12 files changed, 4235 insertions, 4187 deletions
diff --git a/mesalib/src/mesa/program/ir_to_mesa.cpp b/mesalib/src/mesa/program/ir_to_mesa.cpp
index a08aa0af8..326d1fb63 100644
--- a/mesalib/src/mesa/program/ir_to_mesa.cpp
+++ b/mesalib/src/mesa/program/ir_to_mesa.cpp
@@ -585,7 +585,7 @@ ir_to_mesa_visitor::src_reg_for_float(float val)
    src_reg src(PROGRAM_CONSTANT, -1, NULL);
 
    src.index = _mesa_add_unnamed_constant(this->prog->Parameters,
-					  &val, 1, &src.swizzle);
+					  (const gl_constant_value *)&val, 1, &src.swizzle);
 
    return src;
 }
@@ -1795,7 +1795,7 @@ ir_to_mesa_visitor::visit(ir_constant *ir)
 
 	 src = src_reg(PROGRAM_CONSTANT, -1, NULL);
 	 src.index = _mesa_add_unnamed_constant(this->prog->Parameters,
-						values,
+						(gl_constant_value *) values,
 						ir->type->vector_elements,
 						&src.swizzle);
 	 emit(ir, OPCODE_MOV, mat_column, src);
@@ -1833,7 +1833,7 @@ ir_to_mesa_visitor::visit(ir_constant *ir)
 
    this->result = src_reg(PROGRAM_CONSTANT, -1, ir->type);
    this->result.index = _mesa_add_unnamed_constant(this->prog->Parameters,
-						   values,
+						   (gl_constant_value *) values,
 						   ir->type->vector_elements,
 						   &this->result.swizzle);
 }
@@ -2533,7 +2533,7 @@ add_uniforms_to_parameters_list(struct gl_shader_program *shader_program,
 	  */
 	 if (file == PROGRAM_SAMPLER) {
 	    for (unsigned int j = 0; j < size / 4; j++)
-	       prog->Parameters->ParameterValues[index + j][0] = next_sampler++;
+	       prog->Parameters->ParameterValues[index + j][0].f = next_sampler++;
 	 }
 
 	 /* The location chosen in the Parameters list here (returned
diff --git a/mesalib/src/mesa/program/nvfragparse.c b/mesalib/src/mesa/program/nvfragparse.c
index 04538e071..ce72c610d 100644
--- a/mesalib/src/mesa/program/nvfragparse.c
+++ b/mesalib/src/mesa/program/nvfragparse.c
@@ -1,1588 +1,1595 @@
-/*
- * 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.
- */
-
-/**
- * \file nvfragparse.c
- * NVIDIA fragment program parser.
- * \author Brian Paul
- */
-
-/*
- * Regarding GL_NV_fragment_program:
- *
- * 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/context.h"
-#include "main/imports.h"
-#include "main/macros.h"
-#include "program.h"
-#include "prog_parameter.h"
-#include "prog_print.h"
-#include "prog_instruction.h"
-#include "nvfragparse.h"
-
-
-#define INPUT_1V     1
-#define INPUT_2V     2
-#define INPUT_3V     3
-#define INPUT_1S     4
-#define INPUT_2S     5
-#define INPUT_CC     6
-#define INPUT_1V_T   7  /* one source vector, plus textureId */
-#define INPUT_3V_T   8  /* one source vector, plus textureId */
-#define INPUT_NONE   9
-#define INPUT_1V_S  10  /* a string and a vector register */
-#define OUTPUT_V    20
-#define OUTPUT_S    21
-#define OUTPUT_NONE 22
-
-/* IRIX defines some of these */
-#undef _R
-#undef _H
-#undef _X
-#undef _C
-#undef _S
-
-/* Optional suffixes */
-#define _R  FLOAT32  /* float */
-#define _H  FLOAT16  /* half-float */
-#define _X  FIXED12  /* fixed */
-#define _C  0x08     /* set cond codes */
-#define _S  0x10     /* saturate, clamp result to [0,1] */
-
-struct instruction_pattern {
-   const char *name;
-   enum prog_opcode opcode;
-   GLuint inputs;
-   GLuint outputs;
-   GLuint suffixes;
-};
-
-static const struct instruction_pattern Instructions[] = {
-   { "ADD", OPCODE_ADD, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "COS", OPCODE_COS, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
-   { "DDX", OPCODE_DDX, INPUT_1V, OUTPUT_V, _R | _H |      _C | _S },
-   { "DDY", OPCODE_DDY, INPUT_1V, OUTPUT_V, _R | _H |      _C | _S },
-   { "DP3", OPCODE_DP3, INPUT_2V, OUTPUT_S, _R | _H | _X | _C | _S },
-   { "DP4", OPCODE_DP4, INPUT_2V, OUTPUT_S, _R | _H | _X | _C | _S },
-   { "DST", OPCODE_DP4, INPUT_2V, OUTPUT_V, _R | _H |      _C | _S },
-   { "EX2", OPCODE_DP4, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
-   { "FLR", OPCODE_FLR, INPUT_1V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "FRC", OPCODE_FRC, INPUT_1V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "KIL", OPCODE_KIL_NV, INPUT_CC, OUTPUT_NONE, 0                },
-   { "LG2", OPCODE_LG2, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
-   { "LIT", OPCODE_LIT, INPUT_1V, OUTPUT_V, _R | _H |      _C | _S },
-   { "LRP", OPCODE_LRP, INPUT_3V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "MAD", OPCODE_MAD, INPUT_3V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "MAX", OPCODE_MAX, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "MIN", OPCODE_MIN, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "MOV", OPCODE_MOV, INPUT_1V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "MUL", OPCODE_MUL, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "PK2H",  OPCODE_PK2H,  INPUT_1V, OUTPUT_S, 0                  },
-   { "PK2US", OPCODE_PK2US, INPUT_1V, OUTPUT_S, 0                  },
-   { "PK4B",  OPCODE_PK4B,  INPUT_1V, OUTPUT_S, 0                  },
-   { "PK4UB", OPCODE_PK4UB, INPUT_1V, OUTPUT_S, 0                  },
-   { "POW", OPCODE_POW, INPUT_2S, OUTPUT_S, _R | _H |      _C | _S },
-   { "RCP", OPCODE_RCP, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
-   { "RFL", OPCODE_RFL, INPUT_2V, OUTPUT_V, _R | _H |      _C | _S },
-   { "RSQ", OPCODE_RSQ, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
-   { "SEQ", OPCODE_SEQ, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "SFL", OPCODE_SFL, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "SGE", OPCODE_SGE, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "SGT", OPCODE_SGT, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "SIN", OPCODE_SIN, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
-   { "SLE", OPCODE_SLE, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "SLT", OPCODE_SLT, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "SNE", OPCODE_SNE, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "STR", OPCODE_STR, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "SUB", OPCODE_SUB, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
-   { "TEX", OPCODE_TEX, INPUT_1V_T, OUTPUT_V,              _C | _S },
-   { "TXD", OPCODE_TXD, INPUT_3V_T, OUTPUT_V,              _C | _S },
-   { "TXP", OPCODE_TXP_NV, INPUT_1V_T, OUTPUT_V,           _C | _S },
-   { "UP2H",  OPCODE_UP2H,  INPUT_1S, OUTPUT_V,            _C | _S },
-   { "UP2US", OPCODE_UP2US, INPUT_1S, OUTPUT_V,            _C | _S },
-   { "UP4B",  OPCODE_UP4B,  INPUT_1S, OUTPUT_V,            _C | _S },
-   { "UP4UB", OPCODE_UP4UB, INPUT_1S, OUTPUT_V,            _C | _S },
-   { "X2D", OPCODE_X2D, INPUT_3V, OUTPUT_V, _R | _H |      _C | _S },
-   { "PRINT", OPCODE_PRINT, INPUT_1V_S, OUTPUT_NONE, 0               },
-   { NULL, (enum prog_opcode) -1, 0, 0, 0 }
-};
-
-
-/*
- * Information needed or computed during parsing.
- * Remember, we can't modify the target program object until we've
- * _successfully_ parsed the program text.
- */
-struct parse_state {
-   struct gl_context *ctx;
-   const GLubyte *start;              /* start of program string */
-   const GLubyte *pos;                /* current position */
-   const GLubyte *curLine;
-   struct gl_fragment_program *program;  /* current program */
-
-   struct gl_program_parameter_list *parameters;
-
-   GLuint numInst;                    /* number of instructions parsed */
-   GLuint inputsRead;                 /* bitmask of input registers used */
-   GLuint outputsWritten;             /* bitmask of 1 << FRAG_OUTPUT_* bits */
-   GLuint texturesUsed[MAX_TEXTURE_IMAGE_UNITS];
-};
-
-
-
-/*
- * Called whenever we find an error during parsing.
- */
-static void
-record_error(struct parse_state *parseState, const char *msg, int lineNo)
-{
-#ifdef DEBUG
-   GLint line, column;
-   const GLubyte *lineStr;
-   lineStr = _mesa_find_line_column(parseState->start,
-                                    parseState->pos, &line, &column);
-   _mesa_debug(parseState->ctx,
-               "nvfragparse.c(%d): line %d, column %d:%s (%s)\n",
-               lineNo, line, column, (char *) lineStr, msg);
-   free((void *) lineStr);
-#else
-   (void) lineNo;
-#endif
-
-   /* Check that no error was already recorded.  Only record the first one. */
-   if (parseState->ctx->Program.ErrorString[0] == 0) {
-      _mesa_set_program_error(parseState->ctx,
-                              parseState->pos - parseState->start,
-                              msg);
-   }
-}
-
-
-#define RETURN_ERROR							\
-do {									\
-   record_error(parseState, "Unexpected end of input.", __LINE__);	\
-   return GL_FALSE;							\
-} while(0)
-
-#define RETURN_ERROR1(msg)						\
-do {									\
-   record_error(parseState, msg, __LINE__);				\
-   return GL_FALSE;							\
-} while(0)
-
-#define RETURN_ERROR2(msg1, msg2)					\
-do {									\
-   char err[1000];							\
-   sprintf(err, "%s %s", msg1, msg2);				\
-   record_error(parseState, err, __LINE__);				\
-   return GL_FALSE;							\
-} while(0)
-
-
-
-
-/*
- * Search a list of instruction structures for a match.
- */
-static struct instruction_pattern
-MatchInstruction(const GLubyte *token)
-{
-   const struct instruction_pattern *inst;
-   struct instruction_pattern result;
-
-   result.name = NULL;
-   result.opcode = MAX_OPCODE; /* i.e. invalid instruction */
-   result.inputs = 0;
-   result.outputs = 0;
-   result.suffixes = 0;
-
-   for (inst = Instructions; inst->name; inst++) {
-      if (strncmp((const char *) token, inst->name, 3) == 0) {
-         /* matched! */
-         int i = 3;
-         result = *inst;
-         result.suffixes = 0;
-         /* look at suffix */
-         if (token[i] == 'R') {
-            result.suffixes |= _R;
-            i++;
-         }
-         else if (token[i] == 'H') {
-            result.suffixes |= _H;
-            i++;
-         }
-         else if (token[i] == 'X') {
-            result.suffixes |= _X;
-            i++;
-         }
-         if (token[i] == 'C') {
-            result.suffixes |= _C;
-            i++;
-         }
-         if (token[i] == '_' && token[i+1] == 'S' &&
-             token[i+2] == 'A' && token[i+3] == 'T') {
-            result.suffixes |= _S;
-         }
-         return result;
-      }
-   }
-
-   return result;
-}
-
-
-
-
-/**********************************************************************/
-
-
-static GLboolean IsLetter(GLubyte b)
-{
-   return (b >= 'a' && b <= 'z') ||
-          (b >= 'A' && b <= 'Z') ||
-          (b == '_') ||
-          (b == '$');
-}
-
-
-static GLboolean IsDigit(GLubyte b)
-{
-   return b >= '0' && b <= '9';
-}
-
-
-static GLboolean IsWhitespace(GLubyte b)
-{
-   return b == ' ' || b == '\t' || b == '\n' || b == '\r';
-}
-
-
-/**
- * Starting at 'str' find the next token.  A token can be an integer,
- * an identifier or punctuation symbol.
- * \return <= 0 we found an error, else, return number of characters parsed.
- */
-static GLint
-GetToken(struct parse_state *parseState, GLubyte *token)
-{
-   const GLubyte *str = parseState->pos;
-   GLint i = 0, j = 0;
-
-   token[0] = 0;
-
-   /* skip whitespace and comments */
-   while (str[i] && (IsWhitespace(str[i]) || str[i] == '#')) {
-      if (str[i] == '#') {
-         /* skip comment */
-         while (str[i] && (str[i] != '\n' && str[i] != '\r')) {
-            i++;
-         }
-         if (str[i] == '\n' || str[i] == '\r')
-            parseState->curLine = str + i + 1;
-      }
-      else {
-         /* skip whitespace */
-         if (str[i] == '\n' || str[i] == '\r')
-            parseState->curLine = str + i + 1;
-         i++;
-      }
-   }
-
-   if (str[i] == 0)
-      return -i;
-
-   /* try matching an integer */
-   while (str[i] && IsDigit(str[i])) {
-      token[j++] = str[i++];
-   }
-   if (j > 0 || !str[i]) {
-      token[j] = 0;
-      return i;
-   }
-
-   /* try matching an identifier */
-   if (IsLetter(str[i])) {
-      while (str[i] && (IsLetter(str[i]) || IsDigit(str[i]))) {
-         token[j++] = str[i++];
-      }
-      token[j] = 0;
-      return i;
-   }
-
-   /* punctuation character */
-   if (str[i]) {
-      token[0] = str[i++];
-      token[1] = 0;
-      return i;
-   }
-
-   /* end of input */
-   token[0] = 0;
-   return i;
-}
-
-
-/**
- * Get next token from input stream and increment stream pointer past token.
- */
-static GLboolean
-Parse_Token(struct parse_state *parseState, GLubyte *token)
-{
-   GLint i;
-   i = GetToken(parseState, token);
-   if (i <= 0) {
-      parseState->pos += (-i);
-      return GL_FALSE;
-   }
-   parseState->pos += i;
-   return GL_TRUE;
-}
-
-
-/**
- * Get next token from input stream but don't increment stream pointer.
- */
-static GLboolean
-Peek_Token(struct parse_state *parseState, GLubyte *token)
-{
-   GLint i, len;
-   i = GetToken(parseState, token);
-   if (i <= 0) {
-      parseState->pos += (-i);
-      return GL_FALSE;
-   }
-   len = (GLint) strlen((const char *) token);
-   parseState->pos += (i - len);
-   return GL_TRUE;
-}
-
-
-/**********************************************************************/
-
-static const char *InputRegisters[MAX_NV_FRAGMENT_PROGRAM_INPUTS + 1] = {
-   "WPOS", "COL0", "COL1", "FOGC",
-   "TEX0", "TEX1", "TEX2", "TEX3", "TEX4", "TEX5", "TEX6", "TEX7", NULL
-};
-
-
-
-/**********************************************************************/
-
-/**
- * Try to match 'pattern' as the next token after any whitespace/comments.
- */
-static GLboolean
-Parse_String(struct parse_state *parseState, const char *pattern)
-{
-   const GLubyte *m;
-   GLint i;
-
-   /* skip whitespace and comments */
-   while (IsWhitespace(*parseState->pos) || *parseState->pos == '#') {
-      if (*parseState->pos == '#') {
-         while (*parseState->pos && (*parseState->pos != '\n' && *parseState->pos != '\r')) {
-            parseState->pos += 1;
-         }
-         if (*parseState->pos == '\n' || *parseState->pos == '\r')
-            parseState->curLine = parseState->pos + 1;
-      }
-      else {
-         /* skip whitespace */
-         if (*parseState->pos == '\n' || *parseState->pos == '\r')
-            parseState->curLine = parseState->pos + 1;
-         parseState->pos += 1;
-      }
-   }
-
-   /* Try to match the pattern */
-   m = parseState->pos;
-   for (i = 0; pattern[i]; i++) {
-      if (*m != (GLubyte) pattern[i])
-         return GL_FALSE;
-      m += 1;
-   }
-   parseState->pos = m;
-
-   return GL_TRUE; /* success */
-}
-
-
-static GLboolean
-Parse_Identifier(struct parse_state *parseState, GLubyte *ident)
-{
-   if (!Parse_Token(parseState, ident))
-      RETURN_ERROR;
-   if (IsLetter(ident[0]))
-      return GL_TRUE;
-   else
-      RETURN_ERROR1("Expected an identfier");
-}
-
-
-/**
- * Parse a floating point constant, or a defined symbol name.
- * [+/-]N[.N[eN]]
- * Output:  number[0 .. 3] will get the value.
- */
-static GLboolean
-Parse_ScalarConstant(struct parse_state *parseState, GLfloat *number)
-{
-   char *end = NULL;
-
-   *number = (GLfloat) _mesa_strtof((const char *) parseState->pos, &end);
-
-   if (end && end > (char *) parseState->pos) {
-      /* got a number */
-      parseState->pos = (GLubyte *) end;
-      number[1] = *number;
-      number[2] = *number;
-      number[3] = *number;
-      return GL_TRUE;
-   }
-   else {
-      /* should be an identifier */
-      GLubyte ident[100];
-      const GLfloat *constant;
-      if (!Parse_Identifier(parseState, ident))
-         RETURN_ERROR1("Expected an identifier");
-      constant = _mesa_lookup_parameter_value(parseState->parameters,
-                                              -1, (const char *) ident);
-      /* XXX Check that it's a constant and not a parameter */
-      if (!constant) {
-         RETURN_ERROR1("Undefined symbol");
-      }
-      else {
-         COPY_4V(number, constant);
-         return GL_TRUE;
-      }
-   }
-}
-
-
-
-/**
- * Parse a vector constant, one of:
- *   { float }
- *   { float, float }
- *   { float, float, float }
- *   { float, float, float, float }
- */
-static GLboolean
-Parse_VectorConstant(struct parse_state *parseState, GLfloat *vec)
-{
-   /* "{" was already consumed */
-
-   ASSIGN_4V(vec, 0.0, 0.0, 0.0, 1.0);
-
-   if (!Parse_ScalarConstant(parseState, vec+0))  /* X */
-      return GL_FALSE;
-
-   if (Parse_String(parseState, "}")) {
-      return GL_TRUE;
-   }
-
-   if (!Parse_String(parseState, ","))
-      RETURN_ERROR1("Expected comma in vector constant");
-
-   if (!Parse_ScalarConstant(parseState, vec+1))  /* Y */
-      return GL_FALSE;
-
-   if (Parse_String(parseState, "}")) {
-      return GL_TRUE;
-   }
-
-   if (!Parse_String(parseState, ","))
-      RETURN_ERROR1("Expected comma in vector constant");
-
-   if (!Parse_ScalarConstant(parseState, vec+2))  /* Z */
-      return GL_FALSE;
-
-   if (Parse_String(parseState, "}")) {
-      return GL_TRUE;
-   }
-
-   if (!Parse_String(parseState, ","))
-      RETURN_ERROR1("Expected comma in vector constant");
-
-   if (!Parse_ScalarConstant(parseState, vec+3))  /* W */
-      return GL_FALSE;
-
-   if (!Parse_String(parseState, "}"))
-      RETURN_ERROR1("Expected closing brace in vector constant");
-
-   return GL_TRUE;
-}
-
-
-/**
- * Parse <number>, <varname> or {a, b, c, d}.
- * Return number of values in the vector or scalar, or zero if parse error.
- */
-static GLuint
-Parse_VectorOrScalarConstant(struct parse_state *parseState, GLfloat *vec)
-{
-   if (Parse_String(parseState, "{")) {
-      return Parse_VectorConstant(parseState, vec);
-   }
-   else {
-      GLboolean b = Parse_ScalarConstant(parseState, vec);
-      if (b) {
-         vec[1] = vec[2] = vec[3] = vec[0];
-      }
-      return b;
-   }
-}
-
-
-/**
- * Parse a texture image source:
- *    [TEX0 | TEX1 | .. | TEX15] , [1D | 2D | 3D | CUBE | RECT]
- */
-static GLboolean
-Parse_TextureImageId(struct parse_state *parseState,
-                     GLubyte *texUnit, GLubyte *texTargetBit)
-{
-   GLubyte imageSrc[100];
-   GLint unit;
-
-   if (!Parse_Token(parseState, imageSrc))
-      RETURN_ERROR;
-   
-   if (imageSrc[0] != 'T' ||
-       imageSrc[1] != 'E' ||
-       imageSrc[2] != 'X') {
-      RETURN_ERROR1("Expected TEX# source");
-   }
-   unit = atoi((const char *) imageSrc + 3);
-   if ((unit < 0 || unit > MAX_TEXTURE_IMAGE_UNITS) ||
-       (unit == 0 && (imageSrc[3] != '0' || imageSrc[4] != 0))) {
-      RETURN_ERROR1("Invalied TEX# source index");
-   }
-   *texUnit = unit;
-
-   if (!Parse_String(parseState, ","))
-      RETURN_ERROR1("Expected ,");
-
-   if (Parse_String(parseState, "1D")) {
-      *texTargetBit = TEXTURE_1D_BIT;
-   }
-   else if (Parse_String(parseState, "2D")) {
-      *texTargetBit = TEXTURE_2D_BIT;
-   }
-   else if (Parse_String(parseState, "3D")) {
-      *texTargetBit = TEXTURE_3D_BIT;
-   }
-   else if (Parse_String(parseState, "CUBE")) {
-      *texTargetBit = TEXTURE_CUBE_BIT;
-   }
-   else if (Parse_String(parseState, "RECT")) {
-      *texTargetBit = TEXTURE_RECT_BIT;
-   }
-   else {
-      RETURN_ERROR1("Invalid texture target token");
-   }
-
-   /* update record of referenced texture units */
-   parseState->texturesUsed[*texUnit] |= *texTargetBit;
-   if (_mesa_bitcount(parseState->texturesUsed[*texUnit]) > 1) {
-      RETURN_ERROR1("Only one texture target can be used per texture unit.");
-   }
-
-   return GL_TRUE;
-}
-
-
-/**
- * Parse a scalar suffix like .x, .y, .z or .w or parse a swizzle suffix
- * like .wxyz, .xxyy, etc and return the swizzle indexes.
- */
-static GLboolean
-Parse_SwizzleSuffix(const GLubyte *token, GLuint swizzle[4])
-{
-   if (token[1] == 0) {
-      /* single letter swizzle (scalar) */
-      if (token[0] == 'x')
-         ASSIGN_4V(swizzle, 0, 0, 0, 0);
-      else if (token[0] == 'y')
-         ASSIGN_4V(swizzle, 1, 1, 1, 1);
-      else if (token[0] == 'z')
-         ASSIGN_4V(swizzle, 2, 2, 2, 2);
-      else if (token[0] == 'w')
-         ASSIGN_4V(swizzle, 3, 3, 3, 3);
-      else
-         return GL_FALSE;
-   }
-   else {
-      /* 4-component swizzle (vector) */
-      GLint k;
-      for (k = 0; k < 4 && token[k]; k++) {
-         if (token[k] == 'x')
-            swizzle[k] = 0;
-         else if (token[k] == 'y')
-            swizzle[k] = 1;
-         else if (token[k] == 'z')
-            swizzle[k] = 2;
-         else if (token[k] == 'w')
-            swizzle[k] = 3;
-         else
-            return GL_FALSE;
-      }
-      if (k != 4)
-         return GL_FALSE;
-   }
-   return GL_TRUE;
-}
-
-
-static GLboolean
-Parse_CondCodeMask(struct parse_state *parseState,
-                   struct prog_dst_register *dstReg)
-{
-   if (Parse_String(parseState, "EQ"))
-      dstReg->CondMask = COND_EQ;
-   else if (Parse_String(parseState, "GE"))
-      dstReg->CondMask = COND_GE;
-   else if (Parse_String(parseState, "GT"))
-      dstReg->CondMask = COND_GT;
-   else if (Parse_String(parseState, "LE"))
-      dstReg->CondMask = COND_LE;
-   else if (Parse_String(parseState, "LT"))
-      dstReg->CondMask = COND_LT;
-   else if (Parse_String(parseState, "NE"))
-      dstReg->CondMask = COND_NE;
-   else if (Parse_String(parseState, "TR"))
-      dstReg->CondMask = COND_TR;
-   else if (Parse_String(parseState, "FL"))
-      dstReg->CondMask = COND_FL;
-   else
-      RETURN_ERROR1("Invalid condition code mask");
-
-   /* look for optional .xyzw swizzle */
-   if (Parse_String(parseState, ".")) {
-      GLubyte token[100];
-      GLuint swz[4];
-
-      if (!Parse_Token(parseState, token))  /* get xyzw suffix */
-         RETURN_ERROR;
-
-      if (!Parse_SwizzleSuffix(token, swz))
-         RETURN_ERROR1("Invalid swizzle suffix");
-
-      dstReg->CondSwizzle = MAKE_SWIZZLE4(swz[0], swz[1], swz[2], swz[3]);
-   }
-
-   return GL_TRUE;
-}
-
-
-/**
- * Parse a temporary register: Rnn or Hnn
- */
-static GLboolean
-Parse_TempReg(struct parse_state *parseState, GLint *tempRegNum)
-{
-   GLubyte token[100];
-
-   /* Should be 'R##' or 'H##' */
-   if (!Parse_Token(parseState, token))
-      RETURN_ERROR;
-   if (token[0] != 'R' && token[0] != 'H')
-      RETURN_ERROR1("Expected R## or H##");
-
-   if (IsDigit(token[1])) {
-      GLint reg = atoi((const char *) (token + 1));
-      if (token[0] == 'H')
-         reg += 32;
-      if (reg >= MAX_NV_FRAGMENT_PROGRAM_TEMPS)
-         RETURN_ERROR1("Invalid temporary register name");
-      *tempRegNum = reg;
-   }
-   else {
-      RETURN_ERROR1("Invalid temporary register name");
-   }
-
-   return GL_TRUE;
-}
-
-
-/**
- * Parse a write-only dummy register: RC or HC.
- */
-static GLboolean
-Parse_DummyReg(struct parse_state *parseState, GLint *regNum)
-{
-   if (Parse_String(parseState, "RC")) {
-       *regNum = 0;
-   }
-   else if (Parse_String(parseState, "HC")) {
-       *regNum = 1;
-   }
-   else {
-      RETURN_ERROR1("Invalid write-only register name");
-   }
-
-   return GL_TRUE;
-}
-
-
-/**
- * Parse a program local parameter register "p[##]"
- */
-static GLboolean
-Parse_ProgramParamReg(struct parse_state *parseState, GLint *regNum)
-{
-   GLubyte token[100];
-
-   if (!Parse_String(parseState, "p["))
-      RETURN_ERROR1("Expected p[");
-
-   if (!Parse_Token(parseState, token))
-      RETURN_ERROR;
-
-   if (IsDigit(token[0])) {
-      /* a numbered program parameter register */
-      GLint reg = atoi((const char *) token);
-      if (reg >= MAX_NV_FRAGMENT_PROGRAM_PARAMS)
-         RETURN_ERROR1("Invalid constant program number");
-      *regNum = reg;
-   }
-   else {
-      RETURN_ERROR;
-   }
-
-   if (!Parse_String(parseState, "]"))
-      RETURN_ERROR1("Expected ]");
-
-   return GL_TRUE;
-}
-
-
-/**
- * Parse f[name]  - fragment input register
- */
-static GLboolean
-Parse_FragReg(struct parse_state *parseState, GLint *tempRegNum)
-{
-   GLubyte token[100];
-   GLint j;
-
-   /* Match 'f[' */
-   if (!Parse_String(parseState, "f["))
-      RETURN_ERROR1("Expected f[");
-
-   /* get <name> and look for match */
-   if (!Parse_Token(parseState, token)) {
-      RETURN_ERROR;
-   }
-   for (j = 0; InputRegisters[j]; j++) {
-      if (strcmp((const char *) token, InputRegisters[j]) == 0) {
-         *tempRegNum = j;
-         parseState->inputsRead |= (1 << j);
-         break;
-      }
-   }
-   if (!InputRegisters[j]) {
-      /* unknown input register label */
-      RETURN_ERROR2("Invalid register name", token);
-   }
-
-   /* Match '[' */
-   if (!Parse_String(parseState, "]"))
-      RETURN_ERROR1("Expected ]");
-
-   return GL_TRUE;
-}
-
-
-static GLboolean
-Parse_OutputReg(struct parse_state *parseState, GLint *outputRegNum)
-{
-   GLubyte token[100];
-
-   /* Match "o[" */
-   if (!Parse_String(parseState, "o["))
-      RETURN_ERROR1("Expected o[");
-
-   /* Get output reg name */
-   if (!Parse_Token(parseState, token))
-      RETURN_ERROR;
-
-   /* try to match an output register name */
-   if (strcmp((char *) token, "COLR") == 0 ||
-       strcmp((char *) token, "COLH") == 0) {
-      /* note that we don't distinguish between COLR and COLH */
-      *outputRegNum = FRAG_RESULT_COLOR;
-      parseState->outputsWritten |= (1 << FRAG_RESULT_COLOR);
-   }
-   else if (strcmp((char *) token, "DEPR") == 0) {
-      *outputRegNum = FRAG_RESULT_DEPTH;
-      parseState->outputsWritten |= (1 << FRAG_RESULT_DEPTH);
-   }
-   else {
-      RETURN_ERROR1("Invalid output register name");
-   }
-
-   /* Match ']' */
-   if (!Parse_String(parseState, "]"))
-      RETURN_ERROR1("Expected ]");
-
-   return GL_TRUE;
-}
-
-
-static GLboolean
-Parse_MaskedDstReg(struct parse_state *parseState,
-                   struct prog_dst_register *dstReg)
-{
-   GLubyte token[100];
-   GLint idx;
-
-   /* Dst reg can be R<n>, H<n>, o[n], RC or HC */
-   if (!Peek_Token(parseState, token))
-      RETURN_ERROR;
-
-   if (strcmp((const char *) token, "RC") == 0 ||
-       strcmp((const char *) token, "HC") == 0) {
-      /* a write-only register */
-      dstReg->File = PROGRAM_WRITE_ONLY;
-      if (!Parse_DummyReg(parseState, &idx))
-         RETURN_ERROR;
-      dstReg->Index = idx;
-   }
-   else if (token[0] == 'R' || token[0] == 'H') {
-      /* a temporary register */
-      dstReg->File = PROGRAM_TEMPORARY;
-      if (!Parse_TempReg(parseState, &idx))
-         RETURN_ERROR;
-      dstReg->Index = idx;
-   }
-   else if (token[0] == 'o') {
-      /* an output register */
-      dstReg->File = PROGRAM_OUTPUT;
-      if (!Parse_OutputReg(parseState, &idx))
-         RETURN_ERROR;
-      dstReg->Index = idx;
-   }
-   else {
-      RETURN_ERROR1("Invalid destination register name");
-   }
-
-   /* Parse optional write mask */
-   if (Parse_String(parseState, ".")) {
-      /* got a mask */
-      GLint k = 0;
-
-      if (!Parse_Token(parseState, token))  /* get xyzw writemask */
-         RETURN_ERROR;
-
-      dstReg->WriteMask = 0;
-
-      if (token[k] == 'x') {
-         dstReg->WriteMask |= WRITEMASK_X;
-         k++;
-      }
-      if (token[k] == 'y') {
-         dstReg->WriteMask |= WRITEMASK_Y;
-         k++;
-      }
-      if (token[k] == 'z') {
-         dstReg->WriteMask |= WRITEMASK_Z;
-         k++;
-      }
-      if (token[k] == 'w') {
-         dstReg->WriteMask |= WRITEMASK_W;
-         k++;
-      }
-      if (k == 0) {
-         RETURN_ERROR1("Invalid writemask character");
-      }
-
-   }
-   else {
-      dstReg->WriteMask = WRITEMASK_XYZW;
-   }
-
-   /* optional condition code mask */
-   if (Parse_String(parseState, "(")) {
-      /* ("EQ" | "GE" | "GT" | "LE" | "LT" | "NE" | "TR" | "FL".x|y|z|w) */
-      /* ("EQ" | "GE" | "GT" | "LE" | "LT" | "NE" | "TR" | "FL".[xyzw]) */
-      if (!Parse_CondCodeMask(parseState, dstReg))
-         RETURN_ERROR;
-
-      if (!Parse_String(parseState, ")"))  /* consume ")" */
-         RETURN_ERROR1("Expected )");
-
-      return GL_TRUE;
-   }
-   else {
-      /* no cond code mask */
-      dstReg->CondMask = COND_TR;
-      dstReg->CondSwizzle = SWIZZLE_NOOP;
-      return GL_TRUE;
-   }
-}
-
-
-/**
- * Parse a vector source (register, constant, etc):
- *   <vectorSrc>    ::= <absVectorSrc>
- *                    | <baseVectorSrc>
- *   <absVectorSrc> ::= <negate> "|" <baseVectorSrc> "|"
- */
-static GLboolean
-Parse_VectorSrc(struct parse_state *parseState,
-                struct prog_src_register *srcReg)
-{
-   GLfloat sign = 1.0F;
-   GLubyte token[100];
-   GLint idx;
-   GLuint negateBase, negateAbs;
-
-   /*
-    * First, take care of +/- and absolute value stuff.
-    */
-   if (Parse_String(parseState, "-"))
-      sign = -1.0F;
-   else if (Parse_String(parseState, "+"))
-      sign = +1.0F;
-
-   if (Parse_String(parseState, "|")) {
-      srcReg->Abs = GL_TRUE;
-      negateAbs = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE;
-
-      if (Parse_String(parseState, "-"))
-         negateBase = NEGATE_XYZW;
-      else if (Parse_String(parseState, "+"))
-         negateBase = NEGATE_NONE;
-      else
-         negateBase = NEGATE_NONE;
-   }
-   else {
-      srcReg->Abs = GL_FALSE;
-      negateAbs = NEGATE_NONE;
-      negateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE;
-   }
-
-   srcReg->Negate = srcReg->Abs ? negateAbs : negateBase;
-
-   /* This should be the real src vector/register name */
-   if (!Peek_Token(parseState, token))
-      RETURN_ERROR;
-
-   /* Src reg can be Rn, Hn, f[n], p[n], a named parameter, a scalar
-    * literal or vector literal.
-    */
-   if (token[0] == 'R' || token[0] == 'H') {
-      srcReg->File = PROGRAM_TEMPORARY;
-      if (!Parse_TempReg(parseState, &idx))
-         RETURN_ERROR;
-      srcReg->Index = idx;
-   }
-   else if (token[0] == 'f') {
-      /* XXX this might be an identifier! */
-      srcReg->File = PROGRAM_INPUT;
-      if (!Parse_FragReg(parseState, &idx))
-         RETURN_ERROR;
-      srcReg->Index = idx;
-   }
-   else if (token[0] == 'p') {
-      /* XXX this might be an identifier! */
-      srcReg->File = PROGRAM_LOCAL_PARAM;
-      if (!Parse_ProgramParamReg(parseState, &idx))
-         RETURN_ERROR;
-      srcReg->Index = idx;
-   }
-   else if (IsLetter(token[0])){
-      GLubyte ident[100];
-      GLint paramIndex;
-      if (!Parse_Identifier(parseState, ident))
-         RETURN_ERROR;
-      paramIndex = _mesa_lookup_parameter_index(parseState->parameters,
-                                                -1, (const char *) ident);
-      if (paramIndex < 0) {
-         RETURN_ERROR2("Undefined constant or parameter: ", ident);
-      }
-      srcReg->File = PROGRAM_NAMED_PARAM;
-      srcReg->Index = paramIndex;      
-   }
-   else if (IsDigit(token[0]) || token[0] == '-' || token[0] == '+' || token[0] == '.'){
-      /* literal scalar constant */
-      GLfloat values[4];
-      GLuint paramIndex;
-      if (!Parse_ScalarConstant(parseState, values))
-         RETURN_ERROR;
-      paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
-                                              values, 4, NULL);
-      srcReg->File = PROGRAM_NAMED_PARAM;
-      srcReg->Index = paramIndex;
-   }
-   else if (token[0] == '{'){
-      /* literal vector constant */
-      GLfloat values[4];
-      GLuint paramIndex;
-      (void) Parse_String(parseState, "{");
-      if (!Parse_VectorConstant(parseState, values))
-         RETURN_ERROR;
-      paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
-                                              values, 4, NULL);
-      srcReg->File = PROGRAM_NAMED_PARAM;
-      srcReg->Index = paramIndex;      
-   }
-   else {
-      RETURN_ERROR2("Invalid source register name", token);
-   }
-
-   /* init swizzle fields */
-   srcReg->Swizzle = SWIZZLE_NOOP;
-
-   /* Look for optional swizzle suffix */
-   if (Parse_String(parseState, ".")) {
-      GLuint swz[4];
-
-      if (!Parse_Token(parseState, token))
-         RETURN_ERROR;
-
-      if (!Parse_SwizzleSuffix(token, swz))
-         RETURN_ERROR1("Invalid swizzle suffix");
-
-      srcReg->Swizzle = MAKE_SWIZZLE4(swz[0], swz[1], swz[2], swz[3]);
-   }
-
-   /* Finish absolute value */
-   if (srcReg->Abs && !Parse_String(parseState, "|")) {
-      RETURN_ERROR1("Expected |");
-   }
-
-   return GL_TRUE;
-}
-
-
-static GLboolean
-Parse_ScalarSrcReg(struct parse_state *parseState,
-                   struct prog_src_register *srcReg)
-{
-   GLubyte token[100];
-   GLfloat sign = 1.0F;
-   GLboolean needSuffix = GL_TRUE;
-   GLint idx;
-   GLuint negateBase, negateAbs;
-
-   /*
-    * First, take care of +/- and absolute value stuff.
-    */
-   if (Parse_String(parseState, "-"))
-      sign = -1.0F;
-   else if (Parse_String(parseState, "+"))
-      sign = +1.0F;
-
-   if (Parse_String(parseState, "|")) {
-      srcReg->Abs = GL_TRUE;
-      negateAbs = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE;
-
-      if (Parse_String(parseState, "-"))
-         negateBase = NEGATE_XYZW;
-      else if (Parse_String(parseState, "+"))
-         negateBase = NEGATE_NONE;
-      else
-         negateBase = NEGATE_NONE;
-   }
-   else {
-      srcReg->Abs = GL_FALSE;
-      negateAbs = NEGATE_NONE;
-      negateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE;
-   }
-
-   srcReg->Negate = srcReg->Abs ? negateAbs : negateBase;
-
-   if (!Peek_Token(parseState, token))
-      RETURN_ERROR;
-
-   /* Src reg can be R<n>, H<n> or a named fragment attrib */
-   if (token[0] == 'R' || token[0] == 'H') {
-      srcReg->File = PROGRAM_TEMPORARY;
-      if (!Parse_TempReg(parseState, &idx))
-         RETURN_ERROR;
-      srcReg->Index = idx;
-   }
-   else if (token[0] == 'f') {
-      srcReg->File = PROGRAM_INPUT;
-      if (!Parse_FragReg(parseState, &idx))
-         RETURN_ERROR;
-      srcReg->Index = idx;
-   }
-   else if (token[0] == '{') {
-      /* vector literal */
-      GLfloat values[4];
-      GLuint paramIndex;
-      (void) Parse_String(parseState, "{");
-      if (!Parse_VectorConstant(parseState, values))
-         RETURN_ERROR;
-      paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
-                                              values, 4, NULL);
-      srcReg->File = PROGRAM_NAMED_PARAM;
-      srcReg->Index = paramIndex;      
-   }
-   else if (IsLetter(token[0])){
-      /* named param/constant */
-      GLubyte ident[100];
-      GLint paramIndex;
-      if (!Parse_Identifier(parseState, ident))
-         RETURN_ERROR;
-      paramIndex = _mesa_lookup_parameter_index(parseState->parameters,
-                                                -1, (const char *) ident);
-      if (paramIndex < 0) {
-         RETURN_ERROR2("Undefined constant or parameter: ", ident);
-      }
-      srcReg->File = PROGRAM_NAMED_PARAM;
-      srcReg->Index = paramIndex;      
-   }
-   else if (IsDigit(token[0])) {
-      /* scalar literal */
-      GLfloat values[4];
-      GLuint paramIndex;
-      if (!Parse_ScalarConstant(parseState, values))
-         RETURN_ERROR;
-      paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
-                                              values, 4, NULL);
-      srcReg->Index = paramIndex;      
-      srcReg->File = PROGRAM_NAMED_PARAM;
-      needSuffix = GL_FALSE;
-   }
-   else {
-      RETURN_ERROR2("Invalid scalar source argument", token);
-   }
-
-   srcReg->Swizzle = 0;
-   if (needSuffix) {
-      /* parse .[xyzw] suffix */
-      if (!Parse_String(parseState, "."))
-         RETURN_ERROR1("Expected .");
-
-      if (!Parse_Token(parseState, token))
-         RETURN_ERROR;
-
-      if (token[0] == 'x' && token[1] == 0) {
-         srcReg->Swizzle = 0;
-      }
-      else if (token[0] == 'y' && token[1] == 0) {
-         srcReg->Swizzle = 1;
-      }
-      else if (token[0] == 'z' && token[1] == 0) {
-         srcReg->Swizzle = 2;
-      }
-      else if (token[0] == 'w' && token[1] == 0) {
-         srcReg->Swizzle = 3;
-      }
-      else {
-         RETURN_ERROR1("Invalid scalar source suffix");
-      }
-   }
-
-   /* Finish absolute value */
-   if (srcReg->Abs && !Parse_String(parseState, "|")) {
-      RETURN_ERROR1("Expected |");
-   }
-
-   return GL_TRUE;
-}
-
-
-static GLboolean
-Parse_PrintInstruction(struct parse_state *parseState,
-                       struct prog_instruction *inst)
-{
-   const GLubyte *str;
-   GLubyte *msg;
-   GLuint len;
-   GLint idx;
-
-   /* The first argument is a literal string 'just like this' */
-   if (!Parse_String(parseState, "'"))
-      RETURN_ERROR1("Expected '");
-
-   str = parseState->pos;
-   for (len = 0; str[len] != '\''; len++) /* find closing quote */
-      ;
-   parseState->pos += len + 1;
-   msg = (GLubyte*) malloc(len + 1);
-
-   memcpy(msg, str, len);
-   msg[len] = 0;
-   inst->Data = msg;
-
-   if (Parse_String(parseState, ",")) {
-      /* got an optional register to print */
-      GLubyte token[100];
-      GetToken(parseState, token);
-      if (token[0] == 'o') {
-         /* dst reg */
-         if (!Parse_OutputReg(parseState, &idx))
-            RETURN_ERROR;
-	 inst->SrcReg[0].Index = idx;
-         inst->SrcReg[0].File = PROGRAM_OUTPUT;
-      }
-      else {
-         /* src reg */
-         if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
-            RETURN_ERROR;
-      }
-   }
-   else {
-      inst->SrcReg[0].File = PROGRAM_UNDEFINED;
-   }
-
-   inst->SrcReg[0].Swizzle = SWIZZLE_NOOP;
-   inst->SrcReg[0].Abs = GL_FALSE;
-   inst->SrcReg[0].Negate = NEGATE_NONE;
-
-   return GL_TRUE;
-}
-
-
-static GLboolean
-Parse_InstructionSequence(struct parse_state *parseState,
-                          struct prog_instruction program[])
-{
-   while (1) {
-      struct prog_instruction *inst = program + parseState->numInst;
-      struct instruction_pattern instMatch;
-      GLubyte token[100];
-
-      /* Initialize the instruction */
-      _mesa_init_instructions(inst, 1);
-
-      /* special instructions */
-      if (Parse_String(parseState, "DEFINE")) {
-         GLubyte id[100];
-         GLfloat value[7];  /* yes, 7 to be safe */
-         if (!Parse_Identifier(parseState, id))
-            RETURN_ERROR;
-         /* XXX make sure id is not a reserved identifer, like R9 */
-         if (!Parse_String(parseState, "="))
-            RETURN_ERROR1("Expected =");
-         if (!Parse_VectorOrScalarConstant(parseState, value))
-            RETURN_ERROR;
-         if (!Parse_String(parseState, ";"))
-            RETURN_ERROR1("Expected ;");
-         if (_mesa_lookup_parameter_index(parseState->parameters,
-                                          -1, (const char *) id) >= 0) {
-            RETURN_ERROR2(id, "already defined");
-         }
-         _mesa_add_named_parameter(parseState->parameters,
-                                   (const char *) id, value);
-      }
-      else if (Parse_String(parseState, "DECLARE")) {
-         GLubyte id[100];
-         GLfloat value[7] = {0, 0, 0, 0, 0, 0, 0};  /* yes, to be safe */
-         if (!Parse_Identifier(parseState, id))
-            RETURN_ERROR;
-         /* XXX make sure id is not a reserved identifer, like R9 */
-         if (Parse_String(parseState, "=")) {
-            if (!Parse_VectorOrScalarConstant(parseState, value))
-               RETURN_ERROR;
-         }
-         if (!Parse_String(parseState, ";"))
-            RETURN_ERROR1("Expected ;");
-         if (_mesa_lookup_parameter_index(parseState->parameters,
-                                          -1, (const char *) id) >= 0) {
-            RETURN_ERROR2(id, "already declared");
-         }
-         _mesa_add_named_parameter(parseState->parameters,
-                                   (const char *) id, value);
-      }
-      else if (Parse_String(parseState, "END")) {
-         inst->Opcode = OPCODE_END;
-         parseState->numInst++;
-         if (Parse_Token(parseState, token)) {
-            RETURN_ERROR1("Code after END opcode.");
-         }
-         break;
-      }
-      else {
-         /* general/arithmetic instruction */
-
-         /* get token */
-         if (!Parse_Token(parseState, token)) {
-            RETURN_ERROR1("Missing END instruction.");
-         }
-
-         /* try to find matching instuction */
-         instMatch = MatchInstruction(token);
-         if (instMatch.opcode >= MAX_OPCODE) {
-            /* bad instruction name */
-            RETURN_ERROR2("Unexpected token: ", token);
-         }
-
-         inst->Opcode = instMatch.opcode;
-         inst->Precision = instMatch.suffixes & (_R | _H | _X);
-         inst->SaturateMode = (instMatch.suffixes & (_S))
-            ? SATURATE_ZERO_ONE : SATURATE_OFF;
-         inst->CondUpdate = (instMatch.suffixes & (_C)) ? GL_TRUE : GL_FALSE;
-
-         /*
-          * parse the input and output operands
-          */
-         if (instMatch.outputs == OUTPUT_S || instMatch.outputs == OUTPUT_V) {
-            if (!Parse_MaskedDstReg(parseState, &inst->DstReg))
-               RETURN_ERROR;
-            if (!Parse_String(parseState, ","))
-               RETURN_ERROR1("Expected ,");
-         }
-         else if (instMatch.outputs == OUTPUT_NONE) {
-            if (instMatch.opcode == OPCODE_KIL_NV) {
-               /* This is a little weird, the cond code info is in
-                * the dest register.
-                */
-               if (!Parse_CondCodeMask(parseState, &inst->DstReg))
-                  RETURN_ERROR;
-            }
-            else {
-               ASSERT(instMatch.opcode == OPCODE_PRINT);
-            }
-         }
-
-         if (instMatch.inputs == INPUT_1V) {
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
-               RETURN_ERROR;
-         }
-         else if (instMatch.inputs == INPUT_2V) {
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
-               RETURN_ERROR;
-            if (!Parse_String(parseState, ","))
-               RETURN_ERROR1("Expected ,");
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[1]))
-               RETURN_ERROR;
-         }
-         else if (instMatch.inputs == INPUT_3V) {
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
-               RETURN_ERROR;
-            if (!Parse_String(parseState, ","))
-               RETURN_ERROR1("Expected ,");
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[1]))
-               RETURN_ERROR;
-            if (!Parse_String(parseState, ","))
-               RETURN_ERROR1("Expected ,");
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[2]))
-               RETURN_ERROR;
-         }
-         else if (instMatch.inputs == INPUT_1S) {
-            if (!Parse_ScalarSrcReg(parseState, &inst->SrcReg[0]))
-               RETURN_ERROR;
-         }
-         else if (instMatch.inputs == INPUT_2S) {
-            if (!Parse_ScalarSrcReg(parseState, &inst->SrcReg[0]))
-               RETURN_ERROR;
-            if (!Parse_String(parseState, ","))
-               RETURN_ERROR1("Expected ,");
-            if (!Parse_ScalarSrcReg(parseState, &inst->SrcReg[1]))
-               RETURN_ERROR;
-         }
-         else if (instMatch.inputs == INPUT_CC) {
-            /* XXX to-do */
-         }
-         else if (instMatch.inputs == INPUT_1V_T) {
-	    GLubyte unit, idx;
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
-               RETURN_ERROR;
-            if (!Parse_String(parseState, ","))
-               RETURN_ERROR1("Expected ,");
-            if (!Parse_TextureImageId(parseState, &unit, &idx))
-               RETURN_ERROR;
-	    inst->TexSrcUnit = unit;
-	    inst->TexSrcTarget = idx;
-         }
-         else if (instMatch.inputs == INPUT_3V_T) {
-	    GLubyte unit, idx;
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
-               RETURN_ERROR;
-            if (!Parse_String(parseState, ","))
-               RETURN_ERROR1("Expected ,");
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[1]))
-               RETURN_ERROR;
-            if (!Parse_String(parseState, ","))
-               RETURN_ERROR1("Expected ,");
-            if (!Parse_VectorSrc(parseState, &inst->SrcReg[2]))
-               RETURN_ERROR;
-            if (!Parse_String(parseState, ","))
-               RETURN_ERROR1("Expected ,");
-            if (!Parse_TextureImageId(parseState, &unit, &idx))
-               RETURN_ERROR;
-	    inst->TexSrcUnit = unit;
-	    inst->TexSrcTarget = idx;
-         }
-         else if (instMatch.inputs == INPUT_1V_S) {
-            if (!Parse_PrintInstruction(parseState, inst))
-               RETURN_ERROR;
-         }
-
-         /* end of statement semicolon */
-         if (!Parse_String(parseState, ";"))
-            RETURN_ERROR1("Expected ;");
-
-         parseState->numInst++;
-
-         if (parseState->numInst >= MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS)
-            RETURN_ERROR1("Program too long");
-      }
-   }
-   return GL_TRUE;
-}
-
-
-
-/**
- * Parse/compile the 'str' returning the compiled 'program'.
- * ctx->Program.ErrorPos will be -1 if successful.  Otherwise, ErrorPos
- * indicates the position of the error in 'str'.
- */
-void
-_mesa_parse_nv_fragment_program(struct gl_context *ctx, GLenum dstTarget,
-                                const GLubyte *str, GLsizei len,
-                                struct gl_fragment_program *program)
-{
-   struct parse_state parseState;
-   struct prog_instruction instBuffer[MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS];
-   struct prog_instruction *newInst;
-   GLenum target;
-   GLubyte *programString;
-
-   /* Make a null-terminated copy of the program string */
-   programString = (GLubyte *) MALLOC(len + 1);
-   if (!programString) {
-      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glLoadProgramNV");
-      return;
-   }
-   memcpy(programString, str, len);
-   programString[len] = 0;
-
-   /* Get ready to parse */
-   memset(&parseState, 0, sizeof(struct parse_state));
-   parseState.ctx = ctx;
-   parseState.start = programString;
-   parseState.program = program;
-   parseState.numInst = 0;
-   parseState.curLine = programString;
-   parseState.parameters = _mesa_new_parameter_list();
-
-   /* Reset error state */
-   _mesa_set_program_error(ctx, -1, NULL);
-
-   /* check the program header */
-   if (strncmp((const char *) programString, "!!FP1.0", 7) == 0) {
-      target = GL_FRAGMENT_PROGRAM_NV;
-      parseState.pos = programString + 7;
-   }
-   else if (strncmp((const char *) programString, "!!FCP1.0", 8) == 0) {
-      /* fragment / register combiner program - not supported */
-      _mesa_set_program_error(ctx, 0, "Invalid fragment program header");
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV(bad header)");
-      return;
-   }
-   else {
-      /* invalid header */
-      _mesa_set_program_error(ctx, 0, "Invalid fragment program header");
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV(bad header)");
-      return;
-   }
-
-   /* make sure target and header match */
-   if (target != dstTarget) {
-      _mesa_error(ctx, GL_INVALID_OPERATION,
-                  "glLoadProgramNV(target mismatch 0x%x != 0x%x)",
-                  target, dstTarget);
-      return;
-   }
-
-   if (Parse_InstructionSequence(&parseState, instBuffer)) {
-      GLuint u;
-      /* successful parse! */
-
-      if (parseState.outputsWritten == 0) {
-         /* must write at least one output! */
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "Invalid fragment program - no outputs written.");
-         return;
-      }
-
-      /* copy the compiled instructions */
-      assert(parseState.numInst <= MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS);
-      newInst = _mesa_alloc_instructions(parseState.numInst);
-      if (!newInst) {
-         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glLoadProgramNV");
-         return;  /* out of memory */
-      }
-      _mesa_copy_instructions(newInst, instBuffer, parseState.numInst);
-
-      /* install the program */
-      program->Base.Target = target;
-      if (program->Base.String) {
-         FREE(program->Base.String);
-      }
-      program->Base.String = programString;
-      program->Base.Format = GL_PROGRAM_FORMAT_ASCII_ARB;
-      if (program->Base.Instructions) {
-         free(program->Base.Instructions);
-      }
-      program->Base.Instructions = newInst;
-      program->Base.NumInstructions = parseState.numInst;
-      program->Base.InputsRead = parseState.inputsRead;
-      program->Base.OutputsWritten = parseState.outputsWritten;
-      for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++)
-         program->Base.TexturesUsed[u] = parseState.texturesUsed[u];
-
-      /* save program parameters */
-      program->Base.Parameters = parseState.parameters;
-
-      /* allocate registers for declared program parameters */
-#if 00
-      _mesa_assign_program_registers(&(program->SymbolTable));
-#endif
-
-#ifdef DEBUG_foo
-      printf("--- glLoadProgramNV(%d) result ---\n", program->Base.Id);
-      _mesa_fprint_program_opt(stdout, &program->Base, PROG_PRINT_NV, 0);
-      printf("----------------------------------\n");
-#endif
-   }
-   else {
-      /* Error! */
-      _mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV");
-      /* NOTE: _mesa_set_program_error would have been called already */
-   }
-}
-
-
-const char *
-_mesa_nv_fragment_input_register_name(GLuint i)
-{
-   ASSERT(i < MAX_NV_FRAGMENT_PROGRAM_INPUTS);
-   return InputRegisters[i];
-}
-
+/*
+ * 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.
+ */
+
+/**
+ * \file nvfragparse.c
+ * NVIDIA fragment program parser.
+ * \author Brian Paul
+ */
+
+/*
+ * Regarding GL_NV_fragment_program:
+ *
+ * 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/context.h"
+#include "main/imports.h"
+#include "main/macros.h"
+#include "program.h"
+#include "prog_parameter.h"
+#include "prog_print.h"
+#include "prog_instruction.h"
+#include "nvfragparse.h"
+
+
+#define INPUT_1V     1
+#define INPUT_2V     2
+#define INPUT_3V     3
+#define INPUT_1S     4
+#define INPUT_2S     5
+#define INPUT_CC     6
+#define INPUT_1V_T   7  /* one source vector, plus textureId */
+#define INPUT_3V_T   8  /* one source vector, plus textureId */
+#define INPUT_NONE   9
+#define INPUT_1V_S  10  /* a string and a vector register */
+#define OUTPUT_V    20
+#define OUTPUT_S    21
+#define OUTPUT_NONE 22
+
+/* IRIX defines some of these */
+#undef _R
+#undef _H
+#undef _X
+#undef _C
+#undef _S
+
+/* Optional suffixes */
+#define _R  FLOAT32  /* float */
+#define _H  FLOAT16  /* half-float */
+#define _X  FIXED12  /* fixed */
+#define _C  0x08     /* set cond codes */
+#define _S  0x10     /* saturate, clamp result to [0,1] */
+
+struct instruction_pattern {
+   const char *name;
+   enum prog_opcode opcode;
+   GLuint inputs;
+   GLuint outputs;
+   GLuint suffixes;
+};
+
+static const struct instruction_pattern Instructions[] = {
+   { "ADD", OPCODE_ADD, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "COS", OPCODE_COS, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
+   { "DDX", OPCODE_DDX, INPUT_1V, OUTPUT_V, _R | _H |      _C | _S },
+   { "DDY", OPCODE_DDY, INPUT_1V, OUTPUT_V, _R | _H |      _C | _S },
+   { "DP3", OPCODE_DP3, INPUT_2V, OUTPUT_S, _R | _H | _X | _C | _S },
+   { "DP4", OPCODE_DP4, INPUT_2V, OUTPUT_S, _R | _H | _X | _C | _S },
+   { "DST", OPCODE_DP4, INPUT_2V, OUTPUT_V, _R | _H |      _C | _S },
+   { "EX2", OPCODE_DP4, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
+   { "FLR", OPCODE_FLR, INPUT_1V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "FRC", OPCODE_FRC, INPUT_1V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "KIL", OPCODE_KIL_NV, INPUT_CC, OUTPUT_NONE, 0                },
+   { "LG2", OPCODE_LG2, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
+   { "LIT", OPCODE_LIT, INPUT_1V, OUTPUT_V, _R | _H |      _C | _S },
+   { "LRP", OPCODE_LRP, INPUT_3V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "MAD", OPCODE_MAD, INPUT_3V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "MAX", OPCODE_MAX, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "MIN", OPCODE_MIN, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "MOV", OPCODE_MOV, INPUT_1V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "MUL", OPCODE_MUL, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "PK2H",  OPCODE_PK2H,  INPUT_1V, OUTPUT_S, 0                  },
+   { "PK2US", OPCODE_PK2US, INPUT_1V, OUTPUT_S, 0                  },
+   { "PK4B",  OPCODE_PK4B,  INPUT_1V, OUTPUT_S, 0                  },
+   { "PK4UB", OPCODE_PK4UB, INPUT_1V, OUTPUT_S, 0                  },
+   { "POW", OPCODE_POW, INPUT_2S, OUTPUT_S, _R | _H |      _C | _S },
+   { "RCP", OPCODE_RCP, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
+   { "RFL", OPCODE_RFL, INPUT_2V, OUTPUT_V, _R | _H |      _C | _S },
+   { "RSQ", OPCODE_RSQ, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
+   { "SEQ", OPCODE_SEQ, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "SFL", OPCODE_SFL, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "SGE", OPCODE_SGE, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "SGT", OPCODE_SGT, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "SIN", OPCODE_SIN, INPUT_1S, OUTPUT_S, _R | _H |      _C | _S },
+   { "SLE", OPCODE_SLE, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "SLT", OPCODE_SLT, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "SNE", OPCODE_SNE, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "STR", OPCODE_STR, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "SUB", OPCODE_SUB, INPUT_2V, OUTPUT_V, _R | _H | _X | _C | _S },
+   { "TEX", OPCODE_TEX, INPUT_1V_T, OUTPUT_V,              _C | _S },
+   { "TXD", OPCODE_TXD, INPUT_3V_T, OUTPUT_V,              _C | _S },
+   { "TXP", OPCODE_TXP_NV, INPUT_1V_T, OUTPUT_V,           _C | _S },
+   { "UP2H",  OPCODE_UP2H,  INPUT_1S, OUTPUT_V,            _C | _S },
+   { "UP2US", OPCODE_UP2US, INPUT_1S, OUTPUT_V,            _C | _S },
+   { "UP4B",  OPCODE_UP4B,  INPUT_1S, OUTPUT_V,            _C | _S },
+   { "UP4UB", OPCODE_UP4UB, INPUT_1S, OUTPUT_V,            _C | _S },
+   { "X2D", OPCODE_X2D, INPUT_3V, OUTPUT_V, _R | _H |      _C | _S },
+   { "PRINT", OPCODE_PRINT, INPUT_1V_S, OUTPUT_NONE, 0               },
+   { NULL, (enum prog_opcode) -1, 0, 0, 0 }
+};
+
+
+/*
+ * Information needed or computed during parsing.
+ * Remember, we can't modify the target program object until we've
+ * _successfully_ parsed the program text.
+ */
+struct parse_state {
+   struct gl_context *ctx;
+   const GLubyte *start;              /* start of program string */
+   const GLubyte *pos;                /* current position */
+   const GLubyte *curLine;
+   struct gl_fragment_program *program;  /* current program */
+
+   struct gl_program_parameter_list *parameters;
+
+   GLuint numInst;                    /* number of instructions parsed */
+   GLuint inputsRead;                 /* bitmask of input registers used */
+   GLuint outputsWritten;             /* bitmask of 1 << FRAG_OUTPUT_* bits */
+   GLuint texturesUsed[MAX_TEXTURE_IMAGE_UNITS];
+};
+
+
+
+/*
+ * Called whenever we find an error during parsing.
+ */
+static void
+record_error(struct parse_state *parseState, const char *msg, int lineNo)
+{
+#ifdef DEBUG
+   GLint line, column;
+   const GLubyte *lineStr;
+   lineStr = _mesa_find_line_column(parseState->start,
+                                    parseState->pos, &line, &column);
+   _mesa_debug(parseState->ctx,
+               "nvfragparse.c(%d): line %d, column %d:%s (%s)\n",
+               lineNo, line, column, (char *) lineStr, msg);
+   free((void *) lineStr);
+#else
+   (void) lineNo;
+#endif
+
+   /* Check that no error was already recorded.  Only record the first one. */
+   if (parseState->ctx->Program.ErrorString[0] == 0) {
+      _mesa_set_program_error(parseState->ctx,
+                              parseState->pos - parseState->start,
+                              msg);
+   }
+}
+
+
+#define RETURN_ERROR							\
+do {									\
+   record_error(parseState, "Unexpected end of input.", __LINE__);	\
+   return GL_FALSE;							\
+} while(0)
+
+#define RETURN_ERROR1(msg)						\
+do {									\
+   record_error(parseState, msg, __LINE__);				\
+   return GL_FALSE;							\
+} while(0)
+
+#define RETURN_ERROR2(msg1, msg2)					\
+do {									\
+   char err[1000];							\
+   sprintf(err, "%s %s", msg1, msg2);				\
+   record_error(parseState, err, __LINE__);				\
+   return GL_FALSE;							\
+} while(0)
+
+
+
+
+/*
+ * Search a list of instruction structures for a match.
+ */
+static struct instruction_pattern
+MatchInstruction(const GLubyte *token)
+{
+   const struct instruction_pattern *inst;
+   struct instruction_pattern result;
+
+   result.name = NULL;
+   result.opcode = MAX_OPCODE; /* i.e. invalid instruction */
+   result.inputs = 0;
+   result.outputs = 0;
+   result.suffixes = 0;
+
+   for (inst = Instructions; inst->name; inst++) {
+      if (strncmp((const char *) token, inst->name, 3) == 0) {
+         /* matched! */
+         int i = 3;
+         result = *inst;
+         result.suffixes = 0;
+         /* look at suffix */
+         if (token[i] == 'R') {
+            result.suffixes |= _R;
+            i++;
+         }
+         else if (token[i] == 'H') {
+            result.suffixes |= _H;
+            i++;
+         }
+         else if (token[i] == 'X') {
+            result.suffixes |= _X;
+            i++;
+         }
+         if (token[i] == 'C') {
+            result.suffixes |= _C;
+            i++;
+         }
+         if (token[i] == '_' && token[i+1] == 'S' &&
+             token[i+2] == 'A' && token[i+3] == 'T') {
+            result.suffixes |= _S;
+         }
+         return result;
+      }
+   }
+
+   return result;
+}
+
+
+
+
+/**********************************************************************/
+
+
+static GLboolean IsLetter(GLubyte b)
+{
+   return (b >= 'a' && b <= 'z') ||
+          (b >= 'A' && b <= 'Z') ||
+          (b == '_') ||
+          (b == '$');
+}
+
+
+static GLboolean IsDigit(GLubyte b)
+{
+   return b >= '0' && b <= '9';
+}
+
+
+static GLboolean IsWhitespace(GLubyte b)
+{
+   return b == ' ' || b == '\t' || b == '\n' || b == '\r';
+}
+
+
+/**
+ * Starting at 'str' find the next token.  A token can be an integer,
+ * an identifier or punctuation symbol.
+ * \return <= 0 we found an error, else, return number of characters parsed.
+ */
+static GLint
+GetToken(struct parse_state *parseState, GLubyte *token)
+{
+   const GLubyte *str = parseState->pos;
+   GLint i = 0, j = 0;
+
+   token[0] = 0;
+
+   /* skip whitespace and comments */
+   while (str[i] && (IsWhitespace(str[i]) || str[i] == '#')) {
+      if (str[i] == '#') {
+         /* skip comment */
+         while (str[i] && (str[i] != '\n' && str[i] != '\r')) {
+            i++;
+         }
+         if (str[i] == '\n' || str[i] == '\r')
+            parseState->curLine = str + i + 1;
+      }
+      else {
+         /* skip whitespace */
+         if (str[i] == '\n' || str[i] == '\r')
+            parseState->curLine = str + i + 1;
+         i++;
+      }
+   }
+
+   if (str[i] == 0)
+      return -i;
+
+   /* try matching an integer */
+   while (str[i] && IsDigit(str[i])) {
+      token[j++] = str[i++];
+   }
+   if (j > 0 || !str[i]) {
+      token[j] = 0;
+      return i;
+   }
+
+   /* try matching an identifier */
+   if (IsLetter(str[i])) {
+      while (str[i] && (IsLetter(str[i]) || IsDigit(str[i]))) {
+         token[j++] = str[i++];
+      }
+      token[j] = 0;
+      return i;
+   }
+
+   /* punctuation character */
+   if (str[i]) {
+      token[0] = str[i++];
+      token[1] = 0;
+      return i;
+   }
+
+   /* end of input */
+   token[0] = 0;
+   return i;
+}
+
+
+/**
+ * Get next token from input stream and increment stream pointer past token.
+ */
+static GLboolean
+Parse_Token(struct parse_state *parseState, GLubyte *token)
+{
+   GLint i;
+   i = GetToken(parseState, token);
+   if (i <= 0) {
+      parseState->pos += (-i);
+      return GL_FALSE;
+   }
+   parseState->pos += i;
+   return GL_TRUE;
+}
+
+
+/**
+ * Get next token from input stream but don't increment stream pointer.
+ */
+static GLboolean
+Peek_Token(struct parse_state *parseState, GLubyte *token)
+{
+   GLint i, len;
+   i = GetToken(parseState, token);
+   if (i <= 0) {
+      parseState->pos += (-i);
+      return GL_FALSE;
+   }
+   len = (GLint) strlen((const char *) token);
+   parseState->pos += (i - len);
+   return GL_TRUE;
+}
+
+
+/**********************************************************************/
+
+static const char *InputRegisters[MAX_NV_FRAGMENT_PROGRAM_INPUTS + 1] = {
+   "WPOS", "COL0", "COL1", "FOGC",
+   "TEX0", "TEX1", "TEX2", "TEX3", "TEX4", "TEX5", "TEX6", "TEX7", NULL
+};
+
+
+
+/**********************************************************************/
+
+/**
+ * Try to match 'pattern' as the next token after any whitespace/comments.
+ */
+static GLboolean
+Parse_String(struct parse_state *parseState, const char *pattern)
+{
+   const GLubyte *m;
+   GLint i;
+
+   /* skip whitespace and comments */
+   while (IsWhitespace(*parseState->pos) || *parseState->pos == '#') {
+      if (*parseState->pos == '#') {
+         while (*parseState->pos && (*parseState->pos != '\n' && *parseState->pos != '\r')) {
+            parseState->pos += 1;
+         }
+         if (*parseState->pos == '\n' || *parseState->pos == '\r')
+            parseState->curLine = parseState->pos + 1;
+      }
+      else {
+         /* skip whitespace */
+         if (*parseState->pos == '\n' || *parseState->pos == '\r')
+            parseState->curLine = parseState->pos + 1;
+         parseState->pos += 1;
+      }
+   }
+
+   /* Try to match the pattern */
+   m = parseState->pos;
+   for (i = 0; pattern[i]; i++) {
+      if (*m != (GLubyte) pattern[i])
+         return GL_FALSE;
+      m += 1;
+   }
+   parseState->pos = m;
+
+   return GL_TRUE; /* success */
+}
+
+
+static GLboolean
+Parse_Identifier(struct parse_state *parseState, GLubyte *ident)
+{
+   if (!Parse_Token(parseState, ident))
+      RETURN_ERROR;
+   if (IsLetter(ident[0]))
+      return GL_TRUE;
+   else
+      RETURN_ERROR1("Expected an identfier");
+}
+
+
+/**
+ * Parse a floating point constant, or a defined symbol name.
+ * [+/-]N[.N[eN]]
+ * Output:  number[0 .. 3] will get the value.
+ */
+static GLboolean
+Parse_ScalarConstant(struct parse_state *parseState, GLfloat *number)
+{
+   char *end = NULL;
+
+   *number = (GLfloat) _mesa_strtof((const char *) parseState->pos, &end);
+
+   if (end && end > (char *) parseState->pos) {
+      /* got a number */
+      parseState->pos = (GLubyte *) end;
+      number[1] = *number;
+      number[2] = *number;
+      number[3] = *number;
+      return GL_TRUE;
+   }
+   else {
+      /* should be an identifier */
+      GLubyte ident[100];
+      const GLfloat *constant;
+      if (!Parse_Identifier(parseState, ident))
+         RETURN_ERROR1("Expected an identifier");
+      constant = (GLfloat *)_mesa_lookup_parameter_value(parseState->parameters,
+                                                         -1, 
+                                                         (const char *) ident);
+      /* XXX Check that it's a constant and not a parameter */
+      if (!constant) {
+         RETURN_ERROR1("Undefined symbol");
+      }
+      else {
+         COPY_4V(number, constant);
+         return GL_TRUE;
+      }
+   }
+}
+
+
+
+/**
+ * Parse a vector constant, one of:
+ *   { float }
+ *   { float, float }
+ *   { float, float, float }
+ *   { float, float, float, float }
+ */
+static GLboolean
+Parse_VectorConstant(struct parse_state *parseState, GLfloat *vec)
+{
+   /* "{" was already consumed */
+
+   ASSIGN_4V(vec, 0.0, 0.0, 0.0, 1.0);
+
+   if (!Parse_ScalarConstant(parseState, vec+0))  /* X */
+      return GL_FALSE;
+
+   if (Parse_String(parseState, "}")) {
+      return GL_TRUE;
+   }
+
+   if (!Parse_String(parseState, ","))
+      RETURN_ERROR1("Expected comma in vector constant");
+
+   if (!Parse_ScalarConstant(parseState, vec+1))  /* Y */
+      return GL_FALSE;
+
+   if (Parse_String(parseState, "}")) {
+      return GL_TRUE;
+   }
+
+   if (!Parse_String(parseState, ","))
+      RETURN_ERROR1("Expected comma in vector constant");
+
+   if (!Parse_ScalarConstant(parseState, vec+2))  /* Z */
+      return GL_FALSE;
+
+   if (Parse_String(parseState, "}")) {
+      return GL_TRUE;
+   }
+
+   if (!Parse_String(parseState, ","))
+      RETURN_ERROR1("Expected comma in vector constant");
+
+   if (!Parse_ScalarConstant(parseState, vec+3))  /* W */
+      return GL_FALSE;
+
+   if (!Parse_String(parseState, "}"))
+      RETURN_ERROR1("Expected closing brace in vector constant");
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Parse <number>, <varname> or {a, b, c, d}.
+ * Return number of values in the vector or scalar, or zero if parse error.
+ */
+static GLuint
+Parse_VectorOrScalarConstant(struct parse_state *parseState, GLfloat *vec)
+{
+   if (Parse_String(parseState, "{")) {
+      return Parse_VectorConstant(parseState, vec);
+   }
+   else {
+      GLboolean b = Parse_ScalarConstant(parseState, vec);
+      if (b) {
+         vec[1] = vec[2] = vec[3] = vec[0];
+      }
+      return b;
+   }
+}
+
+
+/**
+ * Parse a texture image source:
+ *    [TEX0 | TEX1 | .. | TEX15] , [1D | 2D | 3D | CUBE | RECT]
+ */
+static GLboolean
+Parse_TextureImageId(struct parse_state *parseState,
+                     GLubyte *texUnit, GLubyte *texTargetBit)
+{
+   GLubyte imageSrc[100];
+   GLint unit;
+
+   if (!Parse_Token(parseState, imageSrc))
+      RETURN_ERROR;
+   
+   if (imageSrc[0] != 'T' ||
+       imageSrc[1] != 'E' ||
+       imageSrc[2] != 'X') {
+      RETURN_ERROR1("Expected TEX# source");
+   }
+   unit = atoi((const char *) imageSrc + 3);
+   if ((unit < 0 || unit > MAX_TEXTURE_IMAGE_UNITS) ||
+       (unit == 0 && (imageSrc[3] != '0' || imageSrc[4] != 0))) {
+      RETURN_ERROR1("Invalied TEX# source index");
+   }
+   *texUnit = unit;
+
+   if (!Parse_String(parseState, ","))
+      RETURN_ERROR1("Expected ,");
+
+   if (Parse_String(parseState, "1D")) {
+      *texTargetBit = TEXTURE_1D_BIT;
+   }
+   else if (Parse_String(parseState, "2D")) {
+      *texTargetBit = TEXTURE_2D_BIT;
+   }
+   else if (Parse_String(parseState, "3D")) {
+      *texTargetBit = TEXTURE_3D_BIT;
+   }
+   else if (Parse_String(parseState, "CUBE")) {
+      *texTargetBit = TEXTURE_CUBE_BIT;
+   }
+   else if (Parse_String(parseState, "RECT")) {
+      *texTargetBit = TEXTURE_RECT_BIT;
+   }
+   else {
+      RETURN_ERROR1("Invalid texture target token");
+   }
+
+   /* update record of referenced texture units */
+   parseState->texturesUsed[*texUnit] |= *texTargetBit;
+   if (_mesa_bitcount(parseState->texturesUsed[*texUnit]) > 1) {
+      RETURN_ERROR1("Only one texture target can be used per texture unit.");
+   }
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Parse a scalar suffix like .x, .y, .z or .w or parse a swizzle suffix
+ * like .wxyz, .xxyy, etc and return the swizzle indexes.
+ */
+static GLboolean
+Parse_SwizzleSuffix(const GLubyte *token, GLuint swizzle[4])
+{
+   if (token[1] == 0) {
+      /* single letter swizzle (scalar) */
+      if (token[0] == 'x')
+         ASSIGN_4V(swizzle, 0, 0, 0, 0);
+      else if (token[0] == 'y')
+         ASSIGN_4V(swizzle, 1, 1, 1, 1);
+      else if (token[0] == 'z')
+         ASSIGN_4V(swizzle, 2, 2, 2, 2);
+      else if (token[0] == 'w')
+         ASSIGN_4V(swizzle, 3, 3, 3, 3);
+      else
+         return GL_FALSE;
+   }
+   else {
+      /* 4-component swizzle (vector) */
+      GLint k;
+      for (k = 0; k < 4 && token[k]; k++) {
+         if (token[k] == 'x')
+            swizzle[k] = 0;
+         else if (token[k] == 'y')
+            swizzle[k] = 1;
+         else if (token[k] == 'z')
+            swizzle[k] = 2;
+         else if (token[k] == 'w')
+            swizzle[k] = 3;
+         else
+            return GL_FALSE;
+      }
+      if (k != 4)
+         return GL_FALSE;
+   }
+   return GL_TRUE;
+}
+
+
+static GLboolean
+Parse_CondCodeMask(struct parse_state *parseState,
+                   struct prog_dst_register *dstReg)
+{
+   if (Parse_String(parseState, "EQ"))
+      dstReg->CondMask = COND_EQ;
+   else if (Parse_String(parseState, "GE"))
+      dstReg->CondMask = COND_GE;
+   else if (Parse_String(parseState, "GT"))
+      dstReg->CondMask = COND_GT;
+   else if (Parse_String(parseState, "LE"))
+      dstReg->CondMask = COND_LE;
+   else if (Parse_String(parseState, "LT"))
+      dstReg->CondMask = COND_LT;
+   else if (Parse_String(parseState, "NE"))
+      dstReg->CondMask = COND_NE;
+   else if (Parse_String(parseState, "TR"))
+      dstReg->CondMask = COND_TR;
+   else if (Parse_String(parseState, "FL"))
+      dstReg->CondMask = COND_FL;
+   else
+      RETURN_ERROR1("Invalid condition code mask");
+
+   /* look for optional .xyzw swizzle */
+   if (Parse_String(parseState, ".")) {
+      GLubyte token[100];
+      GLuint swz[4];
+
+      if (!Parse_Token(parseState, token))  /* get xyzw suffix */
+         RETURN_ERROR;
+
+      if (!Parse_SwizzleSuffix(token, swz))
+         RETURN_ERROR1("Invalid swizzle suffix");
+
+      dstReg->CondSwizzle = MAKE_SWIZZLE4(swz[0], swz[1], swz[2], swz[3]);
+   }
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Parse a temporary register: Rnn or Hnn
+ */
+static GLboolean
+Parse_TempReg(struct parse_state *parseState, GLint *tempRegNum)
+{
+   GLubyte token[100];
+
+   /* Should be 'R##' or 'H##' */
+   if (!Parse_Token(parseState, token))
+      RETURN_ERROR;
+   if (token[0] != 'R' && token[0] != 'H')
+      RETURN_ERROR1("Expected R## or H##");
+
+   if (IsDigit(token[1])) {
+      GLint reg = atoi((const char *) (token + 1));
+      if (token[0] == 'H')
+         reg += 32;
+      if (reg >= MAX_NV_FRAGMENT_PROGRAM_TEMPS)
+         RETURN_ERROR1("Invalid temporary register name");
+      *tempRegNum = reg;
+   }
+   else {
+      RETURN_ERROR1("Invalid temporary register name");
+   }
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Parse a write-only dummy register: RC or HC.
+ */
+static GLboolean
+Parse_DummyReg(struct parse_state *parseState, GLint *regNum)
+{
+   if (Parse_String(parseState, "RC")) {
+       *regNum = 0;
+   }
+   else if (Parse_String(parseState, "HC")) {
+       *regNum = 1;
+   }
+   else {
+      RETURN_ERROR1("Invalid write-only register name");
+   }
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Parse a program local parameter register "p[##]"
+ */
+static GLboolean
+Parse_ProgramParamReg(struct parse_state *parseState, GLint *regNum)
+{
+   GLubyte token[100];
+
+   if (!Parse_String(parseState, "p["))
+      RETURN_ERROR1("Expected p[");
+
+   if (!Parse_Token(parseState, token))
+      RETURN_ERROR;
+
+   if (IsDigit(token[0])) {
+      /* a numbered program parameter register */
+      GLint reg = atoi((const char *) token);
+      if (reg >= MAX_NV_FRAGMENT_PROGRAM_PARAMS)
+         RETURN_ERROR1("Invalid constant program number");
+      *regNum = reg;
+   }
+   else {
+      RETURN_ERROR;
+   }
+
+   if (!Parse_String(parseState, "]"))
+      RETURN_ERROR1("Expected ]");
+
+   return GL_TRUE;
+}
+
+
+/**
+ * Parse f[name]  - fragment input register
+ */
+static GLboolean
+Parse_FragReg(struct parse_state *parseState, GLint *tempRegNum)
+{
+   GLubyte token[100];
+   GLint j;
+
+   /* Match 'f[' */
+   if (!Parse_String(parseState, "f["))
+      RETURN_ERROR1("Expected f[");
+
+   /* get <name> and look for match */
+   if (!Parse_Token(parseState, token)) {
+      RETURN_ERROR;
+   }
+   for (j = 0; InputRegisters[j]; j++) {
+      if (strcmp((const char *) token, InputRegisters[j]) == 0) {
+         *tempRegNum = j;
+         parseState->inputsRead |= (1 << j);
+         break;
+      }
+   }
+   if (!InputRegisters[j]) {
+      /* unknown input register label */
+      RETURN_ERROR2("Invalid register name", token);
+   }
+
+   /* Match '[' */
+   if (!Parse_String(parseState, "]"))
+      RETURN_ERROR1("Expected ]");
+
+   return GL_TRUE;
+}
+
+
+static GLboolean
+Parse_OutputReg(struct parse_state *parseState, GLint *outputRegNum)
+{
+   GLubyte token[100];
+
+   /* Match "o[" */
+   if (!Parse_String(parseState, "o["))
+      RETURN_ERROR1("Expected o[");
+
+   /* Get output reg name */
+   if (!Parse_Token(parseState, token))
+      RETURN_ERROR;
+
+   /* try to match an output register name */
+   if (strcmp((char *) token, "COLR") == 0 ||
+       strcmp((char *) token, "COLH") == 0) {
+      /* note that we don't distinguish between COLR and COLH */
+      *outputRegNum = FRAG_RESULT_COLOR;
+      parseState->outputsWritten |= (1 << FRAG_RESULT_COLOR);
+   }
+   else if (strcmp((char *) token, "DEPR") == 0) {
+      *outputRegNum = FRAG_RESULT_DEPTH;
+      parseState->outputsWritten |= (1 << FRAG_RESULT_DEPTH);
+   }
+   else {
+      RETURN_ERROR1("Invalid output register name");
+   }
+
+   /* Match ']' */
+   if (!Parse_String(parseState, "]"))
+      RETURN_ERROR1("Expected ]");
+
+   return GL_TRUE;
+}
+
+
+static GLboolean
+Parse_MaskedDstReg(struct parse_state *parseState,
+                   struct prog_dst_register *dstReg)
+{
+   GLubyte token[100];
+   GLint idx;
+
+   /* Dst reg can be R<n>, H<n>, o[n], RC or HC */
+   if (!Peek_Token(parseState, token))
+      RETURN_ERROR;
+
+   if (strcmp((const char *) token, "RC") == 0 ||
+       strcmp((const char *) token, "HC") == 0) {
+      /* a write-only register */
+      dstReg->File = PROGRAM_WRITE_ONLY;
+      if (!Parse_DummyReg(parseState, &idx))
+         RETURN_ERROR;
+      dstReg->Index = idx;
+   }
+   else if (token[0] == 'R' || token[0] == 'H') {
+      /* a temporary register */
+      dstReg->File = PROGRAM_TEMPORARY;
+      if (!Parse_TempReg(parseState, &idx))
+         RETURN_ERROR;
+      dstReg->Index = idx;
+   }
+   else if (token[0] == 'o') {
+      /* an output register */
+      dstReg->File = PROGRAM_OUTPUT;
+      if (!Parse_OutputReg(parseState, &idx))
+         RETURN_ERROR;
+      dstReg->Index = idx;
+   }
+   else {
+      RETURN_ERROR1("Invalid destination register name");
+   }
+
+   /* Parse optional write mask */
+   if (Parse_String(parseState, ".")) {
+      /* got a mask */
+      GLint k = 0;
+
+      if (!Parse_Token(parseState, token))  /* get xyzw writemask */
+         RETURN_ERROR;
+
+      dstReg->WriteMask = 0;
+
+      if (token[k] == 'x') {
+         dstReg->WriteMask |= WRITEMASK_X;
+         k++;
+      }
+      if (token[k] == 'y') {
+         dstReg->WriteMask |= WRITEMASK_Y;
+         k++;
+      }
+      if (token[k] == 'z') {
+         dstReg->WriteMask |= WRITEMASK_Z;
+         k++;
+      }
+      if (token[k] == 'w') {
+         dstReg->WriteMask |= WRITEMASK_W;
+         k++;
+      }
+      if (k == 0) {
+         RETURN_ERROR1("Invalid writemask character");
+      }
+
+   }
+   else {
+      dstReg->WriteMask = WRITEMASK_XYZW;
+   }
+
+   /* optional condition code mask */
+   if (Parse_String(parseState, "(")) {
+      /* ("EQ" | "GE" | "GT" | "LE" | "LT" | "NE" | "TR" | "FL".x|y|z|w) */
+      /* ("EQ" | "GE" | "GT" | "LE" | "LT" | "NE" | "TR" | "FL".[xyzw]) */
+      if (!Parse_CondCodeMask(parseState, dstReg))
+         RETURN_ERROR;
+
+      if (!Parse_String(parseState, ")"))  /* consume ")" */
+         RETURN_ERROR1("Expected )");
+
+      return GL_TRUE;
+   }
+   else {
+      /* no cond code mask */
+      dstReg->CondMask = COND_TR;
+      dstReg->CondSwizzle = SWIZZLE_NOOP;
+      return GL_TRUE;
+   }
+}
+
+
+/**
+ * Parse a vector source (register, constant, etc):
+ *   <vectorSrc>    ::= <absVectorSrc>
+ *                    | <baseVectorSrc>
+ *   <absVectorSrc> ::= <negate> "|" <baseVectorSrc> "|"
+ */
+static GLboolean
+Parse_VectorSrc(struct parse_state *parseState,
+                struct prog_src_register *srcReg)
+{
+   GLfloat sign = 1.0F;
+   GLubyte token[100];
+   GLint idx;
+   GLuint negateBase, negateAbs;
+
+   /*
+    * First, take care of +/- and absolute value stuff.
+    */
+   if (Parse_String(parseState, "-"))
+      sign = -1.0F;
+   else if (Parse_String(parseState, "+"))
+      sign = +1.0F;
+
+   if (Parse_String(parseState, "|")) {
+      srcReg->Abs = GL_TRUE;
+      negateAbs = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE;
+
+      if (Parse_String(parseState, "-"))
+         negateBase = NEGATE_XYZW;
+      else if (Parse_String(parseState, "+"))
+         negateBase = NEGATE_NONE;
+      else
+         negateBase = NEGATE_NONE;
+   }
+   else {
+      srcReg->Abs = GL_FALSE;
+      negateAbs = NEGATE_NONE;
+      negateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE;
+   }
+
+   srcReg->Negate = srcReg->Abs ? negateAbs : negateBase;
+
+   /* This should be the real src vector/register name */
+   if (!Peek_Token(parseState, token))
+      RETURN_ERROR;
+
+   /* Src reg can be Rn, Hn, f[n], p[n], a named parameter, a scalar
+    * literal or vector literal.
+    */
+   if (token[0] == 'R' || token[0] == 'H') {
+      srcReg->File = PROGRAM_TEMPORARY;
+      if (!Parse_TempReg(parseState, &idx))
+         RETURN_ERROR;
+      srcReg->Index = idx;
+   }
+   else if (token[0] == 'f') {
+      /* XXX this might be an identifier! */
+      srcReg->File = PROGRAM_INPUT;
+      if (!Parse_FragReg(parseState, &idx))
+         RETURN_ERROR;
+      srcReg->Index = idx;
+   }
+   else if (token[0] == 'p') {
+      /* XXX this might be an identifier! */
+      srcReg->File = PROGRAM_LOCAL_PARAM;
+      if (!Parse_ProgramParamReg(parseState, &idx))
+         RETURN_ERROR;
+      srcReg->Index = idx;
+   }
+   else if (IsLetter(token[0])){
+      GLubyte ident[100];
+      GLint paramIndex;
+      if (!Parse_Identifier(parseState, ident))
+         RETURN_ERROR;
+      paramIndex = _mesa_lookup_parameter_index(parseState->parameters,
+                                                -1, (const char *) ident);
+      if (paramIndex < 0) {
+         RETURN_ERROR2("Undefined constant or parameter: ", ident);
+      }
+      srcReg->File = PROGRAM_NAMED_PARAM;
+      srcReg->Index = paramIndex;      
+   }
+   else if (IsDigit(token[0]) || token[0] == '-' || token[0] == '+' || token[0] == '.'){
+      /* literal scalar constant */
+      GLfloat values[4];
+      GLuint paramIndex;
+      if (!Parse_ScalarConstant(parseState, values))
+         RETURN_ERROR;
+      paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
+                                              (gl_constant_value *) values,
+                                              4, NULL);
+      srcReg->File = PROGRAM_NAMED_PARAM;
+      srcReg->Index = paramIndex;
+   }
+   else if (token[0] == '{'){
+      /* literal vector constant */
+      GLfloat values[4];
+      GLuint paramIndex;
+      (void) Parse_String(parseState, "{");
+      if (!Parse_VectorConstant(parseState, values))
+         RETURN_ERROR;
+      paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
+                                              (gl_constant_value *) values,
+                                              4, NULL);
+      srcReg->File = PROGRAM_NAMED_PARAM;
+      srcReg->Index = paramIndex;      
+   }
+   else {
+      RETURN_ERROR2("Invalid source register name", token);
+   }
+
+   /* init swizzle fields */
+   srcReg->Swizzle = SWIZZLE_NOOP;
+
+   /* Look for optional swizzle suffix */
+   if (Parse_String(parseState, ".")) {
+      GLuint swz[4];
+
+      if (!Parse_Token(parseState, token))
+         RETURN_ERROR;
+
+      if (!Parse_SwizzleSuffix(token, swz))
+         RETURN_ERROR1("Invalid swizzle suffix");
+
+      srcReg->Swizzle = MAKE_SWIZZLE4(swz[0], swz[1], swz[2], swz[3]);
+   }
+
+   /* Finish absolute value */
+   if (srcReg->Abs && !Parse_String(parseState, "|")) {
+      RETURN_ERROR1("Expected |");
+   }
+
+   return GL_TRUE;
+}
+
+
+static GLboolean
+Parse_ScalarSrcReg(struct parse_state *parseState,
+                   struct prog_src_register *srcReg)
+{
+   GLubyte token[100];
+   GLfloat sign = 1.0F;
+   GLboolean needSuffix = GL_TRUE;
+   GLint idx;
+   GLuint negateBase, negateAbs;
+
+   /*
+    * First, take care of +/- and absolute value stuff.
+    */
+   if (Parse_String(parseState, "-"))
+      sign = -1.0F;
+   else if (Parse_String(parseState, "+"))
+      sign = +1.0F;
+
+   if (Parse_String(parseState, "|")) {
+      srcReg->Abs = GL_TRUE;
+      negateAbs = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE;
+
+      if (Parse_String(parseState, "-"))
+         negateBase = NEGATE_XYZW;
+      else if (Parse_String(parseState, "+"))
+         negateBase = NEGATE_NONE;
+      else
+         negateBase = NEGATE_NONE;
+   }
+   else {
+      srcReg->Abs = GL_FALSE;
+      negateAbs = NEGATE_NONE;
+      negateBase = (sign < 0.0F) ? NEGATE_XYZW : NEGATE_NONE;
+   }
+
+   srcReg->Negate = srcReg->Abs ? negateAbs : negateBase;
+
+   if (!Peek_Token(parseState, token))
+      RETURN_ERROR;
+
+   /* Src reg can be R<n>, H<n> or a named fragment attrib */
+   if (token[0] == 'R' || token[0] == 'H') {
+      srcReg->File = PROGRAM_TEMPORARY;
+      if (!Parse_TempReg(parseState, &idx))
+         RETURN_ERROR;
+      srcReg->Index = idx;
+   }
+   else if (token[0] == 'f') {
+      srcReg->File = PROGRAM_INPUT;
+      if (!Parse_FragReg(parseState, &idx))
+         RETURN_ERROR;
+      srcReg->Index = idx;
+   }
+   else if (token[0] == '{') {
+      /* vector literal */
+      GLfloat values[4];
+      GLuint paramIndex;
+      (void) Parse_String(parseState, "{");
+      if (!Parse_VectorConstant(parseState, values))
+         RETURN_ERROR;
+      paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
+                                              (gl_constant_value *) values,
+                                              4, NULL);
+      srcReg->File = PROGRAM_NAMED_PARAM;
+      srcReg->Index = paramIndex;      
+   }
+   else if (IsLetter(token[0])){
+      /* named param/constant */
+      GLubyte ident[100];
+      GLint paramIndex;
+      if (!Parse_Identifier(parseState, ident))
+         RETURN_ERROR;
+      paramIndex = _mesa_lookup_parameter_index(parseState->parameters,
+                                                -1, (const char *) ident);
+      if (paramIndex < 0) {
+         RETURN_ERROR2("Undefined constant or parameter: ", ident);
+      }
+      srcReg->File = PROGRAM_NAMED_PARAM;
+      srcReg->Index = paramIndex;      
+   }
+   else if (IsDigit(token[0])) {
+      /* scalar literal */
+      GLfloat values[4];
+      GLuint paramIndex;
+      if (!Parse_ScalarConstant(parseState, values))
+         RETURN_ERROR;
+      paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
+                                              (gl_constant_value *) values,
+                                              4, NULL);
+      srcReg->Index = paramIndex;      
+      srcReg->File = PROGRAM_NAMED_PARAM;
+      needSuffix = GL_FALSE;
+   }
+   else {
+      RETURN_ERROR2("Invalid scalar source argument", token);
+   }
+
+   srcReg->Swizzle = 0;
+   if (needSuffix) {
+      /* parse .[xyzw] suffix */
+      if (!Parse_String(parseState, "."))
+         RETURN_ERROR1("Expected .");
+
+      if (!Parse_Token(parseState, token))
+         RETURN_ERROR;
+
+      if (token[0] == 'x' && token[1] == 0) {
+         srcReg->Swizzle = 0;
+      }
+      else if (token[0] == 'y' && token[1] == 0) {
+         srcReg->Swizzle = 1;
+      }
+      else if (token[0] == 'z' && token[1] == 0) {
+         srcReg->Swizzle = 2;
+      }
+      else if (token[0] == 'w' && token[1] == 0) {
+         srcReg->Swizzle = 3;
+      }
+      else {
+         RETURN_ERROR1("Invalid scalar source suffix");
+      }
+   }
+
+   /* Finish absolute value */
+   if (srcReg->Abs && !Parse_String(parseState, "|")) {
+      RETURN_ERROR1("Expected |");
+   }
+
+   return GL_TRUE;
+}
+
+
+static GLboolean
+Parse_PrintInstruction(struct parse_state *parseState,
+                       struct prog_instruction *inst)
+{
+   const GLubyte *str;
+   GLubyte *msg;
+   GLuint len;
+   GLint idx;
+
+   /* The first argument is a literal string 'just like this' */
+   if (!Parse_String(parseState, "'"))
+      RETURN_ERROR1("Expected '");
+
+   str = parseState->pos;
+   for (len = 0; str[len] != '\''; len++) /* find closing quote */
+      ;
+   parseState->pos += len + 1;
+   msg = (GLubyte*) malloc(len + 1);
+
+   memcpy(msg, str, len);
+   msg[len] = 0;
+   inst->Data = msg;
+
+   if (Parse_String(parseState, ",")) {
+      /* got an optional register to print */
+      GLubyte token[100];
+      GetToken(parseState, token);
+      if (token[0] == 'o') {
+         /* dst reg */
+         if (!Parse_OutputReg(parseState, &idx))
+            RETURN_ERROR;
+	 inst->SrcReg[0].Index = idx;
+         inst->SrcReg[0].File = PROGRAM_OUTPUT;
+      }
+      else {
+         /* src reg */
+         if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
+            RETURN_ERROR;
+      }
+   }
+   else {
+      inst->SrcReg[0].File = PROGRAM_UNDEFINED;
+   }
+
+   inst->SrcReg[0].Swizzle = SWIZZLE_NOOP;
+   inst->SrcReg[0].Abs = GL_FALSE;
+   inst->SrcReg[0].Negate = NEGATE_NONE;
+
+   return GL_TRUE;
+}
+
+
+static GLboolean
+Parse_InstructionSequence(struct parse_state *parseState,
+                          struct prog_instruction program[])
+{
+   while (1) {
+      struct prog_instruction *inst = program + parseState->numInst;
+      struct instruction_pattern instMatch;
+      GLubyte token[100];
+
+      /* Initialize the instruction */
+      _mesa_init_instructions(inst, 1);
+
+      /* special instructions */
+      if (Parse_String(parseState, "DEFINE")) {
+         GLubyte id[100];
+         GLfloat value[7];  /* yes, 7 to be safe */
+         if (!Parse_Identifier(parseState, id))
+            RETURN_ERROR;
+         /* XXX make sure id is not a reserved identifer, like R9 */
+         if (!Parse_String(parseState, "="))
+            RETURN_ERROR1("Expected =");
+         if (!Parse_VectorOrScalarConstant(parseState, value))
+            RETURN_ERROR;
+         if (!Parse_String(parseState, ";"))
+            RETURN_ERROR1("Expected ;");
+         if (_mesa_lookup_parameter_index(parseState->parameters,
+                                          -1, (const char *) id) >= 0) {
+            RETURN_ERROR2(id, "already defined");
+         }
+         _mesa_add_named_parameter(parseState->parameters,
+                                   (const char *) id,
+                                   (gl_constant_value *) value);
+      }
+      else if (Parse_String(parseState, "DECLARE")) {
+         GLubyte id[100];
+         GLfloat value[7] = {0, 0, 0, 0, 0, 0, 0};  /* yes, to be safe */
+         if (!Parse_Identifier(parseState, id))
+            RETURN_ERROR;
+         /* XXX make sure id is not a reserved identifer, like R9 */
+         if (Parse_String(parseState, "=")) {
+            if (!Parse_VectorOrScalarConstant(parseState, value))
+               RETURN_ERROR;
+         }
+         if (!Parse_String(parseState, ";"))
+            RETURN_ERROR1("Expected ;");
+         if (_mesa_lookup_parameter_index(parseState->parameters,
+                                          -1, (const char *) id) >= 0) {
+            RETURN_ERROR2(id, "already declared");
+         }
+         _mesa_add_named_parameter(parseState->parameters,
+                                   (const char *) id,
+                                   (gl_constant_value *) value);
+      }
+      else if (Parse_String(parseState, "END")) {
+         inst->Opcode = OPCODE_END;
+         parseState->numInst++;
+         if (Parse_Token(parseState, token)) {
+            RETURN_ERROR1("Code after END opcode.");
+         }
+         break;
+      }
+      else {
+         /* general/arithmetic instruction */
+
+         /* get token */
+         if (!Parse_Token(parseState, token)) {
+            RETURN_ERROR1("Missing END instruction.");
+         }
+
+         /* try to find matching instuction */
+         instMatch = MatchInstruction(token);
+         if (instMatch.opcode >= MAX_OPCODE) {
+            /* bad instruction name */
+            RETURN_ERROR2("Unexpected token: ", token);
+         }
+
+         inst->Opcode = instMatch.opcode;
+         inst->Precision = instMatch.suffixes & (_R | _H | _X);
+         inst->SaturateMode = (instMatch.suffixes & (_S))
+            ? SATURATE_ZERO_ONE : SATURATE_OFF;
+         inst->CondUpdate = (instMatch.suffixes & (_C)) ? GL_TRUE : GL_FALSE;
+
+         /*
+          * parse the input and output operands
+          */
+         if (instMatch.outputs == OUTPUT_S || instMatch.outputs == OUTPUT_V) {
+            if (!Parse_MaskedDstReg(parseState, &inst->DstReg))
+               RETURN_ERROR;
+            if (!Parse_String(parseState, ","))
+               RETURN_ERROR1("Expected ,");
+         }
+         else if (instMatch.outputs == OUTPUT_NONE) {
+            if (instMatch.opcode == OPCODE_KIL_NV) {
+               /* This is a little weird, the cond code info is in
+                * the dest register.
+                */
+               if (!Parse_CondCodeMask(parseState, &inst->DstReg))
+                  RETURN_ERROR;
+            }
+            else {
+               ASSERT(instMatch.opcode == OPCODE_PRINT);
+            }
+         }
+
+         if (instMatch.inputs == INPUT_1V) {
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
+               RETURN_ERROR;
+         }
+         else if (instMatch.inputs == INPUT_2V) {
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
+               RETURN_ERROR;
+            if (!Parse_String(parseState, ","))
+               RETURN_ERROR1("Expected ,");
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[1]))
+               RETURN_ERROR;
+         }
+         else if (instMatch.inputs == INPUT_3V) {
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
+               RETURN_ERROR;
+            if (!Parse_String(parseState, ","))
+               RETURN_ERROR1("Expected ,");
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[1]))
+               RETURN_ERROR;
+            if (!Parse_String(parseState, ","))
+               RETURN_ERROR1("Expected ,");
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[2]))
+               RETURN_ERROR;
+         }
+         else if (instMatch.inputs == INPUT_1S) {
+            if (!Parse_ScalarSrcReg(parseState, &inst->SrcReg[0]))
+               RETURN_ERROR;
+         }
+         else if (instMatch.inputs == INPUT_2S) {
+            if (!Parse_ScalarSrcReg(parseState, &inst->SrcReg[0]))
+               RETURN_ERROR;
+            if (!Parse_String(parseState, ","))
+               RETURN_ERROR1("Expected ,");
+            if (!Parse_ScalarSrcReg(parseState, &inst->SrcReg[1]))
+               RETURN_ERROR;
+         }
+         else if (instMatch.inputs == INPUT_CC) {
+            /* XXX to-do */
+         }
+         else if (instMatch.inputs == INPUT_1V_T) {
+	    GLubyte unit, idx;
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
+               RETURN_ERROR;
+            if (!Parse_String(parseState, ","))
+               RETURN_ERROR1("Expected ,");
+            if (!Parse_TextureImageId(parseState, &unit, &idx))
+               RETURN_ERROR;
+	    inst->TexSrcUnit = unit;
+	    inst->TexSrcTarget = idx;
+         }
+         else if (instMatch.inputs == INPUT_3V_T) {
+	    GLubyte unit, idx;
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[0]))
+               RETURN_ERROR;
+            if (!Parse_String(parseState, ","))
+               RETURN_ERROR1("Expected ,");
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[1]))
+               RETURN_ERROR;
+            if (!Parse_String(parseState, ","))
+               RETURN_ERROR1("Expected ,");
+            if (!Parse_VectorSrc(parseState, &inst->SrcReg[2]))
+               RETURN_ERROR;
+            if (!Parse_String(parseState, ","))
+               RETURN_ERROR1("Expected ,");
+            if (!Parse_TextureImageId(parseState, &unit, &idx))
+               RETURN_ERROR;
+	    inst->TexSrcUnit = unit;
+	    inst->TexSrcTarget = idx;
+         }
+         else if (instMatch.inputs == INPUT_1V_S) {
+            if (!Parse_PrintInstruction(parseState, inst))
+               RETURN_ERROR;
+         }
+
+         /* end of statement semicolon */
+         if (!Parse_String(parseState, ";"))
+            RETURN_ERROR1("Expected ;");
+
+         parseState->numInst++;
+
+         if (parseState->numInst >= MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS)
+            RETURN_ERROR1("Program too long");
+      }
+   }
+   return GL_TRUE;
+}
+
+
+
+/**
+ * Parse/compile the 'str' returning the compiled 'program'.
+ * ctx->Program.ErrorPos will be -1 if successful.  Otherwise, ErrorPos
+ * indicates the position of the error in 'str'.
+ */
+void
+_mesa_parse_nv_fragment_program(struct gl_context *ctx, GLenum dstTarget,
+                                const GLubyte *str, GLsizei len,
+                                struct gl_fragment_program *program)
+{
+   struct parse_state parseState;
+   struct prog_instruction instBuffer[MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS];
+   struct prog_instruction *newInst;
+   GLenum target;
+   GLubyte *programString;
+
+   /* Make a null-terminated copy of the program string */
+   programString = (GLubyte *) MALLOC(len + 1);
+   if (!programString) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glLoadProgramNV");
+      return;
+   }
+   memcpy(programString, str, len);
+   programString[len] = 0;
+
+   /* Get ready to parse */
+   memset(&parseState, 0, sizeof(struct parse_state));
+   parseState.ctx = ctx;
+   parseState.start = programString;
+   parseState.program = program;
+   parseState.numInst = 0;
+   parseState.curLine = programString;
+   parseState.parameters = _mesa_new_parameter_list();
+
+   /* Reset error state */
+   _mesa_set_program_error(ctx, -1, NULL);
+
+   /* check the program header */
+   if (strncmp((const char *) programString, "!!FP1.0", 7) == 0) {
+      target = GL_FRAGMENT_PROGRAM_NV;
+      parseState.pos = programString + 7;
+   }
+   else if (strncmp((const char *) programString, "!!FCP1.0", 8) == 0) {
+      /* fragment / register combiner program - not supported */
+      _mesa_set_program_error(ctx, 0, "Invalid fragment program header");
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV(bad header)");
+      return;
+   }
+   else {
+      /* invalid header */
+      _mesa_set_program_error(ctx, 0, "Invalid fragment program header");
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV(bad header)");
+      return;
+   }
+
+   /* make sure target and header match */
+   if (target != dstTarget) {
+      _mesa_error(ctx, GL_INVALID_OPERATION,
+                  "glLoadProgramNV(target mismatch 0x%x != 0x%x)",
+                  target, dstTarget);
+      return;
+   }
+
+   if (Parse_InstructionSequence(&parseState, instBuffer)) {
+      GLuint u;
+      /* successful parse! */
+
+      if (parseState.outputsWritten == 0) {
+         /* must write at least one output! */
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "Invalid fragment program - no outputs written.");
+         return;
+      }
+
+      /* copy the compiled instructions */
+      assert(parseState.numInst <= MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS);
+      newInst = _mesa_alloc_instructions(parseState.numInst);
+      if (!newInst) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glLoadProgramNV");
+         return;  /* out of memory */
+      }
+      _mesa_copy_instructions(newInst, instBuffer, parseState.numInst);
+
+      /* install the program */
+      program->Base.Target = target;
+      if (program->Base.String) {
+         FREE(program->Base.String);
+      }
+      program->Base.String = programString;
+      program->Base.Format = GL_PROGRAM_FORMAT_ASCII_ARB;
+      if (program->Base.Instructions) {
+         free(program->Base.Instructions);
+      }
+      program->Base.Instructions = newInst;
+      program->Base.NumInstructions = parseState.numInst;
+      program->Base.InputsRead = parseState.inputsRead;
+      program->Base.OutputsWritten = parseState.outputsWritten;
+      for (u = 0; u < ctx->Const.MaxTextureImageUnits; u++)
+         program->Base.TexturesUsed[u] = parseState.texturesUsed[u];
+
+      /* save program parameters */
+      program->Base.Parameters = parseState.parameters;
+
+      /* allocate registers for declared program parameters */
+#if 00
+      _mesa_assign_program_registers(&(program->SymbolTable));
+#endif
+
+#ifdef DEBUG_foo
+      printf("--- glLoadProgramNV(%d) result ---\n", program->Base.Id);
+      _mesa_fprint_program_opt(stdout, &program->Base, PROG_PRINT_NV, 0);
+      printf("----------------------------------\n");
+#endif
+   }
+   else {
+      /* Error! */
+      _mesa_error(ctx, GL_INVALID_OPERATION, "glLoadProgramNV");
+      /* NOTE: _mesa_set_program_error would have been called already */
+   }
+}
+
+
+const char *
+_mesa_nv_fragment_input_register_name(GLuint i)
+{
+   ASSERT(i < MAX_NV_FRAGMENT_PROGRAM_INPUTS);
+   return InputRegisters[i];
+}
+
diff --git a/mesalib/src/mesa/program/prog_execute.c b/mesalib/src/mesa/program/prog_execute.c
index 28a3e1e20..dbfd1b918 100644
--- a/mesalib/src/mesa/program/prog_execute.c
+++ b/mesalib/src/mesa/program/prog_execute.c
@@ -1,1895 +1,1895 @@
-/*
- * Mesa 3-D graphics library
- * Version:  7.3
- *
- * 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.
- */
-
-/**
- * \file prog_execute.c
- * Software interpreter for vertex/fragment programs.
- * \author Brian Paul
- */
-
-/*
- * NOTE: we do everything in single-precision floating point; we don't
- * currently observe the single/half/fixed-precision qualifiers.
- *
- */
-
-
-#include "main/glheader.h"
-#include "main/colormac.h"
-#include "main/macros.h"
-#include "prog_execute.h"
-#include "prog_instruction.h"
-#include "prog_parameter.h"
-#include "prog_print.h"
-#include "prog_noise.h"
-
-
-/* debug predicate */
-#define DEBUG_PROG 0
-
-
-/**
- * Set x to positive or negative infinity.
- */
-#if defined(USE_IEEE) || defined(_WIN32)
-#define SET_POS_INFINITY(x)                  \
-   do {                                      \
-         fi_type fi;                         \
-         fi.i = 0x7F800000;                  \
-         x = fi.f;                           \
-   } while (0)
-#define SET_NEG_INFINITY(x)                  \
-   do {                                      \
-         fi_type fi;                         \
-         fi.i = 0xFF800000;                  \
-         x = fi.f;                           \
-   } while (0)
-#elif defined(VMS)
-#define SET_POS_INFINITY(x)  x = __MAXFLOAT
-#define SET_NEG_INFINITY(x)  x = -__MAXFLOAT
-#else
-#define SET_POS_INFINITY(x)  x = (GLfloat) HUGE_VAL
-#define SET_NEG_INFINITY(x)  x = (GLfloat) -HUGE_VAL
-#endif
-
-#define SET_FLOAT_BITS(x, bits) ((fi_type *) (void *) &(x))->i = bits
-
-
-static const GLfloat ZeroVec[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
-
-
-
-/**
- * Return TRUE for +0 and other positive values, FALSE otherwise.
- * Used for RCC opcode.
- */
-static INLINE GLboolean
-positive(float x)
-{
-   fi_type fi;
-   fi.f = x;
-   if (fi.i & 0x80000000)
-      return GL_FALSE;
-   return GL_TRUE;
-}
-
-
-
-/**
- * Return a pointer to the 4-element float vector specified by the given
- * source register.
- */
-static INLINE const GLfloat *
-get_src_register_pointer(const struct prog_src_register *source,
-                         const struct gl_program_machine *machine)
-{
-   const struct gl_program *prog = machine->CurProgram;
-   GLint reg = source->Index;
-
-   if (source->RelAddr) {
-      /* add address register value to src index/offset */
-      reg += machine->AddressReg[0][0];
-      if (reg < 0) {
-         return ZeroVec;
-      }
-   }
-
-   switch (source->File) {
-   case PROGRAM_TEMPORARY:
-      if (reg >= MAX_PROGRAM_TEMPS)
-         return ZeroVec;
-      return machine->Temporaries[reg];
-
-   case PROGRAM_INPUT:
-      if (prog->Target == GL_VERTEX_PROGRAM_ARB) {
-         if (reg >= VERT_ATTRIB_MAX)
-            return ZeroVec;
-         return machine->VertAttribs[reg];
-      }
-      else {
-         if (reg >= FRAG_ATTRIB_MAX)
-            return ZeroVec;
-         return machine->Attribs[reg][machine->CurElement];
-      }
-
-   case PROGRAM_OUTPUT:
-      if (reg >= MAX_PROGRAM_OUTPUTS)
-         return ZeroVec;
-      return machine->Outputs[reg];
-
-   case PROGRAM_LOCAL_PARAM:
-      if (reg >= MAX_PROGRAM_LOCAL_PARAMS)
-         return ZeroVec;
-      return machine->CurProgram->LocalParams[reg];
-
-   case PROGRAM_ENV_PARAM:
-      if (reg >= MAX_PROGRAM_ENV_PARAMS)
-         return ZeroVec;
-      return machine->EnvParams[reg];
-
-   case PROGRAM_STATE_VAR:
-      /* Fallthrough */
-   case PROGRAM_CONSTANT:
-      /* Fallthrough */
-   case PROGRAM_UNIFORM:
-      /* Fallthrough */
-   case PROGRAM_NAMED_PARAM:
-      if (reg >= (GLint) prog->Parameters->NumParameters)
-         return ZeroVec;
-      return prog->Parameters->ParameterValues[reg];
-
-   case PROGRAM_SYSTEM_VALUE:
-      assert(reg < Elements(machine->SystemValues));
-      return machine->SystemValues[reg];
-
-   default:
-      _mesa_problem(NULL,
-         "Invalid src register file %d in get_src_register_pointer()",
-         source->File);
-      return NULL;
-   }
-}
-
-
-/**
- * Return a pointer to the 4-element float vector specified by the given
- * destination register.
- */
-static INLINE GLfloat *
-get_dst_register_pointer(const struct prog_dst_register *dest,
-                         struct gl_program_machine *machine)
-{
-   static GLfloat dummyReg[4];
-   GLint reg = dest->Index;
-
-   if (dest->RelAddr) {
-      /* add address register value to src index/offset */
-      reg += machine->AddressReg[0][0];
-      if (reg < 0) {
-         return dummyReg;
-      }
-   }
-
-   switch (dest->File) {
-   case PROGRAM_TEMPORARY:
-      if (reg >= MAX_PROGRAM_TEMPS)
-         return dummyReg;
-      return machine->Temporaries[reg];
-
-   case PROGRAM_OUTPUT:
-      if (reg >= MAX_PROGRAM_OUTPUTS)
-         return dummyReg;
-      return machine->Outputs[reg];
-
-   case PROGRAM_WRITE_ONLY:
-      return dummyReg;
-
-   default:
-      _mesa_problem(NULL,
-         "Invalid dest register file %d in get_dst_register_pointer()",
-         dest->File);
-      return NULL;
-   }
-}
-
-
-
-/**
- * Fetch a 4-element float vector from the given source register.
- * Apply swizzling and negating as needed.
- */
-static void
-fetch_vector4(const struct prog_src_register *source,
-              const struct gl_program_machine *machine, GLfloat result[4])
-{
-   const GLfloat *src = get_src_register_pointer(source, machine);
-   ASSERT(src);
-
-   if (source->Swizzle == SWIZZLE_NOOP) {
-      /* no swizzling */
-      COPY_4V(result, src);
-   }
-   else {
-      ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
-      ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
-      ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
-      ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
-      result[0] = src[GET_SWZ(source->Swizzle, 0)];
-      result[1] = src[GET_SWZ(source->Swizzle, 1)];
-      result[2] = src[GET_SWZ(source->Swizzle, 2)];
-      result[3] = src[GET_SWZ(source->Swizzle, 3)];
-   }
-
-   if (source->Abs) {
-      result[0] = FABSF(result[0]);
-      result[1] = FABSF(result[1]);
-      result[2] = FABSF(result[2]);
-      result[3] = FABSF(result[3]);
-   }
-   if (source->Negate) {
-      ASSERT(source->Negate == NEGATE_XYZW);
-      result[0] = -result[0];
-      result[1] = -result[1];
-      result[2] = -result[2];
-      result[3] = -result[3];
-   }
-
-#ifdef NAN_CHECK
-   assert(!IS_INF_OR_NAN(result[0]));
-   assert(!IS_INF_OR_NAN(result[0]));
-   assert(!IS_INF_OR_NAN(result[0]));
-   assert(!IS_INF_OR_NAN(result[0]));
-#endif
-}
-
-
-/**
- * Fetch a 4-element uint vector from the given source register.
- * Apply swizzling but not negation/abs.
- */
-static void
-fetch_vector4ui(const struct prog_src_register *source,
-                const struct gl_program_machine *machine, GLuint result[4])
-{
-   const GLuint *src = (GLuint *) get_src_register_pointer(source, machine);
-   ASSERT(src);
-
-   if (source->Swizzle == SWIZZLE_NOOP) {
-      /* no swizzling */
-      COPY_4V(result, src);
-   }
-   else {
-      ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
-      ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
-      ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
-      ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
-      result[0] = src[GET_SWZ(source->Swizzle, 0)];
-      result[1] = src[GET_SWZ(source->Swizzle, 1)];
-      result[2] = src[GET_SWZ(source->Swizzle, 2)];
-      result[3] = src[GET_SWZ(source->Swizzle, 3)];
-   }
-
-   /* Note: no Negate or Abs here */
-}
-
-
-
-/**
- * Fetch the derivative with respect to X or Y for the given register.
- * XXX this currently only works for fragment program input attribs.
- */
-static void
-fetch_vector4_deriv(struct gl_context * ctx,
-                    const struct prog_src_register *source,
-                    const struct gl_program_machine *machine,
-                    char xOrY, GLfloat result[4])
-{
-   if (source->File == PROGRAM_INPUT &&
-       source->Index < (GLint) machine->NumDeriv) {
-      const GLint col = machine->CurElement;
-      const GLfloat w = machine->Attribs[FRAG_ATTRIB_WPOS][col][3];
-      const GLfloat invQ = 1.0f / w;
-      GLfloat deriv[4];
-
-      if (xOrY == 'X') {
-         deriv[0] = machine->DerivX[source->Index][0] * invQ;
-         deriv[1] = machine->DerivX[source->Index][1] * invQ;
-         deriv[2] = machine->DerivX[source->Index][2] * invQ;
-         deriv[3] = machine->DerivX[source->Index][3] * invQ;
-      }
-      else {
-         deriv[0] = machine->DerivY[source->Index][0] * invQ;
-         deriv[1] = machine->DerivY[source->Index][1] * invQ;
-         deriv[2] = machine->DerivY[source->Index][2] * invQ;
-         deriv[3] = machine->DerivY[source->Index][3] * invQ;
-      }
-
-      result[0] = deriv[GET_SWZ(source->Swizzle, 0)];
-      result[1] = deriv[GET_SWZ(source->Swizzle, 1)];
-      result[2] = deriv[GET_SWZ(source->Swizzle, 2)];
-      result[3] = deriv[GET_SWZ(source->Swizzle, 3)];
-      
-      if (source->Abs) {
-         result[0] = FABSF(result[0]);
-         result[1] = FABSF(result[1]);
-         result[2] = FABSF(result[2]);
-         result[3] = FABSF(result[3]);
-      }
-      if (source->Negate) {
-         ASSERT(source->Negate == NEGATE_XYZW);
-         result[0] = -result[0];
-         result[1] = -result[1];
-         result[2] = -result[2];
-         result[3] = -result[3];
-      }
-   }
-   else {
-      ASSIGN_4V(result, 0.0, 0.0, 0.0, 0.0);
-   }
-}
-
-
-/**
- * As above, but only return result[0] element.
- */
-static void
-fetch_vector1(const struct prog_src_register *source,
-              const struct gl_program_machine *machine, GLfloat result[4])
-{
-   const GLfloat *src = get_src_register_pointer(source, machine);
-   ASSERT(src);
-
-   result[0] = src[GET_SWZ(source->Swizzle, 0)];
-
-   if (source->Abs) {
-      result[0] = FABSF(result[0]);
-   }
-   if (source->Negate) {
-      result[0] = -result[0];
-   }
-}
-
-
-static GLuint
-fetch_vector1ui(const struct prog_src_register *source,
-                const struct gl_program_machine *machine)
-{
-   const GLuint *src = (GLuint *) get_src_register_pointer(source, machine);
-   return src[GET_SWZ(source->Swizzle, 0)];
-}
-
-
-/**
- * Fetch texel from texture.  Use partial derivatives when possible.
- */
-static INLINE void
-fetch_texel(struct gl_context *ctx,
-            const struct gl_program_machine *machine,
-            const struct prog_instruction *inst,
-            const GLfloat texcoord[4], GLfloat lodBias,
-            GLfloat color[4])
-{
-   const GLuint unit = machine->Samplers[inst->TexSrcUnit];
-
-   /* Note: we only have the right derivatives for fragment input attribs.
-    */
-   if (machine->NumDeriv > 0 &&
-       inst->SrcReg[0].File == PROGRAM_INPUT &&
-       inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) {
-      /* simple texture fetch for which we should have derivatives */
-      GLuint attr = inst->SrcReg[0].Index;
-      machine->FetchTexelDeriv(ctx, texcoord,
-                               machine->DerivX[attr],
-                               machine->DerivY[attr],
-                               lodBias, unit, color);
-   }
-   else {
-      machine->FetchTexelLod(ctx, texcoord, lodBias, unit, color);
-   }
-}
-
-
-/**
- * Test value against zero and return GT, LT, EQ or UN if NaN.
- */
-static INLINE GLuint
-generate_cc(float value)
-{
-   if (value != value)
-      return COND_UN;           /* NaN */
-   if (value > 0.0F)
-      return COND_GT;
-   if (value < 0.0F)
-      return COND_LT;
-   return COND_EQ;
-}
-
-
-/**
- * Test if the ccMaskRule is satisfied by the given condition code.
- * Used to mask destination writes according to the current condition code.
- */
-static INLINE GLboolean
-test_cc(GLuint condCode, GLuint ccMaskRule)
-{
-   switch (ccMaskRule) {
-   case COND_EQ: return (condCode == COND_EQ);
-   case COND_NE: return (condCode != COND_EQ);
-   case COND_LT: return (condCode == COND_LT);
-   case COND_GE: return (condCode == COND_GT || condCode == COND_EQ);
-   case COND_LE: return (condCode == COND_LT || condCode == COND_EQ);
-   case COND_GT: return (condCode == COND_GT);
-   case COND_TR: return GL_TRUE;
-   case COND_FL: return GL_FALSE;
-   default:      return GL_TRUE;
-   }
-}
-
-
-/**
- * Evaluate the 4 condition codes against a predicate and return GL_TRUE
- * or GL_FALSE to indicate result.
- */
-static INLINE GLboolean
-eval_condition(const struct gl_program_machine *machine,
-               const struct prog_instruction *inst)
-{
-   const GLuint swizzle = inst->DstReg.CondSwizzle;
-   const GLuint condMask = inst->DstReg.CondMask;
-   if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
-       test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
-       test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
-       test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
-      return GL_TRUE;
-   }
-   else {
-      return GL_FALSE;
-   }
-}
-
-
-
-/**
- * Store 4 floats into a register.  Observe the instructions saturate and
- * set-condition-code flags.
- */
-static void
-store_vector4(const struct prog_instruction *inst,
-              struct gl_program_machine *machine, const GLfloat value[4])
-{
-   const struct prog_dst_register *dstReg = &(inst->DstReg);
-   const GLboolean clamp = inst->SaturateMode == SATURATE_ZERO_ONE;
-   GLuint writeMask = dstReg->WriteMask;
-   GLfloat clampedValue[4];
-   GLfloat *dst = get_dst_register_pointer(dstReg, machine);
-
-#if 0
-   if (value[0] > 1.0e10 ||
-       IS_INF_OR_NAN(value[0]) ||
-       IS_INF_OR_NAN(value[1]) ||
-       IS_INF_OR_NAN(value[2]) || IS_INF_OR_NAN(value[3]))
-      printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
-#endif
-
-   if (clamp) {
-      clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
-      clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
-      clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
-      clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
-      value = clampedValue;
-   }
-
-   if (dstReg->CondMask != COND_TR) {
-      /* condition codes may turn off some writes */
-      if (writeMask & WRITEMASK_X) {
-         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
-                      dstReg->CondMask))
-            writeMask &= ~WRITEMASK_X;
-      }
-      if (writeMask & WRITEMASK_Y) {
-         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
-                      dstReg->CondMask))
-            writeMask &= ~WRITEMASK_Y;
-      }
-      if (writeMask & WRITEMASK_Z) {
-         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
-                      dstReg->CondMask))
-            writeMask &= ~WRITEMASK_Z;
-      }
-      if (writeMask & WRITEMASK_W) {
-         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
-                      dstReg->CondMask))
-            writeMask &= ~WRITEMASK_W;
-      }
-   }
-
-#ifdef NAN_CHECK
-   assert(!IS_INF_OR_NAN(value[0]));
-   assert(!IS_INF_OR_NAN(value[0]));
-   assert(!IS_INF_OR_NAN(value[0]));
-   assert(!IS_INF_OR_NAN(value[0]));
-#endif
-
-   if (writeMask & WRITEMASK_X)
-      dst[0] = value[0];
-   if (writeMask & WRITEMASK_Y)
-      dst[1] = value[1];
-   if (writeMask & WRITEMASK_Z)
-      dst[2] = value[2];
-   if (writeMask & WRITEMASK_W)
-      dst[3] = value[3];
-
-   if (inst->CondUpdate) {
-      if (writeMask & WRITEMASK_X)
-         machine->CondCodes[0] = generate_cc(value[0]);
-      if (writeMask & WRITEMASK_Y)
-         machine->CondCodes[1] = generate_cc(value[1]);
-      if (writeMask & WRITEMASK_Z)
-         machine->CondCodes[2] = generate_cc(value[2]);
-      if (writeMask & WRITEMASK_W)
-         machine->CondCodes[3] = generate_cc(value[3]);
-#if DEBUG_PROG
-      printf("CondCodes=(%s,%s,%s,%s) for:\n",
-             _mesa_condcode_string(machine->CondCodes[0]),
-             _mesa_condcode_string(machine->CondCodes[1]),
-             _mesa_condcode_string(machine->CondCodes[2]),
-             _mesa_condcode_string(machine->CondCodes[3]));
-#endif
-   }
-}
-
-
-/**
- * Store 4 uints into a register.  Observe the set-condition-code flags.
- */
-static void
-store_vector4ui(const struct prog_instruction *inst,
-                struct gl_program_machine *machine, const GLuint value[4])
-{
-   const struct prog_dst_register *dstReg = &(inst->DstReg);
-   GLuint writeMask = dstReg->WriteMask;
-   GLuint *dst = (GLuint *) get_dst_register_pointer(dstReg, machine);
-
-   if (dstReg->CondMask != COND_TR) {
-      /* condition codes may turn off some writes */
-      if (writeMask & WRITEMASK_X) {
-         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
-                      dstReg->CondMask))
-            writeMask &= ~WRITEMASK_X;
-      }
-      if (writeMask & WRITEMASK_Y) {
-         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
-                      dstReg->CondMask))
-            writeMask &= ~WRITEMASK_Y;
-      }
-      if (writeMask & WRITEMASK_Z) {
-         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
-                      dstReg->CondMask))
-            writeMask &= ~WRITEMASK_Z;
-      }
-      if (writeMask & WRITEMASK_W) {
-         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
-                      dstReg->CondMask))
-            writeMask &= ~WRITEMASK_W;
-      }
-   }
-
-   if (writeMask & WRITEMASK_X)
-      dst[0] = value[0];
-   if (writeMask & WRITEMASK_Y)
-      dst[1] = value[1];
-   if (writeMask & WRITEMASK_Z)
-      dst[2] = value[2];
-   if (writeMask & WRITEMASK_W)
-      dst[3] = value[3];
-
-   if (inst->CondUpdate) {
-      if (writeMask & WRITEMASK_X)
-         machine->CondCodes[0] = generate_cc((float)value[0]);
-      if (writeMask & WRITEMASK_Y)
-         machine->CondCodes[1] = generate_cc((float)value[1]);
-      if (writeMask & WRITEMASK_Z)
-         machine->CondCodes[2] = generate_cc((float)value[2]);
-      if (writeMask & WRITEMASK_W)
-         machine->CondCodes[3] = generate_cc((float)value[3]);
-#if DEBUG_PROG
-      printf("CondCodes=(%s,%s,%s,%s) for:\n",
-             _mesa_condcode_string(machine->CondCodes[0]),
-             _mesa_condcode_string(machine->CondCodes[1]),
-             _mesa_condcode_string(machine->CondCodes[2]),
-             _mesa_condcode_string(machine->CondCodes[3]));
-#endif
-   }
-}
-
-
-
-/**
- * Execute the given vertex/fragment program.
- *
- * \param ctx  rendering context
- * \param program  the program to execute
- * \param machine  machine state (must be initialized)
- * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
- */
-GLboolean
-_mesa_execute_program(struct gl_context * ctx,
-                      const struct gl_program *program,
-                      struct gl_program_machine *machine)
-{
-   const GLuint numInst = program->NumInstructions;
-   const GLuint maxExec = 10000;
-   GLuint pc, numExec = 0;
-
-   machine->CurProgram = program;
-
-   if (DEBUG_PROG) {
-      printf("execute program %u --------------------\n", program->Id);
-   }
-
-   if (program->Target == GL_VERTEX_PROGRAM_ARB) {
-      machine->EnvParams = ctx->VertexProgram.Parameters;
-   }
-   else {
-      machine->EnvParams = ctx->FragmentProgram.Parameters;
-   }
-
-   for (pc = 0; pc < numInst; pc++) {
-      const struct prog_instruction *inst = program->Instructions + pc;
-
-      if (DEBUG_PROG) {
-         _mesa_print_instruction(inst);
-      }
-
-      switch (inst->Opcode) {
-      case OPCODE_ABS:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            result[0] = FABSF(a[0]);
-            result[1] = FABSF(a[1]);
-            result[2] = FABSF(a[2]);
-            result[3] = FABSF(a[3]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_ADD:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = a[0] + b[0];
-            result[1] = a[1] + b[1];
-            result[2] = a[2] + b[2];
-            result[3] = a[3] + b[3];
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_AND:     /* bitwise AND */
-         {
-            GLuint a[4], b[4], result[4];
-            fetch_vector4ui(&inst->SrcReg[0], machine, a);
-            fetch_vector4ui(&inst->SrcReg[1], machine, b);
-            result[0] = a[0] & b[0];
-            result[1] = a[1] & b[1];
-            result[2] = a[2] & b[2];
-            result[3] = a[3] & b[3];
-            store_vector4ui(inst, machine, result);
-         }
-         break;
-      case OPCODE_ARL:
-         {
-            GLfloat t[4];
-            fetch_vector4(&inst->SrcReg[0], machine, t);
-            machine->AddressReg[0][0] = IFLOOR(t[0]);
-            if (DEBUG_PROG) {
-               printf("ARL %d\n", machine->AddressReg[0][0]);
-            }
-         }
-         break;
-      case OPCODE_BGNLOOP:
-         /* no-op */
-         ASSERT(program->Instructions[inst->BranchTarget].Opcode
-                == OPCODE_ENDLOOP);
-         break;
-      case OPCODE_ENDLOOP:
-         /* subtract 1 here since pc is incremented by for(pc) loop */
-         ASSERT(program->Instructions[inst->BranchTarget].Opcode
-                == OPCODE_BGNLOOP);
-         pc = inst->BranchTarget - 1;   /* go to matching BNGLOOP */
-         break;
-      case OPCODE_BGNSUB:      /* begin subroutine */
-         break;
-      case OPCODE_ENDSUB:      /* end subroutine */
-         break;
-      case OPCODE_BRA:         /* branch (conditional) */
-         if (eval_condition(machine, inst)) {
-            /* take branch */
-            /* Subtract 1 here since we'll do pc++ below */
-            pc = inst->BranchTarget - 1;
-         }
-         break;
-      case OPCODE_BRK:         /* break out of loop (conditional) */
-         ASSERT(program->Instructions[inst->BranchTarget].Opcode
-                == OPCODE_ENDLOOP);
-         if (eval_condition(machine, inst)) {
-            /* break out of loop */
-            /* pc++ at end of for-loop will put us after the ENDLOOP inst */
-            pc = inst->BranchTarget;
-         }
-         break;
-      case OPCODE_CONT:        /* continue loop (conditional) */
-         ASSERT(program->Instructions[inst->BranchTarget].Opcode
-                == OPCODE_ENDLOOP);
-         if (eval_condition(machine, inst)) {
-            /* continue at ENDLOOP */
-            /* Subtract 1 here since we'll do pc++ at end of for-loop */
-            pc = inst->BranchTarget - 1;
-         }
-         break;
-      case OPCODE_CAL:         /* Call subroutine (conditional) */
-         if (eval_condition(machine, inst)) {
-            /* call the subroutine */
-            if (machine->StackDepth >= MAX_PROGRAM_CALL_DEPTH) {
-               return GL_TRUE;  /* Per GL_NV_vertex_program2 spec */
-            }
-            machine->CallStack[machine->StackDepth++] = pc + 1; /* next inst */
-            /* Subtract 1 here since we'll do pc++ at end of for-loop */
-            pc = inst->BranchTarget - 1;
-         }
-         break;
-      case OPCODE_CMP:
-         {
-            GLfloat a[4], b[4], c[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            fetch_vector4(&inst->SrcReg[2], machine, c);
-            result[0] = a[0] < 0.0F ? b[0] : c[0];
-            result[1] = a[1] < 0.0F ? b[1] : c[1];
-            result[2] = a[2] < 0.0F ? b[2] : c[2];
-            result[3] = a[3] < 0.0F ? b[3] : c[3];
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("CMP (%g %g %g %g) = (%g %g %g %g) < 0 ? (%g %g %g %g) : (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3],
-                      c[0], c[1], c[2], c[3]);
-            }
-         }
-         break;
-      case OPCODE_COS:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-            result[0] = result[1] = result[2] = result[3]
-               = (GLfloat) cos(a[0]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_DDX:         /* Partial derivative with respect to X */
-         {
-            GLfloat result[4];
-            fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
-                                'X', result);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_DDY:         /* Partial derivative with respect to Y */
-         {
-            GLfloat result[4];
-            fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
-                                'Y', result);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_DP2:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = result[1] = result[2] = result[3] = DOT2(a, b);
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("DP2 %g = (%g %g) . (%g %g)\n",
-                      result[0], a[0], a[1], b[0], b[1]);
-            }
-         }
-         break;
-      case OPCODE_DP2A:
-         {
-            GLfloat a[4], b[4], c, result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            fetch_vector1(&inst->SrcReg[1], machine, &c);
-            result[0] = result[1] = result[2] = result[3] = DOT2(a, b) + c;
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("DP2A %g = (%g %g) . (%g %g) + %g\n",
-                      result[0], a[0], a[1], b[0], b[1], c);
-            }
-         }
-         break;
-      case OPCODE_DP3:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = result[1] = result[2] = result[3] = DOT3(a, b);
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
-                      result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
-            }
-         }
-         break;
-      case OPCODE_DP4:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = result[1] = result[2] = result[3] = DOT4(a, b);
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
-                      result[0], a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_DPH:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = result[1] = result[2] = result[3] = DOT3(a, b) + b[3];
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_DST:         /* Distance vector */
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = 1.0F;
-            result[1] = a[1] * b[1];
-            result[2] = a[2];
-            result[3] = b[3];
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_EXP:
-         {
-            GLfloat t[4], q[4], floor_t0;
-            fetch_vector1(&inst->SrcReg[0], machine, t);
-            floor_t0 = FLOORF(t[0]);
-            if (floor_t0 > FLT_MAX_EXP) {
-               SET_POS_INFINITY(q[0]);
-               SET_POS_INFINITY(q[2]);
-            }
-            else if (floor_t0 < FLT_MIN_EXP) {
-               q[0] = 0.0F;
-               q[2] = 0.0F;
-            }
-            else {
-               q[0] = LDEXPF(1.0, (int) floor_t0);
-               /* Note: GL_NV_vertex_program expects 
-                * result.z = result.x * APPX(result.y)
-                * We do what the ARB extension says.
-                */
-               q[2] = (GLfloat) pow(2.0, t[0]);
-            }
-            q[1] = t[0] - floor_t0;
-            q[3] = 1.0F;
-            store_vector4( inst, machine, q );
-         }
-         break;
-      case OPCODE_EX2:         /* Exponential base 2 */
-         {
-            GLfloat a[4], result[4], val;
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-            val = (GLfloat) pow(2.0, a[0]);
-            /*
-            if (IS_INF_OR_NAN(val))
-               val = 1.0e10;
-            */
-            result[0] = result[1] = result[2] = result[3] = val;
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_FLR:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            result[0] = FLOORF(a[0]);
-            result[1] = FLOORF(a[1]);
-            result[2] = FLOORF(a[2]);
-            result[3] = FLOORF(a[3]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_FRC:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            result[0] = a[0] - FLOORF(a[0]);
-            result[1] = a[1] - FLOORF(a[1]);
-            result[2] = a[2] - FLOORF(a[2]);
-            result[3] = a[3] - FLOORF(a[3]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_IF:
-         {
-            GLboolean cond;
-            ASSERT(program->Instructions[inst->BranchTarget].Opcode
-                   == OPCODE_ELSE ||
-                   program->Instructions[inst->BranchTarget].Opcode
-                   == OPCODE_ENDIF);
-            /* eval condition */
-            if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
-               GLfloat a[4];
-               fetch_vector1(&inst->SrcReg[0], machine, a);
-               cond = (a[0] != 0.0);
-            }
-            else {
-               cond = eval_condition(machine, inst);
-            }
-            if (DEBUG_PROG) {
-               printf("IF: %d\n", cond);
-            }
-            /* do if/else */
-            if (cond) {
-               /* do if-clause (just continue execution) */
-            }
-            else {
-               /* go to the instruction after ELSE or ENDIF */
-               assert(inst->BranchTarget >= 0);
-               pc = inst->BranchTarget;
-            }
-         }
-         break;
-      case OPCODE_ELSE:
-         /* goto ENDIF */
-         ASSERT(program->Instructions[inst->BranchTarget].Opcode
-                == OPCODE_ENDIF);
-         assert(inst->BranchTarget >= 0);
-         pc = inst->BranchTarget;
-         break;
-      case OPCODE_ENDIF:
-         /* nothing */
-         break;
-      case OPCODE_KIL_NV:      /* NV_f_p only (conditional) */
-         if (eval_condition(machine, inst)) {
-            return GL_FALSE;
-         }
-         break;
-      case OPCODE_KIL:         /* ARB_f_p only */
-         {
-            GLfloat a[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            if (DEBUG_PROG) {
-               printf("KIL if (%g %g %g %g) <= 0.0\n",
-                      a[0], a[1], a[2], a[3]);
-            }
-
-            if (a[0] < 0.0F || a[1] < 0.0F || a[2] < 0.0F || a[3] < 0.0F) {
-               return GL_FALSE;
-            }
-         }
-         break;
-      case OPCODE_LG2:         /* log base 2 */
-         {
-            GLfloat a[4], result[4], val;
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-	    /* The fast LOG2 macro doesn't meet the precision requirements.
-	     */
-            if (a[0] == 0.0F) {
-               val = -FLT_MAX;
-            }
-            else {
-               val = (float)(log(a[0]) * 1.442695F);
-            }
-            result[0] = result[1] = result[2] = result[3] = val;
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_LIT:
-         {
-            const GLfloat epsilon = 1.0F / 256.0F;      /* from NV VP spec */
-            GLfloat a[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            a[0] = MAX2(a[0], 0.0F);
-            a[1] = MAX2(a[1], 0.0F);
-            /* XXX ARB version clamps a[3], NV version doesn't */
-            a[3] = CLAMP(a[3], -(128.0F - epsilon), (128.0F - epsilon));
-            result[0] = 1.0F;
-            result[1] = a[0];
-            /* XXX we could probably just use pow() here */
-            if (a[0] > 0.0F) {
-               if (a[1] == 0.0 && a[3] == 0.0)
-                  result[2] = 1.0F;
-               else
-                  result[2] = (GLfloat) pow(a[1], a[3]);
-            }
-            else {
-               result[2] = 0.0F;
-            }
-            result[3] = 1.0F;
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3]);
-            }
-         }
-         break;
-      case OPCODE_LOG:
-         {
-            GLfloat t[4], q[4], abs_t0;
-            fetch_vector1(&inst->SrcReg[0], machine, t);
-            abs_t0 = FABSF(t[0]);
-            if (abs_t0 != 0.0F) {
-               /* Since we really can't handle infinite values on VMS
-                * like other OSes we'll use __MAXFLOAT to represent
-                * infinity.  This may need some tweaking.
-                */
-#ifdef VMS
-               if (abs_t0 == __MAXFLOAT)
-#else
-               if (IS_INF_OR_NAN(abs_t0))
-#endif
-               {
-                  SET_POS_INFINITY(q[0]);
-                  q[1] = 1.0F;
-                  SET_POS_INFINITY(q[2]);
-               }
-               else {
-                  int exponent;
-                  GLfloat mantissa = FREXPF(t[0], &exponent);
-                  q[0] = (GLfloat) (exponent - 1);
-                  q[1] = (GLfloat) (2.0 * mantissa); /* map [.5, 1) -> [1, 2) */
-
-		  /* The fast LOG2 macro doesn't meet the precision
-		   * requirements.
-		   */
-                  q[2] = (float)(log(t[0]) * 1.442695F);
-               }
-            }
-            else {
-               SET_NEG_INFINITY(q[0]);
-               q[1] = 1.0F;
-               SET_NEG_INFINITY(q[2]);
-            }
-            q[3] = 1.0;
-            store_vector4(inst, machine, q);
-         }
-         break;
-      case OPCODE_LRP:
-         {
-            GLfloat a[4], b[4], c[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            fetch_vector4(&inst->SrcReg[2], machine, c);
-            result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
-            result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
-            result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
-            result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("LRP (%g %g %g %g) = (%g %g %g %g), "
-                      "(%g %g %g %g), (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
-            }
-         }
-         break;
-      case OPCODE_MAD:
-         {
-            GLfloat a[4], b[4], c[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            fetch_vector4(&inst->SrcReg[2], machine, c);
-            result[0] = a[0] * b[0] + c[0];
-            result[1] = a[1] * b[1] + c[1];
-            result[2] = a[2] * b[2] + c[2];
-            result[3] = a[3] * b[3] + c[3];
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
-                      "(%g %g %g %g) + (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
-            }
-         }
-         break;
-      case OPCODE_MAX:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = MAX2(a[0], b[0]);
-            result[1] = MAX2(a[1], b[1]);
-            result[2] = MAX2(a[2], b[2]);
-            result[3] = MAX2(a[3], b[3]);
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_MIN:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = MIN2(a[0], b[0]);
-            result[1] = MIN2(a[1], b[1]);
-            result[2] = MIN2(a[2], b[2]);
-            result[3] = MIN2(a[3], b[3]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_MOV:
-         {
-            GLfloat result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, result);
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("MOV (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3]);
-            }
-         }
-         break;
-      case OPCODE_MUL:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = a[0] * b[0];
-            result[1] = a[1] * b[1];
-            result[2] = a[2] * b[2];
-            result[3] = a[3] * b[3];
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_NOISE1:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-            result[0] =
-               result[1] =
-               result[2] =
-               result[3] = _mesa_noise1(a[0]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_NOISE2:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            result[0] =
-               result[1] =
-               result[2] = result[3] = _mesa_noise2(a[0], a[1]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_NOISE3:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            result[0] =
-               result[1] =
-               result[2] =
-               result[3] = _mesa_noise3(a[0], a[1], a[2]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_NOISE4:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            result[0] =
-               result[1] =
-               result[2] =
-               result[3] = _mesa_noise4(a[0], a[1], a[2], a[3]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_NOP:
-         break;
-      case OPCODE_NOT:         /* bitwise NOT */
-         {
-            GLuint a[4], result[4];
-            fetch_vector4ui(&inst->SrcReg[0], machine, a);
-            result[0] = ~a[0];
-            result[1] = ~a[1];
-            result[2] = ~a[2];
-            result[3] = ~a[3];
-            store_vector4ui(inst, machine, result);
-         }
-         break;
-      case OPCODE_NRM3:        /* 3-component normalization */
-         {
-            GLfloat a[4], result[4];
-            GLfloat tmp;
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2];
-            if (tmp != 0.0F)
-               tmp = INV_SQRTF(tmp);
-            result[0] = tmp * a[0];
-            result[1] = tmp * a[1];
-            result[2] = tmp * a[2];
-            result[3] = 0.0;  /* undefined, but prevent valgrind warnings */
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_NRM4:        /* 4-component normalization */
-         {
-            GLfloat a[4], result[4];
-            GLfloat tmp;
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2] + a[3] * a[3];
-            if (tmp != 0.0F)
-               tmp = INV_SQRTF(tmp);
-            result[0] = tmp * a[0];
-            result[1] = tmp * a[1];
-            result[2] = tmp * a[2];
-            result[3] = tmp * a[3];
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_OR:          /* bitwise OR */
-         {
-            GLuint a[4], b[4], result[4];
-            fetch_vector4ui(&inst->SrcReg[0], machine, a);
-            fetch_vector4ui(&inst->SrcReg[1], machine, b);
-            result[0] = a[0] | b[0];
-            result[1] = a[1] | b[1];
-            result[2] = a[2] | b[2];
-            result[3] = a[3] | b[3];
-            store_vector4ui(inst, machine, result);
-         }
-         break;
-      case OPCODE_PK2H:        /* pack two 16-bit floats in one 32-bit float */
-         {
-            GLfloat a[4];
-            GLuint result[4];
-            GLhalfNV hx, hy;
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            hx = _mesa_float_to_half(a[0]);
-            hy = _mesa_float_to_half(a[1]);
-            result[0] =
-            result[1] =
-            result[2] =
-            result[3] = hx | (hy << 16);
-            store_vector4ui(inst, machine, result);
-         }
-         break;
-      case OPCODE_PK2US:       /* pack two GLushorts into one 32-bit float */
-         {
-            GLfloat a[4];
-            GLuint result[4], usx, usy;
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            a[0] = CLAMP(a[0], 0.0F, 1.0F);
-            a[1] = CLAMP(a[1], 0.0F, 1.0F);
-            usx = IROUND(a[0] * 65535.0F);
-            usy = IROUND(a[1] * 65535.0F);
-            result[0] =
-            result[1] =
-            result[2] =
-            result[3] = usx | (usy << 16);
-            store_vector4ui(inst, machine, result);
-         }
-         break;
-      case OPCODE_PK4B:        /* pack four GLbytes into one 32-bit float */
-         {
-            GLfloat a[4];
-            GLuint result[4], ubx, uby, ubz, ubw;
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
-            a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
-            a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
-            a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
-            ubx = IROUND(127.0F * a[0] + 128.0F);
-            uby = IROUND(127.0F * a[1] + 128.0F);
-            ubz = IROUND(127.0F * a[2] + 128.0F);
-            ubw = IROUND(127.0F * a[3] + 128.0F);
-            result[0] =
-            result[1] =
-            result[2] =
-            result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
-            store_vector4ui(inst, machine, result);
-         }
-         break;
-      case OPCODE_PK4UB:       /* pack four GLubytes into one 32-bit float */
-         {
-            GLfloat a[4];
-            GLuint result[4], ubx, uby, ubz, ubw;
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            a[0] = CLAMP(a[0], 0.0F, 1.0F);
-            a[1] = CLAMP(a[1], 0.0F, 1.0F);
-            a[2] = CLAMP(a[2], 0.0F, 1.0F);
-            a[3] = CLAMP(a[3], 0.0F, 1.0F);
-            ubx = IROUND(255.0F * a[0]);
-            uby = IROUND(255.0F * a[1]);
-            ubz = IROUND(255.0F * a[2]);
-            ubw = IROUND(255.0F * a[3]);
-            result[0] =
-            result[1] =
-            result[2] =
-            result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
-            store_vector4ui(inst, machine, result);
-         }
-         break;
-      case OPCODE_POW:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-            fetch_vector1(&inst->SrcReg[1], machine, b);
-            result[0] = result[1] = result[2] = result[3]
-               = (GLfloat) pow(a[0], b[0]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_RCC:  /* clamped riciprocal */
-         {
-            const float largest = 1.884467e+19, smallest = 5.42101e-20;
-            GLfloat a[4], r, result[4];
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-            if (DEBUG_PROG) {
-               if (a[0] == 0)
-                  printf("RCC(0)\n");
-               else if (IS_INF_OR_NAN(a[0]))
-                  printf("RCC(inf)\n");
-            }
-            if (a[0] == 1.0F) {
-               r = 1.0F;
-            }
-            else {
-               r = 1.0F / a[0];
-            }
-            if (positive(r)) {
-               if (r > largest) {
-                  r = largest;
-               }
-               else if (r < smallest) {
-                  r = smallest;
-               }
-            }
-            else {
-               if (r < -largest) {
-                  r = -largest;
-               }
-               else if (r > -smallest) {
-                  r = -smallest;
-               }
-            }
-            result[0] = result[1] = result[2] = result[3] = r;
-            store_vector4(inst, machine, result);
-         }
-         break;
-
-      case OPCODE_RCP:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-            if (DEBUG_PROG) {
-               if (a[0] == 0)
-                  printf("RCP(0)\n");
-               else if (IS_INF_OR_NAN(a[0]))
-                  printf("RCP(inf)\n");
-            }
-            result[0] = result[1] = result[2] = result[3] = 1.0F / a[0];
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_RET:         /* return from subroutine (conditional) */
-         if (eval_condition(machine, inst)) {
-            if (machine->StackDepth == 0) {
-               return GL_TRUE;  /* Per GL_NV_vertex_program2 spec */
-            }
-            /* subtract one because of pc++ in the for loop */
-            pc = machine->CallStack[--machine->StackDepth] - 1;
-         }
-         break;
-      case OPCODE_RFL:         /* reflection vector */
-         {
-            GLfloat axis[4], dir[4], result[4], tmpX, tmpW;
-            fetch_vector4(&inst->SrcReg[0], machine, axis);
-            fetch_vector4(&inst->SrcReg[1], machine, dir);
-            tmpW = DOT3(axis, axis);
-            tmpX = (2.0F * DOT3(axis, dir)) / tmpW;
-            result[0] = tmpX * axis[0] - dir[0];
-            result[1] = tmpX * axis[1] - dir[1];
-            result[2] = tmpX * axis[2] - dir[2];
-            /* result[3] is never written! XXX enforce in parser! */
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_RSQ:         /* 1 / sqrt() */
-         {
-            GLfloat a[4], result[4];
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-            a[0] = FABSF(a[0]);
-            result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("RSQ %g = 1/sqrt(|%g|)\n", result[0], a[0]);
-            }
-         }
-         break;
-      case OPCODE_SCS:         /* sine and cos */
-         {
-            GLfloat a[4], result[4];
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-            result[0] = (GLfloat) cos(a[0]);
-            result[1] = (GLfloat) sin(a[0]);
-            result[2] = 0.0;    /* undefined! */
-            result[3] = 0.0;    /* undefined! */
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_SEQ:         /* set on equal */
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
-            result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
-            result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
-            result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("SEQ (%g %g %g %g) = (%g %g %g %g) == (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_SFL:         /* set false, operands ignored */
-         {
-            static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_SGE:         /* set on greater or equal */
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
-            result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
-            result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
-            result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("SGE (%g %g %g %g) = (%g %g %g %g) >= (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_SGT:         /* set on greater */
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
-            result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
-            result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
-            result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("SGT (%g %g %g %g) = (%g %g %g %g) > (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_SIN:
-         {
-            GLfloat a[4], result[4];
-            fetch_vector1(&inst->SrcReg[0], machine, a);
-            result[0] = result[1] = result[2] = result[3]
-               = (GLfloat) sin(a[0]);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_SLE:         /* set on less or equal */
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
-            result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
-            result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
-            result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("SLE (%g %g %g %g) = (%g %g %g %g) <= (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_SLT:         /* set on less */
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
-            result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
-            result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
-            result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("SLT (%g %g %g %g) = (%g %g %g %g) < (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_SNE:         /* set on not equal */
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
-            result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
-            result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
-            result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("SNE (%g %g %g %g) = (%g %g %g %g) != (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3],
-                      b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_SSG:         /* set sign (-1, 0 or +1) */
-         {
-            GLfloat a[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            result[0] = (GLfloat) ((a[0] > 0.0F) - (a[0] < 0.0F));
-            result[1] = (GLfloat) ((a[1] > 0.0F) - (a[1] < 0.0F));
-            result[2] = (GLfloat) ((a[2] > 0.0F) - (a[2] < 0.0F));
-            result[3] = (GLfloat) ((a[3] > 0.0F) - (a[3] < 0.0F));
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_STR:         /* set true, operands ignored */
-         {
-            static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_SUB:
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = a[0] - b[0];
-            result[1] = a[1] - b[1];
-            result[2] = a[2] - b[2];
-            result[3] = a[3] - b[3];
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
-            }
-         }
-         break;
-      case OPCODE_SWZ:         /* extended swizzle */
-         {
-            const struct prog_src_register *source = &inst->SrcReg[0];
-            const GLfloat *src = get_src_register_pointer(source, machine);
-            GLfloat result[4];
-            GLuint i;
-            for (i = 0; i < 4; i++) {
-               const GLuint swz = GET_SWZ(source->Swizzle, i);
-               if (swz == SWIZZLE_ZERO)
-                  result[i] = 0.0;
-               else if (swz == SWIZZLE_ONE)
-                  result[i] = 1.0;
-               else {
-                  ASSERT(swz >= 0);
-                  ASSERT(swz <= 3);
-                  result[i] = src[swz];
-               }
-               if (source->Negate & (1 << i))
-                  result[i] = -result[i];
-            }
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_TEX:         /* Both ARB and NV frag prog */
-         /* Simple texel lookup */
-         {
-            GLfloat texcoord[4], color[4];
-            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
-
-            fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
-
-            if (DEBUG_PROG) {
-               printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n",
-                      color[0], color[1], color[2], color[3],
-                      inst->TexSrcUnit,
-                      texcoord[0], texcoord[1], texcoord[2], texcoord[3]);
-            }
-            store_vector4(inst, machine, color);
-         }
-         break;
-      case OPCODE_TXB:         /* GL_ARB_fragment_program only */
-         /* Texel lookup with LOD bias */
-         {
-            GLfloat texcoord[4], color[4], lodBias;
-
-            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
-
-            /* texcoord[3] is the bias to add to lambda */
-            lodBias = texcoord[3];
-
-            fetch_texel(ctx, machine, inst, texcoord, lodBias, color);
-
-            if (DEBUG_PROG) {
-               printf("TXB (%g, %g, %g, %g) = texture[%d][%g %g %g %g]"
-                      "  bias %g\n",
-                      color[0], color[1], color[2], color[3],
-                      inst->TexSrcUnit,
-                      texcoord[0],
-                      texcoord[1],
-                      texcoord[2],
-                      texcoord[3],
-                      lodBias);
-            }
-
-            store_vector4(inst, machine, color);
-         }
-         break;
-      case OPCODE_TXD:         /* GL_NV_fragment_program only */
-         /* Texture lookup w/ partial derivatives for LOD */
-         {
-            GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
-            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
-            fetch_vector4(&inst->SrcReg[1], machine, dtdx);
-            fetch_vector4(&inst->SrcReg[2], machine, dtdy);
-            machine->FetchTexelDeriv(ctx, texcoord, dtdx, dtdy,
-                                     0.0, /* lodBias */
-                                     inst->TexSrcUnit, color);
-            store_vector4(inst, machine, color);
-         }
-         break;
-      case OPCODE_TXL:
-         /* Texel lookup with explicit LOD */
-         {
-            GLfloat texcoord[4], color[4], lod;
-
-            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
-
-            /* texcoord[3] is the LOD */
-            lod = texcoord[3];
-
-	    machine->FetchTexelLod(ctx, texcoord, lod,
-				   machine->Samplers[inst->TexSrcUnit], color);
-
-            store_vector4(inst, machine, color);
-         }
-         break;
-      case OPCODE_TXP:         /* GL_ARB_fragment_program only */
-         /* Texture lookup w/ projective divide */
-         {
-            GLfloat texcoord[4], color[4];
-
-            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
-            /* Not so sure about this test - if texcoord[3] is
-             * zero, we'd probably be fine except for an ASSERT in
-             * IROUND_POS() which gets triggered by the inf values created.
-             */
-            if (texcoord[3] != 0.0) {
-               texcoord[0] /= texcoord[3];
-               texcoord[1] /= texcoord[3];
-               texcoord[2] /= texcoord[3];
-            }
-
-            fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
-
-            store_vector4(inst, machine, color);
-         }
-         break;
-      case OPCODE_TXP_NV:      /* GL_NV_fragment_program only */
-         /* Texture lookup w/ projective divide, as above, but do not
-          * do the divide by w if sampling from a cube map.
-          */
-         {
-            GLfloat texcoord[4], color[4];
-
-            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
-            if (inst->TexSrcTarget != TEXTURE_CUBE_INDEX &&
-                texcoord[3] != 0.0) {
-               texcoord[0] /= texcoord[3];
-               texcoord[1] /= texcoord[3];
-               texcoord[2] /= texcoord[3];
-            }
-
-            fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
-
-            store_vector4(inst, machine, color);
-         }
-         break;
-      case OPCODE_TRUNC:       /* truncate toward zero */
-         {
-            GLfloat a[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            result[0] = (GLfloat) (GLint) a[0];
-            result[1] = (GLfloat) (GLint) a[1];
-            result[2] = (GLfloat) (GLint) a[2];
-            result[3] = (GLfloat) (GLint) a[3];
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_UP2H:        /* unpack two 16-bit floats */
-         {
-            const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
-            GLfloat result[4];
-            GLushort hx, hy;
-            hx = raw & 0xffff;
-            hy = raw >> 16;
-            result[0] = result[2] = _mesa_half_to_float(hx);
-            result[1] = result[3] = _mesa_half_to_float(hy);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_UP2US:       /* unpack two GLushorts */
-         {
-            const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
-            GLfloat result[4];
-            GLushort usx, usy;
-            usx = raw & 0xffff;
-            usy = raw >> 16;
-            result[0] = result[2] = usx * (1.0f / 65535.0f);
-            result[1] = result[3] = usy * (1.0f / 65535.0f);
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_UP4B:        /* unpack four GLbytes */
-         {
-            const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
-            GLfloat result[4];
-            result[0] = (((raw >> 0) & 0xff) - 128) / 127.0F;
-            result[1] = (((raw >> 8) & 0xff) - 128) / 127.0F;
-            result[2] = (((raw >> 16) & 0xff) - 128) / 127.0F;
-            result[3] = (((raw >> 24) & 0xff) - 128) / 127.0F;
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_UP4UB:       /* unpack four GLubytes */
-         {
-            const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
-            GLfloat result[4];
-            result[0] = ((raw >> 0) & 0xff) / 255.0F;
-            result[1] = ((raw >> 8) & 0xff) / 255.0F;
-            result[2] = ((raw >> 16) & 0xff) / 255.0F;
-            result[3] = ((raw >> 24) & 0xff) / 255.0F;
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_XOR:         /* bitwise XOR */
-         {
-            GLuint a[4], b[4], result[4];
-            fetch_vector4ui(&inst->SrcReg[0], machine, a);
-            fetch_vector4ui(&inst->SrcReg[1], machine, b);
-            result[0] = a[0] ^ b[0];
-            result[1] = a[1] ^ b[1];
-            result[2] = a[2] ^ b[2];
-            result[3] = a[3] ^ b[3];
-            store_vector4ui(inst, machine, result);
-         }
-         break;
-      case OPCODE_XPD:         /* cross product */
-         {
-            GLfloat a[4], b[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            result[0] = a[1] * b[2] - a[2] * b[1];
-            result[1] = a[2] * b[0] - a[0] * b[2];
-            result[2] = a[0] * b[1] - a[1] * b[0];
-            result[3] = 1.0;
-            store_vector4(inst, machine, result);
-            if (DEBUG_PROG) {
-               printf("XPD (%g %g %g %g) = (%g %g %g) X (%g %g %g)\n",
-                      result[0], result[1], result[2], result[3],
-                      a[0], a[1], a[2], b[0], b[1], b[2]);
-            }
-         }
-         break;
-      case OPCODE_X2D:         /* 2-D matrix transform */
-         {
-            GLfloat a[4], b[4], c[4], result[4];
-            fetch_vector4(&inst->SrcReg[0], machine, a);
-            fetch_vector4(&inst->SrcReg[1], machine, b);
-            fetch_vector4(&inst->SrcReg[2], machine, c);
-            result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
-            result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
-            result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
-            result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
-            store_vector4(inst, machine, result);
-         }
-         break;
-      case OPCODE_PRINT:
-         {
-            if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
-               GLfloat a[4];
-               fetch_vector4(&inst->SrcReg[0], machine, a);
-               printf("%s%g, %g, %g, %g\n", (const char *) inst->Data,
-                            a[0], a[1], a[2], a[3]);
-            }
-            else {
-               printf("%s\n", (const char *) inst->Data);
-            }
-         }
-         break;
-      case OPCODE_END:
-         return GL_TRUE;
-      default:
-         _mesa_problem(ctx, "Bad opcode %d in _mesa_execute_program",
-                       inst->Opcode);
-         return GL_TRUE;        /* return value doesn't matter */
-      }
-
-      numExec++;
-      if (numExec > maxExec) {
-	 static GLboolean reported = GL_FALSE;
-	 if (!reported) {
-	    _mesa_problem(ctx, "Infinite loop detected in fragment program");
-	    reported = GL_TRUE;
-	 }
-         return GL_TRUE;
-      }
-
-   } /* for pc */
-
-   return GL_TRUE;
-}
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.3
+ *
+ * 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.
+ */
+
+/**
+ * \file prog_execute.c
+ * Software interpreter for vertex/fragment programs.
+ * \author Brian Paul
+ */
+
+/*
+ * NOTE: we do everything in single-precision floating point; we don't
+ * currently observe the single/half/fixed-precision qualifiers.
+ *
+ */
+
+
+#include "main/glheader.h"
+#include "main/colormac.h"
+#include "main/macros.h"
+#include "prog_execute.h"
+#include "prog_instruction.h"
+#include "prog_parameter.h"
+#include "prog_print.h"
+#include "prog_noise.h"
+
+
+/* debug predicate */
+#define DEBUG_PROG 0
+
+
+/**
+ * Set x to positive or negative infinity.
+ */
+#if defined(USE_IEEE) || defined(_WIN32)
+#define SET_POS_INFINITY(x)                  \
+   do {                                      \
+         fi_type fi;                         \
+         fi.i = 0x7F800000;                  \
+         x = fi.f;                           \
+   } while (0)
+#define SET_NEG_INFINITY(x)                  \
+   do {                                      \
+         fi_type fi;                         \
+         fi.i = 0xFF800000;                  \
+         x = fi.f;                           \
+   } while (0)
+#elif defined(VMS)
+#define SET_POS_INFINITY(x)  x = __MAXFLOAT
+#define SET_NEG_INFINITY(x)  x = -__MAXFLOAT
+#else
+#define SET_POS_INFINITY(x)  x = (GLfloat) HUGE_VAL
+#define SET_NEG_INFINITY(x)  x = (GLfloat) -HUGE_VAL
+#endif
+
+#define SET_FLOAT_BITS(x, bits) ((fi_type *) (void *) &(x))->i = bits
+
+
+static const GLfloat ZeroVec[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
+
+
+
+/**
+ * Return TRUE for +0 and other positive values, FALSE otherwise.
+ * Used for RCC opcode.
+ */
+static INLINE GLboolean
+positive(float x)
+{
+   fi_type fi;
+   fi.f = x;
+   if (fi.i & 0x80000000)
+      return GL_FALSE;
+   return GL_TRUE;
+}
+
+
+
+/**
+ * Return a pointer to the 4-element float vector specified by the given
+ * source register.
+ */
+static INLINE const GLfloat *
+get_src_register_pointer(const struct prog_src_register *source,
+                         const struct gl_program_machine *machine)
+{
+   const struct gl_program *prog = machine->CurProgram;
+   GLint reg = source->Index;
+
+   if (source->RelAddr) {
+      /* add address register value to src index/offset */
+      reg += machine->AddressReg[0][0];
+      if (reg < 0) {
+         return ZeroVec;
+      }
+   }
+
+   switch (source->File) {
+   case PROGRAM_TEMPORARY:
+      if (reg >= MAX_PROGRAM_TEMPS)
+         return ZeroVec;
+      return machine->Temporaries[reg];
+
+   case PROGRAM_INPUT:
+      if (prog->Target == GL_VERTEX_PROGRAM_ARB) {
+         if (reg >= VERT_ATTRIB_MAX)
+            return ZeroVec;
+         return machine->VertAttribs[reg];
+      }
+      else {
+         if (reg >= FRAG_ATTRIB_MAX)
+            return ZeroVec;
+         return machine->Attribs[reg][machine->CurElement];
+      }
+
+   case PROGRAM_OUTPUT:
+      if (reg >= MAX_PROGRAM_OUTPUTS)
+         return ZeroVec;
+      return machine->Outputs[reg];
+
+   case PROGRAM_LOCAL_PARAM:
+      if (reg >= MAX_PROGRAM_LOCAL_PARAMS)
+         return ZeroVec;
+      return machine->CurProgram->LocalParams[reg];
+
+   case PROGRAM_ENV_PARAM:
+      if (reg >= MAX_PROGRAM_ENV_PARAMS)
+         return ZeroVec;
+      return machine->EnvParams[reg];
+
+   case PROGRAM_STATE_VAR:
+      /* Fallthrough */
+   case PROGRAM_CONSTANT:
+      /* Fallthrough */
+   case PROGRAM_UNIFORM:
+      /* Fallthrough */
+   case PROGRAM_NAMED_PARAM:
+      if (reg >= (GLint) prog->Parameters->NumParameters)
+         return ZeroVec;
+      return (GLfloat *) prog->Parameters->ParameterValues[reg];
+
+   case PROGRAM_SYSTEM_VALUE:
+      assert(reg < Elements(machine->SystemValues));
+      return machine->SystemValues[reg];
+
+   default:
+      _mesa_problem(NULL,
+         "Invalid src register file %d in get_src_register_pointer()",
+         source->File);
+      return NULL;
+   }
+}
+
+
+/**
+ * Return a pointer to the 4-element float vector specified by the given
+ * destination register.
+ */
+static INLINE GLfloat *
+get_dst_register_pointer(const struct prog_dst_register *dest,
+                         struct gl_program_machine *machine)
+{
+   static GLfloat dummyReg[4];
+   GLint reg = dest->Index;
+
+   if (dest->RelAddr) {
+      /* add address register value to src index/offset */
+      reg += machine->AddressReg[0][0];
+      if (reg < 0) {
+         return dummyReg;
+      }
+   }
+
+   switch (dest->File) {
+   case PROGRAM_TEMPORARY:
+      if (reg >= MAX_PROGRAM_TEMPS)
+         return dummyReg;
+      return machine->Temporaries[reg];
+
+   case PROGRAM_OUTPUT:
+      if (reg >= MAX_PROGRAM_OUTPUTS)
+         return dummyReg;
+      return machine->Outputs[reg];
+
+   case PROGRAM_WRITE_ONLY:
+      return dummyReg;
+
+   default:
+      _mesa_problem(NULL,
+         "Invalid dest register file %d in get_dst_register_pointer()",
+         dest->File);
+      return NULL;
+   }
+}
+
+
+
+/**
+ * Fetch a 4-element float vector from the given source register.
+ * Apply swizzling and negating as needed.
+ */
+static void
+fetch_vector4(const struct prog_src_register *source,
+              const struct gl_program_machine *machine, GLfloat result[4])
+{
+   const GLfloat *src = get_src_register_pointer(source, machine);
+   ASSERT(src);
+
+   if (source->Swizzle == SWIZZLE_NOOP) {
+      /* no swizzling */
+      COPY_4V(result, src);
+   }
+   else {
+      ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
+      ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
+      ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
+      ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
+      result[0] = src[GET_SWZ(source->Swizzle, 0)];
+      result[1] = src[GET_SWZ(source->Swizzle, 1)];
+      result[2] = src[GET_SWZ(source->Swizzle, 2)];
+      result[3] = src[GET_SWZ(source->Swizzle, 3)];
+   }
+
+   if (source->Abs) {
+      result[0] = FABSF(result[0]);
+      result[1] = FABSF(result[1]);
+      result[2] = FABSF(result[2]);
+      result[3] = FABSF(result[3]);
+   }
+   if (source->Negate) {
+      ASSERT(source->Negate == NEGATE_XYZW);
+      result[0] = -result[0];
+      result[1] = -result[1];
+      result[2] = -result[2];
+      result[3] = -result[3];
+   }
+
+#ifdef NAN_CHECK
+   assert(!IS_INF_OR_NAN(result[0]));
+   assert(!IS_INF_OR_NAN(result[0]));
+   assert(!IS_INF_OR_NAN(result[0]));
+   assert(!IS_INF_OR_NAN(result[0]));
+#endif
+}
+
+
+/**
+ * Fetch a 4-element uint vector from the given source register.
+ * Apply swizzling but not negation/abs.
+ */
+static void
+fetch_vector4ui(const struct prog_src_register *source,
+                const struct gl_program_machine *machine, GLuint result[4])
+{
+   const GLuint *src = (GLuint *) get_src_register_pointer(source, machine);
+   ASSERT(src);
+
+   if (source->Swizzle == SWIZZLE_NOOP) {
+      /* no swizzling */
+      COPY_4V(result, src);
+   }
+   else {
+      ASSERT(GET_SWZ(source->Swizzle, 0) <= 3);
+      ASSERT(GET_SWZ(source->Swizzle, 1) <= 3);
+      ASSERT(GET_SWZ(source->Swizzle, 2) <= 3);
+      ASSERT(GET_SWZ(source->Swizzle, 3) <= 3);
+      result[0] = src[GET_SWZ(source->Swizzle, 0)];
+      result[1] = src[GET_SWZ(source->Swizzle, 1)];
+      result[2] = src[GET_SWZ(source->Swizzle, 2)];
+      result[3] = src[GET_SWZ(source->Swizzle, 3)];
+   }
+
+   /* Note: no Negate or Abs here */
+}
+
+
+
+/**
+ * Fetch the derivative with respect to X or Y for the given register.
+ * XXX this currently only works for fragment program input attribs.
+ */
+static void
+fetch_vector4_deriv(struct gl_context * ctx,
+                    const struct prog_src_register *source,
+                    const struct gl_program_machine *machine,
+                    char xOrY, GLfloat result[4])
+{
+   if (source->File == PROGRAM_INPUT &&
+       source->Index < (GLint) machine->NumDeriv) {
+      const GLint col = machine->CurElement;
+      const GLfloat w = machine->Attribs[FRAG_ATTRIB_WPOS][col][3];
+      const GLfloat invQ = 1.0f / w;
+      GLfloat deriv[4];
+
+      if (xOrY == 'X') {
+         deriv[0] = machine->DerivX[source->Index][0] * invQ;
+         deriv[1] = machine->DerivX[source->Index][1] * invQ;
+         deriv[2] = machine->DerivX[source->Index][2] * invQ;
+         deriv[3] = machine->DerivX[source->Index][3] * invQ;
+      }
+      else {
+         deriv[0] = machine->DerivY[source->Index][0] * invQ;
+         deriv[1] = machine->DerivY[source->Index][1] * invQ;
+         deriv[2] = machine->DerivY[source->Index][2] * invQ;
+         deriv[3] = machine->DerivY[source->Index][3] * invQ;
+      }
+
+      result[0] = deriv[GET_SWZ(source->Swizzle, 0)];
+      result[1] = deriv[GET_SWZ(source->Swizzle, 1)];
+      result[2] = deriv[GET_SWZ(source->Swizzle, 2)];
+      result[3] = deriv[GET_SWZ(source->Swizzle, 3)];
+      
+      if (source->Abs) {
+         result[0] = FABSF(result[0]);
+         result[1] = FABSF(result[1]);
+         result[2] = FABSF(result[2]);
+         result[3] = FABSF(result[3]);
+      }
+      if (source->Negate) {
+         ASSERT(source->Negate == NEGATE_XYZW);
+         result[0] = -result[0];
+         result[1] = -result[1];
+         result[2] = -result[2];
+         result[3] = -result[3];
+      }
+   }
+   else {
+      ASSIGN_4V(result, 0.0, 0.0, 0.0, 0.0);
+   }
+}
+
+
+/**
+ * As above, but only return result[0] element.
+ */
+static void
+fetch_vector1(const struct prog_src_register *source,
+              const struct gl_program_machine *machine, GLfloat result[4])
+{
+   const GLfloat *src = get_src_register_pointer(source, machine);
+   ASSERT(src);
+
+   result[0] = src[GET_SWZ(source->Swizzle, 0)];
+
+   if (source->Abs) {
+      result[0] = FABSF(result[0]);
+   }
+   if (source->Negate) {
+      result[0] = -result[0];
+   }
+}
+
+
+static GLuint
+fetch_vector1ui(const struct prog_src_register *source,
+                const struct gl_program_machine *machine)
+{
+   const GLuint *src = (GLuint *) get_src_register_pointer(source, machine);
+   return src[GET_SWZ(source->Swizzle, 0)];
+}
+
+
+/**
+ * Fetch texel from texture.  Use partial derivatives when possible.
+ */
+static INLINE void
+fetch_texel(struct gl_context *ctx,
+            const struct gl_program_machine *machine,
+            const struct prog_instruction *inst,
+            const GLfloat texcoord[4], GLfloat lodBias,
+            GLfloat color[4])
+{
+   const GLuint unit = machine->Samplers[inst->TexSrcUnit];
+
+   /* Note: we only have the right derivatives for fragment input attribs.
+    */
+   if (machine->NumDeriv > 0 &&
+       inst->SrcReg[0].File == PROGRAM_INPUT &&
+       inst->SrcReg[0].Index == FRAG_ATTRIB_TEX0 + inst->TexSrcUnit) {
+      /* simple texture fetch for which we should have derivatives */
+      GLuint attr = inst->SrcReg[0].Index;
+      machine->FetchTexelDeriv(ctx, texcoord,
+                               machine->DerivX[attr],
+                               machine->DerivY[attr],
+                               lodBias, unit, color);
+   }
+   else {
+      machine->FetchTexelLod(ctx, texcoord, lodBias, unit, color);
+   }
+}
+
+
+/**
+ * Test value against zero and return GT, LT, EQ or UN if NaN.
+ */
+static INLINE GLuint
+generate_cc(float value)
+{
+   if (value != value)
+      return COND_UN;           /* NaN */
+   if (value > 0.0F)
+      return COND_GT;
+   if (value < 0.0F)
+      return COND_LT;
+   return COND_EQ;
+}
+
+
+/**
+ * Test if the ccMaskRule is satisfied by the given condition code.
+ * Used to mask destination writes according to the current condition code.
+ */
+static INLINE GLboolean
+test_cc(GLuint condCode, GLuint ccMaskRule)
+{
+   switch (ccMaskRule) {
+   case COND_EQ: return (condCode == COND_EQ);
+   case COND_NE: return (condCode != COND_EQ);
+   case COND_LT: return (condCode == COND_LT);
+   case COND_GE: return (condCode == COND_GT || condCode == COND_EQ);
+   case COND_LE: return (condCode == COND_LT || condCode == COND_EQ);
+   case COND_GT: return (condCode == COND_GT);
+   case COND_TR: return GL_TRUE;
+   case COND_FL: return GL_FALSE;
+   default:      return GL_TRUE;
+   }
+}
+
+
+/**
+ * Evaluate the 4 condition codes against a predicate and return GL_TRUE
+ * or GL_FALSE to indicate result.
+ */
+static INLINE GLboolean
+eval_condition(const struct gl_program_machine *machine,
+               const struct prog_instruction *inst)
+{
+   const GLuint swizzle = inst->DstReg.CondSwizzle;
+   const GLuint condMask = inst->DstReg.CondMask;
+   if (test_cc(machine->CondCodes[GET_SWZ(swizzle, 0)], condMask) ||
+       test_cc(machine->CondCodes[GET_SWZ(swizzle, 1)], condMask) ||
+       test_cc(machine->CondCodes[GET_SWZ(swizzle, 2)], condMask) ||
+       test_cc(machine->CondCodes[GET_SWZ(swizzle, 3)], condMask)) {
+      return GL_TRUE;
+   }
+   else {
+      return GL_FALSE;
+   }
+}
+
+
+
+/**
+ * Store 4 floats into a register.  Observe the instructions saturate and
+ * set-condition-code flags.
+ */
+static void
+store_vector4(const struct prog_instruction *inst,
+              struct gl_program_machine *machine, const GLfloat value[4])
+{
+   const struct prog_dst_register *dstReg = &(inst->DstReg);
+   const GLboolean clamp = inst->SaturateMode == SATURATE_ZERO_ONE;
+   GLuint writeMask = dstReg->WriteMask;
+   GLfloat clampedValue[4];
+   GLfloat *dst = get_dst_register_pointer(dstReg, machine);
+
+#if 0
+   if (value[0] > 1.0e10 ||
+       IS_INF_OR_NAN(value[0]) ||
+       IS_INF_OR_NAN(value[1]) ||
+       IS_INF_OR_NAN(value[2]) || IS_INF_OR_NAN(value[3]))
+      printf("store %g %g %g %g\n", value[0], value[1], value[2], value[3]);
+#endif
+
+   if (clamp) {
+      clampedValue[0] = CLAMP(value[0], 0.0F, 1.0F);
+      clampedValue[1] = CLAMP(value[1], 0.0F, 1.0F);
+      clampedValue[2] = CLAMP(value[2], 0.0F, 1.0F);
+      clampedValue[3] = CLAMP(value[3], 0.0F, 1.0F);
+      value = clampedValue;
+   }
+
+   if (dstReg->CondMask != COND_TR) {
+      /* condition codes may turn off some writes */
+      if (writeMask & WRITEMASK_X) {
+         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
+                      dstReg->CondMask))
+            writeMask &= ~WRITEMASK_X;
+      }
+      if (writeMask & WRITEMASK_Y) {
+         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
+                      dstReg->CondMask))
+            writeMask &= ~WRITEMASK_Y;
+      }
+      if (writeMask & WRITEMASK_Z) {
+         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
+                      dstReg->CondMask))
+            writeMask &= ~WRITEMASK_Z;
+      }
+      if (writeMask & WRITEMASK_W) {
+         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
+                      dstReg->CondMask))
+            writeMask &= ~WRITEMASK_W;
+      }
+   }
+
+#ifdef NAN_CHECK
+   assert(!IS_INF_OR_NAN(value[0]));
+   assert(!IS_INF_OR_NAN(value[0]));
+   assert(!IS_INF_OR_NAN(value[0]));
+   assert(!IS_INF_OR_NAN(value[0]));
+#endif
+
+   if (writeMask & WRITEMASK_X)
+      dst[0] = value[0];
+   if (writeMask & WRITEMASK_Y)
+      dst[1] = value[1];
+   if (writeMask & WRITEMASK_Z)
+      dst[2] = value[2];
+   if (writeMask & WRITEMASK_W)
+      dst[3] = value[3];
+
+   if (inst->CondUpdate) {
+      if (writeMask & WRITEMASK_X)
+         machine->CondCodes[0] = generate_cc(value[0]);
+      if (writeMask & WRITEMASK_Y)
+         machine->CondCodes[1] = generate_cc(value[1]);
+      if (writeMask & WRITEMASK_Z)
+         machine->CondCodes[2] = generate_cc(value[2]);
+      if (writeMask & WRITEMASK_W)
+         machine->CondCodes[3] = generate_cc(value[3]);
+#if DEBUG_PROG
+      printf("CondCodes=(%s,%s,%s,%s) for:\n",
+             _mesa_condcode_string(machine->CondCodes[0]),
+             _mesa_condcode_string(machine->CondCodes[1]),
+             _mesa_condcode_string(machine->CondCodes[2]),
+             _mesa_condcode_string(machine->CondCodes[3]));
+#endif
+   }
+}
+
+
+/**
+ * Store 4 uints into a register.  Observe the set-condition-code flags.
+ */
+static void
+store_vector4ui(const struct prog_instruction *inst,
+                struct gl_program_machine *machine, const GLuint value[4])
+{
+   const struct prog_dst_register *dstReg = &(inst->DstReg);
+   GLuint writeMask = dstReg->WriteMask;
+   GLuint *dst = (GLuint *) get_dst_register_pointer(dstReg, machine);
+
+   if (dstReg->CondMask != COND_TR) {
+      /* condition codes may turn off some writes */
+      if (writeMask & WRITEMASK_X) {
+         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 0)],
+                      dstReg->CondMask))
+            writeMask &= ~WRITEMASK_X;
+      }
+      if (writeMask & WRITEMASK_Y) {
+         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 1)],
+                      dstReg->CondMask))
+            writeMask &= ~WRITEMASK_Y;
+      }
+      if (writeMask & WRITEMASK_Z) {
+         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 2)],
+                      dstReg->CondMask))
+            writeMask &= ~WRITEMASK_Z;
+      }
+      if (writeMask & WRITEMASK_W) {
+         if (!test_cc(machine->CondCodes[GET_SWZ(dstReg->CondSwizzle, 3)],
+                      dstReg->CondMask))
+            writeMask &= ~WRITEMASK_W;
+      }
+   }
+
+   if (writeMask & WRITEMASK_X)
+      dst[0] = value[0];
+   if (writeMask & WRITEMASK_Y)
+      dst[1] = value[1];
+   if (writeMask & WRITEMASK_Z)
+      dst[2] = value[2];
+   if (writeMask & WRITEMASK_W)
+      dst[3] = value[3];
+
+   if (inst->CondUpdate) {
+      if (writeMask & WRITEMASK_X)
+         machine->CondCodes[0] = generate_cc((float)value[0]);
+      if (writeMask & WRITEMASK_Y)
+         machine->CondCodes[1] = generate_cc((float)value[1]);
+      if (writeMask & WRITEMASK_Z)
+         machine->CondCodes[2] = generate_cc((float)value[2]);
+      if (writeMask & WRITEMASK_W)
+         machine->CondCodes[3] = generate_cc((float)value[3]);
+#if DEBUG_PROG
+      printf("CondCodes=(%s,%s,%s,%s) for:\n",
+             _mesa_condcode_string(machine->CondCodes[0]),
+             _mesa_condcode_string(machine->CondCodes[1]),
+             _mesa_condcode_string(machine->CondCodes[2]),
+             _mesa_condcode_string(machine->CondCodes[3]));
+#endif
+   }
+}
+
+
+
+/**
+ * Execute the given vertex/fragment program.
+ *
+ * \param ctx  rendering context
+ * \param program  the program to execute
+ * \param machine  machine state (must be initialized)
+ * \return GL_TRUE if program completed or GL_FALSE if program executed KIL.
+ */
+GLboolean
+_mesa_execute_program(struct gl_context * ctx,
+                      const struct gl_program *program,
+                      struct gl_program_machine *machine)
+{
+   const GLuint numInst = program->NumInstructions;
+   const GLuint maxExec = 10000;
+   GLuint pc, numExec = 0;
+
+   machine->CurProgram = program;
+
+   if (DEBUG_PROG) {
+      printf("execute program %u --------------------\n", program->Id);
+   }
+
+   if (program->Target == GL_VERTEX_PROGRAM_ARB) {
+      machine->EnvParams = ctx->VertexProgram.Parameters;
+   }
+   else {
+      machine->EnvParams = ctx->FragmentProgram.Parameters;
+   }
+
+   for (pc = 0; pc < numInst; pc++) {
+      const struct prog_instruction *inst = program->Instructions + pc;
+
+      if (DEBUG_PROG) {
+         _mesa_print_instruction(inst);
+      }
+
+      switch (inst->Opcode) {
+      case OPCODE_ABS:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            result[0] = FABSF(a[0]);
+            result[1] = FABSF(a[1]);
+            result[2] = FABSF(a[2]);
+            result[3] = FABSF(a[3]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_ADD:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = a[0] + b[0];
+            result[1] = a[1] + b[1];
+            result[2] = a[2] + b[2];
+            result[3] = a[3] + b[3];
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("ADD (%g %g %g %g) = (%g %g %g %g) + (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_AND:     /* bitwise AND */
+         {
+            GLuint a[4], b[4], result[4];
+            fetch_vector4ui(&inst->SrcReg[0], machine, a);
+            fetch_vector4ui(&inst->SrcReg[1], machine, b);
+            result[0] = a[0] & b[0];
+            result[1] = a[1] & b[1];
+            result[2] = a[2] & b[2];
+            result[3] = a[3] & b[3];
+            store_vector4ui(inst, machine, result);
+         }
+         break;
+      case OPCODE_ARL:
+         {
+            GLfloat t[4];
+            fetch_vector4(&inst->SrcReg[0], machine, t);
+            machine->AddressReg[0][0] = IFLOOR(t[0]);
+            if (DEBUG_PROG) {
+               printf("ARL %d\n", machine->AddressReg[0][0]);
+            }
+         }
+         break;
+      case OPCODE_BGNLOOP:
+         /* no-op */
+         ASSERT(program->Instructions[inst->BranchTarget].Opcode
+                == OPCODE_ENDLOOP);
+         break;
+      case OPCODE_ENDLOOP:
+         /* subtract 1 here since pc is incremented by for(pc) loop */
+         ASSERT(program->Instructions[inst->BranchTarget].Opcode
+                == OPCODE_BGNLOOP);
+         pc = inst->BranchTarget - 1;   /* go to matching BNGLOOP */
+         break;
+      case OPCODE_BGNSUB:      /* begin subroutine */
+         break;
+      case OPCODE_ENDSUB:      /* end subroutine */
+         break;
+      case OPCODE_BRA:         /* branch (conditional) */
+         if (eval_condition(machine, inst)) {
+            /* take branch */
+            /* Subtract 1 here since we'll do pc++ below */
+            pc = inst->BranchTarget - 1;
+         }
+         break;
+      case OPCODE_BRK:         /* break out of loop (conditional) */
+         ASSERT(program->Instructions[inst->BranchTarget].Opcode
+                == OPCODE_ENDLOOP);
+         if (eval_condition(machine, inst)) {
+            /* break out of loop */
+            /* pc++ at end of for-loop will put us after the ENDLOOP inst */
+            pc = inst->BranchTarget;
+         }
+         break;
+      case OPCODE_CONT:        /* continue loop (conditional) */
+         ASSERT(program->Instructions[inst->BranchTarget].Opcode
+                == OPCODE_ENDLOOP);
+         if (eval_condition(machine, inst)) {
+            /* continue at ENDLOOP */
+            /* Subtract 1 here since we'll do pc++ at end of for-loop */
+            pc = inst->BranchTarget - 1;
+         }
+         break;
+      case OPCODE_CAL:         /* Call subroutine (conditional) */
+         if (eval_condition(machine, inst)) {
+            /* call the subroutine */
+            if (machine->StackDepth >= MAX_PROGRAM_CALL_DEPTH) {
+               return GL_TRUE;  /* Per GL_NV_vertex_program2 spec */
+            }
+            machine->CallStack[machine->StackDepth++] = pc + 1; /* next inst */
+            /* Subtract 1 here since we'll do pc++ at end of for-loop */
+            pc = inst->BranchTarget - 1;
+         }
+         break;
+      case OPCODE_CMP:
+         {
+            GLfloat a[4], b[4], c[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            fetch_vector4(&inst->SrcReg[2], machine, c);
+            result[0] = a[0] < 0.0F ? b[0] : c[0];
+            result[1] = a[1] < 0.0F ? b[1] : c[1];
+            result[2] = a[2] < 0.0F ? b[2] : c[2];
+            result[3] = a[3] < 0.0F ? b[3] : c[3];
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("CMP (%g %g %g %g) = (%g %g %g %g) < 0 ? (%g %g %g %g) : (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3],
+                      c[0], c[1], c[2], c[3]);
+            }
+         }
+         break;
+      case OPCODE_COS:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            result[0] = result[1] = result[2] = result[3]
+               = (GLfloat) cos(a[0]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_DDX:         /* Partial derivative with respect to X */
+         {
+            GLfloat result[4];
+            fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
+                                'X', result);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_DDY:         /* Partial derivative with respect to Y */
+         {
+            GLfloat result[4];
+            fetch_vector4_deriv(ctx, &inst->SrcReg[0], machine,
+                                'Y', result);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_DP2:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = result[1] = result[2] = result[3] = DOT2(a, b);
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("DP2 %g = (%g %g) . (%g %g)\n",
+                      result[0], a[0], a[1], b[0], b[1]);
+            }
+         }
+         break;
+      case OPCODE_DP2A:
+         {
+            GLfloat a[4], b[4], c, result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            fetch_vector1(&inst->SrcReg[1], machine, &c);
+            result[0] = result[1] = result[2] = result[3] = DOT2(a, b) + c;
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("DP2A %g = (%g %g) . (%g %g) + %g\n",
+                      result[0], a[0], a[1], b[0], b[1], c);
+            }
+         }
+         break;
+      case OPCODE_DP3:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = result[1] = result[2] = result[3] = DOT3(a, b);
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("DP3 %g = (%g %g %g) . (%g %g %g)\n",
+                      result[0], a[0], a[1], a[2], b[0], b[1], b[2]);
+            }
+         }
+         break;
+      case OPCODE_DP4:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = result[1] = result[2] = result[3] = DOT4(a, b);
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("DP4 %g = (%g, %g %g %g) . (%g, %g %g %g)\n",
+                      result[0], a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_DPH:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = result[1] = result[2] = result[3] = DOT3(a, b) + b[3];
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_DST:         /* Distance vector */
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = 1.0F;
+            result[1] = a[1] * b[1];
+            result[2] = a[2];
+            result[3] = b[3];
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_EXP:
+         {
+            GLfloat t[4], q[4], floor_t0;
+            fetch_vector1(&inst->SrcReg[0], machine, t);
+            floor_t0 = FLOORF(t[0]);
+            if (floor_t0 > FLT_MAX_EXP) {
+               SET_POS_INFINITY(q[0]);
+               SET_POS_INFINITY(q[2]);
+            }
+            else if (floor_t0 < FLT_MIN_EXP) {
+               q[0] = 0.0F;
+               q[2] = 0.0F;
+            }
+            else {
+               q[0] = LDEXPF(1.0, (int) floor_t0);
+               /* Note: GL_NV_vertex_program expects 
+                * result.z = result.x * APPX(result.y)
+                * We do what the ARB extension says.
+                */
+               q[2] = (GLfloat) pow(2.0, t[0]);
+            }
+            q[1] = t[0] - floor_t0;
+            q[3] = 1.0F;
+            store_vector4( inst, machine, q );
+         }
+         break;
+      case OPCODE_EX2:         /* Exponential base 2 */
+         {
+            GLfloat a[4], result[4], val;
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            val = (GLfloat) pow(2.0, a[0]);
+            /*
+            if (IS_INF_OR_NAN(val))
+               val = 1.0e10;
+            */
+            result[0] = result[1] = result[2] = result[3] = val;
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_FLR:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            result[0] = FLOORF(a[0]);
+            result[1] = FLOORF(a[1]);
+            result[2] = FLOORF(a[2]);
+            result[3] = FLOORF(a[3]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_FRC:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            result[0] = a[0] - FLOORF(a[0]);
+            result[1] = a[1] - FLOORF(a[1]);
+            result[2] = a[2] - FLOORF(a[2]);
+            result[3] = a[3] - FLOORF(a[3]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_IF:
+         {
+            GLboolean cond;
+            ASSERT(program->Instructions[inst->BranchTarget].Opcode
+                   == OPCODE_ELSE ||
+                   program->Instructions[inst->BranchTarget].Opcode
+                   == OPCODE_ENDIF);
+            /* eval condition */
+            if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
+               GLfloat a[4];
+               fetch_vector1(&inst->SrcReg[0], machine, a);
+               cond = (a[0] != 0.0);
+            }
+            else {
+               cond = eval_condition(machine, inst);
+            }
+            if (DEBUG_PROG) {
+               printf("IF: %d\n", cond);
+            }
+            /* do if/else */
+            if (cond) {
+               /* do if-clause (just continue execution) */
+            }
+            else {
+               /* go to the instruction after ELSE or ENDIF */
+               assert(inst->BranchTarget >= 0);
+               pc = inst->BranchTarget;
+            }
+         }
+         break;
+      case OPCODE_ELSE:
+         /* goto ENDIF */
+         ASSERT(program->Instructions[inst->BranchTarget].Opcode
+                == OPCODE_ENDIF);
+         assert(inst->BranchTarget >= 0);
+         pc = inst->BranchTarget;
+         break;
+      case OPCODE_ENDIF:
+         /* nothing */
+         break;
+      case OPCODE_KIL_NV:      /* NV_f_p only (conditional) */
+         if (eval_condition(machine, inst)) {
+            return GL_FALSE;
+         }
+         break;
+      case OPCODE_KIL:         /* ARB_f_p only */
+         {
+            GLfloat a[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            if (DEBUG_PROG) {
+               printf("KIL if (%g %g %g %g) <= 0.0\n",
+                      a[0], a[1], a[2], a[3]);
+            }
+
+            if (a[0] < 0.0F || a[1] < 0.0F || a[2] < 0.0F || a[3] < 0.0F) {
+               return GL_FALSE;
+            }
+         }
+         break;
+      case OPCODE_LG2:         /* log base 2 */
+         {
+            GLfloat a[4], result[4], val;
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+	    /* The fast LOG2 macro doesn't meet the precision requirements.
+	     */
+            if (a[0] == 0.0F) {
+               val = -FLT_MAX;
+            }
+            else {
+               val = (float)(log(a[0]) * 1.442695F);
+            }
+            result[0] = result[1] = result[2] = result[3] = val;
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_LIT:
+         {
+            const GLfloat epsilon = 1.0F / 256.0F;      /* from NV VP spec */
+            GLfloat a[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            a[0] = MAX2(a[0], 0.0F);
+            a[1] = MAX2(a[1], 0.0F);
+            /* XXX ARB version clamps a[3], NV version doesn't */
+            a[3] = CLAMP(a[3], -(128.0F - epsilon), (128.0F - epsilon));
+            result[0] = 1.0F;
+            result[1] = a[0];
+            /* XXX we could probably just use pow() here */
+            if (a[0] > 0.0F) {
+               if (a[1] == 0.0 && a[3] == 0.0)
+                  result[2] = 1.0F;
+               else
+                  result[2] = (GLfloat) pow(a[1], a[3]);
+            }
+            else {
+               result[2] = 0.0F;
+            }
+            result[3] = 1.0F;
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("LIT (%g %g %g %g) : (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3]);
+            }
+         }
+         break;
+      case OPCODE_LOG:
+         {
+            GLfloat t[4], q[4], abs_t0;
+            fetch_vector1(&inst->SrcReg[0], machine, t);
+            abs_t0 = FABSF(t[0]);
+            if (abs_t0 != 0.0F) {
+               /* Since we really can't handle infinite values on VMS
+                * like other OSes we'll use __MAXFLOAT to represent
+                * infinity.  This may need some tweaking.
+                */
+#ifdef VMS
+               if (abs_t0 == __MAXFLOAT)
+#else
+               if (IS_INF_OR_NAN(abs_t0))
+#endif
+               {
+                  SET_POS_INFINITY(q[0]);
+                  q[1] = 1.0F;
+                  SET_POS_INFINITY(q[2]);
+               }
+               else {
+                  int exponent;
+                  GLfloat mantissa = FREXPF(t[0], &exponent);
+                  q[0] = (GLfloat) (exponent - 1);
+                  q[1] = (GLfloat) (2.0 * mantissa); /* map [.5, 1) -> [1, 2) */
+
+		  /* The fast LOG2 macro doesn't meet the precision
+		   * requirements.
+		   */
+                  q[2] = (float)(log(t[0]) * 1.442695F);
+               }
+            }
+            else {
+               SET_NEG_INFINITY(q[0]);
+               q[1] = 1.0F;
+               SET_NEG_INFINITY(q[2]);
+            }
+            q[3] = 1.0;
+            store_vector4(inst, machine, q);
+         }
+         break;
+      case OPCODE_LRP:
+         {
+            GLfloat a[4], b[4], c[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            fetch_vector4(&inst->SrcReg[2], machine, c);
+            result[0] = a[0] * b[0] + (1.0F - a[0]) * c[0];
+            result[1] = a[1] * b[1] + (1.0F - a[1]) * c[1];
+            result[2] = a[2] * b[2] + (1.0F - a[2]) * c[2];
+            result[3] = a[3] * b[3] + (1.0F - a[3]) * c[3];
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("LRP (%g %g %g %g) = (%g %g %g %g), "
+                      "(%g %g %g %g), (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
+            }
+         }
+         break;
+      case OPCODE_MAD:
+         {
+            GLfloat a[4], b[4], c[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            fetch_vector4(&inst->SrcReg[2], machine, c);
+            result[0] = a[0] * b[0] + c[0];
+            result[1] = a[1] * b[1] + c[1];
+            result[2] = a[2] * b[2] + c[2];
+            result[3] = a[3] * b[3] + c[3];
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("MAD (%g %g %g %g) = (%g %g %g %g) * "
+                      "(%g %g %g %g) + (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3], c[0], c[1], c[2], c[3]);
+            }
+         }
+         break;
+      case OPCODE_MAX:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = MAX2(a[0], b[0]);
+            result[1] = MAX2(a[1], b[1]);
+            result[2] = MAX2(a[2], b[2]);
+            result[3] = MAX2(a[3], b[3]);
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("MAX (%g %g %g %g) = (%g %g %g %g), (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_MIN:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = MIN2(a[0], b[0]);
+            result[1] = MIN2(a[1], b[1]);
+            result[2] = MIN2(a[2], b[2]);
+            result[3] = MIN2(a[3], b[3]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_MOV:
+         {
+            GLfloat result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, result);
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("MOV (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3]);
+            }
+         }
+         break;
+      case OPCODE_MUL:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = a[0] * b[0];
+            result[1] = a[1] * b[1];
+            result[2] = a[2] * b[2];
+            result[3] = a[3] * b[3];
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("MUL (%g %g %g %g) = (%g %g %g %g) * (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_NOISE1:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            result[0] =
+               result[1] =
+               result[2] =
+               result[3] = _mesa_noise1(a[0]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_NOISE2:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            result[0] =
+               result[1] =
+               result[2] = result[3] = _mesa_noise2(a[0], a[1]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_NOISE3:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            result[0] =
+               result[1] =
+               result[2] =
+               result[3] = _mesa_noise3(a[0], a[1], a[2]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_NOISE4:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            result[0] =
+               result[1] =
+               result[2] =
+               result[3] = _mesa_noise4(a[0], a[1], a[2], a[3]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_NOP:
+         break;
+      case OPCODE_NOT:         /* bitwise NOT */
+         {
+            GLuint a[4], result[4];
+            fetch_vector4ui(&inst->SrcReg[0], machine, a);
+            result[0] = ~a[0];
+            result[1] = ~a[1];
+            result[2] = ~a[2];
+            result[3] = ~a[3];
+            store_vector4ui(inst, machine, result);
+         }
+         break;
+      case OPCODE_NRM3:        /* 3-component normalization */
+         {
+            GLfloat a[4], result[4];
+            GLfloat tmp;
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2];
+            if (tmp != 0.0F)
+               tmp = INV_SQRTF(tmp);
+            result[0] = tmp * a[0];
+            result[1] = tmp * a[1];
+            result[2] = tmp * a[2];
+            result[3] = 0.0;  /* undefined, but prevent valgrind warnings */
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_NRM4:        /* 4-component normalization */
+         {
+            GLfloat a[4], result[4];
+            GLfloat tmp;
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            tmp = a[0] * a[0] + a[1] * a[1] + a[2] * a[2] + a[3] * a[3];
+            if (tmp != 0.0F)
+               tmp = INV_SQRTF(tmp);
+            result[0] = tmp * a[0];
+            result[1] = tmp * a[1];
+            result[2] = tmp * a[2];
+            result[3] = tmp * a[3];
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_OR:          /* bitwise OR */
+         {
+            GLuint a[4], b[4], result[4];
+            fetch_vector4ui(&inst->SrcReg[0], machine, a);
+            fetch_vector4ui(&inst->SrcReg[1], machine, b);
+            result[0] = a[0] | b[0];
+            result[1] = a[1] | b[1];
+            result[2] = a[2] | b[2];
+            result[3] = a[3] | b[3];
+            store_vector4ui(inst, machine, result);
+         }
+         break;
+      case OPCODE_PK2H:        /* pack two 16-bit floats in one 32-bit float */
+         {
+            GLfloat a[4];
+            GLuint result[4];
+            GLhalfNV hx, hy;
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            hx = _mesa_float_to_half(a[0]);
+            hy = _mesa_float_to_half(a[1]);
+            result[0] =
+            result[1] =
+            result[2] =
+            result[3] = hx | (hy << 16);
+            store_vector4ui(inst, machine, result);
+         }
+         break;
+      case OPCODE_PK2US:       /* pack two GLushorts into one 32-bit float */
+         {
+            GLfloat a[4];
+            GLuint result[4], usx, usy;
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            a[0] = CLAMP(a[0], 0.0F, 1.0F);
+            a[1] = CLAMP(a[1], 0.0F, 1.0F);
+            usx = IROUND(a[0] * 65535.0F);
+            usy = IROUND(a[1] * 65535.0F);
+            result[0] =
+            result[1] =
+            result[2] =
+            result[3] = usx | (usy << 16);
+            store_vector4ui(inst, machine, result);
+         }
+         break;
+      case OPCODE_PK4B:        /* pack four GLbytes into one 32-bit float */
+         {
+            GLfloat a[4];
+            GLuint result[4], ubx, uby, ubz, ubw;
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            a[0] = CLAMP(a[0], -128.0F / 127.0F, 1.0F);
+            a[1] = CLAMP(a[1], -128.0F / 127.0F, 1.0F);
+            a[2] = CLAMP(a[2], -128.0F / 127.0F, 1.0F);
+            a[3] = CLAMP(a[3], -128.0F / 127.0F, 1.0F);
+            ubx = IROUND(127.0F * a[0] + 128.0F);
+            uby = IROUND(127.0F * a[1] + 128.0F);
+            ubz = IROUND(127.0F * a[2] + 128.0F);
+            ubw = IROUND(127.0F * a[3] + 128.0F);
+            result[0] =
+            result[1] =
+            result[2] =
+            result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
+            store_vector4ui(inst, machine, result);
+         }
+         break;
+      case OPCODE_PK4UB:       /* pack four GLubytes into one 32-bit float */
+         {
+            GLfloat a[4];
+            GLuint result[4], ubx, uby, ubz, ubw;
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            a[0] = CLAMP(a[0], 0.0F, 1.0F);
+            a[1] = CLAMP(a[1], 0.0F, 1.0F);
+            a[2] = CLAMP(a[2], 0.0F, 1.0F);
+            a[3] = CLAMP(a[3], 0.0F, 1.0F);
+            ubx = IROUND(255.0F * a[0]);
+            uby = IROUND(255.0F * a[1]);
+            ubz = IROUND(255.0F * a[2]);
+            ubw = IROUND(255.0F * a[3]);
+            result[0] =
+            result[1] =
+            result[2] =
+            result[3] = ubx | (uby << 8) | (ubz << 16) | (ubw << 24);
+            store_vector4ui(inst, machine, result);
+         }
+         break;
+      case OPCODE_POW:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            fetch_vector1(&inst->SrcReg[1], machine, b);
+            result[0] = result[1] = result[2] = result[3]
+               = (GLfloat) pow(a[0], b[0]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_RCC:  /* clamped riciprocal */
+         {
+            const float largest = 1.884467e+19, smallest = 5.42101e-20;
+            GLfloat a[4], r, result[4];
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            if (DEBUG_PROG) {
+               if (a[0] == 0)
+                  printf("RCC(0)\n");
+               else if (IS_INF_OR_NAN(a[0]))
+                  printf("RCC(inf)\n");
+            }
+            if (a[0] == 1.0F) {
+               r = 1.0F;
+            }
+            else {
+               r = 1.0F / a[0];
+            }
+            if (positive(r)) {
+               if (r > largest) {
+                  r = largest;
+               }
+               else if (r < smallest) {
+                  r = smallest;
+               }
+            }
+            else {
+               if (r < -largest) {
+                  r = -largest;
+               }
+               else if (r > -smallest) {
+                  r = -smallest;
+               }
+            }
+            result[0] = result[1] = result[2] = result[3] = r;
+            store_vector4(inst, machine, result);
+         }
+         break;
+
+      case OPCODE_RCP:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            if (DEBUG_PROG) {
+               if (a[0] == 0)
+                  printf("RCP(0)\n");
+               else if (IS_INF_OR_NAN(a[0]))
+                  printf("RCP(inf)\n");
+            }
+            result[0] = result[1] = result[2] = result[3] = 1.0F / a[0];
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_RET:         /* return from subroutine (conditional) */
+         if (eval_condition(machine, inst)) {
+            if (machine->StackDepth == 0) {
+               return GL_TRUE;  /* Per GL_NV_vertex_program2 spec */
+            }
+            /* subtract one because of pc++ in the for loop */
+            pc = machine->CallStack[--machine->StackDepth] - 1;
+         }
+         break;
+      case OPCODE_RFL:         /* reflection vector */
+         {
+            GLfloat axis[4], dir[4], result[4], tmpX, tmpW;
+            fetch_vector4(&inst->SrcReg[0], machine, axis);
+            fetch_vector4(&inst->SrcReg[1], machine, dir);
+            tmpW = DOT3(axis, axis);
+            tmpX = (2.0F * DOT3(axis, dir)) / tmpW;
+            result[0] = tmpX * axis[0] - dir[0];
+            result[1] = tmpX * axis[1] - dir[1];
+            result[2] = tmpX * axis[2] - dir[2];
+            /* result[3] is never written! XXX enforce in parser! */
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_RSQ:         /* 1 / sqrt() */
+         {
+            GLfloat a[4], result[4];
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            a[0] = FABSF(a[0]);
+            result[0] = result[1] = result[2] = result[3] = INV_SQRTF(a[0]);
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("RSQ %g = 1/sqrt(|%g|)\n", result[0], a[0]);
+            }
+         }
+         break;
+      case OPCODE_SCS:         /* sine and cos */
+         {
+            GLfloat a[4], result[4];
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            result[0] = (GLfloat) cos(a[0]);
+            result[1] = (GLfloat) sin(a[0]);
+            result[2] = 0.0;    /* undefined! */
+            result[3] = 0.0;    /* undefined! */
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_SEQ:         /* set on equal */
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = (a[0] == b[0]) ? 1.0F : 0.0F;
+            result[1] = (a[1] == b[1]) ? 1.0F : 0.0F;
+            result[2] = (a[2] == b[2]) ? 1.0F : 0.0F;
+            result[3] = (a[3] == b[3]) ? 1.0F : 0.0F;
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("SEQ (%g %g %g %g) = (%g %g %g %g) == (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_SFL:         /* set false, operands ignored */
+         {
+            static const GLfloat result[4] = { 0.0F, 0.0F, 0.0F, 0.0F };
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_SGE:         /* set on greater or equal */
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = (a[0] >= b[0]) ? 1.0F : 0.0F;
+            result[1] = (a[1] >= b[1]) ? 1.0F : 0.0F;
+            result[2] = (a[2] >= b[2]) ? 1.0F : 0.0F;
+            result[3] = (a[3] >= b[3]) ? 1.0F : 0.0F;
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("SGE (%g %g %g %g) = (%g %g %g %g) >= (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_SGT:         /* set on greater */
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = (a[0] > b[0]) ? 1.0F : 0.0F;
+            result[1] = (a[1] > b[1]) ? 1.0F : 0.0F;
+            result[2] = (a[2] > b[2]) ? 1.0F : 0.0F;
+            result[3] = (a[3] > b[3]) ? 1.0F : 0.0F;
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("SGT (%g %g %g %g) = (%g %g %g %g) > (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_SIN:
+         {
+            GLfloat a[4], result[4];
+            fetch_vector1(&inst->SrcReg[0], machine, a);
+            result[0] = result[1] = result[2] = result[3]
+               = (GLfloat) sin(a[0]);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_SLE:         /* set on less or equal */
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = (a[0] <= b[0]) ? 1.0F : 0.0F;
+            result[1] = (a[1] <= b[1]) ? 1.0F : 0.0F;
+            result[2] = (a[2] <= b[2]) ? 1.0F : 0.0F;
+            result[3] = (a[3] <= b[3]) ? 1.0F : 0.0F;
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("SLE (%g %g %g %g) = (%g %g %g %g) <= (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_SLT:         /* set on less */
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = (a[0] < b[0]) ? 1.0F : 0.0F;
+            result[1] = (a[1] < b[1]) ? 1.0F : 0.0F;
+            result[2] = (a[2] < b[2]) ? 1.0F : 0.0F;
+            result[3] = (a[3] < b[3]) ? 1.0F : 0.0F;
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("SLT (%g %g %g %g) = (%g %g %g %g) < (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_SNE:         /* set on not equal */
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = (a[0] != b[0]) ? 1.0F : 0.0F;
+            result[1] = (a[1] != b[1]) ? 1.0F : 0.0F;
+            result[2] = (a[2] != b[2]) ? 1.0F : 0.0F;
+            result[3] = (a[3] != b[3]) ? 1.0F : 0.0F;
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("SNE (%g %g %g %g) = (%g %g %g %g) != (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3],
+                      b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_SSG:         /* set sign (-1, 0 or +1) */
+         {
+            GLfloat a[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            result[0] = (GLfloat) ((a[0] > 0.0F) - (a[0] < 0.0F));
+            result[1] = (GLfloat) ((a[1] > 0.0F) - (a[1] < 0.0F));
+            result[2] = (GLfloat) ((a[2] > 0.0F) - (a[2] < 0.0F));
+            result[3] = (GLfloat) ((a[3] > 0.0F) - (a[3] < 0.0F));
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_STR:         /* set true, operands ignored */
+         {
+            static const GLfloat result[4] = { 1.0F, 1.0F, 1.0F, 1.0F };
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_SUB:
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = a[0] - b[0];
+            result[1] = a[1] - b[1];
+            result[2] = a[2] - b[2];
+            result[3] = a[3] - b[3];
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("SUB (%g %g %g %g) = (%g %g %g %g) - (%g %g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], a[3], b[0], b[1], b[2], b[3]);
+            }
+         }
+         break;
+      case OPCODE_SWZ:         /* extended swizzle */
+         {
+            const struct prog_src_register *source = &inst->SrcReg[0];
+            const GLfloat *src = get_src_register_pointer(source, machine);
+            GLfloat result[4];
+            GLuint i;
+            for (i = 0; i < 4; i++) {
+               const GLuint swz = GET_SWZ(source->Swizzle, i);
+               if (swz == SWIZZLE_ZERO)
+                  result[i] = 0.0;
+               else if (swz == SWIZZLE_ONE)
+                  result[i] = 1.0;
+               else {
+                  ASSERT(swz >= 0);
+                  ASSERT(swz <= 3);
+                  result[i] = src[swz];
+               }
+               if (source->Negate & (1 << i))
+                  result[i] = -result[i];
+            }
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_TEX:         /* Both ARB and NV frag prog */
+         /* Simple texel lookup */
+         {
+            GLfloat texcoord[4], color[4];
+            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
+
+            fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
+
+            if (DEBUG_PROG) {
+               printf("TEX (%g, %g, %g, %g) = texture[%d][%g, %g, %g, %g]\n",
+                      color[0], color[1], color[2], color[3],
+                      inst->TexSrcUnit,
+                      texcoord[0], texcoord[1], texcoord[2], texcoord[3]);
+            }
+            store_vector4(inst, machine, color);
+         }
+         break;
+      case OPCODE_TXB:         /* GL_ARB_fragment_program only */
+         /* Texel lookup with LOD bias */
+         {
+            GLfloat texcoord[4], color[4], lodBias;
+
+            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
+
+            /* texcoord[3] is the bias to add to lambda */
+            lodBias = texcoord[3];
+
+            fetch_texel(ctx, machine, inst, texcoord, lodBias, color);
+
+            if (DEBUG_PROG) {
+               printf("TXB (%g, %g, %g, %g) = texture[%d][%g %g %g %g]"
+                      "  bias %g\n",
+                      color[0], color[1], color[2], color[3],
+                      inst->TexSrcUnit,
+                      texcoord[0],
+                      texcoord[1],
+                      texcoord[2],
+                      texcoord[3],
+                      lodBias);
+            }
+
+            store_vector4(inst, machine, color);
+         }
+         break;
+      case OPCODE_TXD:         /* GL_NV_fragment_program only */
+         /* Texture lookup w/ partial derivatives for LOD */
+         {
+            GLfloat texcoord[4], dtdx[4], dtdy[4], color[4];
+            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
+            fetch_vector4(&inst->SrcReg[1], machine, dtdx);
+            fetch_vector4(&inst->SrcReg[2], machine, dtdy);
+            machine->FetchTexelDeriv(ctx, texcoord, dtdx, dtdy,
+                                     0.0, /* lodBias */
+                                     inst->TexSrcUnit, color);
+            store_vector4(inst, machine, color);
+         }
+         break;
+      case OPCODE_TXL:
+         /* Texel lookup with explicit LOD */
+         {
+            GLfloat texcoord[4], color[4], lod;
+
+            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
+
+            /* texcoord[3] is the LOD */
+            lod = texcoord[3];
+
+	    machine->FetchTexelLod(ctx, texcoord, lod,
+				   machine->Samplers[inst->TexSrcUnit], color);
+
+            store_vector4(inst, machine, color);
+         }
+         break;
+      case OPCODE_TXP:         /* GL_ARB_fragment_program only */
+         /* Texture lookup w/ projective divide */
+         {
+            GLfloat texcoord[4], color[4];
+
+            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
+            /* Not so sure about this test - if texcoord[3] is
+             * zero, we'd probably be fine except for an ASSERT in
+             * IROUND_POS() which gets triggered by the inf values created.
+             */
+            if (texcoord[3] != 0.0) {
+               texcoord[0] /= texcoord[3];
+               texcoord[1] /= texcoord[3];
+               texcoord[2] /= texcoord[3];
+            }
+
+            fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
+
+            store_vector4(inst, machine, color);
+         }
+         break;
+      case OPCODE_TXP_NV:      /* GL_NV_fragment_program only */
+         /* Texture lookup w/ projective divide, as above, but do not
+          * do the divide by w if sampling from a cube map.
+          */
+         {
+            GLfloat texcoord[4], color[4];
+
+            fetch_vector4(&inst->SrcReg[0], machine, texcoord);
+            if (inst->TexSrcTarget != TEXTURE_CUBE_INDEX &&
+                texcoord[3] != 0.0) {
+               texcoord[0] /= texcoord[3];
+               texcoord[1] /= texcoord[3];
+               texcoord[2] /= texcoord[3];
+            }
+
+            fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
+
+            store_vector4(inst, machine, color);
+         }
+         break;
+      case OPCODE_TRUNC:       /* truncate toward zero */
+         {
+            GLfloat a[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            result[0] = (GLfloat) (GLint) a[0];
+            result[1] = (GLfloat) (GLint) a[1];
+            result[2] = (GLfloat) (GLint) a[2];
+            result[3] = (GLfloat) (GLint) a[3];
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_UP2H:        /* unpack two 16-bit floats */
+         {
+            const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
+            GLfloat result[4];
+            GLushort hx, hy;
+            hx = raw & 0xffff;
+            hy = raw >> 16;
+            result[0] = result[2] = _mesa_half_to_float(hx);
+            result[1] = result[3] = _mesa_half_to_float(hy);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_UP2US:       /* unpack two GLushorts */
+         {
+            const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
+            GLfloat result[4];
+            GLushort usx, usy;
+            usx = raw & 0xffff;
+            usy = raw >> 16;
+            result[0] = result[2] = usx * (1.0f / 65535.0f);
+            result[1] = result[3] = usy * (1.0f / 65535.0f);
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_UP4B:        /* unpack four GLbytes */
+         {
+            const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
+            GLfloat result[4];
+            result[0] = (((raw >> 0) & 0xff) - 128) / 127.0F;
+            result[1] = (((raw >> 8) & 0xff) - 128) / 127.0F;
+            result[2] = (((raw >> 16) & 0xff) - 128) / 127.0F;
+            result[3] = (((raw >> 24) & 0xff) - 128) / 127.0F;
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_UP4UB:       /* unpack four GLubytes */
+         {
+            const GLuint raw = fetch_vector1ui(&inst->SrcReg[0], machine);
+            GLfloat result[4];
+            result[0] = ((raw >> 0) & 0xff) / 255.0F;
+            result[1] = ((raw >> 8) & 0xff) / 255.0F;
+            result[2] = ((raw >> 16) & 0xff) / 255.0F;
+            result[3] = ((raw >> 24) & 0xff) / 255.0F;
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_XOR:         /* bitwise XOR */
+         {
+            GLuint a[4], b[4], result[4];
+            fetch_vector4ui(&inst->SrcReg[0], machine, a);
+            fetch_vector4ui(&inst->SrcReg[1], machine, b);
+            result[0] = a[0] ^ b[0];
+            result[1] = a[1] ^ b[1];
+            result[2] = a[2] ^ b[2];
+            result[3] = a[3] ^ b[3];
+            store_vector4ui(inst, machine, result);
+         }
+         break;
+      case OPCODE_XPD:         /* cross product */
+         {
+            GLfloat a[4], b[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            result[0] = a[1] * b[2] - a[2] * b[1];
+            result[1] = a[2] * b[0] - a[0] * b[2];
+            result[2] = a[0] * b[1] - a[1] * b[0];
+            result[3] = 1.0;
+            store_vector4(inst, machine, result);
+            if (DEBUG_PROG) {
+               printf("XPD (%g %g %g %g) = (%g %g %g) X (%g %g %g)\n",
+                      result[0], result[1], result[2], result[3],
+                      a[0], a[1], a[2], b[0], b[1], b[2]);
+            }
+         }
+         break;
+      case OPCODE_X2D:         /* 2-D matrix transform */
+         {
+            GLfloat a[4], b[4], c[4], result[4];
+            fetch_vector4(&inst->SrcReg[0], machine, a);
+            fetch_vector4(&inst->SrcReg[1], machine, b);
+            fetch_vector4(&inst->SrcReg[2], machine, c);
+            result[0] = a[0] + b[0] * c[0] + b[1] * c[1];
+            result[1] = a[1] + b[0] * c[2] + b[1] * c[3];
+            result[2] = a[2] + b[0] * c[0] + b[1] * c[1];
+            result[3] = a[3] + b[0] * c[2] + b[1] * c[3];
+            store_vector4(inst, machine, result);
+         }
+         break;
+      case OPCODE_PRINT:
+         {
+            if (inst->SrcReg[0].File != PROGRAM_UNDEFINED) {
+               GLfloat a[4];
+               fetch_vector4(&inst->SrcReg[0], machine, a);
+               printf("%s%g, %g, %g, %g\n", (const char *) inst->Data,
+                            a[0], a[1], a[2], a[3]);
+            }
+            else {
+               printf("%s\n", (const char *) inst->Data);
+            }
+         }
+         break;
+      case OPCODE_END:
+         return GL_TRUE;
+      default:
+         _mesa_problem(ctx, "Bad opcode %d in _mesa_execute_program",
+                       inst->Opcode);
+         return GL_TRUE;        /* return value doesn't matter */
+      }
+
+      numExec++;
+      if (numExec > maxExec) {
+	 static GLboolean reported = GL_FALSE;
+	 if (!reported) {
+	    _mesa_problem(ctx, "Infinite loop detected in fragment program");
+	    reported = GL_TRUE;
+	 }
+         return GL_TRUE;
+      }
+
+   } /* for pc */
+
+   return GL_TRUE;
+}
diff --git a/mesalib/src/mesa/program/prog_parameter.c b/mesalib/src/mesa/program/prog_parameter.c
index 157e31b56..49b3ffbdd 100644
--- a/mesalib/src/mesa/program/prog_parameter.c
+++ b/mesalib/src/mesa/program/prog_parameter.c
@@ -1,656 +1,680 @@
-/*
- * Mesa 3-D graphics library
- * Version:  7.3
- *
- * 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.
- */
-
-/**
- * \file prog_parameter.c
- * Program parameter lists and functions.
- * \author Brian Paul
- */
-
-
-#include "main/glheader.h"
-#include "main/imports.h"
-#include "main/macros.h"
-#include "prog_instruction.h"
-#include "prog_parameter.h"
-#include "prog_statevars.h"
-
-
-struct gl_program_parameter_list *
-_mesa_new_parameter_list(void)
-{
-   return CALLOC_STRUCT(gl_program_parameter_list);
-}
-
-
-struct gl_program_parameter_list *
-_mesa_new_parameter_list_sized(unsigned size)
-{
-   struct gl_program_parameter_list *p = _mesa_new_parameter_list();
-
-   if ((p != NULL) && (size != 0)) {
-      p->Size = size;
-
-      /* alloc arrays */
-      p->Parameters = (struct gl_program_parameter *)
-	 calloc(1, size * sizeof(struct gl_program_parameter));
-
-      p->ParameterValues = (GLfloat (*)[4])
-         _mesa_align_malloc(size * 4 *sizeof(GLfloat), 16);
-
-
-      if ((p->Parameters == NULL) || (p->ParameterValues == NULL)) {
-	 free(p->Parameters);
-	 _mesa_align_free(p->ParameterValues);
-	 free(p);
-	 p = NULL;
-      }
-   }
-
-   return p;
-}
-
-
-/**
- * Free a parameter list and all its parameters
- */
-void
-_mesa_free_parameter_list(struct gl_program_parameter_list *paramList)
-{
-   GLuint i;
-   for (i = 0; i < paramList->NumParameters; i++) {
-      if (paramList->Parameters[i].Name)
-	 free((void *) paramList->Parameters[i].Name);
-   }
-   free(paramList->Parameters);
-   if (paramList->ParameterValues)
-      _mesa_align_free(paramList->ParameterValues);
-   free(paramList);
-}
-
-
-/**
- * Add a new parameter to a parameter list.
- * Note that parameter values are usually 4-element GLfloat vectors.
- * When size > 4 we'll allocate a sequential block of parameters to
- * store all the values (in blocks of 4).
- *
- * \param paramList  the list to add the parameter to
- * \param type  type of parameter, such as 
- * \param name  the parameter name, will be duplicated/copied!
- * \param size  number of elements in 'values' vector (1..4, or more)
- * \param datatype  GL_FLOAT, GL_FLOAT_VECx, GL_INT, GL_INT_VECx or GL_NONE.
- * \param values  initial parameter value, up to 4 GLfloats, or NULL
- * \param state  state indexes, or NULL
- * \return  index of new parameter in the list, or -1 if error (out of mem)
- */
-GLint
-_mesa_add_parameter(struct gl_program_parameter_list *paramList,
-                    gl_register_file type, const char *name,
-                    GLuint size, GLenum datatype, const GLfloat *values,
-                    const gl_state_index state[STATE_LENGTH],
-                    GLbitfield flags)
-{
-   const GLuint oldNum = paramList->NumParameters;
-   const GLuint sz4 = (size + 3) / 4; /* no. of new param slots needed */
-
-   assert(size > 0);
-
-   if (oldNum + sz4 > paramList->Size) {
-      /* Need to grow the parameter list array (alloc some extra) */
-      paramList->Size = paramList->Size + 4 * sz4;
-
-      /* realloc arrays */
-      paramList->Parameters = (struct gl_program_parameter *)
-	 _mesa_realloc(paramList->Parameters,
-		       oldNum * sizeof(struct gl_program_parameter),
-		       paramList->Size * sizeof(struct gl_program_parameter));
-
-      paramList->ParameterValues = (GLfloat (*)[4])
-         _mesa_align_realloc(paramList->ParameterValues,         /* old buf */
-                             oldNum * 4 * sizeof(GLfloat),      /* old size */
-                             paramList->Size * 4 *sizeof(GLfloat), /* new sz */
-                             16);
-   }
-
-   if (!paramList->Parameters ||
-       !paramList->ParameterValues) {
-      /* out of memory */
-      paramList->NumParameters = 0;
-      paramList->Size = 0;
-      return -1;
-   }
-   else {
-      GLuint i;
-
-      paramList->NumParameters = oldNum + sz4;
-
-      memset(&paramList->Parameters[oldNum], 0,
-             sz4 * sizeof(struct gl_program_parameter));
-
-      for (i = 0; i < sz4; i++) {
-         struct gl_program_parameter *p = paramList->Parameters + oldNum + i;
-         p->Name = name ? _mesa_strdup(name) : NULL;
-         p->Type = type;
-         p->Size = size;
-         p->DataType = datatype;
-         p->Flags = flags;
-         if (values) {
-            COPY_4V(paramList->ParameterValues[oldNum + i], values);
-            values += 4;
-            p->Initialized = GL_TRUE;
-         }
-         else {
-            /* silence valgrind */
-            ASSIGN_4V(paramList->ParameterValues[oldNum + i], 0, 0, 0, 0);
-         }
-         size -= 4;
-      }
-
-      if (state) {
-         for (i = 0; i < STATE_LENGTH; i++)
-            paramList->Parameters[oldNum].StateIndexes[i] = state[i];
-      }
-
-      return (GLint) oldNum;
-   }
-}
-
-
-/**
- * Add a new named program parameter (Ex: NV_fragment_program DEFINE statement)
- * \return index of the new entry in the parameter list
- */
-GLint
-_mesa_add_named_parameter(struct gl_program_parameter_list *paramList,
-                          const char *name, const GLfloat values[4])
-{
-   return _mesa_add_parameter(paramList, PROGRAM_NAMED_PARAM, name,
-                              4, GL_NONE, values, NULL, 0x0);
-                              
-}
-
-
-/**
- * Add a new named constant to the parameter list.
- * This will be used when the program contains something like this:
- *    PARAM myVals = { 0, 1, 2, 3 };
- *
- * \param paramList  the parameter list
- * \param name  the name for the constant
- * \param values  four float values
- * \return index/position of the new parameter in the parameter list
- */
-GLint
-_mesa_add_named_constant(struct gl_program_parameter_list *paramList,
-                         const char *name, const GLfloat values[4],
-                         GLuint size)
-{
-   /* first check if this is a duplicate constant */
-   GLint pos;
-   for (pos = 0; pos < (GLint)paramList->NumParameters; pos++) {
-      const GLfloat *pvals = paramList->ParameterValues[pos];
-      if (pvals[0] == values[0] &&
-          pvals[1] == values[1] &&
-          pvals[2] == values[2] &&
-          pvals[3] == values[3] &&
-          strcmp(paramList->Parameters[pos].Name, name) == 0) {
-         /* Same name and value is already in the param list - reuse it */
-         return pos;
-      }
-   }
-   /* not found, add new parameter */
-   return _mesa_add_parameter(paramList, PROGRAM_CONSTANT, name,
-                              size, GL_NONE, values, NULL, 0x0);
-}
-
-
-/**
- * Add a new unnamed constant to the parameter list.  This will be used
- * when a fragment/vertex program contains something like this:
- *    MOV r, { 0, 1, 2, 3 };
- * If swizzleOut is non-null we'll search the parameter list for an
- * existing instance of the constant which matches with a swizzle.
- *
- * \param paramList  the parameter list
- * \param values  four float values
- * \param swizzleOut  returns swizzle mask for accessing the constant
- * \return index/position of the new parameter in the parameter list.
- */
-GLint
-_mesa_add_unnamed_constant(struct gl_program_parameter_list *paramList,
-                           const GLfloat values[4], GLuint size,
-                           GLuint *swizzleOut)
-{
-   GLint pos;
-   ASSERT(size >= 1);
-   ASSERT(size <= 4);
-
-   if (swizzleOut &&
-       _mesa_lookup_parameter_constant(paramList, values,
-                                       size, &pos, swizzleOut)) {
-      return pos;
-   }
-
-   /* Look for empty space in an already unnamed constant parameter
-    * to add this constant.  This will only work for single-element
-    * constants because we rely on smearing (i.e. .yyyy or .zzzz).
-    */
-   if (size == 1 && swizzleOut) {
-      for (pos = 0; pos < (GLint) paramList->NumParameters; pos++) {
-         struct gl_program_parameter *p = paramList->Parameters + pos;
-         if (p->Type == PROGRAM_CONSTANT && p->Size + size <= 4) {
-            /* ok, found room */
-            GLfloat *pVal = paramList->ParameterValues[pos];
-            GLuint swz = p->Size; /* 1, 2 or 3 for Y, Z, W */
-            pVal[p->Size] = values[0];
-            p->Size++;
-            *swizzleOut = MAKE_SWIZZLE4(swz, swz, swz, swz);
-            return pos;
-         }
-      }
-   }
-
-   /* add a new parameter to store this constant */
-   pos = _mesa_add_parameter(paramList, PROGRAM_CONSTANT, NULL,
-                             size, GL_NONE, values, NULL, 0x0);
-   if (pos >= 0 && swizzleOut) {
-      if (size == 1)
-         *swizzleOut = SWIZZLE_XXXX;
-      else
-         *swizzleOut = SWIZZLE_NOOP;
-   }
-   return pos;
-}
-
-/**
- * Add parameter representing a varying variable.
- */
-GLint
-_mesa_add_varying(struct gl_program_parameter_list *paramList,
-                  const char *name, GLuint size, GLenum datatype,
-                  GLbitfield flags)
-{
-   GLint i = _mesa_lookup_parameter_index(paramList, -1, name);
-   if (i >= 0 && paramList->Parameters[i].Type == PROGRAM_VARYING) {
-      /* already in list */
-      return i;
-   }
-   else {
-      /*assert(size == 4);*/
-      i = _mesa_add_parameter(paramList, PROGRAM_VARYING, name,
-                              size, datatype, NULL, NULL, flags);
-      return i;
-   }
-}
-
-
-/**
- * Add parameter representing a vertex program attribute.
- * \param size  size of attribute (in floats), may be -1 if unknown
- * \param attrib  the attribute index, or -1 if unknown
- */
-GLint
-_mesa_add_attribute(struct gl_program_parameter_list *paramList,
-                    const char *name, GLint size, GLenum datatype, GLint attrib)
-{
-   GLint i = _mesa_lookup_parameter_index(paramList, -1, name);
-   if (i >= 0) {
-      /* replace */
-      if (attrib < 0)
-         attrib = i;
-      paramList->Parameters[i].StateIndexes[0] = attrib;
-   }
-   else {
-      /* add */
-      gl_state_index state[STATE_LENGTH];
-      state[0] = (gl_state_index) attrib;
-      if (size < 0)
-         size = 4;
-      i = _mesa_add_parameter(paramList, PROGRAM_INPUT, name,
-                              size, datatype, NULL, state, 0x0);
-   }
-   return i;
-}
-
-
-
-#if 0 /* not used yet */
-/**
- * Returns the number of 4-component registers needed to store a piece
- * of GL state.  For matrices this may be as many as 4 registers,
- * everything else needs
- * just 1 register.
- */
-static GLuint
-sizeof_state_reference(const GLint *stateTokens)
-{
-   if (stateTokens[0] == STATE_MATRIX) {
-      GLuint rows = stateTokens[4] - stateTokens[3] + 1;
-      assert(rows >= 1);
-      assert(rows <= 4);
-      return rows;
-   }
-   else {
-      return 1;
-   }
-}
-#endif
-
-
-/**
- * Add a new state reference to the parameter list.
- * This will be used when the program contains something like this:
- *    PARAM ambient = state.material.front.ambient;
- *
- * \param paramList  the parameter list
- * \param stateTokens  an array of 5 (STATE_LENGTH) state tokens
- * \return index of the new parameter.
- */
-GLint
-_mesa_add_state_reference(struct gl_program_parameter_list *paramList,
-                          const gl_state_index stateTokens[STATE_LENGTH])
-{
-   const GLuint size = 4; /* XXX fix */
-   char *name;
-   GLint index;
-
-   /* Check if the state reference is already in the list */
-   for (index = 0; index < (GLint) paramList->NumParameters; index++) {
-      if (!memcmp(paramList->Parameters[index].StateIndexes,
-		  stateTokens, STATE_LENGTH * sizeof(gl_state_index))) {
-	 return index;
-      }
-   }
-
-   name = _mesa_program_state_string(stateTokens);
-   index = _mesa_add_parameter(paramList, PROGRAM_STATE_VAR, name,
-                               size, GL_NONE,
-                               NULL, (gl_state_index *) stateTokens, 0x0);
-   paramList->StateFlags |= _mesa_program_state_flags(stateTokens);
-
-   /* free name string here since we duplicated it in add_parameter() */
-   free(name);
-
-   return index;
-}
-
-
-/**
- * Lookup a parameter value by name in the given parameter list.
- * \return pointer to the float[4] values.
- */
-GLfloat *
-_mesa_lookup_parameter_value(const struct gl_program_parameter_list *paramList,
-                             GLsizei nameLen, const char *name)
-{
-   GLint i = _mesa_lookup_parameter_index(paramList, nameLen, name);
-   if (i < 0)
-      return NULL;
-   else
-      return paramList->ParameterValues[i];
-}
-
-
-/**
- * Given a program parameter name, find its position in the list of parameters.
- * \param paramList  the parameter list to search
- * \param nameLen  length of name (in chars).
- *                 If length is negative, assume that name is null-terminated.
- * \param name  the name to search for
- * \return index of parameter in the list.
- */
-GLint
-_mesa_lookup_parameter_index(const struct gl_program_parameter_list *paramList,
-                             GLsizei nameLen, const char *name)
-{
-   GLint i;
-
-   if (!paramList)
-      return -1;
-
-   if (nameLen == -1) {
-      /* name is null-terminated */
-      for (i = 0; i < (GLint) paramList->NumParameters; i++) {
-         if (paramList->Parameters[i].Name &&
-	     strcmp(paramList->Parameters[i].Name, name) == 0)
-            return i;
-      }
-   }
-   else {
-      /* name is not null-terminated, use nameLen */
-      for (i = 0; i < (GLint) paramList->NumParameters; i++) {
-         if (paramList->Parameters[i].Name &&
-	     strncmp(paramList->Parameters[i].Name, name, nameLen) == 0
-             && strlen(paramList->Parameters[i].Name) == (size_t)nameLen)
-            return i;
-      }
-   }
-   return -1;
-}
-
-
-/**
- * Look for a float vector in the given parameter list.  The float vector
- * may be of length 1, 2, 3 or 4.  If swizzleOut is non-null, we'll try
- * swizzling to find a match.
- * \param list  the parameter list to search
- * \param v  the float vector to search for
- * \param vSize  number of element in v
- * \param posOut  returns the position of the constant, if found
- * \param swizzleOut  returns a swizzle mask describing location of the
- *                    vector elements if found.
- * \return GL_TRUE if found, GL_FALSE if not found
- */
-GLboolean
-_mesa_lookup_parameter_constant(const struct gl_program_parameter_list *list,
-                                const GLfloat v[], GLuint vSize,
-                                GLint *posOut, GLuint *swizzleOut)
-{
-   GLuint i;
-
-   assert(vSize >= 1);
-   assert(vSize <= 4);
-
-   if (!list) {
-      *posOut = -1;
-      return GL_FALSE;
-   }
-
-   for (i = 0; i < list->NumParameters; i++) {
-      if (list->Parameters[i].Type == PROGRAM_CONSTANT) {
-         if (!swizzleOut) {
-            /* swizzle not allowed */
-            GLuint j, match = 0;
-            for (j = 0; j < vSize; j++) {
-               if (v[j] == list->ParameterValues[i][j])
-                  match++;
-            }
-            if (match == vSize) {
-               *posOut = i;
-               return GL_TRUE;
-            }
-         }
-         else {
-            /* try matching w/ swizzle */
-             if (vSize == 1) {
-                /* look for v[0] anywhere within float[4] value */
-                GLuint j;
-                for (j = 0; j < list->Parameters[i].Size; j++) {
-                   if (list->ParameterValues[i][j] == v[0]) {
-                      /* found it */
-                      *posOut = i;
-                      *swizzleOut = MAKE_SWIZZLE4(j, j, j, j);
-                      return GL_TRUE;
-                   }
-                }
-             }
-             else if (vSize <= list->Parameters[i].Size) {
-                /* see if we can match this constant (with a swizzle) */
-                GLuint swz[4];
-                GLuint match = 0, j, k;
-                for (j = 0; j < vSize; j++) {
-                   if (v[j] == list->ParameterValues[i][j]) {
-                      swz[j] = j;
-                      match++;
-                   }
-                   else {
-                      for (k = 0; k < list->Parameters[i].Size; k++) {
-                         if (v[j] == list->ParameterValues[i][k]) {
-                            swz[j] = k;
-                            match++;
-                            break;
-                         }
-                      }
-                   }
-                }
-                /* smear last value to remaining positions */
-                for (; j < 4; j++)
-                   swz[j] = swz[j-1];
-
-                if (match == vSize) {
-                   *posOut = i;
-                   *swizzleOut = MAKE_SWIZZLE4(swz[0], swz[1], swz[2], swz[3]);
-                   return GL_TRUE;
-                }
-             }
-         }
-      }
-   }
-
-   *posOut = -1;
-   return GL_FALSE;
-}
-
-
-struct gl_program_parameter_list *
-_mesa_clone_parameter_list(const struct gl_program_parameter_list *list)
-{
-   struct gl_program_parameter_list *clone;
-   GLuint i;
-
-   clone = _mesa_new_parameter_list();
-   if (!clone)
-      return NULL;
-
-   /** Not too efficient, but correct */
-   for (i = 0; i < list->NumParameters; i++) {
-      struct gl_program_parameter *p = list->Parameters + i;
-      struct gl_program_parameter *pCopy;
-      GLuint size = MIN2(p->Size, 4);
-      GLint j = _mesa_add_parameter(clone, p->Type, p->Name, size, p->DataType,
-                                    list->ParameterValues[i], NULL, 0x0);
-      ASSERT(j >= 0);
-      pCopy = clone->Parameters + j;
-      pCopy->Flags = p->Flags;
-      /* copy state indexes */
-      if (p->Type == PROGRAM_STATE_VAR) {
-         GLint k;
-         for (k = 0; k < STATE_LENGTH; k++) {
-            pCopy->StateIndexes[k] = p->StateIndexes[k];
-         }
-      }
-      else {
-         clone->Parameters[j].Size = p->Size;
-      }
-      
-   }
-
-   clone->StateFlags = list->StateFlags;
-
-   return clone;
-}
-
-
-/**
- * Return a new parameter list which is listA + listB.
- */
-struct gl_program_parameter_list *
-_mesa_combine_parameter_lists(const struct gl_program_parameter_list *listA,
-                              const struct gl_program_parameter_list *listB)
-{
-   struct gl_program_parameter_list *list;
-
-   if (listA) {
-      list = _mesa_clone_parameter_list(listA);
-      if (list && listB) {
-         GLuint i;
-         for (i = 0; i < listB->NumParameters; i++) {
-            struct gl_program_parameter *param = listB->Parameters + i;
-            _mesa_add_parameter(list, param->Type, param->Name, param->Size,
-                                param->DataType,
-                                listB->ParameterValues[i],
-                                param->StateIndexes,
-                                param->Flags);
-         }
-      }
-   }
-   else if (listB) {
-      list = _mesa_clone_parameter_list(listB);
-   }
-   else {
-      list = NULL;
-   }
-   return list;
-}
-
-
-
-/**
- * Find longest name of all uniform parameters in list.
- */
-GLuint
-_mesa_longest_parameter_name(const struct gl_program_parameter_list *list,
-                             gl_register_file type)
-{
-   GLuint i, maxLen = 0;
-   if (!list)
-      return 0;
-   for (i = 0; i < list->NumParameters; i++) {
-      if (list->Parameters[i].Type == type) {
-         GLuint len = strlen(list->Parameters[i].Name);
-         if (len > maxLen)
-            maxLen = len;
-      }
-   }
-   return maxLen;
-}
-
-
-/**
- * Count the number of parameters in the last that match the given type.
- */
-GLuint
-_mesa_num_parameters_of_type(const struct gl_program_parameter_list *list,
-                             gl_register_file type)
-{
-   GLuint i, count = 0;
-   if (list) {
-      for (i = 0; i < list->NumParameters; i++) {
-         if (list->Parameters[i].Type == type)
-            count++;
-      }
-   }
-   return count;
-}
+/*
+ * Mesa 3-D graphics library
+ * Version:  7.3
+ *
+ * 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.
+ */
+
+/**
+ * \file prog_parameter.c
+ * Program parameter lists and functions.
+ * \author Brian Paul
+ */
+
+
+#include "main/glheader.h"
+#include "main/imports.h"
+#include "main/macros.h"
+#include "prog_instruction.h"
+#include "prog_parameter.h"
+#include "prog_statevars.h"
+
+
+struct gl_program_parameter_list *
+_mesa_new_parameter_list(void)
+{
+   return CALLOC_STRUCT(gl_program_parameter_list);
+}
+
+
+struct gl_program_parameter_list *
+_mesa_new_parameter_list_sized(unsigned size)
+{
+   struct gl_program_parameter_list *p = _mesa_new_parameter_list();
+
+   if ((p != NULL) && (size != 0)) {
+      p->Size = size;
+
+      /* alloc arrays */
+      p->Parameters = (struct gl_program_parameter *)
+	 calloc(1, size * sizeof(struct gl_program_parameter));
+
+      p->ParameterValues = (gl_constant_value (*)[4])
+         _mesa_align_malloc(size * 4 *sizeof(gl_constant_value), 16);
+
+
+      if ((p->Parameters == NULL) || (p->ParameterValues == NULL)) {
+	 free(p->Parameters);
+	 _mesa_align_free(p->ParameterValues);
+	 free(p);
+	 p = NULL;
+      }
+   }
+
+   return p;
+}
+
+
+/**
+ * Free a parameter list and all its parameters
+ */
+void
+_mesa_free_parameter_list(struct gl_program_parameter_list *paramList)
+{
+   GLuint i;
+   for (i = 0; i < paramList->NumParameters; i++) {
+      if (paramList->Parameters[i].Name)
+	 free((void *) paramList->Parameters[i].Name);
+   }
+   free(paramList->Parameters);
+   if (paramList->ParameterValues)
+      _mesa_align_free(paramList->ParameterValues);
+   free(paramList);
+}
+
+
+/**
+ * Add a new parameter to a parameter list.
+ * Note that parameter values are usually 4-element GLfloat vectors.
+ * When size > 4 we'll allocate a sequential block of parameters to
+ * store all the values (in blocks of 4).
+ *
+ * \param paramList  the list to add the parameter to
+ * \param type  type of parameter, such as 
+ * \param name  the parameter name, will be duplicated/copied!
+ * \param size  number of elements in 'values' vector (1..4, or more)
+ * \param datatype  GL_FLOAT, GL_FLOAT_VECx, GL_INT, GL_INT_VECx or GL_NONE.
+ * \param values  initial parameter value, up to 4 gl_constant_values, or NULL
+ * \param state  state indexes, or NULL
+ * \return  index of new parameter in the list, or -1 if error (out of mem)
+ */
+GLint
+_mesa_add_parameter(struct gl_program_parameter_list *paramList,
+                    gl_register_file type, const char *name,
+                    GLuint size, GLenum datatype,
+                    const gl_constant_value *values,
+                    const gl_state_index state[STATE_LENGTH],
+                    GLbitfield flags)
+{
+   const GLuint oldNum = paramList->NumParameters;
+   const GLuint sz4 = (size + 3) / 4; /* no. of new param slots needed */
+
+   assert(size > 0);
+
+   if (oldNum + sz4 > paramList->Size) {
+      /* Need to grow the parameter list array (alloc some extra) */
+      paramList->Size = paramList->Size + 4 * sz4;
+
+      /* realloc arrays */
+      paramList->Parameters = (struct gl_program_parameter *)
+	 _mesa_realloc(paramList->Parameters,
+		       oldNum * sizeof(struct gl_program_parameter),
+		       paramList->Size * sizeof(struct gl_program_parameter));
+
+      paramList->ParameterValues = (gl_constant_value (*)[4])
+         _mesa_align_realloc(paramList->ParameterValues,         /* old buf */
+                             oldNum * 4 * sizeof(gl_constant_value),/* old sz */
+                             paramList->Size*4*sizeof(gl_constant_value),/*new*/
+                             16);
+   }
+
+   if (!paramList->Parameters ||
+       !paramList->ParameterValues) {
+      /* out of memory */
+      paramList->NumParameters = 0;
+      paramList->Size = 0;
+      return -1;
+   }
+   else {
+      GLuint i, j;
+
+      paramList->NumParameters = oldNum + sz4;
+
+      memset(&paramList->Parameters[oldNum], 0,
+             sz4 * sizeof(struct gl_program_parameter));
+
+      for (i = 0; i < sz4; i++) {
+         struct gl_program_parameter *p = paramList->Parameters + oldNum + i;
+         p->Name = name ? _mesa_strdup(name) : NULL;
+         p->Type = type;
+         p->Size = size;
+         p->DataType = datatype;
+         p->Flags = flags;
+         if (values) {
+            COPY_4V(paramList->ParameterValues[oldNum + i], values);
+            values += 4;
+            p->Initialized = GL_TRUE;
+         }
+         else {
+            /* silence valgrind */
+            for (j = 0; j < 4; j++)
+            	paramList->ParameterValues[oldNum + i][j].f = 0;
+         }
+         size -= 4;
+      }
+
+      if (state) {
+         for (i = 0; i < STATE_LENGTH; i++)
+            paramList->Parameters[oldNum].StateIndexes[i] = state[i];
+      }
+
+      return (GLint) oldNum;
+   }
+}
+
+
+/**
+ * Add a new named program parameter (Ex: NV_fragment_program DEFINE statement)
+ * \return index of the new entry in the parameter list
+ */
+GLint
+_mesa_add_named_parameter(struct gl_program_parameter_list *paramList,
+                          const char *name, const gl_constant_value values[4])
+{
+   return _mesa_add_parameter(paramList, PROGRAM_NAMED_PARAM, name,
+                              4, GL_NONE, values, NULL, 0x0);
+                              
+}
+
+
+/**
+ * Add a new named constant to the parameter list.
+ * This will be used when the program contains something like this:
+ *    PARAM myVals = { 0, 1, 2, 3 };
+ *
+ * \param paramList  the parameter list
+ * \param name  the name for the constant
+ * \param values  four float values
+ * \return index/position of the new parameter in the parameter list
+ */
+GLint
+_mesa_add_named_constant(struct gl_program_parameter_list *paramList,
+                         const char *name, const gl_constant_value values[4],
+                         GLuint size)
+{
+   /* first check if this is a duplicate constant */
+   GLint pos;
+   for (pos = 0; pos < (GLint)paramList->NumParameters; pos++) {
+      const gl_constant_value *pvals = paramList->ParameterValues[pos];
+      if (pvals[0].u == values[0].u &&
+          pvals[1].u == values[1].u &&
+          pvals[2].u == values[2].u &&
+          pvals[3].u == values[3].u &&
+          strcmp(paramList->Parameters[pos].Name, name) == 0) {
+         /* Same name and value is already in the param list - reuse it */
+         return pos;
+      }
+   }
+   /* not found, add new parameter */
+   return _mesa_add_parameter(paramList, PROGRAM_CONSTANT, name,
+                              size, GL_NONE, values, NULL, 0x0);
+}
+
+
+/**
+ * Add a new unnamed constant to the parameter list.  This will be used
+ * when a fragment/vertex program contains something like this:
+ *    MOV r, { 0, 1, 2, 3 };
+ * If swizzleOut is non-null we'll search the parameter list for an
+ * existing instance of the constant which matches with a swizzle.
+ *
+ * \param paramList  the parameter list
+ * \param values  four float values
+ * \param swizzleOut  returns swizzle mask for accessing the constant
+ * \return index/position of the new parameter in the parameter list.
+ */
+GLint
+_mesa_add_typed_unnamed_constant(struct gl_program_parameter_list *paramList,
+                           const gl_constant_value values[4], GLuint size,
+                           GLenum datatype, GLuint *swizzleOut)
+{
+   GLint pos;
+   ASSERT(size >= 1);
+   ASSERT(size <= 4);
+
+   if (swizzleOut &&
+       _mesa_lookup_parameter_constant(paramList, values,
+                                       size, &pos, swizzleOut)) {
+      return pos;
+   }
+
+   /* Look for empty space in an already unnamed constant parameter
+    * to add this constant.  This will only work for single-element
+    * constants because we rely on smearing (i.e. .yyyy or .zzzz).
+    */
+   if (size == 1 && swizzleOut) {
+      for (pos = 0; pos < (GLint) paramList->NumParameters; pos++) {
+         struct gl_program_parameter *p = paramList->Parameters + pos;
+         if (p->Type == PROGRAM_CONSTANT && p->Size + size <= 4) {
+            /* ok, found room */
+            gl_constant_value *pVal = paramList->ParameterValues[pos];
+            GLuint swz = p->Size; /* 1, 2 or 3 for Y, Z, W */
+            pVal[p->Size] = values[0];
+            p->Size++;
+            *swizzleOut = MAKE_SWIZZLE4(swz, swz, swz, swz);
+            return pos;
+         }
+      }
+   }
+
+   /* add a new parameter to store this constant */
+   pos = _mesa_add_parameter(paramList, PROGRAM_CONSTANT, NULL,
+                             size, datatype, values, NULL, 0x0);
+   if (pos >= 0 && swizzleOut) {
+      if (size == 1)
+         *swizzleOut = SWIZZLE_XXXX;
+      else
+         *swizzleOut = SWIZZLE_NOOP;
+   }
+   return pos;
+}
+
+/**
+ * Add a new unnamed constant to the parameter list.  This will be used
+ * when a fragment/vertex program contains something like this:
+ *    MOV r, { 0, 1, 2, 3 };
+ * If swizzleOut is non-null we'll search the parameter list for an
+ * existing instance of the constant which matches with a swizzle.
+ *
+ * \param paramList  the parameter list
+ * \param values  four float values
+ * \param swizzleOut  returns swizzle mask for accessing the constant
+ * \return index/position of the new parameter in the parameter list.
+ * \sa _mesa_add_typed_unnamed_constant
+ */
+GLint
+_mesa_add_unnamed_constant(struct gl_program_parameter_list *paramList,
+                           const gl_constant_value values[4], GLuint size,
+                           GLuint *swizzleOut)
+{
+   return _mesa_add_typed_unnamed_constant(paramList, values, size, GL_NONE,
+                                           swizzleOut);
+}
+
+/**
+ * Add parameter representing a varying variable.
+ */
+GLint
+_mesa_add_varying(struct gl_program_parameter_list *paramList,
+                  const char *name, GLuint size, GLenum datatype,
+                  GLbitfield flags)
+{
+   GLint i = _mesa_lookup_parameter_index(paramList, -1, name);
+   if (i >= 0 && paramList->Parameters[i].Type == PROGRAM_VARYING) {
+      /* already in list */
+      return i;
+   }
+   else {
+      /*assert(size == 4);*/
+      i = _mesa_add_parameter(paramList, PROGRAM_VARYING, name,
+                              size, datatype, NULL, NULL, flags);
+      return i;
+   }
+}
+
+
+/**
+ * Add parameter representing a vertex program attribute.
+ * \param size  size of attribute (in floats), may be -1 if unknown
+ * \param attrib  the attribute index, or -1 if unknown
+ */
+GLint
+_mesa_add_attribute(struct gl_program_parameter_list *paramList,
+                    const char *name, GLint size, GLenum datatype, GLint attrib)
+{
+   GLint i = _mesa_lookup_parameter_index(paramList, -1, name);
+   if (i >= 0) {
+      /* replace */
+      if (attrib < 0)
+         attrib = i;
+      paramList->Parameters[i].StateIndexes[0] = attrib;
+   }
+   else {
+      /* add */
+      gl_state_index state[STATE_LENGTH];
+      state[0] = (gl_state_index) attrib;
+      if (size < 0)
+         size = 4;
+      i = _mesa_add_parameter(paramList, PROGRAM_INPUT, name,
+                              size, datatype, NULL, state, 0x0);
+   }
+   return i;
+}
+
+
+
+#if 0 /* not used yet */
+/**
+ * Returns the number of 4-component registers needed to store a piece
+ * of GL state.  For matrices this may be as many as 4 registers,
+ * everything else needs
+ * just 1 register.
+ */
+static GLuint
+sizeof_state_reference(const GLint *stateTokens)
+{
+   if (stateTokens[0] == STATE_MATRIX) {
+      GLuint rows = stateTokens[4] - stateTokens[3] + 1;
+      assert(rows >= 1);
+      assert(rows <= 4);
+      return rows;
+   }
+   else {
+      return 1;
+   }
+}
+#endif
+
+
+/**
+ * Add a new state reference to the parameter list.
+ * This will be used when the program contains something like this:
+ *    PARAM ambient = state.material.front.ambient;
+ *
+ * \param paramList  the parameter list
+ * \param stateTokens  an array of 5 (STATE_LENGTH) state tokens
+ * \return index of the new parameter.
+ */
+GLint
+_mesa_add_state_reference(struct gl_program_parameter_list *paramList,
+                          const gl_state_index stateTokens[STATE_LENGTH])
+{
+   const GLuint size = 4; /* XXX fix */
+   char *name;
+   GLint index;
+
+   /* Check if the state reference is already in the list */
+   for (index = 0; index < (GLint) paramList->NumParameters; index++) {
+      if (!memcmp(paramList->Parameters[index].StateIndexes,
+		  stateTokens, STATE_LENGTH * sizeof(gl_state_index))) {
+	 return index;
+      }
+   }
+
+   name = _mesa_program_state_string(stateTokens);
+   index = _mesa_add_parameter(paramList, PROGRAM_STATE_VAR, name,
+                               size, GL_NONE,
+                               NULL, (gl_state_index *) stateTokens, 0x0);
+   paramList->StateFlags |= _mesa_program_state_flags(stateTokens);
+
+   /* free name string here since we duplicated it in add_parameter() */
+   free(name);
+
+   return index;
+}
+
+
+/**
+ * Lookup a parameter value by name in the given parameter list.
+ * \return pointer to the float[4] values.
+ */
+gl_constant_value *
+_mesa_lookup_parameter_value(const struct gl_program_parameter_list *paramList,
+                             GLsizei nameLen, const char *name)
+{
+   GLint i = _mesa_lookup_parameter_index(paramList, nameLen, name);
+   if (i < 0)
+      return NULL;
+   else
+      return paramList->ParameterValues[i];
+}
+
+
+/**
+ * Given a program parameter name, find its position in the list of parameters.
+ * \param paramList  the parameter list to search
+ * \param nameLen  length of name (in chars).
+ *                 If length is negative, assume that name is null-terminated.
+ * \param name  the name to search for
+ * \return index of parameter in the list.
+ */
+GLint
+_mesa_lookup_parameter_index(const struct gl_program_parameter_list *paramList,
+                             GLsizei nameLen, const char *name)
+{
+   GLint i;
+
+   if (!paramList)
+      return -1;
+
+   if (nameLen == -1) {
+      /* name is null-terminated */
+      for (i = 0; i < (GLint) paramList->NumParameters; i++) {
+         if (paramList->Parameters[i].Name &&
+	     strcmp(paramList->Parameters[i].Name, name) == 0)
+            return i;
+      }
+   }
+   else {
+      /* name is not null-terminated, use nameLen */
+      for (i = 0; i < (GLint) paramList->NumParameters; i++) {
+         if (paramList->Parameters[i].Name &&
+	     strncmp(paramList->Parameters[i].Name, name, nameLen) == 0
+             && strlen(paramList->Parameters[i].Name) == (size_t)nameLen)
+            return i;
+      }
+   }
+   return -1;
+}
+
+
+/**
+ * Look for a float vector in the given parameter list.  The float vector
+ * may be of length 1, 2, 3 or 4.  If swizzleOut is non-null, we'll try
+ * swizzling to find a match.
+ * \param list  the parameter list to search
+ * \param v  the float vector to search for
+ * \param vSize  number of element in v
+ * \param posOut  returns the position of the constant, if found
+ * \param swizzleOut  returns a swizzle mask describing location of the
+ *                    vector elements if found.
+ * \return GL_TRUE if found, GL_FALSE if not found
+ */
+GLboolean
+_mesa_lookup_parameter_constant(const struct gl_program_parameter_list *list,
+                                const gl_constant_value v[], GLuint vSize,
+                                GLint *posOut, GLuint *swizzleOut)
+{
+   GLuint i;
+
+   assert(vSize >= 1);
+   assert(vSize <= 4);
+
+   if (!list) {
+      *posOut = -1;
+      return GL_FALSE;
+   }
+
+   for (i = 0; i < list->NumParameters; i++) {
+      if (list->Parameters[i].Type == PROGRAM_CONSTANT) {
+         if (!swizzleOut) {
+            /* swizzle not allowed */
+            GLuint j, match = 0;
+            for (j = 0; j < vSize; j++) {
+               if (v[j].u == list->ParameterValues[i][j].u)
+                  match++;
+            }
+            if (match == vSize) {
+               *posOut = i;
+               return GL_TRUE;
+            }
+         }
+         else {
+            /* try matching w/ swizzle */
+             if (vSize == 1) {
+                /* look for v[0] anywhere within float[4] value */
+                GLuint j;
+                for (j = 0; j < list->Parameters[i].Size; j++) {
+                   if (list->ParameterValues[i][j].u == v[0].u) {
+                      /* found it */
+                      *posOut = i;
+                      *swizzleOut = MAKE_SWIZZLE4(j, j, j, j);
+                      return GL_TRUE;
+                   }
+                }
+             }
+             else if (vSize <= list->Parameters[i].Size) {
+                /* see if we can match this constant (with a swizzle) */
+                GLuint swz[4];
+                GLuint match = 0, j, k;
+                for (j = 0; j < vSize; j++) {
+                   if (v[j].u == list->ParameterValues[i][j].u) {
+                      swz[j] = j;
+                      match++;
+                   }
+                   else {
+                      for (k = 0; k < list->Parameters[i].Size; k++) {
+                         if (v[j].u == list->ParameterValues[i][k].u) {
+                            swz[j] = k;
+                            match++;
+                            break;
+                         }
+                      }
+                   }
+                }
+                /* smear last value to remaining positions */
+                for (; j < 4; j++)
+                   swz[j] = swz[j-1];
+
+                if (match == vSize) {
+                   *posOut = i;
+                   *swizzleOut = MAKE_SWIZZLE4(swz[0], swz[1], swz[2], swz[3]);
+                   return GL_TRUE;
+                }
+             }
+         }
+      }
+   }
+
+   *posOut = -1;
+   return GL_FALSE;
+}
+
+
+struct gl_program_parameter_list *
+_mesa_clone_parameter_list(const struct gl_program_parameter_list *list)
+{
+   struct gl_program_parameter_list *clone;
+   GLuint i;
+
+   clone = _mesa_new_parameter_list();
+   if (!clone)
+      return NULL;
+
+   /** Not too efficient, but correct */
+   for (i = 0; i < list->NumParameters; i++) {
+      struct gl_program_parameter *p = list->Parameters + i;
+      struct gl_program_parameter *pCopy;
+      GLuint size = MIN2(p->Size, 4);
+      GLint j = _mesa_add_parameter(clone, p->Type, p->Name, size, p->DataType,
+                                    list->ParameterValues[i], NULL, 0x0);
+      ASSERT(j >= 0);
+      pCopy = clone->Parameters + j;
+      pCopy->Flags = p->Flags;
+      /* copy state indexes */
+      if (p->Type == PROGRAM_STATE_VAR) {
+         GLint k;
+         for (k = 0; k < STATE_LENGTH; k++) {
+            pCopy->StateIndexes[k] = p->StateIndexes[k];
+         }
+      }
+      else {
+         clone->Parameters[j].Size = p->Size;
+      }
+      
+   }
+
+   clone->StateFlags = list->StateFlags;
+
+   return clone;
+}
+
+
+/**
+ * Return a new parameter list which is listA + listB.
+ */
+struct gl_program_parameter_list *
+_mesa_combine_parameter_lists(const struct gl_program_parameter_list *listA,
+                              const struct gl_program_parameter_list *listB)
+{
+   struct gl_program_parameter_list *list;
+
+   if (listA) {
+      list = _mesa_clone_parameter_list(listA);
+      if (list && listB) {
+         GLuint i;
+         for (i = 0; i < listB->NumParameters; i++) {
+            struct gl_program_parameter *param = listB->Parameters + i;
+            _mesa_add_parameter(list, param->Type, param->Name, param->Size,
+                                param->DataType,
+                                listB->ParameterValues[i],
+                                param->StateIndexes,
+                                param->Flags);
+         }
+      }
+   }
+   else if (listB) {
+      list = _mesa_clone_parameter_list(listB);
+   }
+   else {
+      list = NULL;
+   }
+   return list;
+}
+
+
+
+/**
+ * Find longest name of all uniform parameters in list.
+ */
+GLuint
+_mesa_longest_parameter_name(const struct gl_program_parameter_list *list,
+                             gl_register_file type)
+{
+   GLuint i, maxLen = 0;
+   if (!list)
+      return 0;
+   for (i = 0; i < list->NumParameters; i++) {
+      if (list->Parameters[i].Type == type) {
+         GLuint len = strlen(list->Parameters[i].Name);
+         if (len > maxLen)
+            maxLen = len;
+      }
+   }
+   return maxLen;
+}
+
+
+/**
+ * Count the number of parameters in the last that match the given type.
+ */
+GLuint
+_mesa_num_parameters_of_type(const struct gl_program_parameter_list *list,
+                             gl_register_file type)
+{
+   GLuint i, count = 0;
+   if (list) {
+      for (i = 0; i < list->NumParameters; i++) {
+         if (list->Parameters[i].Type == type)
+            count++;
+      }
+   }
+   return count;
+}
diff --git a/mesalib/src/mesa/program/prog_parameter.h b/mesalib/src/mesa/program/prog_parameter.h
index 10cbbe57a..f858cf0fa 100644
--- a/mesalib/src/mesa/program/prog_parameter.h
+++ b/mesalib/src/mesa/program/prog_parameter.h
@@ -46,7 +46,15 @@
 #define PROG_PARAM_BIT_CYL_WRAP  0x10  /**< XXX gallium debug */
 /*@}*/
 
-
+/**
+ * Actual data for constant values of parameters.
+ */
+typedef union gl_constant_value {
+	GLfloat f;
+	GLboolean b;
+	GLint i;
+	GLuint u;
+} gl_constant_value;
 
 /**
  * Program parameter.
@@ -81,7 +89,7 @@ struct gl_program_parameter_list
    GLuint Size;           /**< allocated size of Parameters, ParameterValues */
    GLuint NumParameters;  /**< number of parameters in arrays */
    struct gl_program_parameter *Parameters; /**< Array [Size] */
-   GLfloat (*ParameterValues)[4];        /**< Array [Size] of GLfloat[4] */
+   gl_constant_value (*ParameterValues)[4]; /**< Array [Size] of constant[4] */
    GLbitfield StateFlags; /**< _NEW_* flags indicating which state changes
                                might invalidate ParameterValues[] */
 };
@@ -112,22 +120,28 @@ _mesa_num_parameters(const struct gl_program_parameter_list *list)
 extern GLint
 _mesa_add_parameter(struct gl_program_parameter_list *paramList,
                     gl_register_file type, const char *name,
-                    GLuint size, GLenum datatype, const GLfloat *values,
+                    GLuint size, GLenum datatype,
+                    const gl_constant_value *values,
                     const gl_state_index state[STATE_LENGTH],
                     GLbitfield flags);
 
 extern GLint
 _mesa_add_named_parameter(struct gl_program_parameter_list *paramList,
-                          const char *name, const GLfloat values[4]);
+                          const char *name, const gl_constant_value values[4]);
 
 extern GLint
 _mesa_add_named_constant(struct gl_program_parameter_list *paramList,
-                         const char *name, const GLfloat values[4],
+                         const char *name, const gl_constant_value values[4],
                          GLuint size);
 
 extern GLint
+_mesa_add_typed_unnamed_constant(struct gl_program_parameter_list *paramList,
+                           const gl_constant_value values[4], GLuint size,
+                           GLenum datatype, GLuint *swizzleOut);
+
+extern GLint
 _mesa_add_unnamed_constant(struct gl_program_parameter_list *paramList,
-                           const GLfloat values[4], GLuint size,
+                           const gl_constant_value values[4], GLuint size,
                            GLuint *swizzleOut);
 
 extern GLint
@@ -143,7 +157,7 @@ extern GLint
 _mesa_add_state_reference(struct gl_program_parameter_list *paramList,
                           const gl_state_index stateTokens[STATE_LENGTH]);
 
-extern GLfloat *
+extern gl_constant_value *
 _mesa_lookup_parameter_value(const struct gl_program_parameter_list *paramList,
                              GLsizei nameLen, const char *name);
 
@@ -153,7 +167,7 @@ _mesa_lookup_parameter_index(const struct gl_program_parameter_list *paramList,
 
 extern GLboolean
 _mesa_lookup_parameter_constant(const struct gl_program_parameter_list *list,
-                                const GLfloat v[], GLuint vSize,
+                                const gl_constant_value v[], GLuint vSize,
                                 GLint *posOut, GLuint *swizzleOut);
 
 extern GLuint
diff --git a/mesalib/src/mesa/program/prog_parameter_layout.c b/mesalib/src/mesa/program/prog_parameter_layout.c
index 90a977108..28fca3b92 100644
--- a/mesalib/src/mesa/program/prog_parameter_layout.c
+++ b/mesalib/src/mesa/program/prog_parameter_layout.c
@@ -182,7 +182,7 @@ _mesa_layout_parameters(struct asm_parser_state *state)
 
 	 switch (p->Type) {
 	 case PROGRAM_CONSTANT: {
-	    const float *const v =
+	    const gl_constant_value *const v =
 	       state->prog->Parameters->ParameterValues[idx];
 
 	    inst->Base.SrcReg[i].Index =
diff --git a/mesalib/src/mesa/program/prog_print.c b/mesalib/src/mesa/program/prog_print.c
index 248f4d730..2cfec13ec 100644
--- a/mesalib/src/mesa/program/prog_print.c
+++ b/mesalib/src/mesa/program/prog_print.c
@@ -985,7 +985,7 @@ _mesa_fprint_parameter_list(FILE *f,
    fprintf(f, "dirty state flags: 0x%x\n", list->StateFlags);
    for (i = 0; i < list->NumParameters; i++){
       struct gl_program_parameter *param = list->Parameters + i;
-      const GLfloat *v = list->ParameterValues[i];
+      const GLfloat *v = (GLfloat *) list->ParameterValues[i];
       fprintf(f, "param[%d] sz=%d %s %s = {%.3g, %.3g, %.3g, %.3g}",
 	      i, param->Size,
 	      _mesa_register_file_name(list->Parameters[i].Type),
diff --git a/mesalib/src/mesa/program/prog_statevars.c b/mesalib/src/mesa/program/prog_statevars.c
index ddda78e19..c68e1643a 100644
--- a/mesalib/src/mesa/program/prog_statevars.c
+++ b/mesalib/src/mesa/program/prog_statevars.c
@@ -1111,7 +1111,7 @@ _mesa_load_state_parameters(struct gl_context *ctx,
       if (paramList->Parameters[i].Type == PROGRAM_STATE_VAR) {
          _mesa_fetch_state(ctx,
 			   paramList->Parameters[i].StateIndexes,
-                           paramList->ParameterValues[i]);
+                           &paramList->ParameterValues[i][0].f);
       }
    }
 }
diff --git a/mesalib/src/mesa/program/program.c b/mesalib/src/mesa/program/program.c
index 2393a6386..5a9bc0fae 100644
--- a/mesalib/src/mesa/program/program.c
+++ b/mesalib/src/mesa/program/program.c
@@ -388,8 +388,9 @@ _mesa_delete_program(struct gl_context *ctx, struct gl_program *prog)
    if (prog->String)
       free(prog->String);
 
-   _mesa_free_instructions(prog->Instructions, prog->NumInstructions);
-
+   if (prog->Instructions) {
+      _mesa_free_instructions(prog->Instructions, prog->NumInstructions);
+   }
    if (prog->Parameters) {
       _mesa_free_parameter_list(prog->Parameters);
    }
@@ -1031,7 +1032,8 @@ _mesa_postprocess_program(struct gl_context *ctx, struct gl_program *prog)
    GLuint i;
    GLuint whiteSwizzle;
    GLint whiteIndex = _mesa_add_unnamed_constant(prog->Parameters,
-                                                 white, 4, &whiteSwizzle);
+                                                 (gl_constant_value *) white,
+                                                 4, &whiteSwizzle);
 
    (void) whiteIndex;
 
diff --git a/mesalib/src/mesa/program/program_parse.y b/mesalib/src/mesa/program/program_parse.y
index e7752fcbd..4efdc0184 100644
--- a/mesalib/src/mesa/program/program_parse.y
+++ b/mesalib/src/mesa/program/program_parse.y
@@ -1854,64 +1854,64 @@ paramConstUse: paramConstScalarUse | paramConstVector;
 paramConstScalarDecl: signedFloatConstant
 	{
 	   $$.count = 4;
-	   $$.data[0] = $1;
-	   $$.data[1] = $1;
-	   $$.data[2] = $1;
-	   $$.data[3] = $1;
+	   $$.data[0].f = $1;
+	   $$.data[1].f = $1;
+	   $$.data[2].f = $1;
+	   $$.data[3].f = $1;
 	}
 	;
 
 paramConstScalarUse: REAL
 	{
 	   $$.count = 1;
-	   $$.data[0] = $1;
-	   $$.data[1] = $1;
-	   $$.data[2] = $1;
-	   $$.data[3] = $1;
+	   $$.data[0].f = $1;
+	   $$.data[1].f = $1;
+	   $$.data[2].f = $1;
+	   $$.data[3].f = $1;
 	}
 	| INTEGER
 	{
 	   $$.count = 1;
-	   $$.data[0] = (float) $1;
-	   $$.data[1] = (float) $1;
-	   $$.data[2] = (float) $1;
-	   $$.data[3] = (float) $1;
+	   $$.data[0].f = (float) $1;
+	   $$.data[1].f = (float) $1;
+	   $$.data[2].f = (float) $1;
+	   $$.data[3].f = (float) $1;
 	}
 	;
 
 paramConstVector: '{' signedFloatConstant '}'
 	{
 	   $$.count = 4;
-	   $$.data[0] = $2;
-	   $$.data[1] = 0.0f;
-	   $$.data[2] = 0.0f;
-	   $$.data[3] = 1.0f;
+	   $$.data[0].f = $2;
+	   $$.data[1].f = 0.0f;
+	   $$.data[2].f = 0.0f;
+	   $$.data[3].f = 1.0f;
 	}
 	| '{' signedFloatConstant ',' signedFloatConstant '}'
 	{
 	   $$.count = 4;
-	   $$.data[0] = $2;
-	   $$.data[1] = $4;
-	   $$.data[2] = 0.0f;
-	   $$.data[3] = 1.0f;
+	   $$.data[0].f = $2;
+	   $$.data[1].f = $4;
+	   $$.data[2].f = 0.0f;
+	   $$.data[3].f = 1.0f;
 	}
 	| '{' signedFloatConstant ',' signedFloatConstant ','
               signedFloatConstant '}'
 	{
 	   $$.count = 4;
-	   $$.data[0] = $2;
-	   $$.data[1] = $4;
-	   $$.data[2] = $6;
-	   $$.data[3] = 1.0f;
+	   $$.data[0].f = $2;
+	   $$.data[1].f = $4;
+	   $$.data[2].f = $6;
+	   $$.data[3].f = 1.0f;
 	}
 	| '{' signedFloatConstant ',' signedFloatConstant ','
               signedFloatConstant ',' signedFloatConstant '}'
 	{
 	   $$.count = 4;
-	   $$.data[0] = $2;
-	   $$.data[1] = $4;
-	   $$.data[2] = $6;
-	   $$.data[3] = $8;
+	   $$.data[0].f = $2;
+	   $$.data[1].f = $4;
+	   $$.data[2].f = $6;
+	   $$.data[3].f = $8;
 	}
 	;
 
diff --git a/mesalib/src/mesa/program/program_parser.h b/mesalib/src/mesa/program/program_parser.h
index cbbb2677d..b6853a229 100644
--- a/mesalib/src/mesa/program/program_parser.h
+++ b/mesalib/src/mesa/program/program_parser.h
@@ -23,6 +23,7 @@
 #pragma once
 
 #include "main/config.h"
+#include "program/prog_parameter.h"
 
 struct gl_context;
 
@@ -96,7 +97,7 @@ struct asm_symbol {
 
 struct asm_vector {
    unsigned count;
-   float    data[4];
+   gl_constant_value data[4];
 };
 
 
diff --git a/mesalib/src/mesa/program/sampler.cpp b/mesalib/src/mesa/program/sampler.cpp
index b219d7016..9a2616743 100644
--- a/mesalib/src/mesa/program/sampler.cpp
+++ b/mesalib/src/mesa/program/sampler.cpp
@@ -132,6 +132,6 @@ _mesa_get_sampler_uniform_value(class ir_dereference *sampler,
 
    index += getname.offset;
 
-   return prog->Parameters->ParameterValues[index][0];
+   return prog->Parameters->ParameterValues[index][0].f;
 }
 }