svn merge ^/branches/released . --username marha

27 files changed, 9465 insertions, 0 deletions

diff --git a/mesalib/src/mesa/tnl/descrip.mms b/mesalib/src/mesa/tnl/descrip.mms
new file mode 100644
index 000000000..25dd1aecb
--- /dev/null
+++ b/mesalib/src/mesa/tnl/descrip.mms
@@ -0,0 +1,68 @@
+# Makefile for core library for VMS
+# contributed by Jouk Jansen  joukj@hrem.nano.tudelft.nl
+# Last revision : 39 September 2008
+
+.first
+	define gl [---.include.gl]
+	define math [-.math]
+	define vbo [-.vbo]
+	define shader [-.shader]
+	define swrast [-.swrast]
+	define array_cache [-.array_cache]
+	define main [-.main]
+	define glapi [-.glapi]
+	define tnl [-.tnl]
+
+.include [---]mms-config.
+
+##### MACROS #####
+
+VPATH = RCS
+
+INCDIR = [---.include],[-.main],[-.glapi],[-.shader],[-.shader.slang]
+LIBDIR = [---.lib]
+CFLAGS = /include=($(INCDIR),[])/define=(PTHREADS=1)/name=(as_is,short)/float=ieee/ieee=denorm
+
+SOURCES = t_context.c t_draw.c \
+	t_pipeline.c t_vb_fog.c \
+	t_vb_light.c t_vb_normals.c t_vb_points.c t_vb_program.c \
+	t_vb_render.c t_vb_texgen.c t_vb_texmat.c t_vb_vertex.c \
+	t_vertex.c t_rasterpos.c\
+	t_vertex_generic.c t_vp_build.c
+
+OBJECTS = t_context.obj,t_draw.obj,\
+	t_pipeline.obj,t_vb_fog.obj,t_vb_light.obj,t_vb_normals.obj,\
+	t_vb_points.obj,t_vb_program.obj,t_vb_render.obj,t_vb_texgen.obj,\
+	t_vb_texmat.obj,t_vb_vertex.obj,t_rasterpos.obj,\
+	t_vertex.obj,t_vertex_generic.obj,\
+	t_vp_build.obj
+
+##### RULES #####
+
+VERSION=Mesa V3.4
+
+##### TARGETS #####
+# Make the library
+$(LIBDIR)$(GL_LIB) : $(OBJECTS)
+  @ library $(LIBDIR)$(GL_LIB) $(OBJECTS)
+
+clean :
+	purge
+	delete *.obj;*
+
+t_context.obj : t_context.c
+t_draw.obj : t_draw.c
+t_pipeline.obj : t_pipeline.c
+t_vb_fog.obj : t_vb_fog.c
+t_vb_light.obj : t_vb_light.c
+t_vb_normals.obj : t_vb_normals.c
+t_vb_points.obj : t_vb_points.c
+t_vb_program.obj : t_vb_program.c
+t_vb_render.obj : t_vb_render.c
+t_vb_texgen.obj : t_vb_texgen.c
+t_vb_texmat.obj : t_vb_texmat.c
+t_vb_vertex.obj : t_vb_vertex.c
+t_vertex.obj : t_vertex.c
+t_vertex_generic.obj : t_vertex_generic.c
+t_vp_build.obj : t_vp_build.c
+t_rasterpos.obj : t_rasterpos.c
diff --git a/mesalib/src/mesa/tnl/t_context.c b/mesalib/src/mesa/tnl/t_context.c
new file mode 100644
index 000000000..f2771cde0
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_context.c
@@ -0,0 +1,231 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.2
+ *
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * 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 "main/glheader.h"
+#include "main/imports.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/mtypes.h"
+#include "main/light.h"
+#include "math/m_translate.h"
+#include "math/m_xform.h"
+
+#include "tnl.h"
+#include "t_context.h"
+#include "t_pipeline.h"
+#include "t_vp_build.h"
+
+#include "vbo/vbo.h"
+
+GLboolean
+_tnl_CreateContext( GLcontext *ctx )
+{
+   TNLcontext *tnl;
+
+   /* Create the TNLcontext structure
+    */
+   ctx->swtnl_context = tnl = (TNLcontext *) CALLOC( sizeof(TNLcontext) );
+
+   if (!tnl) {
+      return GL_FALSE;
+   }
+
+   /* Initialize the VB.
+    */
+   tnl->vb.Size = ctx->Const.MaxArrayLockSize + MAX_CLIPPED_VERTICES;
+
+
+   /* Initialize tnl state.
+    */
+   if (ctx->VertexProgram._MaintainTnlProgram) {
+      _tnl_install_pipeline( ctx, _tnl_vp_pipeline );
+   } else {
+      _tnl_install_pipeline( ctx, _tnl_default_pipeline );
+   }
+
+   tnl->NeedNdcCoords = GL_TRUE;
+   tnl->AllowVertexFog = GL_TRUE;
+   tnl->AllowPixelFog = GL_TRUE;
+
+   /* Set a few default values in the driver struct.
+    */
+   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;
+
+   /* plug in the VBO drawing function */
+   vbo_set_draw_func(ctx, _tnl_vbo_draw_prims);
+
+   _math_init_transformation();
+   _math_init_translate();
+
+   return GL_TRUE;
+}
+
+
+void
+_tnl_DestroyContext( GLcontext *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+
+   _tnl_destroy_pipeline( ctx );
+
+   FREE(tnl);
+   ctx->swtnl_context = NULL;
+}
+
+
+void
+_tnl_InvalidateState( GLcontext *ctx, GLuint new_state )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   const struct gl_vertex_program *vp = ctx->VertexProgram._Current;
+   const struct gl_fragment_program *fp = ctx->FragmentProgram._Current;
+
+   if (new_state & (_NEW_HINT | _NEW_PROGRAM)) {
+      ASSERT(tnl->AllowVertexFog || tnl->AllowPixelFog);
+      tnl->_DoVertexFog = ((tnl->AllowVertexFog && (ctx->Hint.Fog != GL_NICEST))
+         || !tnl->AllowPixelFog) && !fp;
+   }
+
+   tnl->pipeline.new_state |= new_state;
+
+   /* Calculate tnl->render_inputs.  This bitmask indicates which vertex
+    * attributes need to be emitted to the rasterizer.
+    */
+   if (ctx->Visual.rgbMode) {
+      GLuint i;
+
+      RENDERINPUTS_ZERO( tnl->render_inputs_bitset );
+      RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_POS );
+
+      if (!fp || (fp->Base.InputsRead & FRAG_BIT_COL0)) {
+         RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR0 );
+      }
+
+      if (NEED_SECONDARY_COLOR(ctx))
+         RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR1 );
+
+      for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+         if (ctx->Texture._EnabledCoordUnits & (1 << i) ||
+             (fp && fp->Base.InputsRead & FRAG_BIT_TEX(i))) {
+            RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_TEX(i) );
+         }
+      }
+   }
+   else {
+      RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_POS );
+      RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_COLOR_INDEX );
+   }
+
+   if (ctx->Fog.Enabled) {
+      /* fixed-function fog */
+      RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_FOG );
+   }
+   else if (fp) {
+      if (fp->FogOption != GL_NONE || (fp->Base.InputsRead & FRAG_BIT_FOGC)) {
+         /* fragment program needs fog coord */
+         RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_FOG );
+      }
+   }
+
+   if (ctx->Polygon.FrontMode != GL_FILL || 
+       ctx->Polygon.BackMode != GL_FILL)
+      RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_EDGEFLAG );
+
+   if (ctx->RenderMode == GL_FEEDBACK)
+      RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_TEX0 );
+
+   if (ctx->Point._Attenuated ||
+       (ctx->VertexProgram._Enabled && ctx->VertexProgram.PointSizeEnabled))
+      RENDERINPUTS_SET( tnl->render_inputs_bitset, _TNL_ATTRIB_POINTSIZE );
+
+   /* check for varying vars which are written by the vertex program */
+   if (vp) {
+      GLuint i;
+      for (i = 0; i < MAX_VARYING; i++) {
+         if (vp->Base.OutputsWritten & (1 << (VERT_RESULT_VAR0 + i))) {
+            RENDERINPUTS_SET(tnl->render_inputs_bitset,
+                             _TNL_ATTRIB_GENERIC(i));
+         }
+      }
+   }
+}
+
+
+void
+_tnl_wakeup( GLcontext *ctx )
+{
+   /* 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
+}
+
+
+
+
+/**
+ * Drivers call this function to tell the TCL module whether or not
+ * it wants Normalized Device Coords (NDC) computed.  I.e. whether
+ * we should "Divide-by-W".  Software renders will want that.
+ */
+void
+_tnl_need_projected_coords( GLcontext *ctx, GLboolean mode )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   tnl->NeedNdcCoords = mode;
+}
+
+void
+_tnl_allow_vertex_fog( GLcontext *ctx, GLboolean value )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   tnl->AllowVertexFog = value;
+   tnl->_DoVertexFog = ((tnl->AllowVertexFog && (ctx->Hint.Fog != GL_NICEST))
+      || !tnl->AllowPixelFog) && !ctx->FragmentProgram._Current;
+
+}
+
+void
+_tnl_allow_pixel_fog( GLcontext *ctx, GLboolean value )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   tnl->AllowPixelFog = value;
+   tnl->_DoVertexFog = ((tnl->AllowVertexFog && (ctx->Hint.Fog != GL_NICEST))
+      || !tnl->AllowPixelFog) && !ctx->FragmentProgram._Current;
+}
+
diff --git a/mesalib/src/mesa/tnl/t_context.h b/mesalib/src/mesa/tnl/t_context.h
new file mode 100644
index 000000000..c19eb3df3
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_context.h
@@ -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.
+ */
+
+/**
+ * \file t_context.h
+ * \brief TnL module datatypes and definitions.
+ * \author Keith Whitwell
+ */
+
+
+/**
+ * \mainpage The TNL-module
+ *
+ * TNL stands for "transform and lighting", i.e. this module implements
+ * a pipeline that receives as input a buffer of vertices and does all
+ * necessary transformations (rotations, clipping, vertex shader etc.)
+ * and passes then the output to the rasterizer.
+ *
+ * The tnl_pipeline contains the array of all stages, which should be
+ * applied. Each stage is a black-box, which is described by an
+ * tnl_pipeline_stage. The function ::_tnl_run_pipeline applies all the
+ * stages to the vertex_buffer TNLcontext::vb, where the vertex data
+ * is stored. The last stage in the pipeline is the rasterizer.
+ *
+ */
+
+
+#ifndef _T_CONTEXT_H
+#define _T_CONTEXT_H
+
+#include "main/glheader.h"
+#include "main/bitset.h"
+#include "main/mtypes.h"
+
+#include "math/m_matrix.h"
+#include "math/m_vector.h"
+#include "math/m_xform.h"
+
+#include "vbo/vbo.h"
+
+#define MAX_PIPELINE_STAGES     30
+
+/*
+ * Note: The first attributes match the VERT_ATTRIB_* definitions
+ * in mtypes.h.  However, the tnl module has additional attributes
+ * for materials, color indexes, edge flags, etc.
+ */
+/* Although it's nice to use these as bit indexes in a DWORD flag, we
+ * could manage without if necessary.  Another limit currently is the
+ * number of bits allocated for these numbers in places like vertex
+ * program instruction formats and register layouts.
+ */
+/* The bit space exhaustion is a fact now, done by _TNL_ATTRIB_ATTRIBUTE* for
+ * GLSL vertex shader which cannot be aliased with conventional vertex attribs.
+ * Compacting _TNL_ATTRIB_MAT_* attribs would not work, they would not give
+ * as many free bits (11 plus already 1 free bit) as _TNL_ATTRIB_ATTRIBUTE*
+ * attribs want (16).
+ */
+enum {
+	_TNL_ATTRIB_POS = 0,
+	_TNL_ATTRIB_WEIGHT = 1,
+	_TNL_ATTRIB_NORMAL = 2,
+	_TNL_ATTRIB_COLOR0 = 3,
+	_TNL_ATTRIB_COLOR1 = 4,
+	_TNL_ATTRIB_FOG = 5,
+	_TNL_ATTRIB_COLOR_INDEX = 6,
+	_TNL_ATTRIB_EDGEFLAG = 7,
+	_TNL_ATTRIB_TEX0 = 8,
+	_TNL_ATTRIB_TEX1 = 9,
+	_TNL_ATTRIB_TEX2 = 10,
+	_TNL_ATTRIB_TEX3 = 11,
+	_TNL_ATTRIB_TEX4 = 12,
+	_TNL_ATTRIB_TEX5 = 13,
+	_TNL_ATTRIB_TEX6 = 14,
+	_TNL_ATTRIB_TEX7 = 15,
+
+	_TNL_ATTRIB_GENERIC0 = 16, /* doesn't really exist! */
+	_TNL_ATTRIB_GENERIC1 = 17,
+	_TNL_ATTRIB_GENERIC2 = 18,
+	_TNL_ATTRIB_GENERIC3 = 19,
+	_TNL_ATTRIB_GENERIC4 = 20,
+	_TNL_ATTRIB_GENERIC5 = 21,
+	_TNL_ATTRIB_GENERIC6 = 22,
+	_TNL_ATTRIB_GENERIC7 = 23,
+	_TNL_ATTRIB_GENERIC8 = 24,
+	_TNL_ATTRIB_GENERIC9 = 25,
+	_TNL_ATTRIB_GENERIC10 = 26,
+	_TNL_ATTRIB_GENERIC11 = 27,
+	_TNL_ATTRIB_GENERIC12 = 28,
+	_TNL_ATTRIB_GENERIC13 = 29,
+	_TNL_ATTRIB_GENERIC14 = 30,
+	_TNL_ATTRIB_GENERIC15 = 31,
+
+	/* 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))
+#define _TNL_ATTRIB_GENERIC(n) (_TNL_ATTRIB_GENERIC0 + (n))
+
+/* special index used for handing invalid glVertexAttribute() indices */
+#define _TNL_ATTRIB_ERROR    (_TNL_ATTRIB_GENERIC15 + 1)
+
+/**
+ * Handy attribute ranges:
+ */
+#define _TNL_FIRST_PROG      _TNL_ATTRIB_WEIGHT
+#define _TNL_LAST_PROG       _TNL_ATTRIB_TEX7
+
+#define _TNL_FIRST_TEX       _TNL_ATTRIB_TEX0
+#define _TNL_LAST_TEX        _TNL_ATTRIB_TEX7
+
+#define _TNL_FIRST_GENERIC _TNL_ATTRIB_GENERIC0
+#define _TNL_LAST_GENERIC  _TNL_ATTRIB_GENERIC15
+
+#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
+
+/* Number of attributes used for evaluators */
+#define _TNL_NUM_EVAL 16
+
+
+#define PRIM_BEGIN     0x10
+#define PRIM_END       0x20
+#define PRIM_MODE_MASK 0x0f
+
+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;
+}
+
+
+
+
+/**
+ * Contains the current state of a running pipeline.
+ */
+struct vertex_buffer
+{
+   GLuint Size;  /**< Max vertices per vertex buffer, constant */
+
+   /* Constant over the pipeline.
+    */
+   GLuint Count;  /**< Number of vertices currently in buffer */
+
+   /* Pointers to current data.
+    * XXX some of these fields alias AttribPtr below and should be removed
+    * such as NormalPtr, TexCoordPtr, FogCoordPtr, etc.
+    */
+   GLuint      *Elts;		                
+   GLvector4f  *ObjPtr;		                /* _TNL_BIT_POS */
+   GLvector4f  *EyePtr;		                /* _TNL_BIT_POS */
+   GLvector4f  *ClipPtr;	                /* _TNL_BIT_POS */
+   GLvector4f  *NdcPtr;                         /* _TNL_BIT_POS */
+   GLubyte     ClipOrMask;	                /* _TNL_BIT_POS */
+   GLubyte     ClipAndMask;	                /* _TNL_BIT_POS */
+   GLubyte     *ClipMask;		        /* _TNL_BIT_POS */
+   GLvector4f  *NormalPtr;	                /* _TNL_BIT_NORMAL */
+   GLfloat     *NormalLengthPtr;	        /* _TNL_BIT_NORMAL */
+   GLboolean   *EdgeFlag;	                /* _TNL_BIT_EDGEFLAG */
+   GLvector4f  *TexCoordPtr[MAX_TEXTURE_COORD_UNITS]; /* VERT_TEX_0..n */
+   GLvector4f  *IndexPtr[2];	                /* _TNL_BIT_INDEX */
+   GLvector4f  *ColorPtr[2];	                /* _TNL_BIT_COLOR0 */
+   GLvector4f  *SecondaryColorPtr[2];           /* _TNL_BIT_COLOR1 */
+   GLvector4f  *FogCoordPtr;	                /* _TNL_BIT_FOG */
+
+   const struct _mesa_prim  *Primitive;	              
+   GLuint      PrimitiveCount;	      
+
+   /* Inputs to the vertex program stage */
+   GLvector4f *AttribPtr[_TNL_ATTRIB_MAX];      /* GL_NV_vertex_program */
+};
+
+
+/**
+ * Describes an individual operation on the pipeline.
+ */
+struct tnl_pipeline_stage
+{
+   const char *name;
+
+   /* Private data for the pipeline stage:
+    */
+   void *privatePtr;
+
+   /* Allocate private data
+    */
+   GLboolean (*create)( GLcontext *ctx, struct tnl_pipeline_stage * );
+
+   /* Free private data.
+    */
+   void (*destroy)( struct tnl_pipeline_stage * );
+
+   /* Called on any statechange or input array size change or
+    * input array change to/from zero stride.
+    */
+   void (*validate)( GLcontext *ctx, struct tnl_pipeline_stage * );
+
+   /* Called from _tnl_run_pipeline().  The stage.changed_inputs value
+    * encodes all inputs to thee struct which have changed.  If
+    * non-zero, recompute all affected outputs of the stage, otherwise
+    * execute any 'sideeffects' of the stage.
+    *
+    * Return value: GL_TRUE - keep going
+    *               GL_FALSE - finished pipeline
+    */
+   GLboolean (*run)( GLcontext *ctx, struct tnl_pipeline_stage * );
+};
+
+
+
+/** Contains the array of all pipeline stages.
+ * The default values are defined at the end of t_pipeline.c 
+ */
+struct tnl_pipeline {
+   
+   GLuint last_attrib_stride[_TNL_ATTRIB_MAX];
+   GLuint last_attrib_size[_TNL_ATTRIB_MAX];
+   GLuint input_changes;
+   GLuint new_state;
+
+   struct tnl_pipeline_stage stages[MAX_PIPELINE_STAGES+1];
+   GLuint nr_stages;
+};
+
+struct tnl_clipspace;
+struct tnl_clipspace_attr;
+
+typedef void (*tnl_extract_func)( const struct tnl_clipspace_attr *a, 
+				  GLfloat *out, 
+				  const GLubyte *v );
+
+typedef void (*tnl_insert_func)( const struct tnl_clipspace_attr *a, 
+				 GLubyte *v, 
+				 const GLfloat *in );
+
+typedef void (*tnl_emit_func)( GLcontext *ctx, 
+			       GLuint count, 
+			       GLubyte *dest );
+
+
+/**
+ * Describes how to convert/move a vertex attribute from a vertex array
+ * to a vertex structure.
+ */
+struct tnl_clipspace_attr
+{
+   GLuint attrib;          /* which vertex attrib (0=position, etc) */
+   GLuint format;
+   GLuint vertoffset;      /* position of the attrib in the vertex struct */
+   GLuint vertattrsize;    /* size of the attribute in bytes */
+   GLubyte *inputptr;
+   GLuint inputstride;
+   GLuint inputsize;
+   const tnl_insert_func *insert;
+   tnl_insert_func emit;
+   tnl_extract_func extract;
+   const GLfloat *vp;   /* NDC->Viewport mapping matrix */
+};
+
+
+
+
+typedef void (*tnl_points_func)( GLcontext *ctx, GLuint first, GLuint last );
+typedef void (*tnl_line_func)( GLcontext *ctx, GLuint v1, GLuint v2 );
+typedef void (*tnl_triangle_func)( GLcontext *ctx,
+				   GLuint v1, GLuint v2, GLuint v3 );
+typedef void (*tnl_quad_func)( GLcontext *ctx, GLuint v1, GLuint v2,
+			       GLuint v3, GLuint v4 );
+typedef void (*tnl_render_func)( GLcontext *ctx, GLuint start, GLuint count,
+				 GLuint flags );
+typedef void (*tnl_interp_func)( GLcontext *ctx,
+				 GLfloat t, GLuint dst, GLuint out, GLuint in,
+				 GLboolean force_boundary );
+typedef void (*tnl_copy_pv_func)( GLcontext *ctx, GLuint dst, GLuint src );
+typedef void (*tnl_setup_func)( GLcontext *ctx,
+				GLuint start, GLuint end,
+				GLuint new_inputs);
+
+
+struct tnl_attr_type {
+   GLuint format;
+   GLuint size;
+   GLuint stride;
+   GLuint offset;
+};
+
+struct tnl_clipspace_fastpath {
+   GLuint vertex_size;
+   GLuint attr_count;
+   GLboolean match_strides;
+
+   struct tnl_attr_type *attr;
+
+   tnl_emit_func func;
+   struct tnl_clipspace_fastpath *next;
+};
+
+/**
+ * Used to describe conversion of vertex arrays to vertex structures.
+ * I.e. Structure of arrays to arrays of structs.
+ */
+struct tnl_clipspace
+{
+   GLboolean need_extras;
+   
+   GLuint new_inputs;
+
+   GLubyte *vertex_buf;
+   GLuint vertex_size;
+   GLuint max_vertex_size;
+
+   struct tnl_clipspace_attr attr[_TNL_ATTRIB_MAX];
+   GLuint attr_count;
+
+   tnl_emit_func emit;
+   tnl_interp_func interp;
+   tnl_copy_pv_func copy_pv;
+
+   /* Parameters and constants for codegen:
+    */
+   GLboolean need_viewport;
+   GLfloat vp_scale[4];		
+   GLfloat vp_xlate[4];
+   GLfloat chan_scale[4];
+   GLfloat identity[4];
+
+   struct tnl_clipspace_fastpath *fastpath;
+   
+   void (*codegen_emit)( GLcontext *ctx );
+};
+
+
+struct tnl_device_driver
+{
+   /***
+    *** TNL Pipeline
+    ***/
+
+   void (*RunPipeline)(GLcontext *ctx);
+   /* Replaces PipelineStart/PipelineFinish -- intended to allow
+    * drivers to wrap _tnl_run_pipeline() with code to validate state
+    * and grab/release hardware locks.  
+    */
+
+   void (*NotifyMaterialChange)(GLcontext *ctx);
+   /* Alert tnl-aware drivers of changes to material.
+    */
+
+   void (*NotifyInputChanges)(GLcontext *ctx, GLuint bitmask);
+   /* Alert tnl-aware drivers of changes to size and stride of input
+    * arrays.
+    */
+
+   /***
+    *** Rendering -- These functions called only from t_vb_render.c
+    ***/
+   struct
+   {
+      void (*Start)(GLcontext *ctx);
+      void (*Finish)(GLcontext *ctx);
+      /* Called before and after all rendering operations, including DrawPixels,
+       * ReadPixels, Bitmap, span functions, and CopyTexImage, etc commands.
+       * These are a suitable place for grabbing/releasing hardware locks.
+       */
+
+      void (*PrimitiveNotify)(GLcontext *ctx, GLenum mode);
+      /* Called between RenderStart() and RenderFinish() to indicate the
+       * type of primitive we're about to draw.  Mode will be one of the
+       * modes accepted by glBegin().
+       */
+
+      tnl_interp_func Interp;
+      /* The interp function is called by the clipping routines when we need
+       * to generate an interpolated vertex.  All pertinant vertex ancilliary
+       * data should be computed by interpolating between the 'in' and 'out'
+       * vertices.
+       */
+
+      tnl_copy_pv_func CopyPV;
+      /* The copy function is used to make a copy of a vertex.  All pertinant
+       * vertex attributes should be copied.
+       */
+
+      void (*ClippedPolygon)( GLcontext *ctx, const GLuint *elts, GLuint n );
+      /* Render a polygon with <n> vertices whose indexes are in the <elts>
+       * array.
+       */
+
+      void (*ClippedLine)( GLcontext *ctx, GLuint v0, GLuint v1 );
+      /* Render a line between the two vertices given by indexes v0 and v1. */
+
+      tnl_points_func           Points; /* must now respect vb->elts */
+      tnl_line_func             Line;
+      tnl_triangle_func         Triangle;
+      tnl_quad_func             Quad;
+      /* These functions are called in order to render points, lines,
+       * triangles and quads.  These are only called via the T&L module.
+       */
+
+      tnl_render_func          *PrimTabVerts;
+      tnl_render_func          *PrimTabElts;
+      /* Render whole unclipped primitives (points, lines, linestrips,
+       * lineloops, etc).  The tables are indexed by the GL enum of the
+       * primitive to be rendered.  RenderTabVerts is used for non-indexed
+       * arrays of vertices.  RenderTabElts is used for indexed arrays of
+       * vertices.
+       */
+
+      void (*ResetLineStipple)( GLcontext *ctx );
+      /* Reset the hardware's line stipple counter.
+       */
+
+      tnl_setup_func BuildVertices;
+      /* This function is called whenever new vertices are required for
+       * rendering.  The vertices in question are those n such that start
+       * <= n < end.  The new_inputs parameter indicates those fields of
+       * the vertex which need to be updated, if only a partial repair of
+       * the vertex is required.
+       *
+       * This function is called only from _tnl_render_stage in tnl/t_render.c.
+       */
+      
+
+      GLboolean (*Multipass)( GLcontext *ctx, GLuint passno );
+      /* Driver may request additional render passes by returning GL_TRUE
+       * when this function is called.  This function will be called
+       * after the first pass, and passes will be made until the function
+       * returns GL_FALSE.  If no function is registered, only one pass
+       * is made.
+       *
+       * This function will be first invoked with passno == 1.
+       */
+   } Render;
+};
+
+
+#define DECLARE_RENDERINPUTS(name) BITSET64_DECLARE(name, _TNL_ATTRIB_MAX)
+#define RENDERINPUTS_COPY BITSET64_COPY
+#define RENDERINPUTS_EQUAL BITSET64_EQUAL
+#define RENDERINPUTS_ZERO BITSET64_ZERO
+#define RENDERINPUTS_ONES BITSET64_ONES
+#define RENDERINPUTS_TEST BITSET64_TEST
+#define RENDERINPUTS_SET BITSET64_SET
+#define RENDERINPUTS_CLEAR BITSET64_CLEAR
+#define RENDERINPUTS_TEST_RANGE BITSET64_TEST_RANGE
+#define RENDERINPUTS_SET_RANGE BITSET64_SET_RANGE
+#define RENDERINPUTS_CLEAR_RANGE BITSET64_CLEAR_RANGE
+
+
+/**
+ * Context state for T&L context.
+ */
+typedef struct
+{
+   /* Driver interface.
+    */
+   struct tnl_device_driver Driver;
+
+   /* Pipeline
+    */
+   struct tnl_pipeline pipeline;
+   struct vertex_buffer vb;
+
+   /* Clipspace/ndc/window vertex managment:
+    */
+   struct tnl_clipspace clipspace;
+
+   /* Probably need a better configuration mechanism:
+    */
+   GLboolean NeedNdcCoords;
+   GLboolean AllowVertexFog;
+   GLboolean AllowPixelFog;
+   GLboolean _DoVertexFog;  /* eval fog function at each vertex? */
+
+   DECLARE_RENDERINPUTS(render_inputs_bitset);
+
+   GLvector4f tmp_inputs[VERT_ATTRIB_MAX];
+
+   /* Temp storage for t_draw.c: 
+    */
+   GLubyte *block[VERT_ATTRIB_MAX];
+   GLuint nr_blocks;
+
+} TNLcontext;
+
+
+
+#define TNL_CONTEXT(ctx) ((TNLcontext *)((ctx)->swtnl_context))
+
+
+#define TYPE_IDX(t) ((t) & 0xf)
+#define MAX_TYPES TYPE_IDX(GL_DOUBLE)+1      /* 0xa + 1 */
+
+
+#endif
diff --git a/mesalib/src/mesa/tnl/t_draw.c b/mesalib/src/mesa/tnl/t_draw.c
new file mode 100644
index 000000000..c64c2c207
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_draw.c
@@ -0,0 +1,451 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ *
+ * 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"),
+ * 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 "main/glheader.h"
+#include "main/context.h"
+#include "main/imports.h"
+#include "main/state.h"
+#include "main/mtypes.h"
+#include "main/macros.h"
+#include "main/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)
+
+
+/**
+ * Convert array of BGRA/GLubyte[4] values to RGBA/float[4]
+ * \param ptr  input/ubyte array
+ * \param fptr  output/float array
+ */
+static void
+convert_bgra_to_float(const struct gl_client_array *input,
+                      const GLubyte *ptr, GLfloat *fptr,
+                      GLuint count )
+{
+   GLuint i;
+   assert(input->Normalized);
+   assert(input->Size == 4);
+   for (i = 0; i < count; i++) {
+      const GLubyte *in = (GLubyte *) ptr;  /* in is in BGRA order */
+      *fptr++ = UBYTE_TO_FLOAT(in[2]);  /* red */
+      *fptr++ = UBYTE_TO_FLOAT(in[1]);  /* green */
+      *fptr++ = UBYTE_TO_FLOAT(in[0]);  /* blue */
+      *fptr++ = UBYTE_TO_FLOAT(in[3]);  /* alpha */
+      ptr += input->StrideB;
+   }
+}
+
+
+/* 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: 
+         if (input->Format == GL_BGRA) {
+            /* See GL_EXT_vertex_array_bgra */
+            convert_bgra_to_float(input, ptr, fptr, count);
+         }
+         else {
+            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 );
+   }
+   else {
+      /* the data previously pointed to by EdgeFlag may have been freed */
+      VB->EdgeFlag = NULL;
+   }
+}
+
+
+/* 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]);
+   }
+}
+
+
+void _tnl_vbo_draw_prims(GLcontext *ctx,
+			 const struct gl_client_array *arrays[],
+			 const struct _mesa_prim *prim,
+			 GLuint nr_prims,
+			 const struct _mesa_index_buffer *ib,
+			 GLboolean index_bounds_valid,
+			 GLuint min_index,
+			 GLuint max_index)
+{
+   if (!index_bounds_valid)
+      vbo_get_minmax_index(ctx, prim, ib, &min_index, &max_index);
+
+   _tnl_draw_prims(ctx, arrays, prim, nr_prims, ib, min_index, max_index);
+}
+
+/* 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_vbo_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_vbo_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/mesalib/src/mesa/tnl/t_pipeline.c b/mesalib/src/mesa/tnl/t_pipeline.c
new file mode 100644
index 000000000..357ef1e24
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_pipeline.c
@@ -0,0 +1,215 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5.3
+ *
+ * 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"),
+ * 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 "main/glheader.h"
+#include "main/context.h"
+#include "main/imports.h"
+#include "main/state.h"
+#include "main/mtypes.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+#include "t_vp_build.h"
+#include "t_vertex.h"
+
+void _tnl_install_pipeline( GLcontext *ctx,
+			    const struct tnl_pipeline_stage **stages )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLuint i;
+
+   tnl->pipeline.new_state = ~0;
+
+   /* Create a writeable copy of each stage.
+    */
+   for (i = 0 ; i < MAX_PIPELINE_STAGES && stages[i] ; i++) {
+      struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];
+      MEMCPY(s, stages[i], sizeof(*s));
+      if (s->create)
+	 s->create(ctx, s);
+   }
+
+   tnl->pipeline.nr_stages = i;
+}
+
+void _tnl_destroy_pipeline( GLcontext *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLuint i;
+
+   for (i = 0 ; i < tnl->pipeline.nr_stages ; i++) {
+      struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];
+      if (s->destroy)
+	 s->destroy(s);
+   }
+
+   tnl->pipeline.nr_stages = 0;
+}
+
+
+
+static GLuint check_input_changes( GLcontext *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   GLuint i;
+   
+   for (i = 0; i <= _TNL_LAST_MAT; i++) {
+      if (tnl->vb.AttribPtr[i]->size != tnl->pipeline.last_attrib_size[i] ||
+	  tnl->vb.AttribPtr[i]->stride != tnl->pipeline.last_attrib_stride[i]) {
+	 tnl->pipeline.last_attrib_size[i] = tnl->vb.AttribPtr[i]->size;
+	 tnl->pipeline.last_attrib_stride[i] = tnl->vb.AttribPtr[i]->stride;
+	 tnl->pipeline.input_changes |= 1<<i;
+      }
+   }
+
+   if (tnl->pipeline.input_changes &&
+      tnl->Driver.NotifyInputChanges) 
+      tnl->Driver.NotifyInputChanges( ctx, tnl->pipeline.input_changes );
+
+   return tnl->pipeline.input_changes;
+}
+
+
+static GLuint check_output_changes( GLcontext *ctx )
+{
+#if 0
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   
+   for (i = 0; i < VERT_RESULT_MAX; i++) {
+      if (tnl->vb.ResultPtr[i]->size != tnl->last_result_size[i] ||
+	  tnl->vb.ResultPtr[i]->stride != tnl->last_result_stride[i]) {
+	 tnl->last_result_size[i] = tnl->vb.ResultPtr[i]->size;
+	 tnl->last_result_stride[i] = tnl->vb.ResultPtr[i]->stride;
+	 tnl->pipeline.output_changes |= 1<<i;
+      }
+   }
+
+   if (tnl->pipeline.output_changes) 
+      tnl->Driver.NotifyOutputChanges( ctx, tnl->pipeline.output_changes );
+   
+   return tnl->pipeline.output_changes;
+#else
+   return ~0;
+#endif
+}
+
+
+void _tnl_run_pipeline( GLcontext *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   unsigned short __tmp;
+   GLuint i;
+
+   if (!tnl->vb.Count)
+      return;
+
+   /* Check for changed input sizes or change in stride to/from zero
+    * (ie const or non-const).
+    */
+   if (check_input_changes( ctx ) || tnl->pipeline.new_state) {
+      if (ctx->VertexProgram._MaintainTnlProgram)
+	 _tnl_UpdateFixedFunctionProgram( ctx );
+
+      for (i = 0; i < tnl->pipeline.nr_stages ; i++) {
+	 struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];
+	 if (s->validate)
+	    s->validate( ctx, s );
+      }
+      
+      tnl->pipeline.new_state = 0;
+      tnl->pipeline.input_changes = 0;
+      
+      /* Pipeline can only change its output in response to either a
+       * statechange or an input size/stride change.  No other changes
+       * are allowed.
+       */
+      if (check_output_changes( ctx ))
+	 _tnl_notify_pipeline_output_change( ctx );
+   }
+
+   START_FAST_MATH(__tmp);
+
+   for (i = 0; i < tnl->pipeline.nr_stages ; i++) {
+      struct tnl_pipeline_stage *s = &tnl->pipeline.stages[i];
+      if (!s->run( ctx, s ))
+	 break;
+   }
+
+   END_FAST_MATH(__tmp);
+}
+
+
+
+/* The default pipeline.  This is useful for software rasterizers, and
+ * simple hardware rasterizers.  For customization, I don't recommend
+ * tampering with the internals of these stages in the way that
+ * drivers did in Mesa 3.4.  These stages are basically black boxes,
+ * and should be left intact.
+ *
+ * To customize the pipeline, consider:
+ *
+ * - removing redundant stages (making sure that the software rasterizer
+ *   can cope with this on fallback paths).  An example is fog
+ *   coordinate generation, which is not required in the FX driver.
+ *
+ * - replacing general-purpose machine-independent stages with
+ *   general-purpose machine-specific stages.  There is no example of
+ *   this to date, though it must be borne in mind that all subsequent
+ *   stages that reference the output of the new stage must cope with
+ *   any machine-specific data introduced.  This may not be easy
+ *   unless there are no such stages (ie the new stage is the last in
+ *   the pipe).
+ *
+ * - inserting optimized (but specialized) stages ahead of the
+ *   general-purpose fallback implementation.  For example, the old
+ *   fastpath mechanism, which only works when the VB->Elts input is
+ *   available, can be duplicated by placing the fastpath stage at the
+ *   head of this pipeline.  Such specialized stages are currently
+ *   constrained to have no outputs (ie. they must either finish the *
+ *   pipeline by returning GL_FALSE from run(), or do nothing).
+ *
+ * Some work can be done to lift some of the restrictions in the final
+ * case, if it becomes necessary to do so.
+ */
+const struct tnl_pipeline_stage *_tnl_default_pipeline[] = {
+   &_tnl_vertex_transform_stage,
+   &_tnl_normal_transform_stage,
+   &_tnl_lighting_stage,
+   &_tnl_texgen_stage,
+   &_tnl_texture_transform_stage,
+   &_tnl_point_attenuation_stage,
+   &_tnl_vertex_program_stage, 
+   &_tnl_fog_coordinate_stage,
+   &_tnl_render_stage,
+   NULL 
+};
+
+const struct tnl_pipeline_stage *_tnl_vp_pipeline[] = {
+   &_tnl_vertex_program_stage,
+   &_tnl_render_stage,
+   NULL
+};
diff --git a/mesalib/src/mesa/tnl/t_pipeline.h b/mesalib/src/mesa/tnl/t_pipeline.h
new file mode 100644
index 000000000..d110010f0
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_pipeline.h
@@ -0,0 +1,74 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5.3
+ *
+ * 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"),
+ * 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 _T_PIPELINE_H_
+#define _T_PIPELINE_H_
+
+#include "main/mtypes.h"
+#include "t_context.h"
+
+extern void _tnl_run_pipeline( GLcontext *ctx );
+
+extern void _tnl_destroy_pipeline( GLcontext *ctx );
+
+extern void _tnl_install_pipeline( GLcontext *ctx,
+				   const struct tnl_pipeline_stage **stages );
+
+
+/* These are implemented in the t_vb_*.c files:
+ */
+extern const struct tnl_pipeline_stage _tnl_vertex_transform_stage;
+extern const struct tnl_pipeline_stage _tnl_vertex_cull_stage;
+extern const struct tnl_pipeline_stage _tnl_normal_transform_stage;
+extern const struct tnl_pipeline_stage _tnl_lighting_stage;
+extern const struct tnl_pipeline_stage _tnl_fog_coordinate_stage;
+extern const struct tnl_pipeline_stage _tnl_texgen_stage;
+extern const struct tnl_pipeline_stage _tnl_texture_transform_stage;
+extern const struct tnl_pipeline_stage _tnl_point_attenuation_stage;
+extern const struct tnl_pipeline_stage _tnl_vertex_program_stage;
+extern const struct tnl_pipeline_stage _tnl_render_stage;
+
+/* Shorthand to plug in the default pipeline:
+ */
+extern const struct tnl_pipeline_stage *_tnl_default_pipeline[];
+extern const struct tnl_pipeline_stage *_tnl_vp_pipeline[];
+
+
+/* Convenience routines provided by t_vb_render.c:
+ */
+extern tnl_render_func _tnl_render_tab_elts[];
+extern tnl_render_func _tnl_render_tab_verts[];
+
+extern void _tnl_RenderClippedPolygon( GLcontext *ctx, 
+				       const GLuint *elts, GLuint n );
+
+extern void _tnl_RenderClippedLine( GLcontext *ctx, GLuint ii, GLuint jj );
+
+
+#endif
diff --git a/mesalib/src/mesa/tnl/t_rasterpos.c b/mesalib/src/mesa/tnl/t_rasterpos.c
new file mode 100644
index 000000000..f1fdddf0f
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_rasterpos.c
@@ -0,0 +1,505 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ *
+ * 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"),
+ * 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.
+ */
+
+
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/feedback.h"
+#include "main/light.h"
+#include "main/macros.h"
+#include "main/rastpos.h"
+#include "main/simple_list.h"
+#include "main/mtypes.h"
+
+#include "math/m_matrix.h"
+#include "tnl/tnl.h"
+
+
+
+/**
+ * Clip a point against the view volume.
+ *
+ * \param v vertex vector describing the point to clip.
+ * 
+ * \return zero if outside view volume, or one if inside.
+ */
+static GLuint
+viewclip_point( const GLfloat v[] )
+{
+   if (   v[0] > v[3] || v[0] < -v[3]
+       || v[1] > v[3] || v[1] < -v[3]
+       || v[2] > v[3] || v[2] < -v[3] ) {
+      return 0;
+   }
+   else {
+      return 1;
+   }
+}
+
+
+/**
+ * Clip a point against the far/near Z clipping planes.
+ *
+ * \param v vertex vector describing the point to clip.
+ * 
+ * \return zero if outside view volume, or one if inside.
+ */
+static GLuint
+viewclip_point_z( const GLfloat v[] )
+{
+   if (v[2] > v[3] || v[2] < -v[3] ) {
+      return 0;
+   }
+   else {
+      return 1;
+   }
+}
+
+
+/**
+ * Clip a point against the user clipping planes.
+ * 
+ * \param ctx GL context.
+ * \param v vertex vector describing the point to clip.
+ * 
+ * \return zero if the point was clipped, or one otherwise.
+ */
+static GLuint
+userclip_point( GLcontext *ctx, const GLfloat v[] )
+{
+   GLuint p;
+
+   for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
+      if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
+	 GLfloat dot = v[0] * ctx->Transform._ClipUserPlane[p][0]
+		     + v[1] * ctx->Transform._ClipUserPlane[p][1]
+		     + v[2] * ctx->Transform._ClipUserPlane[p][2]
+		     + v[3] * ctx->Transform._ClipUserPlane[p][3];
+         if (dot < 0.0F) {
+            return 0;
+         }
+      }
+   }
+
+   return 1;
+}
+
+
+/**
+ * Compute lighting for the raster position.  Both RGB and CI modes computed.
+ * \param ctx the context
+ * \param vertex vertex location
+ * \param normal normal vector
+ * \param Rcolor returned color
+ * \param Rspec returned specular color (if separate specular enabled)
+ * \param Rindex returned color index
+ */
+static void
+shade_rastpos(GLcontext *ctx,
+              const GLfloat vertex[4],
+              const GLfloat normal[3],
+              GLfloat Rcolor[4],
+              GLfloat Rspec[4],
+              GLfloat *Rindex)
+{
+   /*const*/ GLfloat (*base)[3] = ctx->Light._BaseColor;
+   const struct gl_light *light;
+   GLfloat diffuseColor[4], specularColor[4];  /* for RGB mode only */
+   GLfloat diffuseCI = 0.0, specularCI = 0.0;  /* for CI mode only */
+
+   _mesa_validate_all_lighting_tables( ctx );
+
+   COPY_3V(diffuseColor, base[0]);
+   diffuseColor[3] = CLAMP( 
+      ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3], 0.0F, 1.0F );
+   ASSIGN_4V(specularColor, 0.0, 0.0, 0.0, 1.0);
+
+   foreach (light, &ctx->Light.EnabledList) {
+      GLfloat attenuation = 1.0;
+      GLfloat VP[3]; /* vector from vertex to light pos */
+      GLfloat n_dot_VP;
+      GLfloat diffuseContrib[3], specularContrib[3];
+
+      if (!(light->_Flags & LIGHT_POSITIONAL)) {
+         /* light at infinity */
+	 COPY_3V(VP, light->_VP_inf_norm);
+	 attenuation = light->_VP_inf_spot_attenuation;
+      }
+      else {
+         /* local/positional light */
+	 GLfloat d;
+
+         /* VP = vector from vertex pos to light[i].pos */
+	 SUB_3V(VP, light->_Position, vertex);
+         /* d = length(VP) */
+	 d = (GLfloat) LEN_3FV( VP );
+	 if (d > 1.0e-6) {
+            /* normalize VP */
+	    GLfloat invd = 1.0F / d;
+	    SELF_SCALE_SCALAR_3V(VP, invd);
+	 }
+
+         /* atti */
+	 attenuation = 1.0F / (light->ConstantAttenuation + d *
+			       (light->LinearAttenuation + d *
+				light->QuadraticAttenuation));
+
+	 if (light->_Flags & LIGHT_SPOT) {
+	    GLfloat PV_dot_dir = - DOT3(VP, light->_NormSpotDirection);
+
+	    if (PV_dot_dir<light->_CosCutoff) {
+	       continue;
+	    }
+	    else {
+	       double x = PV_dot_dir * (EXP_TABLE_SIZE-1);
+	       int k = (int) x;
+	       GLfloat spot = (GLfloat) (light->_SpotExpTable[k][0]
+			       + (x-k)*light->_SpotExpTable[k][1]);
+	       attenuation *= spot;
+	    }
+	 }
+      }
+
+      if (attenuation < 1e-3)
+	 continue;
+
+      n_dot_VP = DOT3( normal, VP );
+
+      if (n_dot_VP < 0.0F) {
+	 ACC_SCALE_SCALAR_3V(diffuseColor, attenuation, light->_MatAmbient[0]);
+	 continue;
+      }
+
+      /* Ambient + diffuse */
+      COPY_3V(diffuseContrib, light->_MatAmbient[0]);
+      ACC_SCALE_SCALAR_3V(diffuseContrib, n_dot_VP, light->_MatDiffuse[0]);
+      diffuseCI += n_dot_VP * light->_dli * attenuation;
+
+      /* Specular */
+      {
+         const GLfloat *h;
+         GLfloat n_dot_h;
+
+         ASSIGN_3V(specularContrib, 0.0, 0.0, 0.0);
+
+	 if (ctx->Light.Model.LocalViewer) {
+	    GLfloat v[3];
+	    COPY_3V(v, vertex);
+	    NORMALIZE_3FV(v);
+	    SUB_3V(VP, VP, v);
+            NORMALIZE_3FV(VP);
+	    h = VP;
+	 }
+	 else if (light->_Flags & LIGHT_POSITIONAL) {
+	    ACC_3V(VP, ctx->_EyeZDir);
+            NORMALIZE_3FV(VP);
+	    h = VP;
+	 }
+         else {
+	    h = light->_h_inf_norm;
+	 }
+
+	 n_dot_h = DOT3(normal, h);
+
+	 if (n_dot_h > 0.0F) {
+	    GLfloat spec_coef;
+	    GET_SHINE_TAB_ENTRY( ctx->_ShineTable[0], n_dot_h, spec_coef );
+
+	    if (spec_coef > 1.0e-10) {
+               if (ctx->Light.Model.ColorControl==GL_SEPARATE_SPECULAR_COLOR) {
+                  ACC_SCALE_SCALAR_3V( specularContrib, spec_coef,
+                                       light->_MatSpecular[0]);
+               }
+               else {
+                  ACC_SCALE_SCALAR_3V( diffuseContrib, spec_coef,
+                                       light->_MatSpecular[0]);
+               }
+               /*assert(light->_sli > 0.0);*/
+               specularCI += spec_coef * light->_sli * attenuation;
+	    }
+	 }
+      }
+
+      ACC_SCALE_SCALAR_3V( diffuseColor, attenuation, diffuseContrib );
+      ACC_SCALE_SCALAR_3V( specularColor, attenuation, specularContrib );
+   }
+
+   if (ctx->Visual.rgbMode) {
+      Rcolor[0] = CLAMP(diffuseColor[0], 0.0F, 1.0F);
+      Rcolor[1] = CLAMP(diffuseColor[1], 0.0F, 1.0F);
+      Rcolor[2] = CLAMP(diffuseColor[2], 0.0F, 1.0F);
+      Rcolor[3] = CLAMP(diffuseColor[3], 0.0F, 1.0F);
+      Rspec[0] = CLAMP(specularColor[0], 0.0F, 1.0F);
+      Rspec[1] = CLAMP(specularColor[1], 0.0F, 1.0F);
+      Rspec[2] = CLAMP(specularColor[2], 0.0F, 1.0F);
+      Rspec[3] = CLAMP(specularColor[3], 0.0F, 1.0F);
+   }
+   else {
+      GLfloat *ind = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_INDEXES];
+      GLfloat d_a = ind[MAT_INDEX_DIFFUSE] - ind[MAT_INDEX_AMBIENT];
+      GLfloat s_a = ind[MAT_INDEX_SPECULAR] - ind[MAT_INDEX_AMBIENT];
+      GLfloat i = (ind[MAT_INDEX_AMBIENT]
+		   + diffuseCI * (1.0F-specularCI) * d_a
+		   + specularCI * s_a);
+      if (i > ind[MAT_INDEX_SPECULAR]) {
+	 i = ind[MAT_INDEX_SPECULAR];
+      }
+      *Rindex = i;
+   }
+}
+
+
+/**
+ * Do texgen needed for glRasterPos.
+ * \param ctx  rendering context
+ * \param vObj  object-space vertex coordinate
+ * \param vEye  eye-space vertex coordinate
+ * \param normal  vertex normal
+ * \param unit  texture unit number
+ * \param texcoord  incoming texcoord and resulting texcoord
+ */
+static void
+compute_texgen(GLcontext *ctx, const GLfloat vObj[4], const GLfloat vEye[4],
+               const GLfloat normal[3], GLuint unit, GLfloat texcoord[4])
+{
+   const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+
+   /* always compute sphere map terms, just in case */
+   GLfloat u[3], two_nu, rx, ry, rz, m, mInv;
+   COPY_3V(u, vEye);
+   NORMALIZE_3FV(u);
+   two_nu = 2.0F * DOT3(normal, u);
+   rx = u[0] - normal[0] * two_nu;
+   ry = u[1] - normal[1] * two_nu;
+   rz = u[2] - normal[2] * two_nu;
+   m = rx * rx + ry * ry + (rz + 1.0F) * (rz + 1.0F);
+   if (m > 0.0F)
+      mInv = 0.5F * _mesa_inv_sqrtf(m);
+   else
+      mInv = 0.0F;
+
+   if (texUnit->TexGenEnabled & S_BIT) {
+      switch (texUnit->GenS.Mode) {
+         case GL_OBJECT_LINEAR:
+            texcoord[0] = DOT4(vObj, texUnit->GenS.ObjectPlane);
+            break;
+         case GL_EYE_LINEAR:
+            texcoord[0] = DOT4(vEye, texUnit->GenS.EyePlane);
+            break;
+         case GL_SPHERE_MAP:
+            texcoord[0] = rx * mInv + 0.5F;
+            break;
+         case GL_REFLECTION_MAP:
+            texcoord[0] = rx;
+            break;
+         case GL_NORMAL_MAP:
+            texcoord[0] = normal[0];
+            break;
+         default:
+            _mesa_problem(ctx, "Bad S texgen in compute_texgen()");
+            return;
+      }
+   }
+
+   if (texUnit->TexGenEnabled & T_BIT) {
+      switch (texUnit->GenT.Mode) {
+         case GL_OBJECT_LINEAR:
+            texcoord[1] = DOT4(vObj, texUnit->GenT.ObjectPlane);
+            break;
+         case GL_EYE_LINEAR:
+            texcoord[1] = DOT4(vEye, texUnit->GenT.EyePlane);
+            break;
+         case GL_SPHERE_MAP:
+            texcoord[1] = ry * mInv + 0.5F;
+            break;
+         case GL_REFLECTION_MAP:
+            texcoord[1] = ry;
+            break;
+         case GL_NORMAL_MAP:
+            texcoord[1] = normal[1];
+            break;
+         default:
+            _mesa_problem(ctx, "Bad T texgen in compute_texgen()");
+            return;
+      }
+   }
+
+   if (texUnit->TexGenEnabled & R_BIT) {
+      switch (texUnit->GenR.Mode) {
+         case GL_OBJECT_LINEAR:
+            texcoord[2] = DOT4(vObj, texUnit->GenR.ObjectPlane);
+            break;
+         case GL_EYE_LINEAR:
+            texcoord[2] = DOT4(vEye, texUnit->GenR.EyePlane);
+            break;
+         case GL_REFLECTION_MAP:
+            texcoord[2] = rz;
+            break;
+         case GL_NORMAL_MAP:
+            texcoord[2] = normal[2];
+            break;
+         default:
+            _mesa_problem(ctx, "Bad R texgen in compute_texgen()");
+            return;
+      }
+   }
+
+   if (texUnit->TexGenEnabled & Q_BIT) {
+      switch (texUnit->GenQ.Mode) {
+         case GL_OBJECT_LINEAR:
+            texcoord[3] = DOT4(vObj, texUnit->GenQ.ObjectPlane);
+            break;
+         case GL_EYE_LINEAR:
+            texcoord[3] = DOT4(vEye, texUnit->GenQ.EyePlane);
+            break;
+         default:
+            _mesa_problem(ctx, "Bad Q texgen in compute_texgen()");
+            return;
+      }
+   }
+}
+
+
+/**
+ * glRasterPos transformation.  Typically called via ctx->Driver.RasterPos().
+ * XXX some of this code (such as viewport xform, clip testing and setting
+ * of ctx->Current.Raster* fields) could get lifted up into the
+ * main/rasterpos.c code.
+ *
+ * \param vObj  vertex position in object space
+ */
+void
+_tnl_RasterPos(GLcontext *ctx, const GLfloat vObj[4])
+{
+   if (ctx->VertexProgram._Enabled) {
+      /* XXX implement this */
+      _mesa_problem(ctx, "Vertex programs not implemented for glRasterPos");
+      return;
+   }
+   else {
+      GLfloat eye[4], clip[4], ndc[3], d;
+      GLfloat *norm, eyenorm[3];
+      GLfloat *objnorm = ctx->Current.Attrib[VERT_ATTRIB_NORMAL];
+
+      /* apply modelview matrix:  eye = MV * obj */
+      TRANSFORM_POINT( eye, ctx->ModelviewMatrixStack.Top->m, vObj );
+      /* apply projection matrix:  clip = Proj * eye */
+      TRANSFORM_POINT( clip, ctx->ProjectionMatrixStack.Top->m, eye );
+
+      /* clip to view volume */
+      if (ctx->Transform.RasterPositionUnclipped) {
+         /* GL_IBM_rasterpos_clip: only clip against Z */
+         if (viewclip_point_z(clip) == 0) {
+            ctx->Current.RasterPosValid = GL_FALSE;
+            return;
+         }
+      }
+      else if (viewclip_point(clip) == 0) {
+         /* Normal OpenGL behaviour */
+         ctx->Current.RasterPosValid = GL_FALSE;
+         return;
+      }
+
+      /* clip to user clipping planes */
+      if (ctx->Transform.ClipPlanesEnabled && !userclip_point(ctx, clip)) {
+         ctx->Current.RasterPosValid = GL_FALSE;
+         return;
+      }
+
+      /* ndc = clip / W */
+      d = (clip[3] == 0.0F) ? 1.0F : 1.0F / clip[3];
+      ndc[0] = clip[0] * d;
+      ndc[1] = clip[1] * d;
+      ndc[2] = clip[2] * d;
+      /* wincoord = viewport_mapping(ndc) */
+      ctx->Current.RasterPos[0] = (ndc[0] * ctx->Viewport._WindowMap.m[MAT_SX]
+                                   + ctx->Viewport._WindowMap.m[MAT_TX]);
+      ctx->Current.RasterPos[1] = (ndc[1] * ctx->Viewport._WindowMap.m[MAT_SY]
+                                   + ctx->Viewport._WindowMap.m[MAT_TY]);
+      ctx->Current.RasterPos[2] = (ndc[2] * ctx->Viewport._WindowMap.m[MAT_SZ]
+                                   + ctx->Viewport._WindowMap.m[MAT_TZ])
+                                  / ctx->DrawBuffer->_DepthMaxF;
+      ctx->Current.RasterPos[3] = clip[3];
+
+      /* compute raster distance */
+      if (ctx->Fog.FogCoordinateSource == GL_FOG_COORDINATE_EXT)
+         ctx->Current.RasterDistance = ctx->Current.Attrib[VERT_ATTRIB_FOG][0];
+      else
+         ctx->Current.RasterDistance =
+                        SQRTF( eye[0]*eye[0] + eye[1]*eye[1] + eye[2]*eye[2] );
+
+      /* compute transformed normal vector (for lighting or texgen) */
+      if (ctx->_NeedEyeCoords) {
+         const GLfloat *inv = ctx->ModelviewMatrixStack.Top->inv;
+         TRANSFORM_NORMAL( eyenorm, objnorm, inv );
+         norm = eyenorm;
+      }
+      else {
+         norm = objnorm;
+      }
+
+      /* update raster color */
+      if (ctx->Light.Enabled) {
+         /* lighting */
+         shade_rastpos( ctx, vObj, norm,
+                        ctx->Current.RasterColor,
+                        ctx->Current.RasterSecondaryColor,
+                        &ctx->Current.RasterIndex );
+      }
+      else {
+         /* use current color or index */
+         if (ctx->Visual.rgbMode) {
+            COPY_4FV(ctx->Current.RasterColor,
+                     ctx->Current.Attrib[VERT_ATTRIB_COLOR0]);
+            COPY_4FV(ctx->Current.RasterSecondaryColor,
+                     ctx->Current.Attrib[VERT_ATTRIB_COLOR1]);
+         }
+         else {
+            ctx->Current.RasterIndex
+               = ctx->Current.Attrib[VERT_ATTRIB_COLOR_INDEX][0];
+         }
+      }
+
+      /* texture coords */
+      {
+         GLuint u;
+         for (u = 0; u < ctx->Const.MaxTextureCoordUnits; u++) {
+            GLfloat tc[4];
+            COPY_4V(tc, ctx->Current.Attrib[VERT_ATTRIB_TEX0 + u]);
+            if (ctx->Texture.Unit[u].TexGenEnabled) {
+               compute_texgen(ctx, vObj, eye, norm, u, tc);
+            }
+            TRANSFORM_POINT(ctx->Current.RasterTexCoords[u],
+                            ctx->TextureMatrixStack[u].Top->m, tc);
+         }
+      }
+
+      ctx->Current.RasterPosValid = GL_TRUE;
+   }
+
+   if (ctx->RenderMode == GL_SELECT) {
+      _mesa_update_hitflag( ctx, ctx->Current.RasterPos[2] );
+   }
+}
diff --git a/mesalib/src/mesa/tnl/t_vb_cliptmp.h b/mesalib/src/mesa/tnl/t_vb_cliptmp.h
new file mode 100644
index 000000000..618b8b313
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_cliptmp.h
@@ -0,0 +1,320 @@
+/*
+ * 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.
+ *
+ * Authors:
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#define CLIP_DOTPROD(K, A, B, C, D) X(K)*A + Y(K)*B + Z(K)*C + W(K)*D
+
+#define POLY_CLIP( PLANE_BIT, A, B, C, D )				\
+do {									\
+   if (mask & PLANE_BIT) {						\
+      GLuint idxPrev = inlist[0];					\
+      GLfloat dpPrev = CLIP_DOTPROD(idxPrev, A, B, C, D );		\
+      GLuint outcount = 0;						\
+      GLuint i;								\
+									\
+      inlist[n] = inlist[0]; /* prevent rotation of vertices */		\
+      for (i = 1; i <= n; i++) {					\
+	 GLuint idx = inlist[i];					\
+	 GLfloat dp = CLIP_DOTPROD(idx, A, B, C, D );			\
+									\
+	 if (!IS_NEGATIVE(dpPrev)) {					\
+	    outlist[outcount++] = idxPrev;				\
+	 }								\
+									\
+	 if (DIFFERENT_SIGNS(dp, dpPrev)) {				\
+	    if (IS_NEGATIVE(dp)) {					\
+	       /* Going out of bounds.  Avoid division by zero as we	\
+		* know dp != dpPrev from DIFFERENT_SIGNS, above.	\
+		*/							\
+	       GLfloat t = dp / (dp - dpPrev);				\
+               INTERP_4F( t, coord[newvert], coord[idx], coord[idxPrev]); \
+      	       interp( ctx, t, newvert, idx, idxPrev, GL_TRUE );	\
+	    } else {							\
+	       /* Coming back in.					\
+		*/							\
+	       GLfloat t = dpPrev / (dpPrev - dp);			\
+               INTERP_4F( t, coord[newvert], coord[idxPrev], coord[idx]); \
+	       interp( ctx, t, newvert, idxPrev, idx, GL_FALSE );	\
+	    }								\
+            outlist[outcount++] = newvert++;				\
+	 }								\
+									\
+	 idxPrev = idx;							\
+	 dpPrev = dp;							\
+      }									\
+									\
+      if (outcount < 3)							\
+	 return;							\
+									\
+      {									\
+	 GLuint *tmp = inlist;						\
+	 inlist = outlist;						\
+	 outlist = tmp;							\
+	 n = outcount;							\
+      }									\
+   }									\
+} while (0)
+
+
+#define LINE_CLIP(PLANE_BIT, A, B, C, D )				\
+do {									\
+   if (mask & PLANE_BIT) {						\
+      const GLfloat dp0 = CLIP_DOTPROD( v0, A, B, C, D );		\
+      const GLfloat dp1 = CLIP_DOTPROD( v1, A, B, C, D );		\
+      const GLboolean neg_dp0 = IS_NEGATIVE(dp0);			\
+      const GLboolean neg_dp1 = IS_NEGATIVE(dp1);			\
+      									\
+      /* For regular clipping, we know from the clipmask that one	\
+       * (or both) of these must be negative (otherwise we wouldn't	\
+       * be here).							\
+       * For userclip, there is only a single bit for all active	\
+       * planes, so we can end up here when there is nothing to do,	\
+       * hence the second IS_NEGATIVE() test:				\
+       */								\
+      if (neg_dp0 && neg_dp1)						\
+         return; /* both vertices outside clip plane: discard */	\
+									\
+      if (neg_dp1) {							\
+	 GLfloat t = dp1 / (dp1 - dp0);					\
+	 if (t > t1) t1 = t;						\
+      } else if (neg_dp0) {						\
+	 GLfloat t = dp0 / (dp0 - dp1);					\
+	 if (t > t0) t0 = t;						\
+      }									\
+      if (t0 + t1 >= 1.0)						\
+	 return; /* discard */						\
+   }									\
+} while (0)
+
+
+
+/* Clip a line against the viewport and user clip planes.
+ */
+static INLINE void
+TAG(clip_line)( GLcontext *ctx, GLuint v0, GLuint v1, GLubyte mask )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   tnl_interp_func interp = tnl->Driver.Render.Interp;
+   GLfloat (*coord)[4] = VB->ClipPtr->data;
+   GLuint newvert = VB->Count;
+   GLfloat t0 = 0;
+   GLfloat t1 = 0;
+   GLuint p;
+   const GLuint v0_orig = v0;
+
+   if (mask & CLIP_FRUSTUM_BITS) {
+      LINE_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
+      LINE_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
+      LINE_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
+      LINE_CLIP( CLIP_BOTTOM_BIT,  0,  1,  0, 1 );
+      LINE_CLIP( CLIP_FAR_BIT,     0,  0, -1, 1 );
+      LINE_CLIP( CLIP_NEAR_BIT,    0,  0,  1, 1 );
+   }
+
+   if (mask & CLIP_USER_BIT) {
+      for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
+	 if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
+            const GLfloat a = ctx->Transform._ClipUserPlane[p][0];
+            const GLfloat b = ctx->Transform._ClipUserPlane[p][1];
+            const GLfloat c = ctx->Transform._ClipUserPlane[p][2];
+            const GLfloat d = ctx->Transform._ClipUserPlane[p][3];
+	    LINE_CLIP( CLIP_USER_BIT, a, b, c, d );
+	 }
+      }
+   }
+
+   if (VB->ClipMask[v0]) {
+      INTERP_4F( t0, coord[newvert], coord[v0], coord[v1] );
+      interp( ctx, t0, newvert, v0, v1, GL_FALSE );
+      v0 = newvert;
+      newvert++;
+   }
+   else {
+      ASSERT(t0 == 0.0);
+   }
+
+   /* Note: we need to use vertex v0_orig when computing the new
+    * interpolated/clipped vertex position, not the current v0 which
+    * may have got set when we clipped the other end of the line!
+    */
+   if (VB->ClipMask[v1]) {
+      INTERP_4F( t1, coord[newvert], coord[v1], coord[v0_orig] );
+      interp( ctx, t1, newvert, v1, v0_orig, GL_FALSE );
+
+      if (ctx->Light.ShadeModel == GL_FLAT)
+	 tnl->Driver.Render.CopyPV( ctx, newvert, v1 );
+
+      v1 = newvert;
+
+      newvert++;
+   }
+   else {
+      ASSERT(t1 == 0.0);
+   }
+
+   tnl->Driver.Render.ClippedLine( ctx, v0, v1 );
+}
+
+
+/* Clip a triangle against the viewport and user clip planes.
+ */
+static INLINE void
+TAG(clip_tri)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLubyte mask )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   tnl_interp_func interp = tnl->Driver.Render.Interp;
+   GLuint newvert = VB->Count;
+   GLfloat (*coord)[4] = VB->ClipPtr->data;
+   GLuint pv = v2;
+   GLuint vlist[2][MAX_CLIPPED_VERTICES];
+   GLuint *inlist = vlist[0], *outlist = vlist[1];
+   GLuint p;
+   GLuint n = 3;
+
+   ASSIGN_3V(inlist, v2, v0, v1 ); /* pv rotated to slot zero */
+
+   if (0) {
+      /* print pre-clip vertex coords */
+      GLuint i, j;
+      _mesa_printf("pre clip:\n");
+      for (i = 0; i < n; i++) {
+         j = inlist[i];
+         _mesa_printf("  %u: %u: %f, %f, %f, %f\n",
+                      i, j,
+                      coord[j][0], coord[j][1], coord[j][2], coord[j][3]);
+         assert(!IS_INF_OR_NAN(coord[j][0]));
+         assert(!IS_INF_OR_NAN(coord[j][1]));
+         assert(!IS_INF_OR_NAN(coord[j][2]));
+         assert(!IS_INF_OR_NAN(coord[j][3]));
+      }
+   }
+
+
+   if (mask & CLIP_FRUSTUM_BITS) {
+      POLY_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
+      POLY_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
+      POLY_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
+      POLY_CLIP( CLIP_BOTTOM_BIT,  0,  1,  0, 1 );
+      POLY_CLIP( CLIP_FAR_BIT,     0,  0, -1, 1 );
+      POLY_CLIP( CLIP_NEAR_BIT,    0,  0,  1, 1 );
+   }
+
+   if (mask & CLIP_USER_BIT) {
+      for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
+         if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
+            const GLfloat a = ctx->Transform._ClipUserPlane[p][0];
+            const GLfloat b = ctx->Transform._ClipUserPlane[p][1];
+            const GLfloat c = ctx->Transform._ClipUserPlane[p][2];
+            const GLfloat d = ctx->Transform._ClipUserPlane[p][3];
+            POLY_CLIP( CLIP_USER_BIT, a, b, c, d );
+         }
+      }
+   }
+
+   if (ctx->Light.ShadeModel == GL_FLAT) {
+      if (pv != inlist[0]) {
+	 ASSERT( inlist[0] >= VB->Count );
+	 tnl->Driver.Render.CopyPV( ctx, inlist[0], pv );
+      }
+   }
+
+   if (0) {
+      /* print post-clip vertex coords */
+      GLuint i, j;
+      _mesa_printf("post clip:\n");
+      for (i = 0; i < n; i++) {
+         j = inlist[i];
+         _mesa_printf("  %u: %u: %f, %f, %f, %f\n",
+                      i, j,
+                      coord[j][0], coord[j][1], coord[j][2], coord[j][3]);
+      }
+   }
+
+   tnl->Driver.Render.ClippedPolygon( ctx, inlist, n );
+}
+
+
+/* Clip a quad against the viewport and user clip planes.
+ */
+static INLINE void
+TAG(clip_quad)( GLcontext *ctx, GLuint v0, GLuint v1, GLuint v2, GLuint v3,
+                GLubyte mask )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   tnl_interp_func interp = tnl->Driver.Render.Interp;
+   GLuint newvert = VB->Count;
+   GLfloat (*coord)[4] = VB->ClipPtr->data;
+   GLuint pv = v3;
+   GLuint vlist[2][MAX_CLIPPED_VERTICES];
+   GLuint *inlist = vlist[0], *outlist = vlist[1];
+   GLuint p;
+   GLuint n = 4;
+
+   ASSIGN_4V(inlist, v3, v0, v1, v2 ); /* pv rotated to slot zero */
+
+   if (mask & CLIP_FRUSTUM_BITS) {
+      POLY_CLIP( CLIP_RIGHT_BIT,  -1,  0,  0, 1 );
+      POLY_CLIP( CLIP_LEFT_BIT,    1,  0,  0, 1 );
+      POLY_CLIP( CLIP_TOP_BIT,     0, -1,  0, 1 );
+      POLY_CLIP( CLIP_BOTTOM_BIT,  0,  1,  0, 1 );
+      POLY_CLIP( CLIP_FAR_BIT,     0,  0, -1, 1 );
+      POLY_CLIP( CLIP_NEAR_BIT,    0,  0,  1, 1 );
+   }
+
+   if (mask & CLIP_USER_BIT) {
+      for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
+	 if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
+            const GLfloat a = ctx->Transform._ClipUserPlane[p][0];
+            const GLfloat b = ctx->Transform._ClipUserPlane[p][1];
+            const GLfloat c = ctx->Transform._ClipUserPlane[p][2];
+            const GLfloat d = ctx->Transform._ClipUserPlane[p][3];
+	    POLY_CLIP( CLIP_USER_BIT, a, b, c, d );
+	 }
+      }
+   }
+
+   if (ctx->Light.ShadeModel == GL_FLAT) {
+      if (pv != inlist[0]) {
+	 ASSERT( inlist[0] >= VB->Count );
+	 tnl->Driver.Render.CopyPV( ctx, inlist[0], pv );
+      }
+   }
+
+   tnl->Driver.Render.ClippedPolygon( ctx, inlist, n );
+}
+
+#undef W
+#undef Z
+#undef Y
+#undef X
+#undef SIZE
+#undef TAG
+#undef POLY_CLIP
+#undef LINE_CLIP
diff --git a/mesalib/src/mesa/tnl/t_vb_cull.c b/mesalib/src/mesa/tnl/t_vb_cull.c
new file mode 100644
index 000000000..712901acf
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_cull.c
@@ -0,0 +1,97 @@
+/*
+ * 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 "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "math/m_xform.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+
+
+
+/* EXT_vertex_cull.  Not really a big win, but probably depends on
+ * your application.  This stage not included in the default pipeline.
+ */
+static GLboolean run_cull_stage( GLcontext *ctx,
+				 struct tnl_pipeline_stage *stage )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+
+   const GLfloat a = ctx->Transform.CullObjPos[0];
+   const GLfloat b = ctx->Transform.CullObjPos[1];
+   const GLfloat c = ctx->Transform.CullObjPos[2];
+   GLfloat *norm = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data;
+   GLuint stride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride;
+   GLuint count = VB->Count;
+   GLuint i;
+
+   if (ctx->VertexProgram._Current ||
+       !ctx->Transform.CullVertexFlag) 
+      return GL_TRUE;
+
+   VB->ClipOrMask &= ~CLIP_CULL_BIT;
+   VB->ClipAndMask |= CLIP_CULL_BIT;
+
+   for (i = 0 ; i < count ; i++) {
+      GLfloat dp = (norm[0] * a + 
+		    norm[1] * b +
+		    norm[2] * c);
+
+      if (dp < 0) {
+	 VB->ClipMask[i] |= CLIP_CULL_BIT;
+	 VB->ClipOrMask |= CLIP_CULL_BIT;
+      }
+      else {
+	 VB->ClipMask[i] &= ~CLIP_CULL_BIT;
+	 VB->ClipAndMask &= ~CLIP_CULL_BIT;
+      }
+
+      STRIDE_F(norm, stride);
+   }
+
+   return !(VB->ClipAndMask & CLIP_CULL_BIT);
+}
+
+
+
+const struct tnl_pipeline_stage _tnl_vertex_cull_stage =
+{
+   "EXT_cull_vertex",
+   NULL,			/* private data */
+   NULL,				/* ctr */
+   NULL,				/* destructor */
+   NULL,
+   run_cull_stage		/* run -- initially set to init */
+};
diff --git a/mesalib/src/mesa/tnl/t_vb_fog.c b/mesalib/src/mesa/tnl/t_vb_fog.c
new file mode 100644
index 000000000..f3a7bd49f
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_fog.c
@@ -0,0 +1,277 @@
+/*
+ * 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 "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "math/m_xform.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+
+
+struct fog_stage_data {
+   GLvector4f fogcoord;		/* has actual storage allocated */
+};
+
+#define FOG_STAGE_DATA(stage) ((struct fog_stage_data *)stage->privatePtr)
+
+#define FOG_EXP_TABLE_SIZE 256
+#define FOG_MAX (10.0)
+#define EXP_FOG_MAX .0006595
+#define FOG_INCR (FOG_MAX/FOG_EXP_TABLE_SIZE)
+static GLfloat exp_table[FOG_EXP_TABLE_SIZE];
+static GLfloat inited = 0;
+
+#if 1
+#define NEG_EXP( result, narg )						\
+do {									\
+   GLfloat f = (GLfloat) (narg * (1.0/FOG_INCR));			\
+   GLint k = (GLint) f;							\
+   if (k > FOG_EXP_TABLE_SIZE-2) 					\
+      result = (GLfloat) EXP_FOG_MAX;					\
+   else									\
+      result = exp_table[k] + (f-k)*(exp_table[k+1]-exp_table[k]);	\
+} while (0)
+#else
+#define NEG_EXP( result, narg )					\
+do {								\
+   result = exp(-narg);						\
+} while (0)
+#endif
+
+
+/**
+ * Initialize the exp_table[] lookup table for approximating exp().
+ */
+static void
+init_static_data( void )
+{
+   GLfloat f = 0.0F;
+   GLint i = 0;
+   for ( ; i < FOG_EXP_TABLE_SIZE ; i++, f += FOG_INCR) {
+      exp_table[i] = EXPF(-f);
+   }
+   inited = 1;
+}
+
+
+/**
+ * Compute per-vertex fog blend factors from fog coordinates by
+ * evaluating the GL_LINEAR, GL_EXP or GL_EXP2 fog function.
+ * Fog coordinates are distances from the eye (typically between the
+ * near and far clip plane distances).
+ * Note that fogcoords may be negative, if eye z is source absolute
+ * value must be taken earlier.
+ * Fog blend factors are in the range [0,1].
+ */
+static void
+compute_fog_blend_factors(GLcontext *ctx, GLvector4f *out, const GLvector4f *in)
+{
+   GLfloat end  = ctx->Fog.End;
+   GLfloat *v = in->start;
+   GLuint stride = in->stride;
+   GLuint n = in->count;
+   GLfloat (*data)[4] = out->data;
+   GLfloat d;
+   GLuint i;
+
+   out->count = in->count;
+
+   switch (ctx->Fog.Mode) {
+   case GL_LINEAR:
+      if (ctx->Fog.Start == ctx->Fog.End)
+         d = 1.0F;
+      else
+         d = 1.0F / (ctx->Fog.End - ctx->Fog.Start);
+      for ( i = 0 ; i < n ; i++, STRIDE_F(v, stride)) {
+         const GLfloat z = *v;
+         GLfloat f = (end - z) * d;
+	 data[i][0] = CLAMP(f, 0.0F, 1.0F);
+      }
+      break;
+   case GL_EXP:
+      d = ctx->Fog.Density;
+      for ( i = 0 ; i < n ; i++, STRIDE_F(v,stride)) {
+         const GLfloat z = *v;
+         NEG_EXP( data[i][0], d * z );
+      }
+      break;
+   case GL_EXP2:
+      d = ctx->Fog.Density*ctx->Fog.Density;
+      for ( i = 0 ; i < n ; i++, STRIDE_F(v, stride)) {
+         const GLfloat z = *v;
+         NEG_EXP( data[i][0], d * z * z );
+      }
+      break;
+   default:
+      _mesa_problem(ctx, "Bad fog mode in make_fog_coord");
+      return;
+   }
+}
+
+
+static GLboolean
+run_fog_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   struct fog_stage_data *store = FOG_STAGE_DATA(stage);
+   GLvector4f *input;
+
+
+   if (!ctx->Fog.Enabled)
+      return GL_TRUE;
+
+   if (ctx->Fog.FogCoordinateSource == GL_FRAGMENT_DEPTH_EXT && !ctx->VertexProgram._Current) {
+      GLuint i;
+      GLfloat *coord;
+      /* Fog is computed from vertex or fragment Z values */
+      /* source = VB->ObjPtr or VB->EyePtr coords */
+      /* dest = VB->AttribPtr[_TNL_ATTRIB_FOG] = fog stage private storage */
+      VB->AttribPtr[_TNL_ATTRIB_FOG] = &store->fogcoord;
+
+      if (!ctx->_NeedEyeCoords) {
+         /* compute fog coords from object coords */
+	 const GLfloat *m = ctx->ModelviewMatrixStack.Top->m;
+	 GLfloat plane[4];
+
+	 /* Use this to store calculated eye z values:
+	  */
+	 input = &store->fogcoord;
+
+	 plane[0] = m[2];
+	 plane[1] = m[6];
+	 plane[2] = m[10];
+	 plane[3] = m[14];
+	 /* Full eye coords weren't required, just calculate the
+	  * eye Z values.
+	  */
+	 _mesa_dotprod_tab[VB->ObjPtr->size]( (GLfloat *) input->data,
+					      4 * sizeof(GLfloat),
+					      VB->ObjPtr, plane );
+
+	 input->count = VB->ObjPtr->count;
+
+	 /* make sure coords are really positive
+	    NOTE should avoid going through array twice */
+	 coord = input->start;
+	 for (i = 0; i < input->count; i++) {
+	    *coord = FABSF(*coord);
+	    STRIDE_F(coord, input->stride);
+	 }
+      }
+      else {
+         /* fog coordinates = eye Z coordinates - need to copy for ABS */
+	 input = &store->fogcoord;
+
+	 if (VB->EyePtr->size < 2)
+	    _mesa_vector4f_clean_elem( VB->EyePtr, VB->Count, 2 );
+
+	 input->stride = 4 * sizeof(GLfloat);
+	 input->count = VB->EyePtr->count;
+	 coord = VB->EyePtr->start;
+	 for (i = 0 ; i < VB->EyePtr->count; i++) {
+	    input->data[i][0] = FABSF(coord[2]);
+	    STRIDE_F(coord, VB->EyePtr->stride);
+	 }
+      }
+   }
+   else {
+      /* use glFogCoord() coordinates */
+      input = VB->AttribPtr[_TNL_ATTRIB_FOG];  /* source data */
+
+      /* input->count may be one if glFogCoord was only called once
+       * before glBegin.  But we need to compute fog for all vertices.
+       */
+      input->count = VB->ObjPtr->count;
+
+      VB->AttribPtr[_TNL_ATTRIB_FOG] = &store->fogcoord;  /* dest data */
+   }
+
+   if (tnl->_DoVertexFog) {
+      /* compute blend factors from fog coordinates */
+      compute_fog_blend_factors( ctx, VB->AttribPtr[_TNL_ATTRIB_FOG], input );
+   }
+   else {
+      /* results = incoming fog coords (compute fog per-fragment later) */
+      VB->AttribPtr[_TNL_ATTRIB_FOG] = input;
+   }
+
+   VB->FogCoordPtr = VB->AttribPtr[_TNL_ATTRIB_FOG];
+   return GL_TRUE;
+}
+
+
+
+/* Called the first time stage->run() is invoked.
+ */
+static GLboolean
+alloc_fog_data(GLcontext *ctx, struct tnl_pipeline_stage *stage)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct fog_stage_data *store;
+   stage->privatePtr = MALLOC(sizeof(*store));
+   store = FOG_STAGE_DATA(stage);
+   if (!store)
+      return GL_FALSE;
+
+   _mesa_vector4f_alloc( &store->fogcoord, 0, tnl->vb.Size, 32 );
+
+   if (!inited)
+      init_static_data();
+
+   return GL_TRUE;
+}
+
+
+static void
+free_fog_data(struct tnl_pipeline_stage *stage)
+{
+   struct fog_stage_data *store = FOG_STAGE_DATA(stage);
+   if (store) {
+      _mesa_vector4f_free( &store->fogcoord );
+      FREE( store );
+      stage->privatePtr = NULL;
+   }
+}
+
+
+const struct tnl_pipeline_stage _tnl_fog_coordinate_stage =
+{
+   "build fog coordinates",	/* name */
+   NULL,			/* private_data */
+   alloc_fog_data,		/* dtr */
+   free_fog_data,		/* dtr */
+   NULL,		/* check */
+   run_fog_stage		/* run -- initially set to init. */
+};
diff --git a/mesalib/src/mesa/tnl/t_vb_light.c b/mesalib/src/mesa/tnl/t_vb_light.c
new file mode 100644
index 000000000..f47f99397
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_light.c
@@ -0,0 +1,362 @@
+/*
+ * 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.
+ */
+
+
+
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/light.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/simple_list.h"
+#include "main/mtypes.h"
+
+#include "math/m_translate.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+
+#define LIGHT_TWOSIDE       0x1
+#define LIGHT_MATERIAL      0x2
+#define MAX_LIGHT_FUNC      0x4
+
+typedef void (*light_func)( GLcontext *ctx,
+			    struct vertex_buffer *VB,
+			    struct tnl_pipeline_stage *stage,
+			    GLvector4f *input );
+
+/**
+ * Information for updating current material attributes from vertex color,
+ * for GL_COLOR_MATERIAL.
+ */
+struct material_cursor {
+   const GLfloat *ptr;    /* points to src vertex color (in VB array) */
+   GLuint stride;         /* stride to next vertex color (bytes) */
+   GLfloat *current;      /* points to material attribute to update */
+   GLuint size;           /* vertex/color size: 1, 2, 3 or 4 */
+};
+
+/**
+ * Data private to this pipeline stage.
+ */
+struct light_stage_data {
+   GLvector4f Input;
+   GLvector4f LitColor[2];
+   GLvector4f LitSecondary[2];
+   GLvector4f LitIndex[2];
+   light_func *light_func_tab;
+
+   struct material_cursor mat[MAT_ATTRIB_MAX];
+   GLuint mat_count;
+   GLuint mat_bitmask;
+};
+
+
+#define LIGHT_STAGE_DATA(stage) ((struct light_stage_data *)(stage->privatePtr))
+
+
+
+/**
+ * In the case of colormaterial, the effected material attributes
+ * should already have been bound to point to the incoming color data,
+ * prior to running the pipeline.
+ * This function copies the vertex's color to the material attributes
+ * which are tracking glColor.
+ * It's called per-vertex in the lighting loop.
+ */
+static void
+update_materials(GLcontext *ctx, struct light_stage_data *store)
+{
+   GLuint i;
+
+   for (i = 0 ; i < store->mat_count ; i++) {
+      /* update the material */
+      COPY_CLEAN_4V(store->mat[i].current, store->mat[i].size, store->mat[i].ptr);
+      /* increment src vertex color pointer */
+      STRIDE_F(store->mat[i].ptr, store->mat[i].stride);
+   }
+      
+   /* recompute derived light/material values */
+   _mesa_update_material( ctx, store->mat_bitmask );
+   /* XXX we should only call this if we're tracking/changing the specular
+    * exponent.
+    */
+   _mesa_validate_all_lighting_tables( ctx );
+}
+
+
+/**
+ * Prepare things prior to running the lighting stage.
+ * Return number of material attributes which will track vertex color.
+ */
+static GLuint
+prepare_materials(GLcontext *ctx,
+                  struct vertex_buffer *VB, struct light_stage_data *store)
+{
+   GLuint i;
+   
+   store->mat_count = 0;
+   store->mat_bitmask = 0;
+
+   /* Examine the ColorMaterialBitmask to determine which materials
+    * track vertex color.  Override the material attribute's pointer
+    * with the color pointer for each one.
+    */
+   if (ctx->Light.ColorMaterialEnabled) {
+      const GLuint bitmask = ctx->Light.ColorMaterialBitmask;
+      for (i = 0 ; i < MAT_ATTRIB_MAX ; i++)
+	 if (bitmask & (1<<i))
+	    VB->AttribPtr[_TNL_ATTRIB_MAT_FRONT_AMBIENT + i] = VB->ColorPtr[0];
+   }
+
+   /* Now, for each material attribute that's tracking vertex color, save
+    * some values (ptr, stride, size, current) that we'll need in
+    * update_materials(), above, that'll actually copy the vertex color to
+    * the material attribute(s).
+    */
+   for (i = _TNL_FIRST_MAT; i <= _TNL_LAST_MAT; i++) {
+      if (VB->AttribPtr[i]->stride) {
+	 const GLuint j = store->mat_count++;
+	 const GLuint attr = i - _TNL_ATTRIB_MAT_FRONT_AMBIENT;
+	 store->mat[j].ptr    = VB->AttribPtr[i]->start;
+	 store->mat[j].stride = VB->AttribPtr[i]->stride;
+	 store->mat[j].size   = VB->AttribPtr[i]->size;
+	 store->mat[j].current = ctx->Light.Material.Attrib[attr];
+	 store->mat_bitmask |= (1<<attr);
+      }
+   }
+
+   /* FIXME: Is this already done?
+    */
+   _mesa_update_material( ctx, ~0 );
+   _mesa_validate_all_lighting_tables( ctx );
+
+   return store->mat_count;
+}
+
+/* Tables for all the shading functions.
+ */
+static light_func _tnl_light_tab[MAX_LIGHT_FUNC];
+static light_func _tnl_light_fast_tab[MAX_LIGHT_FUNC];
+static light_func _tnl_light_fast_single_tab[MAX_LIGHT_FUNC];
+static light_func _tnl_light_spec_tab[MAX_LIGHT_FUNC];
+static light_func _tnl_light_ci_tab[MAX_LIGHT_FUNC];
+
+#define TAG(x)           x
+#define IDX              (0)
+#include "t_vb_lighttmp.h"
+
+#define TAG(x)           x##_twoside
+#define IDX              (LIGHT_TWOSIDE)
+#include "t_vb_lighttmp.h"
+
+#define TAG(x)           x##_material
+#define IDX              (LIGHT_MATERIAL)
+#include "t_vb_lighttmp.h"
+
+#define TAG(x)           x##_twoside_material
+#define IDX              (LIGHT_TWOSIDE|LIGHT_MATERIAL)
+#include "t_vb_lighttmp.h"
+
+
+static void init_lighting_tables( void )
+{
+   static int done;
+
+   if (!done) {
+      init_light_tab();
+      init_light_tab_twoside();
+      init_light_tab_material();
+      init_light_tab_twoside_material();
+      done = 1;
+   }
+}
+
+
+static GLboolean run_lighting( GLcontext *ctx, 
+			       struct tnl_pipeline_stage *stage )
+{
+   struct light_stage_data *store = LIGHT_STAGE_DATA(stage);
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLvector4f *input = ctx->_NeedEyeCoords ? VB->EyePtr : VB->ObjPtr;
+   GLuint idx;
+
+   if (!ctx->Light.Enabled || ctx->VertexProgram._Current)
+      return GL_TRUE;
+
+   /* Make sure we can talk about position x,y and z:
+    */
+   if (input->size <= 2 && input == VB->ObjPtr) {
+
+      _math_trans_4f( store->Input.data,
+		      VB->ObjPtr->data,
+		      VB->ObjPtr->stride,
+		      GL_FLOAT,
+		      VB->ObjPtr->size,
+		      0,
+		      VB->Count );
+
+      if (input->size <= 2) {
+	 /* Clean z.
+	  */
+	 _mesa_vector4f_clean_elem(&store->Input, VB->Count, 2);
+      }
+	 
+      if (input->size <= 1) {
+	 /* Clean y.
+	  */
+	 _mesa_vector4f_clean_elem(&store->Input, VB->Count, 1);
+      }
+
+      input = &store->Input;
+   }
+   
+   idx = 0;
+
+   if (prepare_materials( ctx, VB, store ))
+      idx |= LIGHT_MATERIAL;
+
+   if (ctx->Light.Model.TwoSide)
+      idx |= LIGHT_TWOSIDE;
+
+   /* The individual functions know about replaying side-effects
+    * vs. full re-execution. 
+    */
+   store->light_func_tab[idx]( ctx, VB, stage, input );
+
+   VB->AttribPtr[_TNL_ATTRIB_COLOR0] = VB->ColorPtr[0];
+   VB->AttribPtr[_TNL_ATTRIB_COLOR1] = VB->SecondaryColorPtr[0];
+   VB->AttribPtr[_TNL_ATTRIB_COLOR_INDEX] = VB->IndexPtr[0];
+
+   return GL_TRUE;
+}
+
+
+/* Called in place of do_lighting when the light table may have changed.
+ */
+static void validate_lighting( GLcontext *ctx,
+					struct tnl_pipeline_stage *stage )
+{
+   light_func *tab;
+
+   if (!ctx->Light.Enabled || ctx->VertexProgram._Current)
+      return;
+
+   if (ctx->Visual.rgbMode) {
+      if (ctx->Light._NeedVertices) {
+	 if (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR)
+	    tab = _tnl_light_spec_tab;
+	 else
+	    tab = _tnl_light_tab;
+      }
+      else {
+	 if (ctx->Light.EnabledList.next == ctx->Light.EnabledList.prev)
+	    tab = _tnl_light_fast_single_tab;
+	 else
+	    tab = _tnl_light_fast_tab;
+      }
+   }
+   else
+      tab = _tnl_light_ci_tab;
+
+
+   LIGHT_STAGE_DATA(stage)->light_func_tab = tab;
+
+   /* This and the above should only be done on _NEW_LIGHT:
+    */
+   TNL_CONTEXT(ctx)->Driver.NotifyMaterialChange( ctx );
+}
+
+
+
+/* Called the first time stage->run is called.  In effect, don't
+ * allocate data until the first time the stage is run.
+ */
+static GLboolean init_lighting( GLcontext *ctx,
+				struct tnl_pipeline_stage *stage )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct light_stage_data *store;
+   GLuint size = tnl->vb.Size;
+
+   stage->privatePtr = MALLOC(sizeof(*store));
+   store = LIGHT_STAGE_DATA(stage);
+   if (!store)
+      return GL_FALSE;
+
+   /* Do onetime init.
+    */
+   init_lighting_tables();
+
+   _mesa_vector4f_alloc( &store->Input, 0, size, 32 );
+   _mesa_vector4f_alloc( &store->LitColor[0], 0, size, 32 );
+   _mesa_vector4f_alloc( &store->LitColor[1], 0, size, 32 );
+   _mesa_vector4f_alloc( &store->LitSecondary[0], 0, size, 32 );
+   _mesa_vector4f_alloc( &store->LitSecondary[1], 0, size, 32 );
+   _mesa_vector4f_alloc( &store->LitIndex[0], 0, size, 32 );
+   _mesa_vector4f_alloc( &store->LitIndex[1], 0, size, 32 );
+
+   store->LitColor[0].size = 4;
+   store->LitColor[1].size = 4;
+   store->LitSecondary[0].size = 3;
+   store->LitSecondary[1].size = 3;
+
+   store->LitIndex[0].size = 1;
+   store->LitIndex[0].stride = sizeof(GLfloat);
+   store->LitIndex[1].size = 1;
+   store->LitIndex[1].stride = sizeof(GLfloat);
+
+   return GL_TRUE;
+}
+
+
+
+
+static void dtr( struct tnl_pipeline_stage *stage )
+{
+   struct light_stage_data *store = LIGHT_STAGE_DATA(stage);
+
+   if (store) {
+      _mesa_vector4f_free( &store->Input );
+      _mesa_vector4f_free( &store->LitColor[0] );
+      _mesa_vector4f_free( &store->LitColor[1] );
+      _mesa_vector4f_free( &store->LitSecondary[0] );
+      _mesa_vector4f_free( &store->LitSecondary[1] );
+      _mesa_vector4f_free( &store->LitIndex[0] );
+      _mesa_vector4f_free( &store->LitIndex[1] );
+      FREE( store );
+      stage->privatePtr = NULL;
+   }
+}
+
+const struct tnl_pipeline_stage _tnl_lighting_stage =
+{
+   "lighting",			/* name */
+   NULL,			/* private_data */
+   init_lighting,
+   dtr,				/* destroy */
+   validate_lighting,
+   run_lighting
+};
diff --git a/mesalib/src/mesa/tnl/t_vb_lighttmp.h b/mesalib/src/mesa/tnl/t_vb_lighttmp.h
new file mode 100644
index 000000000..124ca3c74
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_lighttmp.h
@@ -0,0 +1,824 @@
+/*
+ * 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:
+ *    Brian Paul
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+
+#if IDX & LIGHT_TWOSIDE
+#  define NR_SIDES 2
+#else
+#  define NR_SIDES 1
+#endif
+
+
+/* define TRACE to trace lighting code */
+/* #define TRACE 1 */
+
+/*
+ * ctx is the current context
+ * VB is the vertex buffer
+ * stage is the lighting stage-private data
+ * input is the vector of eye or object-space vertex coordinates
+ */
+static void TAG(light_rgba_spec)( GLcontext *ctx,
+				  struct vertex_buffer *VB,
+				  struct tnl_pipeline_stage *stage,
+				  GLvector4f *input )
+{
+   struct light_stage_data *store = LIGHT_STAGE_DATA(stage);
+   GLfloat (*base)[3] = ctx->Light._BaseColor;
+   GLfloat sumA[2];
+   GLuint j;
+
+   const GLuint vstride = input->stride;
+   const GLfloat *vertex = (GLfloat *)input->data;
+   const GLuint nstride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride;
+   const GLfloat *normal = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data;
+
+   GLfloat (*Fcolor)[4] = (GLfloat (*)[4]) store->LitColor[0].data;
+   GLfloat (*Fspec)[4] = (GLfloat (*)[4]) store->LitSecondary[0].data;
+#if IDX & LIGHT_TWOSIDE
+   GLfloat (*Bcolor)[4] = (GLfloat (*)[4]) store->LitColor[1].data;
+   GLfloat (*Bspec)[4] = (GLfloat (*)[4]) store->LitSecondary[1].data;
+#endif
+
+   const GLuint nr = VB->Count;
+
+#ifdef TRACE
+   fprintf(stderr, "%s\n", __FUNCTION__ );
+#endif
+
+   VB->ColorPtr[0] = &store->LitColor[0];
+   VB->SecondaryColorPtr[0] = &store->LitSecondary[0];
+   sumA[0] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3];
+
+#if IDX & LIGHT_TWOSIDE
+   VB->ColorPtr[1] = &store->LitColor[1];
+   VB->SecondaryColorPtr[1] = &store->LitSecondary[1];
+   sumA[1] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3];
+#endif
+
+
+   store->LitColor[0].stride = 16;
+   store->LitColor[1].stride = 16;
+
+   for (j = 0; j < nr; j++,STRIDE_F(vertex,vstride),STRIDE_F(normal,nstride)) {
+      GLfloat sum[2][3], spec[2][3];
+      struct gl_light *light;
+
+#if IDX & LIGHT_MATERIAL
+      update_materials( ctx, store );
+      sumA[0] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3];
+#if IDX & LIGHT_TWOSIDE
+      sumA[1] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3];
+#endif
+#endif
+
+      COPY_3V(sum[0], base[0]);
+      ZERO_3V(spec[0]);
+
+#if IDX & LIGHT_TWOSIDE
+      COPY_3V(sum[1], base[1]);
+      ZERO_3V(spec[1]);
+#endif
+
+      /* Add contribution from each enabled light source */
+      foreach (light, &ctx->Light.EnabledList) {
+	 GLfloat n_dot_h;
+	 GLfloat correction;
+	 GLint side;
+	 GLfloat contrib[3];
+	 GLfloat attenuation;
+	 GLfloat VP[3];  /* unit vector from vertex to light */
+	 GLfloat n_dot_VP;       /* n dot VP */
+	 GLfloat *h;
+
+	 /* compute VP and attenuation */
+	 if (!(light->_Flags & LIGHT_POSITIONAL)) {
+	    /* directional light */
+	    COPY_3V(VP, light->_VP_inf_norm);
+	    attenuation = light->_VP_inf_spot_attenuation;
+	 }
+	 else {
+	    GLfloat d;     /* distance from vertex to light */
+
+	    SUB_3V(VP, light->_Position, vertex);
+
+	    d = (GLfloat) LEN_3FV( VP );
+
+	    if (d > 1e-6) {
+	       GLfloat invd = 1.0F / d;
+	       SELF_SCALE_SCALAR_3V(VP, invd);
+	    }
+
+	    attenuation = 1.0F / (light->ConstantAttenuation + d *
+				  (light->LinearAttenuation + d *
+				   light->QuadraticAttenuation));
+
+	    /* spotlight attenuation */
+	    if (light->_Flags & LIGHT_SPOT) {
+	       GLfloat PV_dot_dir = - DOT3(VP, light->_NormSpotDirection);
+
+	       if (PV_dot_dir<light->_CosCutoff) {
+		  continue; /* this light makes no contribution */
+	       }
+	       else {
+		  GLdouble x = PV_dot_dir * (EXP_TABLE_SIZE-1);
+		  GLint k = (GLint) x;
+		  GLfloat spot = (GLfloat) (light->_SpotExpTable[k][0]
+				    + (x-k)*light->_SpotExpTable[k][1]);
+		  attenuation *= spot;
+	       }
+	    }
+	 }
+
+	 if (attenuation < 1e-3)
+	    continue;		/* this light makes no contribution */
+
+	 /* Compute dot product or normal and vector from V to light pos */
+	 n_dot_VP = DOT3( normal, VP );
+
+	 /* Which side gets the diffuse & specular terms? */
+	 if (n_dot_VP < 0.0F) {
+	    ACC_SCALE_SCALAR_3V(sum[0], attenuation, light->_MatAmbient[0]);
+#if IDX & LIGHT_TWOSIDE
+	    side = 1;
+	    correction = -1;
+	    n_dot_VP = -n_dot_VP;
+#else
+            continue;
+#endif
+	 }
+         else {
+#if IDX & LIGHT_TWOSIDE
+            ACC_SCALE_SCALAR_3V( sum[1], attenuation, light->_MatAmbient[1]);
+#endif
+	    side = 0;
+	    correction = 1;
+	 }
+
+	 /* diffuse term */
+	 COPY_3V(contrib, light->_MatAmbient[side]);
+	 ACC_SCALE_SCALAR_3V(contrib, n_dot_VP, light->_MatDiffuse[side]);
+	 ACC_SCALE_SCALAR_3V(sum[side], attenuation, contrib );
+
+	 /* specular term - cannibalize VP... */
+	 if (ctx->Light.Model.LocalViewer) {
+	    GLfloat v[3];
+	    COPY_3V(v, vertex);
+	    NORMALIZE_3FV(v);
+	    SUB_3V(VP, VP, v);                /* h = VP + VPe */
+	    h = VP;
+	    NORMALIZE_3FV(h);
+	 }
+	 else if (light->_Flags & LIGHT_POSITIONAL) {
+	    h = VP;
+	    ACC_3V(h, ctx->_EyeZDir);
+	    NORMALIZE_3FV(h);
+	 }
+         else {
+	    h = light->_h_inf_norm;
+	 }
+
+	 n_dot_h = correction * DOT3(normal, h);
+
+	 if (n_dot_h > 0.0F) {
+	    GLfloat spec_coef;
+	    struct gl_shine_tab *tab = ctx->_ShineTable[side];
+	    GET_SHINE_TAB_ENTRY( tab, n_dot_h, spec_coef );
+
+	    if (spec_coef > 1.0e-10) {
+	       spec_coef *= attenuation;
+	       ACC_SCALE_SCALAR_3V( spec[side], spec_coef,
+				    light->_MatSpecular[side]);
+	    }
+	 }
+      } /*loop over lights*/
+
+      COPY_3V( Fcolor[j], sum[0] );
+      COPY_3V( Fspec[j], spec[0] );
+      Fcolor[j][3] = sumA[0];
+
+#if IDX & LIGHT_TWOSIDE
+      COPY_3V( Bcolor[j], sum[1] );
+      COPY_3V( Bspec[j], spec[1] );
+      Bcolor[j][3] = sumA[1];
+#endif
+   }
+}
+
+
+static void TAG(light_rgba)( GLcontext *ctx,
+			     struct vertex_buffer *VB,
+			     struct tnl_pipeline_stage *stage,
+			     GLvector4f *input )
+{
+   struct light_stage_data *store = LIGHT_STAGE_DATA(stage);
+   GLuint j;
+
+   GLfloat (*base)[3] = ctx->Light._BaseColor;
+   GLfloat sumA[2];
+
+   const GLuint vstride = input->stride;
+   const GLfloat *vertex = (GLfloat *) input->data;
+   const GLuint nstride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride;
+   const GLfloat *normal = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data;
+
+   GLfloat (*Fcolor)[4] = (GLfloat (*)[4]) store->LitColor[0].data;
+#if IDX & LIGHT_TWOSIDE
+   GLfloat (*Bcolor)[4] = (GLfloat (*)[4]) store->LitColor[1].data;
+#endif
+
+   const GLuint nr = VB->Count;
+
+#ifdef TRACE
+   fprintf(stderr, "%s\n", __FUNCTION__ );
+#endif
+
+   VB->ColorPtr[0] = &store->LitColor[0];
+   sumA[0] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3];
+
+#if IDX & LIGHT_TWOSIDE
+   VB->ColorPtr[1] = &store->LitColor[1];
+   sumA[1] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3];
+#endif
+
+   store->LitColor[0].stride = 16;
+   store->LitColor[1].stride = 16;
+
+   for (j = 0; j < nr; j++,STRIDE_F(vertex,vstride),STRIDE_F(normal,nstride)) {
+      GLfloat sum[2][3];
+      struct gl_light *light;
+
+#if IDX & LIGHT_MATERIAL
+      update_materials( ctx, store );
+      sumA[0] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3];
+#if IDX & LIGHT_TWOSIDE
+      sumA[1] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3];
+#endif
+#endif
+
+      COPY_3V(sum[0], base[0]);
+
+#if IDX & LIGHT_TWOSIDE
+      COPY_3V(sum[1], base[1]);
+#endif
+
+      /* Add contribution from each enabled light source */
+      foreach (light, &ctx->Light.EnabledList) {
+
+	 GLfloat n_dot_h;
+	 GLfloat correction;
+	 GLint side;
+	 GLfloat contrib[3];
+	 GLfloat attenuation = 1.0;
+	 GLfloat VP[3];          /* unit vector from vertex to light */
+	 GLfloat n_dot_VP;       /* n dot VP */
+	 GLfloat *h;
+
+	 /* compute VP and attenuation */
+	 if (!(light->_Flags & LIGHT_POSITIONAL)) {
+	    /* directional light */
+	    COPY_3V(VP, light->_VP_inf_norm);
+	    attenuation = light->_VP_inf_spot_attenuation;
+	 }
+	 else {
+	    GLfloat d;     /* distance from vertex to light */
+
+
+	    SUB_3V(VP, light->_Position, vertex);
+
+	    d = (GLfloat) LEN_3FV( VP );
+
+	    if ( d > 1e-6) {
+	       GLfloat invd = 1.0F / d;
+	       SELF_SCALE_SCALAR_3V(VP, invd);
+	    }
+
+            attenuation = 1.0F / (light->ConstantAttenuation + d *
+                                  (light->LinearAttenuation + d *
+                                   light->QuadraticAttenuation));
+
+	    /* spotlight attenuation */
+	    if (light->_Flags & LIGHT_SPOT) {
+	       GLfloat PV_dot_dir = - DOT3(VP, light->_NormSpotDirection);
+
+	       if (PV_dot_dir<light->_CosCutoff) {
+		  continue; /* this light makes no contribution */
+	       }
+	       else {
+		  GLdouble x = PV_dot_dir * (EXP_TABLE_SIZE-1);
+		  GLint k = (GLint) x;
+		  GLfloat spot = (GLfloat) (light->_SpotExpTable[k][0]
+				  + (x-k)*light->_SpotExpTable[k][1]);
+		  attenuation *= spot;
+	       }
+	    }
+	 }
+
+	 if (attenuation < 1e-3)
+	    continue;		/* this light makes no contribution */
+
+	 /* Compute dot product or normal and vector from V to light pos */
+	 n_dot_VP = DOT3( normal, VP );
+
+	 /* which side are we lighting? */
+	 if (n_dot_VP < 0.0F) {
+	    ACC_SCALE_SCALAR_3V(sum[0], attenuation, light->_MatAmbient[0]);
+#if IDX & LIGHT_TWOSIDE
+	    side = 1;
+	    correction = -1;
+	    n_dot_VP = -n_dot_VP;
+#else
+            continue;
+#endif
+	 }
+         else {
+#if IDX & LIGHT_TWOSIDE
+            ACC_SCALE_SCALAR_3V( sum[1], attenuation, light->_MatAmbient[1]);
+#endif
+	    side = 0;
+	    correction = 1;
+	 }
+
+	 COPY_3V(contrib, light->_MatAmbient[side]);
+
+	 /* diffuse term */
+	 ACC_SCALE_SCALAR_3V(contrib, n_dot_VP, light->_MatDiffuse[side]);
+
+	 /* specular term - cannibalize VP... */
+	 {
+	    if (ctx->Light.Model.LocalViewer) {
+	       GLfloat v[3];
+	       COPY_3V(v, vertex);
+	       NORMALIZE_3FV(v);
+	       SUB_3V(VP, VP, v);                /* h = VP + VPe */
+	       h = VP;
+	       NORMALIZE_3FV(h);
+	    }
+	    else if (light->_Flags & LIGHT_POSITIONAL) {
+	       h = VP;
+	       ACC_3V(h, ctx->_EyeZDir);
+	       NORMALIZE_3FV(h);
+	    }
+            else {
+	       h = light->_h_inf_norm;
+	    }
+
+	    n_dot_h = correction * DOT3(normal, h);
+
+	    if (n_dot_h > 0.0F)
+	    {
+	       GLfloat spec_coef;
+	       struct gl_shine_tab *tab = ctx->_ShineTable[side];
+
+	       GET_SHINE_TAB_ENTRY( tab, n_dot_h, spec_coef );
+
+	       ACC_SCALE_SCALAR_3V( contrib, spec_coef,
+				    light->_MatSpecular[side]);
+	    }
+	 }
+
+	 ACC_SCALE_SCALAR_3V( sum[side], attenuation, contrib );
+      }
+
+      COPY_3V( Fcolor[j], sum[0] );
+      Fcolor[j][3] = sumA[0];
+
+#if IDX & LIGHT_TWOSIDE
+      COPY_3V( Bcolor[j], sum[1] );
+      Bcolor[j][3] = sumA[1];
+#endif
+   }
+}
+
+
+
+
+/* As below, but with just a single light.
+ */
+static void TAG(light_fast_rgba_single)( GLcontext *ctx,
+					 struct vertex_buffer *VB,
+					 struct tnl_pipeline_stage *stage,
+					 GLvector4f *input )
+
+{
+   struct light_stage_data *store = LIGHT_STAGE_DATA(stage);
+   const GLuint nstride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride;
+   const GLfloat *normal = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data;
+   GLfloat (*Fcolor)[4] = (GLfloat (*)[4]) store->LitColor[0].data;
+#if IDX & LIGHT_TWOSIDE
+   GLfloat (*Bcolor)[4] = (GLfloat (*)[4]) store->LitColor[1].data;
+#endif
+   const struct gl_light *light = ctx->Light.EnabledList.next;
+   GLuint j = 0;
+   GLfloat base[2][4];
+#if IDX & LIGHT_MATERIAL
+   const GLuint nr = VB->Count;
+#else
+   const GLuint nr = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->count;
+#endif
+
+#ifdef TRACE
+   fprintf(stderr, "%s\n", __FUNCTION__ );
+#endif
+
+   (void) input;		/* doesn't refer to Eye or Obj */
+
+   VB->ColorPtr[0] = &store->LitColor[0];
+#if IDX & LIGHT_TWOSIDE
+   VB->ColorPtr[1] = &store->LitColor[1];
+#endif
+
+   if (nr > 1) {
+      store->LitColor[0].stride = 16;
+      store->LitColor[1].stride = 16;
+   }
+   else {
+      store->LitColor[0].stride = 0;
+      store->LitColor[1].stride = 0;
+   }
+
+   for (j = 0; j < nr; j++, STRIDE_F(normal,nstride)) {
+
+      GLfloat n_dot_VP;
+
+#if IDX & LIGHT_MATERIAL
+      update_materials( ctx, store );
+#endif
+
+      /* No attenuation, so incoporate _MatAmbient into base color.
+       */
+#if !(IDX & LIGHT_MATERIAL)
+      if ( j == 0 )
+#endif
+      {
+	 COPY_3V(base[0], light->_MatAmbient[0]);
+	 ACC_3V(base[0], ctx->Light._BaseColor[0] );
+	 base[0][3] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3];
+
+#if IDX & LIGHT_TWOSIDE
+         COPY_3V(base[1], light->_MatAmbient[1]);
+         ACC_3V(base[1], ctx->Light._BaseColor[1]);
+         base[1][3] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3];
+#endif
+      }
+
+      n_dot_VP = DOT3(normal, light->_VP_inf_norm);
+
+      if (n_dot_VP < 0.0F) {
+#if IDX & LIGHT_TWOSIDE
+         GLfloat n_dot_h = -DOT3(normal, light->_h_inf_norm);
+         GLfloat sum[3];
+         COPY_3V(sum, base[1]);
+         ACC_SCALE_SCALAR_3V(sum, -n_dot_VP, light->_MatDiffuse[1]);
+         if (n_dot_h > 0.0F) {
+            GLfloat spec;
+            GET_SHINE_TAB_ENTRY( ctx->_ShineTable[1], n_dot_h, spec );
+            ACC_SCALE_SCALAR_3V(sum, spec, light->_MatSpecular[1]);
+         }
+         COPY_3V(Bcolor[j], sum );
+         Bcolor[j][3] = base[1][3];
+#endif
+	 COPY_4FV(Fcolor[j], base[0]);
+      }
+      else {
+	 GLfloat n_dot_h = DOT3(normal, light->_h_inf_norm);
+	 GLfloat sum[3];
+	 COPY_3V(sum, base[0]);
+	 ACC_SCALE_SCALAR_3V(sum, n_dot_VP, light->_MatDiffuse[0]);
+	 if (n_dot_h > 0.0F) {
+	    GLfloat spec;
+	    GET_SHINE_TAB_ENTRY( ctx->_ShineTable[0], n_dot_h, spec );
+	    ACC_SCALE_SCALAR_3V(sum, spec, light->_MatSpecular[0]);
+
+	 }
+	 COPY_3V(Fcolor[j], sum );
+	 Fcolor[j][3] = base[0][3];
+#if IDX & LIGHT_TWOSIDE
+         COPY_4FV(Bcolor[j], base[1]);
+#endif
+      }
+   }
+}
+
+
+/* Light infinite lights
+ */
+static void TAG(light_fast_rgba)( GLcontext *ctx,
+				  struct vertex_buffer *VB,
+				  struct tnl_pipeline_stage *stage,
+				  GLvector4f *input )
+{
+   struct light_stage_data *store = LIGHT_STAGE_DATA(stage);
+   GLfloat sumA[2];
+   const GLuint nstride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride;
+   const GLfloat *normal = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data;
+   GLfloat (*Fcolor)[4] = (GLfloat (*)[4]) store->LitColor[0].data;
+#if IDX & LIGHT_TWOSIDE
+   GLfloat (*Bcolor)[4] = (GLfloat (*)[4]) store->LitColor[1].data;
+#endif
+   GLuint j = 0;
+#if IDX & LIGHT_MATERIAL
+   const GLuint nr = VB->Count;
+#else
+   const GLuint nr = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->count;
+#endif
+   const struct gl_light *light;
+
+#ifdef TRACE
+   fprintf(stderr, "%s %d\n", __FUNCTION__, nr );
+#endif
+
+   (void) input;
+
+   sumA[0] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3];
+   sumA[1] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3];
+
+   VB->ColorPtr[0] = &store->LitColor[0];
+#if IDX & LIGHT_TWOSIDE
+   VB->ColorPtr[1] = &store->LitColor[1];
+#endif
+
+   if (nr > 1) {
+      store->LitColor[0].stride = 16;
+      store->LitColor[1].stride = 16;
+   }
+   else {
+      store->LitColor[0].stride = 0;
+      store->LitColor[1].stride = 0;
+   }
+
+   for (j = 0; j < nr; j++, STRIDE_F(normal,nstride)) {
+
+      GLfloat sum[2][3];
+
+#if IDX & LIGHT_MATERIAL
+      update_materials( ctx, store );
+
+      sumA[0] = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_DIFFUSE][3];
+#if IDX & LIGHT_TWOSIDE
+      sumA[1] = ctx->Light.Material.Attrib[MAT_ATTRIB_BACK_DIFFUSE][3];
+#endif
+#endif
+
+
+      COPY_3V(sum[0], ctx->Light._BaseColor[0]);
+#if IDX & LIGHT_TWOSIDE
+      COPY_3V(sum[1], ctx->Light._BaseColor[1]);
+#endif
+
+      foreach (light, &ctx->Light.EnabledList) {
+	 GLfloat n_dot_h, n_dot_VP, spec;
+
+	 ACC_3V(sum[0], light->_MatAmbient[0]);
+#if IDX & LIGHT_TWOSIDE
+         ACC_3V(sum[1], light->_MatAmbient[1]);
+#endif
+
+	 n_dot_VP = DOT3(normal, light->_VP_inf_norm);
+
+	 if (n_dot_VP > 0.0F) {
+	    ACC_SCALE_SCALAR_3V(sum[0], n_dot_VP, light->_MatDiffuse[0]);
+	    n_dot_h = DOT3(normal, light->_h_inf_norm);
+	    if (n_dot_h > 0.0F) {
+	       struct gl_shine_tab *tab = ctx->_ShineTable[0];
+	       GET_SHINE_TAB_ENTRY( tab, n_dot_h, spec );
+	       ACC_SCALE_SCALAR_3V( sum[0], spec, light->_MatSpecular[0]);
+	    }
+	 }
+#if IDX & LIGHT_TWOSIDE
+         else {
+	    ACC_SCALE_SCALAR_3V(sum[1], -n_dot_VP, light->_MatDiffuse[1]);
+	    n_dot_h = -DOT3(normal, light->_h_inf_norm);
+	    if (n_dot_h > 0.0F) {
+	       struct gl_shine_tab *tab = ctx->_ShineTable[1];
+	       GET_SHINE_TAB_ENTRY( tab, n_dot_h, spec );
+	       ACC_SCALE_SCALAR_3V( sum[1], spec, light->_MatSpecular[1]);
+	    }
+	 }
+#endif
+      }
+
+      COPY_3V( Fcolor[j], sum[0] );
+      Fcolor[j][3] = sumA[0];
+
+#if IDX & LIGHT_TWOSIDE
+      COPY_3V( Bcolor[j], sum[1] );
+      Bcolor[j][3] = sumA[1];
+#endif
+   }
+}
+
+
+
+
+
+/*
+ * Use current lighting/material settings to compute the color indexes
+ * for an array of vertices.
+ * Input:  n - number of vertices to light
+ *         side - 0=use front material, 1=use back material
+ *         vertex - array of [n] vertex position in eye coordinates
+ *         normal - array of [n] surface normal vector
+ * Output:  indexResult - resulting array of [n] color indexes
+ */
+static void TAG(light_ci)( GLcontext *ctx,
+			   struct vertex_buffer *VB,
+			   struct tnl_pipeline_stage *stage,
+			   GLvector4f *input )
+{
+   struct light_stage_data *store = LIGHT_STAGE_DATA(stage);
+   GLuint j;
+   const GLuint vstride = input->stride;
+   const GLfloat *vertex = (GLfloat *) input->data;
+   const GLuint nstride = VB->AttribPtr[_TNL_ATTRIB_NORMAL]->stride;
+   const GLfloat *normal = (GLfloat *)VB->AttribPtr[_TNL_ATTRIB_NORMAL]->data;
+   GLfloat *indexResult[2];
+   const GLuint nr = VB->Count;
+
+#ifdef TRACE
+   fprintf(stderr, "%s\n", __FUNCTION__ );
+#endif
+
+   VB->IndexPtr[0] = &store->LitIndex[0];
+#if IDX & LIGHT_TWOSIDE
+   VB->IndexPtr[1] = &store->LitIndex[1];
+#endif
+
+   indexResult[0] = (GLfloat *)VB->IndexPtr[0]->data;
+#if IDX & LIGHT_TWOSIDE
+   indexResult[1] = (GLfloat *)VB->IndexPtr[1]->data;
+#endif
+
+   /* loop over vertices */
+   for (j=0; j<nr; j++,STRIDE_F(vertex,vstride),STRIDE_F(normal, nstride)) {
+      GLfloat diffuse[2], specular[2];
+      GLuint side = 0;
+      struct gl_light *light;
+
+#if IDX & LIGHT_MATERIAL
+      update_materials( ctx, store );
+#endif
+
+      diffuse[0] = specular[0] = 0.0F;
+
+#if IDX & LIGHT_TWOSIDE
+	 diffuse[1] = specular[1] = 0.0F;
+#endif
+
+      /* Accumulate diffuse and specular from each light source */
+      foreach (light, &ctx->Light.EnabledList) {
+
+	 GLfloat attenuation = 1.0F;
+	 GLfloat VP[3];  /* unit vector from vertex to light */
+	 GLfloat n_dot_VP;  /* dot product of l and n */
+	 GLfloat *h, n_dot_h, correction = 1.0;
+
+	 /* compute l and attenuation */
+	 if (!(light->_Flags & LIGHT_POSITIONAL)) {
+	    /* directional light */
+	    COPY_3V(VP, light->_VP_inf_norm);
+	 }
+	 else {
+	    GLfloat d;     /* distance from vertex to light */
+
+	    SUB_3V(VP, light->_Position, vertex);
+
+	    d = (GLfloat) LEN_3FV( VP );
+	    if ( d > 1e-6) {
+	       GLfloat invd = 1.0F / d;
+	       SELF_SCALE_SCALAR_3V(VP, invd);
+	    }
+
+	    attenuation = 1.0F / (light->ConstantAttenuation + d *
+				  (light->LinearAttenuation + d *
+				   light->QuadraticAttenuation));
+
+	    /* spotlight attenuation */
+	    if (light->_Flags & LIGHT_SPOT) {
+	       GLfloat PV_dot_dir = - DOT3(VP, light->_NormSpotDirection);
+	       if (PV_dot_dir < light->_CosCutoff) {
+		  continue; /* this light makes no contribution */
+	       }
+	       else {
+		  GLdouble x = PV_dot_dir * (EXP_TABLE_SIZE-1);
+		  GLint k = (GLint) x;
+		  GLfloat spot = (GLfloat) (light->_SpotExpTable[k][0]
+				  + (x-k)*light->_SpotExpTable[k][1]);
+		  attenuation *= spot;
+	       }
+	    }
+	 }
+
+	 if (attenuation < 1e-3)
+	    continue;		/* this light makes no contribution */
+
+	 n_dot_VP = DOT3( normal, VP );
+
+	 /* which side are we lighting? */
+	 if (n_dot_VP < 0.0F) {
+#if IDX & LIGHT_TWOSIDE
+	    side = 1;
+	    correction = -1;
+	    n_dot_VP = -n_dot_VP;
+#else
+            continue;
+#endif
+	 }
+
+	 /* accumulate diffuse term */
+	 diffuse[side] += n_dot_VP * light->_dli * attenuation;
+
+	 /* specular term */
+	 if (ctx->Light.Model.LocalViewer) {
+	    GLfloat v[3];
+	    COPY_3V(v, vertex);
+	    NORMALIZE_3FV(v);
+	    SUB_3V(VP, VP, v);                /* h = VP + VPe */
+	    h = VP;
+	    NORMALIZE_3FV(h);
+	 }
+	 else if (light->_Flags & LIGHT_POSITIONAL) {
+	    h = VP;
+            /* Strangely, disabling this addition fixes a conformance
+             * problem.  If this code is enabled, l_sed.c fails.
+             */
+	    /*ACC_3V(h, ctx->_EyeZDir);*/
+	    NORMALIZE_3FV(h);
+	 }
+         else {
+	    h = light->_h_inf_norm;
+	 }
+
+	 n_dot_h = correction * DOT3(normal, h);
+	 if (n_dot_h > 0.0F) {
+	    GLfloat spec_coef;
+	    struct gl_shine_tab *tab = ctx->_ShineTable[side];
+	    GET_SHINE_TAB_ENTRY( tab, n_dot_h, spec_coef);
+	    specular[side] += spec_coef * light->_sli * attenuation;
+	 }
+      } /*loop over lights*/
+
+      /* Now compute final color index */
+      for (side = 0 ; side < NR_SIDES ; side++) {
+	 const GLfloat *ind = ctx->Light.Material.Attrib[MAT_ATTRIB_FRONT_INDEXES + side];
+	 GLfloat index;
+
+	 if (specular[side] > 1.0F) {
+	    index = ind[MAT_INDEX_SPECULAR];
+	 }
+	 else {
+	    GLfloat d_a = ind[MAT_INDEX_DIFFUSE] - ind[MAT_INDEX_AMBIENT];
+	    GLfloat s_a = ind[MAT_INDEX_SPECULAR] - ind[MAT_INDEX_AMBIENT];
+	    index = (ind[MAT_INDEX_AMBIENT]
+		     + diffuse[side] * (1.0F-specular[side]) * d_a
+		     + specular[side] * s_a);
+	    if (index > ind[MAT_INDEX_SPECULAR]) {
+	       index = ind[MAT_INDEX_SPECULAR];
+	    }
+	 }
+	 indexResult[side][j] = index;
+      }
+   } /*for vertex*/
+}
+
+
+
+static void TAG(init_light_tab)( void )
+{
+   _tnl_light_tab[IDX] = TAG(light_rgba);
+   _tnl_light_fast_tab[IDX] = TAG(light_fast_rgba);
+   _tnl_light_fast_single_tab[IDX] = TAG(light_fast_rgba_single);
+   _tnl_light_spec_tab[IDX] = TAG(light_rgba_spec);
+   _tnl_light_ci_tab[IDX] = TAG(light_ci);
+}
+
+
+#undef TAG
+#undef IDX
+#undef NR_SIDES
diff --git a/mesalib/src/mesa/tnl/t_vb_normals.c b/mesalib/src/mesa/tnl/t_vb_normals.c
new file mode 100644
index 000000000..a4821cc1c
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_normals.c
@@ -0,0 +1,189 @@
+/*
+ * 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 "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "math/m_xform.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+
+
+struct normal_stage_data {
+   normal_func NormalTransform;
+   GLvector4f normal;
+};
+
+#define NORMAL_STAGE_DATA(stage) ((struct normal_stage_data *)stage->privatePtr)
+
+
+static GLboolean
+run_normal_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage)
+{
+   struct normal_stage_data *store = NORMAL_STAGE_DATA(stage);
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   const GLfloat *lengths;
+
+   if (!store->NormalTransform)
+      return GL_TRUE;
+
+   /* We can only use the display list's saved normal lengths if we've
+    * got a transformation matrix with uniform scaling.
+    */
+   if (_math_matrix_is_general_scale(ctx->ModelviewMatrixStack.Top))
+      lengths = NULL;
+   else
+      lengths = VB->NormalLengthPtr;
+
+   store->NormalTransform( ctx->ModelviewMatrixStack.Top,
+			   ctx->_ModelViewInvScale,
+			   VB->AttribPtr[_TNL_ATTRIB_NORMAL],  /* input normals */
+			   lengths,
+			   &store->normal ); /* resulting normals */
+
+   if (VB->AttribPtr[_TNL_ATTRIB_NORMAL]->count > 1) {
+      store->normal.stride = 4 * sizeof(GLfloat);
+   }
+   else {
+      store->normal.stride = 0;
+   }
+
+   VB->AttribPtr[_TNL_ATTRIB_NORMAL] = &store->normal;
+   VB->NormalPtr = &store->normal;
+
+   VB->NormalLengthPtr = NULL;	/* no longer valid */
+   return GL_TRUE;
+}
+
+
+/**
+ * Examine current GL state and set the store->NormalTransform pointer
+ * to point to the appropriate normal transformation routine.
+ */
+static void
+validate_normal_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage)
+{
+   struct normal_stage_data *store = NORMAL_STAGE_DATA(stage);
+
+   if (ctx->VertexProgram._Current ||
+       (!ctx->Light.Enabled &&
+	!(ctx->Texture._GenFlags & TEXGEN_NEED_NORMALS))) {
+      store->NormalTransform = NULL;
+      return;
+   }
+
+   if (ctx->_NeedEyeCoords) {
+      /* Eye coordinates are needed, for whatever reasons.
+       * Do lighting in eye coordinates, as the GL spec says.
+       */
+      GLuint transform = NORM_TRANSFORM_NO_ROT;
+
+      if (_math_matrix_has_rotation(ctx->ModelviewMatrixStack.Top)) {
+         /* need to do full (3x3) matrix transform */
+	 transform = NORM_TRANSFORM;
+      }
+
+      if (ctx->Transform.Normalize) {
+	 store->NormalTransform = _mesa_normal_tab[transform | NORM_NORMALIZE];
+      }
+      else if (ctx->Transform.RescaleNormals &&
+               ctx->_ModelViewInvScale != 1.0) {
+	 store->NormalTransform = _mesa_normal_tab[transform | NORM_RESCALE];
+      }
+      else {
+	 store->NormalTransform = _mesa_normal_tab[transform];
+      }
+   }
+   else {
+      /* We don't need eye coordinates.
+       * Do lighting in object coordinates.  Thus, we don't need to fully
+       * transform normal vectors (just leave them in object coordinates)
+       * but we still need to do normalization/rescaling if enabled.
+       */
+      if (ctx->Transform.Normalize) {
+	 store->NormalTransform = _mesa_normal_tab[NORM_NORMALIZE];
+      }
+      else if (!ctx->Transform.RescaleNormals &&
+	       ctx->_ModelViewInvScale != 1.0) {
+	 store->NormalTransform = _mesa_normal_tab[NORM_RESCALE];
+      }
+      else {
+	 store->NormalTransform = NULL;
+      }
+   }
+}
+
+
+/**
+ * Allocate stage's private data (storage for transformed normals).
+ */
+static GLboolean
+alloc_normal_data(GLcontext *ctx, struct tnl_pipeline_stage *stage)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct normal_stage_data *store;
+
+   stage->privatePtr = _mesa_malloc(sizeof(*store));
+   store = NORMAL_STAGE_DATA(stage);
+   if (!store)
+      return GL_FALSE;
+
+   _mesa_vector4f_alloc( &store->normal, 0, tnl->vb.Size, 32 );
+   return GL_TRUE;
+}
+
+
+/**
+ * Free stage's private data.
+ */
+static void
+free_normal_data(struct tnl_pipeline_stage *stage)
+{
+   struct normal_stage_data *store = NORMAL_STAGE_DATA(stage);
+   if (store) {
+      _mesa_vector4f_free( &store->normal );
+      _mesa_free( store );
+      stage->privatePtr = NULL;
+   }
+}
+
+
+const struct tnl_pipeline_stage _tnl_normal_transform_stage =
+{
+   "normal transform",		/* name */
+   NULL,			/* privatePtr */
+   alloc_normal_data,		/* create */
+   free_normal_data,		/* destroy */
+   validate_normal_stage,	/* validate */
+   run_normal_stage             /* run */
+};
diff --git a/mesalib/src/mesa/tnl/t_vb_points.c b/mesalib/src/mesa/tnl/t_vb_points.c
new file mode 100644
index 000000000..a52505b4b
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_points.c
@@ -0,0 +1,114 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.0
+ *
+ * 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"),
+ * 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:
+ *    Brian Paul
+ */
+
+#include "main/glheader.h"
+#include "main/mtypes.h"
+#include "main/dd.h"
+#include "main/imports.h"
+#include "t_context.h"
+#include "t_pipeline.h"
+
+
+struct point_stage_data {
+   GLvector4f PointSize;
+};
+
+#define POINT_STAGE_DATA(stage) ((struct point_stage_data *)stage->privatePtr)
+
+
+/**
+ * Compute point size for each vertex from the vertex eye-space Z
+ * coordinate and the point size attenuation factors.
+ * Only done when point size attenuation is enabled and vertex program is
+ * disabled.
+ */
+static GLboolean
+run_point_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage)
+{
+   if (ctx->Point._Attenuated && !ctx->VertexProgram._Current) {
+      struct point_stage_data *store = POINT_STAGE_DATA(stage);
+      struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+      const GLfloat *eyeCoord = (GLfloat *) VB->EyePtr->data + 2;
+      const GLint eyeCoordStride = VB->EyePtr->stride / sizeof(GLfloat);
+      const GLfloat p0 = ctx->Point.Params[0];
+      const GLfloat p1 = ctx->Point.Params[1];
+      const GLfloat p2 = ctx->Point.Params[2];
+      const GLfloat pointSize = ctx->Point.Size;
+      GLfloat (*size)[4] = store->PointSize.data;
+      GLuint i;
+
+      for (i = 0; i < VB->Count; i++) {
+         const GLfloat dist = FABSF(*eyeCoord);
+         const GLfloat q = p0 + dist * (p1 + dist * p2);
+         const GLfloat atten = (q != 0.0) ? SQRTF(1.0 / q) : 1.0;
+         size[i][0] = pointSize * atten; /* clamping done in rasterization */
+         eyeCoord += eyeCoordStride;
+      }
+
+      VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->PointSize;
+   }
+
+   return GL_TRUE;
+}
+
+
+static GLboolean
+alloc_point_data(GLcontext *ctx, struct tnl_pipeline_stage *stage)
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct point_stage_data *store;
+   stage->privatePtr = _mesa_malloc(sizeof(*store));
+   store = POINT_STAGE_DATA(stage);
+   if (!store)
+      return GL_FALSE;
+
+   _mesa_vector4f_alloc( &store->PointSize, 0, VB->Size, 32 );
+   return GL_TRUE;
+}
+
+
+static void
+free_point_data(struct tnl_pipeline_stage *stage)
+{
+   struct point_stage_data *store = POINT_STAGE_DATA(stage);
+   if (store) {
+      _mesa_vector4f_free( &store->PointSize );
+      _mesa_free( store );
+      stage->privatePtr = NULL;
+   }
+}
+
+
+const struct tnl_pipeline_stage _tnl_point_attenuation_stage =
+{
+   "point size attenuation",	/* name */
+   NULL,			/* stage private data */
+   alloc_point_data,		/* alloc data */
+   free_point_data,		/* destructor */
+   NULL,
+   run_point_stage		/* run */
+};
diff --git a/mesalib/src/mesa/tnl/t_vb_program.c b/mesalib/src/mesa/tnl/t_vb_program.c
new file mode 100644
index 000000000..dc954bcba
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_program.c
@@ -0,0 +1,563 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.6
+ *
+ * Copyright (C) 1999-2008  Brian Paul   All Rights Reserved.
+ * Copyright (C) 2009  VMware, Inc.  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 tnl/t_vb_program.c
+ * \brief Pipeline stage for executing vertex programs.
+ * \author Brian Paul,  Keith Whitwell
+ */
+
+
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "shader/prog_instruction.h"
+#include "shader/prog_statevars.h"
+#include "shader/prog_execute.h"
+#include "swrast/s_context.h"
+#include "swrast/s_texfilter.h"
+
+#include "tnl/tnl.h"
+#include "tnl/t_context.h"
+#include "tnl/t_pipeline.h"
+
+
+#ifdef NAN_CHECK
+/** Check for NaNs and very large values */
+static INLINE void
+check_float(float x)
+{
+   assert(!IS_INF_OR_NAN(x));
+   assert(1.0e-15 <= x && x <= 1.0e15);
+}
+#endif
+
+
+/*!
+ * Private storage for the vertex program pipeline stage.
+ */
+struct vp_stage_data {
+   /** The results of running the vertex program go into these arrays. */
+   GLvector4f results[VERT_RESULT_MAX];
+
+   GLvector4f ndcCoords;              /**< normalized device coords */
+   GLubyte *clipmask;                 /**< clip flags */
+   GLubyte ormask, andmask;           /**< for clipping */
+};
+
+
+#define VP_STAGE_DATA(stage) ((struct vp_stage_data *)(stage->privatePtr))
+
+
+static void
+userclip( GLcontext *ctx,
+          GLvector4f *clip,
+          GLubyte *clipmask,
+          GLubyte *clipormask,
+          GLubyte *clipandmask )
+{
+   GLuint p;
+
+   for (p = 0; p < ctx->Const.MaxClipPlanes; p++) {
+      if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {
+	 GLuint nr, i;
+	 const GLfloat a = ctx->Transform._ClipUserPlane[p][0];
+	 const GLfloat b = ctx->Transform._ClipUserPlane[p][1];
+	 const GLfloat c = ctx->Transform._ClipUserPlane[p][2];
+	 const GLfloat d = ctx->Transform._ClipUserPlane[p][3];
+         GLfloat *coord = (GLfloat *)clip->data;
+         GLuint stride = clip->stride;
+         GLuint count = clip->count;
+
+	 for (nr = 0, i = 0 ; i < count ; i++) {
+	    GLfloat dp = (coord[0] * a + 
+			  coord[1] * b +
+			  coord[2] * c +
+			  coord[3] * d);
+
+	    if (dp < 0) {
+	       nr++;
+	       clipmask[i] |= CLIP_USER_BIT;
+	    }
+
+	    STRIDE_F(coord, stride);
+	 }
+
+	 if (nr > 0) {
+	    *clipormask |= CLIP_USER_BIT;
+	    if (nr == count) {
+	       *clipandmask |= CLIP_USER_BIT;
+	       return;
+	    }
+	 }
+      }
+   }
+}
+
+
+static GLboolean
+do_ndc_cliptest(GLcontext *ctx, struct vp_stage_data *store)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   /* Cliptest and perspective divide.  Clip functions must clear
+    * the clipmask.
+    */
+   store->ormask = 0;
+   store->andmask = CLIP_FRUSTUM_BITS;
+
+   if (tnl->NeedNdcCoords) {
+      VB->NdcPtr =
+         _mesa_clip_tab[VB->ClipPtr->size]( VB->ClipPtr,
+                                            &store->ndcCoords,
+                                            store->clipmask,
+                                            &store->ormask,
+                                            &store->andmask );
+   }
+   else {
+      VB->NdcPtr = NULL;
+      _mesa_clip_np_tab[VB->ClipPtr->size]( VB->ClipPtr,
+                                            NULL,
+                                            store->clipmask,
+                                            &store->ormask,
+                                            &store->andmask );
+   }
+
+   if (store->andmask) {
+      /* All vertices are outside the frustum */
+      return GL_FALSE;
+   }
+
+   /* Test userclip planes.  This contributes to VB->ClipMask.
+    */
+   /** XXX NEW_SLANG _Enabled ??? */
+   if (ctx->Transform.ClipPlanesEnabled && (!ctx->VertexProgram._Enabled ||
+      ctx->VertexProgram.Current->IsPositionInvariant)) {
+      userclip( ctx,
+		VB->ClipPtr,
+		store->clipmask,
+		&store->ormask,
+		&store->andmask );
+
+      if (store->andmask) {
+	 return GL_FALSE;
+      }
+   }
+
+   VB->ClipAndMask = store->andmask;
+   VB->ClipOrMask = store->ormask;
+   VB->ClipMask = store->clipmask;
+
+   return GL_TRUE;
+}
+
+
+/**
+ * XXX the texture sampling code in this module is a bit of a hack.
+ * The texture sampling code is in swrast, though it doesn't have any
+ * real dependencies on the rest of swrast.  It should probably be
+ * moved into main/ someday.
+ */
+static void
+vp_fetch_texel(GLcontext *ctx, const GLfloat texcoord[4], GLfloat lambda,
+               GLuint unit, GLfloat color[4])
+{
+   SWcontext *swrast = SWRAST_CONTEXT(ctx);
+
+   /* XXX use a float-valued TextureSample routine here!!! */
+   swrast->TextureSample[unit](ctx, ctx->Texture.Unit[unit]._Current,
+                               1, (const GLfloat (*)[4]) texcoord,
+                               &lambda,  (GLfloat (*)[4]) color);
+}
+
+
+/**
+ * Called via ctx->Driver.ProgramStringNotify() after a new vertex program
+ * string has been parsed.
+ */
+void
+_tnl_program_string(GLcontext *ctx, GLenum target, struct gl_program *program)
+{
+   /* No-op.
+    * If we had derived anything from the program that was private to this
+    * stage we'd recompute/validate it here.
+    */
+}
+
+
+/**
+ * Initialize virtual machine state prior to executing vertex program.
+ */
+static void
+init_machine(GLcontext *ctx, struct gl_program_machine *machine)
+{
+   /* Input registers get initialized from the current vertex attribs */
+   MEMCPY(machine->VertAttribs, ctx->Current.Attrib,
+          MAX_VERTEX_GENERIC_ATTRIBS * 4 * sizeof(GLfloat));
+
+   if (ctx->VertexProgram._Current->IsNVProgram) {
+      GLuint i;
+      /* Output/result regs are initialized to [0,0,0,1] */
+      for (i = 0; i < MAX_NV_VERTEX_PROGRAM_OUTPUTS; i++) {
+         ASSIGN_4V(machine->Outputs[i], 0.0F, 0.0F, 0.0F, 1.0F);
+      }
+      /* Temp regs are initialized to [0,0,0,0] */
+      for (i = 0; i < MAX_NV_VERTEX_PROGRAM_TEMPS; i++) {
+         ASSIGN_4V(machine->Temporaries[i], 0.0F, 0.0F, 0.0F, 0.0F);
+      }
+      for (i = 0; i < MAX_VERTEX_PROGRAM_ADDRESS_REGS; i++) {
+         ASSIGN_4V(machine->AddressReg[i], 0, 0, 0, 0);
+      }
+   }
+
+   machine->NumDeriv = 0;
+
+   /* init condition codes */
+   machine->CondCodes[0] = COND_EQ;
+   machine->CondCodes[1] = COND_EQ;
+   machine->CondCodes[2] = COND_EQ;
+   machine->CondCodes[3] = COND_EQ;
+
+   /* init call stack */
+   machine->StackDepth = 0;
+
+   machine->FetchTexelLod = vp_fetch_texel;
+   machine->FetchTexelDeriv = NULL; /* not used by vertex programs */
+
+   machine->Samplers = ctx->VertexProgram._Current->Base.SamplerUnits;
+}
+
+
+/**
+ * Map the texture images which the vertex program will access (if any).
+ */
+static void
+map_textures(GLcontext *ctx, const struct gl_vertex_program *vp)
+{
+   GLuint u;
+
+   if (!ctx->Driver.MapTexture)
+      return;
+
+   for (u = 0; u < ctx->Const.MaxVertexTextureImageUnits; u++) {
+      if (vp->Base.TexturesUsed[u]) {
+         /* Note: _Current *should* correspond to the target indicated
+          * in TexturesUsed[u].
+          */
+         ctx->Driver.MapTexture(ctx, ctx->Texture.Unit[u]._Current);
+      }
+   }
+}
+
+
+/**
+ * Unmap the texture images which were used by the vertex program (if any).
+ */
+static void
+unmap_textures(GLcontext *ctx, const struct gl_vertex_program *vp)
+{
+   GLuint u;
+
+   if (!ctx->Driver.MapTexture)
+      return;
+
+   for (u = 0; u < ctx->Const.MaxVertexTextureImageUnits; u++) {
+      if (vp->Base.TexturesUsed[u]) {
+         /* Note: _Current *should* correspond to the target indicated
+          * in TexturesUsed[u].
+          */
+         ctx->Driver.UnmapTexture(ctx, ctx->Texture.Unit[u]._Current);
+      }
+   }
+}
+
+
+/**
+ * This function executes vertex programs
+ */
+static GLboolean
+run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vp_stage_data *store = VP_STAGE_DATA(stage);
+   struct vertex_buffer *VB = &tnl->vb;
+   struct gl_vertex_program *program = ctx->VertexProgram._Current;
+   struct gl_program_machine machine;
+   GLuint outputs[VERT_RESULT_MAX], numOutputs;
+   GLuint i, j;
+
+   if (!program)
+      return GL_TRUE;
+
+   if (program->IsNVProgram) {
+      _mesa_load_tracked_matrices(ctx);
+   }
+   else {
+      /* ARB program or vertex shader */
+      _mesa_load_state_parameters(ctx, program->Base.Parameters);
+   }
+
+   /* make list of outputs to save some time below */
+   numOutputs = 0;
+   for (i = 0; i < VERT_RESULT_MAX; i++) {
+      if (program->Base.OutputsWritten & (1 << i)) {
+         outputs[numOutputs++] = i;
+      }
+   }
+
+   map_textures(ctx, program);
+
+   for (i = 0; i < VB->Count; i++) {
+      GLuint attr;
+
+      init_machine(ctx, &machine);
+
+#if 0
+      printf("Input  %d: %f, %f, %f, %f\n", i,
+             VB->AttribPtr[0]->data[i][0],
+             VB->AttribPtr[0]->data[i][1],
+             VB->AttribPtr[0]->data[i][2],
+             VB->AttribPtr[0]->data[i][3]);
+      printf("   color: %f, %f, %f, %f\n",
+             VB->AttribPtr[3]->data[i][0],
+             VB->AttribPtr[3]->data[i][1],
+             VB->AttribPtr[3]->data[i][2],
+             VB->AttribPtr[3]->data[i][3]);
+      printf("  normal: %f, %f, %f, %f\n",
+             VB->AttribPtr[2]->data[i][0],
+             VB->AttribPtr[2]->data[i][1],
+             VB->AttribPtr[2]->data[i][2],
+             VB->AttribPtr[2]->data[i][3]);
+#endif
+
+      /* the vertex array case */
+      for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
+	 if (program->Base.InputsRead & (1 << attr)) {
+	    const GLubyte *ptr = (const GLubyte*) VB->AttribPtr[attr]->data;
+	    const GLuint size = VB->AttribPtr[attr]->size;
+	    const GLuint stride = VB->AttribPtr[attr]->stride;
+	    const GLfloat *data = (GLfloat *) (ptr + stride * i);
+#ifdef NAN_CHECK
+            check_float(data[0]);
+            check_float(data[1]);
+            check_float(data[2]);
+            check_float(data[3]);
+#endif
+	    COPY_CLEAN_4V(machine.VertAttribs[attr], size, data);
+	 }
+      }
+
+      /* execute the program */
+      _mesa_execute_program(ctx, &program->Base, &machine);
+
+      /* copy the output registers into the VB->attribs arrays */
+      for (j = 0; j < numOutputs; j++) {
+         const GLuint attr = outputs[j];
+#ifdef NAN_CHECK
+         check_float(machine.Outputs[attr][0]);
+         check_float(machine.Outputs[attr][1]);
+         check_float(machine.Outputs[attr][2]);
+         check_float(machine.Outputs[attr][3]);
+#endif
+         COPY_4V(store->results[attr].data[i], machine.Outputs[attr]);
+      }
+#ifdef NAN_CHECK
+      ASSERT(machine.Outputs[0][3] != 0.0F);
+#endif
+#if 0
+      printf("HPOS: %f %f %f %f\n",
+             machine.Outputs[0][0], 
+             machine.Outputs[0][1], 
+             machine.Outputs[0][2], 
+             machine.Outputs[0][3]);
+#endif
+   }
+
+   unmap_textures(ctx, program);
+
+   /* Fixup fog and point size results if needed */
+   if (program->IsNVProgram) {
+      if (ctx->Fog.Enabled &&
+          (program->Base.OutputsWritten & (1 << VERT_RESULT_FOGC)) == 0) {
+         for (i = 0; i < VB->Count; i++) {
+            store->results[VERT_RESULT_FOGC].data[i][0] = 1.0;
+         }
+      }
+
+      if (ctx->VertexProgram.PointSizeEnabled &&
+          (program->Base.OutputsWritten & (1 << VERT_RESULT_PSIZ)) == 0) {
+         for (i = 0; i < VB->Count; i++) {
+            store->results[VERT_RESULT_PSIZ].data[i][0] = ctx->Point.Size;
+         }
+      }
+   }
+
+   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( &store->results[0],
+				  &ctx->_ModelProjectMatrix,
+				  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 {
+      /* Setup the VB pointers so that the next pipeline stages get
+       * their data from the right place (the program output arrays).
+       */
+      VB->ClipPtr = &store->results[VERT_RESULT_HPOS];
+      VB->ClipPtr->size = 4;
+      VB->ClipPtr->count = VB->Count;
+   }
+
+   VB->ColorPtr[0] = &store->results[VERT_RESULT_COL0];
+   VB->ColorPtr[1] = &store->results[VERT_RESULT_BFC0];
+   VB->SecondaryColorPtr[0] = &store->results[VERT_RESULT_COL1];
+   VB->SecondaryColorPtr[1] = &store->results[VERT_RESULT_BFC1];
+   VB->FogCoordPtr = &store->results[VERT_RESULT_FOGC];
+
+   VB->AttribPtr[VERT_ATTRIB_COLOR0] = &store->results[VERT_RESULT_COL0];
+   VB->AttribPtr[VERT_ATTRIB_COLOR1] = &store->results[VERT_RESULT_COL1];
+   VB->AttribPtr[VERT_ATTRIB_FOG] = &store->results[VERT_RESULT_FOGC];
+   VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->results[VERT_RESULT_PSIZ];
+
+   for (i = 0; i < ctx->Const.MaxTextureCoordUnits; i++) {
+      VB->TexCoordPtr[i] = 
+      VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]
+         = &store->results[VERT_RESULT_TEX0 + i];
+   }
+
+   for (i = 0; i < ctx->Const.MaxVarying; i++) {
+      if (program->Base.OutputsWritten & (1 << (VERT_RESULT_VAR0 + i))) {
+         /* Note: varying results get put into the generic attributes */
+	 VB->AttribPtr[VERT_ATTRIB_GENERIC0+i]
+            = &store->results[VERT_RESULT_VAR0 + i];
+      }
+   }
+
+
+   /* Perform NDC and cliptest operations:
+    */
+   return do_ndc_cliptest(ctx, store);
+}
+
+
+/**
+ * Called the first time stage->run is called.  In effect, don't
+ * allocate data until the first time the stage is run.
+ */
+static GLboolean
+init_vp(GLcontext *ctx, struct tnl_pipeline_stage *stage)
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &(tnl->vb);
+   struct vp_stage_data *store;
+   const GLuint size = VB->Size;
+   GLuint i;
+
+   stage->privatePtr = MALLOC(sizeof(*store));
+   store = VP_STAGE_DATA(stage);
+   if (!store)
+      return GL_FALSE;
+
+   /* Allocate arrays of vertex output values */
+   for (i = 0; i < VERT_RESULT_MAX; i++) {
+      _mesa_vector4f_alloc( &store->results[i], 0, size, 32 );
+      store->results[i].size = 4;
+   }
+
+   /* a few other misc allocations */
+   _mesa_vector4f_alloc( &store->ndcCoords, 0, size, 32 );
+   store->clipmask = (GLubyte *) ALIGN_MALLOC(sizeof(GLubyte)*size, 32 );
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Destructor for this pipeline stage.
+ */
+static void
+dtr(struct tnl_pipeline_stage *stage)
+{
+   struct vp_stage_data *store = VP_STAGE_DATA(stage);
+
+   if (store) {
+      GLuint i;
+
+      /* free the vertex program result arrays */
+      for (i = 0; i < VERT_RESULT_MAX; i++)
+         _mesa_vector4f_free( &store->results[i] );
+
+      /* free misc arrays */
+      _mesa_vector4f_free( &store->ndcCoords );
+      ALIGN_FREE( store->clipmask );
+
+      FREE( store );
+      stage->privatePtr = NULL;
+   }
+}
+
+
+static void
+validate_vp_stage(GLcontext *ctx, struct tnl_pipeline_stage *stage)
+{
+   if (ctx->VertexProgram._Current) {
+      _swrast_update_texture_samplers(ctx);
+   }
+}
+
+
+
+/**
+ * Public description of this pipeline stage.
+ */
+const struct tnl_pipeline_stage _tnl_vertex_program_stage =
+{
+   "vertex-program",
+   NULL,			/* private_data */
+   init_vp,			/* create */
+   dtr,				/* destroy */
+   validate_vp_stage, 		/* validate */
+   run_vp			/* run -- initially set to ctr */
+};
diff --git a/mesalib/src/mesa/tnl/t_vb_render.c b/mesalib/src/mesa/tnl/t_vb_render.c
new file mode 100644
index 000000000..c1bebc994
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_render.c
@@ -0,0 +1,347 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * 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>
+ */
+
+
+/*
+ * Render whole vertex buffers, including projection of vertices from
+ * clip space and clipping of primitives.
+ *
+ * This file makes calls to project vertices and to the point, line
+ * and triangle rasterizers via the function pointers:
+ *
+ *    context->Driver.Render.*
+ *
+ */
+
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/enums.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "t_pipeline.h"
+
+
+
+/**********************************************************************/
+/*                        Clip single primitives                      */
+/**********************************************************************/
+
+
+#define W(i) coord[i][3]
+#define Z(i) coord[i][2]
+#define Y(i) coord[i][1]
+#define X(i) coord[i][0]
+#define SIZE 4
+#define TAG(x) x##_4
+#include "t_vb_cliptmp.h"
+
+
+
+/**********************************************************************/
+/*              Clip and render whole begin/end objects               */
+/**********************************************************************/
+
+#define NEED_EDGEFLAG_SETUP (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL)
+#define EDGEFLAG_GET(idx) VB->EdgeFlag[idx]
+#define EDGEFLAG_SET(idx, val) VB->EdgeFlag[idx] = val
+
+
+/* This does NOT include the CLIP_USER_BIT! */
+#define CLIPMASK (CLIP_FRUSTUM_BITS | CLIP_CULL_BIT)
+
+
+/* Vertices, with the possibility of clipping.
+ */
+#define RENDER_POINTS( start, count ) \
+   tnl->Driver.Render.Points( ctx, start, count )
+
+#define RENDER_LINE( v1, v2 )			\
+do {						\
+   GLubyte c1 = mask[v1], c2 = mask[v2];	\
+   GLubyte ormask = c1|c2;			\
+   if (!ormask)					\
+      LineFunc( ctx, v1, v2 );			\
+   else if (!(c1 & c2 & CLIPMASK))			\
+      clip_line_4( ctx, v1, v2, ormask );	\
+} while (0)
+
+#define RENDER_TRI( v1, v2, v3 )			\
+do {							\
+   GLubyte c1 = mask[v1], c2 = mask[v2], c3 = mask[v3];	\
+   GLubyte ormask = c1|c2|c3;				\
+   if (!ormask)						\
+      TriangleFunc( ctx, v1, v2, v3 );			\
+   else if (!(c1 & c2 & c3 & CLIPMASK)) 			\
+      clip_tri_4( ctx, v1, v2, v3, ormask );    	\
+} while (0)
+
+#define RENDER_QUAD( v1, v2, v3, v4 )			\
+do {							\
+   GLubyte c1 = mask[v1], c2 = mask[v2];		\
+   GLubyte c3 = mask[v3], c4 = mask[v4];		\
+   GLubyte ormask = c1|c2|c3|c4;			\
+   if (!ormask)						\
+      QuadFunc( ctx, v1, v2, v3, v4 );			\
+   else if (!(c1 & c2 & c3 & c4 & CLIPMASK)) 		\
+      clip_quad_4( ctx, v1, v2, v3, v4, ormask );	\
+} while (0)
+
+
+#define LOCAL_VARS						\
+   TNLcontext *tnl = TNL_CONTEXT(ctx);				\
+   struct vertex_buffer *VB = &tnl->vb;				\
+   const GLuint * const elt = VB->Elts;				\
+   const GLubyte *mask = VB->ClipMask;				\
+   const GLuint sz = VB->ClipPtr->size;				\
+   const tnl_line_func LineFunc = tnl->Driver.Render.Line;		\
+   const tnl_triangle_func TriangleFunc = tnl->Driver.Render.Triangle;	\
+   const tnl_quad_func QuadFunc = tnl->Driver.Render.Quad;		\
+   const GLboolean stipple = ctx->Line.StippleFlag;		\
+   (void) (LineFunc && TriangleFunc && QuadFunc);		\
+   (void) elt; (void) mask; (void) sz; (void) stipple;
+
+#define TAG(x) clip_##x##_verts
+#define INIT(x) tnl->Driver.Render.PrimitiveNotify( ctx, x )
+#define RESET_STIPPLE if (stipple) tnl->Driver.Render.ResetLineStipple( ctx )
+#define PRESERVE_VB_DEFS
+#include "t_vb_rendertmp.h"
+
+
+
+/* Elts, with the possibility of clipping.
+ */
+#undef ELT
+#undef TAG
+#define ELT(x) elt[x]
+#define TAG(x) clip_##x##_elts
+#include "t_vb_rendertmp.h"
+
+/* TODO: do this for all primitives, verts and elts:
+ */
+static void clip_elt_triangles( GLcontext *ctx,
+				GLuint start,
+				GLuint count,
+				GLuint flags )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   tnl_render_func render_tris = tnl->Driver.Render.PrimTabElts[GL_TRIANGLES];
+   struct vertex_buffer *VB = &tnl->vb;
+   const GLuint * const elt = VB->Elts;
+   GLubyte *mask = VB->ClipMask;
+   GLuint last = count-2;
+   GLuint j;
+   (void) flags;
+
+   tnl->Driver.Render.PrimitiveNotify( ctx, GL_TRIANGLES );
+
+   for (j=start; j < last; j+=3 ) {
+      GLubyte c1 = mask[elt[j]];
+      GLubyte c2 = mask[elt[j+1]];
+      GLubyte c3 = mask[elt[j+2]];
+      GLubyte ormask = c1|c2|c3;
+      if (ormask) {
+	 if (start < j)
+	    render_tris( ctx, start, j, 0 );
+	 if (!(c1&c2&c3&CLIPMASK))
+	    clip_tri_4( ctx, elt[j], elt[j+1], elt[j+2], ormask );
+	 start = j+3;
+      }
+   }
+
+   if (start < j)
+      render_tris( ctx, start, j, 0 );
+}
+
+/**********************************************************************/
+/*                  Render whole begin/end objects                    */
+/**********************************************************************/
+
+#define NEED_EDGEFLAG_SETUP (ctx->Polygon.FrontMode != GL_FILL || ctx->Polygon.BackMode != GL_FILL)
+#define EDGEFLAG_GET(idx) VB->EdgeFlag[idx]
+#define EDGEFLAG_SET(idx, val) VB->EdgeFlag[idx] = val
+
+
+/* Vertices, no clipping.
+ */
+#define RENDER_POINTS( start, count ) \
+   tnl->Driver.Render.Points( ctx, start, count )
+
+#define RENDER_LINE( v1, v2 ) \
+   LineFunc( ctx, v1, v2 )
+
+#define RENDER_TRI( v1, v2, v3 ) \
+   TriangleFunc( ctx, v1, v2, v3 )
+
+#define RENDER_QUAD( v1, v2, v3, v4 ) \
+   QuadFunc( ctx, v1, v2, v3, v4 )
+
+#define TAG(x) _tnl_##x##_verts
+
+#define LOCAL_VARS						\
+   TNLcontext *tnl = TNL_CONTEXT(ctx);				\
+   struct vertex_buffer *VB = &tnl->vb;				\
+   const GLuint * const elt = VB->Elts;				\
+   const tnl_line_func LineFunc = tnl->Driver.Render.Line;		\
+   const tnl_triangle_func TriangleFunc = tnl->Driver.Render.Triangle;	\
+   const tnl_quad_func QuadFunc = tnl->Driver.Render.Quad;		\
+   const GLboolean stipple = ctx->Line.StippleFlag;		\
+   (void) (LineFunc && TriangleFunc && QuadFunc);		\
+   (void) elt; (void) stipple
+
+#define RESET_STIPPLE if (stipple) tnl->Driver.Render.ResetLineStipple( ctx )
+#define INIT(x) tnl->Driver.Render.PrimitiveNotify( ctx, x )
+#define RENDER_TAB_QUALIFIER
+#define PRESERVE_VB_DEFS
+#include "t_vb_rendertmp.h"
+
+
+/* Elts, no clipping.
+ */
+#undef ELT
+#define TAG(x) _tnl_##x##_elts
+#define ELT(x) elt[x]
+#include "t_vb_rendertmp.h"
+
+
+/**********************************************************************/
+/*              Helper functions for drivers                  */
+/**********************************************************************/
+
+void _tnl_RenderClippedPolygon( 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;
+   tnl->Driver.Render.PrimTabElts[GL_POLYGON]( ctx, 0, n, PRIM_BEGIN|PRIM_END );
+   VB->Elts = tmp;
+}
+
+void _tnl_RenderClippedLine( GLcontext *ctx, GLuint ii, GLuint jj )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   tnl->Driver.Render.Line( ctx, ii, jj );
+}
+
+
+
+/**********************************************************************/
+/*              Clip and render whole vertex buffers                  */
+/**********************************************************************/
+
+
+static GLboolean run_render( GLcontext *ctx,
+			     struct tnl_pipeline_stage *stage )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   tnl_render_func *tab;
+   GLint pass = 0;
+
+   /* Allow the drivers to lock before projected verts are built so
+    * that window coordinates are guarenteed not to change before
+    * rendering.
+    */
+   ASSERT(tnl->Driver.Render.Start);
+
+   tnl->Driver.Render.Start( ctx );
+
+   ASSERT(tnl->Driver.Render.BuildVertices);
+   ASSERT(tnl->Driver.Render.PrimitiveNotify);
+   ASSERT(tnl->Driver.Render.Points);
+   ASSERT(tnl->Driver.Render.Line);
+   ASSERT(tnl->Driver.Render.Triangle);
+   ASSERT(tnl->Driver.Render.Quad);
+   ASSERT(tnl->Driver.Render.ResetLineStipple);
+   ASSERT(tnl->Driver.Render.Interp);
+   ASSERT(tnl->Driver.Render.CopyPV);
+   ASSERT(tnl->Driver.Render.ClippedLine);
+   ASSERT(tnl->Driver.Render.ClippedPolygon);
+   ASSERT(tnl->Driver.Render.Finish);
+
+   tnl->Driver.Render.BuildVertices( ctx, 0, VB->Count, ~0 );
+
+   if (VB->ClipOrMask) {
+      tab = VB->Elts ? clip_render_tab_elts : clip_render_tab_verts;
+      clip_render_tab_elts[GL_TRIANGLES] = clip_elt_triangles;
+   }
+   else {
+      tab = (VB->Elts ? 
+	     tnl->Driver.Render.PrimTabElts : 
+	     tnl->Driver.Render.PrimTabVerts);
+   }
+
+   do
+   {
+      GLuint i;
+
+      for (i = 0 ; i < VB->PrimitiveCount ; i++)
+      {
+	 GLuint prim = _tnl_translate_prim(&VB->Primitive[i]);
+	 GLuint start = VB->Primitive[i].start;
+	 GLuint length = VB->Primitive[i].count;
+
+	 assert((prim & PRIM_MODE_MASK) <= GL_POLYGON);
+
+	 if (MESA_VERBOSE & VERBOSE_PRIMS) 
+	    _mesa_debug(NULL, "MESA prim %s %d..%d\n", 
+			_mesa_lookup_enum_by_nr(prim & PRIM_MODE_MASK), 
+			start, start+length);
+
+	 if (length)
+	    tab[prim & PRIM_MODE_MASK]( ctx, start, start + length, prim );
+      }
+   } while (tnl->Driver.Render.Multipass &&
+	    tnl->Driver.Render.Multipass( ctx, ++pass ));
+
+   tnl->Driver.Render.Finish( ctx );
+
+   return GL_FALSE;		/* finished the pipe */
+}
+
+
+/**********************************************************************/
+/*                          Render pipeline stage                     */
+/**********************************************************************/
+
+
+
+
+
+const struct tnl_pipeline_stage _tnl_render_stage =
+{
+   "render",			/* name */
+   NULL,			/* private data */
+   NULL,			/* creator */
+   NULL,			/* destructor */
+   NULL,			/* validate */
+   run_render			/* run */
+};
diff --git a/mesalib/src/mesa/tnl/t_vb_rendertmp.h b/mesalib/src/mesa/tnl/t_vb_rendertmp.h
new file mode 100644
index 000000000..75f6f55bd
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_rendertmp.h
@@ -0,0 +1,486 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.5
+ *
+ * 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>
+ */
+
+
+#ifndef POSTFIX
+#define POSTFIX
+#endif
+
+#ifndef INIT
+#define INIT(x)
+#endif
+
+#ifndef NEED_EDGEFLAG_SETUP
+#define NEED_EDGEFLAG_SETUP 0
+#define EDGEFLAG_GET(a) 0
+#define EDGEFLAG_SET(a,b) (void)b
+#endif
+
+#ifndef RESET_STIPPLE
+#define RESET_STIPPLE
+#endif
+
+#ifndef TEST_PRIM_END
+#define TEST_PRIM_END(prim) (flags & PRIM_END)
+#define TEST_PRIM_BEGIN(prim) (flags & PRIM_BEGIN)
+#endif
+
+#ifndef ELT
+#define ELT(x) x
+#endif
+
+#ifndef RENDER_TAB_QUALIFIER
+#define RENDER_TAB_QUALIFIER static
+#endif
+
+static void TAG(render_points)( GLcontext *ctx,
+				GLuint start,
+				GLuint count,
+				GLuint flags )
+{
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_POINTS);
+   RENDER_POINTS( start, count );
+   POSTFIX;
+}
+
+static void TAG(render_lines)( GLcontext *ctx,
+			       GLuint start,
+			       GLuint count,
+			       GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_LINES);
+   for (j=start+1; j<count; j+=2 ) {
+      RESET_STIPPLE;
+      if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+         RENDER_LINE( ELT(j-1), ELT(j) );
+      else
+         RENDER_LINE( ELT(j), ELT(j-1) );
+   }
+   POSTFIX;
+}
+
+
+static void TAG(render_line_strip)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_LINE_STRIP);
+
+   if (TEST_PRIM_BEGIN(flags)) {
+      RESET_STIPPLE;
+   }
+
+   for (j=start+1; j<count; j++ ) {
+      if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+         RENDER_LINE( ELT(j-1), ELT(j) );
+      else
+         RENDER_LINE( ELT(j), ELT(j-1) );
+   }
+   POSTFIX;
+}
+
+
+static void TAG(render_line_loop)( GLcontext *ctx,
+				   GLuint start,
+				   GLuint count,
+				   GLuint flags )
+{
+   GLuint i;
+   LOCAL_VARS;
+
+   (void) flags;
+
+   INIT(GL_LINE_LOOP);
+
+   if (start+1 < count) {
+      if (TEST_PRIM_BEGIN(flags)) {
+	 RESET_STIPPLE;
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_LINE( ELT(start), ELT(start+1) );
+         else
+            RENDER_LINE( ELT(start+1), ELT(start) );
+      }
+
+      for ( i = start+2 ; i < count ; i++) {
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_LINE( ELT(i-1), ELT(i) );
+         else
+            RENDER_LINE( ELT(i), ELT(i-1) );
+      }
+
+      if ( TEST_PRIM_END(flags)) {
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_LINE( ELT(count-1), ELT(start) );
+         else
+            RENDER_LINE( ELT(start), ELT(count-1) );
+      }
+   }
+
+   POSTFIX;
+}
+
+
+static void TAG(render_triangles)( GLcontext *ctx,
+				   GLuint start,
+				   GLuint count,
+				   GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_TRIANGLES);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+2; j<count; j+=3) {
+	 /* Leave the edgeflags as supplied by the user.
+	  */
+	 RESET_STIPPLE;
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_TRI( ELT(j-1), ELT(j), ELT(j-2) );
+      }
+   } else {
+      for (j=start+2; j<count; j+=3) {
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_TRI( ELT(j-1), ELT(j), ELT(j-2) );
+      }
+   }
+   POSTFIX;
+}
+
+
+
+static void TAG(render_tri_strip)( GLcontext *ctx,
+				   GLuint start,
+				   GLuint count,
+				   GLuint flags )
+{
+   GLuint j;
+   GLuint parity = 0;
+   LOCAL_VARS;
+
+   INIT(GL_TRIANGLE_STRIP);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+2;j<count;j++,parity^=1) {
+         GLuint ej2, ej1, ej;
+         GLboolean ef2, ef1, ef;
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT) {
+            ej2 = ELT(j-2+parity);
+            ej1 = ELT(j-1-parity);
+            ej = ELT(j);
+         }
+         else {
+            ej2 = ELT(j-1+parity);
+            ej1 = ELT(j-parity);
+            ej = ELT(j-2);
+         }
+	 ef2 = EDGEFLAG_GET( ej2 );
+	 ef1 = EDGEFLAG_GET( ej1 );
+	 ef = EDGEFLAG_GET( ej );
+	 if (TEST_PRIM_BEGIN(flags)) {
+	    RESET_STIPPLE;
+	 }
+	 EDGEFLAG_SET( ej2, GL_TRUE );
+	 EDGEFLAG_SET( ej1, GL_TRUE );
+	 EDGEFLAG_SET( ej, GL_TRUE );
+         RENDER_TRI( ej2, ej1, ej );
+	 EDGEFLAG_SET( ej2, ef2 );
+	 EDGEFLAG_SET( ej1, ef1 );
+	 EDGEFLAG_SET( ej, ef );
+      }
+   } else {
+      for (j=start+2; j<count ; j++, parity^=1) {
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(j-2+parity), ELT(j-1-parity), ELT(j) );
+         else
+            RENDER_TRI( ELT(j-1+parity), ELT(j-parity), ELT(j-2) );
+      }
+   }
+   POSTFIX;
+}
+
+
+static void TAG(render_tri_fan)( GLcontext *ctx,
+				 GLuint start,
+				 GLuint count,
+				 GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_TRIANGLE_FAN);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+2;j<count;j++) {
+	 /* For trifans, all edges are boundary.
+	  */
+	 GLuint ejs = ELT(start);
+	 GLuint ej1 = ELT(j-1);
+	 GLuint ej = ELT(j);
+	 GLboolean efs = EDGEFLAG_GET( ejs );
+	 GLboolean ef1 = EDGEFLAG_GET( ej1 );
+	 GLboolean ef = EDGEFLAG_GET( ej );
+	 if (TEST_PRIM_BEGIN(flags)) {
+	    RESET_STIPPLE;
+	 }
+	 EDGEFLAG_SET( ejs, GL_TRUE );
+	 EDGEFLAG_SET( ej1, GL_TRUE );
+	 EDGEFLAG_SET( ej, GL_TRUE );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ejs, ej1, ej);
+         else
+            RENDER_TRI( ej, ejs, ej1);
+	 EDGEFLAG_SET( ejs, efs );
+	 EDGEFLAG_SET( ej1, ef1 );
+	 EDGEFLAG_SET( ej, ef );
+      }
+   } else {
+      for (j=start+2;j<count;j++) {
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT)
+            RENDER_TRI( ELT(start), ELT(j-1), ELT(j) );
+         else
+            RENDER_TRI( ELT(j), ELT(start), ELT(j-1) );
+      }
+   }
+
+   POSTFIX;
+}
+
+
+static void TAG(render_poly)( GLcontext *ctx,
+			      GLuint start,
+			      GLuint count,
+			      GLuint flags )
+{
+   GLuint j = start+2;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_POLYGON);
+   if (NEED_EDGEFLAG_SETUP) {
+      GLboolean efstart = EDGEFLAG_GET( ELT(start) );
+      GLboolean efcount = EDGEFLAG_GET( ELT(count-1) );
+
+      /* If the primitive does not begin here, the first edge
+       * is non-boundary.
+       */
+      if (!TEST_PRIM_BEGIN(flags))
+	 EDGEFLAG_SET( ELT(start), GL_FALSE );
+      else {
+	 RESET_STIPPLE;
+      }
+
+      /* If the primitive does not end here, the final edge is
+       * non-boundary.
+       */
+      if (!TEST_PRIM_END(flags))
+	 EDGEFLAG_SET( ELT(count-1), GL_FALSE );
+
+      /* Draw the first triangles (possibly zero)
+       */
+      if (j+1<count) {
+	 GLboolean ef = EDGEFLAG_GET( ELT(j) );
+	 EDGEFLAG_SET( ELT(j), GL_FALSE );
+	 RENDER_TRI( ELT(j-1), ELT(j), ELT(start) );
+	 EDGEFLAG_SET( ELT(j), ef );
+	 j++;
+
+	 /* Don't render the first edge again:
+	  */
+	 EDGEFLAG_SET( ELT(start), GL_FALSE );
+
+	 for (;j+1<count;j++) {
+	    GLboolean efj = EDGEFLAG_GET( ELT(j) );
+	    EDGEFLAG_SET( ELT(j), GL_FALSE );
+	    RENDER_TRI( ELT(j-1), ELT(j), ELT(start) );
+	    EDGEFLAG_SET( ELT(j), efj );
+	 }
+      }
+
+      /* Draw the last or only triangle
+       */
+      if (j < count)
+	 RENDER_TRI( ELT(j-1), ELT(j), ELT(start) );
+
+      /* Restore the first and last edgeflags:
+       */
+      EDGEFLAG_SET( ELT(count-1), efcount );
+      EDGEFLAG_SET( ELT(start), efstart );
+
+   }
+   else {
+      for (j=start+2;j<count;j++) {
+	 RENDER_TRI( ELT(j-1), ELT(j), ELT(start) );
+      }
+   }
+   POSTFIX;
+}
+
+static void TAG(render_quads)( GLcontext *ctx,
+			       GLuint start,
+			       GLuint count,
+			       GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_QUADS);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+3; j<count; j+=4) {
+	 /* Use user-specified edgeflags for quads.
+	  */
+	 RESET_STIPPLE;
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j-1), ELT(j), ELT(j-3) );
+      }
+   } else {
+      for (j=start+3; j<count; j+=4) {
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-3), ELT(j-2), ELT(j-1), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j-1), ELT(j), ELT(j-3) );
+      }
+   }
+   POSTFIX;
+}
+
+static void TAG(render_quad_strip)( GLcontext *ctx,
+				    GLuint start,
+				    GLuint count,
+				    GLuint flags )
+{
+   GLuint j;
+   LOCAL_VARS;
+   (void) flags;
+
+   INIT(GL_QUAD_STRIP);
+   if (NEED_EDGEFLAG_SETUP) {
+      for (j=start+3;j<count;j+=2) {
+	 /* All edges are boundary.  Set edgeflags to 1, draw the
+	  * quad, and restore them to the original values.
+	  */
+	 GLboolean ef3 = EDGEFLAG_GET( ELT(j-3) );
+	 GLboolean ef2 = EDGEFLAG_GET( ELT(j-2) );
+	 GLboolean ef1 = EDGEFLAG_GET( ELT(j-1) );
+	 GLboolean ef = EDGEFLAG_GET( ELT(j) );
+	 if (TEST_PRIM_BEGIN(flags)) {
+	    RESET_STIPPLE;
+	 }
+	 EDGEFLAG_SET( ELT(j-3), GL_TRUE );
+	 EDGEFLAG_SET( ELT(j-2), GL_TRUE );
+	 EDGEFLAG_SET( ELT(j-1), GL_TRUE );
+	 EDGEFLAG_SET( ELT(j), GL_TRUE );
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j), ELT(j-1), ELT(j-3) );
+	 EDGEFLAG_SET( ELT(j-3), ef3 );
+	 EDGEFLAG_SET( ELT(j-2), ef2 );
+	 EDGEFLAG_SET( ELT(j-1), ef1 );
+	 EDGEFLAG_SET( ELT(j), ef );
+      }
+   } else {
+      for (j=start+3;j<count;j+=2) {
+         if (ctx->Light.ProvokingVertex == GL_LAST_VERTEX_CONVENTION_EXT ||
+             !ctx->Const.QuadsFollowProvokingVertexConvention)
+            RENDER_QUAD( ELT(j-1), ELT(j-3), ELT(j-2), ELT(j) );
+         else
+            RENDER_QUAD( ELT(j-2), ELT(j), ELT(j-1), ELT(j-3) );
+      }
+   }
+   POSTFIX;
+}
+
+static void TAG(render_noop)( GLcontext *ctx,
+			      GLuint start,
+			      GLuint count,
+			      GLuint flags )
+{
+   (void)(ctx && start && count && flags);
+}
+
+RENDER_TAB_QUALIFIER void (*TAG(render_tab)[GL_POLYGON+2])(GLcontext *,
+							   GLuint,
+							   GLuint,
+							   GLuint) =
+{
+   TAG(render_points),
+   TAG(render_lines),
+   TAG(render_line_loop),
+   TAG(render_line_strip),
+   TAG(render_triangles),
+   TAG(render_tri_strip),
+   TAG(render_tri_fan),
+   TAG(render_quads),
+   TAG(render_quad_strip),
+   TAG(render_poly),
+   TAG(render_noop),
+};
+
+
+
+#ifndef PRESERVE_VB_DEFS
+#undef RENDER_TRI
+#undef RENDER_QUAD
+#undef RENDER_LINE
+#undef RENDER_POINTS
+#undef LOCAL_VARS
+#undef INIT
+#undef POSTFIX
+#undef RESET_STIPPLE
+#undef DBG
+#undef ELT
+#undef RENDER_TAB_QUALIFIER
+#endif
+
+#ifndef PRESERVE_TAG
+#undef TAG
+#endif
+
+#undef PRESERVE_VB_DEFS
+#undef PRESERVE_TAG
diff --git a/mesalib/src/mesa/tnl/t_vb_texgen.c b/mesalib/src/mesa/tnl/t_vb_texgen.c
new file mode 100644
index 000000000..7c1819b22
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_texgen.c
@@ -0,0 +1,611 @@
+/*
+ * 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:
+ *    Brian Paul
+ *    Keith Whitwell <keith@tungstengraphics.com>
+ */
+
+/*
+ * Regarding GL_NV_texgen_reflection:
+ *
+ * Portions of this software may use or implement intellectual
+ * property owned and licensed by NVIDIA Corporation. NVIDIA disclaims
+ * any and all warranties with respect to such intellectual property,
+ * including any use thereof or modifications thereto.
+ */
+
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "math/m_xform.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+
+
+/***********************************************************************
+ * Automatic texture coordinate generation (texgen) code.
+ */
+
+
+struct texgen_stage_data;
+
+typedef void (*texgen_func)( GLcontext *ctx,
+			     struct texgen_stage_data *store,
+			     GLuint unit);
+
+
+struct texgen_stage_data {
+
+   /* Per-texunit derived state.
+    */
+   GLuint TexgenSize[MAX_TEXTURE_COORD_UNITS];
+   texgen_func TexgenFunc[MAX_TEXTURE_COORD_UNITS];
+
+   /* Temporary values used in texgen.
+    */
+   GLfloat (*tmp_f)[3];
+   GLfloat *tmp_m;
+
+   /* Buffered outputs of the stage.
+    */
+   GLvector4f texcoord[MAX_TEXTURE_COORD_UNITS];
+};
+
+
+#define TEXGEN_STAGE_DATA(stage) ((struct texgen_stage_data *)stage->privatePtr)
+
+
+
+static GLuint all_bits[5] = {
+   0,
+   VEC_SIZE_1,
+   VEC_SIZE_2,
+   VEC_SIZE_3,
+   VEC_SIZE_4,
+};
+
+#define VEC_SIZE_FLAGS (VEC_SIZE_1|VEC_SIZE_2|VEC_SIZE_3|VEC_SIZE_4)
+
+#define TEXGEN_NEED_M            (TEXGEN_SPHERE_MAP)
+#define TEXGEN_NEED_F            (TEXGEN_SPHERE_MAP        | \
+				  TEXGEN_REFLECTION_MAP_NV)
+
+
+
+static void build_m3( GLfloat f[][3], GLfloat m[],
+		      const GLvector4f *normal,
+		      const GLvector4f *eye )
+{
+   GLuint stride = eye->stride;
+   GLfloat *coord = (GLfloat *)eye->start;
+   GLuint count = eye->count;
+   const GLfloat *norm = normal->start;
+   GLuint i;
+
+   for (i=0;i<count;i++,STRIDE_F(coord,stride),STRIDE_F(norm,normal->stride)) {
+      GLfloat u[3], two_nu, fx, fy, fz;
+      COPY_3V( u, coord );
+      NORMALIZE_3FV( u );
+      two_nu = 2.0F * DOT3(norm,u);
+      fx = f[i][0] = u[0] - norm[0] * two_nu;
+      fy = f[i][1] = u[1] - norm[1] * two_nu;
+      fz = f[i][2] = u[2] - norm[2] * two_nu;
+      m[i] = fx * fx + fy * fy + (fz + 1.0F) * (fz + 1.0F);
+      if (m[i] != 0.0F) {
+	 m[i] = 0.5F * _mesa_inv_sqrtf(m[i]);
+      }
+   }
+}
+
+
+
+static void build_m2( GLfloat f[][3], GLfloat m[],
+		      const GLvector4f *normal,
+		      const GLvector4f *eye )
+{
+   GLuint stride = eye->stride;
+   GLfloat *coord = eye->start;
+   GLuint count = eye->count;
+
+   GLfloat *norm = normal->start;
+   GLuint i;
+
+   for (i=0;i<count;i++,STRIDE_F(coord,stride),STRIDE_F(norm,normal->stride)) {
+      GLfloat u[3], two_nu, fx, fy, fz;
+      COPY_2V( u, coord );
+      u[2] = 0;
+      NORMALIZE_3FV( u );
+      two_nu = 2.0F * DOT3(norm,u);
+      fx = f[i][0] = u[0] - norm[0] * two_nu;
+      fy = f[i][1] = u[1] - norm[1] * two_nu;
+      fz = f[i][2] = u[2] - norm[2] * two_nu;
+      m[i] = fx * fx + fy * fy + (fz + 1.0F) * (fz + 1.0F);
+      if (m[i] != 0.0F) {
+	 m[i] = 0.5F * _mesa_inv_sqrtf(m[i]);
+      }
+   }
+}
+
+
+
+typedef void (*build_m_func)( GLfloat f[][3],
+			      GLfloat m[],
+			      const GLvector4f *normal,
+			      const GLvector4f *eye );
+
+
+static build_m_func build_m_tab[5] = {
+   NULL,
+   NULL,
+   build_m2,
+   build_m3,
+   build_m3
+};
+
+
+/* This is unusual in that we respect the stride of the output vector
+ * (f).  This allows us to pass in either a texcoord vector4f, or a
+ * temporary vector3f.
+ */
+static void build_f3( GLfloat *f,
+		      GLuint fstride,
+		      const GLvector4f *normal,
+		      const GLvector4f *eye )
+{
+   GLuint stride = eye->stride;
+   GLfloat *coord = eye->start;
+   GLuint count = eye->count;
+
+   GLfloat *norm = normal->start;
+   GLuint i;
+
+   for (i=0;i<count;i++) {
+      GLfloat u[3], two_nu;
+      COPY_3V( u, coord );
+      NORMALIZE_3FV( u );
+      two_nu = 2.0F * DOT3(norm,u);
+      f[0] = u[0] - norm[0] * two_nu;
+      f[1] = u[1] - norm[1] * two_nu;
+      f[2] = u[2] - norm[2] * two_nu;
+      STRIDE_F(coord,stride);
+      STRIDE_F(f,fstride);
+      STRIDE_F(norm, normal->stride);
+   }
+}
+
+
+static void build_f2( GLfloat *f,
+		      GLuint fstride,
+		      const GLvector4f *normal,
+		      const GLvector4f *eye )
+{
+   GLuint stride = eye->stride;
+   GLfloat *coord = eye->start;
+   GLuint count = eye->count;
+   GLfloat *norm = normal->start;
+   GLuint i;
+
+   for (i=0;i<count;i++) {
+
+      GLfloat u[3], two_nu;
+      COPY_2V( u, coord );
+      u[2] = 0;
+      NORMALIZE_3FV( u );
+      two_nu = 2.0F * DOT3(norm,u);
+      f[0] = u[0] - norm[0] * two_nu;
+      f[1] = u[1] - norm[1] * two_nu;
+      f[2] = u[2] - norm[2] * two_nu;
+
+      STRIDE_F(coord,stride);
+      STRIDE_F(f,fstride);
+      STRIDE_F(norm, normal->stride);
+   }
+}
+
+typedef void (*build_f_func)( GLfloat *f,
+			      GLuint fstride,
+			      const GLvector4f *normal_vec,
+			      const GLvector4f *eye );
+
+
+
+/* Just treat 4-vectors as 3-vectors.
+ */
+static build_f_func build_f_tab[5] = {
+   NULL,
+   NULL,
+   build_f2,
+   build_f3,
+   build_f3
+};
+
+
+
+/* Special case texgen functions.
+ */
+static void texgen_reflection_map_nv( GLcontext *ctx,
+				      struct texgen_stage_data *store,
+				      GLuint unit )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLvector4f *in = VB->AttribPtr[VERT_ATTRIB_TEX0 + unit];
+   GLvector4f *out = &store->texcoord[unit];
+
+   build_f_tab[VB->EyePtr->size]( out->start,
+				  out->stride,
+				  VB->AttribPtr[_TNL_ATTRIB_NORMAL],
+				  VB->EyePtr );
+
+   out->flags |= (in->flags & VEC_SIZE_FLAGS) | VEC_SIZE_3;
+   out->count = VB->Count;
+   out->size = MAX2(in->size, 3);
+   if (in->size == 4) 
+      _mesa_copy_tab[0x8]( out, in );
+}
+
+
+
+static void texgen_normal_map_nv( GLcontext *ctx,
+				  struct texgen_stage_data *store,
+				  GLuint unit )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLvector4f *in = VB->AttribPtr[VERT_ATTRIB_TEX0 + unit];
+   GLvector4f *out = &store->texcoord[unit];
+   GLvector4f *normal = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
+   GLfloat (*texcoord)[4] = (GLfloat (*)[4])out->start;
+   GLuint count = VB->Count;
+   GLuint i;
+   const GLfloat *norm = normal->start;
+
+   for (i=0;i<count;i++, STRIDE_F(norm, normal->stride)) {
+      texcoord[i][0] = norm[0];
+      texcoord[i][1] = norm[1];
+      texcoord[i][2] = norm[2];
+   }
+
+
+   out->flags |= (in->flags & VEC_SIZE_FLAGS) | VEC_SIZE_3;
+   out->count = count;
+   out->size = MAX2(in->size, 3);
+   if (in->size == 4) 
+      _mesa_copy_tab[0x8]( out, in );
+}
+
+
+static void texgen_sphere_map( GLcontext *ctx,
+			       struct texgen_stage_data *store,
+			       GLuint unit )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLvector4f *in = VB->AttribPtr[VERT_ATTRIB_TEX0 + unit];
+   GLvector4f *out = &store->texcoord[unit];
+   GLfloat (*texcoord)[4] = (GLfloat (*)[4]) out->start;
+   GLuint count = VB->Count;
+   GLuint i;
+   GLfloat (*f)[3] = store->tmp_f;
+   GLfloat *m = store->tmp_m;
+
+   (build_m_tab[VB->EyePtr->size])( store->tmp_f,
+				    store->tmp_m,
+				    VB->AttribPtr[_TNL_ATTRIB_NORMAL],
+				    VB->EyePtr );
+
+   out->size = MAX2(in->size,2);
+
+   for (i=0;i<count;i++) {
+      texcoord[i][0] = f[i][0] * m[i] + 0.5F;
+      texcoord[i][1] = f[i][1] * m[i] + 0.5F;
+   }
+
+   out->count = count;
+   out->flags |= (in->flags & VEC_SIZE_FLAGS) | VEC_SIZE_2;
+   if (in->size > 2)
+      _mesa_copy_tab[all_bits[in->size] & ~0x3]( out, in );
+}
+
+
+
+static void texgen( GLcontext *ctx,
+		    struct texgen_stage_data *store,
+		    GLuint unit )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   GLvector4f *in = VB->AttribPtr[VERT_ATTRIB_TEX0 + unit];
+   GLvector4f *out = &store->texcoord[unit];
+   const struct gl_texture_unit *texUnit = &ctx->Texture.Unit[unit];
+   const GLvector4f *obj = VB->ObjPtr;
+   const GLvector4f *eye = VB->EyePtr;
+   const GLvector4f *normal = VB->AttribPtr[_TNL_ATTRIB_NORMAL];
+   const GLfloat *m = store->tmp_m;
+   const GLuint count = VB->Count;
+   GLfloat (*texcoord)[4] = (GLfloat (*)[4])out->data;
+   GLfloat (*f)[3] = store->tmp_f;
+   GLuint copy;
+
+   if (texUnit->_GenFlags & TEXGEN_NEED_M) {
+      build_m_tab[eye->size]( store->tmp_f, store->tmp_m, normal, eye );
+   } else if (texUnit->_GenFlags & TEXGEN_NEED_F) {
+      build_f_tab[eye->size]( (GLfloat *)store->tmp_f, 3, normal, eye );
+   }
+
+
+   out->size = MAX2(in->size, store->TexgenSize[unit]);
+   out->flags |= (in->flags & VEC_SIZE_FLAGS) | texUnit->TexGenEnabled;
+   out->count = count;
+
+   copy = (all_bits[in->size] & ~texUnit->TexGenEnabled);
+   if (copy)
+      _mesa_copy_tab[copy]( out, in );
+
+   if (texUnit->TexGenEnabled & S_BIT) {
+      GLuint i;
+      switch (texUnit->GenS.Mode) {
+      case GL_OBJECT_LINEAR:
+	 _mesa_dotprod_tab[obj->size]( (GLfloat *)out->data,
+				       sizeof(out->data[0]), obj,
+				       texUnit->GenS.ObjectPlane );
+	 break;
+      case GL_EYE_LINEAR:
+	 _mesa_dotprod_tab[eye->size]( (GLfloat *)out->data,
+				       sizeof(out->data[0]), eye,
+				       texUnit->GenS.EyePlane );
+	 break;
+      case GL_SPHERE_MAP:
+         for (i = 0; i < count; i++)
+            texcoord[i][0] = f[i][0] * m[i] + 0.5F;
+	 break;
+      case GL_REFLECTION_MAP_NV:
+	 for (i=0;i<count;i++)
+	     texcoord[i][0] = f[i][0];
+	 break;
+      case GL_NORMAL_MAP_NV: {
+	 const GLfloat *norm = normal->start;
+	 for (i=0;i<count;i++, STRIDE_F(norm, normal->stride)) {
+	     texcoord[i][0] = norm[0];
+	 }
+	 break;
+      }
+      default:
+	 _mesa_problem(ctx, "Bad S texgen");
+      }
+   }
+
+   if (texUnit->TexGenEnabled & T_BIT) {
+      GLuint i;
+      switch (texUnit->GenT.Mode) {
+      case GL_OBJECT_LINEAR:
+	 _mesa_dotprod_tab[obj->size]( &(out->data[0][1]),
+				       sizeof(out->data[0]), obj,
+				       texUnit->GenT.ObjectPlane );
+	 break;
+      case GL_EYE_LINEAR:
+	 _mesa_dotprod_tab[eye->size]( &(out->data[0][1]),
+				       sizeof(out->data[0]), eye,
+				       texUnit->GenT.EyePlane );
+	 break;
+      case GL_SPHERE_MAP:
+         for (i = 0; i < count; i++)
+            texcoord[i][1] = f[i][1] * m[i] + 0.5F;
+	 break;
+      case GL_REFLECTION_MAP_NV:
+	 for (i=0;i<count;i++)
+	     texcoord[i][1] = f[i][1];
+	 break;
+      case GL_NORMAL_MAP_NV: {
+	 const GLfloat *norm = normal->start;
+	 for (i=0;i<count;i++, STRIDE_F(norm, normal->stride)) {
+	     texcoord[i][1] = norm[1];
+	 }
+	 break;
+      }
+      default:
+	 _mesa_problem(ctx, "Bad T texgen");
+      }
+   }
+
+   if (texUnit->TexGenEnabled & R_BIT) {
+      GLuint i;
+      switch (texUnit->GenR.Mode) {
+      case GL_OBJECT_LINEAR:
+	 _mesa_dotprod_tab[obj->size]( &(out->data[0][2]),
+				       sizeof(out->data[0]), obj,
+				       texUnit->GenR.ObjectPlane );
+	 break;
+      case GL_EYE_LINEAR:
+	 _mesa_dotprod_tab[eye->size]( &(out->data[0][2]),
+				       sizeof(out->data[0]), eye,
+				       texUnit->GenR.EyePlane );
+	 break;
+      case GL_REFLECTION_MAP_NV:
+	 for (i=0;i<count;i++)
+	     texcoord[i][2] = f[i][2];
+	 break;
+      case GL_NORMAL_MAP_NV: {
+	 const GLfloat *norm = normal->start;
+	 for (i=0;i<count;i++,STRIDE_F(norm, normal->stride)) {
+	     texcoord[i][2] = norm[2];
+	 }
+	 break;
+      }
+      default:
+	 _mesa_problem(ctx, "Bad R texgen");
+      }
+   }
+
+   if (texUnit->TexGenEnabled & Q_BIT) {
+      switch (texUnit->GenQ.Mode) {
+      case GL_OBJECT_LINEAR:
+	 _mesa_dotprod_tab[obj->size]( &(out->data[0][3]),
+				       sizeof(out->data[0]), obj,
+				       texUnit->GenQ.ObjectPlane );
+	 break;
+      case GL_EYE_LINEAR:
+	 _mesa_dotprod_tab[eye->size]( &(out->data[0][3]),
+				       sizeof(out->data[0]), eye,
+				       texUnit->GenQ.EyePlane );
+	 break;
+      default:
+	 _mesa_problem(ctx, "Bad Q texgen");
+      }
+   }
+}
+
+
+
+
+static GLboolean run_texgen_stage( GLcontext *ctx,
+				   struct tnl_pipeline_stage *stage )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct texgen_stage_data *store = TEXGEN_STAGE_DATA(stage);
+   GLuint i;
+
+   if (!ctx->Texture._TexGenEnabled || ctx->VertexProgram._Current) 
+      return GL_TRUE;
+
+   for (i = 0 ; i < ctx->Const.MaxTextureCoordUnits ; i++) {
+      struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
+
+      if (texUnit->TexGenEnabled) {
+
+	 store->TexgenFunc[i]( ctx, store, i );
+
+         VB->TexCoordPtr[i] =
+         VB->AttribPtr[VERT_ATTRIB_TEX0 + i] = &store->texcoord[i];
+      }
+   }
+
+   return GL_TRUE;
+}
+
+
+static void validate_texgen_stage( GLcontext *ctx,
+				   struct tnl_pipeline_stage *stage )
+{
+   struct texgen_stage_data *store = TEXGEN_STAGE_DATA(stage);
+   GLuint i;
+
+   if (!ctx->Texture._TexGenEnabled || ctx->VertexProgram._Current) 
+      return;
+
+   for (i = 0 ; i < ctx->Const.MaxTextureCoordUnits ; i++) {
+      struct gl_texture_unit *texUnit = &ctx->Texture.Unit[i];
+
+      if (texUnit->TexGenEnabled) {
+	 GLuint sz;
+
+	 if (texUnit->TexGenEnabled & Q_BIT)
+	    sz = 4;
+	 else if (texUnit->TexGenEnabled & R_BIT)
+	    sz = 3;
+	 else if (texUnit->TexGenEnabled & T_BIT)
+	    sz = 2;
+	 else
+	    sz = 1;
+
+	 store->TexgenSize[i] = sz;
+	 store->TexgenFunc[i] = texgen; /* general solution */
+
+         /* look for special texgen cases */
+	 if (texUnit->TexGenEnabled == (S_BIT|T_BIT|R_BIT)) {
+	    if (texUnit->_GenFlags == TEXGEN_REFLECTION_MAP_NV) {
+	       store->TexgenFunc[i] = texgen_reflection_map_nv;
+	    }
+	    else if (texUnit->_GenFlags == TEXGEN_NORMAL_MAP_NV) {
+	       store->TexgenFunc[i] = texgen_normal_map_nv;
+	    }
+	 }
+	 else if (texUnit->TexGenEnabled == (S_BIT|T_BIT) &&
+		  texUnit->_GenFlags == TEXGEN_SPHERE_MAP) {
+	    store->TexgenFunc[i] = texgen_sphere_map;
+	 }
+      }
+   }
+}
+
+
+
+
+
+/* Called the first time stage->run() is invoked.
+ */
+static GLboolean alloc_texgen_data( GLcontext *ctx,
+				    struct tnl_pipeline_stage *stage )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct texgen_stage_data *store;
+   GLuint i;
+
+   stage->privatePtr = CALLOC(sizeof(*store));
+   store = TEXGEN_STAGE_DATA(stage);
+   if (!store)
+      return GL_FALSE;
+
+   for (i = 0 ; i < ctx->Const.MaxTextureCoordUnits ; i++)
+      _mesa_vector4f_alloc( &store->texcoord[i], 0, VB->Size, 32 );
+
+   store->tmp_f = (GLfloat (*)[3]) MALLOC(VB->Size * sizeof(GLfloat) * 3);
+   store->tmp_m = (GLfloat *) MALLOC(VB->Size * sizeof(GLfloat));
+
+   return GL_TRUE;
+}
+
+
+static void free_texgen_data( struct tnl_pipeline_stage *stage )
+
+{
+   struct texgen_stage_data *store = TEXGEN_STAGE_DATA(stage);
+   GLuint i;
+
+   if (store) {
+      for (i = 0 ; i < MAX_TEXTURE_COORD_UNITS ; i++)
+	 if (store->texcoord[i].data)
+	    _mesa_vector4f_free( &store->texcoord[i] );
+
+
+      if (store->tmp_f) FREE( store->tmp_f );
+      if (store->tmp_m) FREE( store->tmp_m );
+      FREE( store );
+      stage->privatePtr = NULL;
+   }
+}
+
+
+
+const struct tnl_pipeline_stage _tnl_texgen_stage =
+{
+   "texgen",			/* name */
+   NULL,			/* private data */
+   alloc_texgen_data,		/* destructor */
+   free_texgen_data,		/* destructor */
+   validate_texgen_stage,		/* check */
+   run_texgen_stage		/* run -- initially set to alloc data */
+};
diff --git a/mesalib/src/mesa/tnl/t_vb_texmat.c b/mesalib/src/mesa/tnl/t_vb_texmat.c
new file mode 100644
index 000000000..0abe8cc35
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_texmat.c
@@ -0,0 +1,130 @@
+/*
+ * 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 "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "math/m_xform.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+
+/* Is there any real benefit seperating texmat from texgen?  It means
+ * we need two lots of intermediate storage.  Any changes to
+ * _NEW_TEXTURE will invalidate both sets -- it's only on changes to
+ * *only* _NEW_TEXTURE_MATRIX that texgen survives but texmat doesn't.
+ *
+ * However, the seperation of this code from the complex texgen stuff
+ * is very appealing.
+ */
+struct texmat_stage_data {
+   GLvector4f texcoord[MAX_TEXTURE_COORD_UNITS];
+};
+
+#define TEXMAT_STAGE_DATA(stage) ((struct texmat_stage_data *)stage->privatePtr)
+
+
+
+static GLboolean run_texmat_stage( GLcontext *ctx,
+				   struct tnl_pipeline_stage *stage )
+{
+   struct texmat_stage_data *store = TEXMAT_STAGE_DATA(stage);
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   GLuint i;
+
+   if (!ctx->Texture._TexMatEnabled || ctx->VertexProgram._Current) 
+      return GL_TRUE;
+
+   /* ENABLE_TEXMAT implies that the texture matrix is not the
+    * identity, so we don't have to check that here.
+    */
+   for (i = 0 ; i < ctx->Const.MaxTextureCoordUnits ; i++) {
+      if (ctx->Texture._TexMatEnabled & ENABLE_TEXMAT(i)) {
+	 (void) TransformRaw( &store->texcoord[i],
+			      ctx->TextureMatrixStack[i].Top,
+			      VB->AttribPtr[_TNL_ATTRIB_TEX0 + i]);
+
+         VB->TexCoordPtr[i] = 
+	 VB->AttribPtr[VERT_ATTRIB_TEX0+i] = &store->texcoord[i];
+      }
+   }
+
+   return GL_TRUE;
+}
+
+
+/* Called the first time stage->run() is invoked.
+ */
+static GLboolean alloc_texmat_data( GLcontext *ctx,
+				    struct tnl_pipeline_stage *stage )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct texmat_stage_data *store;
+   GLuint i;
+
+   stage->privatePtr = CALLOC(sizeof(*store));
+   store = TEXMAT_STAGE_DATA(stage);
+   if (!store)
+      return GL_FALSE;
+
+   for (i = 0 ; i < ctx->Const.MaxTextureCoordUnits ; i++)
+      _mesa_vector4f_alloc( &store->texcoord[i], 0, VB->Size, 32 );
+
+   return GL_TRUE;
+}
+
+
+static void free_texmat_data( struct tnl_pipeline_stage *stage )
+{
+   struct texmat_stage_data *store = TEXMAT_STAGE_DATA(stage);
+   GLuint i;
+
+   if (store) {
+      for (i = 0; i < MAX_TEXTURE_COORD_UNITS; i++)
+	 if (store->texcoord[i].data)
+	    _mesa_vector4f_free( &store->texcoord[i] );
+      FREE( store );
+      stage->privatePtr = NULL;
+   }
+}
+
+
+
+const struct tnl_pipeline_stage _tnl_texture_transform_stage =
+{
+   "texture transform",			/* name */
+   NULL,				/* private data */
+   alloc_texmat_data,
+   free_texmat_data,			/* destructor */
+   NULL,
+   run_texmat_stage,
+};
diff --git a/mesalib/src/mesa/tnl/t_vb_vertex.c b/mesalib/src/mesa/tnl/t_vb_vertex.c
new file mode 100644
index 000000000..30aa7c408
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vb_vertex.c
@@ -0,0 +1,264 @@
+/*
+ * 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 "main/glheader.h"
+#include "main/colormac.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "main/imports.h"
+#include "main/mtypes.h"
+
+#include "math/m_xform.h"
+
+#include "t_context.h"
+#include "t_pipeline.h"
+
+
+
+struct vertex_stage_data {
+   GLvector4f eye;
+   GLvector4f clip;
+   GLvector4f proj;
+   GLubyte *clipmask;
+   GLubyte ormask;
+   GLubyte andmask;
+};
+
+#define VERTEX_STAGE_DATA(stage) ((struct vertex_stage_data *)stage->privatePtr)
+
+
+
+
+/* This function implements cliptesting for user-defined clip planes.
+ * The clipping of primitives to these planes is implemented in
+ * t_render_clip.h.
+ */
+#define USER_CLIPTEST(NAME, SZ)					\
+static void NAME( GLcontext *ctx,				\
+		  GLvector4f *clip,				\
+		  GLubyte *clipmask,				\
+		  GLubyte *clipormask,				\
+		  GLubyte *clipandmask )			\
+{								\
+   GLuint p;							\
+								\
+   for (p = 0; p < ctx->Const.MaxClipPlanes; p++)		\
+      if (ctx->Transform.ClipPlanesEnabled & (1 << p)) {	\
+	 GLuint nr, i;						\
+	 const GLfloat a = ctx->Transform._ClipUserPlane[p][0];	\
+	 const GLfloat b = ctx->Transform._ClipUserPlane[p][1];	\
+	 const GLfloat c = ctx->Transform._ClipUserPlane[p][2];	\
+	 const GLfloat d = ctx->Transform._ClipUserPlane[p][3];	\
+         GLfloat *coord = (GLfloat *)clip->data;		\
+         GLuint stride = clip->stride;				\
+         GLuint count = clip->count;				\
+								\
+	 for (nr = 0, i = 0 ; i < count ; i++) {		\
+	    GLfloat dp = coord[0] * a + coord[1] * b;		\
+	    if (SZ > 2) dp += coord[2] * c;			\
+	    if (SZ > 3) dp += coord[3] * d; else dp += d;	\
+								\
+	    if (dp < 0) {					\
+	       nr++;						\
+	       clipmask[i] |= CLIP_USER_BIT;			\
+	    }							\
+								\
+	    STRIDE_F(coord, stride);				\
+	 }							\
+								\
+	 if (nr > 0) {						\
+	    *clipormask |= CLIP_USER_BIT;			\
+	    if (nr == count) {					\
+	       *clipandmask |= CLIP_USER_BIT;			\
+	       return;						\
+	    }							\
+	 }							\
+      }								\
+}
+
+
+USER_CLIPTEST(userclip2, 2)
+USER_CLIPTEST(userclip3, 3)
+USER_CLIPTEST(userclip4, 4)
+
+static void (*(usercliptab[5]))( GLcontext *,
+				 GLvector4f *, GLubyte *,
+				 GLubyte *, GLubyte * ) =
+{
+   NULL,
+   NULL,
+   userclip2,
+   userclip3,
+   userclip4
+};
+
+
+
+static GLboolean run_vertex_stage( GLcontext *ctx,
+				   struct tnl_pipeline_stage *stage )
+{
+   struct vertex_stage_data *store = (struct vertex_stage_data *)stage->privatePtr;
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+
+   if (ctx->VertexProgram._Current) 
+      return GL_TRUE;
+
+   if (ctx->_NeedEyeCoords) {
+      /* Separate modelview transformation:
+       * Use combined ModelProject to avoid some depth artifacts
+       */
+      if (ctx->ModelviewMatrixStack.Top->type == MATRIX_IDENTITY)
+	 VB->EyePtr = VB->ObjPtr;
+      else
+	 VB->EyePtr = TransformRaw( &store->eye,
+				    ctx->ModelviewMatrixStack.Top,
+				    VB->ObjPtr);
+   }
+
+   VB->ClipPtr = TransformRaw( &store->clip,
+			       &ctx->_ModelProjectMatrix,
+			       VB->ObjPtr );
+
+   /* 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;
+   }
+
+
+   /* Cliptest and perspective divide.  Clip functions must clear
+    * the clipmask.
+    */
+   store->ormask = 0;
+   store->andmask = CLIP_FRUSTUM_BITS;
+
+   if (tnl->NeedNdcCoords) {
+      VB->NdcPtr =
+	 _mesa_clip_tab[VB->ClipPtr->size]( VB->ClipPtr,
+					    &store->proj,
+					    store->clipmask,
+					    &store->ormask,
+					    &store->andmask );
+   }
+   else {
+      VB->NdcPtr = NULL;
+      _mesa_clip_np_tab[VB->ClipPtr->size]( VB->ClipPtr,
+					    NULL,
+					    store->clipmask,
+					    &store->ormask,
+					    &store->andmask );
+   }
+
+   if (store->andmask)
+      return GL_FALSE;
+
+
+   /* Test userclip planes.  This contributes to VB->ClipMask, so
+    * is essentially required to be in this stage.
+    */
+   if (ctx->Transform.ClipPlanesEnabled) {
+      usercliptab[VB->ClipPtr->size]( ctx,
+				      VB->ClipPtr,
+				      store->clipmask,
+				      &store->ormask,
+				      &store->andmask );
+
+      if (store->andmask)
+	 return GL_FALSE;
+   }
+
+   VB->ClipAndMask = store->andmask;
+   VB->ClipOrMask = store->ormask;
+   VB->ClipMask = store->clipmask;
+
+   return GL_TRUE;
+}
+
+
+static GLboolean init_vertex_stage( GLcontext *ctx,
+				    struct tnl_pipeline_stage *stage )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct vertex_stage_data *store;
+   GLuint size = VB->Size;
+
+   stage->privatePtr = CALLOC(sizeof(*store));
+   store = VERTEX_STAGE_DATA(stage);
+   if (!store)
+      return GL_FALSE;
+
+   _mesa_vector4f_alloc( &store->eye, 0, size, 32 );
+   _mesa_vector4f_alloc( &store->clip, 0, size, 32 );
+   _mesa_vector4f_alloc( &store->proj, 0, size, 32 );
+
+   store->clipmask = (GLubyte *) ALIGN_MALLOC(sizeof(GLubyte)*size, 32 );
+
+   if (!store->clipmask ||
+       !store->eye.data ||
+       !store->clip.data ||
+       !store->proj.data)
+      return GL_FALSE;
+
+   return GL_TRUE;
+}
+
+static void dtr( struct tnl_pipeline_stage *stage )
+{
+   struct vertex_stage_data *store = VERTEX_STAGE_DATA(stage);
+
+   if (store) {
+      _mesa_vector4f_free( &store->eye );
+      _mesa_vector4f_free( &store->clip );
+      _mesa_vector4f_free( &store->proj );
+      ALIGN_FREE( store->clipmask );
+      FREE(store);
+      stage->privatePtr = NULL;
+      stage->run = init_vertex_stage;
+   }
+}
+
+
+const struct tnl_pipeline_stage _tnl_vertex_transform_stage =
+{
+   "modelview/project/cliptest/divide",
+   NULL,			/* private data */
+   init_vertex_stage,
+   dtr,				/* destructor */
+   NULL,
+   run_vertex_stage		/* run -- initially set to init */
+};
diff --git a/mesalib/src/mesa/tnl/t_vertex.c b/mesalib/src/mesa/tnl/t_vertex.c
new file mode 100644
index 000000000..fe4209ae5
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vertex.c
@@ -0,0 +1,564 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * 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 <keithw@tungstengraphics.com>
+ */
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+
+#include "t_context.h"
+#include "t_vertex.h"
+
+#define DBG 0
+
+/* Build and manage clipspace/ndc/window vertices.
+ */
+
+static GLboolean match_fastpath( struct tnl_clipspace *vtx,
+				 const struct tnl_clipspace_fastpath *fp)
+{
+   GLuint j;
+
+   if (vtx->attr_count != fp->attr_count) 
+      return GL_FALSE;
+
+   for (j = 0; j < vtx->attr_count; j++) 
+      if (vtx->attr[j].format != fp->attr[j].format ||
+	  vtx->attr[j].inputsize != fp->attr[j].size ||
+	  vtx->attr[j].vertoffset != fp->attr[j].offset) 
+	 return GL_FALSE;
+      
+   if (fp->match_strides) {
+      if (vtx->vertex_size != fp->vertex_size)
+	 return GL_FALSE;
+
+      for (j = 0; j < vtx->attr_count; j++) 
+	 if (vtx->attr[j].inputstride != fp->attr[j].stride) 
+	    return GL_FALSE;
+   }
+   
+   return GL_TRUE;
+}
+
+static GLboolean search_fastpath_emit( struct tnl_clipspace *vtx )
+{
+   struct tnl_clipspace_fastpath *fp = vtx->fastpath;
+
+   for ( ; fp ; fp = fp->next) {
+      if (match_fastpath(vtx, fp)) {
+         vtx->emit = fp->func;
+	 return GL_TRUE;
+      }
+   }
+
+   return GL_FALSE;
+}
+
+void _tnl_register_fastpath( struct tnl_clipspace *vtx,
+			     GLboolean match_strides )
+{
+   struct tnl_clipspace_fastpath *fastpath = CALLOC_STRUCT(tnl_clipspace_fastpath);
+   GLuint i;
+
+   fastpath->vertex_size = vtx->vertex_size;
+   fastpath->attr_count = vtx->attr_count;
+   fastpath->match_strides = match_strides;
+   fastpath->func = vtx->emit;
+   fastpath->attr = (struct tnl_attr_type *)
+      _mesa_malloc(vtx->attr_count * sizeof(fastpath->attr[0]));
+
+   for (i = 0; i < vtx->attr_count; i++) {
+      fastpath->attr[i].format = vtx->attr[i].format;
+      fastpath->attr[i].stride = vtx->attr[i].inputstride;
+      fastpath->attr[i].size = vtx->attr[i].inputsize;
+      fastpath->attr[i].offset = vtx->attr[i].vertoffset;
+   }
+
+   fastpath->next = vtx->fastpath;
+   vtx->fastpath = fastpath;
+}
+
+
+
+/***********************************************************************
+ * Build codegen functions or return generic ones:
+ */
+static void choose_emit_func( GLcontext *ctx, GLuint count, GLubyte *dest)
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   struct tnl_clipspace_attr *a = vtx->attr;
+   const GLuint attr_count = vtx->attr_count;
+   GLuint j;
+
+   for (j = 0; j < attr_count; j++) {
+      GLvector4f *vptr = VB->AttribPtr[a[j].attrib];
+      a[j].inputstride = vptr->stride;
+      a[j].inputsize = vptr->size;
+      a[j].emit = a[j].insert[vptr->size - 1]; /* not always used */
+   }
+
+   vtx->emit = NULL;
+   
+   /* Does this match an existing (hardwired, codegen or known-bad)
+    * fastpath?
+    */
+   if (search_fastpath_emit(vtx)) {
+      /* Use this result.  If it is null, then it is already known
+       * that the current state will fail for codegen and there is no
+       * point trying again.
+       */
+   }
+   else if (vtx->codegen_emit) {
+      vtx->codegen_emit(ctx);
+   }
+
+   if (!vtx->emit) {
+      _tnl_generate_hardwired_emit(ctx);
+   }
+
+   /* Otherwise use the generic version:
+    */
+   if (!vtx->emit)
+      vtx->emit = _tnl_generic_emit;
+
+   vtx->emit( ctx, count, dest );
+}
+
+
+
+static void choose_interp_func( GLcontext *ctx,
+				GLfloat t,
+				GLuint edst, GLuint eout, GLuint ein,
+				GLboolean force_boundary )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+
+   if (vtx->need_extras && 
+       (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
+      vtx->interp = _tnl_generic_interp_extras;
+   } else {
+      vtx->interp = _tnl_generic_interp;
+   }
+
+   vtx->interp( ctx, t, edst, eout, ein, force_boundary );
+}
+
+
+static void choose_copy_pv_func(  GLcontext *ctx, GLuint edst, GLuint esrc )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+
+   if (vtx->need_extras && 
+       (ctx->_TriangleCaps & (DD_TRI_LIGHT_TWOSIDE|DD_TRI_UNFILLED))) {
+      vtx->copy_pv = _tnl_generic_copy_pv_extras;
+   } else {
+      vtx->copy_pv = _tnl_generic_copy_pv;
+   }
+
+   vtx->copy_pv( ctx, edst, esrc );
+}
+
+
+/***********************************************************************
+ * Public entrypoints, mostly dispatch to the above:
+ */
+
+
+/* Interpolate between two vertices to produce a third:
+ */
+void _tnl_interp( GLcontext *ctx,
+		  GLfloat t,
+		  GLuint edst, GLuint eout, GLuint ein,
+		  GLboolean force_boundary )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   vtx->interp( ctx, t, edst, eout, ein, force_boundary );
+}
+
+/* Copy colors from one vertex to another:
+ */
+void _tnl_copy_pv(  GLcontext *ctx, GLuint edst, GLuint esrc )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   vtx->copy_pv( ctx, edst, esrc );
+}
+
+
+/* Extract a named attribute from a hardware vertex.  Will have to
+ * reverse any viewport transformation, swizzling or other conversions
+ * which may have been applied:
+ */
+void _tnl_get_attr( GLcontext *ctx, const void *vin,
+			      GLenum attr, GLfloat *dest )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   const struct tnl_clipspace_attr *a = vtx->attr;
+   const GLuint attr_count = vtx->attr_count;
+   GLuint j;
+
+   for (j = 0; j < attr_count; j++) {
+      if (a[j].attrib == attr) {
+	 a[j].extract( &a[j], dest, (GLubyte *)vin + a[j].vertoffset );
+	 return;
+      }
+   }
+
+   /* Else return the value from ctx->Current.
+    */
+   if (attr == _TNL_ATTRIB_POINTSIZE) {
+      /* If the hardware vertex doesn't have point size then use size from
+       * GLcontext.  XXX this will be wrong if drawing attenuated points!
+       */
+      dest[0] = ctx->Point.Size;
+   }
+   else {
+      _mesa_memcpy( dest, ctx->Current.Attrib[attr], 4*sizeof(GLfloat));
+   }
+}
+
+
+/* Complementary operation to the above.
+ */
+void _tnl_set_attr( GLcontext *ctx, void *vout,
+		    GLenum attr, const GLfloat *src )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   const struct tnl_clipspace_attr *a = vtx->attr;
+   const GLuint attr_count = vtx->attr_count;
+   GLuint j;
+
+   for (j = 0; j < attr_count; j++) {
+      if (a[j].attrib == attr) {
+	 a[j].insert[4-1]( &a[j], (GLubyte *)vout + a[j].vertoffset, src );
+	 return;
+      }
+   }
+}
+
+
+void *_tnl_get_vertex( GLcontext *ctx, GLuint nr )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+
+   return vtx->vertex_buf + nr * vtx->vertex_size;
+}
+
+void _tnl_invalidate_vertex_state( GLcontext *ctx, GLuint new_state )
+{
+   if (new_state & (_DD_NEW_TRI_LIGHT_TWOSIDE|_DD_NEW_TRI_UNFILLED) ) {
+      struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+      vtx->new_inputs = ~0;
+      vtx->interp = choose_interp_func;
+      vtx->copy_pv = choose_copy_pv_func;
+   }
+}
+
+static void invalidate_funcs( struct tnl_clipspace *vtx )
+{
+   vtx->emit = choose_emit_func;
+   vtx->interp = choose_interp_func;
+   vtx->copy_pv = choose_copy_pv_func;
+   vtx->new_inputs = ~0;
+}
+
+GLuint _tnl_install_attrs( GLcontext *ctx, const struct tnl_attr_map *map,
+			   GLuint nr, const GLfloat *vp, 
+			   GLuint unpacked_size )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   GLuint offset = 0;
+   GLuint i, j;
+
+   assert(nr < _TNL_ATTRIB_MAX);
+   assert(nr == 0 || map[0].attrib == VERT_ATTRIB_POS);
+
+   vtx->new_inputs = ~0;
+   vtx->need_viewport = GL_FALSE;
+
+   if (vp) {
+      vtx->need_viewport = GL_TRUE;
+   }
+
+   for (j = 0, i = 0; i < nr; i++) {
+      const GLuint format = map[i].format;
+      if (format == EMIT_PAD) {
+	 if (DBG)
+	    _mesa_printf("%d: pad %d, offset %d\n", i,  
+			 map[i].offset, offset);  
+
+	 offset += map[i].offset;
+
+      }
+      else {
+	 GLuint tmpoffset;
+
+	 if (unpacked_size) 
+	    tmpoffset = map[i].offset;
+	 else
+	    tmpoffset = offset;
+
+	 if (vtx->attr_count != j ||
+	     vtx->attr[j].attrib != map[i].attrib ||
+	     vtx->attr[j].format != format ||
+	     vtx->attr[j].vertoffset != tmpoffset) {
+	    invalidate_funcs(vtx);
+
+	    vtx->attr[j].attrib = map[i].attrib;
+	    vtx->attr[j].format = format;
+	    vtx->attr[j].vp = vp;
+	    vtx->attr[j].insert = _tnl_format_info[format].insert;
+	    vtx->attr[j].extract = _tnl_format_info[format].extract;
+	    vtx->attr[j].vertattrsize = _tnl_format_info[format].attrsize;
+	    vtx->attr[j].vertoffset = tmpoffset;
+	 }
+
+	 
+	 if (DBG)
+	    _mesa_printf("%d: %s, vp %p, offset %d\n", i,  
+			 _tnl_format_info[format].name, (void *)vp,
+			 vtx->attr[j].vertoffset);   
+
+	 offset += _tnl_format_info[format].attrsize;
+	 j++;
+      }
+   }
+
+   vtx->attr_count = j;
+
+   if (unpacked_size)
+      vtx->vertex_size = unpacked_size;
+   else
+      vtx->vertex_size = offset;
+
+   assert(vtx->vertex_size <= vtx->max_vertex_size);
+   return vtx->vertex_size;
+}
+
+
+
+void _tnl_invalidate_vertices( GLcontext *ctx, GLuint newinputs )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   vtx->new_inputs |= newinputs;
+}
+
+
+/* This event has broader use beyond this file - will move elsewhere
+ * and probably invoke a driver callback.
+ */
+void _tnl_notify_pipeline_output_change( GLcontext *ctx )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   invalidate_funcs(vtx);
+}
+
+
+static void adjust_input_ptrs( GLcontext *ctx, GLint diff)
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   struct tnl_clipspace_attr *a = vtx->attr;
+   const GLuint count = vtx->attr_count;
+   int j;
+
+   diff -= 1;
+   for (j=0; j<count; ++j) {
+           register GLvector4f *vptr = VB->AttribPtr[a->attrib];
+	   (a++)->inputptr += diff*vptr->stride;
+   }
+}
+
+static void update_input_ptrs( GLcontext *ctx, GLuint start )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   struct tnl_clipspace_attr *a = vtx->attr;
+   const GLuint count = vtx->attr_count;
+   GLuint j;
+   
+   for (j = 0; j < count; j++) {
+      GLvector4f *vptr = VB->AttribPtr[a[j].attrib];
+
+      if (vtx->emit != choose_emit_func) {
+	 assert(a[j].inputstride == vptr->stride);
+	 assert(a[j].inputsize == vptr->size);
+      }
+
+      a[j].inputptr = ((GLubyte *)vptr->data) + start * vptr->stride;
+   }
+   
+   if (a->vp) {
+      vtx->vp_scale[0] = a->vp[MAT_SX];
+      vtx->vp_scale[1] = a->vp[MAT_SY];
+      vtx->vp_scale[2] = a->vp[MAT_SZ];
+      vtx->vp_scale[3] = 1.0;
+      vtx->vp_xlate[0] = a->vp[MAT_TX];
+      vtx->vp_xlate[1] = a->vp[MAT_TY];
+      vtx->vp_xlate[2] = a->vp[MAT_TZ];
+      vtx->vp_xlate[3] = 0.0;
+   }
+}
+
+
+void _tnl_build_vertices( GLcontext *ctx,
+			  GLuint start,
+			  GLuint end,
+			  GLuint newinputs )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);  
+   update_input_ptrs( ctx, start );      
+   vtx->emit( ctx, end - start, 
+	      (GLubyte *)(vtx->vertex_buf + 
+			  start * vtx->vertex_size));
+}
+
+/* Emit VB vertices start..end to dest.  Note that VB vertex at
+ * postion start will be emitted to dest at position zero.
+ */
+void *_tnl_emit_vertices_to_buffer( GLcontext *ctx,
+				    GLuint start,
+				    GLuint end,
+				    void *dest )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+
+   update_input_ptrs(ctx, start);
+   /* Note: dest should not be adjusted for non-zero 'start' values:
+    */
+   vtx->emit( ctx, end - start, (GLubyte*) dest );	
+   return (void *)((GLubyte *)dest + vtx->vertex_size * (end - start));
+}
+
+/* Emit indexed VB vertices start..end to dest.  Note that VB vertex at
+ * postion start will be emitted to dest at position zero.
+ */
+
+void *_tnl_emit_indexed_vertices_to_buffer( GLcontext *ctx,
+					    const GLuint *elts,
+					    GLuint start,
+					    GLuint end,
+					    void *dest )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   GLuint oldIndex;
+   GLubyte *cdest = dest;
+
+   update_input_ptrs(ctx, oldIndex = elts[start++]);
+   vtx->emit( ctx, 1, cdest );
+   cdest += vtx->vertex_size;
+
+   for (; start < end; ++start) {
+      adjust_input_ptrs(ctx, elts[start] - oldIndex);
+      oldIndex = elts[start];
+      vtx->emit( ctx, 1, cdest);
+      cdest += vtx->vertex_size;
+   }
+
+   return (void *) cdest;
+}
+
+
+void _tnl_init_vertices( GLcontext *ctx, 
+			GLuint vb_size,
+			GLuint max_vertex_size )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);  
+
+   _tnl_install_attrs( ctx, NULL, 0, NULL, 0 );
+
+   vtx->need_extras = GL_TRUE;
+   if (max_vertex_size > vtx->max_vertex_size) {
+      _tnl_free_vertices( ctx );
+      vtx->max_vertex_size = max_vertex_size;
+      vtx->vertex_buf = (GLubyte *)ALIGN_CALLOC(vb_size * max_vertex_size, 32 );
+      invalidate_funcs(vtx);
+   }
+
+   switch(CHAN_TYPE) {
+   case GL_UNSIGNED_BYTE:
+      vtx->chan_scale[0] = 255.0;
+      vtx->chan_scale[1] = 255.0;
+      vtx->chan_scale[2] = 255.0;
+      vtx->chan_scale[3] = 255.0;
+      break;
+   case GL_UNSIGNED_SHORT:
+      vtx->chan_scale[0] = 65535.0;
+      vtx->chan_scale[1] = 65535.0;
+      vtx->chan_scale[2] = 65535.0;
+      vtx->chan_scale[3] = 65535.0;
+      break;
+   default:
+      vtx->chan_scale[0] = 1.0;
+      vtx->chan_scale[1] = 1.0;
+      vtx->chan_scale[2] = 1.0;
+      vtx->chan_scale[3] = 1.0;
+      break;
+   }
+
+   vtx->identity[0] = 0.0;
+   vtx->identity[1] = 0.0;
+   vtx->identity[2] = 0.0;
+   vtx->identity[3] = 1.0;
+
+   vtx->codegen_emit = NULL;
+
+#ifdef USE_SSE_ASM
+   if (!_mesa_getenv("MESA_NO_CODEGEN"))
+      vtx->codegen_emit = _tnl_generate_sse_emit;
+#endif
+}
+
+
+void _tnl_free_vertices( GLcontext *ctx )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   if (tnl) {
+      struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+      struct tnl_clipspace_fastpath *fp, *tmp;
+
+      if (vtx->vertex_buf) {
+         ALIGN_FREE(vtx->vertex_buf);
+         vtx->vertex_buf = NULL;
+      }
+
+      for (fp = vtx->fastpath ; fp ; fp = tmp) {
+         tmp = fp->next;
+         FREE(fp->attr);
+
+         /* KW: At the moment, fp->func is constrained to be allocated by
+          * _mesa_exec_alloc(), as the hardwired fastpaths in
+          * t_vertex_generic.c are handled specially.  It would be nice
+          * to unify them, but this probably won't change until this
+          * module gets another overhaul.
+          */
+         _mesa_exec_free((void *) fp->func);
+         FREE(fp);
+      }
+
+      vtx->fastpath = NULL;
+   }
+}
diff --git a/mesalib/src/mesa/tnl/t_vertex.h b/mesalib/src/mesa/tnl/t_vertex.h
new file mode 100644
index 000000000..2dfd7b57f
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vertex.h
@@ -0,0 +1,181 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * 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 <keithw@tungstengraphics.com>
+ */
+
+#ifndef _TNL_VERTEX_H
+#define _TNL_VERTEX_H
+
+#include "main/mtypes.h"
+#include "t_context.h"
+
+/* New mechanism to specify hardware vertices so that tnl can build
+ * and manipulate them directly.  
+ */
+
+
+/* It will probably be necessary to allow drivers to specify new
+ * emit-styles to cover all the wierd and wacky things out there.
+ */
+enum tnl_attr_format {
+   EMIT_1F,
+   EMIT_2F,
+   EMIT_3F,
+   EMIT_4F,
+   EMIT_2F_VIEWPORT,		/* do viewport transform and emit */
+   EMIT_3F_VIEWPORT,		/* do viewport transform and emit */
+   EMIT_4F_VIEWPORT,		/* do viewport transform and emit */
+   EMIT_3F_XYW,			/* for projective texture */
+   EMIT_1UB_1F,			/* for fog coordinate */
+   EMIT_3UB_3F_RGB,		/* for specular color */
+   EMIT_3UB_3F_BGR,		/* for specular color */
+   EMIT_4UB_4F_RGBA,		/* for color */
+   EMIT_4UB_4F_BGRA,		/* for color */
+   EMIT_4UB_4F_ARGB,		/* for color */
+   EMIT_4UB_4F_ABGR,		/* for color */
+   EMIT_4CHAN_4F_RGBA,		/* for swrast color */
+   EMIT_PAD,			/* leave a hole of 'offset' bytes */
+   EMIT_MAX
+};
+
+struct tnl_attr_map {
+   GLuint attrib;			/* _TNL_ATTRIB_ enum */
+   enum tnl_attr_format format;
+   GLuint offset;
+};
+
+struct tnl_format_info {
+   const char *name;
+   tnl_extract_func extract;
+   tnl_insert_func insert[4];
+   const GLuint attrsize;
+};
+
+extern const struct tnl_format_info _tnl_format_info[EMIT_MAX];
+
+
+/* Interpolate between two vertices to produce a third:
+ */
+extern void _tnl_interp( GLcontext *ctx,
+			 GLfloat t,
+			 GLuint edst, GLuint eout, GLuint ein,
+			 GLboolean force_boundary );
+
+/* Copy colors from one vertex to another:
+ */
+extern void _tnl_copy_pv(  GLcontext *ctx, GLuint edst, GLuint esrc );
+
+
+/* Extract a named attribute from a hardware vertex.  Will have to
+ * reverse any viewport transformation, swizzling or other conversions
+ * which may have been applied:
+ */
+extern void _tnl_get_attr( GLcontext *ctx, const void *vertex, GLenum attrib,
+			   GLfloat *dest );
+
+/* Complementary to the above.
+ */
+extern void _tnl_set_attr( GLcontext *ctx, void *vout, GLenum attrib, 
+			   const GLfloat *src );
+
+
+extern void *_tnl_get_vertex( GLcontext *ctx, GLuint nr );
+
+extern GLuint _tnl_install_attrs( GLcontext *ctx,
+				  const struct tnl_attr_map *map,
+				  GLuint nr, const GLfloat *vp,
+				  GLuint unpacked_size );
+
+extern void _tnl_free_vertices( GLcontext *ctx );
+
+extern void _tnl_init_vertices( GLcontext *ctx, 
+				GLuint vb_size,
+				GLuint max_vertex_size );
+
+extern void *_tnl_emit_vertices_to_buffer( GLcontext *ctx,
+					   GLuint start,
+					   GLuint end,
+					   void *dest );
+
+/* This function isn't optimal. Check out 
+ * gallium/auxilary/translate for a more comprehensive implementation of
+ * the same functionality.
+ */
+  
+extern void *_tnl_emit_indexed_vertices_to_buffer( GLcontext *ctx,
+						   const GLuint *elts,
+						   GLuint start,
+						   GLuint end,
+						   void *dest );
+
+
+extern void _tnl_build_vertices( GLcontext *ctx,
+				 GLuint start,
+				 GLuint end,
+				 GLuint newinputs );
+
+extern void _tnl_invalidate_vertices( GLcontext *ctx, GLuint newinputs );
+
+extern void _tnl_invalidate_vertex_state( GLcontext *ctx, GLuint new_state );
+
+extern void _tnl_notify_pipeline_output_change( GLcontext *ctx );
+
+
+#define GET_VERTEX_STATE(ctx)  &(TNL_CONTEXT(ctx)->clipspace)
+
+/* Internal function:
+ */
+void _tnl_register_fastpath( struct tnl_clipspace *vtx,
+			     GLboolean match_strides );
+
+
+/* t_vertex_generic.c -- Internal functions for t_vertex.c
+ */
+void _tnl_generic_copy_pv_extras( GLcontext *ctx, 
+				  GLuint dst, GLuint src );
+
+void _tnl_generic_interp_extras( GLcontext *ctx,
+				 GLfloat t,
+				 GLuint dst, GLuint out, GLuint in,
+				 GLboolean force_boundary );
+
+void _tnl_generic_copy_pv( GLcontext *ctx, GLuint edst, GLuint esrc );
+
+void _tnl_generic_interp( GLcontext *ctx,
+			  GLfloat t,
+			  GLuint edst, GLuint eout, GLuint ein,
+			  GLboolean force_boundary );
+
+void _tnl_generic_emit( GLcontext *ctx,
+			GLuint count,
+			GLubyte *v );
+
+void _tnl_generate_hardwired_emit( GLcontext *ctx );
+
+/* t_vertex_sse.c -- Internal functions for t_vertex.c
+ */
+void _tnl_generate_sse_emit( GLcontext *ctx );
+
+#endif
diff --git a/mesalib/src/mesa/tnl/t_vertex_generic.c b/mesalib/src/mesa/tnl/t_vertex_generic.c
new file mode 100644
index 000000000..9812f8c80
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vertex_generic.c
@@ -0,0 +1,1155 @@
+
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * 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 <keithw@tungstengraphics.com>
+ */
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+#include "main/simple_list.h"
+#include "t_context.h"
+#include "t_vertex.h"
+
+
+#if 0
+#define DEBUG_INSERT printf("%s\n", __FUNCTION__)
+#else
+#define DEBUG_INSERT
+#endif
+
+
+/*
+ * These functions take the NDC coordinates pointed to by 'in', apply the
+ * NDC->Viewport mapping and store the results at 'v'.
+ */
+
+static INLINE void insert_4f_viewport_4( const struct tnl_clipspace_attr *a, GLubyte *v,
+                      const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = vp[0] * in[0] + vp[12];
+   out[1] = vp[5] * in[1] + vp[13];
+   out[2] = vp[10] * in[2] + vp[14];
+   out[3] = in[3];
+}
+
+static INLINE void insert_4f_viewport_3( const struct tnl_clipspace_attr *a, GLubyte *v,
+				const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = vp[0] * in[0] + vp[12];
+   out[1] = vp[5] * in[1] + vp[13];
+   out[2] = vp[10] * in[2] + vp[14];
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_viewport_2( const struct tnl_clipspace_attr *a, GLubyte *v,
+				const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = vp[0] * in[0] + vp[12];
+   out[1] = vp[5] * in[1] + vp[13];
+   out[2] = vp[14];
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_viewport_1( const struct tnl_clipspace_attr *a, GLubyte *v,
+				const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = vp[0] * in[0] + vp[12];
+   out[1] = vp[13];
+   out[2] = vp[14];
+   out[3] = 1;
+}
+
+static INLINE void insert_3f_viewport_3( const struct tnl_clipspace_attr *a, GLubyte *v,
+				const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = vp[0] * in[0] + vp[12];
+   out[1] = vp[5] * in[1] + vp[13];
+   out[2] = vp[10] * in[2] + vp[14];
+}
+
+static INLINE void insert_3f_viewport_2( const struct tnl_clipspace_attr *a, GLubyte *v,
+				const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = vp[0] * in[0] + vp[12];
+   out[1] = vp[5] * in[1] + vp[13];
+   out[2] = vp[14];
+}
+
+static INLINE void insert_3f_viewport_1( const struct tnl_clipspace_attr *a, GLubyte *v,
+				const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = vp[0] * in[0] + vp[12];
+   out[1] = vp[13];
+   out[2] = vp[14];
+}
+
+static INLINE void insert_2f_viewport_2( const struct tnl_clipspace_attr *a, GLubyte *v,
+				const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = vp[0] * in[0] + vp[12];
+   out[1] = vp[5] * in[1] + vp[13];
+}
+
+static INLINE void insert_2f_viewport_1( const struct tnl_clipspace_attr *a, GLubyte *v,
+				const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = vp[0] * in[0] + vp[12];
+   out[1] = vp[13];
+}
+
+
+/*
+ * These functions do the same as above, except for the viewport mapping.
+ */
+
+static INLINE void insert_4f_4( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = in[3];
+}
+
+static INLINE void insert_4f_3( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_2( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static INLINE void insert_4f_1( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = 0;
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static INLINE void insert_3f_xyw_4( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[3];
+}
+
+static INLINE void insert_3f_xyw_err( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   (void) a; (void) v; (void) in;
+   DEBUG_INSERT;
+   _mesa_exit(1);
+}
+
+static INLINE void insert_3f_3( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+}
+
+static INLINE void insert_3f_2( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+}
+
+static INLINE void insert_3f_1( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = 0;
+   out[2] = 0;
+}
+
+
+static INLINE void insert_2f_2( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = in[1];
+}
+
+static INLINE void insert_2f_1( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+   out[1] = 0;
+}
+
+static INLINE void insert_1f_1( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   GLfloat *out = (GLfloat *)(v);
+   (void) a;
+   DEBUG_INSERT;
+   out[0] = in[0];
+}
+
+static INLINE void insert_null( const struct tnl_clipspace_attr *a, GLubyte *v, const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a; (void) v; (void) in;
+}
+
+static INLINE void insert_4chan_4f_rgba_4( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				  const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[3], in[3]);
+}
+
+static INLINE void insert_4chan_4f_rgba_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				  const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[2], in[2]); 
+   c[3] = CHAN_MAX;
+}
+
+static INLINE void insert_4chan_4f_rgba_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				  const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   UNCLAMPED_FLOAT_TO_CHAN(c[1], in[1]); 
+   c[2] = 0;
+   c[3] = CHAN_MAX;
+}
+
+static INLINE void insert_4chan_4f_rgba_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				  const GLfloat *in )
+{
+   GLchan *c = (GLchan *)v;
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_CHAN(c[0], in[0]); 
+   c[1] = 0;
+   c[2] = 0;
+   c[3] = CHAN_MAX;
+}
+
+static INLINE void insert_4ub_4f_rgba_4( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
+}
+
+static INLINE void insert_4ub_4f_rgba_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_rgba_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[2] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_rgba_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   v[1] = 0;
+   v[2] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_bgra_4( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[3]);
+}
+
+static INLINE void insert_4ub_4f_bgra_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_bgra_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[0] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_bgra_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   v[1] = 0;
+   v[0] = 0;
+   v[3] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_argb_4( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]);
+}
+
+static INLINE void insert_4ub_4f_argb_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[2]);
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_argb_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   v[3] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_argb_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[0]);
+   v[2] = 0x00;
+   v[3] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_abgr_4( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[3]);
+}
+
+static INLINE void insert_4ub_4f_abgr_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[2]);
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_abgr_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[1]);
+   v[1] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_4ub_4f_abgr_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[3], in[0]);
+   v[2] = 0x00;
+   v[1] = 0x00;
+   v[0] = 0xff;
+}
+
+static INLINE void insert_3ub_3f_rgb_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
+			       const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[2]);
+}
+
+static INLINE void insert_3ub_3f_rgb_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
+			       const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[2] = 0;
+}
+
+static INLINE void insert_3ub_3f_rgb_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
+			       const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+   v[1] = 0;
+   v[2] = 0;
+}
+
+static INLINE void insert_3ub_3f_bgr_3( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				 const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[2]);
+}
+
+static INLINE void insert_3ub_3f_bgr_2( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				 const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   UNCLAMPED_FLOAT_TO_UBYTE(v[1], in[1]);
+   v[0] = 0;
+}
+
+static INLINE void insert_3ub_3f_bgr_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
+				 const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[2], in[0]);
+   v[1] = 0;
+   v[0] = 0;
+}
+
+
+static INLINE void insert_1ub_1f_1( const struct tnl_clipspace_attr *a, GLubyte *v, 
+			   const GLfloat *in )
+{
+   DEBUG_INSERT;
+   (void) a;
+   UNCLAMPED_FLOAT_TO_UBYTE(v[0], in[0]);
+}
+
+
+/***********************************************************************
+ * Functions to perform the reverse operations to the above, for
+ * swrast translation and clip-interpolation.
+ * 
+ * Currently always extracts a full 4 floats.
+ */
+
+static void extract_4f_viewport( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   
+   /* Although included for completeness, the position coordinate is
+    * usually handled differently during clipping.
+    */
+   DEBUG_INSERT;
+   out[0] = (in[0] - vp[12]) / vp[0];
+   out[1] = (in[1] - vp[13]) / vp[5];
+   out[2] = (in[2] - vp[14]) / vp[10];
+   out[3] = in[3];
+}
+
+static void extract_3f_viewport( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = (in[0] - vp[12]) / vp[0];
+   out[1] = (in[1] - vp[13]) / vp[5];
+   out[2] = (in[2] - vp[14]) / vp[10];
+   out[3] = 1;
+}
+
+
+static void extract_2f_viewport( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   const GLfloat * const vp = a->vp;
+   DEBUG_INSERT;
+   out[0] = (in[0] - vp[12]) / vp[0];
+   out[1] = (in[1] - vp[13]) / vp[5];
+   out[2] = 0;
+   out[3] = 1;
+}
+
+
+static void extract_4f( const struct tnl_clipspace_attr *a, GLfloat *out, const GLubyte *v  )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = in[3];
+}
+
+static void extract_3f_xyw( const struct tnl_clipspace_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+   out[3] = in[2];
+}
+
+
+static void extract_3f( const struct tnl_clipspace_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = in[2];
+   out[3] = 1;
+}
+
+
+static void extract_2f( const struct tnl_clipspace_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = in[1];
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static void extract_1f( const struct tnl_clipspace_attr *a, GLfloat *out, const GLubyte *v )
+{
+   const GLfloat *in = (const GLfloat *)v;
+   (void) a;
+   
+   out[0] = in[0];
+   out[1] = 0;
+   out[2] = 0;
+   out[3] = 1;
+}
+
+static void extract_4chan_4f_rgba( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   GLchan *c = (GLchan *)v;
+   (void) a;
+
+   out[0] = CHAN_TO_FLOAT(c[0]);
+   out[1] = CHAN_TO_FLOAT(c[1]);
+   out[2] = CHAN_TO_FLOAT(c[2]);
+   out[3] = CHAN_TO_FLOAT(c[3]);
+}
+
+static void extract_4ub_4f_rgba( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[0] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[2] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_4ub_4f_bgra( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[2] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[0] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_4ub_4f_argb( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[3] = UBYTE_TO_FLOAT(v[0]);
+   out[0] = UBYTE_TO_FLOAT(v[1]);
+   out[1] = UBYTE_TO_FLOAT(v[2]);
+   out[2] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_4ub_4f_abgr( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				 const GLubyte *v )
+{
+   (void) a;
+   out[3] = UBYTE_TO_FLOAT(v[0]);
+   out[2] = UBYTE_TO_FLOAT(v[1]);
+   out[1] = UBYTE_TO_FLOAT(v[2]);
+   out[0] = UBYTE_TO_FLOAT(v[3]);
+}
+
+static void extract_3ub_3f_rgb( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				const GLubyte *v )
+{
+   (void) a;
+   out[0] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[2] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = 1;
+}
+
+static void extract_3ub_3f_bgr( const struct tnl_clipspace_attr *a, GLfloat *out, 
+				const GLubyte *v )
+{
+   (void) a;
+   out[2] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = UBYTE_TO_FLOAT(v[1]);
+   out[0] = UBYTE_TO_FLOAT(v[2]);
+   out[3] = 1;
+}
+
+static void extract_1ub_1f( const struct tnl_clipspace_attr *a, GLfloat *out, const GLubyte *v )
+{
+   (void) a;
+   out[0] = UBYTE_TO_FLOAT(v[0]);
+   out[1] = 0;
+   out[2] = 0;
+   out[3] = 1;
+}
+
+
+const struct tnl_format_info _tnl_format_info[EMIT_MAX] = 
+{
+   { "1f",
+     extract_1f,
+     { insert_1f_1, insert_1f_1, insert_1f_1, insert_1f_1 },
+     sizeof(GLfloat) },
+
+   { "2f",
+     extract_2f,
+     { insert_2f_1, insert_2f_2, insert_2f_2, insert_2f_2 },
+     2 * sizeof(GLfloat) },
+
+   { "3f",
+     extract_3f,
+     { insert_3f_1, insert_3f_2, insert_3f_3, insert_3f_3 },
+     3 * sizeof(GLfloat) },
+
+   { "4f",
+     extract_4f,
+     { insert_4f_1, insert_4f_2, insert_4f_3, insert_4f_4 },
+     4 * sizeof(GLfloat) },
+
+   { "2f_viewport",
+     extract_2f_viewport,
+     { insert_2f_viewport_1, insert_2f_viewport_2, insert_2f_viewport_2,
+       insert_2f_viewport_2 },
+     2 * sizeof(GLfloat) },
+
+   { "3f_viewport",
+     extract_3f_viewport,
+     { insert_3f_viewport_1, insert_3f_viewport_2, insert_3f_viewport_3,
+       insert_3f_viewport_3 },
+     3 * sizeof(GLfloat) },
+
+   { "4f_viewport",
+     extract_4f_viewport,
+     { insert_4f_viewport_1, insert_4f_viewport_2, insert_4f_viewport_3,
+       insert_4f_viewport_4 }, 
+     4 * sizeof(GLfloat) },
+
+   { "3f_xyw",
+     extract_3f_xyw,
+     { insert_3f_xyw_err, insert_3f_xyw_err, insert_3f_xyw_err, 
+       insert_3f_xyw_4 },
+     3 * sizeof(GLfloat) },
+
+   { "1ub_1f",
+     extract_1ub_1f,
+     { insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1, insert_1ub_1f_1 },
+     sizeof(GLubyte) },
+
+   { "3ub_3f_rgb",
+     extract_3ub_3f_rgb,
+     { insert_3ub_3f_rgb_1, insert_3ub_3f_rgb_2, insert_3ub_3f_rgb_3,
+       insert_3ub_3f_rgb_3 },
+     3 * sizeof(GLubyte) },
+
+   { "3ub_3f_bgr",
+     extract_3ub_3f_bgr,
+     { insert_3ub_3f_bgr_1, insert_3ub_3f_bgr_2, insert_3ub_3f_bgr_3,
+       insert_3ub_3f_bgr_3 },
+     3 * sizeof(GLubyte) },
+
+   { "4ub_4f_rgba",
+     extract_4ub_4f_rgba,
+     { insert_4ub_4f_rgba_1, insert_4ub_4f_rgba_2, insert_4ub_4f_rgba_3, 
+       insert_4ub_4f_rgba_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4ub_4f_bgra",
+     extract_4ub_4f_bgra,
+     { insert_4ub_4f_bgra_1, insert_4ub_4f_bgra_2, insert_4ub_4f_bgra_3,
+       insert_4ub_4f_bgra_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4ub_4f_argb",
+     extract_4ub_4f_argb,
+     { insert_4ub_4f_argb_1, insert_4ub_4f_argb_2, insert_4ub_4f_argb_3,
+       insert_4ub_4f_argb_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4ub_4f_abgr",
+     extract_4ub_4f_abgr,
+     { insert_4ub_4f_abgr_1, insert_4ub_4f_abgr_2, insert_4ub_4f_abgr_3,
+       insert_4ub_4f_abgr_4 },
+     4 * sizeof(GLubyte) },
+
+   { "4chan_4f_rgba",
+     extract_4chan_4f_rgba,
+     { insert_4chan_4f_rgba_1, insert_4chan_4f_rgba_2, insert_4chan_4f_rgba_3,
+       insert_4chan_4f_rgba_4 },
+     4 * sizeof(GLchan) },
+
+   { "pad",
+     NULL,
+     { NULL, NULL, NULL, NULL },
+     0 }
+
+};
+
+
+
+    
+/***********************************************************************
+ * Hardwired fastpaths for emitting whole vertices or groups of
+ * vertices
+ */
+#define EMIT5(NR, F0, F1, F2, F3, F4, NAME)				\
+static void NAME( GLcontext *ctx,					\
+		  GLuint count,						\
+		  GLubyte *v )						\
+{									\
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);			\
+   struct tnl_clipspace_attr *a = vtx->attr;				\
+   GLuint i;								\
+									\
+   for (i = 0 ; i < count ; i++, v += vtx->vertex_size) {		\
+      if (NR > 0) {							\
+	 F0( &a[0], v + a[0].vertoffset, (GLfloat *)a[0].inputptr );	\
+	 a[0].inputptr += a[0].inputstride;				\
+      }									\
+      									\
+      if (NR > 1) {							\
+	 F1( &a[1], v + a[1].vertoffset, (GLfloat *)a[1].inputptr );	\
+	 a[1].inputptr += a[1].inputstride;				\
+      }									\
+      									\
+      if (NR > 2) {							\
+	 F2( &a[2], v + a[2].vertoffset, (GLfloat *)a[2].inputptr );	\
+	 a[2].inputptr += a[2].inputstride;				\
+      }									\
+      									\
+      if (NR > 3) {							\
+	 F3( &a[3], v + a[3].vertoffset, (GLfloat *)a[3].inputptr );	\
+	 a[3].inputptr += a[3].inputstride;				\
+      }									\
+									\
+      if (NR > 4) {							\
+	 F4( &a[4], v + a[4].vertoffset, (GLfloat *)a[4].inputptr );	\
+	 a[4].inputptr += a[4].inputstride;				\
+      }									\
+   }									\
+}
+
+   
+#define EMIT2(F0, F1, NAME) EMIT5(2, F0, F1, insert_null, \
+				  insert_null, insert_null, NAME)
+
+#define EMIT3(F0, F1, F2, NAME) EMIT5(3, F0, F1, F2, insert_null, \
+				      insert_null, NAME)
+   
+#define EMIT4(F0, F1, F2, F3, NAME) EMIT5(4, F0, F1, F2, F3, \
+				          insert_null, NAME)
+   
+
+EMIT2(insert_3f_viewport_3, insert_4ub_4f_rgba_4, emit_viewport3_rgba4)
+EMIT2(insert_3f_viewport_3, insert_4ub_4f_bgra_4, emit_viewport3_bgra4)
+EMIT2(insert_3f_3, insert_4ub_4f_rgba_4, emit_xyz3_rgba4)
+
+EMIT3(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_viewport4_rgba4_st2)
+EMIT3(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2,  emit_viewport4_bgra4_st2)
+EMIT3(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, emit_xyzw4_rgba4_st2)
+
+EMIT4(insert_4f_viewport_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_viewport4_rgba4_st2_st2)
+EMIT4(insert_4f_viewport_4, insert_4ub_4f_bgra_4, insert_2f_2, insert_2f_2,  emit_viewport4_bgra4_st2_st2)
+EMIT4(insert_4f_4, insert_4ub_4f_rgba_4, insert_2f_2, insert_2f_2, emit_xyzw4_rgba4_st2_st2)
+
+
+/* Use the codegen paths to select one of a number of hardwired
+ * fastpaths.
+ */
+void _tnl_generate_hardwired_emit( GLcontext *ctx )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   tnl_emit_func func = NULL;
+
+   /* Does it fit a hardwired fastpath?  Help! this is growing out of
+    * control!
+    */
+   switch (vtx->attr_count) {
+   case 2:
+      if (vtx->attr[0].emit == insert_3f_viewport_3) {
+	 if (vtx->attr[1].emit == insert_4ub_4f_bgra_4) 
+	    func = emit_viewport3_bgra4;
+	 else if (vtx->attr[1].emit == insert_4ub_4f_rgba_4) 
+	    func = emit_viewport3_rgba4;
+      }
+      else if (vtx->attr[0].emit == insert_3f_3 &&
+	       vtx->attr[1].emit == insert_4ub_4f_rgba_4) {
+ 	 func = emit_xyz3_rgba4; 
+      }
+      break;
+   case 3:
+      if (vtx->attr[2].emit == insert_2f_2) {
+	 if (vtx->attr[1].emit == insert_4ub_4f_rgba_4) {
+	    if (vtx->attr[0].emit == insert_4f_viewport_4)
+	       func = emit_viewport4_rgba4_st2;
+	    else if (vtx->attr[0].emit == insert_4f_4) 
+	       func = emit_xyzw4_rgba4_st2;
+	 }
+	 else if (vtx->attr[1].emit == insert_4ub_4f_bgra_4 &&
+		  vtx->attr[0].emit == insert_4f_viewport_4)
+	    func = emit_viewport4_bgra4_st2;
+      }
+      break;
+   case 4:
+      if (vtx->attr[2].emit == insert_2f_2 &&
+	  vtx->attr[3].emit == insert_2f_2) {
+	 if (vtx->attr[1].emit == insert_4ub_4f_rgba_4) {
+	    if (vtx->attr[0].emit == insert_4f_viewport_4)
+	       func = emit_viewport4_rgba4_st2_st2;
+	    else if (vtx->attr[0].emit == insert_4f_4) 
+	       func = emit_xyzw4_rgba4_st2_st2;
+	 }
+	 else if (vtx->attr[1].emit == insert_4ub_4f_bgra_4 &&
+		  vtx->attr[0].emit == insert_4f_viewport_4)
+	    func = emit_viewport4_bgra4_st2_st2;
+      }
+      break;
+   }
+
+   vtx->emit = func;
+}
+
+/***********************************************************************
+ * Generic (non-codegen) functions for whole vertices or groups of
+ * vertices
+ */
+
+void _tnl_generic_emit( GLcontext *ctx,
+			GLuint count,
+			GLubyte *v )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   struct tnl_clipspace_attr *a = vtx->attr;
+   const GLuint attr_count = vtx->attr_count;
+   const GLuint stride = vtx->vertex_size;
+   GLuint i, j;
+
+   for (i = 0 ; i < count ; i++, v += stride) {
+      for (j = 0; j < attr_count; j++) {
+	 GLfloat *in = (GLfloat *)a[j].inputptr;
+	 a[j].inputptr += a[j].inputstride;
+	 a[j].emit( &a[j], v + a[j].vertoffset, in );
+      }
+   }
+}
+
+
+void _tnl_generic_interp( GLcontext *ctx,
+			    GLfloat t,
+			    GLuint edst, GLuint eout, GLuint ein,
+			    GLboolean force_boundary )
+{
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct vertex_buffer *VB = &tnl->vb;
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   const GLubyte *vin  = vtx->vertex_buf + ein  * vtx->vertex_size;
+   const GLubyte *vout = vtx->vertex_buf + eout * vtx->vertex_size;
+   GLubyte *vdst = vtx->vertex_buf + edst * vtx->vertex_size;
+   const struct tnl_clipspace_attr *a = vtx->attr;
+   const GLuint attr_count = vtx->attr_count;
+   GLuint j;
+   (void) force_boundary;
+
+   if (tnl->NeedNdcCoords) {
+      const GLfloat *dstclip = VB->ClipPtr->data[edst];
+      if (dstclip[3] != 0.0) {
+	 const GLfloat w = 1.0f / dstclip[3];
+	 GLfloat pos[4];
+
+	 pos[0] = dstclip[0] * w;
+	 pos[1] = dstclip[1] * w;
+	 pos[2] = dstclip[2] * w;
+	 pos[3] = w;
+
+	 a[0].insert[4-1]( &a[0], vdst, pos );
+      }
+   }
+   else {
+      a[0].insert[4-1]( &a[0], vdst, VB->ClipPtr->data[edst] );
+   }
+
+
+   for (j = 1; j < attr_count; j++) {
+      GLfloat fin[4], fout[4], fdst[4];
+	 
+      a[j].extract( &a[j], fin, vin + a[j].vertoffset );
+      a[j].extract( &a[j], fout, vout + a[j].vertoffset );
+
+      INTERP_F( t, fdst[3], fout[3], fin[3] );
+      INTERP_F( t, fdst[2], fout[2], fin[2] );
+      INTERP_F( t, fdst[1], fout[1], fin[1] );
+      INTERP_F( t, fdst[0], fout[0], fin[0] );
+
+      a[j].insert[4-1]( &a[j], vdst + a[j].vertoffset, fdst );
+   }
+}
+
+
+/* Extract color attributes from one vertex and insert them into
+ * another.  (Shortcircuit extract/insert with memcpy).
+ */
+void _tnl_generic_copy_pv( GLcontext *ctx, GLuint edst, GLuint esrc )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   GLubyte *vsrc = vtx->vertex_buf + esrc * vtx->vertex_size;
+   GLubyte *vdst = vtx->vertex_buf + edst * vtx->vertex_size;
+   const struct tnl_clipspace_attr *a = vtx->attr;
+   const GLuint attr_count = vtx->attr_count;
+   GLuint j;
+
+   for (j = 0; j < attr_count; j++) {
+      if (a[j].attrib == VERT_ATTRIB_COLOR0 ||
+	  a[j].attrib == VERT_ATTRIB_COLOR1) {
+
+	 _mesa_memcpy( vdst + a[j].vertoffset,
+                       vsrc + a[j].vertoffset,
+                       a[j].vertattrsize );
+      }
+   }
+}
+
+
+/* Helper functions for hardware which doesn't put back colors and/or
+ * edgeflags into vertices.
+ */
+void _tnl_generic_interp_extras( GLcontext *ctx,
+				   GLfloat t,
+				   GLuint dst, GLuint out, GLuint in,
+				   GLboolean force_boundary )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+
+   /* If stride is zero, ColorPtr[1] is constant across the VB, so
+    * there is no point interpolating between two values as they will
+    * be identical.  In all other cases, this value is generated by
+    * t_vb_lighttmp.h and has a stride of 4 dwords.
+    */
+   if (VB->ColorPtr[1] && VB->ColorPtr[1]->stride) {
+      assert(VB->ColorPtr[1]->stride == 4 * sizeof(GLfloat));
+
+      INTERP_4F( t,
+		 VB->ColorPtr[1]->data[dst],
+		 VB->ColorPtr[1]->data[out],
+		 VB->ColorPtr[1]->data[in] );
+   }
+
+   if (VB->SecondaryColorPtr[1]) {
+      assert(VB->SecondaryColorPtr[1]->stride == 4 * sizeof(GLfloat));
+      
+      INTERP_3F( t,
+		 VB->SecondaryColorPtr[1]->data[dst],
+		 VB->SecondaryColorPtr[1]->data[out],
+		 VB->SecondaryColorPtr[1]->data[in] );
+   }
+   
+   if (VB->IndexPtr[1]) {
+      VB->IndexPtr[1]->data[dst][0] = LINTERP( t,
+					       VB->IndexPtr[1]->data[out][0],
+					       VB->IndexPtr[1]->data[in][0] );
+   }
+
+   if (VB->EdgeFlag) {
+      VB->EdgeFlag[dst] = VB->EdgeFlag[out] || force_boundary;
+   }
+
+   _tnl_generic_interp(ctx, t, dst, out, in, force_boundary);
+}
+
+void _tnl_generic_copy_pv_extras( GLcontext *ctx, 
+				  GLuint dst, GLuint src )
+{
+   struct vertex_buffer *VB = &TNL_CONTEXT(ctx)->vb;
+
+   /* See above comment:
+    */
+   if (VB->ColorPtr[1] && VB->ColorPtr[1]->stride) {
+      COPY_4FV( VB->ColorPtr[1]->data[dst], 
+		VB->ColorPtr[1]->data[src] );
+   }
+
+   if (VB->SecondaryColorPtr[1]) {
+      COPY_4FV( VB->SecondaryColorPtr[1]->data[dst], 
+		VB->SecondaryColorPtr[1]->data[src] );
+   }
+
+   if (VB->IndexPtr[1]) {
+      VB->IndexPtr[1]->data[dst][0] = VB->IndexPtr[1]->data[src][0];
+   }
+
+   _tnl_generic_copy_pv(ctx, dst, src);
+}
+
+
diff --git a/mesalib/src/mesa/tnl/t_vertex_sse.c b/mesalib/src/mesa/tnl/t_vertex_sse.c
new file mode 100644
index 000000000..7a255d680
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vertex_sse.c
@@ -0,0 +1,684 @@
+/*
+ * Copyright 2003 Tungsten Graphics, inc.
+ * 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 <keithw@tungstengraphics.com>
+ */
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/colormac.h"
+#include "main/simple_list.h"
+#include "main/enums.h"
+#include "t_context.h"
+#include "t_vertex.h"
+
+#if defined(USE_SSE_ASM)
+
+#include "x86/rtasm/x86sse.h"
+#include "x86/common_x86_asm.h"
+
+
+/**
+ * Number of bytes to allocate for generated SSE functions
+ */
+#define MAX_SSE_CODE_SIZE 1024
+
+
+#define X    0
+#define Y    1
+#define Z    2
+#define W    3
+
+
+struct x86_program {
+   struct x86_function func;
+
+   GLcontext *ctx;
+   GLboolean inputs_safe;
+   GLboolean outputs_safe;
+   GLboolean have_sse2;
+   
+   struct x86_reg identity;
+   struct x86_reg chan0;
+};
+
+
+static struct x86_reg get_identity( struct x86_program *p )
+{
+   return p->identity;
+}
+
+static void emit_load4f_4( struct x86_program *p, 			   
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movups(&p->func, dest, arg0);
+}
+
+static void emit_load4f_3( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Have to jump through some hoops:
+    *
+    * c 0 0 0
+    * c 0 0 1
+    * 0 0 c 1
+    * a b c 1
+    */
+   sse_movss(&p->func, dest, x86_make_disp(arg0, 8));
+   sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) );
+   sse_shufps(&p->func, dest, dest, SHUF(Y,Z,X,W) );
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_load4f_2( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Initialize from identity, then pull in low two words:
+    */
+   sse_movups(&p->func, dest, get_identity(p));
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_load4f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Pull in low word, then swizzle in identity */
+   sse_movss(&p->func, dest, arg0);
+   sse_shufps(&p->func, dest, get_identity(p), SHUF(X,Y,Z,W) );
+}
+
+
+
+static void emit_load3f_3( struct x86_program *p, 			   
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Over-reads by 1 dword - potential SEGV if input is a vertex
+    * array.
+    */
+   if (p->inputs_safe) {
+      sse_movups(&p->func, dest, arg0);
+   } 
+   else {
+      /* c 0 0 0
+       * c c c c
+       * a b c c 
+       */
+      sse_movss(&p->func, dest, x86_make_disp(arg0, 8));
+      sse_shufps(&p->func, dest, dest, SHUF(X,X,X,X));
+      sse_movlps(&p->func, dest, arg0);
+   }
+}
+
+static void emit_load3f_2( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   emit_load4f_2(p, dest, arg0);
+}
+
+static void emit_load3f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Loading from memory erases the upper bits. */
+   sse_movss(&p->func, dest, arg0);
+}
+
+static void emit_load2f_2( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_load2f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   /* Loading from memory erases the upper bits. */
+   sse_movss(&p->func, dest, arg0);
+}
+
+static void emit_load1f_1( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movss(&p->func, dest, arg0);
+}
+
+static void (*load[4][4])( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 ) = {
+   { emit_load1f_1, 
+     emit_load1f_1, 
+     emit_load1f_1, 
+     emit_load1f_1 },
+
+   { emit_load2f_1, 
+     emit_load2f_2, 
+     emit_load2f_2, 
+     emit_load2f_2 },
+
+   { emit_load3f_1, 
+     emit_load3f_2, 
+     emit_load3f_3, 
+     emit_load3f_3 },
+
+   { emit_load4f_1, 
+     emit_load4f_2, 
+     emit_load4f_3, 
+     emit_load4f_4 } 
+};
+
+static void emit_load( struct x86_program *p,
+		       struct x86_reg dest,
+		       GLuint sz,
+		       struct x86_reg src,
+		       GLuint src_sz)
+{
+   load[sz-1][src_sz-1](p, dest, src);
+}
+
+static void emit_store4f( struct x86_program *p, 			   
+			  struct x86_reg dest,
+			  struct x86_reg arg0 )
+{
+   sse_movups(&p->func, dest, arg0);
+}
+
+static void emit_store3f( struct x86_program *p, 
+			  struct x86_reg dest,
+			  struct x86_reg arg0 )
+{
+   if (p->outputs_safe) {
+      /* Emit the extra dword anyway.  This may hurt writecombining,
+       * may cause other problems.
+       */
+      sse_movups(&p->func, dest, arg0);
+   }
+   else {
+      /* Alternate strategy - emit two, shuffle, emit one.
+       */
+      sse_movlps(&p->func, dest, arg0);
+      sse_shufps(&p->func, arg0, arg0, SHUF(Z,Z,Z,Z) ); /* NOTE! destructive */
+      sse_movss(&p->func, x86_make_disp(dest,8), arg0);
+   }
+}
+
+static void emit_store2f( struct x86_program *p, 
+			   struct x86_reg dest,
+			   struct x86_reg arg0 )
+{
+   sse_movlps(&p->func, dest, arg0);
+}
+
+static void emit_store1f( struct x86_program *p, 
+			  struct x86_reg dest,
+			  struct x86_reg arg0 )
+{
+   sse_movss(&p->func, dest, arg0);
+}
+
+
+static void (*store[4])( struct x86_program *p, 
+			 struct x86_reg dest,
+			 struct x86_reg arg0 ) = 
+{
+   emit_store1f, 
+   emit_store2f, 
+   emit_store3f, 
+   emit_store4f 
+};
+
+static void emit_store( struct x86_program *p,
+			struct x86_reg dest,
+			GLuint sz,
+			struct x86_reg temp )
+
+{
+   store[sz-1](p, dest, temp);
+}
+
+static void emit_pack_store_4ub( struct x86_program *p,
+				 struct x86_reg dest,
+				 struct x86_reg temp )
+{
+   /* Scale by 255.0
+    */
+   sse_mulps(&p->func, temp, p->chan0);
+
+   if (p->have_sse2) {
+      sse2_cvtps2dq(&p->func, temp, temp);
+      sse2_packssdw(&p->func, temp, temp);
+      sse2_packuswb(&p->func, temp, temp);
+      sse_movss(&p->func, dest, temp);
+   }
+   else {
+      struct x86_reg mmx0 = x86_make_reg(file_MMX, 0);
+      struct x86_reg mmx1 = x86_make_reg(file_MMX, 1);
+      sse_cvtps2pi(&p->func, mmx0, temp);
+      sse_movhlps(&p->func, temp, temp);
+      sse_cvtps2pi(&p->func, mmx1, temp);
+      mmx_packssdw(&p->func, mmx0, mmx1);
+      mmx_packuswb(&p->func, mmx0, mmx0);
+      mmx_movd(&p->func, dest, mmx0);
+   }
+}
+
+static GLint get_offset( const void *a, const void *b )
+{
+   return (const char *)b - (const char *)a;
+}
+
+/* Not much happens here.  Eventually use this function to try and
+ * avoid saving/reloading the source pointers each vertex (if some of
+ * them can fit in registers).
+ */
+static void get_src_ptr( struct x86_program *p,
+			 struct x86_reg srcREG,
+			 struct x86_reg vtxREG,
+			 struct tnl_clipspace_attr *a )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(p->ctx);
+   struct x86_reg ptr_to_src = x86_make_disp(vtxREG, get_offset(vtx, &a->inputptr));
+
+   /* Load current a[j].inputptr
+    */
+   x86_mov(&p->func, srcREG, ptr_to_src);
+}
+
+static void update_src_ptr( struct x86_program *p,
+			 struct x86_reg srcREG,
+			 struct x86_reg vtxREG,
+			 struct tnl_clipspace_attr *a )
+{
+   if (a->inputstride) {
+      struct tnl_clipspace *vtx = GET_VERTEX_STATE(p->ctx);
+      struct x86_reg ptr_to_src = x86_make_disp(vtxREG, get_offset(vtx, &a->inputptr));
+
+      /* add a[j].inputstride (hardcoded value - could just as easily
+       * pull the stride value from memory each time).
+       */
+      x86_lea(&p->func, srcREG, x86_make_disp(srcREG, a->inputstride));
+      
+      /* save new value of a[j].inputptr 
+       */
+      x86_mov(&p->func, ptr_to_src, srcREG);
+   }
+}
+
+
+/* Lots of hardcoding
+ *
+ * EAX -- pointer to current output vertex
+ * ECX -- pointer to current attribute 
+ * 
+ */
+static GLboolean build_vertex_emit( struct x86_program *p )
+{
+   GLcontext *ctx = p->ctx;
+   TNLcontext *tnl = TNL_CONTEXT(ctx);
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   GLuint j = 0;
+
+   struct x86_reg vertexEAX = x86_make_reg(file_REG32, reg_AX);
+   struct x86_reg srcECX = x86_make_reg(file_REG32, reg_CX);
+   struct x86_reg countEBP = x86_make_reg(file_REG32, reg_BP);
+   struct x86_reg vtxESI = x86_make_reg(file_REG32, reg_SI);
+   struct x86_reg temp = x86_make_reg(file_XMM, 0);
+   struct x86_reg vp0 = x86_make_reg(file_XMM, 1);
+   struct x86_reg vp1 = x86_make_reg(file_XMM, 2);
+   struct x86_reg temp2 = x86_make_reg(file_XMM, 3);
+   GLubyte *fixup, *label;
+
+   /* Push a few regs?
+    */
+   x86_push(&p->func, countEBP);
+   x86_push(&p->func, vtxESI);
+
+
+   /* Get vertex count, compare to zero
+    */
+   x86_xor(&p->func, srcECX, srcECX);
+   x86_mov(&p->func, countEBP, x86_fn_arg(&p->func, 2));
+   x86_cmp(&p->func, countEBP, srcECX);
+   fixup = x86_jcc_forward(&p->func, cc_E);
+
+   /* Initialize destination register. 
+    */
+   x86_mov(&p->func, vertexEAX, x86_fn_arg(&p->func, 3));
+
+   /* Dereference ctx to get tnl, then vtx:
+    */
+   x86_mov(&p->func, vtxESI, x86_fn_arg(&p->func, 1));
+   x86_mov(&p->func, vtxESI, x86_make_disp(vtxESI, get_offset(ctx, &ctx->swtnl_context)));
+   vtxESI = x86_make_disp(vtxESI, get_offset(tnl, &tnl->clipspace));
+
+   
+   /* Possibly load vp0, vp1 for viewport calcs:
+    */
+   if (vtx->need_viewport) {
+      sse_movups(&p->func, vp0, x86_make_disp(vtxESI, get_offset(vtx, &vtx->vp_scale[0])));
+      sse_movups(&p->func, vp1, x86_make_disp(vtxESI, get_offset(vtx, &vtx->vp_xlate[0])));
+   }
+
+   /* always load, needed or not:
+    */
+   sse_movups(&p->func, p->chan0, x86_make_disp(vtxESI, get_offset(vtx, &vtx->chan_scale[0])));
+   sse_movups(&p->func, p->identity, x86_make_disp(vtxESI, get_offset(vtx, &vtx->identity[0])));
+
+   /* Note address for loop jump */
+   label = x86_get_label(&p->func);
+
+   /* Emit code for each of the attributes.  Currently routes
+    * everything through SSE registers, even when it might be more
+    * efficient to stick with regular old x86.  No optimization or
+    * other tricks - enough new ground to cover here just getting
+    * things working.
+    */
+   while (j < vtx->attr_count) {
+      struct tnl_clipspace_attr *a = &vtx->attr[j];
+      struct x86_reg dest = x86_make_disp(vertexEAX, a->vertoffset);
+
+      /* Now, load an XMM reg from src, perhaps transform, then save.
+       * Could be shortcircuited in specific cases:
+       */
+      switch (a->format) {
+      case EMIT_1F:
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize);
+	 emit_store(p, dest, 1, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_2F:
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize);
+	 emit_store(p, dest, 2, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_3F:
+	 /* Potentially the worst case - hardcode 2+1 copying:
+	  */
+	 if (0) {
+	    get_src_ptr(p, srcECX, vtxESI, a);
+	    emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	    emit_store(p, dest, 3, temp);
+	    update_src_ptr(p, srcECX, vtxESI, a);
+	 }
+	 else {
+	    get_src_ptr(p, srcECX, vtxESI, a);
+	    emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize);
+	    emit_store(p, dest, 2, temp);
+	    if (a->inputsize > 2) {
+	       emit_load(p, temp, 1, x86_make_disp(srcECX, 8), 1);
+	       emit_store(p, x86_make_disp(dest,8), 1, temp);
+	    }
+	    else {
+	       sse_movss(&p->func, x86_make_disp(dest,8), get_identity(p));
+	    }
+	    update_src_ptr(p, srcECX, vtxESI, a);
+	 }
+	 break;
+      case EMIT_4F:
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 emit_store(p, dest, 4, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_2F_VIEWPORT: 
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 2, x86_deref(srcECX), a->inputsize);
+	 sse_mulps(&p->func, temp, vp0);
+	 sse_addps(&p->func, temp, vp1);
+	 emit_store(p, dest, 2, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_3F_VIEWPORT: 
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	 sse_mulps(&p->func, temp, vp0);
+	 sse_addps(&p->func, temp, vp1);
+	 emit_store(p, dest, 3, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_4F_VIEWPORT: 
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_mulps(&p->func, temp, vp0);
+	 sse_addps(&p->func, temp, vp1);
+	 emit_store(p, dest, 4, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_3F_XYW:
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(X,Y,W,Z));
+	 emit_store(p, dest, 3, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+
+      case EMIT_1UB_1F:	 
+	 /* Test for PAD3 + 1UB:
+	  */
+	 if (j > 0 &&
+	     a[-1].vertoffset + a[-1].vertattrsize <= a->vertoffset - 3)
+	 {
+	    get_src_ptr(p, srcECX, vtxESI, a);
+	    emit_load(p, temp, 1, x86_deref(srcECX), a->inputsize);
+	    sse_shufps(&p->func, temp, temp, SHUF(X,X,X,X));
+	    emit_pack_store_4ub(p, x86_make_disp(dest, -3), temp); /* overkill! */
+	    update_src_ptr(p, srcECX, vtxESI, a);
+	 }
+	 else {
+	    _mesa_printf("Can't emit 1ub %x %x %d\n", a->vertoffset, a[-1].vertoffset, a[-1].vertattrsize );
+	    return GL_FALSE;
+	 }
+	 break;
+      case EMIT_3UB_3F_RGB:
+      case EMIT_3UB_3F_BGR:
+	 /* Test for 3UB + PAD1:
+	  */
+	 if (j == vtx->attr_count - 1 ||
+	     a[1].vertoffset >= a->vertoffset + 4) {
+	    get_src_ptr(p, srcECX, vtxESI, a);
+	    emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	    if (a->format == EMIT_3UB_3F_BGR)
+	       sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W));
+	    emit_pack_store_4ub(p, dest, temp);
+	    update_src_ptr(p, srcECX, vtxESI, a);
+	 }
+	 /* Test for 3UB + 1UB:
+	  */
+	 else if (j < vtx->attr_count - 1 &&
+		  a[1].format == EMIT_1UB_1F &&
+		  a[1].vertoffset == a->vertoffset + 3) {
+	    get_src_ptr(p, srcECX, vtxESI, a);
+	    emit_load(p, temp, 3, x86_deref(srcECX), a->inputsize);
+	    update_src_ptr(p, srcECX, vtxESI, a);
+
+	    /* Make room for incoming value:
+	     */
+	    sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z));
+
+	    get_src_ptr(p, srcECX, vtxESI, &a[1]);
+	    emit_load(p, temp2, 1, x86_deref(srcECX), a[1].inputsize);
+	    sse_movss(&p->func, temp, temp2);
+	    update_src_ptr(p, srcECX, vtxESI, &a[1]);
+
+	    /* Rearrange and possibly do BGR conversion:
+	     */
+	    if (a->format == EMIT_3UB_3F_BGR)
+	       sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X));
+	    else
+	       sse_shufps(&p->func, temp, temp, SHUF(Y,Z,W,X));
+
+	    emit_pack_store_4ub(p, dest, temp);
+	    j++;		/* NOTE: two attrs consumed */
+	 }
+	 else {
+	    _mesa_printf("Can't emit 3ub\n");
+	    return GL_FALSE;	/* add this later */
+	 }
+	 break;
+
+      case EMIT_4UB_4F_RGBA:
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_4UB_4F_BGRA:
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(Z,Y,X,W));
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_4UB_4F_ARGB:
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(W,X,Y,Z));
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_4UB_4F_ABGR:
+	 get_src_ptr(p, srcECX, vtxESI, a);
+	 emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	 sse_shufps(&p->func, temp, temp, SHUF(W,Z,Y,X));
+	 emit_pack_store_4ub(p, dest, temp);
+	 update_src_ptr(p, srcECX, vtxESI, a);
+	 break;
+      case EMIT_4CHAN_4F_RGBA:
+	 switch (CHAN_TYPE) {
+	 case GL_UNSIGNED_BYTE:
+	    get_src_ptr(p, srcECX, vtxESI, a);
+	    emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	    emit_pack_store_4ub(p, dest, temp);
+	    update_src_ptr(p, srcECX, vtxESI, a);
+	    break;
+	 case GL_FLOAT:
+	    get_src_ptr(p, srcECX, vtxESI, a);
+	    emit_load(p, temp, 4, x86_deref(srcECX), a->inputsize);
+	    emit_store(p, dest, 4, temp);
+	    update_src_ptr(p, srcECX, vtxESI, a);
+	    break;
+	 case GL_UNSIGNED_SHORT:
+	 default:
+	    _mesa_printf("unknown CHAN_TYPE %s\n", _mesa_lookup_enum_by_nr(CHAN_TYPE));
+	    return GL_FALSE;
+	 }
+	 break;
+      default:
+	 _mesa_printf("unknown a[%d].format %d\n", j, a->format);
+	 return GL_FALSE;	/* catch any new opcodes */
+      }
+      
+      /* Increment j by at least 1 - may have been incremented above also:
+       */
+      j++;
+   }
+
+   /* Next vertex:
+    */
+   x86_lea(&p->func, vertexEAX, x86_make_disp(vertexEAX, vtx->vertex_size));
+
+   /* decr count, loop if not zero
+    */
+   x86_dec(&p->func, countEBP);
+   x86_test(&p->func, countEBP, countEBP); 
+   x86_jcc(&p->func, cc_NZ, label);
+
+   /* Exit mmx state?
+    */
+   if (p->func.need_emms)
+      mmx_emms(&p->func);
+
+   /* Land forward jump here:
+    */
+   x86_fixup_fwd_jump(&p->func, fixup);
+
+   /* Pop regs and return
+    */
+   x86_pop(&p->func, x86_get_base_reg(vtxESI));
+   x86_pop(&p->func, countEBP);
+   x86_ret(&p->func);
+
+   assert(!vtx->emit);
+   vtx->emit = (tnl_emit_func)x86_get_func(&p->func);
+
+   assert( (char *) p->func.csr - (char *) p->func.store <= MAX_SSE_CODE_SIZE );
+   return GL_TRUE;
+}
+
+
+
+void _tnl_generate_sse_emit( GLcontext *ctx )
+{
+   struct tnl_clipspace *vtx = GET_VERTEX_STATE(ctx);
+   struct x86_program p;   
+
+   if (!cpu_has_xmm) {
+      vtx->codegen_emit = NULL;
+      return;
+   }
+
+   _mesa_memset(&p, 0, sizeof(p));
+
+   p.ctx = ctx;
+   p.inputs_safe = 0;		/* for now */
+   p.outputs_safe = 0;		/* for now */
+   p.have_sse2 = cpu_has_xmm2;
+   p.identity = x86_make_reg(file_XMM, 6);
+   p.chan0 = x86_make_reg(file_XMM, 7);
+
+   if (!x86_init_func_size(&p.func, MAX_SSE_CODE_SIZE)) {
+      vtx->emit = NULL;
+      return;
+   }
+
+   if (build_vertex_emit(&p)) {
+      _tnl_register_fastpath( vtx, GL_TRUE );
+   }
+   else {
+      /* Note the failure so that we don't keep trying to codegen an
+       * impossible state:
+       */
+      _tnl_register_fastpath( vtx, GL_FALSE );
+      x86_release_func(&p.func);
+   }
+}
+
+#else
+
+void _tnl_generate_sse_emit( GLcontext *ctx )
+{
+   /* Dummy version for when USE_SSE_ASM not defined */
+}
+
+#endif
diff --git a/mesalib/src/mesa/tnl/t_vp_build.c b/mesalib/src/mesa/tnl/t_vp_build.c
new file mode 100644
index 000000000..735937bbe
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vp_build.c
@@ -0,0 +1,60 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ *
+ * Copyright (C) 2007  Tungsten Graphics   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
+ * TUNGSTEN GRAPHICS 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_vp_build.c
+ * Create a vertex program to execute the current fixed function T&L pipeline.
+ * \author Keith Whitwell
+ */
+
+
+#include "main/glheader.h"
+#include "main/ffvertex_prog.h"
+#include "main/dd.h"
+#include "t_vp_build.h"
+
+
+/**
+ * XXX This should go away someday, but still referenced by some drivers...
+ */
+void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx )
+{
+   const struct gl_vertex_program *prev = ctx->VertexProgram._Current;
+
+   if (!ctx->VertexProgram._Current ||
+       ctx->VertexProgram._Current == ctx->VertexProgram._TnlProgram) {
+      ctx->VertexProgram._Current
+         = ctx->VertexProgram._TnlProgram
+         = _mesa_get_fixed_func_vertex_program(ctx);
+   }
+
+   /* Tell the driver about the change.  Could define a new target for
+    * this?
+    */
+   if (ctx->VertexProgram._Current != prev && ctx->Driver.BindProgram) {
+      ctx->Driver.BindProgram(ctx, GL_VERTEX_PROGRAM_ARB,
+                            (struct gl_program *) ctx->VertexProgram._Current);
+   }
+}
diff --git a/mesalib/src/mesa/tnl/t_vp_build.h b/mesalib/src/mesa/tnl/t_vp_build.h
new file mode 100644
index 000000000..d6ebc66c0
--- /dev/null
+++ b/mesalib/src/mesa/tnl/t_vp_build.h
@@ -0,0 +1,42 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  6.3
+ *
+ * Copyright (C) 2005  Tungsten Graphics   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
+ * TUNGSTEN GRAPHICS 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_VP_BUILD_H
+#define T_VP_BUILD_H
+
+#include "main/mtypes.h"
+
+#define TNL_FIXED_FUNCTION_STATE_FLAGS (_NEW_PROGRAM |		\
+					_NEW_LIGHT |		\
+					_NEW_TEXTURE |		\
+					_NEW_TEXTURE_MATRIX |		\
+					_NEW_TRANSFORM |	\
+					_NEW_FOG |		\
+					_NEW_POINT)
+
+extern void _tnl_UpdateFixedFunctionProgram( GLcontext *ctx );
+
+#endif
diff --git a/mesalib/src/mesa/tnl/tnl.h b/mesalib/src/mesa/tnl/tnl.h
new file mode 100644
index 000000000..9c66d3b01
--- /dev/null
+++ b/mesalib/src/mesa/tnl/tnl.h
@@ -0,0 +1,100 @@
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.1
+ *
+ * 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"),
+ * 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 _TNL_H
+#define _TNL_H
+
+#include "main/mtypes.h"
+
+
+
+/* These are the public-access functions exported from tnl.  (A few
+ * more are currently hooked into dispatch directly by the module
+ * itself.)
+ */
+extern GLboolean
+_tnl_CreateContext( GLcontext *ctx );
+
+extern void
+_tnl_DestroyContext( GLcontext *ctx );
+
+extern void
+_tnl_InvalidateState( GLcontext *ctx, GLuint new_state );
+
+/* Functions to revive the tnl module after being unhooked from
+ * dispatch and/or driver callbacks.
+ */
+
+extern void
+_tnl_wakeup( GLcontext *ctx );
+
+/* Driver configuration options:
+ */
+extern void
+_tnl_need_projected_coords( GLcontext *ctx, GLboolean flag );
+
+
+/* Control whether T&L does per-vertex fog
+ */
+extern void
+_tnl_allow_vertex_fog( GLcontext *ctx, GLboolean value );
+
+extern void
+_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);
+
+void
+_tnl_vbo_draw_prims( GLcontext *ctx,
+		     const struct gl_client_array *arrays[],
+		     const struct _mesa_prim *prim,
+		     GLuint nr_prims,
+		     const struct _mesa_index_buffer *ib,
+		     GLboolean index_bounds_valid,
+		     GLuint min_index,
+		     GLuint max_index);
+
+extern void
+_mesa_load_tracked_matrices(GLcontext *ctx);
+
+extern void
+_tnl_RasterPos(GLcontext *ctx, const GLfloat vObj[4]);
+
+#endif