbo_exec_FlushVertices;
 
-   exec->eval.recalculate_maps = 1;
+   vbo_exec_invalidate_state( ctx, ~0 );
 }
 
 
-void brw_exec_destroy( GLcontext *ctx )
+void vbo_exec_destroy( GLcontext *ctx )
 {
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    if (ctx->aelt_context) {
       _ae_destroy_context( ctx );
       ctx->aelt_context = NULL;
    }
 
-   brw_exec_vtx_destroy( exec );
-   brw_exec_array_destroy( exec );
-
-   if (exec) {
-      FREE(exec);
-      IMM_CONTEXT(ctx)->exec = NULL;
-   }
-
-   if (IMM_CONTEXT(ctx)->exec == NULL &&
-       IMM_CONTEXT(ctx)->save == NULL) {
-      FREE(IMM_CONTEXT(ctx));
-      ctx->swtnl_im = NULL;
-   }
+   vbo_exec_vtx_destroy( exec );
+   vbo_exec_array_destroy( exec );
 }
 
 /* Really want to install these callbacks to a central facility to be
  * invoked according to the state flags.  That will have to wait for a
  * mesa rework:
  */ 
-void brw_exec_invalidate_state( GLcontext *ctx, GLuint new_state )
+void vbo_exec_invalidate_state( GLcontext *ctx, GLuint new_state )
 {
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    if (new_state & (_NEW_PROGRAM|_NEW_EVAL))
       exec->eval.recalculate_maps = 1;
@@ -105,21 +92,6 @@ void brw_exec_invalidate_state( GLcontext *ctx, GLuint new_state )
 }
 
 
-void brw_exec_wakeup( GLcontext *ctx )
-{
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
-
-   ctx->Driver.FlushVertices = brw_exec_FlushVertices;
-   ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
-
-   /* Hook our functions into exec and compile dispatch tables.
-    */
-   _mesa_install_exec_vtxfmt( ctx, &exec->vtxfmt );
-
-   /* Assume we haven't been getting state updates either:
-    */
-   brw_exec_invalidate_state( ctx, ~0 );
-}
 
 
 
diff --git a/src/mesa/vbo/vbo_exec.h b/src/mesa/vbo/vbo_exec.h
new file mode 100644
index 00000000000..a9b01e08e6a
--- /dev/null
+++ b/src/mesa/vbo/vbo_exec.h
@@ -0,0 +1,165 @@
+/**************************************************************************
+
+Copyright 2002 Tungsten Graphics Inc., Cedar Park, Texas.
+
+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"),
+to deal in the Software without restriction, including without limitation
+on the rights to use, copy, modify, merge, publish, distribute, sub
+license, 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+TUNGSTEN GRAPHICS AND/OR THEIR SUPPLIERS 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:
+ *   Keith Whitwell <keith@tungstengraphics.com>
+ *
+ */
+
+#ifndef __VBO_EXEC_H__
+#define __VBO_EXEC_H__
+
+#include "mtypes.h"
+#include "vbo.h"
+#include "vbo_attrib.h"
+
+
+#define VBO_MAX_PRIM 64
+
+/* Wierd implementation stuff:
+ */
+#define VBO_VERT_BUFFER_SIZE (1024*16)	/* dwords == 64k */
+#define VBO_MAX_ATTR_CODEGEN 16 
+#define ERROR_ATTRIB 16
+
+
+
+
+struct vbo_exec_eval1_map {
+   struct gl_1d_map *map;
+   GLuint sz;
+};
+
+struct vbo_exec_eval2_map {
+   struct gl_2d_map *map;
+   GLuint sz;
+};
+
+
+
+struct vbo_exec_copied_vtx {
+   GLfloat buffer[VBO_ATTRIB_MAX * 4 * VBO_MAX_COPIED_VERTS];
+   GLuint nr;
+};
+
+
+typedef void (*vbo_attrfv_func)( const GLfloat * );
+
+
+struct vbo_exec_context
+{
+   GLcontext *ctx;   
+   GLvertexformat vtxfmt;
+
+   struct {
+      struct gl_buffer_object *bufferobj;
+      GLubyte *buffer_map;
+
+      GLuint vertex_size;
+
+      struct _mesa_prim prim[VBO_MAX_PRIM];
+      GLuint prim_count;
+
+      GLfloat *vbptr;		     /* cursor, points into buffer */
+      GLfloat vertex[VBO_ATTRIB_MAX*4]; /* current vertex */
+
+      GLuint vert_count;
+      GLuint max_vert;
+      struct vbo_exec_copied_vtx copied;
+
+      GLubyte attrsz[VBO_ATTRIB_MAX];
+      GLubyte active_sz[VBO_ATTRIB_MAX];
+
+      GLfloat *attrptr[VBO_ATTRIB_MAX]; 
+      struct gl_client_array arrays[VBO_ATTRIB_MAX];
+
+      /* According to program mode, the values above plus current
+       * values are squashed down to the 32 attributes passed to the
+       * vertex program below:
+       */
+      GLuint program_mode;
+      GLuint enabled_flags;
+      const struct gl_client_array *inputs[VERT_ATTRIB_MAX];
+   } vtx;
+
+   
+   struct {
+      GLboolean recalculate_maps;
+      struct vbo_exec_eval1_map map1[VERT_ATTRIB_MAX];
+      struct vbo_exec_eval2_map map2[VERT_ATTRIB_MAX];
+   } eval;
+
+   struct {
+      GLuint program_mode;
+      GLuint enabled_flags;
+      GLuint array_obj;
+
+      /* These just mirror the current arrayobj (todo: make arrayobj
+       * look like this and remove the mirror):
+       */
+      const struct gl_client_array *legacy_array[16];
+      const struct gl_client_array *generic_array[16];
+
+      /* Arrays and current values manipulated according to program
+       * mode, etc.  These are the attributes as seen by vertex
+       * programs:
+       */
+      const struct gl_client_array *inputs[VERT_ATTRIB_MAX];
+   } array;
+};
+
+
+
+/* External API:
+ */
+void vbo_exec_init( GLcontext *ctx );
+void vbo_exec_destroy( GLcontext *ctx );
+void vbo_exec_invalidate_state( GLcontext *ctx, GLuint new_state );
+void vbo_exec_FlushVertices( GLcontext *ctx, GLuint flags );
+
+
+/* Internal functions:
+ */
+void vbo_exec_array_init( struct vbo_exec_context *exec );
+void vbo_exec_array_destroy( struct vbo_exec_context *exec );
+
+
+void vbo_exec_vtx_init( struct vbo_exec_context *exec );
+void vbo_exec_vtx_destroy( struct vbo_exec_context *exec );
+void vbo_exec_vtx_flush( struct vbo_exec_context *exec );
+void vbo_exec_vtx_wrap( struct vbo_exec_context *exec );
+
+void vbo_exec_eval_update( struct vbo_exec_context *exec );
+
+void vbo_exec_do_EvalCoord2f( struct vbo_exec_context *exec, 
+				     GLfloat u, GLfloat v );
+
+void vbo_exec_do_EvalCoord1f( struct vbo_exec_context *exec,
+				     GLfloat u);
+
+#endif
diff --git a/src/mesa/drivers/dri/i965/brw_exec_api.c b/src/mesa/vbo/vbo_exec_api.c
index 470fa6f4177..71fee8ca16e 100644
--- a/src/mesa/drivers/dri/i965/brw_exec_api.c
+++ b/src/mesa/vbo/vbo_exec_api.c
@@ -41,15 +41,15 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "api_noop.h"
 #include "dispatch.h"
 
-#include "brw_exec.h"
+#include "vbo_context.h"
 
-static void reset_attrfv( struct brw_exec_context *exec );
+static void reset_attrfv( struct vbo_exec_context *exec );
 
 
 /* Close off the last primitive, execute the buffer, restart the
  * primitive.  
  */
-static void brw_exec_wrap_buffers( struct brw_exec_context *exec )
+static void vbo_exec_wrap_buffers( struct vbo_exec_context *exec )
 {
    if (exec->vtx.prim_count == 0) {
       exec->vtx.copied.nr = 0;
@@ -72,7 +72,7 @@ static void brw_exec_wrap_buffers( struct brw_exec_context *exec )
       /* Execute the buffer and save copied vertices.
        */
       if (exec->vtx.vert_count)
-	 brw_exec_vtx_flush( exec );
+	 vbo_exec_vtx_flush( exec );
       else {
 	 exec->vtx.prim_count = 0;
 	 exec->vtx.copied.nr = 0;
@@ -98,7 +98,7 @@ static void brw_exec_wrap_buffers( struct brw_exec_context *exec )
 /* Deal with buffer wrapping where provoked by the vertex buffer
  * filling up, as opposed to upgrade_vertex().
  */
-void brw_exec_vtx_wrap( struct brw_exec_context *exec )
+void vbo_exec_vtx_wrap( struct vbo_exec_context *exec )
 {
    GLfloat *data = exec->vtx.copied.buffer;
    GLuint i;
@@ -106,7 +106,7 @@ void brw_exec_vtx_wrap( struct brw_exec_context *exec )
    /* Run pipeline on current vertices, copy wrapped vertices
     * to exec->vtx.copied.
     */
-   brw_exec_wrap_buffers( exec );
+   vbo_exec_wrap_buffers( exec );
    
    /* Copy stored stored vertices to start of new list. 
     */
@@ -127,72 +127,68 @@ void brw_exec_vtx_wrap( struct brw_exec_context *exec )
 /*
  * Copy the active vertex's values to the ctx->Current fields.
  */
-static void brw_exec_copy_to_current( struct brw_exec_context *exec )
+static void vbo_exec_copy_to_current( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLuint i;
 
-   for (i = BRW_ATTRIB_POS+1 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       if (exec->vtx.attrsz[i]) {
+	 GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
+
          /* Note: the exec->vtx.current[i] pointers point into the
           * ctx->Current.Attrib and ctx->Light.Material.Attrib arrays.
           */
-	 COPY_CLEAN_4V(exec->vtx.current[i], 
+	 COPY_CLEAN_4V(current, 
 		       exec->vtx.attrsz[i], 
 		       exec->vtx.attrptr[i]);
 
+	 
+	 /* Given that we explicitly state size here, there is no need
+	  * for the COPY_CLEAN above, could just copy 16 bytes and be
+	  * done.  The only problem is when Mesa accesses ctx->Current
+	  * directly.
+	  */
+	 vbo->currval[i].Size = exec->vtx.attrsz[i];
+
 	 /* This triggers rather too much recalculation of Mesa state
 	  * that doesn't get used (eg light positions).
 	  */
-	 if (i >= BRW_ATTRIB_MAT_FRONT_AMBIENT &&
-	     i <= BRW_ATTRIB_MAT_BACK_INDEXES)
+	 if (i >= VBO_ATTRIB_MAT_FRONT_AMBIENT &&
+	     i <= VBO_ATTRIB_MAT_BACK_INDEXES)
 	    ctx->NewState |= _NEW_LIGHT;
       }
    }
 
-   /* color index is special (it's not a float[4] so COPY_CLEAN_4V above
-    * will trash adjacent memory!)
-    */
-   if (exec->vtx.attrsz[BRW_ATTRIB_INDEX]) {
-      ctx->Current.Index = exec->vtx.attrptr[BRW_ATTRIB_INDEX][0];
-   }
-
-   /* Edgeflag requires additional treatment:
-    */
-   if (exec->vtx.attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = (exec->vtx.CurrentFloatEdgeFlag == 1.0);
-   }
-
    /* Colormaterial -- this kindof sucks.
     */
    if (ctx->Light.ColorMaterialEnabled &&
-       exec->vtx.attrsz[BRW_ATTRIB_COLOR0]) {
+       exec->vtx.attrsz[VBO_ATTRIB_COLOR0]) {
       _mesa_update_color_material(ctx, 
-				  ctx->Current.Attrib[BRW_ATTRIB_COLOR0]);
+				  ctx->Current.Attrib[VBO_ATTRIB_COLOR0]);
    }
 
    ctx->Driver.NeedFlush &= ~FLUSH_UPDATE_CURRENT;
 }
 
 
-static void brw_exec_copy_from_current( struct brw_exec_context *exec )
+static void vbo_exec_copy_from_current( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLint i;
 
-   /* Edgeflag requires additional treatment:
-    */
-   exec->vtx.CurrentFloatEdgeFlag = 
-      (GLfloat)ctx->Current.EdgeFlag;
-   
-   for (i = BRW_ATTRIB_POS+1 ; i < BRW_ATTRIB_MAX ; i++) 
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
+      const GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
       switch (exec->vtx.attrsz[i]) {
-      case 4: exec->vtx.attrptr[i][3] = exec->vtx.current[i][3];
-      case 3: exec->vtx.attrptr[i][2] = exec->vtx.current[i][2];
-      case 2: exec->vtx.attrptr[i][1] = exec->vtx.current[i][1];
-      case 1: exec->vtx.attrptr[i][0] = exec->vtx.current[i][0];
+      case 4: exec->vtx.attrptr[i][3] = current[3];
+      case 3: exec->vtx.attrptr[i][2] = current[2];
+      case 2: exec->vtx.attrptr[i][1] = current[1];
+      case 1: exec->vtx.attrptr[i][0] = current[0];
 	 break;
       }
+   }
 
    ctx->Driver.NeedFlush |= FLUSH_UPDATE_CURRENT;
 }
@@ -200,11 +196,12 @@ static void brw_exec_copy_from_current( struct brw_exec_context *exec )
 
 /* Flush existing data, set new attrib size, replay copied vertices.
  */ 
-static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
+static void vbo_exec_wrap_upgrade_vertex( struct vbo_exec_context *exec,
 					  GLuint attr,
 					  GLuint newsz )
 {
    GLcontext *ctx = exec->ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLint lastcount = exec->vtx.vert_count;
    GLfloat *tmp;
    GLuint oldsz;
@@ -213,14 +210,14 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
    /* Run pipeline on current vertices, copy wrapped vertices
     * to exec->vtx.copied.
     */
-   brw_exec_wrap_buffers( exec );
+   vbo_exec_wrap_buffers( exec );
 
 
    /* Do a COPY_TO_CURRENT to ensure back-copying works for the case
     * when the attribute already exists in the vertex and is having
     * its size increased.  
     */
-   brw_exec_copy_to_current( exec );
+   vbo_exec_copy_to_current( exec );
 
 
    /* Heuristic: Attempt to isolate attributes received outside
@@ -239,14 +236,14 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
    exec->vtx.attrsz[attr] = newsz;
 
    exec->vtx.vertex_size += newsz - oldsz;
-   exec->vtx.max_vert = BRW_VERT_BUFFER_SIZE / exec->vtx.vertex_size;
+   exec->vtx.max_vert = VBO_VERT_BUFFER_SIZE / exec->vtx.vertex_size;
    exec->vtx.vert_count = 0;
    exec->vtx.vbptr = (GLfloat *)exec->vtx.buffer_map;
    
 
    /* Recalculate all the attrptr[] values
     */
-   for (i = 0, tmp = exec->vtx.vertex ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0, tmp = exec->vtx.vertex ; i < VBO_ATTRIB_MAX ; i++) {
       if (exec->vtx.attrsz[i]) {
 	 exec->vtx.attrptr[i] = tmp;
 	 tmp += exec->vtx.attrsz[i];
@@ -257,7 +254,7 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
 
    /* Copy from current to repopulate the vertex with correct values.
     */
-   brw_exec_copy_from_current( exec );
+   vbo_exec_copy_from_current( exec );
 
    /* Replay stored vertices to translate them
     * to new format here.
@@ -273,7 +270,7 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
       assert(exec->vtx.vbptr == (GLfloat *)exec->vtx.buffer_map);
       
       for (i = 0 ; i < exec->vtx.copied.nr ; i++) {
-	 for (j = 0 ; j < BRW_ATTRIB_MAX ; j++) {
+	 for (j = 0 ; j < VBO_ATTRIB_MAX ; j++) {
 	    if (exec->vtx.attrsz[j]) {
 	       if (j == attr) {
 		  if (oldsz) {
@@ -281,7 +278,8 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
 		     data += oldsz;
 		     dest += newsz;
 		  } else {
-		     COPY_SZ_4V( dest, newsz, exec->vtx.current[j] );
+		     const GLfloat *current = (const GLfloat *)vbo->currval[j].Ptr;
+		     COPY_SZ_4V( dest, newsz, current );
 		     dest += newsz;
 		  }
 	       }
@@ -302,17 +300,17 @@ static void brw_exec_wrap_upgrade_vertex( struct brw_exec_context *exec,
 }
 
 
-static void brw_exec_fixup_vertex( GLcontext *ctx,
+static void vbo_exec_fixup_vertex( GLcontext *ctx,
 				   GLuint attr, GLuint sz )
 {
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
    int i;
 
    if (sz > exec->vtx.attrsz[attr]) {
       /* New size is larger.  Need to flush existing vertices and get
        * an enlarged vertex format.
        */
-      brw_exec_wrap_upgrade_vertex( exec, attr, sz );
+      vbo_exec_wrap_upgrade_vertex( exec, attr, sz );
    }
    else if (sz < exec->vtx.active_sz[attr]) {
       static const GLfloat id[4] = { 0, 0, 0, 1 };
@@ -343,10 +341,10 @@ static void brw_exec_fixup_vertex( GLcontext *ctx,
  */
 #define ATTR( A, N, V0, V1, V2, V3 )				\
 do {								\
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;	\
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;	\
 								\
    if (exec->vtx.active_sz[A] != N)				\
-      brw_exec_fixup_vertex(ctx, A, N);			\
+      vbo_exec_fixup_vertex(ctx, A, N);			\
 								\
    {								\
       GLfloat *dest = exec->vtx.attrptr[A];			\
@@ -366,15 +364,15 @@ do {								\
       exec->ctx->Driver.NeedFlush |= FLUSH_STORED_VERTICES;	\
 								\
       if (++exec->vtx.vert_count >= exec->vtx.max_vert)		\
-	 brw_exec_vtx_wrap( exec );				\
+	 vbo_exec_vtx_wrap( exec );				\
    }								\
 } while (0)
 
 
 #define ERROR() _mesa_error( ctx, GL_INVALID_ENUM, __FUNCTION__ )
-#define TAG(x) brw_##x
+#define TAG(x) vbo_##x
 
-#include "brw_attrib_tmp.h"
+#include "vbo_attrib_tmp.h"
 
 
 
@@ -382,20 +380,20 @@ do {								\
 
 /* Eval
  */
-static void GLAPIENTRY brw_exec_EvalCoord1f( GLfloat u )
+static void GLAPIENTRY vbo_exec_EvalCoord1f( GLfloat u )
 {
    GET_CURRENT_CONTEXT( ctx );
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    {
       GLint i;
       if (exec->eval.recalculate_maps) 
-	 brw_exec_eval_update( exec );
+	 vbo_exec_eval_update( exec );
 
-      for (i = 0 ; i <= BRW_ATTRIB_INDEX ; i++) {
+      for (i = 0 ; i <= VBO_ATTRIB_INDEX ; i++) {
 	 if (exec->eval.map1[i].map) 
 	    if (exec->vtx.active_sz[i] != exec->eval.map1[i].sz)
-	       brw_exec_fixup_vertex( ctx, i, exec->eval.map1[i].sz );
+	       vbo_exec_fixup_vertex( ctx, i, exec->eval.map1[i].sz );
       }
    }
 
@@ -403,64 +401,64 @@ static void GLAPIENTRY brw_exec_EvalCoord1f( GLfloat u )
    _mesa_memcpy( exec->vtx.copied.buffer, exec->vtx.vertex, 
                  exec->vtx.vertex_size * sizeof(GLfloat));
 
-   brw_exec_do_EvalCoord1f( exec, u );
+   vbo_exec_do_EvalCoord1f( exec, u );
 
    _mesa_memcpy( exec->vtx.vertex, exec->vtx.copied.buffer,
                  exec->vtx.vertex_size * sizeof(GLfloat));
 }
 
-static void GLAPIENTRY brw_exec_EvalCoord2f( GLfloat u, GLfloat v )
+static void GLAPIENTRY vbo_exec_EvalCoord2f( GLfloat u, GLfloat v )
 {
    GET_CURRENT_CONTEXT( ctx );
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    {
       GLint i;
       if (exec->eval.recalculate_maps) 
-	 brw_exec_eval_update( exec );
+	 vbo_exec_eval_update( exec );
 
-      for (i = 0 ; i <= BRW_ATTRIB_INDEX ; i++) {
+      for (i = 0 ; i <= VBO_ATTRIB_INDEX ; i++) {
 	 if (exec->eval.map2[i].map) 
 	    if (exec->vtx.active_sz[i] != exec->eval.map2[i].sz)
-	       brw_exec_fixup_vertex( ctx, i, exec->eval.map2[i].sz );
+	       vbo_exec_fixup_vertex( ctx, i, exec->eval.map2[i].sz );
       }
 
       if (ctx->Eval.AutoNormal) 
-	 if (exec->vtx.active_sz[BRW_ATTRIB_NORMAL] != 3)
-	    brw_exec_fixup_vertex( ctx, BRW_ATTRIB_NORMAL, 3 );
+	 if (exec->vtx.active_sz[VBO_ATTRIB_NORMAL] != 3)
+	    vbo_exec_fixup_vertex( ctx, VBO_ATTRIB_NORMAL, 3 );
    }
 
    _mesa_memcpy( exec->vtx.copied.buffer, exec->vtx.vertex, 
                  exec->vtx.vertex_size * sizeof(GLfloat));
 
-   brw_exec_do_EvalCoord2f( exec, u, v );
+   vbo_exec_do_EvalCoord2f( exec, u, v );
 
    _mesa_memcpy( exec->vtx.vertex, exec->vtx.copied.buffer, 
                  exec->vtx.vertex_size * sizeof(GLfloat));
 }
 
-static void GLAPIENTRY brw_exec_EvalCoord1fv( const GLfloat *u )
+static void GLAPIENTRY vbo_exec_EvalCoord1fv( const GLfloat *u )
 {
-   brw_exec_EvalCoord1f( u[0] );
+   vbo_exec_EvalCoord1f( u[0] );
 }
 
-static void GLAPIENTRY brw_exec_EvalCoord2fv( const GLfloat *u )
+static void GLAPIENTRY vbo_exec_EvalCoord2fv( const GLfloat *u )
 {
-   brw_exec_EvalCoord2f( u[0], u[1] );
+   vbo_exec_EvalCoord2f( u[0], u[1] );
 }
 
-static void GLAPIENTRY brw_exec_EvalPoint1( GLint i )
+static void GLAPIENTRY vbo_exec_EvalPoint1( GLint i )
 {
    GET_CURRENT_CONTEXT( ctx );
    GLfloat du = ((ctx->Eval.MapGrid1u2 - ctx->Eval.MapGrid1u1) /
 		 (GLfloat) ctx->Eval.MapGrid1un);
    GLfloat u = i * du + ctx->Eval.MapGrid1u1;
 
-   brw_exec_EvalCoord1f( u );
+   vbo_exec_EvalCoord1f( u );
 }
 
 
-static void GLAPIENTRY brw_exec_EvalPoint2( GLint i, GLint j )
+static void GLAPIENTRY vbo_exec_EvalPoint2( GLint i, GLint j )
 {
    GET_CURRENT_CONTEXT( ctx );
    GLfloat du = ((ctx->Eval.MapGrid2u2 - ctx->Eval.MapGrid2u1) / 
@@ -470,19 +468,19 @@ static void GLAPIENTRY brw_exec_EvalPoint2( GLint i, GLint j )
    GLfloat u = i * du + ctx->Eval.MapGrid2u1;
    GLfloat v = j * dv + ctx->Eval.MapGrid2v1;
 
-   brw_exec_EvalCoord2f( u, v );
+   vbo_exec_EvalCoord2f( u, v );
 }
 
 
 /* Build a list of primitives on the fly.  Keep
  * ctx->Driver.CurrentExecPrimitive uptodate as well.
  */
-static void GLAPIENTRY brw_exec_Begin( GLenum mode )
+static void GLAPIENTRY vbo_exec_Begin( GLenum mode )
 {
    GET_CURRENT_CONTEXT( ctx ); 
 
    if (ctx->Driver.CurrentExecPrimitive == GL_POLYGON+1) {
-      struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+      struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
       int i;
 
       if (ctx->NewState) {
@@ -503,7 +501,7 @@ static void GLAPIENTRY brw_exec_Begin( GLenum mode )
        * begin/end pairs.
        */
       if (exec->vtx.vertex_size && !exec->vtx.attrsz[0]) 
-	 brw_exec_FlushVertices( ctx, ~0 );
+	 vbo_exec_FlushVertices( ctx, ~0 );
 
       i = exec->vtx.prim_count++;
       exec->vtx.prim[i].mode = mode;
@@ -522,12 +520,12 @@ static void GLAPIENTRY brw_exec_Begin( GLenum mode )
       
 }
 
-static void GLAPIENTRY brw_exec_End( void )
+static void GLAPIENTRY vbo_exec_End( void )
 {
    GET_CURRENT_CONTEXT( ctx ); 
 
    if (ctx->Driver.CurrentExecPrimitive != GL_POLYGON+1) {
-      struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+      struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
       int idx = exec->vtx.vert_count;
       int i = exec->vtx.prim_count - 1;
 
@@ -536,29 +534,29 @@ static void GLAPIENTRY brw_exec_End( void )
 
       ctx->Driver.CurrentExecPrimitive = GL_POLYGON+1;
 
-      if (exec->vtx.prim_count == BRW_MAX_PRIM)
-	 brw_exec_vtx_flush( exec );	
+      if (exec->vtx.prim_count == VBO_MAX_PRIM)
+	 vbo_exec_vtx_flush( exec );	
    }
    else 
       _mesa_error( ctx, GL_INVALID_OPERATION, "glEnd" );
 }
 
 
-static void brw_exec_vtxfmt_init( struct brw_exec_context *exec )
+static void vbo_exec_vtxfmt_init( struct vbo_exec_context *exec )
 {
    GLvertexformat *vfmt = &exec->vtxfmt;
 
    vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
-   vfmt->Begin = brw_exec_Begin;
+   vfmt->Begin = vbo_exec_Begin;
    vfmt->CallList = _mesa_CallList;
    vfmt->CallLists = _mesa_CallLists;
-   vfmt->End = brw_exec_End;
-   vfmt->EvalCoord1f = brw_exec_EvalCoord1f;
-   vfmt->EvalCoord1fv = brw_exec_EvalCoord1fv;
-   vfmt->EvalCoord2f = brw_exec_EvalCoord2f;
-   vfmt->EvalCoord2fv = brw_exec_EvalCoord2fv;
-   vfmt->EvalPoint1 = brw_exec_EvalPoint1;
-   vfmt->EvalPoint2 = brw_exec_EvalPoint2;
+   vfmt->End = vbo_exec_End;
+   vfmt->EvalCoord1f = vbo_exec_EvalCoord1f;
+   vfmt->EvalCoord1fv = vbo_exec_EvalCoord1fv;
+   vfmt->EvalCoord2f = vbo_exec_EvalCoord2f;
+   vfmt->EvalCoord2fv = vbo_exec_EvalCoord2fv;
+   vfmt->EvalPoint1 = vbo_exec_EvalPoint1;
+   vfmt->EvalPoint2 = vbo_exec_EvalPoint2;
 
    vfmt->Rectf = _mesa_noop_Rectf;
    vfmt->EvalMesh1 = _mesa_noop_EvalMesh1;
@@ -567,113 +565,101 @@ static void brw_exec_vtxfmt_init( struct brw_exec_context *exec )
 
    /* from attrib_tmp.h:
     */
-   vfmt->Color3f = brw_Color3f;
-   vfmt->Color3fv = brw_Color3fv;
-   vfmt->Color4f = brw_Color4f;
-   vfmt->Color4fv = brw_Color4fv;
-   vfmt->FogCoordfEXT = brw_FogCoordfEXT;
-   vfmt->FogCoordfvEXT = brw_FogCoordfvEXT;
-   vfmt->MultiTexCoord1fARB = brw_MultiTexCoord1f;
-   vfmt->MultiTexCoord1fvARB = brw_MultiTexCoord1fv;
-   vfmt->MultiTexCoord2fARB = brw_MultiTexCoord2f;
-   vfmt->MultiTexCoord2fvARB = brw_MultiTexCoord2fv;
-   vfmt->MultiTexCoord3fARB = brw_MultiTexCoord3f;
-   vfmt->MultiTexCoord3fvARB = brw_MultiTexCoord3fv;
-   vfmt->MultiTexCoord4fARB = brw_MultiTexCoord4f;
-   vfmt->MultiTexCoord4fvARB = brw_MultiTexCoord4fv;
-   vfmt->Normal3f = brw_Normal3f;
-   vfmt->Normal3fv = brw_Normal3fv;
-   vfmt->SecondaryColor3fEXT = brw_SecondaryColor3fEXT;
-   vfmt->SecondaryColor3fvEXT = brw_SecondaryColor3fvEXT;
-   vfmt->TexCoord1f = brw_TexCoord1f;
-   vfmt->TexCoord1fv = brw_TexCoord1fv;
-   vfmt->TexCoord2f = brw_TexCoord2f;
-   vfmt->TexCoord2fv = brw_TexCoord2fv;
-   vfmt->TexCoord3f = brw_TexCoord3f;
-   vfmt->TexCoord3fv = brw_TexCoord3fv;
-   vfmt->TexCoord4f = brw_TexCoord4f;
-   vfmt->TexCoord4fv = brw_TexCoord4fv;
-   vfmt->Vertex2f = brw_Vertex2f;
-   vfmt->Vertex2fv = brw_Vertex2fv;
-   vfmt->Vertex3f = brw_Vertex3f;
-   vfmt->Vertex3fv = brw_Vertex3fv;
-   vfmt->Vertex4f = brw_Vertex4f;
-   vfmt->Vertex4fv = brw_Vertex4fv;
+   vfmt->Color3f = vbo_Color3f;
+   vfmt->Color3fv = vbo_Color3fv;
+   vfmt->Color4f = vbo_Color4f;
+   vfmt->Color4fv = vbo_Color4fv;
+   vfmt->FogCoordfEXT = vbo_FogCoordfEXT;
+   vfmt->FogCoordfvEXT = vbo_FogCoordfvEXT;
+   vfmt->MultiTexCoord1fARB = vbo_MultiTexCoord1f;
+   vfmt->MultiTexCoord1fvARB = vbo_MultiTexCoord1fv;
+   vfmt->MultiTexCoord2fARB = vbo_MultiTexCoord2f;
+   vfmt->MultiTexCoord2fvARB = vbo_MultiTexCoord2fv;
+   vfmt->MultiTexCoord3fARB = vbo_MultiTexCoord3f;
+   vfmt->MultiTexCoord3fvARB = vbo_MultiTexCoord3fv;
+   vfmt->MultiTexCoord4fARB = vbo_MultiTexCoord4f;
+   vfmt->MultiTexCoord4fvARB = vbo_MultiTexCoord4fv;
+   vfmt->Normal3f = vbo_Normal3f;
+   vfmt->Normal3fv = vbo_Normal3fv;
+   vfmt->SecondaryColor3fEXT = vbo_SecondaryColor3fEXT;
+   vfmt->SecondaryColor3fvEXT = vbo_SecondaryColor3fvEXT;
+   vfmt->TexCoord1f = vbo_TexCoord1f;
+   vfmt->TexCoord1fv = vbo_TexCoord1fv;
+   vfmt->TexCoord2f = vbo_TexCoord2f;
+   vfmt->TexCoord2fv = vbo_TexCoord2fv;
+   vfmt->TexCoord3f = vbo_TexCoord3f;
+   vfmt->TexCoord3fv = vbo_TexCoord3fv;
+   vfmt->TexCoord4f = vbo_TexCoord4f;
+   vfmt->TexCoord4fv = vbo_TexCoord4fv;
+   vfmt->Vertex2f = vbo_Vertex2f;
+   vfmt->Vertex2fv = vbo_Vertex2fv;
+   vfmt->Vertex3f = vbo_Vertex3f;
+   vfmt->Vertex3fv = vbo_Vertex3fv;
+   vfmt->Vertex4f = vbo_Vertex4f;
+   vfmt->Vertex4fv = vbo_Vertex4fv;
    
-   vfmt->VertexAttrib1fARB = brw_VertexAttrib1fARB;
-   vfmt->VertexAttrib1fvARB = brw_VertexAttrib1fvARB;
-   vfmt->VertexAttrib2fARB = brw_VertexAttrib2fARB;
-   vfmt->VertexAttrib2fvARB = brw_VertexAttrib2fvARB;
-   vfmt->VertexAttrib3fARB = brw_VertexAttrib3fARB;
-   vfmt->VertexAttrib3fvARB = brw_VertexAttrib3fvARB;
-   vfmt->VertexAttrib4fARB = brw_VertexAttrib4fARB;
-   vfmt->VertexAttrib4fvARB = brw_VertexAttrib4fvARB;
-
-   vfmt->VertexAttrib1fNV = brw_VertexAttrib1fNV;
-   vfmt->VertexAttrib1fvNV = brw_VertexAttrib1fvNV;
-   vfmt->VertexAttrib2fNV = brw_VertexAttrib2fNV;
-   vfmt->VertexAttrib2fvNV = brw_VertexAttrib2fvNV;
-   vfmt->VertexAttrib3fNV = brw_VertexAttrib3fNV;
-   vfmt->VertexAttrib3fvNV = brw_VertexAttrib3fvNV;
-   vfmt->VertexAttrib4fNV = brw_VertexAttrib4fNV;
-   vfmt->VertexAttrib4fvNV = brw_VertexAttrib4fvNV;
-
-   vfmt->Materialfv = brw_Materialfv;
-
-   vfmt->EdgeFlag = brw_EdgeFlag;
-   vfmt->Indexf = brw_Indexf;
-   vfmt->Indexfv = brw_Indexfv;
+   vfmt->VertexAttrib1fARB = vbo_VertexAttrib1fARB;
+   vfmt->VertexAttrib1fvARB = vbo_VertexAttrib1fvARB;
+   vfmt->VertexAttrib2fARB = vbo_VertexAttrib2fARB;
+   vfmt->VertexAttrib2fvARB = vbo_VertexAttrib2fvARB;
+   vfmt->VertexAttrib3fARB = vbo_VertexAttrib3fARB;
+   vfmt->VertexAttrib3fvARB = vbo_VertexAttrib3fvARB;
+   vfmt->VertexAttrib4fARB = vbo_VertexAttrib4fARB;
+   vfmt->VertexAttrib4fvARB = vbo_VertexAttrib4fvARB;
+
+   vfmt->VertexAttrib1fNV = vbo_VertexAttrib1fNV;
+   vfmt->VertexAttrib1fvNV = vbo_VertexAttrib1fvNV;
+   vfmt->VertexAttrib2fNV = vbo_VertexAttrib2fNV;
+   vfmt->VertexAttrib2fvNV = vbo_VertexAttrib2fvNV;
+   vfmt->VertexAttrib3fNV = vbo_VertexAttrib3fNV;
+   vfmt->VertexAttrib3fvNV = vbo_VertexAttrib3fvNV;
+   vfmt->VertexAttrib4fNV = vbo_VertexAttrib4fNV;
+   vfmt->VertexAttrib4fvNV = vbo_VertexAttrib4fvNV;
+
+   vfmt->Materialfv = vbo_Materialfv;
+
+   vfmt->EdgeFlag = vbo_EdgeFlag;
+   vfmt->Indexf = vbo_Indexf;
+   vfmt->Indexfv = vbo_Indexfv;
 
 }
 
 
-static void brw_exec_current_init( struct brw_exec_context *exec ) 
-{
-   GLcontext *ctx = exec->ctx;
-   GLint i;
-
-   /* setup the pointers for the typical 16 vertex attributes */
-   for (i = 0; i < BRW_ATTRIB_FIRST_MATERIAL; i++) 
-      exec->vtx.current[i] = ctx->Current.Attrib[i];
-
-   /* setup pointers for the 12 material attributes */
-   for (i = 0; i < MAT_ATTRIB_MAX; i++)
-      exec->vtx.current[BRW_ATTRIB_FIRST_MATERIAL + i] = 
-	 ctx->Light.Material.Attrib[i];
-
-   exec->vtx.current[BRW_ATTRIB_INDEX] = &ctx->Current.Index;
-   exec->vtx.current[BRW_ATTRIB_EDGEFLAG] = &exec->vtx.CurrentFloatEdgeFlag;
-}
-
-void brw_exec_vtx_init( struct brw_exec_context *exec )
+void vbo_exec_vtx_init( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
+   struct vbo_context *vbo = vbo_context(ctx);
    GLuint i;
 
    /* Allocate a buffer object.  Will just reuse this object
     * continuously.
     */
    exec->vtx.bufferobj = ctx->Array.NullBufferObj;
-   exec->vtx.buffer_map = ALIGN_MALLOC(BRW_VERT_BUFFER_SIZE * sizeof(GLfloat), 64);
+   exec->vtx.buffer_map = ALIGN_MALLOC(VBO_VERT_BUFFER_SIZE * sizeof(GLfloat), 64);
 
-   brw_exec_current_init( exec );
-   brw_exec_vtxfmt_init( exec );
+   vbo_exec_vtxfmt_init( exec );
 
    /* Hook our functions into the dispatch table.
     */
    _mesa_install_exec_vtxfmt( exec->ctx, &exec->vtxfmt );
 
-   for (i = 0 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0 ; i < VBO_ATTRIB_MAX ; i++) {
       exec->vtx.attrsz[i] = 0;
       exec->vtx.active_sz[i] = 0;
       exec->vtx.inputs[i] = &exec->vtx.arrays[i];
    }
- 
+   
+   {
+      struct gl_client_array *arrays = exec->vtx.arrays;
+      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+   }
+
    exec->vtx.vertex_size = 0;
 }
 
 
-void brw_exec_vtx_destroy( struct brw_exec_context *exec )
+void vbo_exec_vtx_destroy( struct vbo_exec_context *exec )
 {
    if (exec->vtx.buffer_map) {
       ALIGN_FREE(exec->vtx.buffer_map);
@@ -682,19 +668,19 @@ void brw_exec_vtx_destroy( struct brw_exec_context *exec )
 }
 
 
-void brw_exec_FlushVertices( GLcontext *ctx, GLuint flags )
+void vbo_exec_FlushVertices( GLcontext *ctx, GLuint flags )
 {
-   struct brw_exec_context *exec = IMM_CONTEXT(ctx)->exec;
+   struct vbo_exec_context *exec = &vbo_context(ctx)->exec;
 
    if (exec->ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END)
       return;
 
    if (exec->vtx.vert_count) {
-      brw_exec_vtx_flush( exec );
+      vbo_exec_vtx_flush( exec );
    }
 
    if (exec->vtx.vertex_size) {
-      brw_exec_copy_to_current( exec );
+      vbo_exec_copy_to_current( exec );
       reset_attrfv( exec );
    }
 
@@ -702,11 +688,11 @@ void brw_exec_FlushVertices( GLcontext *ctx, GLuint flags )
 }
 
 
-static void reset_attrfv( struct brw_exec_context *exec )
+static void reset_attrfv( struct vbo_exec_context *exec )
 {   
    GLuint i;
 
-   for (i = 0 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0 ; i < VBO_ATTRIB_MAX ; i++) {
       exec->vtx.attrsz[i] = 0;
       exec->vtx.active_sz[i] = 0;
    }
diff --git a/src/mesa/vbo/vbo_exec_array.c b/src/mesa/vbo/vbo_exec_array.c
new file mode 100644
index 00000000000..1e4c310203b
--- /dev/null
+++ b/src/mesa/vbo/vbo_exec_array.c
@@ -0,0 +1,387 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * 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"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, 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 NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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.
+ * 
+ **************************************************************************/
+
+#include "glheader.h"
+#include "context.h"
+#include "state.h"
+#include "api_validate.h"
+#include "api_noop.h"
+#include "dispatch.h"
+
+#include "vbo_context.h"
+
+/* Compute min and max elements for drawelements calls.
+ */
+static void get_minmax_index( GLuint count, GLuint type, 
+			      const GLvoid *indices,
+			      GLuint *min_index,
+			      GLuint *max_index)
+{
+   GLint i;
+
+   switch(type) {
+   case GL_UNSIGNED_INT: {
+      const GLuint *ui_indices = (const GLuint *)indices;
+      GLuint max_ui = ui_indices[0];
+      GLuint min_ui = ui_indices[0];
+      for (i = 1; i < count; i++) {
+	 if (ui_indices[i] > max_ui) max_ui = ui_indices[i];
+	 if (ui_indices[i] < min_ui) min_ui = ui_indices[i];
+      }
+      *min_index = min_ui;
+      *max_index = max_ui;
+      break;
+   }
+   case GL_UNSIGNED_SHORT: {
+      const GLushort *us_indices = (const GLushort *)indices;
+      GLuint max_us = us_indices[0];
+      GLuint min_us = us_indices[0];
+      for (i = 1; i < count; i++) {
+	 if (us_indices[i] > max_us) max_us = us_indices[i];
+	 if (us_indices[i] < min_us) min_us = us_indices[i];
+      }
+      *min_index = min_us;
+      *max_index = max_us;
+      break;
+   }
+   case GL_UNSIGNED_BYTE: {
+      const GLubyte *ub_indices = (const GLubyte *)indices;
+      GLuint max_ub = ub_indices[0];
+      GLuint min_ub = ub_indices[0];
+      for (i = 1; i < count; i++) {
+	 if (ub_indices[i] > max_ub) max_ub = ub_indices[i];
+	 if (ub_indices[i] < min_ub) min_ub = ub_indices[i];
+      }
+      *min_index = min_ub;
+      *max_index = max_ub;
+      break;
+   }
+   default:
+      assert(0);
+      break;
+   }
+}
+
+
+/* Just translate the arrayobj into a sane layout.
+ */
+static void bind_array_obj( GLcontext *ctx )
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   GLuint i;
+
+   /* TODO: Fix the ArrayObj struct to keep legacy arrays in an array
+    * rather than as individual named arrays.  Then this function can
+    * go away.
+    */
+   exec->array.legacy_array[VERT_ATTRIB_POS] = &ctx->Array.ArrayObj->Vertex;
+   exec->array.legacy_array[VERT_ATTRIB_WEIGHT] = &vbo->legacy_currval[VERT_ATTRIB_WEIGHT];
+   exec->array.legacy_array[VERT_ATTRIB_NORMAL] = &ctx->Array.ArrayObj->Normal;
+   exec->array.legacy_array[VERT_ATTRIB_COLOR0] = &ctx->Array.ArrayObj->Color;
+   exec->array.legacy_array[VERT_ATTRIB_COLOR1] = &ctx->Array.ArrayObj->SecondaryColor;
+   exec->array.legacy_array[VERT_ATTRIB_FOG] = &ctx->Array.ArrayObj->FogCoord;
+   exec->array.legacy_array[VERT_ATTRIB_COLOR_INDEX] = &ctx->Array.ArrayObj->Index;
+   exec->array.legacy_array[VERT_ATTRIB_EDGEFLAG] = &ctx->Array.ArrayObj->EdgeFlag;
+
+   for (i = 0; i < 8; i++)
+      exec->array.legacy_array[VERT_ATTRIB_TEX0 + i] = &ctx->Array.ArrayObj->TexCoord[i];
+
+   for (i = 0; i < VERT_ATTRIB_MAX; i++)
+      exec->array.generic_array[i] = &ctx->Array.ArrayObj->VertexAttrib[i];
+   
+   exec->array.array_obj = ctx->Array.ArrayObj->Name;
+}
+
+static void recalculate_input_bindings( GLcontext *ctx )
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   const struct gl_client_array **inputs = &exec->array.inputs[0];
+   GLuint i;
+
+   exec->array.program_mode = get_program_mode(ctx);
+   exec->array.enabled_flags = ctx->Array.ArrayObj->_Enabled;
+
+   switch (exec->array.program_mode) {
+   case VP_NONE:
+      /* When no vertex program is active, we put the material values
+       * into the generic slots.  This is the only situation where
+       * material values are available as per-vertex attributes.
+       */
+      for (i = 0; i <= VERT_ATTRIB_TEX7; i++) {
+	 if (exec->array.legacy_array[i]->Enabled)
+	    inputs[i] = exec->array.legacy_array[i];
+	 else
+	    inputs[i] = &vbo->legacy_currval[i];
+      }
+
+      for (i = 0; i < MAT_ATTRIB_MAX; i++) {
+	 inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->mat_currval[i];
+      }
+
+      /* Could use just about anything, just to fill in the empty
+       * slots:
+       */
+      for (i = MAT_ATTRIB_MAX; i < VERT_ATTRIB_MAX - VERT_ATTRIB_GENERIC0; i++)
+	 inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->generic_currval[i];
+
+      break;
+   case VP_NV:
+      /* NV_vertex_program - attribute arrays alias and override
+       * conventional, legacy arrays.  No materials, and the generic
+       * slots are vacant.
+       */
+      for (i = 0; i <= VERT_ATTRIB_TEX7; i++) {
+	 if (exec->array.generic_array[i]->Enabled)
+	    inputs[i] = exec->array.generic_array[i];
+	 else if (exec->array.legacy_array[i]->Enabled)
+	    inputs[i] = exec->array.legacy_array[i];
+	 else
+	    inputs[i] = &vbo->legacy_currval[i];
+      }
+
+      /* Could use just about anything, just to fill in the empty
+       * slots:
+       */
+      for (i = VERT_ATTRIB_GENERIC0; i < VERT_ATTRIB_MAX; i++)
+	 inputs[i] = &vbo->generic_currval[i - VERT_ATTRIB_GENERIC0];
+
+      break;
+   case VP_ARB:
+      /* ARB_vertex_program - Only the attribute zero (position) array
+       * aliases and overrides the legacy position array.  
+       *
+       * Otherwise, legacy attributes available in the legacy slots,
+       * generic attributes in the generic slots and materials are not
+       * available as per-vertex attributes.
+       */
+      if (exec->array.generic_array[0]->Enabled)
+	 inputs[0] = exec->array.generic_array[0];
+      else if (exec->array.legacy_array[0]->Enabled)
+	 inputs[0] = exec->array.legacy_array[0];
+      else
+	 inputs[0] = &vbo->legacy_currval[0];
+
+
+      for (i = 1; i <= VERT_ATTRIB_TEX7; i++) {
+	 if (exec->array.legacy_array[i]->Enabled)
+	    inputs[i] = exec->array.legacy_array[i];
+	 else
+	    inputs[i] = &vbo->legacy_currval[i];
+      }
+
+      for (i = 0; i < 16; i++) {
+	 if (exec->array.generic_array[i]->Enabled)
+	    inputs[VERT_ATTRIB_GENERIC0 + i] = exec->array.generic_array[i];
+	 else
+	    inputs[VERT_ATTRIB_GENERIC0 + i] = &vbo->generic_currval[i];
+      }
+      break;
+   }
+}
+
+static void bind_arrays( GLcontext *ctx )
+{
+#if 0
+   if (ctx->Array.ArrayObj.Name != exec->array.array_obj) {
+      bind_array_obj(ctx);
+      recalculate_input_bindings(ctx);
+   }
+   else if (exec->array.program_mode != get_program_mode(ctx) ||
+	    exec->array.enabled_flags != ctx->Array.ArrayObj->_Enabled) {
+      
+      recalculate_input_bindings(ctx);
+   }
+#else
+   bind_array_obj(ctx);
+   recalculate_input_bindings(ctx);
+#endif
+}
+
+
+
+/***********************************************************************
+ * API functions.
+ */
+
+static void GLAPIENTRY
+vbo_exec_DrawArrays(GLenum mode, GLint start, GLsizei count)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   struct _mesa_prim prim[1];
+
+   if (!_mesa_validate_DrawArrays( ctx, mode, start, count ))
+      return;
+
+   FLUSH_CURRENT( ctx, 0 );
+
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+      
+   bind_arrays( ctx );
+
+   prim[0].begin = 1;
+   prim[0].end = 1;
+   prim[0].weak = 0;
+   prim[0].pad = 0;
+   prim[0].mode = mode;
+   prim[0].start = start;
+   prim[0].count = count;
+   prim[0].indexed = 0;
+
+   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, NULL, start, start + count - 1 );
+}
+
+
+
+static void GLAPIENTRY
+vbo_exec_DrawRangeElements(GLenum mode,
+			   GLuint start, GLuint end,
+			   GLsizei count, GLenum type, const GLvoid *indices)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
+   struct _mesa_index_buffer ib;
+   struct _mesa_prim prim[1];
+
+   if (!_mesa_validate_DrawRangeElements( ctx, mode, start, end, count, type, indices ))
+      return;
+
+   FLUSH_CURRENT( ctx, 0 );
+
+   if (ctx->NewState)
+      _mesa_update_state( ctx );
+
+   bind_arrays( ctx );
+
+   ib.count = count;
+   ib.type = type; 
+   ib.obj = ctx->Array.ElementArrayBufferObj;
+   ib.ptr = indices;
+
+   prim[0].begin = 1;
+   prim[0].end = 1;
+   prim[0].weak = 0;
+   prim[0].pad = 0;
+   prim[0].mode = mode;
+   prim[0].start = 0;
+   prim[0].count = count;
+   prim[0].indexed = 1;
+
+   /* Need to give special consideration to rendering a range of
+    * indices starting somewhere above zero.  Typically the
+    * application is issuing multiple DrawRangeElements() to draw
+    * successive primitives layed out linearly in the vertex arrays.
+    * Unless the vertex arrays are all in a VBO (or locked as with
+    * CVA), the OpenGL semantics imply that we need to re-read or
+    * re-upload the vertex data on each draw call.  
+    *
+    * In the case of hardware tnl, we want to avoid starting the
+    * upload at zero, as it will mean every draw call uploads an
+    * increasing amount of not-used vertex data.  Worse - in the
+    * software tnl module, all those vertices might be transformed and
+    * lit but never rendered.
+    *
+    * If we just upload or transform the vertices in start..end,
+    * however, the indices will be incorrect.
+    *
+    * At this level, we don't know exactly what the requirements of
+    * the backend are going to be, though it will likely boil down to
+    * either:
+    *
+    * 1) Do nothing, everything is in a VBO and is processed once
+    *       only.
+    *
+    * 2) Adjust the indices and vertex arrays so that start becomes
+    *    zero.
+    *
+    * Rather than doing anything here, I'll provide a helper function
+    * for the latter case elsewhere.
+    */
+
+   vbo->draw_prims( ctx, exec->array.inputs, prim, 1, &ib, start, end );
+}
+
+static void GLAPIENTRY
+vbo_exec_DrawElements(GLenum mode, GLsizei count, GLenum type, const GLvoid *indices)
+{
+   GET_CURRENT_CONTEXT(ctx);
+   GLuint min_index = 0;
+   GLuint max_index = 0;
+
+   if (!_mesa_validate_DrawElements( ctx, mode, count, type, indices ))
+      return;
+
+   if (ctx->Array.ElementArrayBufferObj->Name) {
+      const GLvoid *map = ctx->Driver.MapBuffer(ctx,
+						 GL_ELEMENT_ARRAY_BUFFER_ARB,
+						 GL_READ_ONLY,
+						 ctx->Array.ElementArrayBufferObj);
+
+      get_minmax_index(count, type, ADD_POINTERS(map, indices), &min_index, &max_index);
+
+      ctx->Driver.UnmapBuffer(ctx,
+			      GL_ELEMENT_ARRAY_BUFFER_ARB,
+			      ctx->Array.ElementArrayBufferObj);
+   }
+   else {
+      get_minmax_index(count, type, indices, &min_index, &max_index);
+   }
+
+   vbo_exec_DrawRangeElements(mode, min_index, max_index, count, type, indices);
+}
+
+
+/***********************************************************************
+ * Initialization
+ */
+
+
+
+
+void vbo_exec_array_init( struct vbo_exec_context *exec )
+{
+#if 1
+   exec->vtxfmt.DrawArrays = vbo_exec_DrawArrays;
+   exec->vtxfmt.DrawElements = vbo_exec_DrawElements;
+   exec->vtxfmt.DrawRangeElements = vbo_exec_DrawRangeElements;
+#else
+   exec->vtxfmt.DrawArrays = _mesa_noop_DrawArrays;
+   exec->vtxfmt.DrawElements = _mesa_noop_DrawElements;
+   exec->vtxfmt.DrawRangeElements = _mesa_noop_DrawRangeElements;
+#endif
+}
+
+
+void vbo_exec_array_destroy( struct vbo_exec_context *exec )
+{
+   /* nothing to do */
+}
diff --git a/src/mesa/drivers/dri/i965/brw_exec_draw.c b/src/mesa/vbo/vbo_exec_draw.c
index 62bda9845b9..2b52d7a1fff 100644
--- a/src/mesa/drivers/dri/i965/brw_exec_draw.c
+++ b/src/mesa/vbo/vbo_exec_draw.c
@@ -31,12 +31,10 @@
 #include "state.h"
 #include "macros.h"
 
-#include "brw_exec.h"
-#include "brw_draw.h"
-#include "brw_fallback.h"
+#include "vbo_context.h"
 
 
-static void brw_exec_debug_verts( struct brw_exec_context *exec )
+static void vbo_exec_debug_verts( struct vbo_exec_context *exec )
 {
    GLuint count = exec->vtx.vert_count;
    GLuint i;
@@ -48,7 +46,7 @@ static void brw_exec_debug_verts( struct brw_exec_context *exec )
 		exec->vtx.vertex_size);
 
    for (i = 0 ; i < exec->vtx.prim_count ; i++) {
-      struct brw_draw_prim *prim = &exec->vtx.prim[i];
+      struct _mesa_prim *prim = &exec->vtx.prim[i];
       _mesa_printf("   prim %d: %s%s %d..%d %s %s\n",
 		   i, 
 		   _mesa_lookup_enum_by_nr(prim->mode),
@@ -65,7 +63,7 @@ static void brw_exec_debug_verts( struct brw_exec_context *exec )
  * NOTE: Need to have calculated primitives by this point -- do it on the fly.
  * NOTE: Old 'parity' issue is gone.
  */
-static GLuint brw_copy_vertices( struct brw_exec_context *exec )
+static GLuint vbo_copy_vertices( struct vbo_exec_context *exec )
 {
    GLuint nr = exec->vtx.prim[exec->vtx.prim_count-1].count;
    GLuint ovf, i;
@@ -116,6 +114,11 @@ static GLuint brw_copy_vertices( struct brw_exec_context *exec )
 	 return 2;
       }
    case GL_TRIANGLE_STRIP:
+      /* no parity issue, but need to make sure the tri is not drawn twice */
+      if (nr & 1) {
+	 exec->vtx.prim[exec->vtx.prim_count-1].count--;
+      }
+      /* fallthrough */
    case GL_QUAD_STRIP:
       switch (nr) {
       case 0: ovf = 0; break;
@@ -134,24 +137,49 @@ static GLuint brw_copy_vertices( struct brw_exec_context *exec )
 }
 
 
+
 /* TODO: populate these as the vertex is defined:
  */
-static void brw_exec_bind_arrays( struct brw_exec_context *exec )
+static void vbo_exec_bind_arrays( GLcontext *ctx )
 {
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_exec_context *exec = &vbo->exec;
    struct gl_client_array *arrays = exec->vtx.arrays;
    GLuint count = exec->vtx.vert_count;
    GLubyte *data = exec->vtx.buffer_map;
+   const GLuint *map;
    GLuint attr;
 
-   memset(arrays, 0, BRW_ATTRIB_MAX * sizeof(arrays[0]));
+   /* Install the default (ie Current) attributes first, then overlay
+    * all active ones.
+    */
+   switch (get_program_mode(exec->ctx)) {
+   case VP_NONE:
+      memcpy(arrays,      vbo->legacy_currval, 16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->mat_currval,    MAT_ATTRIB_MAX * sizeof(arrays[0]));
+      map = vbo->map_vp_none;
+      break;
+   case VP_NV:
+   case VP_ARB:
+      /* The aliasing of attributes for NV vertex programs has already
+       * occurred.  NV vertex programs cannot access material values,
+       * nor attributes greater than VERT_ATTRIB_TEX7.  
+       */
+      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+      map = vbo->map_vp_arb;
+      break;
+   }
 
    /* Make all active attributes (including edgeflag) available as
     * arrays of floats.
     */
-   for (attr = 0; attr < BRW_ATTRIB_MAX ; attr++) {
-      if (exec->vtx.attrsz[attr]) {
+   for (attr = 0; attr < VERT_ATTRIB_MAX ; attr++) {
+      GLuint src = map[attr];
+
+      if (exec->vtx.attrsz[src]) {
 	 arrays[attr].Ptr = (void *)data;
-	 arrays[attr].Size = exec->vtx.attrsz[attr];
+	 arrays[attr].Size = exec->vtx.attrsz[src];
 	 arrays[attr].StrideB = exec->vtx.vertex_size * sizeof(GLfloat);
 	 arrays[attr].Stride = exec->vtx.vertex_size * sizeof(GLfloat);
 	 arrays[attr].Type = GL_FLOAT;
@@ -165,59 +193,32 @@ static void brw_exec_bind_arrays( struct brw_exec_context *exec )
 }
 
 
-static
-void brw_exec_loopback_vertex_list( GLcontext *ctx,
-				    struct brw_exec_context *exec )
-{
-   const GLfloat *buffer = (const GLfloat *)exec->vtx.buffer_map;
-
-   brw_fallback(ctx);
-
-   brw_loopback_vertex_list( ctx,
-			     buffer,
-			     exec->vtx.attrsz,
-			     exec->vtx.prim,
-			     exec->vtx.prim_count,
-			     0,	/* XXX - copied.nr */
-			     exec->vtx.vertex_size);
-
-
-   brw_unfallback(ctx);
-}
-
-
 /**
  * Execute the buffer and save copied verts.
  */
-void brw_exec_vtx_flush( struct brw_exec_context *exec )
+void vbo_exec_vtx_flush( struct vbo_exec_context *exec )
 {
    if (0)
-      brw_exec_debug_verts( exec );
+      vbo_exec_debug_verts( exec );
 
 
    if (exec->vtx.prim_count && 
        exec->vtx.vert_count) {
 
-      exec->vtx.copied.nr = brw_copy_vertices( exec ); 
+      exec->vtx.copied.nr = vbo_copy_vertices( exec ); 
 
       if (exec->vtx.copied.nr != exec->vtx.vert_count) {
 	 GLcontext *ctx = exec->ctx;
 
-	 brw_exec_bind_arrays( exec );
-
-
-	 if (!brw_draw_prims( ctx, 
-			      exec->vtx.inputs, 
-			      exec->vtx.prim, 
-			      exec->vtx.prim_count,
-			      NULL,
-			      0,
-			      exec->vtx.vert_count,
-			      0 )) {
-	    /* Fallback path: 
-	     */
-	    brw_exec_loopback_vertex_list(ctx, exec);
-	 }
+	 vbo_exec_bind_arrays( ctx );
+
+	 vbo_context(ctx)->draw_prims( ctx, 
+				       exec->vtx.inputs, 
+				       exec->vtx.prim, 
+				       exec->vtx.prim_count,
+				       NULL,
+				       0,
+				       exec->vtx.vert_count );
       }
    }
 
diff --git a/src/mesa/drivers/dri/i965/brw_exec_eval.c b/src/mesa/vbo/vbo_exec_eval.c
index 9dbeb1b58e4..99a80445763 100644
--- a/src/mesa/drivers/dri/i965/brw_exec_eval.c
+++ b/src/mesa/vbo/vbo_exec_eval.c
@@ -30,21 +30,21 @@
 #include "context.h"
 #include "macros.h"
 #include "math/m_eval.h"
-#include "brw_exec.h"
+#include "vbo_exec.h"
 #include "dispatch.h"
 
 
-static void clear_active_eval1( struct brw_exec_context *exec, GLuint attr ) 
+static void clear_active_eval1( struct vbo_exec_context *exec, GLuint attr ) 
 {
    exec->eval.map1[attr].map = NULL;
 }
 
-static void clear_active_eval2( struct brw_exec_context *exec, GLuint attr ) 
+static void clear_active_eval2( struct vbo_exec_context *exec, GLuint attr ) 
 {
    exec->eval.map2[attr].map = NULL;
 }
 
-static void set_active_eval1( struct brw_exec_context *exec, GLuint attr, GLuint dim, 
+static void set_active_eval1( struct vbo_exec_context *exec, GLuint attr, GLuint dim, 
 			      struct gl_1d_map *map )
 {
    if (!exec->eval.map1[attr].map) {
@@ -53,7 +53,7 @@ static void set_active_eval1( struct brw_exec_context *exec, GLuint attr, GLuint
    }
 } 
 
-static void set_active_eval2( struct brw_exec_context *exec, GLuint attr, GLuint dim, 
+static void set_active_eval2( struct vbo_exec_context *exec, GLuint attr, GLuint dim, 
 			      struct gl_2d_map *map )
 {
    if (!exec->eval.map2[attr].map) {
@@ -62,21 +62,21 @@ static void set_active_eval2( struct brw_exec_context *exec, GLuint attr, GLuint
    }
 } 
 
-void brw_exec_eval_update( struct brw_exec_context *exec )
+void vbo_exec_eval_update( struct vbo_exec_context *exec )
 {
    GLcontext *ctx = exec->ctx;
    GLuint attr;
 
    /* Vertex program maps have priority over conventional attribs */
 
-   for (attr = 0; attr < BRW_ATTRIB_FIRST_MATERIAL; attr++) {
+   for (attr = 0; attr < VBO_ATTRIB_FIRST_MATERIAL; attr++) {
       clear_active_eval1( exec, attr );
       clear_active_eval2( exec, attr );
    }
 
    /* _NEW_PROGRAM */
    if (ctx->VertexProgram._Enabled) {
-      for (attr = 0; attr < BRW_ATTRIB_FIRST_MATERIAL; attr++) {
+      for (attr = 0; attr < VBO_ATTRIB_FIRST_MATERIAL; attr++) {
 	 /* _NEW_EVAL */
 	 if (ctx->Eval.Map1Attrib[attr]) 
 	    set_active_eval1( exec, attr, 4, &ctx->EvalMap.Map1Attrib[attr] );
@@ -87,55 +87,55 @@ void brw_exec_eval_update( struct brw_exec_context *exec )
    }
 
    if (ctx->Eval.Map1Color4) 
-      set_active_eval1( exec, BRW_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map1Color4 );
+      set_active_eval1( exec, VBO_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map1Color4 );
       
    if (ctx->Eval.Map2Color4) 
-      set_active_eval2( exec, BRW_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map2Color4 );
+      set_active_eval2( exec, VBO_ATTRIB_COLOR0, 4, &ctx->EvalMap.Map2Color4 );
 
    if (ctx->Eval.Map1TextureCoord4) 
-      set_active_eval1( exec, BRW_ATTRIB_TEX0, 4, &ctx->EvalMap.Map1Texture4 );
+      set_active_eval1( exec, VBO_ATTRIB_TEX0, 4, &ctx->EvalMap.Map1Texture4 );
    else if (ctx->Eval.Map1TextureCoord3) 
-      set_active_eval1( exec, BRW_ATTRIB_TEX0, 3, &ctx->EvalMap.Map1Texture3 );
+      set_active_eval1( exec, VBO_ATTRIB_TEX0, 3, &ctx->EvalMap.Map1Texture3 );
    else if (ctx->Eval.Map1TextureCoord2) 
-      set_active_eval1( exec, BRW_ATTRIB_TEX0, 2, &ctx->EvalMap.Map1Texture2 );
+      set_active_eval1( exec, VBO_ATTRIB_TEX0, 2, &ctx->EvalMap.Map1Texture2 );
    else if (ctx->Eval.Map1TextureCoord1) 
-      set_active_eval1( exec, BRW_ATTRIB_TEX0, 1, &ctx->EvalMap.Map1Texture1 );
+      set_active_eval1( exec, VBO_ATTRIB_TEX0, 1, &ctx->EvalMap.Map1Texture1 );
 
    if (ctx->Eval.Map2TextureCoord4) 
-      set_active_eval2( exec, BRW_ATTRIB_TEX0, 4, &ctx->EvalMap.Map2Texture4 );
+      set_active_eval2( exec, VBO_ATTRIB_TEX0, 4, &ctx->EvalMap.Map2Texture4 );
    else if (ctx->Eval.Map2TextureCoord3) 
-      set_active_eval2( exec, BRW_ATTRIB_TEX0, 3, &ctx->EvalMap.Map2Texture3 );
+      set_active_eval2( exec, VBO_ATTRIB_TEX0, 3, &ctx->EvalMap.Map2Texture3 );
    else if (ctx->Eval.Map2TextureCoord2) 
-      set_active_eval2( exec, BRW_ATTRIB_TEX0, 2, &ctx->EvalMap.Map2Texture2 );
+      set_active_eval2( exec, VBO_ATTRIB_TEX0, 2, &ctx->EvalMap.Map2Texture2 );
    else if (ctx->Eval.Map2TextureCoord1) 
-      set_active_eval2( exec, BRW_ATTRIB_TEX0, 1, &ctx->EvalMap.Map2Texture1 );
+      set_active_eval2( exec, VBO_ATTRIB_TEX0, 1, &ctx->EvalMap.Map2Texture1 );
 
    if (ctx->Eval.Map1Normal) 
-      set_active_eval1( exec, BRW_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map1Normal );
+      set_active_eval1( exec, VBO_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map1Normal );
 
    if (ctx->Eval.Map2Normal) 
-      set_active_eval2( exec, BRW_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map2Normal );
+      set_active_eval2( exec, VBO_ATTRIB_NORMAL, 3, &ctx->EvalMap.Map2Normal );
 
    if (ctx->Eval.Map1Vertex4) 
-      set_active_eval1( exec, BRW_ATTRIB_POS, 4, &ctx->EvalMap.Map1Vertex4 );
+      set_active_eval1( exec, VBO_ATTRIB_POS, 4, &ctx->EvalMap.Map1Vertex4 );
    else if (ctx->Eval.Map1Vertex3) 
-      set_active_eval1( exec, BRW_ATTRIB_POS, 3, &ctx->EvalMap.Map1Vertex3 );
+      set_active_eval1( exec, VBO_ATTRIB_POS, 3, &ctx->EvalMap.Map1Vertex3 );
 
    if (ctx->Eval.Map2Vertex4) 
-      set_active_eval2( exec, BRW_ATTRIB_POS, 4, &ctx->EvalMap.Map2Vertex4 );
+      set_active_eval2( exec, VBO_ATTRIB_POS, 4, &ctx->EvalMap.Map2Vertex4 );
    else if (ctx->Eval.Map2Vertex3) 
-      set_active_eval2( exec, BRW_ATTRIB_POS, 3, &ctx->EvalMap.Map2Vertex3 );
+      set_active_eval2( exec, VBO_ATTRIB_POS, 3, &ctx->EvalMap.Map2Vertex3 );
 
    exec->eval.recalculate_maps = 0;
 }
 
 
 
-void brw_exec_do_EvalCoord1f(struct brw_exec_context *exec, GLfloat u)
+void vbo_exec_do_EvalCoord1f(struct vbo_exec_context *exec, GLfloat u)
 {
    GLuint attr;
 
-   for (attr = 1; attr <= BRW_ATTRIB_INDEX; attr++) {
+   for (attr = 1; attr <= VBO_ATTRIB_INDEX; attr++) {
       struct gl_1d_map *map = exec->eval.map1[attr].map;
       if (map) {
 	 GLfloat uu = (u - map->u1) * map->du;
@@ -175,12 +175,12 @@ void brw_exec_do_EvalCoord1f(struct brw_exec_context *exec, GLfloat u)
 
 
 
-void brw_exec_do_EvalCoord2f( struct brw_exec_context *exec, 
+void vbo_exec_do_EvalCoord2f( struct vbo_exec_context *exec, 
 			      GLfloat u, GLfloat v )
 {   
    GLuint attr;
 
-   for (attr = 1; attr <= BRW_ATTRIB_INDEX; attr++) {
+   for (attr = 1; attr <= VBO_ATTRIB_INDEX; attr++) {
       struct gl_2d_map *map = exec->eval.map2[attr].map;
       if (map) {
 	 GLfloat uu = (u - map->u1) * map->du;
@@ -234,8 +234,8 @@ void brw_exec_do_EvalCoord2f( struct brw_exec_context *exec,
          NORMALIZE_3FV(normal);
 	 normal[3] = 1.0;
 
-	 COPY_SZ_4V( exec->vtx.attrptr[BRW_ATTRIB_NORMAL],
-		     exec->vtx.attrsz[BRW_ATTRIB_NORMAL],
+	 COPY_SZ_4V( exec->vtx.attrptr[VBO_ATTRIB_NORMAL],
+		     exec->vtx.attrsz[VBO_ATTRIB_NORMAL],
 		     normal );
 
       }
diff --git a/src/mesa/vbo/vbo_rebase.c b/src/mesa/vbo/vbo_rebase.c
new file mode 100644
index 00000000000..bc4211d8529
--- /dev/null
+++ b/src/mesa/vbo/vbo_rebase.c
@@ -0,0 +1,222 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  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"),
+ * 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 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
+ * BRIAN PAUL 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:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/* Helper for drivers which find themselves rendering a range of
+ * indices starting somewhere above zero.  Typically the application
+ * is issuing multiple DrawArrays() or DrawElements() to draw
+ * successive primitives layed out linearly in the vertex arrays.
+ * Unless the vertex arrays are all in a VBO, the OpenGL semantics
+ * imply that we need to re-upload the vertex data on each draw call.
+ * In that case, we want to avoid starting the upload at zero, as it
+ * will mean every draw call uploads an increasing amount of not-used
+ * vertex data.  Worse - in the software tnl module, all those
+ * vertices will be transformed and lit.
+ *
+ * If we just upload the new data, however, the indices will be
+ * incorrect as we tend to upload each set of vertex data to a new
+ * region.  
+ *
+ * This file provides a helper to adjust the arrays, primitives and
+ * indices of a draw call so that it can be re-issued with a min_index
+ * of zero.
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "mtypes.h"
+
+#include "vbo.h"
+
+
+#define REBASE(TYPE) 						\
+static void *rebase_##TYPE( const void *ptr,			\
+			  GLuint count, 			\
+			  TYPE min_index )			\
+{								\
+   const TYPE *in = (TYPE *)ptr;				\
+   TYPE *tmp_indices = malloc(count * sizeof(TYPE));	\
+   GLuint i;							\
+								\
+   for (i = 0; i < count; i++)  				\
+      tmp_indices[i] = in[i] - min_index;			\
+								\
+   return (void *)tmp_indices;					\
+}
+
+
+REBASE(GLuint)
+REBASE(GLushort)
+REBASE(GLubyte)
+
+GLboolean vbo_all_varyings_in_vbos( const struct gl_client_array *arrays[] )
+{
+   GLuint i;
+   
+   for (i = 0; i < VERT_ATTRIB_MAX; i++)
+      if (arrays[i]->StrideB &&
+	  arrays[i]->BufferObj->Name == 0)
+	 return GL_FALSE;
+
+   return GL_TRUE;
+}
+
+/* Adjust primitives, indices and vertex definitions so that min_index
+ * becomes zero. There are lots of reasons for wanting to do this, eg:
+ *
+ * Software tnl:
+ *    - any time min_index != 0, otherwise unused vertices lower than
+ *      min_index will be transformed.
+ *
+ * Hardware tnl:
+ *    - if ib != NULL and min_index != 0, otherwise vertices lower than 
+ *      min_index will be uploaded.  Requires adjusting index values.
+ *
+ *    - if ib == NULL and min_index != 0, just for convenience so this doesn't
+ *      have to be handled within the driver.
+ *
+ * Hardware tnl with VBO support:
+ *    - as above, but only when vertices are not (all?) in VBO's.
+ *    - can't save time by trying to upload half a vbo - typically it is
+ *      all or nothing.
+ */
+void vbo_rebase_prims( GLcontext *ctx,
+		       const struct gl_client_array *arrays[],
+		       const struct _mesa_prim *prim,
+		       GLuint nr_prims,
+		       const struct _mesa_index_buffer *ib,
+		       GLuint min_index,
+		       GLuint max_index,
+		       vbo_draw_func draw )
+{
+   struct gl_client_array tmp_arrays[VERT_ATTRIB_MAX];
+   const struct gl_client_array *tmp_array_pointers[VERT_ATTRIB_MAX];
+
+   struct _mesa_index_buffer tmp_ib;
+   struct _mesa_prim *tmp_prims = NULL;
+   void *tmp_indices = NULL;
+   GLuint i;
+
+   assert(min_index != 0);
+
+   if (0)
+      _mesa_printf("%s %d..%d\n", __FUNCTION__, min_index, max_index);
+
+   if (ib) {
+      /* Unfortunately need to adjust each index individually.
+       */
+      GLboolean map_ib = ib->obj->Name && !ib->obj->Pointer;
+      void *ptr;
+
+      if (map_ib) 
+	 ctx->Driver.MapBuffer(ctx, 
+			       GL_ELEMENT_ARRAY_BUFFER,
+			       GL_READ_ONLY_ARB,
+			       ib->obj);
+
+
+      ptr = ADD_POINTERS(ib->obj->Pointer, ib->ptr);
+
+      /* Some users might prefer it if we translated elements to
+       * GLuints here.  Others wouldn't...
+       */
+      switch (ib->type) {
+      case GL_UNSIGNED_INT: 
+	 tmp_indices = rebase_GLuint( ptr, ib->count, min_index );
+	 break;
+      case GL_UNSIGNED_SHORT: 
+	 tmp_indices = rebase_GLushort( ptr, ib->count, min_index );
+	 break;
+      case GL_UNSIGNED_BYTE: 
+	 tmp_indices = rebase_GLubyte( ptr, ib->count, min_index );
+	 break;
+      }      
+
+      if (map_ib) 
+	 ctx->Driver.UnmapBuffer(ctx, 
+				 GL_ELEMENT_ARRAY_BUFFER,
+				 ib->obj);
+
+      tmp_ib.obj = ctx->Array.NullBufferObj;
+      tmp_ib.ptr = tmp_indices;
+      tmp_ib.count = ib->count;
+      tmp_ib.type = ib->type;
+
+      ib = &tmp_ib;
+   }
+   else {
+      /* Otherwise the primitives need adjustment.
+       */
+      tmp_prims = (struct _mesa_prim *)_mesa_malloc(sizeof(*prim) * nr_prims);
+
+      for (i = 0; i < nr_prims; i++) {
+	 /* If this fails, it could indicate an application error:
+	  */
+	 assert(prim[i].start >= min_index);
+
+	 tmp_prims[i] = prim[i];
+	 tmp_prims[i].start -= min_index;
+      }
+
+      prim = tmp_prims;
+   }
+
+   /* Just need to adjust the pointer values on each incoming array.
+    * This works for VBO and non-vbo rendering and shouldn't pesimize
+    * VBO-based upload schemes.  However this may still not be a fast
+    * path for hardware tnl for VBO based rendering as most machines
+    * will be happier if you just specify a starting vertex value in
+    * each primitive.
+    *
+    * For drivers with hardware tnl, you only want to do this if you
+    * are forced to, eg non-VBO indexed rendering with start != 0.
+    */
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      tmp_arrays[i] = *arrays[i];
+      tmp_arrays[i].Ptr += min_index * tmp_arrays[i].StrideB;
+      tmp_array_pointers[i] = &tmp_arrays[i];
+   }
+   
+   /* Re-issue the draw call.
+    */
+   draw( ctx, 
+	 tmp_array_pointers, 
+	 prim, 
+	 nr_prims, 
+	 ib, 
+	 0, 
+	 max_index - min_index );
+   
+   if (tmp_indices)
+      _mesa_free(tmp_indices);
+   
+   if (tmp_prims)
+      _mesa_free(tmp_prims);
+}
+
+
+
diff --git a/src/mesa/vbo/vbo_save.c b/src/mesa/vbo/vbo_save.c
new file mode 100644
index 00000000000..e7f46879639
--- /dev/null
+++ b/src/mesa/vbo/vbo_save.c
@@ -0,0 +1,89 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.3
+ *
+ * Copyright (C) 1999-2005  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"),
+ * 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 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
+ * BRIAN PAUL 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:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "mtypes.h"
+#include "dlist.h"
+#include "vtxfmt.h"
+#include "imports.h"
+
+#include "vbo_context.h"
+
+
+
+static void vbo_save_callback_init( GLcontext *ctx )
+{
+   ctx->Driver.NewList = vbo_save_NewList;
+   ctx->Driver.EndList = vbo_save_EndList;
+   ctx->Driver.SaveFlushVertices = vbo_save_SaveFlushVertices;
+   ctx->Driver.BeginCallList = vbo_save_BeginCallList;
+   ctx->Driver.EndCallList = vbo_save_EndCallList;
+   ctx->Driver.NotifySaveBegin = vbo_save_NotifyBegin;
+}
+
+
+
+void vbo_save_init( GLcontext *ctx )
+{
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_save_context *save = &vbo->save;
+
+   save->ctx = ctx;
+
+   vbo_save_api_init( save );
+   vbo_save_callback_init(ctx);
+
+   {
+      struct gl_client_array *arrays = save->arrays;
+      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+   }
+
+   ctx->Driver.CurrentSavePrimitive = PRIM_UNKNOWN;
+}
+
+
+void vbo_save_destroy( GLcontext *ctx )
+{
+}
+
+
+
+
+/* Note that this can occur during the playback of a display list:
+ */
+void vbo_save_fallback( GLcontext *ctx, GLboolean fallback )
+{
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
+
+   if (fallback)
+      save->replay_flags |= VBO_SAVE_FALLBACK;
+   else
+      save->replay_flags &= ~VBO_SAVE_FALLBACK;
+}
+
+
diff --git a/src/mesa/drivers/dri/i965/brw_save.h b/src/mesa/vbo/vbo_save.h
index 41cabe65084..b81f275a602 100644
--- a/src/mesa/drivers/dri/i965/brw_save.h
+++ b/src/mesa/vbo/vbo_save.h
@@ -31,16 +31,16 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  *
  */
 
-#ifndef BRW_SAVE_H
-#define BRW_SAVE_H
+#ifndef VBO_SAVE_H
+#define VBO_SAVE_H
 
 #include "mtypes.h"
-#include "brw_attrib.h"
-#include "brw_draw.h"
+#include "vbo.h"
+#include "vbo_attrib.h"
 
 
-struct brw_save_copied_vtx {
-   GLfloat buffer[BRW_ATTRIB_MAX * 4 * BRW_MAX_COPIED_VERTS];
+struct vbo_save_copied_vtx {
+   GLfloat buffer[VBO_ATTRIB_MAX * 4 * VBO_MAX_COPIED_VERTS];
    GLuint nr;
 };
 
@@ -60,8 +60,8 @@ struct brw_save_copied_vtx {
  * Eval and other commands that don't fit into these vertex lists are
  * compiled using the fallback opcode mechanism provided by dlist.c.
  */
-struct brw_save_vertex_list {
-   GLubyte attrsz[BRW_ATTRIB_MAX];
+struct vbo_save_vertex_list {
+   GLubyte attrsz[VBO_ATTRIB_MAX];
    GLuint vertex_size;
 
    GLuint buffer_offset;
@@ -70,15 +70,15 @@ struct brw_save_vertex_list {
    GLboolean dangling_attr_ref;	/* current attr implicitly referenced 
 				   outside the list */
 
-   struct brw_draw_prim *prim;
+   struct _mesa_prim *prim;
    GLuint prim_count;
 
-   struct brw_save_vertex_store *vertex_store;
-   struct brw_save_primitive_store *prim_store;
+   struct vbo_save_vertex_store *vertex_store;
+   struct vbo_save_primitive_store *prim_store;
 };
 
 /* These buffers should be a reasonable size to support upload to
- * hardware.  Current brw implementation will re-upload on any
+ * hardware.  Current vbo implementation will re-upload on any
  * changes, so don't make too big or apps which dynamically create
  * dlists and use only a few times will suffer.
  *
@@ -87,36 +87,36 @@ struct brw_save_vertex_list {
  * likelyhood as it occurs.  No reason we couldn't change usage
  * internally even though this probably isn't allowed for client VBOs?
  */
-#define BRW_SAVE_BUFFER_SIZE (8*1024) /* dwords */
-#define BRW_SAVE_PRIM_SIZE   128
-#define BRW_SAVE_PRIM_WEAK 0x40
+#define VBO_SAVE_BUFFER_SIZE (8*1024) /* dwords */
+#define VBO_SAVE_PRIM_SIZE   128
+#define VBO_SAVE_PRIM_WEAK 0x40
 
-#define BRW_SAVE_FALLBACK    0x10000000
+#define VBO_SAVE_FALLBACK    0x10000000
 
 /* Storage to be shared among several vertex_lists.
  */
-struct brw_save_vertex_store {
+struct vbo_save_vertex_store {
    struct gl_buffer_object *bufferobj;
    GLfloat *buffer;
    GLuint used;
    GLuint refcount;
 };
 
-struct brw_save_primitive_store {
-   struct brw_draw_prim buffer[BRW_SAVE_PRIM_SIZE];
+struct vbo_save_primitive_store {
+   struct _mesa_prim buffer[VBO_SAVE_PRIM_SIZE];
    GLuint used;
    GLuint refcount;
 };
 
 
-struct brw_save_context {
+struct vbo_save_context {
    GLcontext *ctx;
    GLvertexformat vtxfmt;
-   struct gl_client_array arrays[BRW_ATTRIB_MAX];
-   const struct gl_client_array *inputs[BRW_ATTRIB_MAX];
+   struct gl_client_array arrays[VBO_ATTRIB_MAX];
+   const struct gl_client_array *inputs[VBO_ATTRIB_MAX];
 
-   GLubyte attrsz[BRW_ATTRIB_MAX];
-   GLubyte active_sz[BRW_ATTRIB_MAX];
+   GLubyte attrsz[VBO_ATTRIB_MAX];
+   GLubyte active_sz[VBO_ATTRIB_MAX];
    GLuint vertex_size;
 
    GLfloat *buffer;
@@ -124,15 +124,15 @@ struct brw_save_context {
    GLuint wrap_count;
    GLuint replay_flags;
 
-   struct brw_draw_prim *prim;
+   struct _mesa_prim *prim;
    GLuint prim_count, prim_max;
 
-   struct brw_save_vertex_store *vertex_store;
-   struct brw_save_primitive_store *prim_store;
+   struct vbo_save_vertex_store *vertex_store;
+   struct vbo_save_primitive_store *prim_store;
 
    GLfloat *vbptr;		   /* cursor, points into buffer */
-   GLfloat vertex[BRW_ATTRIB_MAX*4];	   /* current values */
-   GLfloat *attrptr[BRW_ATTRIB_MAX];
+   GLfloat vertex[VBO_ATTRIB_MAX*4];	   /* current values */
+   GLfloat *attrptr[VBO_ATTRIB_MAX];
    GLuint vert_count;
    GLuint max_vert;
    GLboolean dangling_attr_ref;
@@ -140,32 +140,38 @@ struct brw_save_context {
 
    GLuint opcode_vertex_list;
 
-   struct brw_save_copied_vtx copied;
+   struct vbo_save_copied_vtx copied;
    
-   GLfloat CurrentFloatEdgeFlag;
-
-   GLfloat *current[BRW_ATTRIB_MAX]; /* points into ctx->ListState */
-   GLubyte *currentsz[BRW_ATTRIB_MAX];
+   GLfloat *current[VBO_ATTRIB_MAX]; /* points into ctx->ListState */
+   GLubyte *currentsz[VBO_ATTRIB_MAX];
 };
 
 
-void brw_save_init( GLcontext *ctx );
-void brw_save_destroy( GLcontext *ctx );
-void brw_save_wakeup( GLcontext *ctx );
-void brw_save_invalidate_state( GLcontext *ctx, GLuint new_state );
-void brw_save_fallback( GLcontext *ctx, GLboolean fallback );
+void vbo_save_init( GLcontext *ctx );
+void vbo_save_destroy( GLcontext *ctx );
+void vbo_save_fallback( GLcontext *ctx, GLboolean fallback );
+
+/* save_loopback.c:
+ */
+void vbo_loopback_vertex_list( GLcontext *ctx,
+			       const GLfloat *buffer,
+			       const GLubyte *attrsz,
+			       const struct _mesa_prim *prim,
+			       GLuint prim_count,
+			       GLuint wrap_count,
+			       GLuint vertex_size);
 
 /* Callbacks:
  */
-void brw_save_EndList( GLcontext *ctx );
-void brw_save_NewList( GLcontext *ctx, GLuint list, GLenum mode );
-void brw_save_EndCallList( GLcontext *ctx );
-void brw_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *list );
-void brw_save_SaveFlushVertices( GLcontext *ctx );
-GLboolean brw_save_NotifyBegin( GLcontext *ctx, GLenum mode );
+void vbo_save_EndList( GLcontext *ctx );
+void vbo_save_NewList( GLcontext *ctx, GLuint list, GLenum mode );
+void vbo_save_EndCallList( GLcontext *ctx );
+void vbo_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *list );
+void vbo_save_SaveFlushVertices( GLcontext *ctx );
+GLboolean vbo_save_NotifyBegin( GLcontext *ctx, GLenum mode );
 
-void brw_save_playback_vertex_list( GLcontext *ctx, void *data );
+void vbo_save_playback_vertex_list( GLcontext *ctx, void *data );
 
-void brw_save_api_init( struct brw_save_context *save );
+void vbo_save_api_init( struct vbo_save_context *save );
 
 #endif
diff --git a/src/mesa/drivers/dri/i965/brw_save_api.c b/src/mesa/vbo/vbo_save_api.c
index 06ed1d23aa2..c08fd1fe555 100644
--- a/src/mesa/drivers/dri/i965/brw_save_api.c
+++ b/src/mesa/vbo/vbo_save_api.c
@@ -77,9 +77,7 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "vtxfmt.h"
 #include "dispatch.h"
 
-#include "brw_save.h"
-#include "brw_fallback.h"
-
+#include "vbo_context.h"
 
 
 
@@ -88,11 +86,11 @@ USE OR OTHER DEALINGS IN THE SOFTWARE.
  * wrong-footed on replay.
  */
 static GLuint _save_copy_vertices( GLcontext *ctx, 
-				   const struct brw_save_vertex_list *node,
+				   const struct vbo_save_vertex_list *node,
 				   const GLfloat *src_buffer)
 {
-   struct brw_save_context *save = IMM_CONTEXT( ctx )->save;
-   const struct brw_draw_prim *prim = &node->prim[node->prim_count-1];
+   struct vbo_save_context *save = &vbo_context( ctx )->save;
+   const struct _mesa_prim *prim = &node->prim[node->prim_count-1];
    GLuint nr = prim->count;
    GLuint sz = save->vertex_size;
    const GLfloat *src = src_buffer + prim->start * sz;
@@ -158,9 +156,9 @@ static GLuint _save_copy_vertices( GLcontext *ctx,
 }
 
 
-static struct brw_save_vertex_store *alloc_vertex_store( GLcontext *ctx )
+static struct vbo_save_vertex_store *alloc_vertex_store( GLcontext *ctx )
 {
-   struct brw_save_vertex_store *vertex_store = CALLOC_STRUCT(brw_save_vertex_store);
+   struct vbo_save_vertex_store *vertex_store = CALLOC_STRUCT(vbo_save_vertex_store);
 
    /* obj->Name needs to be non-zero, but won't ever be examined more
     * closely than that.  In particular these buffers won't be entered
@@ -172,7 +170,7 @@ static struct brw_save_vertex_store *alloc_vertex_store( GLcontext *ctx )
 
    ctx->Driver.BufferData( ctx, 
 			   GL_ARRAY_BUFFER_ARB, 
-			   BRW_SAVE_BUFFER_SIZE * sizeof(GLfloat),
+			   VBO_SAVE_BUFFER_SIZE * sizeof(GLfloat),
 			   NULL,
 			   GL_STATIC_DRAW_ARB,
 			   vertex_store->bufferobj);
@@ -184,7 +182,7 @@ static struct brw_save_vertex_store *alloc_vertex_store( GLcontext *ctx )
    return vertex_store;
 }
 
-static void free_vertex_store( GLcontext *ctx, struct brw_save_vertex_store *vertex_store )
+static void free_vertex_store( GLcontext *ctx, struct vbo_save_vertex_store *vertex_store )
 {
    assert(!vertex_store->buffer);
 
@@ -194,29 +192,29 @@ static void free_vertex_store( GLcontext *ctx, struct brw_save_vertex_store *ver
    FREE( vertex_store );
 }
 
-static GLfloat *map_vertex_store( GLcontext *ctx, struct brw_save_vertex_store *vertex_store )
+static GLfloat *map_vertex_store( GLcontext *ctx, struct vbo_save_vertex_store *vertex_store )
 {
    assert(vertex_store->bufferobj);
    assert(!vertex_store->buffer);
    vertex_store->buffer = (GLfloat *)ctx->Driver.MapBuffer(ctx, 
 							   GL_ARRAY_BUFFER_ARB,	/* not used */
-							   GL_STATIC_DRAW_ARB, /* not used */
+							   GL_WRITE_ONLY, /* not used */
 							   vertex_store->bufferobj); 
 
    assert(vertex_store->buffer);
    return vertex_store->buffer + vertex_store->used;
 }
 
-static void unmap_vertex_store( GLcontext *ctx, struct brw_save_vertex_store *vertex_store )
+static void unmap_vertex_store( GLcontext *ctx, struct vbo_save_vertex_store *vertex_store )
 {
    ctx->Driver.UnmapBuffer( ctx, GL_ARRAY_BUFFER_ARB, vertex_store->bufferobj );
    vertex_store->buffer = NULL;
 }
 
 
-static struct brw_save_primitive_store *alloc_prim_store( GLcontext *ctx )
+static struct vbo_save_primitive_store *alloc_prim_store( GLcontext *ctx )
 {
-   struct brw_save_primitive_store *store = CALLOC_STRUCT(brw_save_primitive_store);
+   struct vbo_save_primitive_store *store = CALLOC_STRUCT(vbo_save_primitive_store);
    (void) ctx;
    store->used = 0;
    store->refcount = 1;
@@ -225,7 +223,7 @@ static struct brw_save_primitive_store *alloc_prim_store( GLcontext *ctx )
 
 static void _save_reset_counters( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
 
    save->prim = save->prim_store->buffer + save->prim_store->used;
    save->buffer = (save->vertex_store->buffer + 
@@ -234,14 +232,14 @@ static void _save_reset_counters( GLcontext *ctx )
    assert(save->buffer == save->vbptr);
 
    if (save->vertex_size)
-      save->max_vert = ((BRW_SAVE_BUFFER_SIZE - save->vertex_store->used) / 
+      save->max_vert = ((VBO_SAVE_BUFFER_SIZE - save->vertex_store->used) / 
 			 save->vertex_size);
    else
       save->max_vert = 0;
 
    save->vert_count = 0;
    save->prim_count = 0;
-   save->prim_max = BRW_SAVE_PRIM_SIZE - save->prim_store->used;
+   save->prim_max = VBO_SAVE_PRIM_SIZE - save->prim_store->used;
    save->dangling_attr_ref = 0;
 }
 
@@ -251,13 +249,13 @@ static void _save_reset_counters( GLcontext *ctx )
  */
 static void _save_compile_vertex_list( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
-   struct brw_save_vertex_list *node;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
+   struct vbo_save_vertex_list *node;
 
    /* Allocate space for this structure in the display list currently
     * being compiled.
     */
-   node = (struct brw_save_vertex_list *)
+   node = (struct vbo_save_vertex_list *)
       _mesa_alloc_instruction(ctx, save->opcode_vertex_list, sizeof(*node));
 
    if (!node)
@@ -279,7 +277,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
    node->vertex_store->refcount++;
    node->prim_store->refcount++;
 
-   assert(node->attrsz[BRW_ATTRIB_POS] != 0 ||
+   assert(node->attrsz[VBO_ATTRIB_POS] != 0 ||
 	  node->count == 0);
 
    if (save->dangling_attr_ref)
@@ -301,7 +299,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
 
       _glapi_set_dispatch(ctx->Exec);
 
-      brw_loopback_vertex_list( ctx,
+      vbo_loopback_vertex_list( ctx,
 				(const GLfloat *)((const char *)save->vertex_store->buffer + 
 						  node->buffer_offset),
 				node->attrsz,
@@ -318,7 +316,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
     * the next vertex lists as well.
     */
    if (save->vertex_store->used > 
-       BRW_SAVE_BUFFER_SIZE - 16 * (save->vertex_size + 4)) {
+       VBO_SAVE_BUFFER_SIZE - 16 * (save->vertex_size + 4)) {
 
       /* Unmap old store:
        */
@@ -336,7 +334,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
       save->vbptr = map_vertex_store( ctx, save->vertex_store );
    } 
 
-   if (save->prim_store->used > BRW_SAVE_PRIM_SIZE - 6) {
+   if (save->prim_store->used > VBO_SAVE_PRIM_SIZE - 6) {
       save->prim_store->refcount--; 
       assert(save->prim_store->refcount != 0);
       save->prim_store = alloc_prim_store( ctx );
@@ -353,7 +351,7 @@ static void _save_compile_vertex_list( GLcontext *ctx )
  */
 static void _save_wrap_buffers( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLint i = save->prim_count - 1;
    GLenum mode;
    GLboolean weak;
@@ -391,7 +389,7 @@ static void _save_wrap_buffers( GLcontext *ctx )
  */
 static void _save_wrap_filled_vertex( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLfloat *data = save->copied.buffer;
    GLuint i;
 
@@ -414,38 +412,26 @@ static void _save_wrap_filled_vertex( GLcontext *ctx )
 
 static void _save_copy_to_current( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
    GLuint i;
 
-   for (i = BRW_ATTRIB_POS+1 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       if (save->attrsz[i]) {
 	 save->currentsz[i][0] = save->attrsz[i];
 	 COPY_CLEAN_4V(save->current[i], 
-		    save->attrsz[i], 
-		    save->attrptr[i]);
+		       save->attrsz[i], 
+		       save->attrptr[i]);
       }
    }
-
-   /* Edgeflag requires special treatment: 
-    *
-    * TODO: change edgeflag to GLfloat in Mesa.
-    */
-   if (save->attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      ctx->ListState.ActiveEdgeFlag = 1;
-      save->CurrentFloatEdgeFlag = 
-	 save->attrptr[BRW_ATTRIB_EDGEFLAG][0];
-      ctx->ListState.CurrentEdgeFlag = 
-	 (save->CurrentFloatEdgeFlag == 1.0);
-   }
 }
 
 
 static void _save_copy_from_current( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
    GLint i;
 
-   for (i = BRW_ATTRIB_POS+1 ; i < BRW_ATTRIB_MAX ; i++) 
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       switch (save->attrsz[i]) {
       case 4: save->attrptr[i][3] = save->current[i][3];
       case 3: save->attrptr[i][2] = save->current[i][2];
@@ -453,12 +439,6 @@ static void _save_copy_from_current( GLcontext *ctx )
       case 1: save->attrptr[i][0] = save->current[i][0];
       case 0: break;
       }
-
-   /* Edgeflag requires special treatment:
-    */
-   if (save->attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      save->CurrentFloatEdgeFlag = (GLfloat)ctx->ListState.CurrentEdgeFlag;
-      save->attrptr[BRW_ATTRIB_EDGEFLAG][0] = save->CurrentFloatEdgeFlag;
    }
 }
 
@@ -471,7 +451,7 @@ static void _save_upgrade_vertex( GLcontext *ctx,
 				 GLuint attr,
 				 GLuint newsz )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLuint oldsz;
    GLuint i;
    GLfloat *tmp;
@@ -496,13 +476,13 @@ static void _save_upgrade_vertex( GLcontext *ctx,
    save->attrsz[attr] = newsz;
 
    save->vertex_size += newsz - oldsz;
-   save->max_vert = ((BRW_SAVE_BUFFER_SIZE - save->vertex_store->used) / 
+   save->max_vert = ((VBO_SAVE_BUFFER_SIZE - save->vertex_store->used) / 
 		      save->vertex_size);
    save->vert_count = 0;
 
    /* Recalculate all the attrptr[] values:
     */
-   for (i = 0, tmp = save->vertex ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0, tmp = save->vertex ; i < VBO_ATTRIB_MAX ; i++) {
       if (save->attrsz[i]) {
 	 save->attrptr[i] = tmp;
 	 tmp += save->attrsz[i];
@@ -529,13 +509,13 @@ static void _save_upgrade_vertex( GLcontext *ctx,
 
       /* Need to note this and fix up at runtime (or loopback):
        */
-      if (attr != BRW_ATTRIB_POS && save->currentsz[attr][0] == 0) {
+      if (attr != VBO_ATTRIB_POS && save->currentsz[attr][0] == 0) {
 	 assert(oldsz == 0);
 	 save->dangling_attr_ref = GL_TRUE;
       }
 
       for (i = 0 ; i < save->copied.nr ; i++) {
-	 for (j = 0 ; j < BRW_ATTRIB_MAX ; j++) {
+	 for (j = 0 ; j < VBO_ATTRIB_MAX ; j++) {
 	    if (save->attrsz[j]) {
 	       if (j == attr) {
 		  if (oldsz) {
@@ -565,7 +545,7 @@ static void _save_upgrade_vertex( GLcontext *ctx,
 
 static void save_fixup_vertex( GLcontext *ctx, GLuint attr, GLuint sz )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
 
    if (sz > save->attrsz[attr]) {
       /* New size is larger.  Need to flush existing vertices and get
@@ -589,10 +569,10 @@ static void save_fixup_vertex( GLcontext *ctx, GLuint attr, GLuint sz )
 
 static void _save_reset_vertex( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLuint i;
 
-   for (i = 0 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = 0 ; i < VBO_ATTRIB_MAX ; i++) {
       save->attrsz[i] = 0;
       save->active_sz[i] = 0;
    }
@@ -613,7 +593,7 @@ static void _save_reset_vertex( GLcontext *ctx )
  */
 #define ATTR( A, N, V0, V1, V2, V3 )				\
 do {								\
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;	\
+   struct vbo_save_context *save = &vbo_context(ctx)->save;	\
 								\
    if (save->active_sz[A] != N)				\
       save_fixup_vertex(ctx, A, N);				\
@@ -641,7 +621,7 @@ do {								\
 
 #define TAG(x) _save_##x
 
-#include "brw_attrib_tmp.h"
+#include "vbo_attrib_tmp.h"
 
 
 
@@ -652,7 +632,7 @@ do {								\
  */
 #define DO_FALLBACK(ctx) 							\
 do {									\
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;					\
+   struct vbo_save_context *save = &vbo_context(ctx)->save;					\
 									\
    if (save->vert_count || save->prim_count) 						\
       _save_compile_vertex_list( ctx );					\
@@ -726,17 +706,17 @@ static void GLAPIENTRY _save_CallLists( GLsizei n, GLenum type, const GLvoid *v
 /* This begin is hooked into ...  Updating of
  * ctx->Driver.CurrentSavePrimitive is already taken care of.
  */
-GLboolean brw_save_NotifyBegin( GLcontext *ctx, GLenum mode )
+GLboolean vbo_save_NotifyBegin( GLcontext *ctx, GLenum mode )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
 
    GLuint i = save->prim_count++;
 
    assert(i < save->prim_max);
-   save->prim[i].mode = mode & ~BRW_SAVE_PRIM_WEAK;
+   save->prim[i].mode = mode & ~VBO_SAVE_PRIM_WEAK;
    save->prim[i].begin = 1;
    save->prim[i].end = 0;
-   save->prim[i].weak = (mode & BRW_SAVE_PRIM_WEAK) ? 1 : 0;
+   save->prim[i].weak = (mode & VBO_SAVE_PRIM_WEAK) ? 1 : 0;
    save->prim[i].pad = 0;
    save->prim[i].start = save->vert_count;
    save->prim[i].count = 0;   
@@ -751,7 +731,7 @@ GLboolean brw_save_NotifyBegin( GLcontext *ctx, GLenum mode )
 static void GLAPIENTRY _save_End( void )
 {
    GET_CURRENT_CONTEXT( ctx ); 
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
+   struct vbo_save_context *save = &vbo_context(ctx)->save; 
    GLint i = save->prim_count - 1;
 
    ctx->Driver.CurrentSavePrimitive = PRIM_OUTSIDE_BEGIN_END;
@@ -838,7 +818,7 @@ static void GLAPIENTRY _save_Begin( GLenum mode )
 static void GLAPIENTRY _save_OBE_Rectf( GLfloat x1, GLfloat y1, GLfloat x2, GLfloat y2 )
 {
    GET_CURRENT_CONTEXT(ctx);
-   brw_save_NotifyBegin( ctx, GL_QUADS | BRW_SAVE_PRIM_WEAK );
+   vbo_save_NotifyBegin( ctx, GL_QUADS | VBO_SAVE_PRIM_WEAK );
    CALL_Vertex2f(GET_DISPATCH(), ( x1, y1 ));
    CALL_Vertex2f(GET_DISPATCH(), ( x2, y1 ));
    CALL_Vertex2f(GET_DISPATCH(), ( x2, y2 ));
@@ -857,7 +837,8 @@ static void GLAPIENTRY _save_OBE_DrawArrays(GLenum mode, GLint start, GLsizei co
 
    _ae_map_vbos( ctx );
 
-   brw_save_NotifyBegin( ctx, mode | BRW_SAVE_PRIM_WEAK );
+   vbo_save_NotifyBegin( ctx, mode | VBO_SAVE_PRIM_WEAK );
+
    for (i = 0; i < count; i++)
        CALL_ArrayElement(GET_DISPATCH(), (start + i));
    CALL_End(GET_DISPATCH(), ());
@@ -879,7 +860,7 @@ static void GLAPIENTRY _save_OBE_DrawElements(GLenum mode, GLsizei count, GLenum
 
    _ae_map_vbos( ctx );
 
-   brw_save_NotifyBegin( ctx, mode | BRW_SAVE_PRIM_WEAK );
+   vbo_save_NotifyBegin( ctx, mode | VBO_SAVE_PRIM_WEAK );
 
    switch (type) {
    case GL_UNSIGNED_BYTE:
@@ -922,7 +903,7 @@ static void GLAPIENTRY _save_OBE_DrawRangeElements(GLenum mode,
 
 static void _save_vtxfmt_init( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLvertexformat *vfmt = &save->vtxfmt;
 
    vfmt->ArrayElement = _ae_loopback_array_elt;	        /* generic helper */
@@ -1007,9 +988,9 @@ static void _save_vtxfmt_init( GLcontext *ctx )
 }
 
 
-void brw_save_SaveFlushVertices( GLcontext *ctx )
+void vbo_save_SaveFlushVertices( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
 
    /* Noop when we are actually active:
     */
@@ -1027,9 +1008,9 @@ void brw_save_SaveFlushVertices( GLcontext *ctx )
    ctx->Driver.SaveNeedFlush = 0;
 }
 
-void brw_save_NewList( GLcontext *ctx, GLuint list, GLenum mode )
+void vbo_save_NewList( GLcontext *ctx, GLuint list, GLenum mode )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
 
    (void) list; (void) mode;
 
@@ -1046,36 +1027,36 @@ void brw_save_NewList( GLcontext *ctx, GLuint list, GLenum mode )
    ctx->Driver.SaveNeedFlush = 0;
 }
 
-void brw_save_EndList( GLcontext *ctx )
+void vbo_save_EndList( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    unmap_vertex_store( ctx, save->vertex_store );
 
    assert(save->vertex_size == 0);
 }
  
-void brw_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *dlist )
+void vbo_save_BeginCallList( GLcontext *ctx, struct mesa_display_list *dlist )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    save->replay_flags |= dlist->flags;
 }
 
-void brw_save_EndCallList( GLcontext *ctx )
+void vbo_save_EndCallList( GLcontext *ctx )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    
    if (ctx->ListState.CallDepth == 1) {
-      /* This is correct: want to keep only the BRW_SAVE_FALLBACK
+      /* This is correct: want to keep only the VBO_SAVE_FALLBACK
        * flag, if it is set:
        */
-      save->replay_flags &= BRW_SAVE_FALLBACK;
+      save->replay_flags &= VBO_SAVE_FALLBACK;
    }
 }
 
 
-static void brw_destroy_vertex_list( GLcontext *ctx, void *data )
+static void vbo_destroy_vertex_list( GLcontext *ctx, void *data )
 {
-   struct brw_save_vertex_list *node = (struct brw_save_vertex_list *)data;
+   struct vbo_save_vertex_list *node = (struct vbo_save_vertex_list *)data;
    (void) ctx;
 
    if ( --node->vertex_store->refcount == 0 ) 
@@ -1086,19 +1067,19 @@ static void brw_destroy_vertex_list( GLcontext *ctx, void *data )
 }
 
 
-static void brw_print_vertex_list( GLcontext *ctx, void *data )
+static void vbo_print_vertex_list( GLcontext *ctx, void *data )
 {
-   struct brw_save_vertex_list *node = (struct brw_save_vertex_list *)data;
+   struct vbo_save_vertex_list *node = (struct vbo_save_vertex_list *)data;
    GLuint i;
    (void) ctx;
 
-   _mesa_debug(NULL, "BRW-VERTEX-LIST, %u vertices %d primitives, %d vertsize\n",
+   _mesa_debug(NULL, "VBO-VERTEX-LIST, %u vertices %d primitives, %d vertsize\n",
                node->count,
 	       node->prim_count,
 	       node->vertex_size);
 
    for (i = 0 ; i < node->prim_count ; i++) {
-      struct brw_draw_prim *prim = &node->prim[i];
+      struct _mesa_prim *prim = &node->prim[i];
       _mesa_debug(NULL, "   prim %d: %s%s %d..%d %s %s\n",
 		  i, 
 		  _mesa_lookup_enum_by_nr(prim->mode),
@@ -1113,49 +1094,45 @@ static void brw_print_vertex_list( GLcontext *ctx, void *data )
 
 static void _save_current_init( GLcontext *ctx ) 
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
    GLint i;
 
-   for (i = 0; i < BRW_ATTRIB_FIRST_MATERIAL; i++) {
-      save->currentsz[i] = &ctx->ListState.ActiveAttribSize[i];
-      save->current[i] = ctx->ListState.CurrentAttrib[i];
+   for (i = VBO_ATTRIB_POS; i <= VBO_ATTRIB_GENERIC15; i++) {
+      const GLuint j = i - VBO_ATTRIB_POS;
+      ASSERT(j < VERT_ATTRIB_MAX);
+      save->currentsz[i] = &ctx->ListState.ActiveAttribSize[j];
+      save->current[i] = ctx->ListState.CurrentAttrib[j];
    }
 
-   for (i = BRW_ATTRIB_FIRST_MATERIAL; i <= BRW_ATTRIB_LAST_MATERIAL; i++) {
-      const GLuint j = i - BRW_ATTRIB_FIRST_MATERIAL;
+   for (i = VBO_ATTRIB_FIRST_MATERIAL; i <= VBO_ATTRIB_LAST_MATERIAL; i++) {
+      const GLuint j = i - VBO_ATTRIB_FIRST_MATERIAL;
       ASSERT(j < MAT_ATTRIB_MAX);
       save->currentsz[i] = &ctx->ListState.ActiveMaterialSize[j];
       save->current[i] = ctx->ListState.CurrentMaterial[j];
    }
-
-   save->currentsz[BRW_ATTRIB_INDEX] = &ctx->ListState.ActiveIndex;
-   save->current[BRW_ATTRIB_INDEX] = &ctx->ListState.CurrentIndex;
-
-   save->currentsz[BRW_ATTRIB_EDGEFLAG] = &ctx->ListState.ActiveEdgeFlag;
-   save->current[BRW_ATTRIB_EDGEFLAG] = &save->CurrentFloatEdgeFlag;
 }
 
 /**
  * Initialize the display list compiler
  */
-void brw_save_api_init( struct brw_save_context *save )
+void vbo_save_api_init( struct vbo_save_context *save )
 {
    GLcontext *ctx = save->ctx;
    GLuint i;
 
    save->opcode_vertex_list =
       _mesa_alloc_opcode( ctx,
-			  sizeof(struct brw_save_vertex_list),
-			  brw_save_playback_vertex_list,
-			  brw_destroy_vertex_list,
-			  brw_print_vertex_list );
+			  sizeof(struct vbo_save_vertex_list),
+			  vbo_save_playback_vertex_list,
+			  vbo_destroy_vertex_list,
+			  vbo_print_vertex_list );
 
-   ctx->Driver.NotifySaveBegin = brw_save_NotifyBegin;
+   ctx->Driver.NotifySaveBegin = vbo_save_NotifyBegin;
 
    _save_vtxfmt_init( ctx );
    _save_current_init( ctx );
 
-   for (i = 0; i < BRW_ATTRIB_MAX; i++)
+   for (i = 0; i < VBO_ATTRIB_MAX; i++)
       save->inputs[i] = &save->arrays[i];
 
    /* Hook our array functions into the outside-begin-end vtxfmt in 
diff --git a/src/mesa/drivers/dri/i965/brw_save_draw.c b/src/mesa/vbo/vbo_save_draw.c
index cebdd8d0607..a50d1cd9049 100644
--- a/src/mesa/drivers/dri/i965/brw_save_draw.c
+++ b/src/mesa/vbo/vbo_save_draw.c
@@ -34,19 +34,18 @@
 #include "light.h"
 #include "state.h"
 
-#include "brw_save.h"
-#include "brw_draw.h"
-#include "brw_fallback.h"
+#include "vbo_context.h"
+
 
 /*
  * After playback, copy everything but the position from the
  * last vertex to the saved state
  */
 static void _playback_copy_to_current( GLcontext *ctx,
-				       const struct brw_save_vertex_list *node )
+				       const struct vbo_save_vertex_list *node )
 {
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save; 
-   GLfloat vertex[BRW_ATTRIB_MAX * 4], *data = vertex;
+   struct vbo_context *vbo = vbo_context(ctx);
+   GLfloat vertex[VBO_ATTRIB_MAX * 4], *data = vertex;
    GLuint i, offset;
 
    if (node->count)
@@ -59,43 +58,34 @@ static void _playback_copy_to_current( GLcontext *ctx,
 				 node->vertex_size * sizeof(GLfloat), 
 				 data, node->vertex_store->bufferobj );
 
-   for (i = 0 ; i < BRW_ATTRIB_MAX ; i++) {
+   for (i = VBO_ATTRIB_POS+1 ; i < VBO_ATTRIB_MAX ; i++) {
       if (node->attrsz[i]) {
-	 if (i != BRW_ATTRIB_POS)
-	    COPY_CLEAN_4V(save->current[i], node->attrsz[i], data);
+	 GLfloat *current = (GLfloat *)vbo->currval[i].Ptr;
 
-	 if (i >= BRW_ATTRIB_MAT_FRONT_AMBIENT &&
-	     i <= BRW_ATTRIB_MAT_BACK_INDEXES)
-	    ctx->NewState |= _NEW_LIGHT;
+	 COPY_CLEAN_4V(current, 
+		       node->attrsz[i], 
+		       data);
 
-	 /* Edgeflag requires special treatment:
-	  */
-	 if (i == BRW_ATTRIB_EDGEFLAG)
-	    ctx->Current.EdgeFlag = (data[0] == 1.0);
+	 vbo->currval[i].Size = node->attrsz[i];
 
-	 data += node->attrsz[i] * sizeof(GLfloat);
-      }
-   }
+	 data += node->attrsz[i];
 
-   /* Edgeflag requires special treatment:
-    */
-   if (node->attrsz[BRW_ATTRIB_EDGEFLAG]) {
-      ctx->Current.EdgeFlag = (data[0] == 1.0);
+	 if (i >= VBO_ATTRIB_FIRST_MATERIAL &&
+	     i <= VBO_ATTRIB_LAST_MATERIAL)
+	    ctx->NewState |= _NEW_LIGHT;
+      }
    }
 
-
-#if 1
    /* Colormaterial -- this kindof sucks.
     */
    if (ctx->Light.ColorMaterialEnabled) {
-      _mesa_update_color_material(ctx, ctx->Current.Attrib[BRW_ATTRIB_COLOR0]);
+      _mesa_update_color_material(ctx, ctx->Current.Attrib[VBO_ATTRIB_COLOR0]);
    }
-#endif
 
    /* CurrentExecPrimitive
     */
    if (node->prim_count) {
-      const struct brw_draw_prim *prim = &node->prim[node->prim_count - 1];
+      const struct _mesa_prim *prim = &node->prim[node->prim_count - 1];
       if (prim->end)
 	 ctx->Driver.CurrentExecPrimitive = PRIM_OUTSIDE_BEGIN_END;
       else
@@ -108,16 +98,38 @@ static void _playback_copy_to_current( GLcontext *ctx,
 /* Treat the vertex storage as a VBO, define vertex arrays pointing
  * into it:
  */
-static void brw_bind_vertex_list( struct brw_save_context *save,
-                                   const struct brw_save_vertex_list *node )
+static void vbo_bind_vertex_list( GLcontext *ctx,
+                                   const struct vbo_save_vertex_list *node )
 {
+   struct vbo_context *vbo = vbo_context(ctx);
+   struct vbo_save_context *save = &vbo->save;
    struct gl_client_array *arrays = save->arrays;
    GLuint data = node->buffer_offset;
+   const GLuint *map;
    GLuint attr;
 
-   memset(arrays, 0, BRW_ATTRIB_MAX * sizeof(arrays[0]));
+   /* Install the default (ie Current) attributes first, then overlay
+    * all active ones.
+    */
+   switch (get_program_mode(ctx)) {
+   case VP_NONE:
+      memcpy(arrays,      vbo->legacy_currval, 16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->mat_currval,    MAT_ATTRIB_MAX * sizeof(arrays[0]));
+      map = vbo->map_vp_none;
+      break;
+   case VP_NV:
+   case VP_ARB:
+      /* The aliasing of attributes for NV vertex programs has already
+       * occurred.  NV vertex programs cannot access material values,
+       * nor attributes greater than VERT_ATTRIB_TEX7.  
+       */
+      memcpy(arrays,      vbo->legacy_currval,  16 * sizeof(arrays[0]));
+      memcpy(arrays + 16, vbo->generic_currval, 16 * sizeof(arrays[0]));
+      map = vbo->map_vp_arb;
+      break;
+   }
 
-   for (attr = 0; attr < BRW_ATTRIB_MAX; attr++) {
+   for (attr = 0; attr < VBO_ATTRIB_MAX; attr++) {
       if (node->attrsz[attr]) {
 	 arrays[attr].Ptr = (const GLubyte *)data;
 	 arrays[attr].Size = node->attrsz[attr];
@@ -135,15 +147,15 @@ static void brw_bind_vertex_list( struct brw_save_context *save,
    }
 }
 
-static void brw_save_loopback_vertex_list( GLcontext *ctx,
-					   const struct brw_save_vertex_list *list )
+static void vbo_save_loopback_vertex_list( GLcontext *ctx,
+					   const struct vbo_save_vertex_list *list )
 {
    const char *buffer = ctx->Driver.MapBuffer(ctx, 
 					      GL_ARRAY_BUFFER_ARB, 
-					      GL_DYNAMIC_READ_ARB, /* ? */
+					      GL_READ_ONLY, /* ? */
 					       list->vertex_store->bufferobj);
 
-   brw_loopback_vertex_list( ctx,
+   vbo_loopback_vertex_list( ctx,
 			     (const GLfloat *)(buffer + list->buffer_offset),
 			     list->attrsz,
 			     list->prim,
@@ -159,10 +171,10 @@ static void brw_save_loopback_vertex_list( GLcontext *ctx,
 /**
  * Execute the buffer and save copied verts.
  */
-void brw_save_playback_vertex_list( GLcontext *ctx, void *data )
+void vbo_save_playback_vertex_list( GLcontext *ctx, void *data )
 {
-   const struct brw_save_vertex_list *node = (const struct brw_save_vertex_list *) data;
-   struct brw_save_context *save = IMM_CONTEXT(ctx)->save;
+   const struct vbo_save_vertex_list *node = (const struct vbo_save_vertex_list *) data;
+   struct vbo_save_context *save = &vbo_context(ctx)->save;
 
    FLUSH_CURRENT(ctx, 0);
 
@@ -177,14 +189,14 @@ void brw_save_playback_vertex_list( GLcontext *ctx, void *data )
 	 if (0)
 	    _mesa_printf("displaylist recursive begin");
 
-	 brw_save_loopback_vertex_list( ctx, node );
+	 vbo_save_loopback_vertex_list( ctx, node );
 	 return;
       }
       else if (save->replay_flags) {
 	 /* Various degnerate cases: translate into immediate mode
 	  * calls rather than trying to execute in place.
 	  */
-	 brw_save_loopback_vertex_list( ctx, node );
+	 vbo_save_loopback_vertex_list( ctx, node );
 	 return;
       }
       
@@ -198,21 +210,15 @@ void brw_save_playback_vertex_list( GLcontext *ctx, void *data )
          return;
       }
 
-      brw_bind_vertex_list( save, node );
-
-      if (!brw_draw_prims( save->ctx, 
-			   save->inputs, 
-			   node->prim, 
-			   node->prim_count,
-			   NULL,
-			   0,	/* Node is a VBO, so this is ok */
-			   node->count,
-			   0 )) {
-	 brw_fallback(ctx);
-	 brw_save_loopback_vertex_list( ctx, node );
-	 brw_unfallback(ctx);
-	 return;
-      }
+      vbo_bind_vertex_list( ctx, node );
+
+      vbo_context(ctx)->draw_prims( ctx, 
+				    save->inputs, 
+				    node->prim, 
+				    node->prim_count,
+				    NULL,
+				    0,	/* Node is a VBO, so this is ok */
+				    node->count );
    }
 
    /* Copy to current?
diff --git a/src/mesa/vbo/vbo_save_loopback.c b/src/mesa/vbo/vbo_save_loopback.c
new file mode 100644
index 00000000000..430333b84dd
--- /dev/null
+++ b/src/mesa/vbo/vbo_save_loopback.c
@@ -0,0 +1,194 @@
+/**************************************************************************
+ * 
+ * Copyright 2005 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * 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"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, 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 NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS 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.
+ * 
+ **************************************************************************/
+
+#include "swrast_setup/swrast_setup.h"
+#include "swrast/swrast.h"
+#include "tnl/tnl.h"
+#include "context.h"
+
+#include "vbo_context.h"
+
+#include "glheader.h"
+#include "enums.h"
+#include "glapi.h"
+#include "imports.h"
+#include "macros.h"
+#include "mtypes.h"
+#include "dispatch.h"
+
+
+typedef void (*attr_func)( GLcontext *ctx, GLint target, const GLfloat * );
+
+
+/* This file makes heavy use of the aliasing of NV vertex attributes
+ * with the legacy attributes, and also with ARB and Material
+ * attributes as currently implemented.
+ */
+static void VertexAttrib1fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
+{
+   CALL_VertexAttrib1fvNV(ctx->Exec, (target, v));
+}
+
+static void VertexAttrib2fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
+{
+   CALL_VertexAttrib2fvNV(ctx->Exec, (target, v));
+}
+
+static void VertexAttrib3fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
+{
+   CALL_VertexAttrib3fvNV(ctx->Exec, (target, v));
+}
+
+static void VertexAttrib4fvNV(GLcontext *ctx, GLint target, const GLfloat *v)
+{
+   CALL_VertexAttrib4fvNV(ctx->Exec, (target, v));
+}
+
+static attr_func vert_attrfunc[4] = {
+   VertexAttrib1fvNV,
+   VertexAttrib2fvNV,
+   VertexAttrib3fvNV,
+   VertexAttrib4fvNV
+};
+
+struct loopback_attr {
+   GLint target;
+   GLint sz;
+   attr_func func;
+};
+
+/* Don't emit ends and begins on wrapped primitives.  Don't replay
+ * wrapped vertices.  If we get here, it's probably because the the
+ * precalculated wrapping is wrong.
+ */
+static void loopback_prim( GLcontext *ctx,
+			   const GLfloat *buffer,
+			   const struct _mesa_prim *prim,
+			   GLuint wrap_count,
+			   GLuint vertex_size,
+			   const struct loopback_attr *la, GLuint nr )
+{
+   GLint start = prim->start;
+   GLint end = start + prim->count;
+   const GLfloat *data;
+   GLint j;
+   GLuint k;
+
+   if (0)
+      _mesa_printf("loopback prim %s(%s,%s) verts %d..%d\n",
+		   _mesa_lookup_enum_by_nr(prim->mode),
+		   prim->begin ? "begin" : "..",
+		   prim->end ? "end" : "..",
+		   start, 
+		   end);
+
+   if (prim->begin) {
+      CALL_Begin(GET_DISPATCH(), ( prim->mode ));
+   }
+   else {
+      assert(start == 0);
+      start += wrap_count;
+   }
+
+   data = buffer + start * vertex_size;
+
+   for (j = start ; j < end ; j++) {
+      const GLfloat *tmp = data + la[0].sz;
+
+      for (k = 1 ; k < nr ; k++) {
+	 la[k].func( ctx, la[k].target, tmp );
+	 tmp += la[k].sz;
+      }
+	 
+      /* Fire the vertex
+       */
+      la[0].func( ctx, VBO_ATTRIB_POS, data );
+      data = tmp;
+   }
+
+   if (prim->end) {
+      CALL_End(GET_DISPATCH(), ());
+   }
+}
+
+/* Primitives generated by DrawArrays/DrawElements/Rectf may be
+ * caught here.  If there is no primitive in progress, execute them
+ * normally, otherwise need to track and discard the generated
+ * primitives.
+ */
+static void loopback_weak_prim( GLcontext *ctx,
+				const struct _mesa_prim *prim )
+{
+   /* Use the prim_weak flag to ensure that if this primitive
+    * wraps, we don't mistake future vertex_lists for part of the
+    * surrounding primitive.
+    *
+    * While this flag is set, we are simply disposing of data
+    * generated by an operation now known to be a noop.
+    */
+   if (prim->begin)
+      ctx->Driver.CurrentExecPrimitive |= VBO_SAVE_PRIM_WEAK;
+   if (prim->end)
+      ctx->Driver.CurrentExecPrimitive &= ~VBO_SAVE_PRIM_WEAK;
+}
+
+
+void vbo_loopback_vertex_list( GLcontext *ctx,
+			       const GLfloat *buffer,
+			       const GLubyte *attrsz,
+			       const struct _mesa_prim *prim,
+			       GLuint prim_count,
+			       GLuint wrap_count,
+			       GLuint vertex_size)
+{
+   struct loopback_attr la[VBO_ATTRIB_MAX];
+   GLuint i, nr = 0;
+
+   /* All Legacy, NV, ARB and Material attributes are routed through
+    * the NV attributes entrypoints:
+    */
+   for (i = 0 ; i < VBO_ATTRIB_MAX ; i++) {
+      if (attrsz[i]) {
+	 la[nr].target = i;
+	 la[nr].sz = attrsz[i];
+	 la[nr].func = vert_attrfunc[attrsz[i]-1];
+	 nr++;
+      }
+   }
+
+   for (i = 0 ; i < prim_count ; i++) {
+      if ((prim[i].mode & VBO_SAVE_PRIM_WEAK) &&
+	  (ctx->Driver.CurrentExecPrimitive != PRIM_OUTSIDE_BEGIN_END))
+      {
+	 loopback_weak_prim( ctx, &prim[i] );
+      }
+      else
+      {
+	 loopback_prim( ctx, buffer, &prim[i], wrap_count, vertex_size, la, nr );
+      }
+   }
+}
diff --git a/src/mesa/vbo/vbo_split.c b/src/mesa/vbo/vbo_split.c
new file mode 100644
index 00000000000..ef205a3bb1f
--- /dev/null
+++ b/src/mesa/vbo/vbo_split.c
@@ -0,0 +1,161 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  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"),
+ * 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 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
+ * BRIAN PAUL 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:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/* Deal with hardware and/or swtnl maximums:
+ * - maximum number of vertices in buffer
+ * - maximum number of elements (maybe zero)
+ *
+ * The maximums may vary with opengl state (eg if a larger hardware
+ * vertex is required in this state, the maximum number of vertices
+ * may be smaller than in another state).
+ *
+ * We want buffer splitting to be a convenience function for the code
+ * actually drawing the primitives rather than a system-wide maximum,
+ * otherwise it is hard to avoid pessimism.  
+ *
+ * For instance, if a driver has no hardware limits on vertex buffer
+ * dimensions, it would not ordinarily want to split vbos.  But if
+ * there is an unexpected fallback, eg memory manager fails to upload
+ * textures, it will want to pass the drawing commands onto swtnl,
+ * which does have limitations.  A convenience function allows swtnl
+ * to split the drawing and vbos internally without imposing its
+ * limitations on drivers which want to use it as a fallback path.
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "mtypes.h"
+
+#include "vbo_split.h"
+#include "vbo.h"
+
+/* True if a primitive can be split without copying of vertices, false
+ * otherwise.
+ */
+GLboolean split_prim_inplace(GLenum mode, GLuint *first, GLuint *incr)
+{
+   switch (mode) {
+   case GL_POINTS:
+      *first = 1;
+      *incr = 1;
+      return GL_TRUE;
+   case GL_LINES:
+      *first = 2;
+      *incr = 2;
+      return GL_TRUE;
+   case GL_LINE_STRIP:
+      *first = 2;
+      *incr = 1;
+      return GL_TRUE;
+   case GL_TRIANGLES:
+      *first = 3;
+      *incr = 3;
+      return GL_TRUE;
+   case GL_TRIANGLE_STRIP:
+      *first = 3;
+      *incr = 1;
+      return GL_TRUE;
+   case GL_QUADS:
+      *first = 4;
+      *incr = 4;
+      return GL_TRUE;
+   case GL_QUAD_STRIP:
+      *first = 4;
+      *incr = 2;
+      return GL_TRUE;
+   default:
+      *first = 0;
+      *incr = 1;		/* so that count % incr works */
+      return GL_FALSE;
+   }
+}
+
+
+
+void vbo_split_prims( GLcontext *ctx,
+		      const struct gl_client_array *arrays[],
+		      const struct _mesa_prim *prim,
+		      GLuint nr_prims,
+		      const struct _mesa_index_buffer *ib,
+		      GLuint min_index,
+		      GLuint max_index,
+		      vbo_draw_func draw,
+		      const struct split_limits *limits )
+{
+  
+   if (ib) {
+      if (limits->max_indices == 0) {
+	 /* Could traverse the indices, re-emitting vertices in turn.
+	  * But it's hard to see why this case would be needed - for
+	  * software tnl, it is better to convert to non-indexed
+	  * rendering after transformation is complete, as is done in
+	  * the t_dd_rendertmp.h templates.  Are there any devices
+	  * with hardware tnl that cannot do indexed rendering?
+	  *
+	  * For now, this path is disabled.
+	  */
+	 assert(0);
+      }
+      else if (max_index - min_index >= limits->max_verts) {
+	 /* The vertex buffers are too large for hardware (or the
+	  * swtnl module).  Traverse the indices, re-emitting vertices
+	  * in turn.  Use a vertex cache to preserve some of the
+	  * sharing from the original index list.
+	  */
+	 vbo_split_copy(ctx, arrays, prim, nr_prims, ib,
+			draw, limits );
+      }
+      else if (ib->count > limits->max_indices) {
+	 /* The index buffer is too large for hardware.  Try to split
+	  * on whole-primitive boundaries, otherwise try to split the
+	  * individual primitives.
+	  */
+	 vbo_split_inplace(ctx, arrays, prim, nr_prims, ib,
+			   min_index, max_index, draw, limits );
+      }
+      else {
+	 /* Why were we called? */
+	 assert(0);
+      }
+   }
+   else {
+      if (max_index - min_index >= limits->max_verts) {
+	 /* The vertex buffer is too large for hardware (or the swtnl
+	  * module).  Try to split on whole-primitive boundaries,
+	  * otherwise try to split the individual primitives.
+	  */
+	 vbo_split_inplace(ctx, arrays, prim, nr_prims, ib,
+			   min_index, max_index, draw, limits );
+      }
+      else {
+	 /* Why were we called? */
+	 assert(0);
+      }
+   }
+}
+
diff --git a/src/mesa/vbo/vbo_split.h b/src/mesa/vbo/vbo_split.h
new file mode 100644
index 00000000000..05888d048cb
--- /dev/null
+++ b/src/mesa/vbo/vbo_split.h
@@ -0,0 +1,72 @@
+/*
+ * mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  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"),
+ * 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 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
+ * BRIAN PAUL 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.
+ */
+
+/**
+ * \file vbo_context.h
+ * \brief VBO builder module datatypes and definitions.
+ * \author Keith Whitwell
+ */
+
+
+/**
+ * \mainpage The VBO splitter
+ *
+ * This is the private data used internally to the vbo_split_prims()
+ * helper function.  Nobody outside the vbo_split* files needs to
+ * include or know about this structure.
+ */
+
+
+#ifndef _VBO_SPLIT_H
+#define _VBO_SPLIT_H
+
+#include "vbo.h"
+
+
+/* True if a primitive can be split without copying of vertices, false
+ * otherwise.
+ */
+GLboolean split_prim_inplace(GLenum mode, GLuint *first, GLuint *incr);
+
+void vbo_split_inplace( GLcontext *ctx,
+			const struct gl_client_array *arrays[],
+			const struct _mesa_prim *prim,
+			GLuint nr_prims,
+			const struct _mesa_index_buffer *ib,
+			GLuint min_index,
+			GLuint max_index,
+			vbo_draw_func draw,
+			const struct split_limits *limits );
+
+/* Requires ib != NULL:
+ */
+void vbo_split_copy( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     vbo_draw_func draw,
+		     const struct split_limits *limits );
+
+#endif
diff --git a/src/mesa/vbo/vbo_split_copy.c b/src/mesa/vbo/vbo_split_copy.c
new file mode 100644
index 00000000000..e142dde6803
--- /dev/null
+++ b/src/mesa/vbo/vbo_split_copy.c
@@ -0,0 +1,551 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  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"),
+ * 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 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
+ * BRIAN PAUL 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:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/* Split indexed primitives with per-vertex copying.
+ */
+
+#include "glheader.h"
+#include "imports.h"
+#include "macros.h"
+#include "enums.h"
+#include "mtypes.h"
+
+#include "vbo_split.h"
+#include "vbo.h"
+
+
+#define ELT_TABLE_SIZE 16
+
+/* Used for vertex-level splitting of indexed buffers.  Note that
+ * non-indexed primitives may be converted to indexed in some cases
+ * (eg loops, fans) in order to use this splitting path.
+ */
+struct copy_context {
+
+   GLcontext *ctx;
+   const struct gl_client_array **array;
+   const struct _mesa_prim *prim;
+   GLuint nr_prims;
+   const struct _mesa_index_buffer *ib;
+   vbo_draw_func draw;
+
+   const struct split_limits *limits;
+
+   struct {
+      GLuint attr;
+      GLuint size;
+      const struct gl_client_array *array;
+      const GLubyte *src_ptr;
+
+      struct gl_client_array dstarray;
+
+   } varying[VERT_ATTRIB_MAX];
+   GLuint nr_varying;
+
+   const struct gl_client_array *dstarray_ptr[VERT_ATTRIB_MAX];
+   struct _mesa_index_buffer dstib;
+
+   GLuint *translated_elt_buf;
+   const GLuint *srcelt;
+
+   /* A baby hash table to avoid re-emitting (some) duplicate
+    * vertices when splitting indexed primitives.
+    */
+   struct { 
+      GLuint in;
+      GLuint out;
+   } vert_cache[ELT_TABLE_SIZE];
+      
+
+   GLuint vertex_size;
+   GLubyte *dstbuf;
+   GLubyte *dstptr;		/* dstptr == dstbuf + dstelt_max * vertsize */
+   GLuint dstbuf_size;	/* in vertices */
+   GLuint dstbuf_nr;		/* count of emitted vertices, also the
+				 * largest value in dstelt.  Our
+				 * MaxIndex.
+				 */
+
+   GLuint *dstelt;
+   GLuint dstelt_nr;
+   GLuint dstelt_size;
+
+#define MAX_PRIM 32
+   struct _mesa_prim dstprim[MAX_PRIM];
+   GLuint dstprim_nr;
+
+};
+
+
+static GLuint type_size( GLenum type )
+{
+   switch(type) {
+   case GL_BYTE: return sizeof(GLbyte);
+   case GL_UNSIGNED_BYTE: return sizeof(GLubyte);
+   case GL_SHORT: return sizeof(GLshort);
+   case GL_UNSIGNED_SHORT: return sizeof(GLushort);
+   case GL_INT: return sizeof(GLint);
+   case GL_UNSIGNED_INT: return sizeof(GLuint);
+   case GL_FLOAT: return sizeof(GLfloat);
+   case GL_DOUBLE: return sizeof(GLdouble);
+   default: return 0;
+   }
+}
+
+static GLuint attr_size( const struct gl_client_array *array )
+{
+   return array->Size * type_size(array->Type);
+}
+
+
+/* Starts returning true slightly before the buffer fills, to ensure
+ * that there is sufficient room for any remaining vertices to finish
+ * off the prim:
+ */
+static GLboolean check_flush( struct copy_context *copy )
+{
+   if (copy->dstbuf_nr + 4 > copy->dstbuf_size)
+      return GL_TRUE;
+
+   if (copy->dstelt_nr + 4 > copy->dstelt_size)
+      return GL_TRUE;
+
+   return GL_FALSE;
+}
+
+static void flush( struct copy_context *copy )
+{
+   GLuint i;
+
+   /* Set some counters: 
+    */
+   copy->dstib.count = copy->dstelt_nr;
+
+   copy->draw( copy->ctx,
+	       copy->dstarray_ptr,
+	       copy->dstprim,
+	       copy->dstprim_nr,
+	       &copy->dstib,
+	       0,
+	       copy->dstbuf_nr );
+
+   /* Reset all pointers: 
+    */
+   copy->dstprim_nr = 0;
+   copy->dstelt_nr = 0;
+   copy->dstbuf_nr = 0;
+   copy->dstptr = copy->dstbuf;
+
+   /* Clear the vertex cache:
+    */
+   for (i = 0; i < ELT_TABLE_SIZE; i++)
+      copy->vert_cache[i].in = ~0;
+}
+
+
+
+static void begin( struct copy_context *copy, GLenum mode, GLboolean begin_flag )
+{
+   struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
+
+/*    _mesa_printf("begin %s (%d)\n", _mesa_lookup_enum_by_nr(mode), begin_flag); */
+		
+   prim->mode = mode;
+   prim->begin = begin_flag;
+}
+
+
+/* Use a hashtable to attempt to identify recently-emitted vertices
+ * and avoid re-emitting them.
+ */
+static GLuint elt(struct copy_context *copy, GLuint elt_idx)
+{
+   GLuint elt = copy->srcelt[elt_idx];
+   GLuint slot = elt & (ELT_TABLE_SIZE-1);
+
+/*    _mesa_printf("elt %d\n", elt); */
+
+   /* Look up the incoming element in the vertex cache.  Re-emit if
+    * necessary.   
+    */
+   if (copy->vert_cache[slot].in != elt) {
+      GLubyte *csr = copy->dstptr;
+      GLuint i;
+
+/*       _mesa_printf("  --> emit to dstelt %d\n", copy->dstbuf_nr); */
+
+      for (i = 0; i < copy->nr_varying; i++) {
+	 const struct gl_client_array *srcarray = copy->varying[i].array;
+	 const GLubyte *srcptr = copy->varying[i].src_ptr + elt * srcarray->StrideB;
+
+	 memcpy(csr, srcptr, copy->varying[i].size);
+	 csr += copy->varying[i].size;
+
+	 if (0) 
+	 {
+	    const GLuint *f = (const GLuint *)srcptr;
+	    GLuint j;
+	    _mesa_printf("  varying %d: ", i);
+	    for(j = 0; j < copy->varying[i].size / 4; j++)
+	       _mesa_printf("%x ", f[j]);
+	    _mesa_printf("\n");
+	 }
+	       
+      }
+
+      copy->vert_cache[slot].in = elt;
+      copy->vert_cache[slot].out = copy->dstbuf_nr++;
+      copy->dstptr += copy->vertex_size;
+
+      assert(csr == copy->dstptr);
+      assert(copy->dstptr == (copy->dstbuf + 
+				    copy->dstbuf_nr * 
+				    copy->vertex_size));
+   }
+/*    else */
+/*       _mesa_printf("  --> reuse vertex\n"); */
+   
+/*    _mesa_printf("  --> emit %d\n", copy->vert_cache[slot].out); */
+   copy->dstelt[copy->dstelt_nr++] = copy->vert_cache[slot].out;
+   return check_flush(copy);
+}
+
+static void end( struct copy_context *copy, GLboolean end_flag )
+{
+   struct _mesa_prim *prim = &copy->dstprim[copy->dstprim_nr];
+
+/*    _mesa_printf("end (%d)\n", end_flag); */
+
+   prim->end = end_flag;
+   prim->count = copy->dstelt_nr - prim->start;
+
+   if (++copy->dstprim_nr == MAX_PRIM ||
+       check_flush(copy)) 
+      flush(copy);
+}
+
+
+
+static void replay_elts( struct copy_context *copy )
+{
+   GLuint i, j, k;
+   GLboolean split;
+
+   for (i = 0; i < copy->nr_prims; i++) {
+      const struct _mesa_prim *prim = &copy->prim[i];
+      const GLuint start = prim->start;
+      GLuint first, incr;
+
+      switch (prim->mode) {
+	 
+      case GL_LINE_LOOP:
+	 /* Convert to linestrip and emit the final vertex explicitly,
+	  * but only in the resultant strip that requires it.
+	  */
+	 j = 0;
+	 while (j != prim->count) {
+	    begin(copy, GL_LINE_STRIP, prim->begin && j == 0);
+
+	    for (split = GL_FALSE; j != prim->count && !split; j++)
+	       split = elt(copy, start + j);
+
+	    if (j == prim->count) {
+	       /* Done, emit final line.  Split doesn't matter as
+		* it is always raised a bit early so we can emit
+		* the last verts if necessary!
+		*/
+	       if (prim->end) 
+		  (void)elt(copy, start + 0);
+
+	       end(copy, prim->end);
+	    }
+	    else {
+	       /* Wrap
+		*/
+	       assert(split);
+	       end(copy, 0);
+	       j--;
+	    }
+	 }
+	 break;
+
+      case GL_TRIANGLE_FAN:
+      case GL_POLYGON:
+	 j = 2;
+	 while (j != prim->count) {
+	    begin(copy, prim->mode, prim->begin && j == 0);
+
+	    split = elt(copy, start+0); 
+	    assert(!split);
+
+	    split = elt(copy, start+j-1); 
+	    assert(!split);
+
+	    for (; j != prim->count && !split; j++)
+	       split = elt(copy, start+j);
+
+	    end(copy, prim->end && j == prim->count);
+
+	    if (j != prim->count) {
+	       /* Wrapped the primitive, need to repeat some vertices:
+		*/
+	       j -= 1;
+	    }
+	 }
+	 break;
+
+      default:
+	 (void)split_prim_inplace(prim->mode, &first, &incr);
+	 
+	 j = 0;
+	 while (j != prim->count) {
+
+	    begin(copy, prim->mode, prim->begin && j == 0);
+
+	    split = 0;
+	    for (k = 0; k < first; k++, j++)
+	       split |= elt(copy, start+j);
+
+	    assert(!split);
+
+	    for (; j != prim->count && !split; )
+	       for (k = 0; k < incr; k++, j++)
+		  split |= elt(copy, start+j);
+
+	    end(copy, prim->end && j == prim->count);
+
+	    if (j != prim->count) {
+	       /* Wrapped the primitive, need to repeat some vertices:
+		*/
+	       assert(j > first - incr);
+	       j -= (first - incr);
+	    }
+	 }
+	 break;
+      }
+   }
+
+   if (copy->dstprim_nr)
+      flush(copy);
+}
+
+
+static void replay_init( struct copy_context *copy )
+{
+   GLcontext *ctx = copy->ctx;
+   GLuint i;
+   GLuint offset;
+   const GLvoid *srcptr;
+
+   /* Make a list of varying attributes and their vbo's.  Also
+    * calculate vertex size.
+    */
+   copy->vertex_size = 0;
+   for (i = 0; i < VERT_ATTRIB_MAX; i++) {
+      struct gl_buffer_object *vbo = copy->array[i]->BufferObj;
+
+      if (copy->array[i]->StrideB == 0) {
+	 copy->dstarray_ptr[i] = copy->array[i];
+      }
+      else {
+	 GLuint j = copy->nr_varying++;
+	 
+	 copy->varying[j].attr = i;
+	 copy->varying[j].array = copy->array[i];
+	 copy->varying[j].size = attr_size(copy->array[i]);
+	 copy->vertex_size += attr_size(copy->array[i]);
+      
+	 if (vbo->Name && !vbo->Pointer) 
+	    ctx->Driver.MapBuffer(ctx,
+				  GL_ARRAY_BUFFER_ARB, 
+				  GL_WRITE_ONLY, /* XXX */
+				  vbo);
+
+	 copy->varying[j].src_ptr = ADD_POINTERS(vbo->Pointer,
+						 copy->array[i]->Ptr);
+
+	 copy->dstarray_ptr[i] = &copy->varying[j].dstarray;
+      }
+   }
+
+   /* There must always be an index buffer.  Currently require the
+    * caller convert non-indexed prims to indexed.  Could alternately
+    * do it internally.
+    */
+   if (copy->ib->obj->Name && !copy->ib->obj->Pointer) 
+      ctx->Driver.MapBuffer(ctx, 
+			    GL_ARRAY_BUFFER_ARB, /* XXX */
+			    GL_WRITE_ONLY, /* XXX */
+			    copy->ib->obj);
+
+   srcptr = (const GLubyte *)ADD_POINTERS(copy->ib->obj->Pointer, copy->ib->ptr);
+
+   switch (copy->ib->type) {
+   case GL_UNSIGNED_BYTE:
+      copy->translated_elt_buf = _mesa_malloc(sizeof(GLuint) * copy->ib->count);
+      copy->srcelt = copy->translated_elt_buf;
+
+      for (i = 0; i < copy->ib->count; i++)
+	 copy->translated_elt_buf[i] = ((const GLubyte *)srcptr)[i];
+      break;
+
+   case GL_UNSIGNED_SHORT:
+      copy->translated_elt_buf = _mesa_malloc(sizeof(GLuint) * copy->ib->count);
+      copy->srcelt = copy->translated_elt_buf;
+
+      for (i = 0; i < copy->ib->count; i++)
+	 copy->translated_elt_buf[i] = ((const GLushort *)srcptr)[i];
+      break;
+
+   case GL_UNSIGNED_INT:
+      copy->translated_elt_buf = NULL;
+      copy->srcelt = (const GLuint *)srcptr;
+      break;
+   }
+   
+
+   /* Figure out the maximum allowed vertex buffer size:
+    */
+   if (copy->vertex_size * copy->limits->max_verts <= copy->limits->max_vb_size) {
+      copy->dstbuf_size = copy->limits->max_verts;
+   }
+   else {
+      copy->dstbuf_size = copy->limits->max_vb_size / copy->vertex_size;
+   }
+
+   /* Allocate an output vertex buffer:
+    *
+    * XXX:  This should be a VBO!
+    */
+   copy->dstbuf = _mesa_malloc(copy->dstbuf_size * 
+			       copy->vertex_size);   
+   copy->dstptr = copy->dstbuf;
+
+   /* Setup new vertex arrays to point into the output buffer: 
+    */
+   for (offset = 0, i = 0; i < copy->nr_varying; i++) {
+      const struct gl_client_array *src = copy->varying[i].array;
+      struct gl_client_array *dst = &copy->varying[i].dstarray;
+
+      dst->Size = src->Size;
+      dst->Type = src->Type;
+      dst->Stride = copy->vertex_size;
+      dst->StrideB = copy->vertex_size;
+      dst->Ptr = copy->dstbuf + offset;
+      dst->Enabled = GL_TRUE;
+      dst->Normalized = GL_TRUE;
+      dst->BufferObj = ctx->Array.NullBufferObj;
+      dst->_MaxElement = copy->dstbuf_size; /* may be less! */
+
+      offset += copy->varying[i].size;
+   }
+
+   /* Allocate an output element list:
+    */
+   copy->dstelt_size = MIN2(65536,
+			    copy->ib->count * 2 + 3);
+   copy->dstelt_size = MIN2(copy->dstelt_size,
+			    copy->limits->max_indices);
+   copy->dstelt = _mesa_malloc(sizeof(GLuint) * copy->dstelt_size);
+   copy->dstelt_nr = 0;
+
+   /* Setup the new index buffer to point to the allocated element
+    * list:
+    */
+   copy->dstib.count = 0;	/* duplicates dstelt_nr */
+   copy->dstib.type = GL_UNSIGNED_INT;
+   copy->dstib.obj = ctx->Array.NullBufferObj;
+   copy->dstib.ptr = copy->dstelt;
+}
+
+
+static void replay_finish( struct copy_context *copy )
+{
+   GLcontext *ctx = copy->ctx;
+   GLuint i;
+
+   /* Free our vertex and index buffers: 
+    */
+   _mesa_free(copy->translated_elt_buf);
+   _mesa_free(copy->dstbuf);
+   _mesa_free(copy->dstelt);
+   
+   /* Unmap VBO's 
+    */
+   for (i = 0; i < copy->nr_varying; i++) {
+      struct gl_buffer_object *vbo = copy->varying[i].array->BufferObj;
+
+      if (vbo->Name && vbo->Pointer) 
+	 ctx->Driver.UnmapBuffer(ctx, GL_ARRAY_BUFFER_ARB, vbo);
+   }
+
+   /* Unmap index buffer:
+    */
+   if (copy->ib->obj->Name && copy->ib->obj->Pointer) {
+      ctx->Driver.UnmapBuffer(ctx, 
+			      GL_ARRAY_BUFFER_ARB, /* XXX */
+			      copy->ib->obj);
+   }
+}
+
+void vbo_split_copy( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     vbo_draw_func draw,
+		     const struct split_limits *limits )
+{
+   struct copy_context copy;
+   GLuint i;
+
+   memset(&copy, 0, sizeof(copy));
+
+   /* Require indexed primitives:
+    */
+   assert(ib);
+   
+   copy.ctx = ctx;
+   copy.array = arrays;
+   copy.prim = prim;
+   copy.nr_prims = nr_prims;
+   copy.ib = ib;
+   copy.draw = draw;
+   copy.limits = limits;
+
+
+   /* Clear the vertex cache:
+    */
+   for (i = 0; i < ELT_TABLE_SIZE; i++)
+      copy.vert_cache[i].in = ~0;
+
+
+   replay_init(&copy);
+   replay_elts(&copy);
+   replay_finish(&copy);
+}
diff --git a/src/mesa/vbo/vbo_split_inplace.c b/src/mesa/vbo/vbo_split_inplace.c
new file mode 100644
index 00000000000..ea62866e7c9
--- /dev/null
+++ b/src/mesa/vbo/vbo_split_inplace.c
@@ -0,0 +1,286 @@
+
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * Copyright (C) 1999-2006  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"),
+ * 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 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
+ * BRIAN PAUL 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:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#include "mtypes.h"
+#include "macros.h"
+#include "enums.h"
+#include "vbo_split.h"
+
+
+#define MAX_PRIM 32
+
+/* Used for splitting without copying.
+ */
+struct split_context {
+   GLcontext *ctx;
+   const struct gl_client_array **array;
+   const struct _mesa_prim *prim;
+   GLuint nr_prims;
+   const struct _mesa_index_buffer *ib;
+   GLuint min_index;
+   GLuint max_index;
+   vbo_draw_func draw;
+
+   const struct split_limits *limits;
+
+   struct _mesa_prim dstprim[MAX_PRIM];
+   GLuint dstprim_nr;
+};
+
+
+
+
+static void flush_vertex( struct split_context *split )
+{
+   GLint min_index, max_index;
+
+   if (!split->dstprim_nr) 
+      return;
+
+   if (split->ib) {
+      /* This should basically be multipass rendering over the same
+       * unchanging set of VBO's.  Would like the driver not to
+       * re-upload the data, or swtnl not to re-transform the
+       * vertices.
+       */
+      assert(split->max_index - split->min_index < split->limits->max_verts);
+      min_index = split->min_index;
+      max_index = split->max_index;
+   }
+   else {
+      /* Non-indexed rendering.  Cannot assume that the primitives are
+       * ordered by increasing vertex, because of entrypoints like
+       * MultiDrawArrays.
+       */
+      GLuint i;
+      min_index = split->dstprim[0].start;
+      max_index = min_index + split->dstprim[0].count - 1;
+
+      for (i = 1; i < split->dstprim_nr; i++) {
+	 GLuint tmp_min = split->dstprim[i].start;
+	 GLuint tmp_max = tmp_min + split->dstprim[i].count - 1;
+
+	 if (tmp_min < min_index) 
+	    min_index = tmp_min;
+
+	 if (tmp_max > max_index) 
+	    max_index = tmp_max;
+      }
+   }
+
+   assert(max_index >= min_index);
+
+   split->draw( split->ctx, 
+		split->array, 
+		split->dstprim,
+		split->dstprim_nr,
+		NULL,
+		min_index,
+		max_index);
+
+   split->dstprim_nr = 0;
+}
+
+
+static struct _mesa_prim *next_outprim( struct split_context *split )
+{
+   if (split->dstprim_nr == MAX_PRIM-1) {
+      flush_vertex(split);
+   }
+
+   {
+      struct _mesa_prim *prim = &split->dstprim[split->dstprim_nr++];
+      memset(prim, 0, sizeof(*prim));
+      return prim;
+   }
+}
+
+static int align(int value, int alignment)
+{
+   return (value + alignment - 1) & ~(alignment - 1);
+}
+
+
+
+/* Break large primitives into smaller ones.  If not possible, convert
+ * the primitive to indexed and pass to split_elts().
+ */
+static void split_prims( struct split_context *split) 
+{
+   GLuint csr = 0;
+   GLuint i;
+
+   for (i = 0; i < split->nr_prims; i++) {
+      const struct _mesa_prim *prim = &split->prim[i];
+      GLuint first, incr;
+      GLboolean split_inplace = split_prim_inplace(prim->mode, &first, &incr);
+      GLuint count;
+
+      /* Always wrap on an even numbered vertex to avoid problems with
+       * triangle strips.  
+       */
+      GLuint available = align(split->limits->max_verts - csr - 1, 2); 
+      assert(split->limits->max_verts >= csr);
+
+      if (prim->count < first)
+	 continue;
+      
+      count = prim->count - (prim->count - first) % incr; 
+
+
+      if ((available < count && !split_inplace) || 
+	  (available < first && split_inplace)) {
+	 flush_vertex(split);
+	 csr = 0;
+	 available = align(split->limits->max_verts - csr - 1, 2);
+      }
+      
+      if (available >= count) {
+	 struct _mesa_prim *outprim = next_outprim(split);
+	 *outprim = *prim;
+	 csr += prim->count;
+	 available = align(split->limits->max_verts - csr - 1, 2); 
+      } 
+      else if (split_inplace) {
+	 GLuint j, nr;
+
+
+	 for (j = 0 ; j < count ; ) {
+	    GLuint remaining = count - j;
+	    struct _mesa_prim *outprim = next_outprim(split);
+
+	    nr = MIN2( available, remaining );
+	    
+	    nr -= (nr - first) % incr;
+	    
+	    outprim->mode = prim->mode;
+	    outprim->begin = (j == 0 && prim->begin);
+	    outprim->end = (nr == remaining && prim->end);
+	    outprim->start = prim->start + j;
+	    outprim->count = nr;
+	    
+	    if (nr == remaining) {
+	       /* Finished. 
+		*/
+	       j += nr;		
+	       csr += nr;
+	       available = align(split->limits->max_verts - csr - 1, 2); 
+	    }
+	    else {
+	       /* Wrapped the primitive: 
+		*/
+	       j += nr - (first - incr);
+	       flush_vertex(split);
+	       csr = 0;
+	       available = align(split->limits->max_verts - csr - 1, 2); 
+	    }
+	 }
+      }
+      else if (split->ib == NULL) {
+	 /* XXX: could at least send the first max_verts off from the
+	  * inplace buffers.
+	  */
+
+	 /* else convert to indexed primitive and pass to split_elts,
+	  * which will do the necessary copying and turn it back into a
+	  * vertex primitive for rendering...
+	  */
+	 struct _mesa_index_buffer ib;
+	 struct _mesa_prim tmpprim;
+	 GLuint *elts = malloc(count * sizeof(GLuint));
+	 GLuint j;
+	 
+	 for (j = 0; j < count; j++)
+	    elts[j] = prim->start + j;
+
+	 ib.count = count;
+	 ib.type = GL_UNSIGNED_INT;
+	 ib.obj = split->ctx->Array.NullBufferObj;
+	 ib.ptr = elts;
+	    
+	 tmpprim = *prim;
+	 tmpprim.indexed = 1;
+	 tmpprim.start = 0;
+	 tmpprim.count = count;
+
+	 flush_vertex(split);
+
+	 vbo_split_copy(split->ctx,
+			split->array,
+			&tmpprim, 1, 
+			&ib,
+			split->draw,
+			split->limits);
+	    
+	 free(elts);
+      }
+      else {
+	 flush_vertex(split);
+
+	 vbo_split_copy(split->ctx,
+			split->array,
+			prim, 1, 
+			split->ib,
+			split->draw,
+			split->limits);
+      }
+   }
+
+   flush_vertex(split);
+}
+
+
+void vbo_split_inplace( GLcontext *ctx,
+			const struct gl_client_array *arrays[],
+			const struct _mesa_prim *prim,
+			GLuint nr_prims,
+			const struct _mesa_index_buffer *ib,
+			GLuint min_index,
+			GLuint max_index,
+			vbo_draw_func draw,
+			const struct split_limits *limits )
+{
+   struct split_context split;
+
+   memset(&split, 0, sizeof(split));
+
+   split.ctx = ctx;
+   split.array = arrays;
+   split.prim = prim;
+   split.nr_prims = nr_prims;
+   split.ib = ib;
+   split.min_index = min_index;
+   split.max_index = max_index;
+   split.draw = draw;
+   split.limits = limits;
+
+   split_prims( &split );
+}
+
+
diff --git a/src/mesa/x86-64/glapi_x86-64.S b/src/mesa/x86-64/glapi_x86-64.S
index d8ef5eae095..eb54ba4848f 100644
--- a/src/mesa/x86-64/glapi_x86-64.S
+++ b/src/mesa/x86-64/glapi_x86-64.S
@@ -29477,6 +29477,7 @@ GL_PREFIX(_dispatch_stub_771):
 	.globl GL_PREFIX(BindProgramARB) ; .set GL_PREFIX(BindProgramARB), GL_PREFIX(BindProgramNV)
 	.globl GL_PREFIX(DeleteProgramsARB) ; .set GL_PREFIX(DeleteProgramsARB), GL_PREFIX(DeleteProgramsNV)
 	.globl GL_PREFIX(GenProgramsARB) ; .set GL_PREFIX(GenProgramsARB), GL_PREFIX(GenProgramsNV)
+	.globl GL_PREFIX(GetVertexAttribPointerv) ; .set GL_PREFIX(GetVertexAttribPointerv), GL_PREFIX(GetVertexAttribPointervNV)
 	.globl GL_PREFIX(GetVertexAttribPointervARB) ; .set GL_PREFIX(GetVertexAttribPointervARB), GL_PREFIX(GetVertexAttribPointervNV)
 	.globl GL_PREFIX(IsProgramARB) ; .set GL_PREFIX(IsProgramARB), GL_PREFIX(IsProgramNV)
 	.globl GL_PREFIX(PointParameteri) ; .set GL_PREFIX(PointParameteri), GL_PREFIX(PointParameteriNV)
diff --git a/src/mesa/x86/glapi_x86.S b/src/mesa/x86/glapi_x86.S
index 2c01b38aeb7..1106eeede87 100644
--- a/src/mesa/x86/glapi_x86.S
+++ b/src/mesa/x86/glapi_x86.S
@@ -1246,6 +1246,7 @@ GLNAME(gl_dispatch_functions_start):
 	GL_STUB_ALIAS(BindProgramARB, _gloffset_BindProgramNV, BindProgramARB@8, BindProgramNV, BindProgramNV@8)
 	GL_STUB_ALIAS(DeleteProgramsARB, _gloffset_DeleteProgramsNV, DeleteProgramsARB@8, DeleteProgramsNV, DeleteProgramsNV@8)
 	GL_STUB_ALIAS(GenProgramsARB, _gloffset_GenProgramsNV, GenProgramsARB@8, GenProgramsNV, GenProgramsNV@8)
+	GL_STUB_ALIAS(GetVertexAttribPointerv, _gloffset_GetVertexAttribPointervNV, GetVertexAttribPointerv@12, GetVertexAttribPointervNV, GetVertexAttribPointervNV@12)
 	GL_STUB_ALIAS(GetVertexAttribPointervARB, _gloffset_GetVertexAttribPointervNV, GetVertexAttribPointervARB@12, GetVertexAttribPointervNV, GetVertexAttribPointervNV@12)
 	GL_STUB_ALIAS(IsProgramARB, _gloffset_IsProgramNV, IsProgramARB@4, IsProgramNV, IsProgramNV@4)
 	GL_STUB_ALIAS(PointParameteri, _gloffset_PointParameteriNV, PointParameteri@8, PointParameteriNV, PointParameteriNV@8)