X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fmesa%2Ftnl%2Ft_vb_program.c;h=f99401ca6d870f09657ae26c72d4afb82fe7ce1b;hb=87a6fe57e85ea5f273e808211a0e61d2b837b8e2;hp=4b1d2ff00e2987183919d354434b0135d4ac6224;hpb=071a4ddd92229b102b0b67410303505681114cd7;p=mesa.git

diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c
index 4b1d2ff00e2..f99401ca6d8 100644
--- a/src/mesa/tnl/t_vb_program.c
+++ b/src/mesa/tnl/t_vb_program.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  5.1
+ * Version:  7.1
  *
- * Copyright (C) 1999-2003  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -25,53 +25,26 @@
 
 /**
  * \file tnl/t_vb_program.c
- * \brief Pipeline stage for executing NVIDIA vertex programs.
+ * \brief Pipeline stage for executing vertex programs.
  * \author Brian Paul,  Keith Whitwell
  */
 
 
-#include "glheader.h"
-#include "api_noop.h"
-#include "colormac.h"
-#include "context.h"
-#include "dlist.h"
-#include "hash.h"
-#include "light.h"
-#include "macros.h"
-#include "imports.h"
-#include "simple_list.h"
-#include "mtypes.h"
-#include "nvvertprog.h"
-#include "nvvertexec.h"
-#include "nvprogram.h"
+#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 "math/m_translate.h"
+#include "tnl/tnl.h"
+#include "tnl/t_context.h"
+#include "tnl/t_pipeline.h"
 
-#include "t_context.h"
-#include "t_pipeline.h"
-#include "t_imm_api.h"
-#include "t_imm_exec.h"
-
-
-/**
- * \warning These values _MUST_ match the values in the OutputRegisters[]
- * array in vpparse.c!!!
- */
-#define VERT_RESULT_HPOS 0
-#define VERT_RESULT_COL0 1
-#define VERT_RESULT_COL1 2
-#define VERT_RESULT_BFC0 3
-#define VERT_RESULT_BFC1 4
-#define VERT_RESULT_FOGC 5
-#define VERT_RESULT_PSIZ 6
-#define VERT_RESULT_TEX0 7
-#define VERT_RESULT_TEX1 8
-#define VERT_RESULT_TEX2 9
-#define VERT_RESULT_TEX3 10
-#define VERT_RESULT_TEX4 11
-#define VERT_RESULT_TEX5 12
-#define VERT_RESULT_TEX6 13
-#define VERT_RESULT_TEX7 14
 
 
 /*!
@@ -79,11 +52,7 @@
  */
 struct vp_stage_data {
    /** The results of running the vertex program go into these arrays. */
-   GLvector4f attribs[15];
-
-   /* These point to the attribs[VERT_RESULT_COL0, COL1, BFC0, BFC1] arrays */
-   struct gl_client_array color0[2];  /**< diffuse front and back */
-   struct gl_client_array color1[2];  /**< specular front and back */
+   GLvector4f results[VERT_RESULT_MAX];
 
    GLvector4f ndcCoords;              /**< normalized device coords */
    GLubyte *clipmask;                 /**< clip flags */
@@ -94,24 +63,274 @@ struct vp_stage_data {
 #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])
+{
+   GLchan rgba[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, &rgba);
+   color[0] = CHAN_TO_FLOAT(rgba[0]);
+   color[1] = CHAN_TO_FLOAT(rgba[1]);
+   color[2] = CHAN_TO_FLOAT(rgba[2]);
+   color[3] = CHAN_TO_FLOAT(rgba[3]);
+}
+
+
+/**
+ * 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_PROGRAM_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 gl_pipeline_stage *stage )
+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 vp_machine *machine = &(ctx->VertexProgram.Machine);
-   struct vertex_program *program = ctx->VertexProgram.Current;
-   GLuint i;
+   struct gl_vertex_program *program = ctx->VertexProgram._Current;
+   struct gl_program_machine machine;
+   GLuint outputs[VERT_RESULT_MAX], numOutputs;
+   GLuint i, j;
 
-   _mesa_init_tracked_matrices(ctx); /* load registers with matrices */
-   _mesa_init_vp_registers(ctx);     /* init temp and result regs */
+   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],
@@ -130,187 +349,115 @@ static GLboolean run_vp( GLcontext *ctx, struct gl_pipeline_stage *stage )
              VB->AttribPtr[2]->data[i][3]);
 #endif
 
-      /* load the input attribute registers */
-      if (VB->Flag) {
-         /* the traditional glBegin/glVertex/glEnd case */
-         for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
-            if (attr == 0 || (program->InputsRead & (1 << attr))) {
-               COPY_4V(machine->Registers[VP_INPUT_REG_START + attr],
-                       VB->AttribPtr[attr]->data[i]);
-            }
-         }
-      }
-      else {
-         /* the vertex array case */
-         for (attr = 0; attr < VERT_ATTRIB_MAX; attr++) {
-            if (program->InputsRead & (1 << attr)) {
-               const GLubyte *ptr = (const GLubyte*) VB->AttribPtr[attr]->data;
-               const GLuint stride = VB->AttribPtr[attr]->stride;
-               const GLfloat *data = (GLfloat *) (ptr + stride * i);
-               COPY_4V(machine->Registers[VP_INPUT_REG_START + attr], data);
-               /*ASSERT(VB->AttribPtr[attr]->size == 4);*/
-               ASSERT(stride == 4 * sizeof(GLfloat) || stride == 0);
-            }
-         }
+      /* 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);
+	    COPY_CLEAN_4V(machine.VertAttribs[attr], size, data);
+	 }
       }
 
       /* execute the program */
-      ASSERT(program);
-      _mesa_exec_vertex_program(ctx, 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];
+         COPY_4V(store->results[attr].data[i], machine.Outputs[attr]);
+      }
 #if 0
-      printf("Output %d: %f, %f, %f, %f\n", i,
-             machine->Registers[VP_OUTPUT_REG_START + 0][0],
-             machine->Registers[VP_OUTPUT_REG_START + 0][1],
-             machine->Registers[VP_OUTPUT_REG_START + 0][2],
-             machine->Registers[VP_OUTPUT_REG_START + 0][3]);
-      printf("   color: %f, %f, %f, %f\n",
-             machine->Registers[VP_OUTPUT_REG_START +_1][0],
-             machine->Registers[VP_OUTPUT_REG_START + 1][1],
-             machine->Registers[VP_OUTPUT_REG_START + 1][2],
-             machine->Registers[VP_OUTPUT_REG_START + 1][3]);
-      printf("PointSize[%d]: %g\n", i,
-             machine->Registers[VP_OUTPUT_REG_START + VERT_RESULT_PSIZ][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 an point size results if needed */
+   /* Fixup fog and point size results if needed */
+   if (program->IsNVProgram) {
       if (ctx->Fog.Enabled &&
-          (program->OutputsWritten & (1 << VERT_RESULT_FOGC)) == 0) {
-         machine->Registers[VP_OUTPUT_REG_START + VERT_RESULT_FOGC][0] = 1.0;
+          (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->OutputsWritten & (1 << VERT_RESULT_PSIZ)) == 0) {
-         machine->Registers[VP_OUTPUT_REG_START + VERT_RESULT_PSIZ][0]
-            = ctx->Point.Size;
-      }
-
-      /* copy the output registers into the VB->attribs arrays */
-      /* XXX (optimize) could use a conditional and smaller loop limit here */
-      for (attr = 0; attr < 15; attr++) {
-         COPY_4V( store->attribs[attr].data[i],
-                  machine->Registers[VP_OUTPUT_REG_START + attr] );
+          (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;
+         }
       }
    }
 
-   /* Setup the VB pointers so that the next pipeline stages get
-    * their data from the right place (the program output arrays).
-    */
-   VB->ClipPtr = &store->attribs[VERT_RESULT_HPOS];
-   VB->ClipPtr->size = 4;
-   VB->ClipPtr->count = VB->Count;
-   VB->ColorPtr[0] = &store->color0[0];
-   VB->ColorPtr[1] = &store->color0[1];
-   VB->SecondaryColorPtr[0] = &store->color1[0];
-   VB->SecondaryColorPtr[1] = &store->color1[1];
-   VB->FogCoordPtr = &store->attribs[VERT_RESULT_FOGC];
-   VB->PointSizePtr = &store->attribs[VERT_RESULT_PSIZ];
-   for (i = 0; i < ctx->Const.MaxTextureUnits; i++)
-      VB->TexCoordPtr[i] = &store->attribs[VERT_RESULT_TEX0 + i];
-
-   /* Cliptest and perspective divide.  Clip functions must clear
-    * the clipmask.
-    */
-   store->ormask = 0;
-   store->andmask = CLIP_ALL_BITS;
+   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] );
 
-   if (tnl->NeedNdcCoords) {
-      VB->NdcPtr =
-         _mesa_clip_tab[VB->ClipPtr->size]( VB->ClipPtr,
-                                            &store->ndcCoords,
-                                            store->clipmask,
-                                            &store->ormask,
-                                            &store->andmask );
+      /* Drivers expect this to be clean to element 4...
+       */
+      switch (VB->ClipPtr->size) {
+      case 1:
+	 /* impossible */
+      case 2:
+	 _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 2 );
+	 /* fall-through */
+      case 3:
+	 _mesa_vector4f_clean_elem( VB->ClipPtr, VB->Count, 3 );
+	 /* fall-through */
+      case 4:
+	 break;
+      }
    }
    else {
-      VB->NdcPtr = 0;
-      _mesa_clip_np_tab[VB->ClipPtr->size]( VB->ClipPtr,
-                                            0,
-                                            store->clipmask,
-                                            &store->ormask,
-                                            &store->andmask );
+      /* 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;
    }
 
-   if (store->andmask)  /* All vertices are outside the frustum */
-      return GL_FALSE;
-
-
-   /* This is where we'd do clip testing against the user-defined
-    * clipping planes, but they're not supported by vertex programs.
-    */
-
-   VB->ClipOrMask = store->ormask;
-   VB->ClipMask = store->clipmask;
-
-   /* XXXX what's this?
-   if (VB->ClipPtr == VB->ObjPtr && (VB->importable_data & VERT_BIT_POS))
-      VB->importable_data |= VERT_BIT_CLIP;
-   */
-
-   return GL_TRUE;
-}
-
+   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];
+   }
 
-/**
- * This function validates stuff.
- */
-static GLboolean run_validate_program( GLcontext *ctx,
-					struct gl_pipeline_stage *stage )
-{
-#if 000
-   /* XXX do we need any validation for vertex programs? */
-   GLuint ind = 0;
-   light_func *tab;
-
-   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;
+   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];
       }
    }
-   else
-      tab = _tnl_light_ci_tab;
-
-   if (ctx->Light.ColorMaterialEnabled)
-      ind |= LIGHT_COLORMATERIAL;
-
-   if (ctx->Light.Model.TwoSide)
-      ind |= LIGHT_TWOSIDE;
 
-   VP_STAGE_DATA(stage)->light_func_tab = &tab[ind];
-
-   /* This and the above should only be done on _NEW_LIGHT:
-    */
-   _mesa_validate_all_lighting_tables( ctx );
-#endif
 
-   /* Now run the stage...
+   /* Perform NDC and cliptest operations:
     */
-   stage->run = run_vp;
-   return stage->run( ctx, stage );
-}
-
-
-/**
- * Initialize a gl_client_array to point into a GLvector4f color vector.
- */
-static void init_color_array( struct gl_client_array *a, GLvector4f *vec )
-{
-   a->Ptr = vec->data;
-   a->Size = 4;
-   a->Type = GL_FLOAT;
-   a->Stride = 0;
-   a->StrideB = sizeof(GLfloat) * 4;
-   a->Enabled = 0;
-   a->Flags = 0;
+   return do_ndc_cliptest(ctx, store);
 }
 
 
@@ -318,8 +465,8 @@ static void init_color_array( struct gl_client_array *a, GLvector4f *vec )
  * Called the first time stage->run is called.  In effect, don't
  * allocate data until the first time the stage is run.
  */
-static GLboolean run_init_vp( GLcontext *ctx,
-                              struct gl_pipeline_stage *stage )
+static GLboolean
+init_vp(GLcontext *ctx, struct tnl_pipeline_stage *stage)
 {
    TNLcontext *tnl = TNL_CONTEXT(ctx);
    struct vertex_buffer *VB = &(tnl->vb);
@@ -333,66 +480,24 @@ static GLboolean run_init_vp( GLcontext *ctx,
       return GL_FALSE;
 
    /* Allocate arrays of vertex output values */
-   for (i = 0; i < 15; i++) {
-      _mesa_vector4f_alloc( &store->attribs[i], 0, size, 32 );
-      store->attribs[i].size = 4;
+   for (i = 0; i < VERT_RESULT_MAX; i++) {
+      _mesa_vector4f_alloc( &store->results[i], 0, size, 32 );
+      store->results[i].size = 4;
    }
 
-   /* Make the color0[] and color1[] arrays point into the attribs[] arrays */
-   init_color_array( &store->color0[0], &store->attribs[VERT_RESULT_COL0] );
-   init_color_array( &store->color0[1], &store->attribs[VERT_RESULT_BFC0] );
-   init_color_array( &store->color1[0], &store->attribs[VERT_RESULT_COL1] );
-   init_color_array( &store->color1[1], &store->attribs[VERT_RESULT_BFC1] );
-
    /* a few other misc allocations */
    _mesa_vector4f_alloc( &store->ndcCoords, 0, size, 32 );
    store->clipmask = (GLubyte *) ALIGN_MALLOC(sizeof(GLubyte)*size, 32 );
 
-   /* Now validate the stage derived data...
-    */
-   stage->run = run_validate_program;
-   return stage->run( ctx, stage );
-}
-
-
-
-/**
- * Check if vertex program mode is enabled. 
- * If so, configure the pipeline stage's type, inputs, and outputs.
- */
-static void check_vp( GLcontext *ctx, struct gl_pipeline_stage *stage )
-{
-   stage->active = ctx->VertexProgram.Enabled;
-
-   if (stage->active) {
-      /* I believe this is right - Keith?
-       * Set stage->inputs equal to the bitmask of vertex attributes
-       * which the program needs for inputs.
-       */
-
-      stage->inputs = ctx->VertexProgram.Current->InputsRead;
-
-#if 000
-      if (stage->privatePtr)
-	 stage->run = run_validate_program;
-      stage->inputs = VERT_BIT_NORMAL|VERT_BIT_MATERIAL;
-      if (ctx->Light._NeedVertices)
-	 stage->inputs |= VERT_BIT_EYE; /* effectively, even when lighting in obj */
-      if (ctx->Light.ColorMaterialEnabled)
-	 stage->inputs |= VERT_BIT_COLOR0;
-
-      stage->outputs = VERT_BIT_COLOR0;
-      if (ctx->Light.Model.ColorControl == GL_SEPARATE_SPECULAR_COLOR)
-	 stage->outputs |= VERT_BIT_COLOR1;
-#endif
-   }
+   return GL_TRUE;
 }
 
 
 /**
  * Destructor for this pipeline stage.
  */
-static void dtr( struct gl_pipeline_stage *stage )
+static void
+dtr(struct tnl_pipeline_stage *stage)
 {
    struct vp_stage_data *store = VP_STAGE_DATA(stage);
 
@@ -400,34 +505,38 @@ static void dtr( struct gl_pipeline_stage *stage )
       GLuint i;
 
       /* free the vertex program result arrays */
-      for (i = 0; i < 15; i++)
-         _mesa_vector4f_free( &store->attribs[i] );
+      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 = 0;
+      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 gl_pipeline_stage _tnl_vertex_program_stage =
+const struct tnl_pipeline_stage _tnl_vertex_program_stage =
 {
    "vertex-program",
-   _NEW_ALL,	/*XXX FIX */	/* recheck */
-   _NEW_ALL,	/*XXX FIX */    /* recalc -- modelview dependency
-				 * otherwise not captured by inputs
-				 * (which may be VERT_BIT_POS) */
-   GL_FALSE,			/* active */
-   /*0*/ VERT_BIT_POS,				/* inputs  XXX OK? */
-   VERT_BIT_CLIP | VERT_BIT_COLOR0,			/* outputs XXX OK? */
-   0,				/* changed_inputs */
    NULL,			/* private_data */
+   init_vp,			/* create */
    dtr,				/* destroy */
-   check_vp,			/* check */
-   run_init_vp			/* run -- initially set to ctr */
+   validate_vp_stage, 		/* validate */
+   run_vp			/* run -- initially set to ctr */
 };