1 files changed, 193 insertions, 51 deletions

diff --git a/src/mesa/tnl/t_vb_program.c b/src/mesa/tnl/t_vb_program.c
index f11ac616f1..17eefe7032 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:  6.5
+ * Version:  6.5.3
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
  * copy of this software and associated documentation files (the "Software"),
@@ -31,20 +31,14 @@
 
 
 #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 "program_instruction.h"
-#include "nvvertexec.h"
-#include "nvprogram.h"
+#include "prog_instruction.h"
+#include "prog_statevars.h"
+#include "prog_execute.h"
 
+#include "tnl.h"
 #include "t_context.h"
 #include "t_pipeline.h"
 
@@ -55,7 +49,7 @@
  */
 struct vp_stage_data {
    /** The results of running the vertex program go into these arrays. */
-   GLvector4f attribs[15];
+   GLvector4f results[VERT_RESULT_MAX];
 
    GLvector4f ndcCoords;              /**< normalized device coords */
    GLubyte *clipmask;                 /**< clip flags */
@@ -67,6 +61,131 @@ struct vp_stage_data {
 
 
 /**
+ * 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);
+      }
+   }
+
+   /* init condition codes */
+   machine->CondCodes[0] = COND_EQ;
+   machine->CondCodes[1] = COND_EQ;
+   machine->CondCodes[2] = COND_EQ;
+   machine->CondCodes[3] = COND_EQ;
+}
+
+
+/**
+ * Copy the 16 elements of a matrix into four consecutive program
+ * registers starting at 'pos'.
+ */
+static void
+load_matrix(GLfloat registers[][4], GLuint pos, const GLfloat mat[16])
+{
+   GLuint i;
+   for (i = 0; i < 4; i++) {
+      registers[pos + i][0] = mat[0 + i];
+      registers[pos + i][1] = mat[4 + i];
+      registers[pos + i][2] = mat[8 + i];
+      registers[pos + i][3] = mat[12 + i];
+   }
+}
+
+
+/**
+ * As above, but transpose the matrix.
+ */
+static void
+load_transpose_matrix(GLfloat registers[][4], GLuint pos,
+                      const GLfloat mat[16])
+{
+   MEMCPY(registers[pos], mat, 16 * sizeof(GLfloat));
+}
+
+
+/**
+ * Load current vertex program's parameter registers with tracked
+ * matrices (if NV program).  This only needs to be done per
+ * glBegin/glEnd, not per-vertex.
+ */
+void
+_mesa_load_tracked_matrices(GLcontext *ctx)
+{
+   GLuint i;
+
+   for (i = 0; i < MAX_NV_VERTEX_PROGRAM_PARAMS / 4; i++) {
+      /* point 'mat' at source matrix */
+      GLmatrix *mat;
+      if (ctx->VertexProgram.TrackMatrix[i] == GL_MODELVIEW) {
+         mat = ctx->ModelviewMatrixStack.Top;
+      }
+      else if (ctx->VertexProgram.TrackMatrix[i] == GL_PROJECTION) {
+         mat = ctx->ProjectionMatrixStack.Top;
+      }
+      else if (ctx->VertexProgram.TrackMatrix[i] == GL_TEXTURE) {
+         mat = ctx->TextureMatrixStack[ctx->Texture.CurrentUnit].Top;
+      }
+      else if (ctx->VertexProgram.TrackMatrix[i] == GL_COLOR) {
+         mat = ctx->ColorMatrixStack.Top;
+      }
+      else if (ctx->VertexProgram.TrackMatrix[i]==GL_MODELVIEW_PROJECTION_NV) {
+         /* XXX verify the combined matrix is up to date */
+         mat = &ctx->_ModelProjectMatrix;
+      }
+      else if (ctx->VertexProgram.TrackMatrix[i] >= GL_MATRIX0_NV &&
+               ctx->VertexProgram.TrackMatrix[i] <= GL_MATRIX7_NV) {
+         GLuint n = ctx->VertexProgram.TrackMatrix[i] - GL_MATRIX0_NV;
+         ASSERT(n < MAX_PROGRAM_MATRICES);
+         mat = ctx->ProgramMatrixStack[n].Top;
+      }
+      else {
+         /* no matrix is tracked, but we leave the register values as-is */
+         assert(ctx->VertexProgram.TrackMatrix[i] == GL_NONE);
+         continue;
+      }
+
+      /* load the matrix values into sequential registers */
+      if (ctx->VertexProgram.TrackMatrixTransform[i] == GL_IDENTITY_NV) {
+         load_matrix(ctx->VertexProgram.Parameters, i*4, mat->m);
+      }
+      else if (ctx->VertexProgram.TrackMatrixTransform[i] == GL_INVERSE_NV) {
+         _math_matrix_analyse(mat); /* update the inverse */
+         ASSERT(!_math_matrix_is_dirty(mat));
+         load_matrix(ctx->VertexProgram.Parameters, i*4, mat->inv);
+      }
+      else if (ctx->VertexProgram.TrackMatrixTransform[i] == GL_TRANSPOSE_NV) {
+         load_transpose_matrix(ctx->VertexProgram.Parameters, i*4, mat->m);
+      }
+      else {
+         assert(ctx->VertexProgram.TrackMatrixTransform[i]
+                == GL_INVERSE_TRANSPOSE_NV);
+         _math_matrix_analyse(mat); /* update the inverse */
+         ASSERT(!_math_matrix_is_dirty(mat));
+         load_transpose_matrix(ctx->VertexProgram.Parameters, i*4, mat->inv);
+      }
+   }
+}
+
+
+/**
  * This function executes vertex programs
  */
 static GLboolean
@@ -75,24 +194,38 @@ 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 vp_machine machine;
-   GLuint i;
-
-   if (ctx->ShaderObjects._VertexShaderPresent)
+   struct gl_vertex_program *program = ctx->VertexProgram._Current;
+   struct gl_program_machine machine;
+   GLuint outputs[VERT_RESULT_MAX], numOutputs;
+   GLuint i, j;
+
+#define FORCE_PROG_EXECUTE_C 1
+#if FORCE_PROG_EXECUTE_C
+   if (!program)
       return GL_TRUE;
-
-   if (!ctx->VertexProgram._Enabled ||
-       !program->IsNVProgram)
+#else
+   if (!program || !program->IsNVProgram)
       return GL_TRUE;
+#endif
 
-   /* load program parameter registers (they're read-only) */
-   _mesa_init_vp_per_primitive_registers(ctx);
+   if (ctx->VertexProgram.Current->IsNVProgram) {
+      _mesa_load_tracked_matrices(ctx);
+   }
+   else {
+      _mesa_load_state_parameters(ctx, program->Base.Parameters);
+   }
+
+   numOutputs = 0;
+   for (i = 0; i < VERT_RESULT_MAX; i++) {
+      if (program->Base.OutputsWritten & (1 << i)) {
+         outputs[numOutputs++] = i;
+      }
+   }
 
    for (i = 0; i < VB->Count; i++) {
       GLuint attr;
 
-      _mesa_init_vp_per_vertex_registers(ctx, &machine);
+      init_machine(ctx, &machine);
 
 #if 0
       printf("Input  %d: %f, %f, %f, %f\n", i,
@@ -119,13 +252,12 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage )
 	    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.Inputs[attr], size, data);
+	    COPY_CLEAN_4V(machine.VertAttribs[attr], size, data);
 	 }
       }
 
       /* execute the program */
-      ASSERT(program);
-      _mesa_exec_vertex_program(ctx, &machine, program);
+      _mesa_execute_program(ctx, &program->Base, &machine);
 
       /* Fixup fog an point size results if needed */
       if (ctx->Fog.Enabled &&
@@ -139,33 +271,48 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage )
       }
 
       /* 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.Outputs[attr]);
+      for (j = 0; j < numOutputs; j++) {
+         const GLuint attr = outputs[j];
+         COPY_4V(store->results[attr].data[i], machine.Outputs[attr]);
       }
+#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
    }
 
    /* 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 = &store->results[VERT_RESULT_HPOS];
    VB->ClipPtr->size = 4;
    VB->ClipPtr->count = VB->Count;
-   VB->ColorPtr[0] = &store->attribs[VERT_RESULT_COL0];
-   VB->ColorPtr[1] = &store->attribs[VERT_RESULT_BFC0];
-   VB->SecondaryColorPtr[0] = &store->attribs[VERT_RESULT_COL1];
-   VB->SecondaryColorPtr[1] = &store->attribs[VERT_RESULT_BFC1];
-   VB->FogCoordPtr = &store->attribs[VERT_RESULT_FOGC];
+   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->attribs[VERT_RESULT_COL0];
-   VB->AttribPtr[VERT_ATTRIB_COLOR1] = &store->attribs[VERT_RESULT_COL1];
-   VB->AttribPtr[VERT_ATTRIB_FOG] = &store->attribs[VERT_RESULT_FOGC];
-   VB->AttribPtr[_TNL_ATTRIB_POINTSIZE] = &store->attribs[VERT_RESULT_PSIZ];
+   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->attribs[VERT_RESULT_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];
+      }
    }
 
    /* Cliptest and perspective divide.  Clip functions must clear
@@ -206,8 +353,6 @@ run_vp( GLcontext *ctx, struct tnl_pipeline_stage *stage )
 }
 
 
-
-
 /**
  * Called the first time stage->run is called.  In effect, don't
  * allocate data until the first time the stage is run.
@@ -227,10 +372,9 @@ static GLboolean init_vp( GLcontext *ctx,
       return GL_FALSE;
 
    /* Allocate arrays of vertex output values */
-   /* XXX change '15' to a named constant */
-   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;
    }
 
    /* a few other misc allocations */
@@ -241,9 +385,6 @@ static GLboolean init_vp( GLcontext *ctx,
 }
 
 
-
-
-
 /**
  * Destructor for this pipeline stage.
  */
@@ -256,7 +397,7 @@ static void dtr( struct tnl_pipeline_stage *stage )
 
       /* free the vertex program result arrays */
       for (i = 0; i < VERT_RESULT_MAX; i++)
-         _mesa_vector4f_free( &store->attribs[i] );
+         _mesa_vector4f_free( &store->results[i] );
 
       /* free misc arrays */
       _mesa_vector4f_free( &store->ndcCoords );
@@ -267,6 +408,7 @@ static void dtr( struct tnl_pipeline_stage *stage )
    }
 }
 
+
 /**
  * Public description of this pipeline stage.
  */