mesa git update 9 sep 2011

20 files changed, 2600 insertions, 415 deletions

diff --git a/mesalib/src/glsl/ast_to_hir.cpp b/mesalib/src/glsl/ast_to_hir.cpp
index 9e7496b4b..777f190a4 100644
--- a/mesalib/src/glsl/ast_to_hir.cpp
+++ b/mesalib/src/glsl/ast_to_hir.cpp
@@ -679,16 +679,20 @@ do_assignment(exec_list *instructions, struct _mesa_glsl_parse_state *state,
                           lhs->variable_referenced()->name);
          error_emitted = true;
 
+      } else if (state->language_version <= 110 && lhs->type->is_array()) {
+	 /* From page 32 (page 38 of the PDF) of the GLSL 1.10 spec:
+	  *
+	  *    "Other binary or unary expressions, non-dereferenced
+	  *     arrays, function names, swizzles with repeated fields,
+	  *     and constants cannot be l-values."
+	  */
+	 _mesa_glsl_error(&lhs_loc, state, "whole array assignment is not "
+			  "allowed in GLSL 1.10 or GLSL ES 1.00.");
+	 error_emitted = true;
       } else if (!lhs->is_lvalue()) {
 	 _mesa_glsl_error(& lhs_loc, state, "non-lvalue in assignment");
 	 error_emitted = true;
       }
-
-      if (state->es_shader && lhs->type->is_array()) {
-	 _mesa_glsl_error(&lhs_loc, state, "whole array assignment is not "
-			  "allowed in GLSL ES 1.00.");
-	 error_emitted = true;
-      }
    }
 
    ir_rvalue *new_rhs =
@@ -723,6 +727,7 @@ do_assignment(exec_list *instructions, struct _mesa_glsl_parse_state *state,
 						   rhs->type->array_size());
 	 d->type = var->type;
       }
+      mark_whole_array_access(rhs);
       mark_whole_array_access(lhs);
    }
 
@@ -893,6 +898,44 @@ get_scalar_boolean_operand(exec_list *instructions,
    return new(ctx) ir_constant(true);
 }
 
+/**
+ * If name refers to a builtin array whose maximum allowed size is less than
+ * size, report an error and return true.  Otherwise return false.
+ */
+static bool
+check_builtin_array_max_size(const char *name, unsigned size,
+                             YYLTYPE loc, struct _mesa_glsl_parse_state *state)
+{
+   if ((strcmp("gl_TexCoord", name) == 0)
+       && (size > state->Const.MaxTextureCoords)) {
+      /* From page 54 (page 60 of the PDF) of the GLSL 1.20 spec:
+       *
+       *     "The size [of gl_TexCoord] can be at most
+       *     gl_MaxTextureCoords."
+       */
+      _mesa_glsl_error(&loc, state, "`gl_TexCoord' array size cannot "
+                       "be larger than gl_MaxTextureCoords (%u)\n",
+                       state->Const.MaxTextureCoords);
+      return true;
+   } else if (strcmp("gl_ClipDistance", name) == 0
+              && size > state->Const.MaxClipPlanes) {
+      /* From section 7.1 (Vertex Shader Special Variables) of the
+       * GLSL 1.30 spec:
+       *
+       *   "The gl_ClipDistance array is predeclared as unsized and
+       *   must be sized by the shader either redeclaring it with a
+       *   size or indexing it only with integral constant
+       *   expressions. ... The size can be at most
+       *   gl_MaxClipDistances."
+       */
+      _mesa_glsl_error(&loc, state, "`gl_ClipDistance' array size cannot "
+                       "be larger than gl_MaxClipDistances (%u)\n",
+                       state->Const.MaxClipPlanes);
+      return true;
+   }
+   return false;
+}
+
 ir_rvalue *
 ast_expression::hir(exec_list *instructions,
 		    struct _mesa_glsl_parse_state *state)
@@ -1550,8 +1593,15 @@ ast_expression::hir(exec_list *instructions,
 	     * FINISHME: array access limits be added to ir_dereference?
 	     */
 	    ir_variable *const v = array->whole_variable_referenced();
-	    if ((v != NULL) && (unsigned(idx) > v->max_array_access))
+	    if ((v != NULL) && (unsigned(idx) > v->max_array_access)) {
 	       v->max_array_access = idx;
+
+               /* Check whether this access will, as a side effect, implicitly
+                * cause the size of a built-in array to be too large.
+                */
+               if (check_builtin_array_max_size(v->name, idx+1, loc, state))
+                  error_emitted = true;
+            }
 	 }
       } else if (array->type->array_size() == 0) {
 	 _mesa_glsl_error(&loc, state, "unsized array index must be constant");
@@ -2071,6 +2121,21 @@ apply_type_qualifier_to_variable(const struct ast_type_qualifier *qual,
    else
        var->depth_layout = ir_depth_layout_none;
 
+   /* From page 46 (page 52 of the PDF) of the GLSL ES specification:
+    *
+    *    "Array variables are l-values and may be passed to parameters
+    *     declared as out or inout. However, they may not be used as
+    *     the target of an assignment."
+    *
+    * From page 32 (page 38 of the PDF) of the GLSL 1.10 spec:
+    *
+    *    "Other binary or unary expressions, non-dereferenced arrays,
+    *     function names, swizzles with repeated fields, and constants
+    *     cannot be l-values."
+    *
+    * So we only mark 1.10 as non-lvalues, and check for array
+    * assignment in 100 specifically in do_assignment.
+    */
    if (var->type->is_array() && state->language_version != 110) {
       var->array_lvalue = true;
    }
@@ -2121,18 +2186,9 @@ get_variable_being_redeclared(ir_variable *var, ast_declaration *decl,
        * FINISHME: required or not.
        */
 
-      /* From page 54 (page 60 of the PDF) of the GLSL 1.20 spec:
-       *
-       *     "The size [of gl_TexCoord] can be at most
-       *     gl_MaxTextureCoords."
-       */
       const unsigned size = unsigned(var->type->array_size());
-      if ((strcmp("gl_TexCoord", var->name) == 0)
-	  && (size > state->Const.MaxTextureCoords)) {
-	 _mesa_glsl_error(& loc, state, "`gl_TexCoord' array size cannot "
-			  "be larger than gl_MaxTextureCoords (%u)\n",
-			  state->Const.MaxTextureCoords);
-      } else if ((size > 0) && (size <= earlier->max_array_access)) {
+      check_builtin_array_max_size(var->name, size, loc, state);
+      if ((size > 0) && (size <= earlier->max_array_access)) {
 	 _mesa_glsl_error(& loc, state, "array size must be > %u due to "
 			  "previous access",
 			  earlier->max_array_access);
diff --git a/mesalib/src/glsl/glsl_parser_extras.cpp b/mesalib/src/glsl/glsl_parser_extras.cpp
index 8f740e6a8..8faddc578 100644
--- a/mesalib/src/glsl/glsl_parser_extras.cpp
+++ b/mesalib/src/glsl/glsl_parser_extras.cpp
@@ -891,8 +891,8 @@ do_common_optimization(exec_list *ir, bool linked, unsigned max_unroll_iteration
    if (linked) {
       progress = do_function_inlining(ir) || progress;
       progress = do_dead_functions(ir) || progress;
+      progress = do_structure_splitting(ir) || progress;
    }
-   progress = do_structure_splitting(ir) || progress;
    progress = do_if_simplification(ir) || progress;
    progress = do_discard_simplification(ir) || progress;
    progress = do_copy_propagation(ir) || progress;
diff --git a/mesalib/src/glsl/ir_set_program_inouts.cpp b/mesalib/src/glsl/ir_set_program_inouts.cpp
index c5533c8cd..3b10b9097 100644
--- a/mesalib/src/glsl/ir_set_program_inouts.cpp
+++ b/mesalib/src/glsl/ir_set_program_inouts.cpp
@@ -1,166 +1,164 @@
-/*

- * 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.

- */

-

-/**

- * \file ir_set_program_inouts.cpp

- *

- * Sets the InputsRead and OutputsWritten of Mesa programs.

- *

- * Mesa programs (gl_program, not gl_shader_program) have a set of

- * flags indicating which varyings are read and written.  Computing

- * which are actually read from some sort of backend code can be

- * tricky when variable array indexing involved.  So this pass

- * provides support for setting InputsRead and OutputsWritten right

- * from the GLSL IR.

- */

-

-extern "C" {

-#include "main/core.h" /* for struct gl_program */

-#include "program/hash_table.h"

-}

-#include "ir.h"

-#include "ir_visitor.h"

-#include "glsl_types.h"

-

-class ir_set_program_inouts_visitor : public ir_hierarchical_visitor {

-public:

-   ir_set_program_inouts_visitor(struct gl_program *prog)

-   {

-      this->prog = prog;

-      this->ht = hash_table_ctor(0,

-				 hash_table_pointer_hash,

-				 hash_table_pointer_compare);

-   }

-   ~ir_set_program_inouts_visitor()

-   {

-      hash_table_dtor(this->ht);

-   }

-

-   virtual ir_visitor_status visit_enter(ir_dereference_array *);

-   virtual ir_visitor_status visit_enter(ir_function_signature *);

-   virtual ir_visitor_status visit(ir_dereference_variable *);

-   virtual ir_visitor_status visit(ir_variable *);

-

-   struct gl_program *prog;

-   struct hash_table *ht;

-};

-

-static void

-mark(struct gl_program *prog, ir_variable *var, int offset, int len)

-{

-   /* As of GLSL 1.20, varyings can only be floats, floating-point

-    * vectors or matrices, or arrays of them.  For Mesa programs using

-    * InputsRead/OutputsWritten, everything but matrices uses one

-    * slot, while matrices use a slot per column.  Presumably

-    * something doing a more clever packing would use something other

-    * than InputsRead/OutputsWritten.

-    */

-

-   for (int i = 0; i < len; i++) {

-      if (var->mode == ir_var_in)

-	 prog->InputsRead |= BITFIELD64_BIT(var->location + offset + i);

-      else if (var->mode == ir_var_system_value)

-         prog->SystemValuesRead |= (1 << (var->location + offset + i));

-      else

-	 prog->OutputsWritten |= BITFIELD64_BIT(var->location + offset + i);

-   }

-}

-

-/* Default handler: Mark all the locations in the variable as used. */

-ir_visitor_status

-ir_set_program_inouts_visitor::visit(ir_dereference_variable *ir)

-{

-   if (hash_table_find(this->ht, ir->var) == NULL)

-      return visit_continue;

-

-   if (ir->type->is_array()) {

-      for (unsigned int i = 0; i < ir->type->length; i++) {

-	 mark(this->prog, ir->var, i,

-	      ir->type->length * ir->type->fields.array->matrix_columns);

-      }

-   } else {

-      mark(this->prog, ir->var, 0, ir->type->matrix_columns);

-   }

-

-   return visit_continue;

-}

-

-ir_visitor_status

-ir_set_program_inouts_visitor::visit_enter(ir_dereference_array *ir)

-{

-   ir_dereference_variable *deref_var;

-   ir_constant *index = ir->array_index->as_constant();

-   deref_var = ir->array->as_dereference_variable();

-   ir_variable *var = NULL;

-

-   /* Check that we're dereferencing a shader in or out */

-   if (deref_var)

-      var = (ir_variable *)hash_table_find(this->ht, deref_var->var);

-

-   if (index && var) {

-      int width = 1;

-

-      if (deref_var->type->is_array() &&

-	  deref_var->type->fields.array->is_matrix()) {

-	 width = deref_var->type->fields.array->matrix_columns;

-      }

-

-      mark(this->prog, var, index->value.i[0] * width, width);

-      return visit_continue_with_parent;

-   }

-

-   return visit_continue;

-}

-

-ir_visitor_status

-ir_set_program_inouts_visitor::visit(ir_variable *ir)

-{

-   if (ir->mode == ir_var_in ||

-       ir->mode == ir_var_out ||

-       ir->mode == ir_var_system_value) {

-      hash_table_insert(this->ht, ir, ir);

-   }

-

-   return visit_continue;

-}

-

-ir_visitor_status

-ir_set_program_inouts_visitor::visit_enter(ir_function_signature *ir)

-{

-   /* We don't want to descend into the function parameters and

-    * consider them as shader inputs or outputs.

-    */

-   visit_list_elements(this, &ir->body);

-   return visit_continue_with_parent;

-}

-

-void

-do_set_program_inouts(exec_list *instructions, struct gl_program *prog)

-{

-   ir_set_program_inouts_visitor v(prog);

-

-   prog->InputsRead = 0;

-   prog->OutputsWritten = 0;

-   prog->SystemValuesRead = 0;

-   visit_list_elements(&v, instructions);

-}

+/*
+ * 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.
+ */
+
+/**
+ * \file ir_set_program_inouts.cpp
+ *
+ * Sets the InputsRead and OutputsWritten of Mesa programs.
+ *
+ * Mesa programs (gl_program, not gl_shader_program) have a set of
+ * flags indicating which varyings are read and written.  Computing
+ * which are actually read from some sort of backend code can be
+ * tricky when variable array indexing involved.  So this pass
+ * provides support for setting InputsRead and OutputsWritten right
+ * from the GLSL IR.
+ */
+
+extern "C" {
+#include "main/core.h" /* for struct gl_program */
+#include "program/hash_table.h"
+}
+#include "ir.h"
+#include "ir_visitor.h"
+#include "glsl_types.h"
+
+class ir_set_program_inouts_visitor : public ir_hierarchical_visitor {
+public:
+   ir_set_program_inouts_visitor(struct gl_program *prog)
+   {
+      this->prog = prog;
+      this->ht = hash_table_ctor(0,
+				 hash_table_pointer_hash,
+				 hash_table_pointer_compare);
+   }
+   ~ir_set_program_inouts_visitor()
+   {
+      hash_table_dtor(this->ht);
+   }
+
+   virtual ir_visitor_status visit_enter(ir_dereference_array *);
+   virtual ir_visitor_status visit_enter(ir_function_signature *);
+   virtual ir_visitor_status visit(ir_dereference_variable *);
+   virtual ir_visitor_status visit(ir_variable *);
+
+   struct gl_program *prog;
+   struct hash_table *ht;
+};
+
+static void
+mark(struct gl_program *prog, ir_variable *var, int offset, int len)
+{
+   /* As of GLSL 1.20, varyings can only be floats, floating-point
+    * vectors or matrices, or arrays of them.  For Mesa programs using
+    * InputsRead/OutputsWritten, everything but matrices uses one
+    * slot, while matrices use a slot per column.  Presumably
+    * something doing a more clever packing would use something other
+    * than InputsRead/OutputsWritten.
+    */
+
+   for (int i = 0; i < len; i++) {
+      if (var->mode == ir_var_in)
+	 prog->InputsRead |= BITFIELD64_BIT(var->location + offset + i);
+      else if (var->mode == ir_var_system_value)
+         prog->SystemValuesRead |= (1 << (var->location + offset + i));
+      else
+	 prog->OutputsWritten |= BITFIELD64_BIT(var->location + offset + i);
+   }
+}
+
+/* Default handler: Mark all the locations in the variable as used. */
+ir_visitor_status
+ir_set_program_inouts_visitor::visit(ir_dereference_variable *ir)
+{
+   if (hash_table_find(this->ht, ir->var) == NULL)
+      return visit_continue;
+
+   if (ir->type->is_array()) {
+      mark(this->prog, ir->var, 0,
+	   ir->type->length * ir->type->fields.array->matrix_columns);
+   } else {
+      mark(this->prog, ir->var, 0, ir->type->matrix_columns);
+   }
+
+   return visit_continue;
+}
+
+ir_visitor_status
+ir_set_program_inouts_visitor::visit_enter(ir_dereference_array *ir)
+{
+   ir_dereference_variable *deref_var;
+   ir_constant *index = ir->array_index->as_constant();
+   deref_var = ir->array->as_dereference_variable();
+   ir_variable *var = NULL;
+
+   /* Check that we're dereferencing a shader in or out */
+   if (deref_var)
+      var = (ir_variable *)hash_table_find(this->ht, deref_var->var);
+
+   if (index && var) {
+      int width = 1;
+
+      if (deref_var->type->is_array() &&
+	  deref_var->type->fields.array->is_matrix()) {
+	 width = deref_var->type->fields.array->matrix_columns;
+      }
+
+      mark(this->prog, var, index->value.i[0] * width, width);
+      return visit_continue_with_parent;
+   }
+
+   return visit_continue;
+}
+
+ir_visitor_status
+ir_set_program_inouts_visitor::visit(ir_variable *ir)
+{
+   if (ir->mode == ir_var_in ||
+       ir->mode == ir_var_out ||
+       ir->mode == ir_var_system_value) {
+      hash_table_insert(this->ht, ir, ir);
+   }
+
+   return visit_continue;
+}
+
+ir_visitor_status
+ir_set_program_inouts_visitor::visit_enter(ir_function_signature *ir)
+{
+   /* We don't want to descend into the function parameters and
+    * consider them as shader inputs or outputs.
+    */
+   visit_list_elements(this, &ir->body);
+   return visit_continue_with_parent;
+}
+
+void
+do_set_program_inouts(exec_list *instructions, struct gl_program *prog)
+{
+   ir_set_program_inouts_visitor v(prog);
+
+   prog->InputsRead = 0;
+   prog->OutputsWritten = 0;
+   prog->SystemValuesRead = 0;
+   visit_list_elements(&v, instructions);
+}
diff --git a/mesalib/src/glsl/ir_variable.cpp b/mesalib/src/glsl/ir_variable.cpp
index b8487694c..e0b6f38f1 100644
--- a/mesalib/src/glsl/ir_variable.cpp
+++ b/mesalib/src/glsl/ir_variable.cpp
@@ -585,6 +585,17 @@ generate_120_vs_variables(exec_list *instructions,
 
 
 static void
+generate_130_uniforms(exec_list *instructions,
+		      struct _mesa_glsl_parse_state *state)
+{
+   glsl_symbol_table *const symtab = state->symbols;
+
+   add_builtin_constant(instructions, symtab, "gl_MaxClipDistances",
+                        state->Const.MaxClipPlanes);
+}
+
+
+static void
 generate_130_vs_variables(exec_list *instructions,
 			  struct _mesa_glsl_parse_state *state)
 {
@@ -595,9 +606,20 @@ generate_130_vs_variables(exec_list *instructions,
 			   & builtin_130_vs_variables[i]);
    }
 
+   generate_130_uniforms(instructions, state);
+
+   /* From the GLSL 1.30 spec, section 7.1 (Vertex Shader Special
+    * Variables):
+    *
+    *   The gl_ClipDistance array is predeclared as unsized and must
+    *   be sized by the shader either redeclaring it with a size or
+    *   indexing it only with integral constant expressions.
+    *
+    * We represent this in Mesa by initially declaring the array as
+    * size 0.
+    */
    const glsl_type *const clip_distance_array_type =
-      glsl_type::get_array_instance(glsl_type::float_type,
-				    state->Const.MaxClipPlanes);
+      glsl_type::get_array_instance(glsl_type::float_type, 0);
 
    /* FINISHME: gl_ClipDistance needs a real location assigned. */
    add_variable(instructions, state->symbols,
@@ -802,9 +824,22 @@ generate_130_fs_variables(exec_list *instructions,
 {
    generate_120_fs_variables(instructions, state);
 
+   generate_130_uniforms(instructions, state);
+
+   /* From the GLSL 1.30 spec, section 7.2 (Fragment Shader Special
+    * Variables):
+    *
+    *   The built-in input variable gl_ClipDistance array contains linearly
+    *   interpolated values for the vertex values written by the vertex shader
+    *   to the gl_ClipDistance vertex output variable. This array must be
+    *   sized in the fragment shader either implicitly or explicitly to be the
+    *   same size as it was sized in the vertex shader.
+    *
+    * In other words, the array must be pre-declared as implicitly sized.  We
+    * represent this in Mesa by initially declaring the array as size 0.
+    */
    const glsl_type *const clip_distance_array_type =
-      glsl_type::get_array_instance(glsl_type::float_type,
-				    state->Const.MaxClipPlanes);
+      glsl_type::get_array_instance(glsl_type::float_type, 0);
 
    /* FINISHME: gl_ClipDistance needs a real location assigned. */
    add_variable(instructions, state->symbols,
diff --git a/mesalib/src/glsl/linker.cpp b/mesalib/src/glsl/linker.cpp
index ba81c59ff..195f58f29 100644
--- a/mesalib/src/glsl/linker.cpp
+++ b/mesalib/src/glsl/linker.cpp
@@ -262,6 +262,25 @@ validate_vertex_shader_executable(struct gl_shader_program *prog,
       return false;
    }
 
+   if (prog->Version >= 130) {
+      /* From section 7.1 (Vertex Shader Special Variables) of the
+       * GLSL 1.30 spec:
+       *
+       *   "It is an error for a shader to statically write both
+       *   gl_ClipVertex and gl_ClipDistance."
+       */
+      find_assignment_visitor clip_vertex("gl_ClipVertex");
+      find_assignment_visitor clip_distance("gl_ClipDistance");
+
+      clip_vertex.run(shader->ir);
+      clip_distance.run(shader->ir);
+      if (clip_vertex.variable_found() && clip_distance.variable_found()) {
+         linker_error(prog, "vertex shader writes to both `gl_ClipVertex' "
+                      "and `gl_ClipDistance'\n");
+         return false;
+      }
+   }
+
    return true;
 }
 
diff --git a/mesalib/src/mapi/mapi/u_compiler.h b/mesalib/src/mapi/mapi/u_compiler.h
index f1752d16f..2b019ed15 100644
--- a/mesalib/src/mapi/mapi/u_compiler.h
+++ b/mesalib/src/mapi/mapi/u_compiler.h
@@ -2,27 +2,30 @@
 #define _U_COMPILER_H_
 
 /* Function inlining */
-#ifndef INLINE
+#ifndef inline
 #  ifdef __cplusplus
-#    define INLINE inline
+     /* C++ supports inline keyword */
 #  elif defined(__GNUC__)
-#    define INLINE __inline__
+#    define inline __inline__
 #  elif defined(_MSC_VER)
-#    define INLINE __inline
+#    define inline __inline
 #  elif defined(__ICL)
-#    define INLINE __inline
+#    define inline __inline
 #  elif defined(__INTEL_COMPILER)
-#    define INLINE inline
+     /* Intel compiler supports inline keyword */
 #  elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100)
-#    define INLINE __inline
+#    define inline __inline
 #  elif defined(__SUNPRO_C) && defined(__C99FEATURES__)
-#    define INLINE inline
-#  elif (__STDC_VERSION__ >= 199901L) /* C99 */
-#    define INLINE inline
+     /* C99 supports inline keyword */
+#  elif (__STDC_VERSION__ >= 199901L)
+     /* C99 supports inline keyword */
 #  else
-#    define INLINE
+#    define inline
 #  endif
 #endif
+#ifndef INLINE
+#  define INLINE inline
+#endif
 
 /* Function visibility */
 #ifndef PUBLIC
diff --git a/mesalib/src/mesa/SConscript b/mesalib/src/mesa/SConscript
index 64b706534..dfc8bd441 100644
--- a/mesalib/src/mesa/SConscript
+++ b/mesalib/src/mesa/SConscript
@@ -72,6 +72,7 @@ main_sources = [
     'main/ffvertex_prog.c',
     'main/fog.c',
     'main/formats.c',
+    'main/format_unpack.c',
     'main/framebuffer.c',
     'main/get.c',
     'main/getstring.c',
diff --git a/mesalib/src/mesa/drivers/common/driverfuncs.c b/mesalib/src/mesa/drivers/common/driverfuncs.c
index 6484d284d..36ed4f892 100644
--- a/mesalib/src/mesa/drivers/common/driverfuncs.c
+++ b/mesalib/src/mesa/drivers/common/driverfuncs.c
@@ -94,7 +94,7 @@ _mesa_init_driver_functions(struct dd_function_table *driver)
    driver->TexSubImage1D = _mesa_store_texsubimage1d;
    driver->TexSubImage2D = _mesa_store_texsubimage2d;
    driver->TexSubImage3D = _mesa_store_texsubimage3d;
-   driver->GetTexImage = _mesa_get_teximage;
+   driver->GetTexImage = _mesa_meta_GetTexImage;
    driver->CopyTexSubImage1D = _mesa_meta_CopyTexSubImage1D;
    driver->CopyTexSubImage2D = _mesa_meta_CopyTexSubImage2D;
    driver->CopyTexSubImage3D = _mesa_meta_CopyTexSubImage3D;
diff --git a/mesalib/src/mesa/drivers/common/meta.c b/mesalib/src/mesa/drivers/common/meta.c
index ad04d369f..2ebcd35bd 100644
--- a/mesalib/src/mesa/drivers/common/meta.c
+++ b/mesalib/src/mesa/drivers/common/meta.c
@@ -59,6 +59,7 @@
 #include "main/stencil.h"
 #include "main/texobj.h"
 #include "main/texenv.h"
+#include "main/texgetimage.h"
 #include "main/teximage.h"
 #include "main/texparam.h"
 #include "main/texstate.h"
@@ -259,6 +260,18 @@ struct gen_mipmap_state
    GLuint FBO;
 };
 
+
+/**
+ * State for texture decompression
+ */
+struct decompress_state
+{
+   GLuint ArrayObj;
+   GLuint VBO, FBO, RBO;
+   GLint Width, Height;
+};
+
+
 #define MAX_META_OPS_DEPTH      2
 /**
  * All per-context meta state.
@@ -278,6 +291,7 @@ struct gl_meta_state
    struct drawpix_state DrawPix;  /**< For _mesa_meta_DrawPixels() */
    struct bitmap_state Bitmap;    /**< For _mesa_meta_Bitmap() */
    struct gen_mipmap_state Mipmap;    /**< For _mesa_meta_GenerateMipmap() */
+   struct decompress_state Decompress;  /**< For texture decompression */
 };
 
 
@@ -2477,6 +2491,160 @@ _mesa_meta_check_generate_mipmap_fallback(struct gl_context *ctx, GLenum target,
 
 
 /**
+ * Compute the texture coordinates for the four vertices of a quad for
+ * drawing a 2D texture image or slice of a cube/3D texture.
+ * \param faceTarget  GL_TEXTURE_1D/2D/3D or cube face name
+ * \param slice  slice of a 1D/2D array texture or 3D texture
+ * \param width  width of the texture image
+ * \param height  height of the texture image
+ * \param coords0/1/2/3  returns the computed texcoords
+ */
+static void
+setup_texture_coords(GLenum faceTarget,
+                     GLint slice,
+                     GLint width,
+                     GLint height,
+                     GLfloat coords0[3],
+                     GLfloat coords1[3],
+                     GLfloat coords2[3],
+                     GLfloat coords3[3])
+{
+   static const GLfloat st[4][2] = {
+      {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f}
+   };
+   GLuint i;
+   GLfloat r;
+
+   switch (faceTarget) {
+   case GL_TEXTURE_1D:
+   case GL_TEXTURE_2D:
+   case GL_TEXTURE_3D:
+   case GL_TEXTURE_2D_ARRAY:
+      if (faceTarget == GL_TEXTURE_3D)
+         r = 1.0F / slice;
+      else if (faceTarget == GL_TEXTURE_2D_ARRAY)
+         r = slice;
+      else
+         r = 0.0F;
+      coords0[0] = 0.0F; /* s */
+      coords0[1] = 0.0F; /* t */
+      coords0[2] = r; /* r */
+      coords1[0] = 1.0F;
+      coords1[1] = 0.0F;
+      coords1[2] = r;
+      coords2[0] = 1.0F;
+      coords2[1] = 1.0F;
+      coords2[2] = r;
+      coords3[0] = 0.0F;
+      coords3[1] = 1.0F;
+      coords3[2] = r;
+      break;
+   case GL_TEXTURE_RECTANGLE_ARB:
+      coords0[0] = 0.0F; /* s */
+      coords0[1] = 0.0F; /* t */
+      coords0[2] = 0.0F; /* r */
+      coords1[0] = width;
+      coords1[1] = 0.0F;
+      coords1[2] = 0.0F;
+      coords2[0] = width;
+      coords2[1] = height;
+      coords2[2] = 0.0F;
+      coords3[0] = 0.0F;
+      coords3[1] = height;
+      coords3[2] = 0.0F;
+      break;
+   case GL_TEXTURE_1D_ARRAY:
+      coords0[0] = 0.0F; /* s */
+      coords0[1] = slice; /* t */
+      coords0[2] = 0.0F; /* r */
+      coords1[0] = 1.0f;
+      coords1[1] = slice;
+      coords1[2] = 0.0F;
+      coords2[0] = 1.0F;
+      coords2[1] = slice;
+      coords2[2] = 0.0F;
+      coords3[0] = 0.0F;
+      coords3[1] = slice;
+      coords3[2] = 0.0F;
+      break;
+
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
+   case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
+   case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
+      /* loop over quad verts */
+      for (i = 0; i < 4; i++) {
+         /* Compute sc = +/-scale and tc = +/-scale.
+          * Not +/-1 to avoid cube face selection ambiguity near the edges,
+          * though that can still sometimes happen with this scale factor...
+          */
+         const GLfloat scale = 0.9999f;
+         const GLfloat sc = (2.0f * st[i][0] - 1.0f) * scale;
+         const GLfloat tc = (2.0f * st[i][1] - 1.0f) * scale;
+         GLfloat *coord;
+
+         switch (i) {
+         case 0:
+            coord = coords0;
+            break;
+         case 1:
+            coord = coords1;
+            break;
+         case 2:
+            coord = coords2;
+            break;
+         case 3:
+            coord = coords3;
+            break;
+         default:
+            assert(0);
+         }
+
+         switch (faceTarget) {
+         case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
+            coord[0] = 1.0f;
+            coord[1] = -tc;
+            coord[2] = -sc;
+            break;
+         case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
+            coord[0] = -1.0f;
+            coord[1] = -tc;
+            coord[2] = sc;
+            break;
+         case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
+            coord[0] = sc;
+            coord[1] = 1.0f;
+            coord[2] = tc;
+            break;
+         case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
+            coord[0] = sc;
+            coord[1] = -1.0f;
+            coord[2] = -tc;
+            break;
+         case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
+            coord[0] = sc;
+            coord[1] = -tc;
+            coord[2] = 1.0f;
+            break;
+         case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
+            coord[0] = -sc;
+            coord[1] = -tc;
+            coord[2] = -1.0f;
+            break;
+         default:
+            assert(0);
+         }
+      }
+      break;
+   default:
+      assert(0 && "unexpected target in meta setup_texture_coords()");
+   }
+}
+
+
+/**
  * Called via ctx->Driver.GenerateMipmap()
  * Note: We don't yet support 3D textures, 1D/2D array textures or texture
  * borders.
@@ -2487,7 +2655,7 @@ _mesa_meta_GenerateMipmap(struct gl_context *ctx, GLenum target,
 {
    struct gen_mipmap_state *mipmap = &ctx->Meta->Mipmap;
    struct vertex {
-      GLfloat x, y, s, t, r;
+      GLfloat x, y, tex[3];
    };
    struct vertex verts[4];
    const GLuint baseLevel = texObj->BaseLevel;
@@ -2503,7 +2671,8 @@ _mesa_meta_GenerateMipmap(struct gl_context *ctx, GLenum target,
    const GLuint original_active_unit = ctx->Texture.CurrentUnit;
    GLenum faceTarget;
    GLuint dstLevel;
-   GLuint border = 0;
+   const GLuint border = 0;
+   const GLint slice = 0;
 
    if (_mesa_meta_check_generate_mipmap_fallback(ctx, target, texObj)) {
       _mesa_generate_mipmap(ctx, target, texObj);
@@ -2539,7 +2708,7 @@ _mesa_meta_GenerateMipmap(struct gl_context *ctx, GLenum target,
 
       /* setup vertex arrays */
       _mesa_VertexPointer(2, GL_FLOAT, sizeof(struct vertex), OFFSET(x));
-      _mesa_TexCoordPointer(3, GL_FLOAT, sizeof(struct vertex), OFFSET(s));
+      _mesa_TexCoordPointer(3, GL_FLOAT, sizeof(struct vertex), OFFSET(tex));
       _mesa_EnableClientState(GL_VERTEX_ARRAY);
       _mesa_EnableClientState(GL_TEXTURE_COORD_ARRAY);
    }
@@ -2562,98 +2731,27 @@ _mesa_meta_GenerateMipmap(struct gl_context *ctx, GLenum target,
 
    _mesa_set_enable(ctx, target, GL_TRUE);
 
-   /* setup texcoords once (XXX what about border?) */
-   switch (faceTarget) {
-   case GL_TEXTURE_1D:
-   case GL_TEXTURE_2D:
-      verts[0].s = 0.0F;
-      verts[0].t = 0.0F;
-      verts[0].r = 0.0F;
-      verts[1].s = 1.0F;
-      verts[1].t = 0.0F;
-      verts[1].r = 0.0F;
-      verts[2].s = 1.0F;
-      verts[2].t = 1.0F;
-      verts[2].r = 0.0F;
-      verts[3].s = 0.0F;
-      verts[3].t = 1.0F;
-      verts[3].r = 0.0F;
-      break;
-   case GL_TEXTURE_3D:
-      abort();
-      break;
-   default:
-      /* cube face */
-      {
-         static const GLfloat st[4][2] = {
-            {0.0f, 0.0f}, {1.0f, 0.0f}, {1.0f, 1.0f}, {0.0f, 1.0f}
-         };
-         GLuint i;
-
-         /* loop over quad verts */
-         for (i = 0; i < 4; i++) {
-            /* Compute sc = +/-scale and tc = +/-scale.
-             * Not +/-1 to avoid cube face selection ambiguity near the edges,
-             * though that can still sometimes happen with this scale factor...
-             */
-            const GLfloat scale = 0.9999f;
-            const GLfloat sc = (2.0f * st[i][0] - 1.0f) * scale;
-            const GLfloat tc = (2.0f * st[i][1] - 1.0f) * scale;
-
-            switch (faceTarget) {
-            case GL_TEXTURE_CUBE_MAP_POSITIVE_X:
-               verts[i].s = 1.0f;
-               verts[i].t = -tc;
-               verts[i].r = -sc;
-               break;
-            case GL_TEXTURE_CUBE_MAP_NEGATIVE_X:
-               verts[i].s = -1.0f;
-               verts[i].t = -tc;
-               verts[i].r = sc;
-               break;
-            case GL_TEXTURE_CUBE_MAP_POSITIVE_Y:
-               verts[i].s = sc;
-               verts[i].t = 1.0f;
-               verts[i].r = tc;
-               break;
-            case GL_TEXTURE_CUBE_MAP_NEGATIVE_Y:
-               verts[i].s = sc;
-               verts[i].t = -1.0f;
-               verts[i].r = -tc;
-               break;
-            case GL_TEXTURE_CUBE_MAP_POSITIVE_Z:
-               verts[i].s = sc;
-               verts[i].t = -tc;
-               verts[i].r = 1.0f;
-               break;
-            case GL_TEXTURE_CUBE_MAP_NEGATIVE_Z:
-               verts[i].s = -sc;
-               verts[i].t = -tc;
-               verts[i].r = -1.0f;
-               break;
-            default:
-               assert(0);
-            }
-         }
-      }
-   }
-
-   _mesa_set_enable(ctx, target, GL_TRUE);
+   /* setup texcoords (XXX what about border?) */
+   setup_texture_coords(faceTarget,
+                        0.0, 0.0, /* width, height never used here */
+                        slice,
+                        verts[0].tex,
+                        verts[1].tex,
+                        verts[2].tex,
+                        verts[3].tex);
 
    /* setup vertex positions */
-   {
-      verts[0].x = 0.0F;
-      verts[0].y = 0.0F;
-      verts[1].x = 1.0F;
-      verts[1].y = 0.0F;
-      verts[2].x = 1.0F;
-      verts[2].y = 1.0F;
-      verts[3].x = 0.0F;
-      verts[3].y = 1.0F;
+   verts[0].x = 0.0F;
+   verts[0].y = 0.0F;
+   verts[1].x = 1.0F;
+   verts[1].y = 0.0F;
+   verts[2].x = 1.0F;
+   verts[2].y = 1.0F;
+   verts[3].x = 0.0F;
+   verts[3].y = 1.0F;
       
-      /* upload new vertex data */
-      _mesa_BufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, sizeof(verts), verts);
-   }
+   /* upload new vertex data */
+   _mesa_BufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, sizeof(verts), verts);
 
    /* setup projection matrix */
    _mesa_MatrixMode(GL_PROJECTION);
@@ -2939,3 +3037,213 @@ _mesa_meta_CopyTexSubImage3D(struct gl_context *ctx, GLenum target, GLint level,
    copy_tex_sub_image(ctx, 3, target, level, xoffset, yoffset, zoffset,
                       x, y, width, height);
 }
+
+
+/**
+ * Decompress a texture image by drawing a quad with the compressed
+ * texture and reading the pixels out of the color buffer.
+ * \param slice  which slice of a 3D texture or layer of a 1D/2D texture
+ * \param destFormat  format, ala glReadPixels
+ * \param destType  type, ala glReadPixels
+ * \param dest  destination buffer
+ * \param destRowLength  dest image rowLength (ala GL_PACK_ROW_LENGTH)
+ */
+static void
+decompress_texture_image(struct gl_context *ctx,
+                         struct gl_texture_image *texImage,
+                         GLuint slice,
+                         GLenum destFormat, GLenum destType,
+                         GLvoid *dest, GLint destRowLength)
+{
+   struct decompress_state *decompress = &ctx->Meta->Decompress;
+   struct gl_texture_object *texObj = texImage->TexObject;
+   const GLint width = texImage->Width;
+   const GLint height = texImage->Height;
+   const GLenum target = texObj->Target;
+   GLenum faceTarget;
+   struct vertex {
+      GLfloat x, y, tex[3];
+   };
+   struct vertex verts[4];
+   GLuint fboDrawSave, fboReadSave;
+
+   if (slice > 0) {
+      assert(target == GL_TEXTURE_3D ||
+             target == GL_TEXTURE_2D_ARRAY);
+   }
+
+   if (target == GL_TEXTURE_CUBE_MAP) {
+      faceTarget = GL_TEXTURE_CUBE_MAP_POSITIVE_X + texImage->Face;
+   }
+   else {
+      faceTarget = target;
+   }
+
+   /* save fbo bindings (not saved by _mesa_meta_begin()) */
+   fboDrawSave = ctx->DrawBuffer->Name;
+   fboReadSave = ctx->ReadBuffer->Name;
+
+   _mesa_meta_begin(ctx, MESA_META_ALL);
+
+   /* Create/bind FBO/renderbuffer */
+   if (decompress->FBO == 0) {
+      _mesa_GenFramebuffersEXT(1, &decompress->FBO);
+      _mesa_GenRenderbuffersEXT(1, &decompress->RBO);
+      _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, decompress->FBO);
+      _mesa_BindRenderbufferEXT(GL_RENDERBUFFER_EXT, decompress->RBO);
+      _mesa_FramebufferRenderbufferEXT(GL_FRAMEBUFFER_EXT,
+                                       GL_COLOR_ATTACHMENT0_EXT,
+                                       GL_RENDERBUFFER_EXT,
+                                       decompress->RBO);
+   }
+   else {
+      _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, decompress->FBO);
+   }
+
+   /* alloc dest surface */
+   if (width != decompress->Width || height != decompress->Height) {
+      _mesa_RenderbufferStorageEXT(GL_RENDERBUFFER_EXT, GL_RGBA,
+                                   width, height);
+      decompress->Width = width;
+      decompress->Height = height;
+   }
+
+   /* setup VBO data */
+   if (decompress->ArrayObj == 0) {
+      /* create vertex array object */
+      _mesa_GenVertexArrays(1, &decompress->ArrayObj);
+      _mesa_BindVertexArray(decompress->ArrayObj);
+
+      /* create vertex array buffer */
+      _mesa_GenBuffersARB(1, &decompress->VBO);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, decompress->VBO);
+      _mesa_BufferDataARB(GL_ARRAY_BUFFER_ARB, sizeof(verts),
+                          NULL, GL_DYNAMIC_DRAW_ARB);
+
+      /* setup vertex arrays */
+      _mesa_VertexPointer(2, GL_FLOAT, sizeof(struct vertex), OFFSET(x));
+      _mesa_TexCoordPointer(3, GL_FLOAT, sizeof(struct vertex), OFFSET(tex));
+      _mesa_EnableClientState(GL_VERTEX_ARRAY);
+      _mesa_EnableClientState(GL_TEXTURE_COORD_ARRAY);
+   }
+   else {
+      _mesa_BindVertexArray(decompress->ArrayObj);
+      _mesa_BindBufferARB(GL_ARRAY_BUFFER_ARB, decompress->VBO);
+   }
+
+   setup_texture_coords(faceTarget, slice, width, height,
+                        verts[0].tex,
+                        verts[1].tex,
+                        verts[2].tex,
+                        verts[3].tex);
+
+   /* setup vertex positions */
+   verts[0].x = 0.0F;
+   verts[0].y = 0.0F;
+   verts[1].x = width;
+   verts[1].y = 0.0F;
+   verts[2].x = width;
+   verts[2].y = height;
+   verts[3].x = 0.0F;
+   verts[3].y = height;
+
+   /* upload new vertex data */
+   _mesa_BufferSubDataARB(GL_ARRAY_BUFFER_ARB, 0, sizeof(verts), verts);
+
+   /* setup texture state */
+   _mesa_BindTexture(target, texObj->Name);
+   _mesa_Enable(target);
+
+   {
+      /* save texture object state */
+      const GLenum minFilterSave = texObj->Sampler.MinFilter;
+      const GLenum magFilterSave = texObj->Sampler.MagFilter;
+      const GLint baseLevelSave = texObj->BaseLevel;
+      const GLint maxLevelSave = texObj->MaxLevel;
+      const GLenum wrapSSave = texObj->Sampler.WrapS;
+      const GLenum wrapTSave = texObj->Sampler.WrapT;
+      const GLenum srgbSave = texObj->Sampler.sRGBDecode;
+
+      /* restrict sampling to the texture level of interest */
+      _mesa_TexParameteri(target, GL_TEXTURE_BASE_LEVEL, texImage->Level);
+      _mesa_TexParameteri(target, GL_TEXTURE_MAX_LEVEL, texImage->Level);
+      /* nearest filtering */
+      _mesa_TexParameteri(target, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+      _mesa_TexParameteri(target, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+
+      /* No sRGB decode or encode.*/
+      if (ctx->Extensions.EXT_texture_sRGB_decode) {
+         _mesa_TexParameteri(target, GL_TEXTURE_SRGB_DECODE_EXT,
+                             GL_SKIP_DECODE_EXT);
+      }
+      if (ctx->Extensions.EXT_framebuffer_sRGB) {
+         _mesa_Disable(GL_FRAMEBUFFER_SRGB_EXT);
+      }
+
+      /* render quad w/ texture into renderbuffer */
+      _mesa_DrawArrays(GL_TRIANGLE_FAN, 0, 4);
+      
+      /* Restore texture object state, the texture binding will
+       * be restored by _mesa_meta_end().
+       */
+      _mesa_TexParameteri(target, GL_TEXTURE_MIN_FILTER, minFilterSave);
+      _mesa_TexParameteri(target, GL_TEXTURE_MAG_FILTER, magFilterSave);
+      if (target != GL_TEXTURE_RECTANGLE_ARB) {
+         _mesa_TexParameteri(target, GL_TEXTURE_BASE_LEVEL, baseLevelSave);
+         _mesa_TexParameteri(target, GL_TEXTURE_MAX_LEVEL, maxLevelSave);
+      }
+      _mesa_TexParameteri(target, GL_TEXTURE_WRAP_S, wrapSSave);
+      _mesa_TexParameteri(target, GL_TEXTURE_WRAP_T, wrapTSave);
+      if (ctx->Extensions.EXT_texture_sRGB_decode) {
+         _mesa_TexParameteri(target, GL_TEXTURE_SRGB_DECODE_EXT, srgbSave);
+      }
+   }
+
+   /* read pixels from renderbuffer */
+   ctx->Pack.RowLength = destRowLength;
+   _mesa_ReadPixels(0, 0, width, height, destFormat, destType, dest);
+
+   _mesa_meta_end(ctx);
+
+   /* restore fbo bindings */
+   if (fboDrawSave == fboReadSave) {
+      _mesa_BindFramebufferEXT(GL_FRAMEBUFFER_EXT, fboDrawSave);
+   }
+   else {
+      _mesa_BindFramebufferEXT(GL_DRAW_FRAMEBUFFER_EXT, fboDrawSave);
+      _mesa_BindFramebufferEXT(GL_READ_FRAMEBUFFER_EXT, fboReadSave);
+   }
+}
+
+
+/**
+ * This is just a wrapper around _mesa_get_tex_image() and
+ * decompress_texture_image().  Meta functions should not be directly called
+ * from core Mesa.
+ */
+void
+_mesa_meta_GetTexImage(struct gl_context *ctx, GLenum target, GLint level,
+                       GLenum format, GLenum type, GLvoid *pixels,
+                       struct gl_texture_object *texObj,
+                       struct gl_texture_image *texImage)
+{
+   /* We can only use the decompress-with-blit method here if the texels are
+    * unsigned, normalized values.  We could handle signed and unnormalized 
+    * with floating point renderbuffers...
+    */
+   if (_mesa_is_format_compressed(texImage->TexFormat) &&
+       _mesa_get_format_datatype(texImage->TexFormat)
+       == GL_UNSIGNED_NORMALIZED) {
+      const GLuint slice = 0; /* only 2D compressed textures for now */
+      /* Need to unlock the texture here to prevent deadlock... */
+      _mesa_unlock_texture(ctx, texObj);
+      decompress_texture_image(ctx, texImage, slice, format, type, pixels,
+                               ctx->Pack.RowLength);
+      /* ... and relock it */
+      _mesa_lock_texture(ctx, texObj);
+   }
+   else {
+      _mesa_get_teximage(ctx, target, level, format, type, pixels,
+                          texObj, texImage);
+   }
+}
diff --git a/mesalib/src/mesa/drivers/common/meta.h b/mesalib/src/mesa/drivers/common/meta.h
index 9a92613c4..9d634ae56 100644
--- a/mesalib/src/mesa/drivers/common/meta.h
+++ b/mesalib/src/mesa/drivers/common/meta.h
@@ -121,4 +121,11 @@ _mesa_meta_CopyTexSubImage3D(struct gl_context *ctx, GLenum target, GLint level,
                              GLint x, GLint y,
                              GLsizei width, GLsizei height);
 
+extern void
+_mesa_meta_GetTexImage(struct gl_context *ctx, GLenum target, GLint level,
+                       GLenum format, GLenum type, GLvoid *pixels,
+                       struct gl_texture_object *texObj,
+                       struct gl_texture_image *texImage);
+
+
 #endif /* META_H */
diff --git a/mesalib/src/mesa/main/compiler.h b/mesalib/src/mesa/main/compiler.h
index 8ed1c6fa6..89d6cda91 100644
--- a/mesalib/src/mesa/main/compiler.h
+++ b/mesalib/src/mesa/main/compiler.h
@@ -114,29 +114,30 @@ extern "C" {
 /**
  * Function inlining
  */
-#ifndef INLINE
-#  if defined(__GNUC__)
-#    define INLINE __inline__
-#  elif defined(__MSC__)
-#    define INLINE __inline
+#ifndef inline
+#  ifdef __cplusplus
+     /* C++ supports inline keyword */
+#  elif defined(__GNUC__)
+#    define inline __inline__
 #  elif defined(_MSC_VER)
-#    define INLINE __inline
+#    define inline __inline
 #  elif defined(__ICL)
-#    define INLINE __inline
+#    define inline __inline
 #  elif defined(__INTEL_COMPILER)
-#    define INLINE inline
+     /* Intel compiler supports inline keyword */
 #  elif defined(__WATCOMC__) && (__WATCOMC__ >= 1100)
-#    define INLINE __inline
+#    define inline __inline
 #  elif defined(__SUNPRO_C) && defined(__C99FEATURES__)
-#    define INLINE inline
-#    define __inline inline
-#    define __inline__ inline
-#  elif (__STDC_VERSION__ >= 199901L) /* C99 */
-#    define INLINE inline
+     /* C99 supports inline keyword */
+#  elif (__STDC_VERSION__ >= 199901L)
+     /* C99 supports inline keyword */
 #  else
-#    define INLINE
+#    define inline
 #  endif
 #endif
+#ifndef INLINE
+#  define INLINE inline
+#endif
 
 
 /**
diff --git a/mesalib/src/mesa/main/format_unpack.c b/mesalib/src/mesa/main/format_unpack.c
new file mode 100644
index 000000000..c5146f72d
--- /dev/null
+++ b/mesalib/src/mesa/main/format_unpack.c
@@ -0,0 +1,1494 @@
+/*
+ * Mesa 3-D graphics library
+ *
+ * Copyright (c) 2011 VMware, Inc.
+ *
+ * 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
+ * THE AUTHORS 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 "colormac.h"
+#include "format_unpack.h"
+#include "macros.h"
+#include "../../gallium/auxiliary/util/u_format_rgb9e5.h"
+#include "../../gallium/auxiliary/util/u_format_r11g11b10f.h"
+
+
+/**
+ * Convert an 8-bit sRGB value from non-linear space to a
+ * linear RGB value in [0, 1].
+ * Implemented with a 256-entry lookup table.
+ */
+static INLINE GLfloat
+nonlinear_to_linear(GLubyte cs8)
+{
+   static GLfloat table[256];
+   static GLboolean tableReady = GL_FALSE;
+   if (!tableReady) {
+      /* compute lookup table now */
+      GLuint i;
+      for (i = 0; i < 256; i++) {
+         const GLfloat cs = UBYTE_TO_FLOAT(i);
+         if (cs <= 0.04045) {
+            table[i] = cs / 12.92f;
+         }
+         else {
+            table[i] = (GLfloat) pow((cs + 0.055) / 1.055, 2.4);
+         }
+      }
+      tableReady = GL_TRUE;
+   }
+   return table[cs8];
+}
+
+
+typedef void (*unpack_rgba_func)(const void *src, GLfloat dst[4]);
+
+
+static void
+unpack_RGBA8888(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = UBYTE_TO_FLOAT( (s >> 24)        );
+   dst[GCOMP] = UBYTE_TO_FLOAT( (s >> 16) & 0xff );
+   dst[BCOMP] = UBYTE_TO_FLOAT( (s >>  8) & 0xff );
+   dst[ACOMP] = UBYTE_TO_FLOAT( (s      ) & 0xff );
+}
+
+static void
+unpack_RGBA8888_REV(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = UBYTE_TO_FLOAT( (s      ) & 0xff );
+   dst[GCOMP] = UBYTE_TO_FLOAT( (s >>  8) & 0xff );
+   dst[BCOMP] = UBYTE_TO_FLOAT( (s >> 16) & 0xff );
+   dst[ACOMP] = UBYTE_TO_FLOAT( (s >> 24)        );
+}
+
+static void
+unpack_ARGB8888(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = UBYTE_TO_FLOAT( (s >> 16) & 0xff );
+   dst[GCOMP] = UBYTE_TO_FLOAT( (s >>  8) & 0xff );
+   dst[BCOMP] = UBYTE_TO_FLOAT( (s      ) & 0xff );
+   dst[ACOMP] = UBYTE_TO_FLOAT( (s >> 24)        );
+}
+
+static void
+unpack_ARGB8888_REV(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = UBYTE_TO_FLOAT( (s >>  8) & 0xff );
+   dst[GCOMP] = UBYTE_TO_FLOAT( (s >> 16) & 0xff );
+   dst[BCOMP] = UBYTE_TO_FLOAT( (s >> 24)        );
+   dst[ACOMP] = UBYTE_TO_FLOAT( (s      ) & 0xff );
+}
+
+static void
+unpack_XRGB8888(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = UBYTE_TO_FLOAT( (s >> 16) & 0xff );
+   dst[GCOMP] = UBYTE_TO_FLOAT( (s >>  8) & 0xff );
+   dst[BCOMP] = UBYTE_TO_FLOAT( (s      ) & 0xff );
+   dst[ACOMP] = 1.0f;
+}
+
+static void
+unpack_XRGB8888_REV(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = UBYTE_TO_FLOAT( (s >>  8) & 0xff );
+   dst[GCOMP] = UBYTE_TO_FLOAT( (s >> 16) & 0xff );
+   dst[BCOMP] = UBYTE_TO_FLOAT( (s >> 24)        );
+   dst[ACOMP] = 1.0f;
+}
+
+static void
+unpack_RGB888(const void *src, GLfloat dst[4])
+{
+   const GLubyte *s = (const GLubyte *) src;
+   dst[RCOMP] = UBYTE_TO_FLOAT( s[2] );
+   dst[GCOMP] = UBYTE_TO_FLOAT( s[1] );
+   dst[BCOMP] = UBYTE_TO_FLOAT( s[0] );
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_BGR888(const void *src, GLfloat dst[4])
+{
+   const GLubyte *s = (const GLubyte *) src;
+   dst[RCOMP] = UBYTE_TO_FLOAT( s[0] );
+   dst[GCOMP] = UBYTE_TO_FLOAT( s[1] );
+   dst[BCOMP] = UBYTE_TO_FLOAT( s[2] );
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_RGB565(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = ((s >> 11) & 0x1f) * (1.0F / 31.0F);
+   dst[GCOMP] = ((s >> 5 ) & 0x3f) * (1.0F / 63.0F);
+   dst[BCOMP] = ((s      ) & 0x1f) * (1.0F / 31.0F);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_RGB565_REV(const void *src, GLfloat dst[4])
+{
+   GLushort s = *((const GLushort *) src);
+   s = (s >> 8) | (s << 8); /* byte swap */
+   dst[RCOMP] = UBYTE_TO_FLOAT( ((s >> 8) & 0xf8) | ((s >> 13) & 0x7) );
+   dst[GCOMP] = UBYTE_TO_FLOAT( ((s >> 3) & 0xfc) | ((s >>  9) & 0x3) );
+   dst[BCOMP] = UBYTE_TO_FLOAT( ((s << 3) & 0xf8) | ((s >>  2) & 0x7) );
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_ARGB4444(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = ((s >>  8) & 0xf) * (1.0F / 15.0F);
+   dst[GCOMP] = ((s >>  4) & 0xf) * (1.0F / 15.0F);
+   dst[BCOMP] = ((s      ) & 0xf) * (1.0F / 15.0F);
+   dst[ACOMP] = ((s >> 12) & 0xf) * (1.0F / 15.0F);
+}
+
+static void
+unpack_ARGB4444_REV(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = ((s      ) & 0xf) * (1.0F / 15.0F);
+   dst[GCOMP] = ((s >> 12) & 0xf) * (1.0F / 15.0F);
+   dst[BCOMP] = ((s >>  8) & 0xf) * (1.0F / 15.0F);
+   dst[ACOMP] = ((s >>  4) & 0xf) * (1.0F / 15.0F);
+}
+
+static void
+unpack_RGBA5551(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = ((s >> 11) & 0x1f) * (1.0F / 31.0F);
+   dst[GCOMP] = ((s >>  6) & 0x1f) * (1.0F / 31.0F);
+   dst[BCOMP] = ((s >>  1) & 0x1f) * (1.0F / 31.0F);
+   dst[ACOMP] = ((s      ) & 0x01) * 1.0F;
+}
+
+static void
+unpack_ARGB1555(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = ((s >> 10) & 0x1f) * (1.0F / 31.0F);
+   dst[GCOMP] = ((s >>  5) & 0x1f) * (1.0F / 31.0F);
+   dst[BCOMP] = ((s >>  0) & 0x1f) * (1.0F / 31.0F);
+   dst[ACOMP] = ((s >> 15) & 0x01) * 1.0F;
+}
+
+static void
+unpack_ARGB1555_REV(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = UBYTE_TO_FLOAT( ((s >>  7) & 0xf8) | ((s >> 12) & 0x7) );
+   dst[GCOMP] = UBYTE_TO_FLOAT( ((s >>  2) & 0xf8) | ((s >>  7) & 0x7) );
+   dst[BCOMP] = UBYTE_TO_FLOAT( ((s <<  3) & 0xf8) | ((s >>  2) & 0x7) );
+   dst[ACOMP] = UBYTE_TO_FLOAT( ((s >> 15) & 0x01) * 255 );
+}
+
+static void
+unpack_AL44(const void *src, GLfloat dst[4])
+{
+   const GLubyte s = *((const GLubyte *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = (s & 0xf) * (1.0F / 15.0F);
+   dst[ACOMP] = ((s >> 4) & 0xf) * (1.0F / 15.0F);
+}
+
+static void
+unpack_AL88(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = 
+   dst[GCOMP] = 
+   dst[BCOMP] = UBYTE_TO_FLOAT( s & 0xff );
+   dst[ACOMP] = UBYTE_TO_FLOAT( s >> 8 );
+}
+
+static void
+unpack_AL88_REV(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = 
+   dst[GCOMP] = 
+   dst[BCOMP] = UBYTE_TO_FLOAT( s >> 8 );
+   dst[ACOMP] = UBYTE_TO_FLOAT( s & 0xff );
+}
+
+static void
+unpack_AL1616(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = USHORT_TO_FLOAT( s & 0xffff );
+   dst[ACOMP] = USHORT_TO_FLOAT( s >> 16 );
+}
+
+static void
+unpack_AL1616_REV(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = USHORT_TO_FLOAT( s >> 16 );
+   dst[ACOMP] = USHORT_TO_FLOAT( s & 0xffff );
+}
+
+static void
+unpack_RGB332(const void *src, GLfloat dst[4])
+{
+   const GLubyte s = *((const GLubyte *) src);
+   dst[RCOMP] = ((s >> 5) & 0x7) * (1.0F / 7.0F);
+   dst[GCOMP] = ((s >> 2) & 0x7) * (1.0F / 7.0F);
+   dst[BCOMP] = ((s     ) & 0x3) * (1.0F / 3.0F);
+   dst[ACOMP] = 1.0F;
+}
+
+
+static void
+unpack_A8(const void *src, GLfloat dst[4])
+{
+   const GLubyte s = *((const GLubyte *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = UBYTE_TO_FLOAT(s);
+}
+
+static void
+unpack_A16(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = USHORT_TO_FLOAT(s);
+}
+
+static void
+unpack_L8(const void *src, GLfloat dst[4])
+{
+   const GLubyte s = *((const GLubyte *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = UBYTE_TO_FLOAT(s);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_L16(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = USHORT_TO_FLOAT(s);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_I8(const void *src, GLfloat dst[4])
+{
+   const GLubyte s = *((const GLubyte *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] =
+   dst[ACOMP] = UBYTE_TO_FLOAT(s);
+}
+
+static void
+unpack_I16(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] =
+   dst[ACOMP] = USHORT_TO_FLOAT(s);
+}
+
+static void
+unpack_YCBCR(const void *src, GLfloat dst[4])
+{
+   const GLuint i = 0;
+   const GLushort *src0 = (const GLushort *) src;
+   const GLushort *src1 = src0 + 1;                               /* odd */
+   const GLubyte y0 = (*src0 >> 8) & 0xff;  /* luminance */
+   const GLubyte cb = *src0 & 0xff;         /* chroma U */
+   const GLubyte y1 = (*src1 >> 8) & 0xff;  /* luminance */
+   const GLubyte cr = *src1 & 0xff;         /* chroma V */
+   const GLubyte y = (i & 1) ? y1 : y0;     /* choose even/odd luminance */
+   GLfloat r = 1.164F * (y - 16) + 1.596F * (cr - 128);
+   GLfloat g = 1.164F * (y - 16) - 0.813F * (cr - 128) - 0.391F * (cb - 128);
+   GLfloat b = 1.164F * (y - 16) + 2.018F * (cb - 128);
+   r *= (1.0F / 255.0F);
+   g *= (1.0F / 255.0F);
+   b *= (1.0F / 255.0F);
+   dst[RCOMP] = CLAMP(r, 0.0F, 1.0F);
+   dst[GCOMP] = CLAMP(g, 0.0F, 1.0F);
+   dst[BCOMP] = CLAMP(b, 0.0F, 1.0F);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_YCBCR_REV(const void *src, GLfloat dst[4])
+{
+   const GLuint i = 0;
+   const GLushort *src0 = (const GLushort *) src;
+   const GLushort *src1 = src0 + 1;                               /* odd */
+   const GLubyte y0 = *src0 & 0xff;         /* luminance */
+   const GLubyte cr = (*src0 >> 8) & 0xff;  /* chroma V */
+   const GLubyte y1 = *src1 & 0xff;         /* luminance */
+   const GLubyte cb = (*src1 >> 8) & 0xff;  /* chroma U */
+   const GLubyte y = (i & 1) ? y1 : y0;     /* choose even/odd luminance */
+   GLfloat r = 1.164F * (y - 16) + 1.596F * (cr - 128);
+   GLfloat g = 1.164F * (y - 16) - 0.813F * (cr - 128) - 0.391F * (cb - 128);
+   GLfloat b = 1.164F * (y - 16) + 2.018F * (cb - 128);
+   r *= (1.0F / 255.0F);
+   g *= (1.0F / 255.0F);
+   b *= (1.0F / 255.0F);
+   dst[RCOMP] = CLAMP(r, 0.0F, 1.0F);
+   dst[GCOMP] = CLAMP(g, 0.0F, 1.0F);
+   dst[BCOMP] = CLAMP(b, 0.0F, 1.0F);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_R8(const void *src, GLfloat dst[4])
+{
+   const GLubyte s = *((const GLubyte *) src);
+   dst[0] = UBYTE_TO_FLOAT(s);
+   dst[1] = dst[2] = 0.0F;
+   dst[3] = 1.0F;
+}
+
+static void
+unpack_RG88(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = UBYTE_TO_FLOAT( s & 0xff );
+   dst[GCOMP] = UBYTE_TO_FLOAT( s >> 8 );
+   dst[BCOMP] = 0.0;
+   dst[ACOMP] = 1.0;
+}
+
+static void
+unpack_RG88_REV(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = UBYTE_TO_FLOAT( s & 0xff );
+   dst[GCOMP] = UBYTE_TO_FLOAT( s >> 8 );
+   dst[BCOMP] = 0.0;
+   dst[ACOMP] = 1.0;
+}
+
+static void
+unpack_R16(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = USHORT_TO_FLOAT(s);
+   dst[GCOMP] = 0.0;
+   dst[BCOMP] = 0.0;
+   dst[ACOMP] = 1.0;
+}
+
+static void
+unpack_RG1616(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = USHORT_TO_FLOAT( s & 0xffff );
+   dst[GCOMP] = USHORT_TO_FLOAT( s >> 16 );
+   dst[BCOMP] = 0.0;
+   dst[ACOMP] = 1.0;
+}
+
+static void
+unpack_RG1616_REV(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = USHORT_TO_FLOAT( s >> 16 );
+   dst[GCOMP] = USHORT_TO_FLOAT( s & 0xffff );
+   dst[BCOMP] = 0.0;
+   dst[ACOMP] = 1.0;
+}
+
+static void
+unpack_ARGB2101010(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = ((s >> 20) & 0x3ff) * (1.0F / 1023.0F);
+   dst[GCOMP] = ((s >> 10) & 0x3ff) * (1.0F / 1023.0F);
+   dst[BCOMP] = ((s >>  0) & 0x3ff) * (1.0F / 1023.0F);
+   dst[ACOMP] = ((s >> 30) & 0x03) * (1.0F / 3.0F);
+}
+
+
+static void
+unpack_Z24_S8(const void *src, GLfloat dst[4])
+{
+   /* only return Z, not stencil data */
+   const GLuint s = *((const GLuint *) src);
+   const GLfloat scale = 1.0F / (GLfloat) 0xffffff;
+   dst[0] = dst[1] = dst[2] = (s >> 8) * scale;
+   dst[3] = 1.0F;
+   ASSERT(dst[0] >= 0.0F);
+   ASSERT(dst[0] <= 1.0F);
+}
+
+static void
+unpack_S8_Z24(const void *src, GLfloat dst[4])
+{
+   /* only return Z, not stencil data */
+   const GLuint s = *((const GLuint *) src);
+   const GLfloat scale = 1.0F / (GLfloat) 0xffffff;
+   dst[0] = dst[1] = dst[2] = (s & 0x00ffffff) * scale;
+   dst[3] = 1.0F;
+   ASSERT(dst[0] >= 0.0F);
+   ASSERT(dst[0] <= 1.0F);
+}
+
+static void
+unpack_Z16(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[0] = dst[1] = dst[2] = s * (1.0F / 65535.0F);
+   dst[3] = 1.0F;
+}
+
+static void
+unpack_X8_Z24(const void *src, GLfloat dst[4])
+{
+   unpack_S8_Z24(src, dst);
+}
+
+static void
+unpack_Z24_X8(const void *src, GLfloat dst[4])
+{
+   unpack_Z24_S8(src, dst);
+}
+
+static void
+unpack_Z32(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[0] = dst[1] = dst[2] = s * (1.0F / 0xffffffff);
+   dst[3] = 1.0F;
+}
+
+static void
+unpack_Z32_FLOAT(const void *src, GLfloat dst[4])
+{
+   const GLfloat s = *((const GLfloat *) src);
+   dst[0] = dst[1] = dst[2] = s;
+   dst[3] = 1.0F;
+}
+
+static void
+unpack_Z32_FLOAT_X24S8(const void *src, GLfloat dst[4])
+{
+   const GLfloat s = *((const GLfloat *) src);
+   dst[0] = dst[1] = dst[2] = s;
+   dst[3] = 1.0F;
+}
+
+
+static void
+unpack_S8(const void *src, GLfloat dst[4])
+{
+   /* should never be used */
+   dst[0] = dst[1] = dst[2] = 0.0F;
+   dst[3] = 1.0F;
+}
+
+
+static void
+unpack_SRGB8(const void *src, GLfloat dst[4])
+{
+   const GLubyte *s = (const GLubyte *) src;
+   dst[RCOMP] = nonlinear_to_linear(s[2]);
+   dst[GCOMP] = nonlinear_to_linear(s[1]);
+   dst[BCOMP] = nonlinear_to_linear(s[0]);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_SRGBA8(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = nonlinear_to_linear( (s >> 24) );
+   dst[GCOMP] = nonlinear_to_linear( (s >> 16) & 0xff );
+   dst[BCOMP] = nonlinear_to_linear( (s >>  8) & 0xff );
+   dst[ACOMP] = UBYTE_TO_FLOAT( (s      ) & 0xff ); /* linear! */
+}
+
+static void
+unpack_SARGB8(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = nonlinear_to_linear( (s >> 16) & 0xff );
+   dst[GCOMP] = nonlinear_to_linear( (s >>  8) & 0xff );
+   dst[BCOMP] = nonlinear_to_linear( (s      ) & 0xff );
+   dst[ACOMP] = UBYTE_TO_FLOAT( (s >> 24) ); /* linear! */
+}
+
+static void
+unpack_SL8(const void *src, GLfloat dst[4])
+{
+   const GLubyte s = *((const GLubyte *) src);
+   dst[RCOMP] = 
+   dst[GCOMP] = 
+   dst[BCOMP] = nonlinear_to_linear(s);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_SLA8(const void *src, GLfloat dst[4])
+{
+   const GLubyte *s = (const GLubyte *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = nonlinear_to_linear(s[0]);
+   dst[ACOMP] = UBYTE_TO_FLOAT(s[1]); /* linear */
+}
+
+static void
+unpack_SRGB_DXT1(const void *src, GLfloat dst[4])
+{
+}
+
+static void
+unpack_SRGBA_DXT1(const void *src, GLfloat dst[4])
+{
+}
+
+static void
+unpack_SRGBA_DXT3(const void *src, GLfloat dst[4])
+{
+}
+
+static void
+unpack_SRGBA_DXT5(const void *src, GLfloat dst[4])
+{
+}
+
+static void
+unpack_RGB_FXT1(const void *src, GLfloat dst[4])
+{
+}
+
+static void
+unpack_RGBA_FXT1(const void *src, GLfloat dst[4])
+{
+}
+
+static void
+unpack_RGB_DXT1(const void *src, GLfloat dst[4])
+{
+}
+
+static void
+unpack_RGBA_DXT1(const void *src, GLfloat dst[4])
+{
+}
+
+static void
+unpack_RGBA_DXT3(const void *src, GLfloat dst[4])
+{
+}
+
+static void
+unpack_RGBA_DXT5(const void *src, GLfloat dst[4])
+{
+}
+
+
+static void
+unpack_RGBA_FLOAT32(const void *src, GLfloat dst[4])
+{
+   const GLfloat *s = (const GLfloat *) src;
+   dst[RCOMP] = s[0];
+   dst[GCOMP] = s[1];
+   dst[BCOMP] = s[2];
+   dst[ACOMP] = s[3];
+}
+
+static void
+unpack_RGBA_FLOAT16(const void *src, GLfloat dst[4])
+{
+   const GLhalfARB *s = (const GLhalfARB *) src;
+   dst[RCOMP] = _mesa_half_to_float(s[0]);
+   dst[GCOMP] = _mesa_half_to_float(s[1]);
+   dst[BCOMP] = _mesa_half_to_float(s[2]);
+   dst[ACOMP] = _mesa_half_to_float(s[3]);
+}
+
+static void
+unpack_RGB_FLOAT32(const void *src, GLfloat dst[4])
+{
+   const GLfloat *s = (const GLfloat *) src;
+   dst[RCOMP] = s[0];
+   dst[GCOMP] = s[1];
+   dst[BCOMP] = s[2];
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_RGB_FLOAT16(const void *src, GLfloat dst[4])
+{
+   const GLhalfARB *s = (const GLhalfARB *) src;
+   dst[RCOMP] = _mesa_half_to_float(s[0]);
+   dst[GCOMP] = _mesa_half_to_float(s[1]);
+   dst[BCOMP] = _mesa_half_to_float(s[2]);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_ALPHA_FLOAT32(const void *src, GLfloat dst[4])
+{
+   const GLfloat *s = (const GLfloat *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = s[0];
+}
+
+static void
+unpack_ALPHA_FLOAT16(const void *src, GLfloat dst[4])
+{
+   const GLhalfARB *s = (const GLhalfARB *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = _mesa_half_to_float(s[0]);
+}
+
+static void
+unpack_LUMINANCE_FLOAT32(const void *src, GLfloat dst[4])
+{
+   const GLfloat *s = (const GLfloat *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = s[0];
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_LUMINANCE_FLOAT16(const void *src, GLfloat dst[4])
+{
+   const GLhalfARB *s = (const GLhalfARB *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = _mesa_half_to_float(s[0]);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_LUMINANCE_ALPHA_FLOAT32(const void *src, GLfloat dst[4])
+{
+   const GLfloat *s = (const GLfloat *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = s[0];
+   dst[ACOMP] = s[1];
+}
+
+static void
+unpack_LUMINANCE_ALPHA_FLOAT16(const void *src, GLfloat dst[4])
+{
+   const GLhalfARB *s = (const GLhalfARB *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = _mesa_half_to_float(s[0]);
+   dst[ACOMP] = _mesa_half_to_float(s[1]);
+}
+
+static void
+unpack_INTENSITY_FLOAT32(const void *src, GLfloat dst[4])
+{
+   const GLfloat *s = (const GLfloat *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] =
+   dst[ACOMP] = s[0];
+}
+
+static void
+unpack_INTENSITY_FLOAT16(const void *src, GLfloat dst[4])
+{
+   const GLhalfARB *s = (const GLhalfARB *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] =
+   dst[ACOMP] = s[0];
+}
+
+static void
+unpack_R_FLOAT32(const void *src, GLfloat dst[4])
+{
+   const GLfloat *s = (const GLfloat *) src;
+   dst[RCOMP] = s[0];
+   dst[GCOMP] = 0.0F;
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_R_FLOAT16(const void *src, GLfloat dst[4])
+{
+   const GLhalfARB *s = (const GLhalfARB *) src;
+   dst[RCOMP] = _mesa_half_to_float(s[0]);
+   dst[GCOMP] = 0.0F;
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_RG_FLOAT32(const void *src, GLfloat dst[4])
+{
+   const GLfloat *s = (const GLfloat *) src;
+   dst[RCOMP] = s[0];
+   dst[GCOMP] = s[1];
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_RG_FLOAT16(const void *src, GLfloat dst[4])
+{
+   const GLhalfARB *s = (const GLhalfARB *) src;
+   dst[RCOMP] = _mesa_half_to_float(s[0]);
+   dst[GCOMP] = _mesa_half_to_float(s[1]);
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = 1.0F;
+}
+
+
+static void
+unpack_RGBA_INT8(const void *src, GLfloat dst[4])
+{
+   const GLbyte *s = (const GLbyte *) src;
+   dst[RCOMP] = (GLfloat) s[0];
+   dst[GCOMP] = (GLfloat) s[1];
+   dst[BCOMP] = (GLfloat) s[2];
+   dst[ACOMP] = (GLfloat) s[3];
+}
+
+static void
+unpack_RGBA_INT16(const void *src, GLfloat dst[4])
+{
+   const GLshort *s = (const GLshort *) src;
+   dst[RCOMP] = (GLfloat) s[0];
+   dst[GCOMP] = (GLfloat) s[1];
+   dst[BCOMP] = (GLfloat) s[2];
+   dst[ACOMP] = (GLfloat) s[3];
+}
+
+static void
+unpack_RGBA_INT32(const void *src, GLfloat dst[4])
+{
+   const GLint *s = (const GLint *) src;
+   dst[RCOMP] = (GLfloat) s[0];
+   dst[GCOMP] = (GLfloat) s[1];
+   dst[BCOMP] = (GLfloat) s[2];
+   dst[ACOMP] = (GLfloat) s[3];
+}
+
+static void
+unpack_RGBA_UINT8(const void *src, GLfloat dst[4])
+{
+   const GLubyte *s = (const GLubyte *) src;
+   dst[RCOMP] = (GLfloat) s[0];
+   dst[GCOMP] = (GLfloat) s[1];
+   dst[BCOMP] = (GLfloat) s[2];
+   dst[ACOMP] = (GLfloat) s[3];
+}
+
+static void
+unpack_RGBA_UINT16(const void *src, GLfloat dst[4])
+{
+   const GLushort *s = (const GLushort *) src;
+   dst[RCOMP] = (GLfloat) s[0];
+   dst[GCOMP] = (GLfloat) s[1];
+   dst[BCOMP] = (GLfloat) s[2];
+   dst[ACOMP] = (GLfloat) s[3];
+}
+
+static void
+unpack_RGBA_UINT32(const void *src, GLfloat dst[4])
+{
+   const GLuint *s = (const GLuint *) src;
+   dst[RCOMP] = (GLfloat) s[0];
+   dst[GCOMP] = (GLfloat) s[1];
+   dst[BCOMP] = (GLfloat) s[2];
+   dst[ACOMP] = (GLfloat) s[3];
+}
+
+static void
+unpack_DUDV8(const void *src, GLfloat dst[4])
+{
+   const GLbyte *s = (const GLbyte *) src;
+   dst[RCOMP] = BYTE_TO_FLOAT(s[0]);
+   dst[GCOMP] = BYTE_TO_FLOAT(s[1]);
+   dst[BCOMP] = 0;
+   dst[ACOMP] = 0;
+}
+
+static void
+unpack_SIGNED_R8(const void *src, GLfloat dst[4])
+{
+   const GLbyte s = *((const GLbyte *) src);
+   dst[RCOMP] = BYTE_TO_FLOAT_TEX( s );
+   dst[GCOMP] = 0.0F;
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_SIGNED_RG88_REV(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLushort *) src);
+   dst[RCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s & 0xff) );
+   dst[GCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >> 8) );
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_SIGNED_RGBX8888(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >> 24) );
+   dst[GCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >> 16) );
+   dst[BCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >>  8) );
+   dst[ACOMP] = 1.0f;
+}
+
+static void
+unpack_SIGNED_RGBA8888(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >> 24) );
+   dst[GCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >> 16) );
+   dst[BCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >>  8) );
+   dst[ACOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s      ) );
+}
+
+static void
+unpack_SIGNED_RGBA8888_REV(const void *src, GLfloat dst[4])
+{
+   const GLuint s = *((const GLuint *) src);
+   dst[RCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s      ) );
+   dst[GCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >>  8) );
+   dst[BCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >> 16) );
+   dst[ACOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >> 24) );
+}
+
+static void
+unpack_SIGNED_R16(const void *src, GLfloat dst[4])
+{
+   const GLshort s = *((const GLshort *) src);
+   dst[RCOMP] = SHORT_TO_FLOAT_TEX( s );
+   dst[GCOMP] = 0.0F;
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_SIGNED_GR1616(const void *src, GLfloat dst[4])
+{
+   /* XXX TODO */
+}
+
+static void
+unpack_SIGNED_RGB_16(const void *src, GLfloat dst[4])
+{
+   const GLshort *s = (const GLshort *) src;
+   dst[RCOMP] = SHORT_TO_FLOAT_TEX( s[0] );
+   dst[GCOMP] = SHORT_TO_FLOAT_TEX( s[1] );
+   dst[BCOMP] = SHORT_TO_FLOAT_TEX( s[2] );
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_SIGNED_RGBA_16(const void *src, GLfloat dst[4])
+{
+   const GLshort *s = (const GLshort *) src;
+   dst[RCOMP] = SHORT_TO_FLOAT_TEX( s[0] );
+   dst[GCOMP] = SHORT_TO_FLOAT_TEX( s[1] );
+   dst[BCOMP] = SHORT_TO_FLOAT_TEX( s[2] );
+   dst[ACOMP] = SHORT_TO_FLOAT_TEX( s[3] );
+}
+
+static void
+unpack_RGBA_16(const void *src, GLfloat dst[4])
+{
+   const GLshort *s = (const GLshort *) src;
+   dst[RCOMP] = USHORT_TO_FLOAT( s[0] );
+   dst[GCOMP] = USHORT_TO_FLOAT( s[1] );
+   dst[BCOMP] = USHORT_TO_FLOAT( s[2] );
+   dst[ACOMP] = USHORT_TO_FLOAT( s[3] );
+}
+
+static void
+unpack_RED_RGTC1(const void *src, GLfloat dst[4])
+{
+   /* XXX to do */
+}
+
+static void
+unpack_SIGNED_RED_RGTC1(const void *src, GLfloat dst[4])
+{
+   /* XXX to do */
+}
+
+static void
+unpack_RG_RGTC2(const void *src, GLfloat dst[4])
+{
+   /* XXX to do */
+}
+
+static void
+unpack_SIGNED_RG_RGTC2(const void *src, GLfloat dst[4])
+{
+   /* XXX to do */
+}
+
+static void
+unpack_L_LATC1(const void *src, GLfloat dst[4])
+{
+   /* XXX to do */
+}
+
+static void
+unpack_SIGNED_L_LATC1(const void *src, GLfloat dst[4])
+{
+   /* XXX to do */
+}
+
+static void
+unpack_LA_LATC2(const void *src, GLfloat dst[4])
+{
+   /* XXX to do */
+}
+
+static void
+unpack_SIGNED_LA_LATC2(const void *src, GLfloat dst[4])
+{
+   /* XXX to do */
+}
+
+static void
+unpack_SIGNED_A8(const void *src, GLfloat dst[4])
+{
+   const GLbyte s = *((const GLbyte *) src);
+   dst[RCOMP] = 0.0F;
+   dst[GCOMP] = 0.0F;
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = BYTE_TO_FLOAT_TEX( s );
+}
+
+static void
+unpack_SIGNED_L8(const void *src, GLfloat dst[4])
+{
+   const GLbyte s = *((const GLbyte *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = BYTE_TO_FLOAT_TEX( s );
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_SIGNED_AL88(const void *src, GLfloat dst[4])
+{
+   const GLushort s = *((const GLshort *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s & 0xff) );
+   dst[ACOMP] = BYTE_TO_FLOAT_TEX( (GLbyte) (s >> 8) );
+}
+
+static void
+unpack_SIGNED_I8(const void *src, GLfloat dst[4])
+{
+   const GLbyte s = *((const GLbyte *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] =
+   dst[ACOMP] = BYTE_TO_FLOAT_TEX( s );
+}
+
+static void
+unpack_SIGNED_A16(const void *src, GLfloat dst[4])
+{
+   const GLshort s = *((const GLshort *) src);
+   dst[RCOMP] = 0.0F;
+   dst[GCOMP] = 0.0F;
+   dst[BCOMP] = 0.0F;
+   dst[ACOMP] = SHORT_TO_FLOAT_TEX( s );
+}
+
+static void
+unpack_SIGNED_L16(const void *src, GLfloat dst[4])
+{
+   const GLshort s = *((const GLshort *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = SHORT_TO_FLOAT_TEX( s );
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_SIGNED_AL1616(const void *src, GLfloat dst[4])
+{
+   const GLshort *s = (const GLshort *) src;
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] = SHORT_TO_FLOAT_TEX( s[0] );
+   dst[ACOMP] = SHORT_TO_FLOAT_TEX( s[1] );
+}
+
+static void
+unpack_SIGNED_I16(const void *src, GLfloat dst[4])
+{
+   const GLshort s = *((const GLshort *) src);
+   dst[RCOMP] =
+   dst[GCOMP] =
+   dst[BCOMP] =
+   dst[ACOMP] = SHORT_TO_FLOAT_TEX( s );
+}
+
+static void
+unpack_RGB9_E5_FLOAT(const void *src, GLfloat dst[4])
+{
+   const GLuint *s = (const GLuint *) src;
+   rgb9e5_to_float3(*s, dst);
+   dst[ACOMP] = 1.0F;
+}
+
+static void
+unpack_R11_G11_B10_FLOAT(const void *src, GLfloat dst[4])
+{
+   const GLuint *s = (const GLuint *) src;
+   r11g11b10f_to_float3(*s, dst);
+   dst[ACOMP] = 1.0F;
+}
+
+
+/**
+ * Return the unpacker function for the given format.
+ */
+static unpack_rgba_func
+get_unpack_rgba_function(gl_format format)
+{
+   static unpack_rgba_func table[MESA_FORMAT_COUNT];
+   static GLboolean initialized = GL_FALSE;
+
+   if (!initialized) {
+      table[MESA_FORMAT_NONE] = NULL;
+
+      table[MESA_FORMAT_RGBA8888] = unpack_RGBA8888;
+      table[MESA_FORMAT_RGBA8888_REV] = unpack_RGBA8888_REV;
+      table[MESA_FORMAT_ARGB8888] = unpack_ARGB8888;
+      table[MESA_FORMAT_ARGB8888_REV] = unpack_ARGB8888_REV;
+      table[MESA_FORMAT_XRGB8888] = unpack_XRGB8888;
+      table[MESA_FORMAT_XRGB8888_REV] = unpack_XRGB8888_REV;
+      table[MESA_FORMAT_RGB888] = unpack_RGB888;
+      table[MESA_FORMAT_BGR888] = unpack_BGR888;
+      table[MESA_FORMAT_RGB565] = unpack_RGB565;
+      table[MESA_FORMAT_RGB565_REV] = unpack_RGB565_REV;
+      table[MESA_FORMAT_ARGB4444] = unpack_ARGB4444;
+      table[MESA_FORMAT_ARGB4444_REV] = unpack_ARGB4444_REV;
+      table[MESA_FORMAT_RGBA5551] = unpack_RGBA5551;
+      table[MESA_FORMAT_ARGB1555] = unpack_ARGB1555;
+      table[MESA_FORMAT_ARGB1555_REV] = unpack_ARGB1555_REV;
+      table[MESA_FORMAT_AL44] = unpack_AL44;
+      table[MESA_FORMAT_AL88] = unpack_AL88;
+      table[MESA_FORMAT_AL88_REV] = unpack_AL88_REV;
+      table[MESA_FORMAT_AL1616] = unpack_AL1616;
+      table[MESA_FORMAT_AL1616_REV] = unpack_AL1616_REV;
+      table[MESA_FORMAT_RGB332] = unpack_RGB332;
+      table[MESA_FORMAT_A8] = unpack_A8;
+      table[MESA_FORMAT_A16] = unpack_A16;
+      table[MESA_FORMAT_L8] = unpack_L8;
+      table[MESA_FORMAT_L16] = unpack_L16;
+      table[MESA_FORMAT_I8] = unpack_I8;
+      table[MESA_FORMAT_I16] = unpack_I16;
+      table[MESA_FORMAT_YCBCR] = unpack_YCBCR;
+      table[MESA_FORMAT_YCBCR_REV] = unpack_YCBCR_REV;
+      table[MESA_FORMAT_R8] = unpack_R8;
+      table[MESA_FORMAT_RG88] = unpack_RG88;
+      table[MESA_FORMAT_RG88_REV] = unpack_RG88_REV;
+      table[MESA_FORMAT_R16] = unpack_R16;
+      table[MESA_FORMAT_RG1616] = unpack_RG1616;
+      table[MESA_FORMAT_RG1616_REV] = unpack_RG1616_REV;
+      table[MESA_FORMAT_ARGB2101010] = unpack_ARGB2101010;
+      table[MESA_FORMAT_Z24_S8] = unpack_Z24_S8;
+      table[MESA_FORMAT_S8_Z24] = unpack_S8_Z24;
+      table[MESA_FORMAT_Z16] = unpack_Z16;
+      table[MESA_FORMAT_X8_Z24] = unpack_X8_Z24;
+      table[MESA_FORMAT_Z24_X8] = unpack_Z24_X8;
+      table[MESA_FORMAT_Z32] = unpack_Z32;
+      table[MESA_FORMAT_S8] = unpack_S8;
+      table[MESA_FORMAT_SRGB8] = unpack_SRGB8;
+      table[MESA_FORMAT_SRGBA8] = unpack_SRGBA8;
+      table[MESA_FORMAT_SARGB8] = unpack_SARGB8;
+      table[MESA_FORMAT_SL8] = unpack_SL8;
+      table[MESA_FORMAT_SLA8] = unpack_SLA8;
+      table[MESA_FORMAT_SRGB_DXT1] = unpack_SRGB_DXT1;
+      table[MESA_FORMAT_SRGBA_DXT1] = unpack_SRGBA_DXT1;
+      table[MESA_FORMAT_SRGBA_DXT3] = unpack_SRGBA_DXT3;
+      table[MESA_FORMAT_SRGBA_DXT5] = unpack_SRGBA_DXT5;
+
+      table[MESA_FORMAT_RGB_FXT1] = unpack_RGB_FXT1;
+      table[MESA_FORMAT_RGBA_FXT1] = unpack_RGBA_FXT1;
+      table[MESA_FORMAT_RGB_DXT1] = unpack_RGB_DXT1;
+      table[MESA_FORMAT_RGBA_DXT1] = unpack_RGBA_DXT1;
+      table[MESA_FORMAT_RGBA_DXT3] = unpack_RGBA_DXT3;
+      table[MESA_FORMAT_RGBA_DXT5] = unpack_RGBA_DXT5;
+
+      table[MESA_FORMAT_RGBA_FLOAT32] = unpack_RGBA_FLOAT32;
+      table[MESA_FORMAT_RGBA_FLOAT16] = unpack_RGBA_FLOAT16;
+      table[MESA_FORMAT_RGB_FLOAT32] = unpack_RGB_FLOAT32;
+      table[MESA_FORMAT_RGB_FLOAT16] = unpack_RGB_FLOAT16;
+      table[MESA_FORMAT_ALPHA_FLOAT32] = unpack_ALPHA_FLOAT32;
+      table[MESA_FORMAT_ALPHA_FLOAT16] = unpack_ALPHA_FLOAT16;
+      table[MESA_FORMAT_LUMINANCE_FLOAT32] = unpack_LUMINANCE_FLOAT32;
+      table[MESA_FORMAT_LUMINANCE_FLOAT16] = unpack_LUMINANCE_FLOAT16;
+      table[MESA_FORMAT_LUMINANCE_ALPHA_FLOAT32] = unpack_LUMINANCE_ALPHA_FLOAT32;
+      table[MESA_FORMAT_LUMINANCE_ALPHA_FLOAT16] = unpack_LUMINANCE_ALPHA_FLOAT16;
+      table[MESA_FORMAT_INTENSITY_FLOAT32] = unpack_INTENSITY_FLOAT32;
+      table[MESA_FORMAT_INTENSITY_FLOAT16] = unpack_INTENSITY_FLOAT16;
+      table[MESA_FORMAT_R_FLOAT32] = unpack_R_FLOAT32;
+      table[MESA_FORMAT_R_FLOAT16] = unpack_R_FLOAT16;
+      table[MESA_FORMAT_RG_FLOAT32] = unpack_RG_FLOAT32;
+      table[MESA_FORMAT_RG_FLOAT16] = unpack_RG_FLOAT16;
+
+      table[MESA_FORMAT_RGBA_INT8] = unpack_RGBA_INT8;
+      table[MESA_FORMAT_RGBA_INT16] = unpack_RGBA_INT16;
+      table[MESA_FORMAT_RGBA_INT32] = unpack_RGBA_INT32;
+      table[MESA_FORMAT_RGBA_UINT8] = unpack_RGBA_UINT8;
+      table[MESA_FORMAT_RGBA_UINT16] = unpack_RGBA_UINT16;
+      table[MESA_FORMAT_RGBA_UINT32] = unpack_RGBA_UINT32;
+
+      table[MESA_FORMAT_DUDV8] = unpack_DUDV8;
+      table[MESA_FORMAT_SIGNED_R8] = unpack_SIGNED_R8;
+      table[MESA_FORMAT_SIGNED_RG88_REV] = unpack_SIGNED_RG88_REV;
+      table[MESA_FORMAT_SIGNED_RGBX8888] = unpack_SIGNED_RGBX8888;
+      table[MESA_FORMAT_SIGNED_RGBA8888] = unpack_SIGNED_RGBA8888;
+      table[MESA_FORMAT_SIGNED_RGBA8888_REV] = unpack_SIGNED_RGBA8888_REV;
+      table[MESA_FORMAT_SIGNED_R16] = unpack_SIGNED_R16;
+      table[MESA_FORMAT_SIGNED_GR1616] = unpack_SIGNED_GR1616;
+      table[MESA_FORMAT_SIGNED_RGB_16] = unpack_SIGNED_RGB_16;
+      table[MESA_FORMAT_SIGNED_RGBA_16] = unpack_SIGNED_RGBA_16;
+      table[MESA_FORMAT_RGBA_16] = unpack_RGBA_16;
+
+      table[MESA_FORMAT_RED_RGTC1] = unpack_RED_RGTC1;
+      table[MESA_FORMAT_SIGNED_RED_RGTC1] = unpack_SIGNED_RED_RGTC1;
+      table[MESA_FORMAT_RG_RGTC2] = unpack_RG_RGTC2;
+      table[MESA_FORMAT_SIGNED_RG_RGTC2] = unpack_SIGNED_RG_RGTC2;
+
+      table[MESA_FORMAT_L_LATC1] = unpack_L_LATC1;
+      table[MESA_FORMAT_SIGNED_L_LATC1] = unpack_SIGNED_L_LATC1;
+      table[MESA_FORMAT_LA_LATC2] = unpack_LA_LATC2;
+      table[MESA_FORMAT_SIGNED_LA_LATC2] = unpack_SIGNED_LA_LATC2;
+
+      table[MESA_FORMAT_SIGNED_A8] = unpack_SIGNED_A8;
+      table[MESA_FORMAT_SIGNED_L8] = unpack_SIGNED_L8;
+      table[MESA_FORMAT_SIGNED_AL88] = unpack_SIGNED_AL88;
+      table[MESA_FORMAT_SIGNED_I8] = unpack_SIGNED_I8;
+      table[MESA_FORMAT_SIGNED_A16] = unpack_SIGNED_A16;
+      table[MESA_FORMAT_SIGNED_L16] = unpack_SIGNED_L16;
+      table[MESA_FORMAT_SIGNED_AL1616] = unpack_SIGNED_AL1616;
+      table[MESA_FORMAT_SIGNED_I16] = unpack_SIGNED_I16;
+
+      table[MESA_FORMAT_RGB9_E5_FLOAT] = unpack_RGB9_E5_FLOAT;
+      table[MESA_FORMAT_R11_G11_B10_FLOAT] = unpack_R11_G11_B10_FLOAT;
+
+      table[MESA_FORMAT_Z32_FLOAT] = unpack_Z32_FLOAT;
+      table[MESA_FORMAT_Z32_FLOAT_X24S8] = unpack_Z32_FLOAT_X24S8;
+
+      initialized = GL_TRUE;
+   }
+
+   return table[format];
+}
+
+
+void
+_mesa_unpack_rgba_row(gl_format format, GLuint n,
+                      const void *src, GLfloat dst[][4])
+{
+   unpack_rgba_func unpack = get_unpack_rgba_function(format);
+   GLuint srcStride = _mesa_get_format_bytes(format);
+   const GLubyte *srcPtr = (GLubyte *) src;
+   GLuint i;
+
+   for (i = 0; i < n; i++) {
+      unpack(srcPtr, dst[i]);
+      srcPtr += srcStride;
+   }
+}
+
+
+
+/**
+ * Unpack a 2D rect of pixels returning float RGBA colors.
+ * \param format  the source image format
+ * \param src  start address of the source image
+ * \param srcRowStride  source image row stride in bytes
+ * \param dst  start address of the dest image
+ * \param dstRowStride  dest image row stride in bytes
+ * \param x  source image start X pos
+ * \param y  source image start Y pos
+ * \param width  width of rect region to convert
+ * \param height  height of rect region to convert
+ */
+void
+_mesa_unpack_rgba_block(gl_format format,
+                        const void *src, GLint srcRowStride,
+                        GLfloat dst[][4], GLint dstRowStride,
+                        GLuint x, GLuint y, GLuint width, GLuint height)
+{
+   unpack_rgba_func unpack = get_unpack_rgba_function(format);
+   const GLuint srcPixStride = _mesa_get_format_bytes(format);
+   const GLuint dstPixStride = 4 * sizeof(GLfloat);
+   const GLubyte *srcRow, *srcPix;
+   GLubyte *dstRow;
+   GLfloat *dstPix;
+   GLuint i, j;
+
+   /* XXX needs to be fixed for compressed formats */
+
+   srcRow = ((const GLubyte *) src) + srcRowStride * y + srcPixStride * x;
+   dstRow = ((GLubyte *) dst) + dstRowStride * y + dstPixStride * x;
+
+   for (i = 0; i < height; i++) {
+      srcPix = srcRow;
+      dstPix = (GLfloat *) dstRow;
+
+      for (j = 0; j < width; j++) {
+         unpack(srcPix, dstPix);
+         srcPix += srcPixStride;
+         dstPix += dstPixStride;
+      }
+
+      dstRow += dstRowStride;
+      srcRow += srcRowStride;
+   }
+}
+
+
+
+
+typedef void (*unpack_float_z_func)(const void *src, GLfloat *dst);
+
+static void
+unpack_float_z_Z24_S8(const void *src, GLfloat *dst)
+{
+   /* only return Z, not stencil data */
+   const GLuint s = *((const GLuint *) src);
+   const GLfloat scale = 1.0F / (GLfloat) 0xffffff;
+   *dst = (s >> 8) * scale;
+   ASSERT(*dst >= 0.0F);
+   ASSERT(*dst <= 1.0F);
+}
+
+static void
+unpack_float_z_S8_Z24(const void *src, GLfloat *dst)
+{
+   /* only return Z, not stencil data */
+   const GLuint s = *((const GLuint *) src);
+   const GLfloat scale = 1.0F / (GLfloat) 0xffffff;
+   *dst = (s & 0x00ffffff) * scale;
+   ASSERT(*dst >= 0.0F);
+   ASSERT(*dst <= 1.0F);
+}
+
+static void
+unpack_float_z_Z16(const void *src, GLfloat *dst)
+{
+   const GLushort s = *((const GLushort *) src);
+   *dst = s * (1.0F / 65535.0F);
+}
+
+static void
+unpack_float_z_X8_Z24(const void *src, GLfloat *dst)
+{
+   unpack_float_z_S8_Z24(src, dst);
+}
+
+static void
+unpack_float_z_Z24_X8(const void *src, GLfloat *dst)
+{
+   unpack_float_z_Z24_S8(src, dst);
+}
+
+static void
+unpack_float_z_Z32(const void *src, GLfloat *dst)
+{
+   const GLuint s = *((const GLuint *) src);
+   *dst = s * (1.0F / 0xffffffff);
+}
+
+static void
+unpack_float_z_Z32X24S8(const void *src, GLfloat *dst)
+{
+   *dst = *((const GLfloat *) src);
+}
+
+
+
+void
+_mesa_unpack_float_z_row(gl_format format, GLuint n,
+                         const void *src, GLfloat *dst)
+{
+   unpack_float_z_func unpack;
+   GLuint srcStride = _mesa_get_format_bytes(format);
+   const GLubyte *srcPtr = (GLubyte *) src;
+   GLuint i;
+
+   switch (format) {
+   case MESA_FORMAT_Z24_S8:
+      unpack = unpack_float_z_Z24_S8;
+      break;
+   case MESA_FORMAT_S8_Z24:
+      unpack = unpack_float_z_S8_Z24;
+      break;
+   case MESA_FORMAT_Z16:
+      unpack = unpack_float_z_Z16;
+      break;
+   case MESA_FORMAT_X8_Z24:
+      unpack = unpack_float_z_X8_Z24;
+      break;
+   case MESA_FORMAT_Z24_X8:
+      unpack = unpack_float_z_Z24_X8;
+      break;
+   case MESA_FORMAT_Z32:
+      unpack = unpack_float_z_Z32;
+      break;
+   case MESA_FORMAT_Z32_FLOAT_X24S8:
+      unpack = unpack_float_z_Z32X24S8;
+      break;
+   default:
+      _mesa_problem(NULL, "bad format %s in _mesa_unpack_float_z_row",
+                    _mesa_get_format_name(format));
+      return;
+   }
+
+   for (i = 0; i < n; i++) {
+      unpack(srcPtr, &dst[i]);
+      srcPtr += srcStride;
+   }
+}
+
+
+
+typedef void (*unpack_uint_z_func)(const void *src, GLuint *dst);
+
+static void
+unpack_uint_z_Z24_S8(const void *src, GLuint *dst)
+{
+   /* only return Z, not stencil data */
+   const GLuint s = *((const GLuint *) src);
+   *dst = (s >> 8);
+}
+
+static void
+unpack_uint_z_S8_Z24(const void *src, GLuint *dst)
+{
+   /* only return Z, not stencil data */
+   const GLuint s = *((const GLuint *) src);
+   *dst = s & 0x00ffffff;
+}
+
+static void
+unpack_uint_z_Z16(const void *src, GLuint *dst)
+{
+   *dst = *((const GLushort *) src);
+}
+
+static void
+unpack_uint_z_X8_Z24(const void *src, GLuint *dst)
+{
+   unpack_uint_z_S8_Z24(src, dst);
+}
+
+static void
+unpack_uint_z_Z24_X8(const void *src, GLuint *dst)
+{
+   unpack_uint_z_Z24_S8(src, dst);
+}
+
+static void
+unpack_uint_z_Z32(const void *src, GLuint *dst)
+{
+   *dst = *((const GLuint *) src);
+}
+
+
+void
+_mesa_unpack_uint_z_row(gl_format format, GLuint n,
+                        const void *src, GLuint *dst)
+{
+   unpack_uint_z_func unpack;
+   GLuint srcStride = _mesa_get_format_bytes(format);
+   const GLubyte *srcPtr = (GLubyte *) src;
+   GLuint i;
+
+   switch (format) {
+   case MESA_FORMAT_Z24_S8:
+      unpack = unpack_uint_z_Z24_S8;
+      break;
+   case MESA_FORMAT_S8_Z24:
+      unpack = unpack_uint_z_S8_Z24;
+      break;
+   case MESA_FORMAT_Z16:
+      unpack = unpack_uint_z_Z16;
+      break;
+   case MESA_FORMAT_X8_Z24:
+      unpack = unpack_uint_z_X8_Z24;
+      break;
+   case MESA_FORMAT_Z24_X8:
+      unpack = unpack_uint_z_Z24_X8;
+      break;
+   case MESA_FORMAT_Z32:
+      unpack = unpack_uint_z_Z32;
+      break;
+   default:
+      _mesa_problem(NULL, "bad format %s in _mesa_unpack_uint_z_row",
+                    _mesa_get_format_name(format));
+      return;
+   }
+
+   for (i = 0; i < n; i++) {
+      unpack(srcPtr, &dst[i]);
+      srcPtr += srcStride;
+   }
+}
+
+
diff --git a/mesalib/src/mesa/main/format_unpack.h b/mesalib/src/mesa/main/format_unpack.h
new file mode 100644
index 000000000..c37727d91
--- /dev/null
+++ b/mesalib/src/mesa/main/format_unpack.h
@@ -0,0 +1,49 @@
+/*
+ * Mesa 3-D graphics library
+ *
+ * Copyright (c) 2011 VMware, Inc.
+ *
+ * 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
+ * THE AUTHORS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef FORMAT_UNPACK_H
+#define FORMAT_UNPACK_H
+
+extern void
+_mesa_unpack_rgba_row(gl_format format, GLuint n,
+                      const void *src, GLfloat dst[][4]);
+
+
+extern void
+_mesa_unpack_rgba_block(gl_format format,
+                        const void *src, GLint srcRowStride,
+                        GLfloat dst[][4], GLint dstRowStride,
+                        GLuint x, GLuint y, GLuint width, GLuint height);
+
+
+extern void
+_mesa_unpack_float_z_row(gl_format format, GLuint n,
+                         const void *src, GLfloat *dst);
+
+
+void
+_mesa_unpack_uint_z_row(gl_format format, GLuint n,
+                        const void *src, GLuint *dst);
+
+
+#endif /* FORMAT_UNPACK_H */
diff --git a/mesalib/src/mesa/main/formats.c b/mesalib/src/mesa/main/formats.c
index c6634c458..11d670689 100644
--- a/mesalib/src/mesa/main/formats.c
+++ b/mesalib/src/mesa/main/formats.c
@@ -1343,6 +1343,70 @@ _mesa_get_srgb_format_linear(gl_format format)
 
 
 /**
+ * If the given format is a compressed format, return a corresponding
+ * uncompressed format.
+ */
+gl_format
+_mesa_get_uncompressed_format(gl_format format)
+{
+   switch (format) {
+   case MESA_FORMAT_RGB_FXT1:
+      return MESA_FORMAT_RGB888;
+   case MESA_FORMAT_RGBA_FXT1:
+      return MESA_FORMAT_RGBA8888;
+   case MESA_FORMAT_RGB_DXT1:
+   case MESA_FORMAT_SRGB_DXT1:
+      return MESA_FORMAT_RGB888;
+   case MESA_FORMAT_RGBA_DXT1:
+   case MESA_FORMAT_SRGBA_DXT1:
+      return MESA_FORMAT_RGBA8888;
+   case MESA_FORMAT_RGBA_DXT3:
+   case MESA_FORMAT_SRGBA_DXT3:
+      return MESA_FORMAT_RGBA8888;
+   case MESA_FORMAT_RGBA_DXT5:
+   case MESA_FORMAT_SRGBA_DXT5:
+      return MESA_FORMAT_RGBA8888;
+   case MESA_FORMAT_RED_RGTC1:
+      return MESA_FORMAT_R8;
+   case MESA_FORMAT_SIGNED_RED_RGTC1:
+      return MESA_FORMAT_SIGNED_R8;
+   case MESA_FORMAT_RG_RGTC2:
+      return MESA_FORMAT_RG88;
+   case MESA_FORMAT_SIGNED_RG_RGTC2:
+      return MESA_FORMAT_SIGNED_RG88_REV;
+   case MESA_FORMAT_L_LATC1:
+      return MESA_FORMAT_L8;
+   case MESA_FORMAT_SIGNED_L_LATC1:
+      return MESA_FORMAT_SIGNED_L8;
+   case MESA_FORMAT_LA_LATC2:
+      return MESA_FORMAT_AL88;
+   case MESA_FORMAT_SIGNED_LA_LATC2:
+      return MESA_FORMAT_SIGNED_AL88;
+   default:
+#ifdef DEBUG
+      assert(!_mesa_is_format_compressed(format));
+#endif
+      return format;
+   }
+}
+
+
+GLuint
+_mesa_format_num_components(gl_format format)
+{
+   const struct gl_format_info *info = _mesa_get_format_info(format);
+   return ((info->RedBits > 0) +
+           (info->GreenBits > 0) +
+           (info->BlueBits > 0) +
+           (info->AlphaBits > 0) +
+           (info->LuminanceBits > 0) +
+           (info->IntensityBits > 0) +
+           (info->DepthBits > 0) +
+           (info->StencilBits > 0));
+}
+
+
+/**
  * Return number of bytes needed to store an image of the given size
  * in the given format.
  */
diff --git a/mesalib/src/mesa/main/formats.h b/mesalib/src/mesa/main/formats.h
index 75d93deba..610204cb3 100644
--- a/mesalib/src/mesa/main/formats.h
+++ b/mesalib/src/mesa/main/formats.h
@@ -266,4 +266,11 @@ _mesa_test_formats(void);
 extern gl_format
 _mesa_get_srgb_format_linear(gl_format format);
 
+extern gl_format
+_mesa_get_uncompressed_format(gl_format format);
+
+extern GLuint
+_mesa_format_num_components(gl_format format);
+
+
 #endif /* FORMATS_H */
diff --git a/mesalib/src/mesa/main/mipmap.c b/mesalib/src/mesa/main/mipmap.c
index cf9d522f2..869243d1c 100644
--- a/mesalib/src/mesa/main/mipmap.c
+++ b/mesalib/src/mesa/main/mipmap.c
@@ -1985,50 +1985,45 @@ generate_mipmap_uncompressed(struct gl_context *ctx, GLenum target,
 static void
 generate_mipmap_compressed(struct gl_context *ctx, GLenum target,
 			   struct gl_texture_object *texObj,
-			   const struct gl_texture_image *srcImage,
+			   struct gl_texture_image *srcImage,
 			   GLuint maxLevel)
 {
    GLint level;
    gl_format temp_format;
-   GLenum datatype;
-   GLuint comps;
-   GLuint row;
    GLint components;
    GLuint temp_src_stride, temp_dst_stride; /* in bytes */
    GLchan *temp_src = NULL, *temp_dst = NULL;
+   GLenum temp_datatype;
+   GLenum temp_base_format;
 
-   /* Choose the format we will do _mesa_generate_mipmap_level() in,
-    * and uncompress the firstImage into a temporary of that format.
-    */
+   /* only two types of compressed textures at this time */
    assert(texObj->Target == GL_TEXTURE_2D ||
 	  texObj->Target == GL_TEXTURE_CUBE_MAP_ARB);
 
-   if (srcImage->_BaseFormat == GL_RGB) {
-      temp_format = MESA_FORMAT_RGB888;
-      components = 3;
-   } else if (srcImage->_BaseFormat == GL_RED) {
-      temp_format = MESA_FORMAT_R8;
-      components = 1;
-   } else if (srcImage->_BaseFormat == GL_RG) {
-      temp_format = MESA_FORMAT_RG88;
-      components = 2;
-   } else if (srcImage->_BaseFormat == GL_RGBA) {
-      temp_format = MESA_FORMAT_RGBA8888;
-      components = 4;
-   } else if (srcImage->_BaseFormat == GL_LUMINANCE) {
-      temp_format = MESA_FORMAT_L8;
-      components = 1;
-   } else if (srcImage->_BaseFormat == GL_LUMINANCE_ALPHA) {
-      temp_format = MESA_FORMAT_AL88;
-      components = 2;
-   } else {
-      _mesa_problem(ctx, "bad srcImage->_BaseFormat in _mesa_generate_mipmaps");
-      return;
+   /*
+    * Choose a format for the temporary, uncompressed base image.
+    * Then, get number of components, choose temporary image datatype,
+    * and get base format.
+    */
+   temp_format = _mesa_get_uncompressed_format(srcImage->TexFormat);
+
+   components = _mesa_format_num_components(temp_format);
+
+   /* Revisit this if we get compressed formats with >8 bits per component */
+   if (_mesa_get_format_datatype(srcImage->TexFormat)
+       == GL_SIGNED_NORMALIZED) {
+      temp_datatype = GL_BYTE;
+   }
+   else {
+      temp_datatype = GL_UNSIGNED_BYTE;
    }
 
-   /* allocate storage for uncompressed GL_RGB or GL_RGBA images */
-   temp_src_stride = _mesa_format_row_stride(temp_format, srcImage->Width);
+   temp_base_format = _mesa_get_format_base_format(temp_format);
+
+
+   /* allocate storage for the temporary, uncompressed image */
    /* 20 extra bytes, just be safe when calling last FetchTexel */
+   temp_src_stride = _mesa_format_row_stride(temp_format, srcImage->Width);
    temp_src = (GLubyte *) malloc(temp_src_stride * srcImage->Height + 20);
    if (!temp_src) {
       _mesa_error(ctx, GL_OUT_OF_MEMORY, "generate mipmaps");
@@ -2036,16 +2031,20 @@ generate_mipmap_compressed(struct gl_context *ctx, GLenum target,
    }
 
    /* decompress base image to the temporary */
-   for (row = 0; row < srcImage->Height; row++) {
-      GLuint col;
-      GLchan *dst = (GLchan *) temp_src + temp_src_stride * row;
-      for (col = 0; col < srcImage->Width; col++) {
-	 srcImage->FetchTexelc(srcImage, col, row, 0, dst);
-	 dst += components;
-      }
+   {
+      /* save pixel packing mode */
+      struct gl_pixelstore_attrib save = ctx->Pack;
+      /* use default/tight packing parameters */
+      ctx->Pack = ctx->DefaultPacking;
+
+      /* Get the uncompressed image */
+      ctx->Driver.GetTexImage(ctx, target, texObj->BaseLevel,
+                              temp_base_format, temp_datatype,
+                              temp_src, texObj, srcImage);
+      /* restore packing mode */
+      ctx->Pack = save;
    }
 
-   _mesa_format_to_type_and_comps(temp_format, &datatype, &comps);
 
    for (level = texObj->BaseLevel; level < maxLevel; level++) {
       /* generate image[level+1] from image[level] */
@@ -2086,7 +2085,10 @@ generate_mipmap_compressed(struct gl_context *ctx, GLenum target,
          return;
       }
 
-      _mesa_generate_mipmap_level(target, datatype, comps, border,
+      /* Free old image data */
+      ctx->Driver.FreeTextureImageBuffer(ctx, dstImage);
+
+      _mesa_generate_mipmap_level(target, temp_datatype, components, border,
                                   srcWidth, srcHeight, srcDepth,
                                   temp_src, temp_src_stride / components,
                                   dstWidth, dstHeight, dstDepth,
@@ -2100,8 +2102,7 @@ generate_mipmap_compressed(struct gl_context *ctx, GLenum target,
       ctx->Driver.TexImage2D(ctx, target, level + 1,
 			     srcImage->InternalFormat,
 			     dstWidth, dstHeight, border,
-			     _mesa_get_format_base_format(temp_format),
-			     GL_UNSIGNED_BYTE,
+			     temp_base_format, temp_datatype,
 			     temp_dst, &ctx->DefaultPacking, texObj, dstImage);
 
       /* swap src and dest pointers */
@@ -2130,7 +2131,7 @@ void
 _mesa_generate_mipmap(struct gl_context *ctx, GLenum target,
                       struct gl_texture_object *texObj)
 {
-   const struct gl_texture_image *srcImage;
+   struct gl_texture_image *srcImage;
    GLint maxLevel;
 
    ASSERT(texObj);
diff --git a/mesalib/src/mesa/main/texcompress.c b/mesalib/src/mesa/main/texcompress.c
index 42bd1eee5..4e5c14cbb 100644
--- a/mesalib/src/mesa/main/texcompress.c
+++ b/mesalib/src/mesa/main/texcompress.c
@@ -37,6 +37,9 @@
 #include "mfeatures.h"
 #include "mtypes.h"
 #include "texcompress.h"
+#include "texcompress_fxt1.h"
+#include "texcompress_rgtc.h"
+#include "texcompress_s3tc.h"
 
 
 /**
@@ -259,7 +262,6 @@ _mesa_get_compressed_formats(struct gl_context *ctx, GLint *formats)
          n += 4;
       }
    }
-   return n;
 
 #if FEATURE_ES1 || FEATURE_ES2
    if (formats) {
@@ -278,6 +280,8 @@ _mesa_get_compressed_formats(struct gl_context *ctx, GLint *formats)
       n += 10;
    }
 #endif
+
+   return n;
 }
 
 
@@ -437,3 +441,86 @@ _mesa_compressed_image_address(GLint col, GLint row, GLint img,
 
    return (GLubyte *) image + offset;
 }
+
+
+/**
+ * Decompress a compressed texture image, returning a GL_RGBA/GL_FLOAT image.
+ */
+void
+_mesa_decompress_image(gl_format format, GLuint width, GLuint height,
+                       const GLubyte *src, GLint srcRowStride,
+                       GLfloat *dest)
+{
+   void (*fetch)(const struct gl_texture_image *texImage,
+                 GLint i, GLint j, GLint k, GLfloat *texel);
+   struct gl_texture_image texImage;  /* dummy teximage */
+   GLuint i, j;
+
+   /* setup dummy texture image info */
+   memset(&texImage, 0, sizeof(texImage));
+   texImage.Data = (void *) src;
+   texImage.RowStride = srcRowStride;
+
+   switch (format) {
+   /* DXT formats */
+   case MESA_FORMAT_RGB_DXT1:
+      fetch = _mesa_fetch_texel_2d_f_rgb_dxt1;
+      break;
+   case MESA_FORMAT_RGBA_DXT1:
+      fetch = _mesa_fetch_texel_2d_f_rgba_dxt1;
+      break;
+   case MESA_FORMAT_RGBA_DXT3:
+      fetch = _mesa_fetch_texel_2d_f_rgba_dxt3;
+      break;
+   case MESA_FORMAT_RGBA_DXT5:
+      fetch = _mesa_fetch_texel_2d_f_rgba_dxt5;
+      break;
+
+   /* FXT1 formats */
+   case MESA_FORMAT_RGB_FXT1:
+      fetch = _mesa_fetch_texel_2d_f_rgb_fxt1;
+      break;
+   case MESA_FORMAT_RGBA_FXT1:
+      fetch = _mesa_fetch_texel_2d_f_rgba_fxt1;
+      break;
+
+   /* Red/RG formats */
+   case MESA_FORMAT_RED_RGTC1:
+      fetch = _mesa_fetch_texel_2d_f_red_rgtc1;
+      break;
+   case MESA_FORMAT_SIGNED_RED_RGTC1:
+      fetch = _mesa_fetch_texel_2d_f_signed_red_rgtc1;
+      break;
+   case MESA_FORMAT_RG_RGTC2:
+      fetch = _mesa_fetch_texel_2d_f_rg_rgtc2;
+      break;
+   case MESA_FORMAT_SIGNED_RG_RGTC2:
+      fetch = _mesa_fetch_texel_2d_f_signed_rg_rgtc2;
+      break;
+
+   /* L/LA formats */
+   case MESA_FORMAT_L_LATC1:
+      fetch = _mesa_fetch_texel_2d_f_l_latc1;
+      break;
+   case MESA_FORMAT_SIGNED_L_LATC1:
+      fetch = _mesa_fetch_texel_2d_f_signed_l_latc1;
+      break;
+   case MESA_FORMAT_LA_LATC2:
+      fetch = _mesa_fetch_texel_2d_f_la_latc2;
+      break;
+   case MESA_FORMAT_SIGNED_LA_LATC2:
+      fetch = _mesa_fetch_texel_2d_f_signed_la_latc2;
+      break;
+
+   default:
+      _mesa_problem(NULL, "Unexpected format in _mesa_decompress_image()");
+      return;
+   }
+
+   for (j = 0; j < height; j++) {
+      for (i = 0; i < width; i++) {
+         fetch(&texImage, i, j, 0, dest);
+         dest += 4;
+      }
+   }
+}
diff --git a/mesalib/src/mesa/main/texcompress.h b/mesalib/src/mesa/main/texcompress.h
index 375cf90c8..2c357068c 100644
--- a/mesalib/src/mesa/main/texcompress.h
+++ b/mesalib/src/mesa/main/texcompress.h
@@ -50,6 +50,11 @@ _mesa_compressed_image_address(GLint col, GLint row, GLint img,
                                gl_format mesaFormat,
                                GLsizei width, const GLubyte *image);
 
+extern void
+_mesa_decompress_image(gl_format format, GLuint width, GLuint height,
+                       const GLubyte *src, GLint srcRowStride,
+                       GLfloat *dest);
+
 #else /* _HAVE_FULL_GL */
 
 /* no-op macros */
diff --git a/mesalib/src/mesa/main/texgetimage.c b/mesalib/src/mesa/main/texgetimage.c
index 99ace91a9..2830dda86 100644
--- a/mesalib/src/mesa/main/texgetimage.c
+++ b/mesalib/src/mesa/main/texgetimage.c
@@ -34,13 +34,14 @@
 #include "enums.h"
 #include "context.h"
 #include "formats.h"
+#include "format_unpack.h"
 #include "image.h"
 #include "mfeatures.h"
 #include "mtypes.h"
 #include "pack.h"
 #include "pbo.h"
+#include "texcompress.h"
 #include "texgetimage.h"
-#include "texfetch.h"
 #include "teximage.h"
 
 
@@ -75,8 +76,9 @@ get_tex_depth(struct gl_context *ctx, GLuint dimensions,
    const GLint width = texImage->Width;
    const GLint height = texImage->Height;
    const GLint depth = texImage->Depth;
-   GLint img, row, col;
+   GLint img, row;
    GLfloat *depthRow = (GLfloat *) malloc(width * sizeof(GLfloat));
+   const GLint texelSize = _mesa_get_format_bytes(texImage->TexFormat);
 
    if (!depthRow) {
       _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGetTexImage");
@@ -88,11 +90,12 @@ get_tex_depth(struct gl_context *ctx, GLuint dimensions,
          void *dest = _mesa_image_address(dimensions, &ctx->Pack, pixels,
                                           width, height, format, type,
                                           img, row, 0);
-         assert(dest);
+         const GLubyte *src = (GLubyte *) texImage->Data +
+            (texImage->ImageOffsets[img] +
+             texImage->RowStride * row) * texelSize;
+
+         _mesa_unpack_float_z_row(texImage->TexFormat, width, src, depthRow);
 
-         for (col = 0; col < width; col++) {
-            texImage->FetchTexelf(texImage, col, row, img, depthRow + col);
-         }
          _mesa_pack_depth_span(ctx, width, dest, type, depthRow, &ctx->Pack);
       }
    }
@@ -173,95 +176,141 @@ get_tex_ycbcr(struct gl_context *ctx, GLuint dimensions,
 
 
 /**
- * glGetTexImage for (s)RGBA, Luminance, etc. pixels.
- * This is the slow way since we use texture sampling.
+ * glGetTexImage for color formats (RGBA, RGB, alpha, LA, etc).
+ * Compressed textures are handled here as well.
  */
 static void
 get_tex_rgba(struct gl_context *ctx, GLuint dimensions,
              GLenum format, GLenum type, GLvoid *pixels,
              struct gl_texture_image *texImage)
 {
-   const GLint width = texImage->Width;
-   const GLint height = texImage->Height;
-   const GLint depth = texImage->Depth;
-   const GLenum dataType = _mesa_get_format_datatype(texImage->TexFormat);
+   /* don't want to apply sRGB -> RGB conversion here so override the format */
+   const gl_format texFormat = _mesa_get_srgb_format_linear(texImage->TexFormat);
+   const GLuint width = texImage->Width;
+   const GLuint height = texImage->Height;
+   const GLuint depth = texImage->Depth;
+   const GLenum dataType = _mesa_get_format_datatype(texFormat);
+   const GLenum baseFormat = _mesa_get_format_base_format(texFormat);
    /* Normally, no pixel transfer ops are performed during glGetTexImage.
     * The only possible exception is component clamping to [0,1].
     */
    GLbitfield transferOps = 0x0;
-   GLint img, row;
-   GLfloat (*rgba)[4] = (GLfloat (*)[4]) malloc(4 * width * sizeof(GLfloat));
-   const GLboolean is_sampler_srgb_decode =
-       _mesa_get_format_color_encoding(texImage->TexFormat) == GL_SRGB &&
-       texImage->TexObject->Sampler.sRGBDecode == GL_DECODE_EXT;
-
-   if (!rgba) {
-      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGetTexImage");
-      return;
-   }
 
-   /* Clamping does not apply to GetTexImage (final conversion)?
-    * Looks like we need clamp though when going from format
-    * containing negative values to unsigned format.
+   /* In general, clamping does not apply to glGetTexImage, except when
+    * the returned type of the image can't hold negative values.
     */
-   if (format == GL_LUMINANCE || format == GL_LUMINANCE_ALPHA) {
-      transferOps |= IMAGE_CLAMP_BIT;
-   }
-   else if (!type_with_negative_values(type) &&
-            (dataType == GL_FLOAT ||
-             dataType == GL_SIGNED_NORMALIZED)) {
-      transferOps |= IMAGE_CLAMP_BIT;
+   if (!type_with_negative_values(type)) {
+      /* the returned image type can't have negative values */
+      if (dataType == GL_FLOAT ||
+          dataType == GL_SIGNED_NORMALIZED ||
+          format == GL_LUMINANCE ||
+          format == GL_LUMINANCE_ALPHA) {
+         transferOps |= IMAGE_CLAMP_BIT;
+      }
    }
 
-   /* glGetTexImage always returns sRGB data for sRGB textures. Make sure the
-    * fetch functions return sRGB data without linearizing it.
-    */
-   if (is_sampler_srgb_decode) {
-      texImage->TexObject->Sampler.sRGBDecode = GL_SKIP_DECODE_EXT;
-      _mesa_set_fetch_functions(texImage, dimensions);
-   }
+   if (_mesa_is_format_compressed(texFormat)) {
+      /* Decompress into temp buffer, then pack into user buffer */
+      GLfloat *tempImage, *srcRow;
+      GLuint row;
 
-   for (img = 0; img < depth; img++) {
+      tempImage = (GLfloat *) malloc(texImage->Width * texImage->Height *
+                                     texImage->Depth * 4 * sizeof(GLfloat));
+      if (!tempImage) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGetTexImage()");
+         return;
+      }
+
+      _mesa_decompress_image(texFormat, texImage->Width, texImage->Height,
+                             texImage->Data, texImage->RowStride, tempImage);
+
+      if (baseFormat == GL_LUMINANCE ||
+          baseFormat == GL_LUMINANCE_ALPHA) {
+         /* Set green and blue to zero since the pack function here will
+          * compute L=R+G+B.
+          */
+         GLuint i;
+         for (i = 0; i < width * height; i++) {
+            tempImage[i * 4 + GCOMP] = tempImage[i * 4 + BCOMP] = 0.0f;
+         }
+      }
+
+      srcRow = tempImage;
       for (row = 0; row < height; row++) {
          void *dest = _mesa_image_address(dimensions, &ctx->Pack, pixels,
                                           width, height, format, type,
-                                          img, row, 0);
-         GLint col;
+                                          0, row, 0);
+
+         _mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) srcRow,
+                                    format, type, dest, &ctx->Pack, transferOps);
+         srcRow += width * 4;
+      }
+
+      free(tempImage);
+   }
+   else {
+      /* No decompression needed */
+      const GLint texelSize = _mesa_get_format_bytes(texFormat);
+      GLuint img, row;
+      GLfloat (*rgba)[4];
+
+      rgba = (GLfloat (*)[4]) malloc(4 * width * sizeof(GLfloat));
+      if (!rgba) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glGetTexImage()");
+         return;
+      }
+
+      for (img = 0; img < depth; img++) {
+         for (row = 0; row < height; row++) {
+            void *dest = _mesa_image_address(dimensions, &ctx->Pack, pixels,
+                                             width, height, format, type,
+                                             img, row, 0);
+            const GLubyte *src = (const GLubyte *) texImage->Data +
+               (texImage->ImageOffsets[img] +
+                texImage->RowStride * row) * texelSize;
+
+            _mesa_unpack_rgba_row(texFormat, width, src, rgba);
 
-         for (col = 0; col < width; col++) {
-            texImage->FetchTexelf(texImage, col, row, img, rgba[col]);
             if (texImage->_BaseFormat == GL_ALPHA) {
-               rgba[col][RCOMP] = 0.0F;
-               rgba[col][GCOMP] = 0.0F;
-               rgba[col][BCOMP] = 0.0F;
+               GLint col;
+               for (col = 0; col < width; col++) {
+                  rgba[col][RCOMP] = 0.0F;
+                  rgba[col][GCOMP] = 0.0F;
+                  rgba[col][BCOMP] = 0.0F;
+               }
             }
             else if (texImage->_BaseFormat == GL_LUMINANCE) {
-               rgba[col][GCOMP] = 0.0F;
-               rgba[col][BCOMP] = 0.0F;
-               rgba[col][ACOMP] = 1.0F;
+               GLint col;
+               for (col = 0; col < width; col++) {
+                  rgba[col][GCOMP] = 0.0F;
+                  rgba[col][BCOMP] = 0.0F;
+                  rgba[col][ACOMP] = 1.0F;
+               }
             }
             else if (texImage->_BaseFormat == GL_LUMINANCE_ALPHA) {
-               rgba[col][GCOMP] = 0.0F;
-               rgba[col][BCOMP] = 0.0F;
+               GLint col;
+               for (col = 0; col < width; col++) {
+                  rgba[col][GCOMP] = 0.0F;
+                  rgba[col][BCOMP] = 0.0F;
+               }
             }
             else if (texImage->_BaseFormat == GL_INTENSITY) {
-               rgba[col][GCOMP] = 0.0F;
-               rgba[col][BCOMP] = 0.0F;
-               rgba[col][ACOMP] = 1.0F;
+               GLint col;
+               for (col = 0; col < width; col++) {
+                  rgba[col][GCOMP] = 0.0F;
+                  rgba[col][BCOMP] = 0.0F;
+                  rgba[col][ACOMP] = 1.0F;
+               }
             }
+
+            _mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) rgba,
+                                       format, type, dest,
+                                       &ctx->Pack, transferOps);
          }
-         _mesa_pack_rgba_span_float(ctx, width, (GLfloat (*)[4]) rgba,
-                                    format, type, dest,
-                                    &ctx->Pack, transferOps);
       }
-   }
 
-   if (is_sampler_srgb_decode) {
-      texImage->TexObject->Sampler.sRGBDecode = GL_DECODE_EXT;
-      _mesa_set_fetch_functions(texImage, dimensions);
+      free(rgba);
    }
-
-   free(rgba);
 }
 
 
diff --git a/mesalib/src/mesa/sources.mak b/mesalib/src/mesa/sources.mak
index deff7bd78..71f708f1e 100644
--- a/mesalib/src/mesa/sources.mak
+++ b/mesalib/src/mesa/sources.mak
@@ -43,6 +43,7 @@ MAIN_SOURCES = \
 	main/ffvertex_prog.c \
 	main/fog.c \
 	main/formats.c \
+	main/format_unpack.c \
 	main/framebuffer.c \
 	main/get.c \
 	main/getstring.c \