Merge branch 'master' of git://anongit.freedesktop.org/mesa/mesa

17 files changed, 872 insertions, 160 deletions

diff --git a/src/mesa/program/ir_to_mesa.cpp b/src/mesa/program/ir_to_mesa.cpp
index d8e5a3a9772..6820e4c6ba7 100644
--- a/src/mesa/program/ir_to_mesa.cpp
+++ b/src/mesa/program/ir_to_mesa.cpp
@@ -134,7 +134,7 @@ src_reg::src_reg(dst_reg reg)
    this->index = reg.index;
    this->swizzle = SWIZZLE_XYZW;
    this->negate = 0;
-   this->reladdr = NULL;
+   this->reladdr = reg.reladdr;
 }
 
 dst_reg::dst_reg(src_reg reg)
@@ -297,11 +297,11 @@ public:
    /**
     * Emit the correct dot-product instruction for the type of arguments
     */
-   void emit_dp(ir_instruction *ir,
-	        dst_reg dst,
-	        src_reg src0,
-	        src_reg src1,
-	        unsigned elements);
+   ir_to_mesa_instruction * emit_dp(ir_instruction *ir,
+				    dst_reg dst,
+				    src_reg src0,
+				    src_reg src1,
+				    unsigned elements);
 
    void emit_scalar(ir_instruction *ir, enum prog_opcode op,
 		    dst_reg dst, src_reg src0);
@@ -312,9 +312,11 @@ public:
    void emit_scs(ir_instruction *ir, enum prog_opcode op,
 		 dst_reg dst, const src_reg &src);
 
-   GLboolean try_emit_mad(ir_expression *ir,
+   bool try_emit_mad(ir_expression *ir,
 			  int mul_operand);
-   GLboolean try_emit_sat(ir_expression *ir);
+   bool try_emit_mad_for_and_not(ir_expression *ir,
+				 int mul_operand);
+   bool try_emit_sat(ir_expression *ir);
 
    void emit_swz(ir_expression *ir);
 
@@ -331,20 +333,6 @@ dst_reg undef_dst = dst_reg(PROGRAM_UNDEFINED, SWIZZLE_NOOP);
 
 dst_reg address_reg = dst_reg(PROGRAM_ADDRESS, WRITEMASK_X);
 
-static void
-fail_link(struct gl_shader_program *prog, const char *fmt, ...) PRINTFLIKE(2, 3);
-
-static void
-fail_link(struct gl_shader_program *prog, const char *fmt, ...)
-{
-   va_list args;
-   va_start(args, fmt);
-   ralloc_vasprintf_append(&prog->InfoLog, fmt, args);
-   va_end(args);
-
-   prog->LinkStatus = GL_FALSE;
-}
-
 static int
 swizzle_for_size(int size)
 {
@@ -422,7 +410,7 @@ ir_to_mesa_visitor::emit(ir_instruction *ir, enum prog_opcode op)
    return emit(ir, op, undef_dst, undef_src, undef_src, undef_src);
 }
 
-void
+ir_to_mesa_instruction *
 ir_to_mesa_visitor::emit_dp(ir_instruction *ir,
 			    dst_reg dst, src_reg src0, src_reg src1,
 			    unsigned elements)
@@ -431,7 +419,7 @@ ir_to_mesa_visitor::emit_dp(ir_instruction *ir,
       OPCODE_DP2, OPCODE_DP3, OPCODE_DP4
    };
 
-   emit(ir, dot_opcodes[elements - 2], dst, src0, src1);
+   return emit(ir, dot_opcodes[elements - 2], dst, src0, src1);
 }
 
 /**
@@ -593,13 +581,13 @@ ir_to_mesa_visitor::emit_scs(ir_instruction *ir, enum prog_opcode op,
    }
 }
 
-struct src_reg
+src_reg
 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;
 }
@@ -655,8 +643,6 @@ src_reg
 ir_to_mesa_visitor::get_temp(const glsl_type *type)
 {
    src_reg src;
-   int swizzle[4];
-   int i;
 
    src.file = PROGRAM_TEMPORARY;
    src.index = next_temp;
@@ -666,12 +652,7 @@ ir_to_mesa_visitor::get_temp(const glsl_type *type)
    if (type->is_array() || type->is_record()) {
       src.swizzle = SWIZZLE_NOOP;
    } else {
-      for (i = 0; i < type->vector_elements; i++)
-	 swizzle[i] = i;
-      for (; i < 4; i++)
-	 swizzle[i] = type->vector_elements - 1;
-      src.swizzle = MAKE_SWIZZLE4(swizzle[0], swizzle[1],
-				  swizzle[2], swizzle[3]);
+      src.swizzle = swizzle_for_size(type->vector_elements);
    }
    src.negate = 0;
 
@@ -744,7 +725,7 @@ ir_to_mesa_visitor::visit(ir_variable *ir)
 	 }
       }
 
-      struct variable_storage *storage;
+      variable_storage *storage;
       dst_reg dst;
       if (i == ir->num_state_slots) {
 	 /* We'll set the index later. */
@@ -789,10 +770,11 @@ ir_to_mesa_visitor::visit(ir_variable *ir)
 
       if (storage->file == PROGRAM_TEMPORARY &&
 	  dst.index != storage->index + (int) ir->num_state_slots) {
-	 fail_link(this->shader_program,
-		   "failed to load builtin uniform `%s'  (%d/%d regs loaded)\n",
-		   ir->name, dst.index - storage->index,
-		   type_size(ir->type));
+	 linker_error(this->shader_program,
+		      "failed to load builtin uniform `%s' "
+		      "(%d/%d regs loaded)\n",
+		      ir->name, dst.index - storage->index,
+		      type_size(ir->type));
       }
    }
 }
@@ -889,7 +871,7 @@ ir_to_mesa_visitor::visit(ir_function *ir)
    }
 }
 
-GLboolean
+bool
 ir_to_mesa_visitor::try_emit_mad(ir_expression *ir, int mul_operand)
 {
    int nonmul_operand = 1 - mul_operand;
@@ -912,7 +894,47 @@ ir_to_mesa_visitor::try_emit_mad(ir_expression *ir, int mul_operand)
    return true;
 }
 
-GLboolean
+/**
+ * Emit OPCODE_MAD(a, -b, a) instead of AND(a, NOT(b))
+ *
+ * The logic values are 1.0 for true and 0.0 for false.  Logical-and is
+ * implemented using multiplication, and logical-or is implemented using
+ * addition.  Logical-not can be implemented as (true - x), or (1.0 - x).
+ * As result, the logical expression (a & !b) can be rewritten as:
+ *
+ *     - a * !b
+ *     - a * (1 - b)
+ *     - (a * 1) - (a * b)
+ *     - a + -(a * b)
+ *     - a + (a * -b)
+ *
+ * This final expression can be implemented as a single MAD(a, -b, a)
+ * instruction.
+ */
+bool
+ir_to_mesa_visitor::try_emit_mad_for_and_not(ir_expression *ir, int try_operand)
+{
+   const int other_operand = 1 - try_operand;
+   src_reg a, b;
+
+   ir_expression *expr = ir->operands[try_operand]->as_expression();
+   if (!expr || expr->operation != ir_unop_logic_not)
+      return false;
+
+   ir->operands[other_operand]->accept(this);
+   a = this->result;
+   expr->operands[0]->accept(this);
+   b = this->result;
+
+   b.negate = ~b.negate;
+
+   this->result = get_temp(ir->type);
+   emit(ir, OPCODE_MAD, dst_reg(this->result), a, b, a);
+
+   return true;
+}
+
+bool
 ir_to_mesa_visitor::try_emit_sat(ir_expression *ir)
 {
    /* Saturates were only introduced to vertex programs in
@@ -928,10 +950,30 @@ ir_to_mesa_visitor::try_emit_sat(ir_expression *ir)
    sat_src->accept(this);
    src_reg src = this->result;
 
-   this->result = get_temp(ir->type);
-   ir_to_mesa_instruction *inst;
-   inst = emit(ir, OPCODE_MOV, dst_reg(this->result), src);
-   inst->saturate = true;
+   /* If we generated an expression instruction into a temporary in
+    * processing the saturate's operand, apply the saturate to that
+    * instruction.  Otherwise, generate a MOV to do the saturate.
+    *
+    * Note that we have to be careful to only do this optimization if
+    * the instruction in question was what generated src->result.  For
+    * example, ir_dereference_array might generate a MUL instruction
+    * to create the reladdr, and return us a src reg using that
+    * reladdr.  That MUL result is not the value we're trying to
+    * saturate.
+    */
+   ir_expression *sat_src_expr = sat_src->as_expression();
+   ir_to_mesa_instruction *new_inst;
+   new_inst = (ir_to_mesa_instruction *)this->instructions.get_tail();
+   if (sat_src_expr && (sat_src_expr->operation == ir_binop_mul ||
+			sat_src_expr->operation == ir_binop_add ||
+			sat_src_expr->operation == ir_binop_dot)) {
+      new_inst->saturate = true;
+   } else {
+      this->result = get_temp(ir->type);
+      ir_to_mesa_instruction *inst;
+      inst = emit(ir, OPCODE_MOV, dst_reg(this->result), src);
+      inst->saturate = true;
+   }
 
    return true;
 }
@@ -1088,6 +1130,16 @@ ir_to_mesa_visitor::visit(ir_expression *ir)
       if (try_emit_mad(ir, 0))
 	 return;
    }
+
+   /* Quick peephole: Emit OPCODE_MAD(-a, -b, a) instead of AND(a, NOT(b))
+    */
+   if (ir->operation == ir_binop_logic_and) {
+      if (try_emit_mad_for_and_not(ir, 1))
+	 return;
+      if (try_emit_mad_for_and_not(ir, 0))
+	 return;
+   }
+
    if (try_emit_sat(ir))
       return;
 
@@ -1135,7 +1187,13 @@ ir_to_mesa_visitor::visit(ir_expression *ir)
 
    switch (ir->operation) {
    case ir_unop_logic_not:
-      emit(ir, OPCODE_SEQ, result_dst, op[0], src_reg_for_float(0.0));
+      /* Previously 'SEQ dst, src, 0.0' was used for this.  However, many
+       * older GPUs implement SEQ using multiple instructions (i915 uses two
+       * SGE instructions and a MUL instruction).  Since our logic values are
+       * 0.0 and 1.0, 1-x also implements !x.
+       */
+      op[0].negate = ~op[0].negate;
+      emit(ir, OPCODE_ADD, result_dst, op[0], src_reg_for_float(1.0));
       break;
    case ir_unop_neg:
       op[0].negate = ~op[0].negate;
@@ -1231,8 +1289,19 @@ ir_to_mesa_visitor::visit(ir_expression *ir)
 	  ir->operands[1]->type->is_vector()) {
 	 src_reg temp = get_temp(glsl_type::vec4_type);
 	 emit(ir, OPCODE_SNE, dst_reg(temp), op[0], op[1]);
+
+	 /* After the dot-product, the value will be an integer on the
+	  * range [0,4].  Zero becomes 1.0, and positive values become zero.
+	  */
 	 emit_dp(ir, result_dst, temp, temp, vector_elements);
-	 emit(ir, OPCODE_SEQ, result_dst, result_src, src_reg_for_float(0.0));
+
+	 /* Negating the result of the dot-product gives values on the range
+	  * [-4, 0].  Zero becomes 1.0, and negative values become zero.  This
+	  * achieved using SGE.
+	  */
+	 src_reg sge_src = result_src;
+	 sge_src.negate = ~sge_src.negate;
+	 emit(ir, OPCODE_SGE, result_dst, sge_src, src_reg_for_float(0.0));
       } else {
 	 emit(ir, OPCODE_SEQ, result_dst, op[0], op[1]);
       }
@@ -1243,29 +1312,83 @@ ir_to_mesa_visitor::visit(ir_expression *ir)
 	  ir->operands[1]->type->is_vector()) {
 	 src_reg temp = get_temp(glsl_type::vec4_type);
 	 emit(ir, OPCODE_SNE, dst_reg(temp), op[0], op[1]);
-	 emit_dp(ir, result_dst, temp, temp, vector_elements);
-	 emit(ir, OPCODE_SNE, result_dst, result_src, src_reg_for_float(0.0));
+
+	 /* After the dot-product, the value will be an integer on the
+	  * range [0,4].  Zero stays zero, and positive values become 1.0.
+	  */
+	 ir_to_mesa_instruction *const dp =
+	    emit_dp(ir, result_dst, temp, temp, vector_elements);
+	 if (this->prog->Target == GL_FRAGMENT_PROGRAM_ARB) {
+	    /* The clamping to [0,1] can be done for free in the fragment
+	     * shader with a saturate.
+	     */
+	    dp->saturate = true;
+	 } else {
+	    /* Negating the result of the dot-product gives values on the range
+	     * [-4, 0].  Zero stays zero, and negative values become 1.0.  This
+	     * achieved using SLT.
+	     */
+	    src_reg slt_src = result_src;
+	    slt_src.negate = ~slt_src.negate;
+	    emit(ir, OPCODE_SLT, result_dst, slt_src, src_reg_for_float(0.0));
+	 }
       } else {
 	 emit(ir, OPCODE_SNE, result_dst, op[0], op[1]);
       }
       break;
 
-   case ir_unop_any:
+   case ir_unop_any: {
       assert(ir->operands[0]->type->is_vector());
-      emit_dp(ir, result_dst, op[0], op[0],
-	      ir->operands[0]->type->vector_elements);
-      emit(ir, OPCODE_SNE, result_dst, result_src, src_reg_for_float(0.0));
+
+      /* After the dot-product, the value will be an integer on the
+       * range [0,4].  Zero stays zero, and positive values become 1.0.
+       */
+      ir_to_mesa_instruction *const dp =
+	 emit_dp(ir, result_dst, op[0], op[0],
+		 ir->operands[0]->type->vector_elements);
+      if (this->prog->Target == GL_FRAGMENT_PROGRAM_ARB) {
+	 /* The clamping to [0,1] can be done for free in the fragment
+	  * shader with a saturate.
+	  */
+	 dp->saturate = true;
+      } else {
+	 /* Negating the result of the dot-product gives values on the range
+	  * [-4, 0].  Zero stays zero, and negative values become 1.0.  This
+	  * is achieved using SLT.
+	  */
+	 src_reg slt_src = result_src;
+	 slt_src.negate = ~slt_src.negate;
+	 emit(ir, OPCODE_SLT, result_dst, slt_src, src_reg_for_float(0.0));
+      }
       break;
+   }
 
    case ir_binop_logic_xor:
       emit(ir, OPCODE_SNE, result_dst, op[0], op[1]);
       break;
 
-   case ir_binop_logic_or:
-      /* This could be a saturated add and skip the SNE. */
-      emit(ir, OPCODE_ADD, result_dst, op[0], op[1]);
-      emit(ir, OPCODE_SNE, result_dst, result_src, src_reg_for_float(0.0));
+   case ir_binop_logic_or: {
+      /* After the addition, the value will be an integer on the
+       * range [0,2].  Zero stays zero, and positive values become 1.0.
+       */
+      ir_to_mesa_instruction *add =
+	 emit(ir, OPCODE_ADD, result_dst, op[0], op[1]);
+      if (this->prog->Target == GL_FRAGMENT_PROGRAM_ARB) {
+	 /* The clamping to [0,1] can be done for free in the fragment
+	  * shader with a saturate.
+	  */
+	 add->saturate = true;
+      } else {
+	 /* Negating the result of the addition gives values on the range
+	  * [-2, 0].  Zero stays zero, and negative values become 1.0.  This
+	  * is achieved using SLT.
+	  */
+	 src_reg slt_src = result_src;
+	 slt_src.negate = ~slt_src.negate;
+	 emit(ir, OPCODE_SLT, result_dst, slt_src, src_reg_for_float(0.0));
+      }
       break;
+   }
 
    case ir_binop_logic_and:
       /* the bool args are stored as float 0.0 or 1.0, so "mul" gives us "and". */
@@ -1496,6 +1619,18 @@ ir_to_mesa_visitor::visit(ir_dereference_array *ir)
 	      this->result, src_reg_for_float(element_size));
       }
 
+      /* If there was already a relative address register involved, add the
+       * new and the old together to get the new offset.
+       */
+      if (src.reladdr != NULL)  {
+	 src_reg accum_reg = get_temp(glsl_type::float_type);
+
+	 emit(ir, OPCODE_ADD, dst_reg(accum_reg),
+	      index_reg, *src.reladdr);
+
+	 index_reg = accum_reg;
+      }
+
       src.reladdr = ralloc(mem_ctx, src_reg);
       memcpy(src.reladdr, &index_reg, sizeof(index_reg));
    }
@@ -1796,7 +1931,7 @@ 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);
@@ -1834,7 +1969,7 @@ 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);
 }
@@ -1969,7 +2104,10 @@ ir_to_mesa_visitor::visit(ir_texture *ir)
    ir_to_mesa_instruction *inst = NULL;
    prog_opcode opcode = OPCODE_NOP;
 
-   ir->coordinate->accept(this);
+   if (ir->op == ir_txs)
+      this->result = src_reg_for_float(0.0);
+   else
+      ir->coordinate->accept(this);
 
    /* Put our coords in a temp.  We'll need to modify them for shadow,
     * projection, or LOD, so the only case we'd use it as is is if
@@ -1993,6 +2131,7 @@ ir_to_mesa_visitor::visit(ir_texture *ir)
 
    switch (ir->op) {
    case ir_tex:
+   case ir_txs:
       opcode = OPCODE_TEX;
       break;
    case ir_txb:
@@ -2401,29 +2540,32 @@ check_resources(const struct gl_context *ctx,
    case GL_VERTEX_PROGRAM_ARB:
       if (_mesa_bitcount(prog->SamplersUsed) >
           ctx->Const.MaxVertexTextureImageUnits) {
-         fail_link(shader_program, "Too many vertex shader texture samplers");
+         linker_error(shader_program,
+		      "Too many vertex shader texture samplers");
       }
       if (prog->Parameters->NumParameters > MAX_UNIFORMS) {
-         fail_link(shader_program, "Too many vertex shader constants");
+         linker_error(shader_program, "Too many vertex shader constants");
       }
       break;
    case MESA_GEOMETRY_PROGRAM:
       if (_mesa_bitcount(prog->SamplersUsed) >
           ctx->Const.MaxGeometryTextureImageUnits) {
-         fail_link(shader_program, "Too many geometry shader texture samplers");
+         linker_error(shader_program,
+		      "Too many geometry shader texture samplers");
       }
       if (prog->Parameters->NumParameters >
           MAX_GEOMETRY_UNIFORM_COMPONENTS / 4) {
-         fail_link(shader_program, "Too many geometry shader constants");
+         linker_error(shader_program, "Too many geometry shader constants");
       }
       break;
    case GL_FRAGMENT_PROGRAM_ARB:
       if (_mesa_bitcount(prog->SamplersUsed) >
           ctx->Const.MaxTextureImageUnits) {
-         fail_link(shader_program, "Too many fragment shader texture samplers");
+         linker_error(shader_program,
+		      "Too many fragment shader texture samplers");
       }
       if (prog->Parameters->NumParameters > MAX_UNIFORMS) {
-         fail_link(shader_program, "Too many fragment shader constants");
+         linker_error(shader_program, "Too many fragment shader constants");
       }
       break;
    default:
@@ -2531,16 +2673,17 @@ 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
 	  * from _mesa_add_uniform) has to match what the linker chose.
 	  */
 	 if (index != parameter_index) {
-	    fail_link(shader_program, "Allocation of uniform `%s' to target "
-		      "failed (%d vs %d)\n",
-		      uniform->Name, index, parameter_index);
+	    linker_error(shader_program,
+			 "Allocation of uniform `%s' to target failed "
+			 "(%d vs %d)\n",
+			 uniform->Name, index, parameter_index);
 	 }
       }
    }
@@ -2573,8 +2716,8 @@ set_uniform_initializer(struct gl_context *ctx, void *mem_ctx,
    int loc = _mesa_get_uniform_location(ctx, shader_program, name);
 
    if (loc == -1) {
-      fail_link(shader_program,
-		"Couldn't find uniform for initializer %s\n", name);
+      linker_error(shader_program,
+		   "Couldn't find uniform for initializer %s\n", name);
       return;
    }
 
@@ -2974,11 +3117,31 @@ get_mesa_program(struct gl_context *ctx,
          if (mesa_inst->SrcReg[src].RelAddr)
             prog->IndirectRegisterFiles |= 1 << mesa_inst->SrcReg[src].File;
 
-      if (options->EmitNoIfs && mesa_inst->Opcode == OPCODE_IF) {
-	 fail_link(shader_program, "Couldn't flatten if statement\n");
-      }
-
       switch (mesa_inst->Opcode) {
+      case OPCODE_IF:
+	 if (options->EmitNoIfs) {
+	    linker_warning(shader_program,
+			   "Couldn't flatten if-statement.  "
+			   "This will likely result in software "
+			   "rasterization.\n");
+	 }
+	 break;
+      case OPCODE_BGNLOOP:
+	 if (options->EmitNoLoops) {
+	    linker_warning(shader_program,
+			   "Couldn't unroll loop.  "
+			   "This will likely result in software "
+			   "rasterization.\n");
+	 }
+	 break;
+      case OPCODE_CONT:
+	 if (options->EmitNoCont) {
+	    linker_warning(shader_program,
+			   "Couldn't lower continue-statement.  "
+			   "This will likely result in software "
+			   "rasterization.\n");
+	 }
+	 break;
       case OPCODE_BGNSUB:
 	 inst->function->inst = i;
 	 mesa_inst->Comment = strdup(inst->function->sig->function_name());
@@ -3246,7 +3409,7 @@ _mesa_glsl_link_shader(struct gl_context *ctx, struct gl_shader_program *prog)
 
    for (i = 0; i < prog->NumShaders; i++) {
       if (!prog->Shaders[i]->CompileStatus) {
-	 fail_link(prog, "linking with uncompiled shader");
+	 linker_error(prog, "linking with uncompiled shader");
 	 prog->LinkStatus = GL_FALSE;
       }
    }
diff --git a/src/mesa/program/nvfragparse.c b/src/mesa/program/nvfragparse.c
index 8516b5fc1ff..ce72c610d89 100644
--- a/src/mesa/program/nvfragparse.c
+++ b/src/mesa/program/nvfragparse.c
@@ -472,8 +472,9 @@ Parse_ScalarConstant(struct parse_state *parseState, GLfloat *number)
       const GLfloat *constant;
       if (!Parse_Identifier(parseState, ident))
          RETURN_ERROR1("Expected an identifier");
-      constant = _mesa_lookup_parameter_value(parseState->parameters,
-                                              -1, (const char *) ident);
+      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");
@@ -1039,7 +1040,8 @@ Parse_VectorSrc(struct parse_state *parseState,
       if (!Parse_ScalarConstant(parseState, values))
          RETURN_ERROR;
       paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
-                                              values, 4, NULL);
+                                              (gl_constant_value *) values,
+                                              4, NULL);
       srcReg->File = PROGRAM_NAMED_PARAM;
       srcReg->Index = paramIndex;
    }
@@ -1051,7 +1053,8 @@ Parse_VectorSrc(struct parse_state *parseState,
       if (!Parse_VectorConstant(parseState, values))
          RETURN_ERROR;
       paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
-                                              values, 4, NULL);
+                                              (gl_constant_value *) values,
+                                              4, NULL);
       srcReg->File = PROGRAM_NAMED_PARAM;
       srcReg->Index = paramIndex;      
    }
@@ -1145,7 +1148,8 @@ Parse_ScalarSrcReg(struct parse_state *parseState,
       if (!Parse_VectorConstant(parseState, values))
          RETURN_ERROR;
       paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
-                                              values, 4, NULL);
+                                              (gl_constant_value *) values,
+                                              4, NULL);
       srcReg->File = PROGRAM_NAMED_PARAM;
       srcReg->Index = paramIndex;      
    }
@@ -1170,7 +1174,8 @@ Parse_ScalarSrcReg(struct parse_state *parseState,
       if (!Parse_ScalarConstant(parseState, values))
          RETURN_ERROR;
       paramIndex = _mesa_add_unnamed_constant(parseState->parameters,
-                                              values, 4, NULL);
+                                              (gl_constant_value *) values,
+                                              4, NULL);
       srcReg->Index = paramIndex;      
       srcReg->File = PROGRAM_NAMED_PARAM;
       needSuffix = GL_FALSE;
@@ -1296,7 +1301,8 @@ Parse_InstructionSequence(struct parse_state *parseState,
             RETURN_ERROR2(id, "already defined");
          }
          _mesa_add_named_parameter(parseState->parameters,
-                                   (const char *) id, value);
+                                   (const char *) id,
+                                   (gl_constant_value *) value);
       }
       else if (Parse_String(parseState, "DECLARE")) {
          GLubyte id[100];
@@ -1315,7 +1321,8 @@ Parse_InstructionSequence(struct parse_state *parseState,
             RETURN_ERROR2(id, "already declared");
          }
          _mesa_add_named_parameter(parseState->parameters,
-                                   (const char *) id, value);
+                                   (const char *) id,
+                                   (gl_constant_value *) value);
       }
       else if (Parse_String(parseState, "END")) {
          inst->Opcode = OPCODE_END;
diff --git a/src/mesa/program/prog_execute.c b/src/mesa/program/prog_execute.c
index e7553c69dbe..77f842a1630 100644
--- a/src/mesa/program/prog_execute.c
+++ b/src/mesa/program/prog_execute.c
@@ -157,7 +157,7 @@ get_src_register_pointer(const struct prog_src_register *source,
    case PROGRAM_NAMED_PARAM:
       if (reg >= (GLint) prog->Parameters->NumParameters)
          return ZeroVec;
-      return prog->Parameters->ParameterValues[reg];
+      return (GLfloat *) prog->Parameters->ParameterValues[reg];
 
    case PROGRAM_SYSTEM_VALUE:
       assert(reg < Elements(machine->SystemValues));
@@ -639,7 +639,7 @@ _mesa_execute_program(struct gl_context * ctx,
                       struct gl_program_machine *machine)
 {
    const GLuint numInst = program->NumInstructions;
-   const GLuint maxExec = 10000;
+   const GLuint maxExec = 65536;
    GLuint pc, numExec = 0;
 
    machine->CurProgram = program;
@@ -1651,6 +1651,14 @@ _mesa_execute_program(struct gl_context * ctx,
             GLfloat texcoord[4], color[4];
             fetch_vector4(&inst->SrcReg[0], machine, texcoord);
 
+            /* For TEX, texcoord.Q should not be used and its value should not
+             * matter (at most, we pass coord.xyz to texture3D() in GLSL).
+             * Set Q=1 so that FetchTexelDeriv() doesn't get a garbage value
+             * which is effectively what happens when the texcoord swizzle
+             * is .xyzz
+             */
+            texcoord[3] = 1.0f;
+
             fetch_texel(ctx, machine, inst, texcoord, 0.0, color);
 
             if (DEBUG_PROG) {
diff --git a/src/mesa/program/prog_opt_constant_fold.c b/src/mesa/program/prog_opt_constant_fold.c
new file mode 100644
index 00000000000..e2418b55451
--- /dev/null
+++ b/src/mesa/program/prog_opt_constant_fold.c
@@ -0,0 +1,451 @@
+/*
+ * Copyright © 2010 Intel Corporation
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the next
+ * paragraph) shall be included in all copies or substantial portions of the
+ * Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+ * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
+ * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
+ * DEALINGS IN THE SOFTWARE.
+ */
+
+#include "main/glheader.h"
+#include "main/context.h"
+#include "main/macros.h"
+#include "program.h"
+#include "prog_instruction.h"
+#include "prog_optimize.h"
+#include "prog_parameter.h"
+#include <stdbool.h>
+
+static bool
+src_regs_are_constant(const struct prog_instruction *inst, unsigned num_srcs)
+{
+   unsigned i;
+
+   for (i = 0; i < num_srcs; i++) {
+      if (inst->SrcReg[i].File != PROGRAM_CONSTANT)
+	 return false;
+   }
+
+   return true;
+}
+
+static struct prog_src_register
+src_reg_for_float(struct gl_program *prog, float val)
+{
+   struct prog_src_register src;
+   unsigned swiz;
+
+   memset(&src, 0, sizeof(src));
+
+   src.File = PROGRAM_CONSTANT;
+   src.Index = _mesa_add_unnamed_constant(prog->Parameters,
+					  (gl_constant_value *) &val, 1, &swiz);
+   src.Swizzle = swiz;
+   return src;
+}
+
+static struct prog_src_register
+src_reg_for_vec4(struct gl_program *prog, const float *val)
+{
+   struct prog_src_register src;
+   unsigned swiz;
+
+   memset(&src, 0, sizeof(src));
+
+   src.File = PROGRAM_CONSTANT;
+   src.Index = _mesa_add_unnamed_constant(prog->Parameters,
+					  (gl_constant_value *) val, 4, &swiz);
+   src.Swizzle = swiz;
+   return src;
+}
+
+static bool
+src_regs_are_same(const struct prog_src_register *a,
+		  const struct prog_src_register *b)
+{
+   return (a->File == b->File)
+      && (a->Index == b->Index)
+      && (a->Swizzle == b->Swizzle)
+      && (a->Abs == b->Abs)
+      && (a->Negate == b->Negate)
+      && (a->RelAddr == 0)
+      && (b->RelAddr == 0);
+}
+
+static void
+get_value(struct gl_program *prog, struct prog_src_register *r, float *data)
+{
+   const gl_constant_value *const value =
+      prog->Parameters->ParameterValues[r->Index];
+
+   data[0] = value[GET_SWZ(r->Swizzle, 0)].f;
+   data[1] = value[GET_SWZ(r->Swizzle, 1)].f;
+   data[2] = value[GET_SWZ(r->Swizzle, 2)].f;
+   data[3] = value[GET_SWZ(r->Swizzle, 3)].f;
+
+   if (r->Abs) {
+      data[0] = fabsf(data[0]);
+      data[1] = fabsf(data[1]);
+      data[2] = fabsf(data[2]);
+      data[3] = fabsf(data[3]);
+   }
+
+   if (r->Negate & 0x01) {
+      data[0] = -data[0];
+   }
+
+   if (r->Negate & 0x02) {
+      data[1] = -data[1];
+   }
+
+   if (r->Negate & 0x04) {
+      data[2] = -data[2];
+   }
+
+   if (r->Negate & 0x08) {
+      data[3] = -data[3];
+   }
+}
+
+/**
+ * Try to replace instructions that produce a constant result with simple moves
+ *
+ * The hope is that a following copy propagation pass will eliminate the
+ * unnecessary move instructions.
+ */
+GLboolean
+_mesa_constant_fold(struct gl_program *prog)
+{
+   bool progress = false;
+   unsigned i;
+
+   for (i = 0; i < prog->NumInstructions; i++) {
+      struct prog_instruction *const inst = &prog->Instructions[i];
+
+      switch (inst->Opcode) {
+      case OPCODE_ADD:
+	 if (src_regs_are_constant(inst, 2)) {
+	    float a[4];
+	    float b[4];
+	    float result[4];
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], 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];
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_vec4(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      case OPCODE_CMP:
+	 /* FINISHME: We could also optimize CMP instructions where the first
+	  * FINISHME: source is a constant that is either all < 0.0 or all
+	  * FINISHME: >= 0.0.
+	  */
+	 if (src_regs_are_constant(inst, 3)) {
+	    float a[4];
+	    float b[4];
+	    float c[4];
+	    float result[4];
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], b);
+	    get_value(prog, &inst->SrcReg[2], 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];
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_vec4(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+	    inst->SrcReg[2].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[2].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      case OPCODE_DP2:
+      case OPCODE_DP3:
+      case OPCODE_DP4:
+	 if (src_regs_are_constant(inst, 2)) {
+	    float a[4];
+	    float b[4];
+	    float result;
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], b);
+
+	    /* It seems like a loop could be used here, but we cleverly put
+	     * DP2A between DP2 and DP3.  Subtracting DP2 (or similar) from
+	     * the opcode results in various failures of the loop control.
+	     */
+	    result = (a[0] * b[0]) + (a[1] * b[1]);
+
+	    if (inst->Opcode >= OPCODE_DP3)
+	       result += a[2] * b[2];
+
+	    if (inst->Opcode == OPCODE_DP4)
+	       result += a[3] * b[3];
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_float(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      case OPCODE_MUL:
+	 if (src_regs_are_constant(inst, 2)) {
+	    float a[4];
+	    float b[4];
+	    float result[4];
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], 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];
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_vec4(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      case OPCODE_SEQ:
+	 if (src_regs_are_constant(inst, 2)) {
+	    float a[4];
+	    float b[4];
+	    float result[4];
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], 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;
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_vec4(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 } else if (src_regs_are_same(&inst->SrcReg[0], &inst->SrcReg[1])) {
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_float(prog, 1.0f);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      case OPCODE_SGE:
+	 if (src_regs_are_constant(inst, 2)) {
+	    float a[4];
+	    float b[4];
+	    float result[4];
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], 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;
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_vec4(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 } else if (src_regs_are_same(&inst->SrcReg[0], &inst->SrcReg[1])) {
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_float(prog, 1.0f);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      case OPCODE_SGT:
+	 if (src_regs_are_constant(inst, 2)) {
+	    float a[4];
+	    float b[4];
+	    float result[4];
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], 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;
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_vec4(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 } else if (src_regs_are_same(&inst->SrcReg[0], &inst->SrcReg[1])) {
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_float(prog, 0.0f);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      case OPCODE_SLE:
+	 if (src_regs_are_constant(inst, 2)) {
+	    float a[4];
+	    float b[4];
+	    float result[4];
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], 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;
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_vec4(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 } else if (src_regs_are_same(&inst->SrcReg[0], &inst->SrcReg[1])) {
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_float(prog, 1.0f);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      case OPCODE_SLT:
+	 if (src_regs_are_constant(inst, 2)) {
+	    float a[4];
+	    float b[4];
+	    float result[4];
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], 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;
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_vec4(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 } else if (src_regs_are_same(&inst->SrcReg[0], &inst->SrcReg[1])) {
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_float(prog, 0.0f);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      case OPCODE_SNE:
+	 if (src_regs_are_constant(inst, 2)) {
+	    float a[4];
+	    float b[4];
+	    float result[4];
+
+	    get_value(prog, &inst->SrcReg[0], a);
+	    get_value(prog, &inst->SrcReg[1], 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;
+
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_vec4(prog, result);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 } else if (src_regs_are_same(&inst->SrcReg[0], &inst->SrcReg[1])) {
+	    inst->Opcode = OPCODE_MOV;
+	    inst->SrcReg[0] = src_reg_for_float(prog, 0.0f);
+
+	    inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	    inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+
+	    progress = true;
+	 }
+	 break;
+
+      default:
+	 break;
+      }
+   }
+
+   return progress;
+}
diff --git a/src/mesa/program/prog_optimize.c b/src/mesa/program/prog_optimize.c
index 8a40fa69eca..25d9684b137 100644
--- a/src/mesa/program/prog_optimize.c
+++ b/src/mesa/program/prog_optimize.c
@@ -472,8 +472,7 @@ can_downward_mov_be_modifed(const struct prog_instruction *mov)
       mov->SrcReg[0].HasIndex2 == 0 &&
       mov->SrcReg[0].RelAddr2 == 0 &&
       mov->DstReg.RelAddr == 0 &&
-      mov->DstReg.CondMask == COND_TR &&
-      mov->SaturateMode == SATURATE_OFF;
+      mov->DstReg.CondMask == COND_TR;
 }
 
 
@@ -482,7 +481,8 @@ can_upward_mov_be_modifed(const struct prog_instruction *mov)
 {
    return
       can_downward_mov_be_modifed(mov) &&
-      mov->DstReg.File == PROGRAM_TEMPORARY;
+      mov->DstReg.File == PROGRAM_TEMPORARY &&
+      mov->SaturateMode == SATURATE_OFF;
 }
 
 
@@ -657,6 +657,8 @@ _mesa_merge_mov_into_inst(struct prog_instruction *inst,
    if (mask != (inst->DstReg.WriteMask & mask))
       return GL_FALSE;
 
+   inst->SaturateMode |= mov->SaturateMode;
+
    /* Depending on the instruction, we may need to recompute the swizzles.
     * Also, some other instructions (like TEX) are not linear. We will only
     * consider completely active sources and destinations
@@ -1319,6 +1321,15 @@ _mesa_simplify_cmp(struct gl_program * program)
 
          inst->Opcode = OPCODE_MOV;
          inst->SrcReg[0] = inst->SrcReg[1];
+
+	 /* Unused operands are expected to have the file set to
+	  * PROGRAM_UNDEFINED.  This is how _mesa_init_instructions initializes
+	  * all of the sources.
+	  */
+	 inst->SrcReg[1].File = PROGRAM_UNDEFINED;
+	 inst->SrcReg[1].Swizzle = SWIZZLE_NOOP;
+	 inst->SrcReg[2].File = PROGRAM_UNDEFINED;
+	 inst->SrcReg[2].Swizzle = SWIZZLE_NOOP;
       }
    }
    if (dbg) {
@@ -1347,6 +1358,8 @@ _mesa_optimize_program(struct gl_context *ctx, struct gl_program *program)
          any_change = GL_TRUE;
       if (_mesa_remove_dead_code_local(program))
          any_change = GL_TRUE;
+
+      any_change = _mesa_constant_fold(program) || any_change;
       _mesa_reallocate_registers(program);
    } while (any_change);
 }
diff --git a/src/mesa/program/prog_optimize.h b/src/mesa/program/prog_optimize.h
index 463f5fc51c4..9854fb7a491 100644
--- a/src/mesa/program/prog_optimize.h
+++ b/src/mesa/program/prog_optimize.h
@@ -44,4 +44,7 @@ _mesa_find_temp_intervals(const struct prog_instruction *instructions,
 extern void
 _mesa_optimize_program(struct gl_context *ctx, struct gl_program *program);
 
+extern GLboolean
+_mesa_constant_fold(struct gl_program *prog);
+
 #endif
diff --git a/src/mesa/program/prog_parameter.c b/src/mesa/program/prog_parameter.c
index 3570cab118b..49b3ffbdd5c 100644
--- a/src/mesa/program/prog_parameter.c
+++ b/src/mesa/program/prog_parameter.c
@@ -56,8 +56,8 @@ _mesa_new_parameter_list_sized(unsigned size)
       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);
+      p->ParameterValues = (gl_constant_value (*)[4])
+         _mesa_align_malloc(size * 4 *sizeof(gl_constant_value), 16);
 
 
       if ((p->Parameters == NULL) || (p->ParameterValues == NULL)) {
@@ -101,14 +101,15 @@ _mesa_free_parameter_list(struct gl_program_parameter_list *paramList)
  * \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 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 GLfloat *values,
+                    GLuint size, GLenum datatype,
+                    const gl_constant_value *values,
                     const gl_state_index state[STATE_LENGTH],
                     GLbitfield flags)
 {
@@ -127,10 +128,10 @@ _mesa_add_parameter(struct gl_program_parameter_list *paramList,
 		       oldNum * sizeof(struct gl_program_parameter),
 		       paramList->Size * sizeof(struct gl_program_parameter));
 
-      paramList->ParameterValues = (GLfloat (*)[4])
+      paramList->ParameterValues = (gl_constant_value (*)[4])
          _mesa_align_realloc(paramList->ParameterValues,         /* old buf */
-                             oldNum * 4 * sizeof(GLfloat),      /* old size */
-                             paramList->Size * 4 *sizeof(GLfloat), /* new sz */
+                             oldNum * 4 * sizeof(gl_constant_value),/* old sz */
+                             paramList->Size*4*sizeof(gl_constant_value),/*new*/
                              16);
    }
 
@@ -142,7 +143,7 @@ _mesa_add_parameter(struct gl_program_parameter_list *paramList,
       return -1;
    }
    else {
-      GLuint i;
+      GLuint i, j;
 
       paramList->NumParameters = oldNum + sz4;
 
@@ -163,7 +164,8 @@ _mesa_add_parameter(struct gl_program_parameter_list *paramList,
          }
          else {
             /* silence valgrind */
-            ASSIGN_4V(paramList->ParameterValues[oldNum + i], 0, 0, 0, 0);
+            for (j = 0; j < 4; j++)
+            	paramList->ParameterValues[oldNum + i][j].f = 0;
          }
          size -= 4;
       }
@@ -184,7 +186,7 @@ _mesa_add_parameter(struct gl_program_parameter_list *paramList,
  */
 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])
 {
    return _mesa_add_parameter(paramList, PROGRAM_NAMED_PARAM, name,
                               4, GL_NONE, values, NULL, 0x0);
@@ -204,17 +206,17 @@ _mesa_add_named_parameter(struct gl_program_parameter_list *paramList,
  */
 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)
 {
    /* 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] &&
+      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;
@@ -239,9 +241,9 @@ _mesa_add_named_constant(struct gl_program_parameter_list *paramList,
  * \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)
+_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);
@@ -262,7 +264,7 @@ _mesa_add_unnamed_constant(struct gl_program_parameter_list *paramList,
          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];
+            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++;
@@ -274,7 +276,7 @@ _mesa_add_unnamed_constant(struct gl_program_parameter_list *paramList,
 
    /* add a new parameter to store this constant */
    pos = _mesa_add_parameter(paramList, PROGRAM_CONSTANT, NULL,
-                             size, GL_NONE, values, NULL, 0x0);
+                             size, datatype, values, NULL, 0x0);
    if (pos >= 0 && swizzleOut) {
       if (size == 1)
          *swizzleOut = SWIZZLE_XXXX;
@@ -285,6 +287,28 @@ _mesa_add_unnamed_constant(struct gl_program_parameter_list *paramList,
 }
 
 /**
+ * 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
@@ -401,7 +425,7 @@ _mesa_add_state_reference(struct gl_program_parameter_list *paramList,
  * Lookup a parameter value by name in the given parameter list.
  * \return pointer to the float[4] values.
  */
-GLfloat *
+gl_constant_value *
 _mesa_lookup_parameter_value(const struct gl_program_parameter_list *paramList,
                              GLsizei nameLen, const char *name)
 {
@@ -465,7 +489,7 @@ _mesa_lookup_parameter_index(const struct gl_program_parameter_list *paramList,
  */
 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)
 {
    GLuint i;
@@ -484,7 +508,7 @@ _mesa_lookup_parameter_constant(const struct gl_program_parameter_list *list,
             /* swizzle not allowed */
             GLuint j, match = 0;
             for (j = 0; j < vSize; j++) {
-               if (v[j] == list->ParameterValues[i][j])
+               if (v[j].u == list->ParameterValues[i][j].u)
                   match++;
             }
             if (match == vSize) {
@@ -498,7 +522,7 @@ _mesa_lookup_parameter_constant(const struct gl_program_parameter_list *list,
                 /* 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]) {
+                   if (list->ParameterValues[i][j].u == v[0].u) {
                       /* found it */
                       *posOut = i;
                       *swizzleOut = MAKE_SWIZZLE4(j, j, j, j);
@@ -511,13 +535,13 @@ _mesa_lookup_parameter_constant(const struct gl_program_parameter_list *list,
                 GLuint swz[4];
                 GLuint match = 0, j, k;
                 for (j = 0; j < vSize; j++) {
-                   if (v[j] == list->ParameterValues[i][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] == list->ParameterValues[i][k]) {
+                         if (v[j].u == list->ParameterValues[i][k].u) {
                             swz[j] = k;
                             match++;
                             break;
diff --git a/src/mesa/program/prog_parameter.h b/src/mesa/program/prog_parameter.h
index 10cbbe57a6c..1a5ed343937 100644
--- a/src/mesa/program/prog_parameter.h
+++ b/src/mesa/program/prog_parameter.h
@@ -47,6 +47,17 @@
 /*@}*/
 
 
+/**
+ * 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 +92,7 @@ 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 +123,28 @@ _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 +160,7 @@ 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 +170,7 @@ _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/src/mesa/program/prog_parameter_layout.c b/src/mesa/program/prog_parameter_layout.c
index 90a9771080c..28fca3b92d9 100644
--- a/src/mesa/program/prog_parameter_layout.c
+++ b/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/src/mesa/program/prog_print.c b/src/mesa/program/prog_print.c
index 7c3b4909e73..70412b1fa6a 100644
--- a/src/mesa/program/prog_print.c
+++ b/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/src/mesa/program/prog_statevars.c b/src/mesa/program/prog_statevars.c
index 16f9690e865..6aa2409e85e 100644
--- a/src/mesa/program/prog_statevars.c
+++ b/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/src/mesa/program/program.c b/src/mesa/program/program.c
index adca094ee89..ecff2344a44 100644
--- a/src/mesa/program/program.c
+++ b/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/src/mesa/program/program_parse.y b/src/mesa/program/program_parse.y
index dbf5abaa617..dec35038be5 100644
--- a/src/mesa/program/program_parse.y
+++ b/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/src/mesa/program/program_parser.h b/src/mesa/program/program_parser.h
index 8e5aaee95e5..5637598f3b3 100644
--- a/src/mesa/program/program_parser.h
+++ b/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/src/mesa/program/register_allocate.c b/src/mesa/program/register_allocate.c
index de96eb42c9b..f5b5174fc18 100644
--- a/src/mesa/program/register_allocate.c
+++ b/src/mesa/program/register_allocate.c
@@ -200,6 +200,27 @@ ra_add_reg_conflict(struct ra_regs *regs, unsigned int r1, unsigned int r2)
    }
 }
 
+/**
+ * Adds a conflict between base_reg and reg, and also between reg and
+ * anything that base_reg conflicts with.
+ *
+ * This can simplify code for setting up multiple register classes
+ * which are aggregates of some base hardware registers, compared to
+ * explicitly using ra_add_reg_conflict.
+ */
+void
+ra_add_transitive_reg_conflict(struct ra_regs *regs,
+			       unsigned int base_reg, unsigned int reg)
+{
+   int i;
+
+   ra_add_reg_conflict(regs, reg, base_reg);
+
+   for (i = 0; i < regs->regs[base_reg].num_conflicts; i++) {
+      ra_add_reg_conflict(regs, reg, regs->regs[base_reg].conflict_list[i]);
+   }
+}
+
 unsigned int
 ra_alloc_reg_class(struct ra_regs *regs)
 {
diff --git a/src/mesa/program/register_allocate.h b/src/mesa/program/register_allocate.h
index 5b95833f394..ee2e58a4756 100644
--- a/src/mesa/program/register_allocate.h
+++ b/src/mesa/program/register_allocate.h
@@ -40,6 +40,8 @@ struct ra_regs *ra_alloc_reg_set(unsigned int count);
 unsigned int ra_alloc_reg_class(struct ra_regs *regs);
 void ra_add_reg_conflict(struct ra_regs *regs,
 			 unsigned int r1, unsigned int r2);
+void ra_add_transitive_reg_conflict(struct ra_regs *regs,
+				    unsigned int base_reg, unsigned int reg);
 void ra_class_add_reg(struct ra_regs *regs, unsigned int c, unsigned int reg);
 void ra_set_finalize(struct ra_regs *regs);
 /** @} */
diff --git a/src/mesa/program/sampler.cpp b/src/mesa/program/sampler.cpp
index 1457d1199fa..e8d34c670a9 100644
--- a/src/mesa/program/sampler.cpp
+++ b/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;
 }
 }