Merge remote branch 'origin/master' into gallium_draw_llvm

23 files changed, 1213 insertions, 1849 deletions

diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile
index 452eceb7f4..4c629924b9 100644
--- a/src/gallium/auxiliary/Makefile
+++ b/src/gallium/auxiliary/Makefile
@@ -105,7 +105,8 @@ C_SOURCES = \
 	util/u_cpu_detect.c \
 	util/u_dl.c \
 	util/u_draw_quad.c \
-	util/u_format_access.c \
+	util/u_format.c \
+	util/u_format_s3tc.c \
 	util/u_format_table.c \
 	util/u_format_tests.c \
 	util/u_gen_mipmap.c \
@@ -148,8 +149,6 @@ GALLIVM_SOURCES = \
         gallivm/lp_bld_debug.c \
         gallivm/lp_bld_depth.c \
         gallivm/lp_bld_flow.c \
-        gallivm/lp_bld_format_aos.c \
-        gallivm/lp_bld_format_query.c \
         gallivm/lp_bld_format_soa.c \
         gallivm/lp_bld_interp.c \
         gallivm/lp_bld_intr.c \
@@ -169,8 +168,6 @@ GALLIVM_CPP_SOURCES = \
 GENERATED_SOURCES = \
 	indices/u_indices_gen.c \
 	indices/u_unfilled_gen.c \
-	util/u_format_access.c \
-	util/u_format_pack.h \
 	util/u_format_table.c
 
 
@@ -194,12 +191,9 @@ indices/u_indices_gen.c: indices/u_indices_gen.py
 indices/u_unfilled_gen.c: indices/u_unfilled_gen.py
 	python $< > $@
 
-util/u_format_table.c: util/u_format_table.py util/u_format_parse.py util/u_format.csv
+util/u_format_table.c: util/u_format_table.py util/u_format_pack.py util/u_format_parse.py util/u_format.csv
 	python util/u_format_table.py util/u_format.csv > $@
 
-util/u_format_pack.h: util/u_format_pack.py util/u_format_parse.py util/u_format.csv
-	python util/u_format_pack.py util/u_format.csv > $@
-
 util/u_format_access.c: util/u_format_access.py util/u_format_parse.py util/u_format.csv
 	python util/u_format_access.py util/u_format.csv > $@
 
diff --git a/src/gallium/auxiliary/SConscript b/src/gallium/auxiliary/SConscript
index d0785bce16..b4e8eb041a 100644
--- a/src/gallium/auxiliary/SConscript
+++ b/src/gallium/auxiliary/SConscript
@@ -29,20 +29,10 @@ env.CodeGenerate(
     source = ['util/u_format.csv'],
     command = 'python $SCRIPT $SOURCE > $TARGET'
 )
-
-env.CodeGenerate(
-    target = File('util/u_format_pack.h').srcnode(),
-    script = 'util/u_format_pack.py',
-    source = ['util/u_format.csv'],
-    command = 'python $SCRIPT $SOURCE > $TARGET'
-)
-
-env.CodeGenerate(
-    target = 'util/u_format_access.c',
-    script = 'util/u_format_access.py',
-    source = ['util/u_format.csv'],
-    command = 'python $SCRIPT $SOURCE > $TARGET'
-)
+env.Depends('util/u_format_table.c', [
+    'util/u_format_parse.py', 
+    'util/u_format_pack.py', 
+])
 
 source = [
     'cso_cache/cso_context.c',
@@ -149,7 +139,8 @@ source = [
     'util/u_dump_state.c',
     'util/u_dl.c',
     'util/u_draw_quad.c',
-    'util/u_format_access.c',
+    'util/u_format.c',
+    'util/u_format_s3tc.c',
     'util/u_format_table.c',
     'util/u_format_tests.c',
     'util/u_gen_mipmap.c',
@@ -192,8 +183,6 @@ if drawllvm:
     'gallivm/lp_bld_debug.c',
     'gallivm/lp_bld_depth.c',
     'gallivm/lp_bld_flow.c',
-    'gallivm/lp_bld_format_aos.c',
-    'gallivm/lp_bld_format_query.c',
     'gallivm/lp_bld_format_soa.c',
     'gallivm/lp_bld_interp.c',
     'gallivm/lp_bld_intr.c',
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_depth.c b/src/gallium/auxiliary/gallivm/lp_bld_depth.c
index 4ce1a27a06..c6d0e4bc04 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_depth.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_depth.c
@@ -495,8 +495,8 @@ lp_build_depth_stencil_test(LLVMBuilderRef builder,
       assert(format_desc->block.height == 1);
 
       if (stencil[0].enabled) {
-         assert(format_desc->format == PIPE_FORMAT_Z24S8_UNORM ||
-                format_desc->format == PIPE_FORMAT_S8Z24_UNORM);
+         assert(format_desc->format == PIPE_FORMAT_Z24_UNORM_S8_USCALED ||
+                format_desc->format == PIPE_FORMAT_S8_USCALED_Z24_UNORM);
       }
 
       assert(z_swizzle < 4);
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c b/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c
deleted file mode 100644
index a07f7418f2..0000000000
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c
+++ /dev/null
@@ -1,383 +0,0 @@
-/**************************************************************************
- *
- * Copyright 2009 VMware, Inc.
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, 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 NON-INFRINGEMENT.
- * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS 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
- * AoS pixel format manipulation.
- *
- * @author Jose Fonseca <jfonseca@vmware.com>
- */
-
-
-#include "util/u_cpu_detect.h"
-#include "util/u_format.h"
-
-#include "lp_bld_type.h"
-#include "lp_bld_const.h"
-#include "lp_bld_swizzle.h"
-#include "lp_bld_format.h"
-
-
-/**
- * Unpack a single pixel into its RGBA components.
- *
- * @param packed integer.
- *
- * @return RGBA in a 4 floats vector.
- *
- * XXX: This is mostly for reference and testing -- operating a single pixel at
- * a time is rarely if ever needed.
- */
-LLVMValueRef
-lp_build_unpack_rgba_aos(LLVMBuilderRef builder,
-                         const struct util_format_description *desc,
-                         LLVMValueRef packed)
-{
-   LLVMTypeRef type;
-   LLVMValueRef shifted, casted, scaled, masked;
-   LLVMValueRef shifts[4];
-   LLVMValueRef masks[4];
-   LLVMValueRef scales[4];
-   LLVMValueRef swizzles[4];
-   LLVMValueRef aux[4];
-   bool normalized;
-   int empty_channel;
-   unsigned shift;
-   unsigned i;
-
-   /* FIXME: Support more formats */
-   assert(desc->layout == UTIL_FORMAT_LAYOUT_PLAIN);
-   assert(desc->block.width == 1);
-   assert(desc->block.height == 1);
-   assert(desc->block.bits <= 32);
-
-   type = LLVMIntType(desc->block.bits);
-
-   /* Do the intermediate integer computations with 32bit integers since it
-    * matches floating point size */
-   if (desc->block.bits < 32)
-      packed = LLVMBuildZExt(builder, packed, LLVMInt32Type(), "");
-
-   /* Broadcast the packed value to all four channels */
-   packed = LLVMBuildInsertElement(builder,
-                                   LLVMGetUndef(LLVMVectorType(LLVMInt32Type(), 4)),
-                                   packed,
-                                   LLVMConstNull(LLVMInt32Type()),
-                                   "");
-   packed = LLVMBuildShuffleVector(builder,
-                                   packed,
-                                   LLVMGetUndef(LLVMVectorType(LLVMInt32Type(), 4)),
-                                   LLVMConstNull(LLVMVectorType(LLVMInt32Type(), 4)),
-                                   "");
-
-   /* Initialize vector constants */
-   normalized = FALSE;
-   empty_channel = -1;
-   shift = 0;
-   for (i = 0; i < 4; ++i) {
-      unsigned bits = desc->channel[i].size;
-
-      if (desc->channel[i].type == UTIL_FORMAT_TYPE_VOID) {
-         shifts[i] = LLVMGetUndef(LLVMInt32Type());
-         masks[i] = LLVMConstNull(LLVMInt32Type());
-         scales[i] =  LLVMConstNull(LLVMFloatType());
-         empty_channel = i;
-      }
-      else {
-         unsigned mask = (1 << bits) - 1;
-
-         assert(desc->channel[i].type == UTIL_FORMAT_TYPE_UNSIGNED);
-         assert(bits < 32);
-
-         shifts[i] = LLVMConstInt(LLVMInt32Type(), shift, 0);
-         masks[i] = LLVMConstInt(LLVMInt32Type(), mask, 0);
-
-         if (desc->channel[i].normalized) {
-            scales[i] = LLVMConstReal(LLVMFloatType(), 1.0/mask);
-            normalized = TRUE;
-         }
-         else
-            scales[i] =  LLVMConstReal(LLVMFloatType(), 1.0);
-      }
-
-      shift += bits;
-   }
-
-   shifted = LLVMBuildLShr(builder, packed, LLVMConstVector(shifts, 4), "");
-   masked = LLVMBuildAnd(builder, shifted, LLVMConstVector(masks, 4), "");
-   /* UIToFP can't be expressed in SSE2 */
-   casted = LLVMBuildSIToFP(builder, masked, LLVMVectorType(LLVMFloatType(), 4), "");
-
-   if (normalized)
-      scaled = LLVMBuildMul(builder, casted, LLVMConstVector(scales, 4), "");
-   else
-      scaled = casted;
-
-   for (i = 0; i < 4; ++i)
-      aux[i] = LLVMGetUndef(LLVMFloatType());
-
-   for (i = 0; i < 4; ++i) {
-      enum util_format_swizzle swizzle = desc->swizzle[i];
-
-      switch (swizzle) {
-      case UTIL_FORMAT_SWIZZLE_X:
-      case UTIL_FORMAT_SWIZZLE_Y:
-      case UTIL_FORMAT_SWIZZLE_Z:
-      case UTIL_FORMAT_SWIZZLE_W:
-         swizzles[i] = LLVMConstInt(LLVMInt32Type(), swizzle, 0);
-         break;
-      case UTIL_FORMAT_SWIZZLE_0:
-         assert(empty_channel >= 0);
-         swizzles[i] = LLVMConstInt(LLVMInt32Type(), empty_channel, 0);
-         break;
-      case UTIL_FORMAT_SWIZZLE_1:
-         swizzles[i] = LLVMConstInt(LLVMInt32Type(), 4, 0);
-         aux[0] = LLVMConstReal(LLVMFloatType(), 1.0);
-         break;
-      case UTIL_FORMAT_SWIZZLE_NONE:
-         swizzles[i] = LLVMGetUndef(LLVMFloatType());
-         assert(0);
-         break;
-      }
-   }
-
-   return LLVMBuildShuffleVector(builder, scaled, LLVMConstVector(aux, 4), LLVMConstVector(swizzles, 4), "");
-}
-
-
-/**
- * Take a vector with packed pixels and unpack into a rgba8 vector.
- *
- * Formats with bit depth smaller than 32bits are accepted, but they must be
- * padded to 32bits.
- */
-LLVMValueRef
-lp_build_unpack_rgba8_aos(LLVMBuilderRef builder,
-                          const struct util_format_description *desc,
-                          struct lp_type type,
-                          LLVMValueRef packed)
-{
-   struct lp_build_context bld;
-   bool rgba8;
-   LLVMValueRef res;
-   unsigned i;
-
-   lp_build_context_init(&bld, builder, type);
-
-   /* FIXME: Support more formats */
-   assert(desc->layout == UTIL_FORMAT_LAYOUT_PLAIN);
-   assert(desc->block.width == 1);
-   assert(desc->block.height == 1);
-   assert(desc->block.bits <= 32);
-
-   assert(!type.floating);
-   assert(!type.fixed);
-   assert(type.norm);
-   assert(type.width == 8);
-   assert(type.length % 4 == 0);
-
-   rgba8 = TRUE;
-   for(i = 0; i < 4; ++i) {
-      assert(desc->channel[i].type == UTIL_FORMAT_TYPE_UNSIGNED ||
-             desc->channel[i].type == UTIL_FORMAT_TYPE_VOID);
-      if(desc->channel[0].size != 8)
-         rgba8 = FALSE;
-   }
-
-   if(rgba8) {
-      /*
-       * The pixel is already in a rgba8 format variant. All it is necessary
-       * is to swizzle the channels.
-       */
-
-      unsigned char swizzles[4];
-      boolean zeros[4]; /* bitwise AND mask */
-      boolean ones[4]; /* bitwise OR mask */
-      boolean swizzles_needed = FALSE;
-      boolean zeros_needed = FALSE;
-      boolean ones_needed = FALSE;
-
-      for(i = 0; i < 4; ++i) {
-         enum util_format_swizzle swizzle = desc->swizzle[i];
-
-         /* Initialize with the no-op case */
-         swizzles[i] = util_cpu_caps.little_endian ? 3 - i : i;
-         zeros[i] = TRUE;
-         ones[i] = FALSE;
-
-         switch (swizzle) {
-         case UTIL_FORMAT_SWIZZLE_X:
-         case UTIL_FORMAT_SWIZZLE_Y:
-         case UTIL_FORMAT_SWIZZLE_Z:
-         case UTIL_FORMAT_SWIZZLE_W:
-            if(swizzle != swizzles[i]) {
-               swizzles[i] = swizzle;
-               swizzles_needed = TRUE;
-            }
-            break;
-         case UTIL_FORMAT_SWIZZLE_0:
-            zeros[i] = FALSE;
-            zeros_needed = TRUE;
-            break;
-         case UTIL_FORMAT_SWIZZLE_1:
-            ones[i] = TRUE;
-            ones_needed = TRUE;
-            break;
-         case UTIL_FORMAT_SWIZZLE_NONE:
-            assert(0);
-            break;
-         }
-      }
-
-      res = packed;
-
-      if(swizzles_needed)
-         res = lp_build_swizzle1_aos(&bld, res, swizzles);
-
-      if(zeros_needed) {
-         /* Mask out zero channels */
-         LLVMValueRef mask = lp_build_const_mask_aos(type, zeros);
-         res = LLVMBuildAnd(builder, res, mask, "");
-      }
-
-      if(ones_needed) {
-         /* Or one channels */
-         LLVMValueRef mask = lp_build_const_mask_aos(type, ones);
-         res = LLVMBuildOr(builder, res, mask, "");
-      }
-   }
-   else {
-      /* FIXME */
-      assert(0);
-      res = lp_build_undef(type);
-   }
-
-   return res;
-}
-
-
-/**
- * Pack a single pixel.
- *
- * @param rgba 4 float vector with the unpacked components.
- *
- * XXX: This is mostly for reference and testing -- operating a single pixel at
- * a time is rarely if ever needed.
- */
-LLVMValueRef
-lp_build_pack_rgba_aos(LLVMBuilderRef builder,
-                       const struct util_format_description *desc,
-                       LLVMValueRef rgba)
-{
-   LLVMTypeRef type;
-   LLVMValueRef packed = NULL;
-   LLVMValueRef swizzles[4];
-   LLVMValueRef shifted, casted, scaled, unswizzled;
-   LLVMValueRef shifts[4];
-   LLVMValueRef scales[4];
-   bool normalized;
-   unsigned shift;
-   unsigned i, j;
-
-   assert(desc->layout == UTIL_FORMAT_LAYOUT_PLAIN);
-   assert(desc->block.width == 1);
-   assert(desc->block.height == 1);
-
-   type = LLVMIntType(desc->block.bits);
-
-   /* Unswizzle the color components into the source vector. */
-   for (i = 0; i < 4; ++i) {
-      for (j = 0; j < 4; ++j) {
-         if (desc->swizzle[j] == i)
-            break;
-      }
-      if (j < 4)
-         swizzles[i] = LLVMConstInt(LLVMInt32Type(), j, 0);
-      else
-         swizzles[i] = LLVMGetUndef(LLVMInt32Type());
-   }
-
-   unswizzled = LLVMBuildShuffleVector(builder, rgba,
-                                       LLVMGetUndef(LLVMVectorType(LLVMFloatType(), 4)),
-                                       LLVMConstVector(swizzles, 4), "");
-
-   normalized = FALSE;
-   shift = 0;
-   for (i = 0; i < 4; ++i) {
-      unsigned bits = desc->channel[i].size;
-
-      if (desc->channel[i].type == UTIL_FORMAT_TYPE_VOID) {
-         shifts[i] = LLVMGetUndef(LLVMInt32Type());
-         scales[i] =  LLVMGetUndef(LLVMFloatType());
-      }
-      else {
-         unsigned mask = (1 << bits) - 1;
-
-         assert(desc->channel[i].type == UTIL_FORMAT_TYPE_UNSIGNED);
-         assert(bits < 32);
-
-         shifts[i] = LLVMConstInt(LLVMInt32Type(), shift, 0);
-
-         if (desc->channel[i].normalized) {
-            scales[i] = LLVMConstReal(LLVMFloatType(), mask);
-            normalized = TRUE;
-         }
-         else
-            scales[i] =  LLVMConstReal(LLVMFloatType(), 1.0);
-      }
-
-      shift += bits;
-   }
-
-   if (normalized)
-      scaled = LLVMBuildMul(builder, unswizzled, LLVMConstVector(scales, 4), "");
-   else
-      scaled = unswizzled;
-
-   casted = LLVMBuildFPToSI(builder, scaled, LLVMVectorType(LLVMInt32Type(), 4), "");
-
-   shifted = LLVMBuildShl(builder, casted, LLVMConstVector(shifts, 4), "");
-   
-   /* Bitwise or all components */
-   for (i = 0; i < 4; ++i) {
-      if (desc->channel[i].type == UTIL_FORMAT_TYPE_UNSIGNED) {
-         LLVMValueRef component = LLVMBuildExtractElement(builder, shifted, LLVMConstInt(LLVMInt32Type(), i, 0), "");
-         if (packed)
-            packed = LLVMBuildOr(builder, packed, component, "");
-         else
-            packed = component;
-      }
-   }
-
-   if (!packed)
-      packed = LLVMGetUndef(LLVMInt32Type());
-
-   if (desc->block.bits < 32)
-      packed = LLVMBuildTrunc(builder, packed, type, "");
-
-   return packed;
-}
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_query.c b/src/gallium/auxiliary/gallivm/lp_bld_format_query.c
deleted file mode 100644
index f3832d07ff..0000000000
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_query.c
+++ /dev/null
@@ -1,72 +0,0 @@
-/**************************************************************************
- *
- * Copyright 2009 VMware, Inc.
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, 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 NON-INFRINGEMENT.
- * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS 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
- * Utility functions to make assertions about formats.
- *
- * This module centralizes most of logic used when determining what algorithm
- * is most suitable (i.e., most efficient yet correct) for a given format.
- *
- * It might be possible to move some of these functions to u_format module,
- * but since tiny differences in the format my render it more/less
- * appropriate to a given algorithm it is impossible to make any long term
- * guarantee about the semantics of these functions.
- *
- * @author Jose Fonseca <jfonseca@vmware.com>
- */
-
-
-#include "util/u_format.h"
-
-#include "lp_bld_format.h"
-
-
-/**
- * Whether this format is a 4 rgba8 variant
- */
-boolean
-lp_format_is_rgba8(const struct util_format_description *desc)
-{
-   unsigned chan;
-
-   if(desc->block.width != 1 ||
-      desc->block.height != 1 ||
-      desc->block.bits != 32)
-      return FALSE;
-
-   for(chan = 0; chan < 4; ++chan) {
-      if(desc->channel[chan].type != UTIL_FORMAT_TYPE_UNSIGNED &&
-         desc->channel[chan].type != UTIL_FORMAT_TYPE_SIGNED &&
-         desc->channel[chan].type != UTIL_FORMAT_TYPE_VOID)
-         return FALSE;
-      if(desc->channel[chan].size != 8)
-         return FALSE;
-   }
-
-   return TRUE;
-}
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
index 45ee4b12ce..c9fbe0ad10 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
@@ -92,14 +92,14 @@ lp_build_unpack_rgba_soa(LLVMBuilderRef builder,
    unsigned chan;
 
    /* FIXME: Support more formats */
-   assert(format_desc->layout == UTIL_FORMAT_LAYOUT_PLAIN);
+   assert(format_desc->is_bitmask);
    assert(format_desc->block.width == 1);
    assert(format_desc->block.height == 1);
    assert(format_desc->block.bits <= 32);
 
    /* Decode the input vector components */
    start = 0;
-   for (chan = 0; chan < 4; ++chan) {
+   for (chan = 0; chan < format_desc->nr_channels; ++chan) {
       unsigned width = format_desc->channel[chan].size;
       unsigned stop = start + width;
       LLVMValueRef input;
@@ -123,7 +123,7 @@ lp_build_unpack_rgba_soa(LLVMBuilderRef builder,
             if(format_desc->channel[chan].normalized)
                input = lp_build_unsigned_norm_to_float(builder, width, type, input);
             else
-               input = LLVMBuildFPToSI(builder, input, lp_build_vec_type(type), "");
+               input = LLVMBuildSIToFP(builder, input, lp_build_vec_type(type), "");
          }
          else {
             /* FIXME */
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_printf.c b/src/gallium/auxiliary/gallivm/lp_bld_printf.c
index 78c9ec778b..e4fa2c264c 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_printf.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_printf.c
@@ -104,8 +104,15 @@ lp_build_printf(LLVMBuilderRef builder, const char *fmt, ...)
    params[0] = LLVMBuildGEP(builder, fmtarg, index, 2, "");
 
    va_start(arglist, fmt);
-   for (i = 1; i <= argcount; i++)
-      params[i] = va_arg(arglist, LLVMValueRef);
+   for (i = 1; i <= argcount; i++) {
+      LLVMValueRef val = va_arg(arglist, LLVMValueRef);
+      LLVMTypeRef type = LLVMTypeOf(val);
+      /* printf wants doubles, so lets convert so that
+       * we can actually print them */
+      if (LLVMGetTypeKind(type) == LLVMFloatTypeKind)
+         val = LLVMBuildFPExt(builder, val, LLVMDoubleType(), "");
+      params[i] = val;
+   }
    va_end(arglist);
 
    return LLVMBuildCall(builder, func_printf, params, argcount + 1, "");
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
index 995c016b9d..5b56f2cf3b 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
@@ -2037,7 +2037,7 @@ lp_build_sample_soa(LLVMBuilderRef builder,
    height_vec = lp_build_broadcast_scalar(&bld.uint_coord_bld, height);
    depth_vec = lp_build_broadcast_scalar(&bld.uint_coord_bld, depth);
 
-   if (lp_format_is_rgba8(bld.format_desc) &&
+   if (util_format_is_rgba8_variant(bld.format_desc) &&
        static_state->target == PIPE_TEXTURE_2D &&
        static_state->min_img_filter == PIPE_TEX_FILTER_LINEAR &&
        static_state->mag_img_filter == PIPE_TEX_FILTER_LINEAR &&
diff --git a/src/gallium/auxiliary/os/os_thread.h b/src/gallium/auxiliary/os/os_thread.h
index a04df4106f..004dad6b63 100644
--- a/src/gallium/auxiliary/os/os_thread.h
+++ b/src/gallium/auxiliary/os/os_thread.h
@@ -40,7 +40,7 @@
 #include "util/u_debug.h" /* for assert */
 
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_EMBEDDED)
 
 #include <pthread.h> /* POSIX threads headers */
 #include <stdio.h> /* for perror() */
@@ -257,7 +257,7 @@ typedef unsigned pipe_condvar;
  * pipe_barrier
  */
 
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_EMBEDDED)
 
 typedef pthread_barrier_t pipe_barrier;
 
@@ -377,7 +377,7 @@ pipe_semaphore_wait(pipe_semaphore *sema)
  */
 
 typedef struct {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_EMBEDDED)
    pthread_key_t key;
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
    DWORD key;
@@ -392,7 +392,7 @@ typedef struct {
 static INLINE void
 pipe_tsd_init(pipe_tsd *tsd)
 {
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_EMBEDDED)
    if (pthread_key_create(&tsd->key, NULL/*free*/) != 0) {
       perror("pthread_key_create(): failed to allocate key for thread specific data");
       exit(-1);
@@ -409,7 +409,7 @@ pipe_tsd_get(pipe_tsd *tsd)
    if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) {
       pipe_tsd_init(tsd);
    }
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_EMBEDDED)
    return pthread_getspecific(tsd->key);
 #elif defined(PIPE_SUBSYSTEM_WINDOWS_USER)
    assert(0);
@@ -426,7 +426,7 @@ pipe_tsd_set(pipe_tsd *tsd, void *value)
    if (tsd->initMagic != (int) PIPE_TSD_INIT_MAGIC) {
       pipe_tsd_init(tsd);
    }
-#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU)
+#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS) || defined(PIPE_OS_APPLE) || defined(PIPE_OS_HAIKU) || defined(PIPE_OS_EMBEDDED)
    if (pthread_setspecific(tsd->key, value) != 0) {
       perror("pthread_set_specific() failed");
       exit(-1);
diff --git a/src/gallium/auxiliary/util/.gitignore b/src/gallium/auxiliary/util/.gitignore
index 448d2f304f..53fe9b8a65 100644
--- a/src/gallium/auxiliary/util/.gitignore
+++ b/src/gallium/auxiliary/util/.gitignore
@@ -1,3 +1 @@
-u_format_access.c
 u_format_table.c
-u_format_pack.h
diff --git a/src/gallium/auxiliary/util/u_format.c b/src/gallium/auxiliary/util/u_format.c
new file mode 100644
index 0000000000..7f16cf7d01
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format.c
@@ -0,0 +1,118 @@
+/**************************************************************************
+ *
+ * Copyright 2010 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, 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 NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS 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
+ * Pixel format accessor functions.
+ *
+ * @author Jose Fonseca <jfonseca@vmware.com>
+ */
+
+#include "u_format.h"
+
+
+void
+util_format_read_4f(enum pipe_format format,
+                    float *dst, unsigned dst_stride,
+                    const void *src, unsigned src_stride,
+                    unsigned x, unsigned y, unsigned w, unsigned h)
+{
+   const struct util_format_description *format_desc;
+   const uint8_t *src_row;
+   float *dst_row;
+
+   format_desc = util_format_description(format);
+
+   assert(x % format_desc->block.width == 0);
+   assert(y % format_desc->block.height == 0);
+
+   src_row = (const uint8_t *)src + y*src_stride + x*(format_desc->block.bits/8);
+   dst_row = dst;
+
+   format_desc->unpack_float(dst_row, dst_stride, src_row, src_stride, w, h);
+}
+
+
+void
+util_format_write_4f(enum pipe_format format,
+                     const float *src, unsigned src_stride,
+                     void *dst, unsigned dst_stride,
+                     unsigned x, unsigned y, unsigned w, unsigned h)
+{
+   const struct util_format_description *format_desc;
+   uint8_t *dst_row;
+   const float *src_row;
+
+   format_desc = util_format_description(format);
+
+   assert(x % format_desc->block.width == 0);
+   assert(y % format_desc->block.height == 0);
+
+   dst_row = (uint8_t *)dst + y*dst_stride + x*(format_desc->block.bits/8);
+   src_row = src;
+
+   format_desc->pack_float(dst_row, dst_stride, src_row, src_stride, w, h);
+}
+
+
+void
+util_format_read_4ub(enum pipe_format format, uint8_t *dst, unsigned dst_stride, const void *src, unsigned src_stride, unsigned x, unsigned y, unsigned w, unsigned h)
+{
+   const struct util_format_description *format_desc;
+   const uint8_t *src_row;
+   uint8_t *dst_row;
+
+   format_desc = util_format_description(format);
+
+   assert(x % format_desc->block.width == 0);
+   assert(y % format_desc->block.height == 0);
+
+   src_row = (const uint8_t *)src + y*src_stride + x*(format_desc->block.bits/8);
+   dst_row = dst;
+
+   format_desc->unpack_8unorm(dst_row, dst_stride, src_row, src_stride, w, h);
+}
+
+
+void
+util_format_write_4ub(enum pipe_format format, const uint8_t *src, unsigned src_stride, void *dst, unsigned dst_stride, unsigned x, unsigned y, unsigned w, unsigned h)
+{
+   const struct util_format_description *format_desc;
+   uint8_t *dst_row;
+   const uint8_t *src_row;
+
+   format_desc = util_format_description(format);
+
+   assert(x % format_desc->block.width == 0);
+   assert(y % format_desc->block.height == 0);
+
+   dst_row = (uint8_t *)dst + y*dst_stride + x*(format_desc->block.bits/8);
+   src_row = src;
+
+   format_desc->pack_8unorm(dst_row, dst_stride, src_row, src_stride, w, h);
+}
+
diff --git a/src/gallium/auxiliary/util/u_format.csv b/src/gallium/auxiliary/util/u_format.csv
index d819bbbde1..0cf539247e 100644
--- a/src/gallium/auxiliary/util/u_format.csv
+++ b/src/gallium/auxiliary/util/u_format.csv
@@ -56,14 +56,18 @@
 # internal formats to base in- ternal formats, and desired component
 # resolutions for each sized internal format."
 
+# None
+# Described as regular uint_8 bytes, i.e. PIPE_FORMAT_R8_USCALED
+PIPE_FORMAT_NONE                  , plain, 1, 1, u8  ,     ,     ,     , x001, rgb
+
 # Typical rendertarget formats
 PIPE_FORMAT_B8G8R8A8_UNORM        , plain, 1, 1, un8 , un8 , un8 , un8 , zyxw, rgb
-PIPE_FORMAT_B8G8R8X8_UNORM        , plain, 1, 1, un8 , un8 , un8 , un8 , zyx1, rgb
+PIPE_FORMAT_B8G8R8X8_UNORM        , plain, 1, 1, un8 , un8 , un8 , x8  , zyx1, rgb
 PIPE_FORMAT_A8R8G8B8_UNORM        , plain, 1, 1, un8 , un8 , un8 , un8 , yzwx, rgb
-PIPE_FORMAT_X8R8G8B8_UNORM        , plain, 1, 1, un8 , un8 , un8 , un8 , yzw1, rgb
+PIPE_FORMAT_X8R8G8B8_UNORM        , plain, 1, 1, x8  , un8 , un8 , un8 , yzw1, rgb
 PIPE_FORMAT_A8B8G8R8_UNORM        , plain, 1, 1, un8 , un8 , un8 , un8 , wzyx, rgb
-PIPE_FORMAT_X8B8G8R8_UNORM        , plain, 1, 1, un8 , un8 , un8 , un8 , wzy1, rgb
-PIPE_FORMAT_B5G5R5X1_UNORM        , plain, 1, 1, un5 , un5 , un5 , un1 , zyx1, rgb
+PIPE_FORMAT_X8B8G8R8_UNORM        , plain, 1, 1, x8  , un8 , un8 , un8 , wzy1, rgb
+PIPE_FORMAT_B5G5R5X1_UNORM        , plain, 1, 1, un5 , un5 , un5 , x1  , zyx1, rgb
 PIPE_FORMAT_B5G5R5A1_UNORM        , plain, 1, 1, un5 , un5 , un5 , un1 , zyxw, rgb
 PIPE_FORMAT_B4G4R4A4_UNORM        , plain, 1, 1, un4 , un4 , un4 , un4 , zyxw, rgb
 PIPE_FORMAT_B5G6R5_UNORM          , plain, 1, 1, un5 , un6 , un5 ,     , zyx1, rgb
@@ -80,26 +84,28 @@ PIPE_FORMAT_L16_UNORM             , plain, 1, 1, un16,     ,     ,     , xxx1, r
 PIPE_FORMAT_L8_SRGB               , plain, 1, 1, un8 ,     ,     ,     , xxx1, srgb 
 PIPE_FORMAT_L8A8_SRGB             , plain, 1, 1, un8 , un8 ,     ,     , xxxy, srgb 
 PIPE_FORMAT_R8G8B8_SRGB           , plain, 1, 1, un8 , un8 , un8 ,     , xyz1, srgb 
+PIPE_FORMAT_R8G8B8A8_SRGB         , plain, 1, 1, un8 , un8 , un8 , un8 , xyzw, srgb 
 PIPE_FORMAT_A8B8G8R8_SRGB         , plain, 1, 1, un8 , un8 , un8 , un8 , wzyx, srgb
-PIPE_FORMAT_X8B8G8R8_SRGB         , plain, 1, 1, un8 , un8 , un8 , un8 , wzy1, srgb
+PIPE_FORMAT_X8B8G8R8_SRGB         , plain, 1, 1, x8  , un8 , un8 , un8 , wzy1, srgb
 PIPE_FORMAT_B8G8R8A8_SRGB         , plain, 1, 1, un8 , un8 , un8 , un8 , zyxw, srgb
-PIPE_FORMAT_B8G8R8X8_SRGB         , plain, 1, 1, un8 , un8 , un8 , un8 , zyx1, srgb
+PIPE_FORMAT_B8G8R8X8_SRGB         , plain, 1, 1, un8 , un8 , un8 , x8  , zyx1, srgb
 PIPE_FORMAT_A8R8G8B8_SRGB         , plain, 1, 1, un8 , un8 , un8 , un8 , yzwx, srgb
-PIPE_FORMAT_X8R8G8B8_SRGB         , plain, 1, 1, un8 , un8 , un8 , un8 , yzw1, srgb
+PIPE_FORMAT_X8R8G8B8_SRGB         , plain, 1, 1, x8  , un8 , un8 , un8 , yzw1, srgb
 
 # Mixed-sign formats (typically used for bump map textures)
 PIPE_FORMAT_R8SG8SB8UX8U_NORM     , plain, 1, 1, sn8 , sn8 , un8 , x8  , xyz1, rgb
 PIPE_FORMAT_R5SG5SB6U_NORM        , plain, 1, 1, sn5 , sn5 , un6 ,     , xyz1, rgb
 
 # Depth-stencil formats
-PIPE_FORMAT_S8_UNORM              , plain, 1, 1, un8 ,     ,     ,     , _x__, zs 
-PIPE_FORMAT_Z16_UNORM             , plain, 1, 1, un16,     ,     ,     , x___, zs 
-PIPE_FORMAT_Z32_UNORM             , plain, 1, 1, un32,     ,     ,     , x___, zs 
-PIPE_FORMAT_Z32_FLOAT             , plain, 1, 1, f32 ,     ,     ,     , x___, zs 
-PIPE_FORMAT_Z24S8_UNORM           , plain, 1, 1, un24, un8 ,     ,     , xy__, zs 
-PIPE_FORMAT_S8Z24_UNORM           , plain, 1, 1, un8 , un24,     ,     , yx__, zs 
-PIPE_FORMAT_Z24X8_UNORM           , plain, 1, 1, un24, un8 ,     ,     , x___, zs 
-PIPE_FORMAT_X8Z24_UNORM           , plain, 1, 1, un8 , un24,     ,     , y___, zs 
+PIPE_FORMAT_S8_USCALED              , plain, 1, 1, u8  ,     ,     ,     , _x__, zs
+PIPE_FORMAT_Z16_UNORM               , plain, 1, 1, un16,     ,     ,     , x___, zs
+PIPE_FORMAT_Z32_UNORM               , plain, 1, 1, un32,     ,     ,     , x___, zs
+PIPE_FORMAT_Z32_FLOAT               , plain, 1, 1, f32 ,     ,     ,     , x___, zs
+PIPE_FORMAT_Z24_UNORM_S8_USCALED    , plain, 1, 1, un24, u8  ,     ,     , xy__, zs
+PIPE_FORMAT_S8_USCALED_Z24_UNORM    , plain, 1, 1, u8 ,  un24,     ,     , yx__, zs
+PIPE_FORMAT_Z24X8_UNORM             , plain, 1, 1, un24, x8  ,     ,     , x___, zs
+PIPE_FORMAT_X8Z24_UNORM             , plain, 1, 1, x8  , un24,     ,     , y___, zs
+PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED , plain, 1, 1, f32,  u8  , x24 ,     , xy__, zs
 
 # YUV formats
 # http://www.fourcc.org/yuv.php#UYVY
@@ -108,6 +114,15 @@ PIPE_FORMAT_UYVY                 , subsampled, 2, 1, x32 ,     ,     ,     , xyz
 # XXX: u_tile.c's ycbcr_get_tile_rgba actually interprets it as VYUY but the 
 # intent should be to match D3DFMT_YUY2
 PIPE_FORMAT_YUYV                 , subsampled, 2, 1, x32 ,     ,     ,     , xyz1, yuv
+# same subsampling but with rgb channels
+PIPE_FORMAT_R8G8_B8G8_UNORM      , subsampled, 2, 1, x32 ,     ,     ,     , xyz1, rgb
+PIPE_FORMAT_G8R8_G8B8_UNORM      , subsampled, 2, 1, x32 ,     ,     ,     , xyz1, rgb
+
+# some special formats not fitting anywhere else
+PIPE_FORMAT_R10G10B10A2_USCALED   , plain,      1,  1, u10 , u10 , u10 , u2  , xyzw, rgb
+PIPE_FORMAT_R11G11B10_FLOAT       , plain,      1,  1, f11 , f11 , f10 ,     , xyz1, rgb
+PIPE_FORMAT_R9G9B9E5_FLOAT        , other,      1,  1, x32 ,     ,     ,     , xyz1, rgb
+PIPE_FORMAT_R1_UNORM              , other,      8,  1, x8  ,     ,     ,     , x001, rgb
 
 # Compressed formats
 # - http://en.wikipedia.org/wiki/S3_Texture_Compression
@@ -123,6 +138,11 @@ PIPE_FORMAT_DXT1_SRGBA            , s3tc, 4, 4, x64 ,     ,     ,     , xyzw, sr
 PIPE_FORMAT_DXT3_SRGBA            , s3tc, 4, 4, x128,     ,     ,     , xyzw, srgb
 PIPE_FORMAT_DXT5_SRGBA            , s3tc, 4, 4, x128,     ,     ,     , xyzw, srgb
 
+PIPE_FORMAT_RGTC1_UNORM           , rgtc, 4, 4, x64,      ,     ,     , x001, rgb
+PIPE_FORMAT_RGTC1_SNORM           , rgtc, 4, 4, x64,      ,     ,     , x001, rgb
+PIPE_FORMAT_RGTC2_UNORM           , rgtc, 4, 4, x128,     ,     ,     , xy01, rgb
+PIPE_FORMAT_RGTC2_SNORM           , rgtc, 4, 4, x128,     ,     ,     , xy01, rgb
+
 # Straightforward D3D10-like formats (also used for 
 # vertex buffer element description)
 # 
@@ -157,6 +177,10 @@ PIPE_FORMAT_R32_FIXED             , plain, 1, 1, h32 ,     ,     ,     , x001, r
 PIPE_FORMAT_R32G32_FIXED          , plain, 1, 1, h32 , h32 ,     ,     , xy01, rgb
 PIPE_FORMAT_R32G32B32_FIXED       , plain, 1, 1, h32 , h32 , h32 ,     , xyz1, rgb
 PIPE_FORMAT_R32G32B32A32_FIXED    , plain, 1, 1, h32 , h32 , h32 , h32 , xyzw, rgb
+PIPE_FORMAT_R16_FLOAT             , plain, 1, 1, f16 ,     ,     ,     , x001, rgb
+PIPE_FORMAT_R16G16_FLOAT          , plain, 1, 1, f16 , f16 ,     ,     , xy01, rgb
+PIPE_FORMAT_R16G16B16_FLOAT       , plain, 1, 1, f16 , f16 , f16 ,     , xyz1, rgb
+PIPE_FORMAT_R16G16B16A16_FLOAT    , plain, 1, 1, f16 , f16 , f16 , f16 , xyzw, rgb
 PIPE_FORMAT_R16_UNORM             , plain, 1, 1, un16,     ,     ,     , x001, rgb
 PIPE_FORMAT_R16G16_UNORM          , plain, 1, 1, un16, un16,     ,     , xy01, rgb
 PIPE_FORMAT_R16G16B16_UNORM       , plain, 1, 1, un16, un16, un16,     , xyz1, rgb
diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h
index 98d4b98ebb..93818a3161 100644
--- a/src/gallium/auxiliary/util/u_format.h
+++ b/src/gallium/auxiliary/util/u_format.h
@@ -189,6 +189,46 @@ struct util_format_description
     * Colorspace transformation.
     */
    enum util_format_colorspace colorspace;
+
+   /**
+    * Unpack pixel blocks to R8G8B8A8_UNORM.
+    */
+   void
+   (*unpack_8unorm)(uint8_t *dst, unsigned dst_stride,
+                    const uint8_t *src, unsigned src_stride,
+                    unsigned width, unsigned height);
+
+   /**
+    * Pack pixel blocks from R8G8B8A8_UNORM.
+    */
+   void
+   (*pack_8unorm)(uint8_t *dst, unsigned dst_stride,
+                  const uint8_t *src, unsigned src_stride,
+                  unsigned width, unsigned height);
+
+   /**
+    * Unpack pixel blocks to R32G32B32A32_FLOAT.
+    */
+   void
+   (*unpack_float)(float *dst, unsigned dst_stride,
+                   const uint8_t *src, unsigned src_stride,
+                   unsigned width, unsigned height);
+
+   /**
+    * Pack pixel blocks from R32G32B32A32_FLOAT.
+    */
+   void
+   (*pack_float)(uint8_t *dst, unsigned dst_stride,
+                 const float *src, unsigned src_stride,
+                 unsigned width, unsigned height);
+
+   /**
+    * Fetch a single pixel (i, j) from a block.
+    */
+   void
+   (*fetch_float)(float *dst,
+                  const uint8_t *src,
+                  unsigned i, unsigned j);
 };
 
 
@@ -209,8 +249,8 @@ util_format_name(enum pipe_format format)
 {
    const struct util_format_description *desc = util_format_description(format);
 
-   assert(format);
-   if (!format) {
+   assert(desc);
+   if (!desc) {
       return "???";
    }
 
@@ -222,8 +262,8 @@ util_format_is_s3tc(enum pipe_format format)
 {
    const struct util_format_description *desc = util_format_description(format);
 
-   assert(format);
-   if (!format) {
+   assert(desc);
+   if (!desc) {
       return FALSE;
    }
 
@@ -235,8 +275,8 @@ util_format_is_depth_or_stencil(enum pipe_format format)
 {
    const struct util_format_description *desc = util_format_description(format);
 
-   assert(format);
-   if (!format) {
+   assert(desc);
+   if (!desc) {
       return FALSE;
    }
 
@@ -248,8 +288,8 @@ util_format_is_depth_and_stencil(enum pipe_format format)
 {
    const struct util_format_description *desc = util_format_description(format);
 
-   assert(format);
-   if (!format) {
+   assert(desc);
+   if (!desc) {
       return FALSE;
    }
 
@@ -261,6 +301,34 @@ util_format_is_depth_and_stencil(enum pipe_format format)
            desc->swizzle[1] != UTIL_FORMAT_SWIZZLE_NONE) ? TRUE : FALSE;
 }
 
+/**
+ * Whether this format is a rgab8 variant.
+ *
+ * That is, any format that matches the
+ *
+ *   PIPE_FORMAT_?8?8?8?8_UNORM
+ */
+static INLINE boolean
+util_format_is_rgba8_variant(const struct util_format_description *desc)
+{
+   unsigned chan;
+
+   if(desc->block.width != 1 ||
+      desc->block.height != 1 ||
+      desc->block.bits != 32)
+      return FALSE;
+
+   for(chan = 0; chan < 4; ++chan) {
+      if(desc->channel[chan].type != UTIL_FORMAT_TYPE_UNSIGNED &&
+         desc->channel[chan].type != UTIL_FORMAT_TYPE_VOID)
+         return FALSE;
+      if(desc->channel[chan].size != 8)
+         return FALSE;
+   }
+
+   return TRUE;
+}
+
 
 /**
  * Return total bits needed for the pixel format per block.
@@ -270,8 +338,8 @@ util_format_get_blocksizebits(enum pipe_format format)
 {
    const struct util_format_description *desc = util_format_description(format);
 
-   assert(format);
-   if (!format) {
+   assert(desc);
+   if (!desc) {
       return 0;
    }
 
@@ -296,8 +364,8 @@ util_format_get_blockwidth(enum pipe_format format)
 {
    const struct util_format_description *desc = util_format_description(format);
 
-   assert(format);
-   if (!format) {
+   assert(desc);
+   if (!desc) {
       return 1;
    }
 
@@ -309,8 +377,8 @@ util_format_get_blockheight(enum pipe_format format)
 {
    const struct util_format_description *desc = util_format_description(format);
 
-   assert(format);
-   if (!format) {
+   assert(desc);
+   if (!desc) {
       return 1;
    }
 
diff --git a/src/gallium/auxiliary/util/u_format_access.py b/src/gallium/auxiliary/util/u_format_access.py
deleted file mode 100644
index 00424779d2..0000000000
--- a/src/gallium/auxiliary/util/u_format_access.py
+++ /dev/null
@@ -1,341 +0,0 @@
-#!/usr/bin/env python
-
-'''
-/**************************************************************************
- *
- * Copyright 2009 VMware, Inc.
- * All Rights Reserved.
- *
- * Permission is hereby granted, free of charge, to any person obtaining a
- * copy of this software and associated documentation files (the
- * "Software"), to deal in the Software without restriction, including
- * without limitation the rights to use, copy, modify, merge, publish,
- * distribute, sub license, 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 NON-INFRINGEMENT.
- * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS 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
- * Pixel format accessor functions.
- *
- * @author Jose Fonseca <jfonseca@vmware.com>
- */
-'''
-
-
-import math
-import sys
-
-from u_format_pack import *
-
-
-def is_format_supported(format):
-    '''Determines whether we actually have the plumbing necessary to generate the 
-    to read/write to/from this format.'''
-
-    # FIXME: Ideally we would support any format combination here.
-
-    # XXX: It should be straightforward to support srgb
-    if format.colorspace not in ('rgb', 'zs'):
-        return False
-
-    if format.layout != PLAIN:
-        return False
-
-    for i in range(4):
-        channel = format.channels[i]
-        if channel.type not in (VOID, UNSIGNED, FLOAT):
-            return False
-
-    # We can only read a color from a depth/stencil format if the depth channel is present
-    if format.colorspace == 'zs' and format.swizzles[0] == SWIZZLE_NONE:
-        return False
-
-    return True
-
-
-def native_type(format):
-    '''Get the native appropriate for a format.'''
-
-    if format.layout == PLAIN:
-        if not format.is_array():
-            # For arithmetic pixel formats return the integer type that matches the whole pixel
-            return 'uint%u_t' % format.block_size()
-        else:
-            # For array pixel formats return the integer type that matches the color channel
-            channel = format.channels[0]
-            if channel.type == UNSIGNED:
-                return 'uint%u_t' % channel.size
-            elif channel.type == SIGNED:
-                return 'int%u_t' % channel.size
-            elif channel.type == FLOAT:
-                if channel.size == 32:
-                    return 'float'
-                elif channel.size == 64:
-                    return 'double'
-                else:
-                    assert False
-            else:
-                assert False
-    else:
-        assert False
-
-
-def generate_srgb_tables():
-    print 'static ubyte srgb_to_linear[256] = {'
-    for i in range(256):
-        print '   %s,' % (int(math.pow((i / 255.0 + 0.055) / 1.055, 2.4) * 255))
-    print '};'
-    print
-    print 'static ubyte linear_to_srgb[256] = {'
-    print '   0,'
-    for i in range(1, 256):
-        print '   %s,' % (int((1.055 * math.pow(i / 255.0, 0.41666) - 0.055) * 255))
-    print '};'
-    print
-
-
-def generate_format_read(format, dst_channel, dst_native_type, dst_suffix):
-    '''Generate the function to read pixels from a particular format'''
-
-    name = format.short_name()
-
-    src_native_type = native_type(format)
-
-    print 'static void'
-    print 'util_format_%s_read_%s(%s *dst, unsigned dst_stride, const uint8_t *src, unsigned src_stride, unsigned x0, unsigned y0, unsigned w, unsigned h)' % (name, dst_suffix, dst_native_type)
-    print '{'
-    print '   unsigned x, y;'
-    print '   const uint8_t *src_row = src + y0*src_stride;'
-    print '   %s *dst_row = dst;' % dst_native_type
-    print '   for (y = 0; y < h; ++y) {'
-    print '      const %s *src_pixel = (const %s *)(src_row + x0*%u);' % (src_native_type, src_native_type, format.stride())
-    print '      %s *dst_pixel = dst_row;' %dst_native_type
-    print '      for (x = 0; x < w; ++x) {'
-
-    names = ['']*4
-    if format.colorspace == 'rgb':
-        for i in range(4):
-            swizzle = format.swizzles[i]
-            if swizzle < 4:
-                names[swizzle] += 'rgba'[i]
-    elif format.colorspace == 'zs':
-        swizzle = format.swizzles[0]
-        if swizzle < 4:
-            names[swizzle] = 'z'
-        else:
-            assert False
-    else:
-        assert False
-
-    if format.layout == PLAIN:
-        if not format.is_array():
-            print '         %s pixel = *src_pixel++;' % src_native_type
-            shift = 0;
-            for i in range(4):
-                src_channel = format.channels[i]
-                width = src_channel.size
-                if names[i]:
-                    value = 'pixel'
-                    mask = (1 << width) - 1
-                    if shift:
-                        value = '(%s >> %u)' % (value, shift)
-                    if shift + width < format.block_size():
-                        value = '(%s & 0x%x)' % (value, mask)
-                    value = conversion_expr(src_channel, dst_channel, dst_native_type, value)
-                    print '         %s %s = %s;' % (dst_native_type, names[i], value)
-                shift += width
-        else:
-            for i in range(4):
-                src_channel = format.channels[i]
-                if names[i]:
-                    value = 'src_pixel[%u]' % i
-                    value = conversion_expr(src_channel, dst_channel, dst_native_type, value)
-                    print '         %s %s = %s;' % (dst_native_type, names[i], value)
-            print '         src_pixel += %u;' % (format.nr_channels())
-    else:
-        assert False
-
-    for i in range(4):
-        if format.colorspace == 'rgb':
-            swizzle = format.swizzles[i]
-            if swizzle < 4:
-                value = names[swizzle]
-            elif swizzle == SWIZZLE_0:
-                value = '0'
-            elif swizzle == SWIZZLE_1:
-                value = get_one(dst_channel)
-            else:
-                assert False
-        elif format.colorspace == 'zs':
-            if i < 3:
-                value = 'z'
-            else:
-                value = get_one(dst_channel)
-        else:
-            assert False
-        print '         *dst_pixel++ = %s; /* %s */' % (value, 'rgba'[i])
-
-    print '      }'
-    print '      src_row += src_stride;'
-    print '      dst_row += dst_stride/sizeof(*dst_row);'
-    print '   }'
-    print '}'
-    print
-    
-
-def generate_format_write(format, src_channel, src_native_type, src_suffix):
-    '''Generate the function to write pixels to a particular format'''
-
-    name = format.short_name()
-
-    dst_native_type = native_type(format)
-
-    print 'static void'
-    print 'util_format_%s_write_%s(const %s *src, unsigned src_stride, uint8_t *dst, unsigned dst_stride, unsigned x0, unsigned y0, unsigned w, unsigned h)' % (name, src_suffix, src_native_type)
-    print '{'
-    print '   unsigned x, y;'
-    print '   uint8_t *dst_row = dst + y0*dst_stride;'
-    print '   const %s *src_row = src;' % src_native_type
-    print '   for (y = 0; y < h; ++y) {'
-    print '      %s *dst_pixel = (%s *)(dst_row + x0*%u);' % (dst_native_type, dst_native_type, format.stride())
-    print '      const %s *src_pixel = src_row;' %src_native_type
-    print '      for (x = 0; x < w; ++x) {'
-
-    inv_swizzle = format.inv_swizzles()
-
-    if format.layout == PLAIN:
-        if not format.is_array():
-            print '         %s pixel = 0;' % dst_native_type
-            shift = 0;
-            for i in range(4):
-                dst_channel = format.channels[i]
-                width = dst_channel.size
-                if inv_swizzle[i] is not None:
-                    value = 'src_pixel[%u]' % inv_swizzle[i]
-                    value = conversion_expr(src_channel, dst_channel, dst_native_type, value)
-                    if shift:
-                        value = '(%s << %u)' % (value, shift)
-                    print '         pixel |= %s;' % value
-                shift += width
-            print '         *dst_pixel++ = pixel;'
-        else:
-            for i in range(4):
-                dst_channel = format.channels[i]
-                if inv_swizzle[i] is not None:
-                    value = 'src_pixel[%u]' % inv_swizzle[i]
-                    value = conversion_expr(src_channel, dst_channel, dst_native_type, value)
-                    print '         *dst_pixel++ = %s;' % value
-    else:
-        assert False
-    print '         src_pixel += 4;'
-
-    print '      }'
-    print '      dst_row += dst_stride;'
-    print '      src_row += src_stride/sizeof(*src_row);'
-    print '   }'
-    print '}'
-    print
-    
-
-def generate_read(formats, dst_channel, dst_native_type, dst_suffix):
-    '''Generate the dispatch function to read pixels from any format'''
-
-    for format in formats:
-        if is_format_supported(format):
-            generate_format_read(format, dst_channel, dst_native_type, dst_suffix)
-
-    print 'void'
-    print 'util_format_read_%s(enum pipe_format format, %s *dst, unsigned dst_stride, const void *src, unsigned src_stride, unsigned x, unsigned y, unsigned w, unsigned h)' % (dst_suffix, dst_native_type)
-    print '{'
-    print '   void (*func)(%s *dst, unsigned dst_stride, const uint8_t *src, unsigned src_stride, unsigned x0, unsigned y0, unsigned w, unsigned h);' % dst_native_type
-    print '   switch(format) {'
-    for format in formats:
-        if is_format_supported(format):
-            print '   case %s:' % format.name
-            print '      func = &util_format_%s_read_%s;' % (format.short_name(), dst_suffix)
-            print '      break;'
-    print '   default:'
-    print '      debug_printf("unsupported format\\n");'
-    print '      return;'
-    print '   }'
-    print '   func(dst, dst_stride, (const uint8_t *)src, src_stride, x, y, w, h);'
-    print '}'
-    print
-
-
-def generate_write(formats, src_channel, src_native_type, src_suffix):
-    '''Generate the dispatch function to write pixels to any format'''
-
-    for format in formats:
-        if is_format_supported(format):
-            generate_format_write(format, src_channel, src_native_type, src_suffix)
-
-    print 'void'
-    print 'util_format_write_%s(enum pipe_format format, const %s *src, unsigned src_stride, void *dst, unsigned dst_stride, unsigned x, unsigned y, unsigned w, unsigned h)' % (src_suffix, src_native_type)
-    
-    print '{'
-    print '   void (*func)(const %s *src, unsigned src_stride, uint8_t *dst, unsigned dst_stride, unsigned x0, unsigned y0, unsigned w, unsigned h);' % src_native_type
-    print '   switch(format) {'
-    for format in formats:
-        if is_format_supported(format):
-            print '   case %s:' % format.name
-            print '      func = &util_format_%s_write_%s;' % (format.short_name(), src_suffix)
-            print '      break;'
-    print '   default:'
-    print '      debug_printf("unsupported format\\n");'
-    print '      return;'
-    print '   }'
-    print '   func(src, src_stride, (uint8_t *)dst, dst_stride, x, y, w, h);'
-    print '}'
-    print
-
-
-def main():
-    formats = []
-    for arg in sys.argv[1:]:
-        formats.extend(parse(arg))
-
-    print '/* This file is autogenerated by u_format_access.py from u_format.csv. Do not edit directly. */'
-    print
-    # This will print the copyright message on the top of this file
-    print __doc__.strip()
-    print
-    print '#include "pipe/p_compiler.h"'
-    print '#include "u_math.h"'
-    print '#include "u_format_pack.h"'
-    print
-
-    generate_srgb_tables()
-
-    type = Channel(FLOAT, False, 32)
-    native_type = 'float'
-    suffix = '4f'
-
-    generate_read(formats, type, native_type, suffix)
-    generate_write(formats, type, native_type, suffix)
-
-    type = Channel(UNSIGNED, True, 8)
-    native_type = 'uint8_t'
-    suffix = '4ub'
-
-    generate_read(formats, type, native_type, suffix)
-    generate_write(formats, type, native_type, suffix)
-
-
-if __name__ == '__main__':
-    main()
diff --git a/src/gallium/auxiliary/util/u_format_pack.py b/src/gallium/auxiliary/util/u_format_pack.py
index 409d024c63..95c0c79be9 100644
--- a/src/gallium/auxiliary/util/u_format_pack.py
+++ b/src/gallium/auxiliary/util/u_format_pack.py
@@ -38,19 +38,59 @@
 
 
 import sys
+import math
 
 from u_format_parse import *
 
 
+def generate_f16_to_f32():
+    '''Naive implementation, need something faster that operates on bits'''
+
+    print '''
+static float
+f16_to_f32(uint16_t h)
+{
+    unsigned mantissa = h & 0x3ff;
+    unsigned exponent = (h >> 10) & 0x1f;
+    float sign = (h & 0x8000) ? -1.0f : 1.0f;
+
+    if (exponent == 0) {
+        if (mantissa == 0) {
+            return sign * 0.0f;
+        }
+        return sign * powf(2.0f, -14.0f) * (float)mantissa / 1024.0f;
+    }
+    if (exponent == 31) {
+        if (mantissa == 0) {
+            /* XXX: infinity */
+            return sign * 100000.0f;
+        }
+        /* XXX: NaN */
+        return 1000.0f;
+    }
+    return sign * powf(2.0f, (float)exponent - 15.0f) * (1.0f + (float)mantissa / 1024.0f);
+}
+'''
+
+def generate_f32_to_f16():
+    print '''
+static uint16_t
+f32_to_f16(float f)
+{
+    /* TODO */
+    return 0;
+}
+'''
+
 def generate_format_type(format):
     '''Generate a structure that describes the format.'''
 
     print 'union util_format_%s {' % format.short_name()
-    if format.is_bitmask():
+    if format.is_bitmask() or format.short_name() == "r11g11b10_float":
         print '   uint%u_t value;' % (format.block_size(),)
     print '   struct {'
     for channel in format.channels:
-        if format.is_bitmask() and not format.is_array():
+        if (format.is_bitmask() or format.is_mixed()) and not format.is_array() or format.short_name() == "r11g11b10_float":
             if channel.type == VOID:
                 if channel.size:
                     print '      unsigned %s:%u;' % (channel.name, channel.size)
@@ -58,6 +98,11 @@ def generate_format_type(format):
                 print '      unsigned %s:%u;' % (channel.name, channel.size)
             elif channel.type == SIGNED:
                 print '      int %s:%u;' % (channel.name, channel.size)
+            elif channel.type == FLOAT:
+                if channel.size == 32:
+                    print '      float %s;' % (channel.name)
+                else:
+                    print '      unsigned %s:%u;' % (channel.name, channel.size)
             else:
                 assert 0
         else:
@@ -85,6 +130,20 @@ def generate_format_type(format):
     print
 
 
+def generate_srgb_tables():
+    print 'static ubyte srgb_to_linear[256] = {'
+    for i in range(256):
+        print '   %s,' % (int(math.pow((i / 255.0 + 0.055) / 1.055, 2.4) * 255))
+    print '};'
+    print
+    print 'static ubyte linear_to_srgb[256] = {'
+    print '   0,'
+    for i in range(1, 256):
+        print '   %s,' % (int((1.055 * math.pow(i / 255.0, 0.41666) - 0.055) * 255))
+    print '};'
+    print
+
+
 def bswap_format(format):
     '''Generate a structure that describes the format.'''
 
@@ -124,15 +183,17 @@ def native_type(format):
             return 'uint%u_t' % format.block_size()
         else:
             # For array pixel formats return the integer type that matches the color channel
-            type = format.channels[0]
-            if type.type == UNSIGNED:
-                return 'uint%u_t' % type.size
-            elif type.type == SIGNED:
-                return 'int%u_t' % type.size
-            elif type.type == FLOAT:
-                if type.size == 32:
+            channel = format.channels[0]
+            if channel.type in (UNSIGNED, VOID):
+                return 'uint%u_t' % channel.size
+            elif channel.type == SIGNED:
+                return 'int%u_t' % channel.size
+            elif channel.type == FLOAT:
+                if channel.size == 16:
+                    return 'uint16_t'
+                elif channel.size == 32:
                     return 'float'
-                elif type.size == 64:
+                elif channel.size == 64:
                     return 'double'
                 else:
                     assert False
@@ -179,31 +240,6 @@ def get_one(type):
         return (1 << get_one_shift(type)) - 1
 
 
-def generate_clamp():
-    '''Code generate the clamping functions for each type.
-
-    We don't use a macro so that arguments with side effects, 
-    like *src_pixel++ are correctly handled.
-    '''
-
-    for suffix, native_type in [
-        ('', 'double'),
-        ('f', 'float'),
-        ('ui', 'unsigned int'),
-        ('si', 'int'),
-    ]:
-        print 'static INLINE %s' % native_type
-        print 'clamp%s(%s value, %s lbound, %s ubound)' % (suffix, native_type, native_type, native_type)
-        print '{'
-        print '   if(value < lbound)'
-        print '      return lbound;'
-        print '   if(value > ubound)'
-        print '      return ubound;'
-        print '   return value;'
-        print '}'
-        print
-
-
 def clamp_expr(src_channel, dst_channel, dst_native_type, value):
     '''Generate the expression to clamp the value in the source type to the
     destination type range.'''
@@ -211,21 +247,6 @@ def clamp_expr(src_channel, dst_channel, dst_native_type, value):
     if src_channel == dst_channel:
         return value
 
-    # Pick the approriate clamp function
-    if src_channel.type == FLOAT:
-        if src_channel.size == 32:
-            func = 'clampf'
-        elif src_channel.size == 64:
-            func = 'clamp'
-        else:
-            assert False
-    elif src_channel.type == UNSIGNED:
-        func = 'clampui'
-    elif src_channel.type == SIGNED:
-        func = 'clampsi'
-    else:
-        assert False
-
     src_min = src_channel.min()
     src_max = src_channel.max()
     dst_min = dst_channel.min()
@@ -250,7 +271,17 @@ def conversion_expr(src_channel, dst_channel, dst_native_type, value, clamp=True
         return value
 
     if src_channel.type == FLOAT and dst_channel.type == FLOAT:
-        return '(%s)%s' % (dst_native_type, value)
+        if src_channel.size == dst_channel.size:
+            return value
+        if src_channel.size == 64:
+            value = '(float)%s' % (value)
+        elif src_channel.size == 16:
+            value = 'f16_to_f32(%s)' % (value)
+        if dst_channel.size == 16:
+            value = 'f32_to_f16(%s)' % (value)
+        elif dst_channel.size == 64:
+            value = '(double)%s' % (value)
+        return value
     
     if clamp:
         value = clamp_expr(src_channel, dst_channel, dst_native_type, value)
@@ -301,33 +332,29 @@ def conversion_expr(src_channel, dst_channel, dst_native_type, value, clamp=True
     assert False
 
 
-def generate_format_unpack(format, dst_channel, dst_native_type, dst_suffix):
-    '''Generate the function to unpack pixels from a particular format'''
+def generate_unpack_kernel(format, dst_channel, dst_native_type):
 
+    if not is_format_supported(format):
+        return
+    
     assert format.layout == PLAIN
 
-    name = format.short_name()
-
     src_native_type = native_type(format)
 
-    print 'static INLINE void'
-    print 'util_format_%s_unpack_%s(%s *dst, const void *src)' % (name, dst_suffix, dst_native_type)
-    print '{'
-    
     if format.is_bitmask():
         depth = format.block_size()
-        print '   uint%u_t value = *(uint%u_t *)src;' % (depth, depth) 
+        print '         uint%u_t value = *(uint%u_t *)src;' % (depth, depth) 
 
         # Declare the intermediate variables
         for i in range(format.nr_channels()):
             src_channel = format.channels[i]
             if src_channel.type == UNSIGNED:
-                print '   uint%u_t %s;' % (depth, src_channel.name)
+                print '         uint%u_t %s;' % (depth, src_channel.name)
             elif src_channel.type == SIGNED:
-                print '   int%u_t %s;' % (depth, src_channel.name)
+                print '         int%u_t %s;' % (depth, src_channel.name)
 
         print '#ifdef PIPE_ARCH_BIG_ENDIAN'
-        print '   value = util_bswap%u(value);' % depth
+        print '         value = util_bswap%u(value);' % depth
         print '#endif'
 
         # Compute the intermediate unshifted values 
@@ -355,7 +382,7 @@ def generate_format_unpack(format, dst_channel, dst_native_type, dst_suffix):
                 value = None
                 
             if value is not None:
-                print '   %s = %s;' % (src_channel.name, value)
+                print '         %s = %s;' % (src_channel.name, value)
                 
             shift += src_channel.size
 
@@ -379,11 +406,11 @@ def generate_format_unpack(format, dst_channel, dst_native_type, dst_suffix):
                     value = get_one(dst_channel)
                 elif i >= 1:
                     value = 'dst[0]'
-            print '   dst[%u] = %s; /* %s */' % (i, value, 'rgba'[i])
+            print '         dst[%u] = %s; /* %s */' % (i, value, 'rgba'[i])
         
     else:
-        print '   union util_format_%s pixel;' % format.short_name()
-        print '   memcpy(&pixel, src, sizeof pixel);'
+        print '         union util_format_%s pixel;' % format.short_name()
+        print '         memcpy(&pixel, src, sizeof pixel);'
         bswap_format(format)
     
         for i in range(4):
@@ -405,36 +432,29 @@ def generate_format_unpack(format, dst_channel, dst_native_type, dst_suffix):
                     value = get_one(dst_channel)
                 elif i >= 1:
                     value = 'dst[0]'
-            print '   dst[%u] = %s; /* %s */' % (i, value, 'rgba'[i])
-
-    print '}'
-    print
+            print '         dst[%u] = %s; /* %s */' % (i, value, 'rgba'[i])
     
 
-def generate_format_pack(format, src_channel, src_native_type, src_suffix):
-    '''Generate the function to pack pixels to a particular format'''
-
-    name = format.short_name()
+def generate_pack_kernel(format, src_channel, src_native_type):
 
+    if not is_format_supported(format):
+        return
+    
     dst_native_type = native_type(format)
 
     assert format.layout == PLAIN
 
     inv_swizzle = format.inv_swizzles()
-    
-    print 'static INLINE void'
-    print 'util_format_%s_pack_%s(void *dst, %s r, %s g, %s b, %s a)' % (name, src_suffix, src_native_type, src_native_type, src_native_type, src_native_type)
-    print '{'
-    
+
     if format.is_bitmask():
         depth = format.block_size()
-        print '   uint%u_t value = 0;' % depth 
+        print '         uint%u_t value = 0;' % depth 
 
         shift = 0
         for i in range(4):
             dst_channel = format.channels[i]
             if inv_swizzle[i] is not None:
-                value = 'rgba'[inv_swizzle[i]]
+                value ='src[%u]' % inv_swizzle[i]
                 value = conversion_expr(src_channel, dst_channel, dst_native_type, value)
                 if format.colorspace == ZS:
                     if i == 3:
@@ -452,138 +472,141 @@ def generate_format_pack(format, src_channel, src_native_type, src_suffix):
                 else:
                     value = None
                 if value is not None:
-                    print '   value |= %s;' % (value)
+                    print '         value |= %s;' % (value)
                 
             shift += dst_channel.size
 
         print '#ifdef PIPE_ARCH_BIG_ENDIAN'
-        print '   value = util_bswap%u(value);' % depth
+        print '         value = util_bswap%u(value);' % depth
         print '#endif'
         
-        print '   *(uint%u_t *)dst = value;' % depth 
+        print '         *(uint%u_t *)dst = value;' % depth 
 
     else:
-        print '   union util_format_%s pixel;' % format.short_name()
+        print '         union util_format_%s pixel;' % format.short_name()
     
         for i in range(4):
             dst_channel = format.channels[i]
             width = dst_channel.size
             if inv_swizzle[i] is None:
                 continue
-            value = 'rgba'[inv_swizzle[i]]
+            value ='src[%u]' % inv_swizzle[i]
             value = conversion_expr(src_channel, dst_channel, dst_native_type, value)
             if format.colorspace == ZS:
                 if i == 3:
                     value = get_one(dst_channel)
                 elif i >= 1:
                     value = '0'
-            print '   pixel.chan.%s = %s;' % (dst_channel.name, value)
+            print '         pixel.chan.%s = %s;' % (dst_channel.name, value)
     
         bswap_format(format)
-        print '   memcpy(dst, &pixel, sizeof pixel);'
-        
-    print '}'
-    print
+        print '         memcpy(dst, &pixel, sizeof pixel);'
     
 
-def generate_unpack(formats, dst_channel, dst_native_type, dst_suffix):
-    '''Generate the dispatch function to unpack pixels from any format'''
+def generate_format_unpack(format, dst_channel, dst_native_type, dst_suffix):
+    '''Generate the function to unpack pixels from a particular format'''
 
-    for format in formats:
-        if is_format_supported(format):
-            generate_format_unpack(format, dst_channel, dst_native_type, dst_suffix)
+    name = format.short_name()
 
     print 'static INLINE void'
-    print 'util_format_unpack_%s(enum pipe_format format, %s *dst, const void *src)' % (dst_suffix, dst_native_type)
+    print 'util_format_%s_unpack_%s(%s *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)' % (name, dst_suffix, dst_native_type)
     print '{'
-    print '   void (*func)(%s *dst, const void *src);' % dst_native_type
-    print '   switch(format) {'
-    for format in formats:
-        if is_format_supported(format):
-            print '   case %s:' % format.name
-            print '      func = &util_format_%s_unpack_%s;' % (format.short_name(), dst_suffix)
-            print '      break;'
-    print '   default:'
-    print '      debug_printf("unsupported format\\n");'
-    print '      return;'
-    print '   }'
-    print '   func(dst, src);'
+
+    if is_format_supported(format):
+        print '   unsigned x, y;'
+        print '   for(y = 0; y < height; y += %u) {' % (format.block_height,)
+        print '      %s *dst = dst_row;' % (dst_native_type)
+        print '      const uint8_t *src = src_row;'
+        print '      for(x = 0; x < width; x += %u) {' % (format.block_width,)
+        
+        generate_unpack_kernel(format, dst_channel, dst_native_type)
+    
+        print '         src += %u;' % (format.block_size() / 8,)
+        print '         dst += 4;'
+        print '      }'
+        print '      src_row += src_stride;'
+        print '      dst_row += dst_stride/sizeof(*dst_row);'
+        print '   }'
+
     print '}'
     print
+    
 
+def generate_format_pack(format, src_channel, src_native_type, src_suffix):
+    '''Generate the function to pack pixels to a particular format'''
 
-def generate_pack(formats, src_channel, src_native_type, src_suffix):
-    '''Generate the dispatch function to pack pixels to any format'''
-
-    for format in formats:
-        if is_format_supported(format):
-            generate_format_pack(format, src_channel, src_native_type, src_suffix)
+    name = format.short_name()
 
     print 'static INLINE void'
-    print 'util_format_pack_%s(enum pipe_format format, void *dst, %s r, %s g, %s b, %s a)' % (src_suffix, src_native_type, src_native_type, src_native_type, src_native_type)
+    print 'util_format_%s_pack_%s(uint8_t *dst_row, unsigned dst_stride, const %s *src_row, unsigned src_stride, unsigned width, unsigned height)' % (name, src_suffix, src_native_type)
     print '{'
-    print '   void (*func)(void *dst, %s r, %s g, %s b, %s a);' % (src_native_type, src_native_type, src_native_type, src_native_type)
-    print '   switch(format) {'
-    for format in formats:
-        if is_format_supported(format):
-            print '   case %s:' % format.name
-            print '      func = &util_format_%s_pack_%s;' % (format.short_name(), src_suffix)
-            print '      break;'
-    print '   default:'
-    print '      debug_printf("%s: unsupported format\\n", __FUNCTION__);'
-    print '      return;'
-    print '   }'
-    print '   func(dst, r, g, b, a);'
+    
+    if is_format_supported(format):
+        print '   unsigned x, y;'
+        print '   for(y = 0; y < height; y += %u) {' % (format.block_height,)
+        print '      const %s *src = src_row;' % (src_native_type)
+        print '      uint8_t *dst = dst_row;'
+        print '      for(x = 0; x < width; x += %u) {' % (format.block_width,)
+    
+        generate_pack_kernel(format, src_channel, src_native_type)
+            
+        print '         src += 4;'
+        print '         dst += %u;' % (format.block_size() / 8,)
+        print '      }'
+        print '      dst_row += dst_stride;'
+        print '      src_row += src_stride/sizeof(*src_row);'
+        print '   }'
+        
     print '}'
     print
+    
 
+def generate_format_fetch(format, dst_channel, dst_native_type, dst_suffix):
+    '''Generate the function to unpack pixels from a particular format'''
 
-def main():
-    formats = []
-    for arg in sys.argv[1:]:
-        formats.extend(parse(arg))
+    name = format.short_name()
 
-    print '/* This file is autogenerated by u_format_pack.py from u_format.csv. Do not edit directly. */'
-    print
-    # This will print the copyright message on the top of this file
-    print __doc__.strip()
+    print 'static INLINE void'
+    print 'util_format_%s_fetch_%s(%s *dst, const uint8_t *src, unsigned i, unsigned j)' % (name, dst_suffix, dst_native_type)
+    print '{'
+
+    if is_format_supported(format):
+        generate_unpack_kernel(format, dst_channel, dst_native_type)
 
+    print '}'
     print
-    print '#ifndef U_FORMAT_PACK_H'
-    print '#define U_FORMAT_PACK_H'
+
+
+def generate(formats):
     print
     print '#include "pipe/p_compiler.h"'
     print '#include "u_math.h"'
     print '#include "u_format.h"'
     print
 
-    generate_clamp()
+    generate_f16_to_f32()
+    generate_f32_to_f16()
 
     for format in formats:
-        if format.layout == PLAIN:
+        if is_format_supported(format):
             generate_format_type(format)
 
     channel = Channel(FLOAT, False, 32)
     native_type = 'float'
-    suffix = '4f'
+    suffix = 'float'
 
-    generate_unpack(formats, channel, native_type, suffix)
-    generate_pack(formats, channel, native_type, suffix)
+    for format in formats:
+        if format.layout != 's3tc':
+            generate_format_unpack(format, channel, native_type, suffix)
+            generate_format_pack(format, channel, native_type, suffix)
+            generate_format_fetch(format, channel, native_type, suffix)
 
     channel = Channel(UNSIGNED, True, 8)
     native_type = 'uint8_t'
-    suffix = '4ub'
-
-    generate_unpack(formats, channel, native_type, suffix)
-    generate_pack(formats, channel, native_type, suffix)
-
-    print
-    print '#ifdef __cplusplus'
-    print '}'
-    print '#endif'
-    print
-    print '#endif /* ! U_FORMAT_PACK_H */'
+    suffix = '8unorm'
 
+    for format in formats:
+        if format.layout != 's3tc':
+            generate_format_unpack(format, channel, native_type, suffix)
+            generate_format_pack(format, channel, native_type, suffix)
 
-if __name__ == '__main__':
-    main()
diff --git a/src/gallium/auxiliary/util/u_format_parse.py b/src/gallium/auxiliary/util/u_format_parse.py
index f74dc5e88a..03dc62d329 100755
--- a/src/gallium/auxiliary/util/u_format_parse.py
+++ b/src/gallium/auxiliary/util/u_format_parse.py
@@ -142,6 +142,8 @@ class Format:
 
     def is_mixed(self):
         ref_channel = self.channels[0]
+        if ref_channel.type == VOID:
+           ref_channel = self.channels[1]
         for channel in self.channels[1:]:
             if channel.type != VOID:
                 if channel.type != ref_channel.type:
diff --git a/src/gallium/auxiliary/util/u_format_s3tc.c b/src/gallium/auxiliary/util/u_format_s3tc.c
new file mode 100644
index 0000000000..4fe00b0e43
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format_s3tc.c
@@ -0,0 +1,510 @@
+/**************************************************************************
+ *
+ * Copyright (C) 1999-2007  Brian Paul   All Rights Reserved.
+ * Copyright (c) 2008 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
+ * BRIAN PAUL BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN
+ * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+#include "u_dl.h"
+#include "u_math.h"
+#include "u_format.h"
+#include "u_format_s3tc.h"
+
+
+#if defined(_WIN32) || defined(WIN32)
+#define DXTN_LIBNAME "dxtn.dll"
+#else
+#define DXTN_LIBNAME "libtxc_dxtn.so"
+#endif
+
+
+boolean util_format_s3tc_enabled = FALSE;
+
+util_format_dxtn_fetch_t util_format_dxt1_rgb_fetch = NULL;
+util_format_dxtn_fetch_t util_format_dxt1_rgba_fetch = NULL;
+util_format_dxtn_fetch_t util_format_dxt3_rgba_fetch = NULL;
+util_format_dxtn_fetch_t util_format_dxt5_rgba_fetch = NULL;
+
+util_format_dxtn_pack_t util_format_dxtn_pack = NULL;
+
+
+void
+util_format_s3tc_init(void)
+{
+   static struct util_dl_library *
+   library = NULL;
+
+   if (util_format_s3tc_enabled)
+      return;
+
+   if (!library) {
+      library = util_dl_open(DXTN_LIBNAME);
+      if (!library) {
+	 debug_printf("couldn't open " DXTN_LIBNAME ", software DXTn "
+	    "compression/decompression unavailable");
+      }
+      else {
+         util_format_dxt1_rgb_fetch = (util_format_dxtn_fetch_t)
+            util_dl_get_proc_address(library, "fetch_2d_texel_rgb_dxt1");
+         util_format_dxt1_rgba_fetch = (util_format_dxtn_fetch_t)
+            util_dl_get_proc_address(library, "fetch_2d_texel_rgba_dxt1");
+         util_format_dxt3_rgba_fetch = (util_format_dxtn_fetch_t)
+            util_dl_get_proc_address(library, "fetch_2d_texel_rgba_dxt3");
+         util_format_dxt5_rgba_fetch = (util_format_dxtn_fetch_t)
+            util_dl_get_proc_address(library, "fetch_2d_texel_rgba_dxt5");
+         util_format_dxtn_pack = (util_format_dxtn_pack_t)
+            util_dl_get_proc_address(library, "tx_compress_dxtn");
+
+         if (util_format_dxt1_rgb_fetch ||
+             util_format_dxt1_rgba_fetch ||
+             util_format_dxt3_rgba_fetch ||
+             util_format_dxt5_rgba_fetch ||
+             util_format_dxtn_pack) {
+            util_format_s3tc_enabled = TRUE;
+            debug_printf("software DXTn compression/decompression available");
+
+         } else {
+	    debug_printf("couldn't reference all symbols in "
+	       DXTN_LIBNAME ", software DXTn compression/decompression "
+	       "unavailable");
+         }
+      }
+   }
+}
+
+
+/*
+ * Pixel fetch.
+ */
+
+void
+util_format_dxt1_rgb_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   if (util_format_dxt1_rgb_fetch) {
+      util_format_dxt1_rgb_fetch(0, src, i, j, dst);
+   }
+}
+
+void
+util_format_dxt1_rgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   if (util_format_dxt1_rgba_fetch) {
+      util_format_dxt1_rgba_fetch(0, src, i, j, dst);
+   }
+}
+
+void
+util_format_dxt3_rgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   if (util_format_dxt3_rgba_fetch) {
+      util_format_dxt3_rgba_fetch(0, src, i, j, dst);
+   }
+}
+
+void
+util_format_dxt5_rgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   if (util_format_dxt5_rgba_fetch) {
+      util_format_dxt5_rgba_fetch(0, src, i, j, dst);
+   }
+}
+
+void
+util_format_dxt1_rgb_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   if (util_format_dxt1_rgb_fetch) {
+      uint8_t tmp[4];
+      util_format_dxt1_rgb_fetch(0, src, i, j, tmp);
+      dst[0] = ubyte_to_float(tmp[0]);
+      dst[1] = ubyte_to_float(tmp[1]);
+      dst[2] = ubyte_to_float(tmp[2]);
+      dst[3] = 1.0;
+   }
+}
+
+void
+util_format_dxt1_rgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   if (util_format_dxt1_rgba_fetch) {
+      uint8_t tmp[4];
+      util_format_dxt1_rgba_fetch(0, src, i, j, tmp);
+      dst[0] = ubyte_to_float(tmp[0]);
+      dst[1] = ubyte_to_float(tmp[1]);
+      dst[2] = ubyte_to_float(tmp[2]);
+      dst[3] = ubyte_to_float(tmp[3]);
+   }
+}
+
+void
+util_format_dxt3_rgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   if (util_format_dxt3_rgba_fetch) {
+      uint8_t tmp[4];
+      util_format_dxt3_rgba_fetch(0, src, i, j, tmp);
+      dst[0] = ubyte_to_float(tmp[0]);
+      dst[1] = ubyte_to_float(tmp[1]);
+      dst[2] = ubyte_to_float(tmp[2]);
+      dst[3] = ubyte_to_float(tmp[3]);
+   }
+}
+
+void
+util_format_dxt5_rgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   if (util_format_dxt5_rgba_fetch) {
+      uint8_t tmp[4];
+      util_format_dxt5_rgba_fetch(0, src, i, j, tmp);
+      dst[0] = ubyte_to_float(tmp[0]);
+      dst[1] = ubyte_to_float(tmp[1]);
+      dst[2] = ubyte_to_float(tmp[2]);
+      dst[3] = ubyte_to_float(tmp[3]);
+   }
+}
+
+
+/*
+ * Block decompression.
+ *
+ * FIXME
+ */
+
+void
+util_format_dxt1_rgb_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt1_rgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+
+}
+
+void
+util_format_dxt3_rgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt5_rgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt1_rgb_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   if (util_format_dxt1_rgb_fetch) {
+      unsigned x, y, i, j;
+      for(y = 0; y < height; y += 4) {
+         const uint8_t *src = src_row;
+         for(x = 0; x < width; x += 4) {
+            for(j = 0; j < 4; ++j) {
+               for(i = 0; i < 4; ++i) {
+                  float *dst = dst_row + (y + j)*dst_stride/sizeof(*dst_row) + (x + i)*4;
+                  uint8_t tmp[4];
+                  util_format_dxt1_rgb_fetch(0, src, i, j, tmp);
+                  dst[0] = ubyte_to_float(tmp[0]);
+                  dst[1] = ubyte_to_float(tmp[1]);
+                  dst[2] = ubyte_to_float(tmp[2]);
+                  dst[3] = 1.0;
+               }
+            }
+            src += 8;
+         }
+         src_row += src_stride;
+      }
+   }
+}
+
+void
+util_format_dxt1_rgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   if (util_format_dxt1_rgba_fetch) {
+      unsigned x, y, i, j;
+      for(y = 0; y < height; y += 4) {
+         const uint8_t *src = src_row;
+         for(x = 0; x < width; x += 4) {
+            for(j = 0; j < 4; ++j) {
+               for(i = 0; i < 4; ++i) {
+                  float *dst = dst_row + (y + j)*dst_stride/sizeof(*dst_row) + (x + i)*4;
+                  uint8_t tmp[4];
+                  util_format_dxt1_rgba_fetch(0, src, i, j, tmp);
+                  dst[0] = ubyte_to_float(tmp[0]);
+                  dst[1] = ubyte_to_float(tmp[1]);
+                  dst[2] = ubyte_to_float(tmp[2]);
+                  dst[3] = ubyte_to_float(tmp[3]);
+               }
+            }
+            src += 8;
+         }
+         src_row += src_stride;
+      }
+   }
+}
+
+void
+util_format_dxt3_rgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   if (util_format_dxt3_rgba_fetch) {
+      unsigned x, y, i, j;
+      for(y = 0; y < height; y += 4) {
+         const uint8_t *src = src_row;
+         for(x = 0; x < width; x += 4) {
+            for(j = 0; j < 4; ++j) {
+               for(i = 0; i < 4; ++i) {
+                  float *dst = dst_row + (y + j)*dst_stride/sizeof(*dst_row) + (x + i)*4;
+                  uint8_t tmp[4];
+                  util_format_dxt3_rgba_fetch(0, src, i, j, tmp);
+                  dst[0] = ubyte_to_float(tmp[0]);
+                  dst[1] = ubyte_to_float(tmp[1]);
+                  dst[2] = ubyte_to_float(tmp[2]);
+                  dst[3] = ubyte_to_float(tmp[3]);
+               }
+            }
+            src += 16;
+         }
+         src_row += src_stride;
+      }
+   }
+}
+
+void
+util_format_dxt5_rgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   if (util_format_dxt5_rgba_fetch) {
+      unsigned x, y, i, j;
+      for(y = 0; y < height; y += 4) {
+         const uint8_t *src = src_row;
+         for(x = 0; x < width; x += 4) {
+            for(j = 0; j < 4; ++j) {
+               for(i = 0; i < 4; ++i) {
+                  float *dst = dst_row + (y + j)*dst_stride/sizeof(*dst_row) + (x + i)*4;
+                  uint8_t tmp[4];
+                  util_format_dxt5_rgba_fetch(0, src, i, j, tmp);
+                  dst[0] = ubyte_to_float(tmp[0]);
+                  dst[1] = ubyte_to_float(tmp[1]);
+                  dst[2] = ubyte_to_float(tmp[2]);
+                  dst[3] = ubyte_to_float(tmp[3]);
+               }
+            }
+            src += 16;
+         }
+         src_row += src_stride;
+      }
+   }
+}
+
+
+/*
+ * Block compression.
+ *
+ * FIXME
+ */
+
+void
+util_format_dxt1_rgb_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt1_rgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt3_rgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt5_rgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt1_rgb_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt1_rgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt3_rgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+void
+util_format_dxt5_rgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+}
+
+
+/*
+ * SRGB variants.
+ *
+ * FIXME: shunts to RGB for now
+ */
+
+void
+util_format_dxt1_srgb_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt1_rgb_unpack_8unorm(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt1_srgb_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt1_rgb_pack_8unorm(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt1_srgb_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   util_format_dxt1_rgb_fetch_8unorm(dst, src, i, j);
+}
+
+void
+util_format_dxt1_srgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt1_rgba_unpack_8unorm(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt1_srgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt1_rgba_pack_8unorm(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt1_srgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   util_format_dxt1_rgba_fetch_8unorm(dst, src, i, j);
+}
+
+void
+util_format_dxt3_srgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt3_rgba_unpack_8unorm(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt3_srgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt3_rgba_pack_8unorm(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt3_srgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   util_format_dxt3_rgba_fetch_8unorm(dst, src, i, j);
+}
+
+void
+util_format_dxt5_srgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt5_rgba_unpack_8unorm(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt5_srgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt5_rgba_pack_8unorm(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt5_srgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   util_format_dxt5_rgba_fetch_8unorm(dst, src, i, j);
+}
+
+void
+util_format_dxt1_srgb_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt1_rgb_unpack_float(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt1_srgb_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt1_rgb_pack_float(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt1_srgb_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   util_format_dxt1_rgb_fetch_float(dst, src, i, j);
+}
+
+void
+util_format_dxt1_srgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt1_rgba_unpack_float(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt1_srgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt1_rgba_pack_float(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt1_srgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   util_format_dxt1_rgba_fetch_float(dst, src, i, j);
+}
+
+void
+util_format_dxt3_srgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt3_rgba_unpack_float(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt3_srgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt3_rgba_pack_float(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt3_srgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   util_format_dxt3_rgba_fetch_float(dst, src, i, j);
+}
+
+void
+util_format_dxt5_srgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt5_rgba_unpack_float(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt5_srgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height)
+{
+   util_format_dxt5_rgba_pack_float(dst_row, dst_stride, src_row, src_stride, width, height);
+}
+
+void
+util_format_dxt5_srgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j)
+{
+   util_format_dxt5_rgba_fetch_float(dst, src, i, j);
+}
+
diff --git a/src/gallium/auxiliary/util/u_format_s3tc.h b/src/gallium/auxiliary/util/u_format_s3tc.h
new file mode 100644
index 0000000000..5c3a9dd22c
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_format_s3tc.h
@@ -0,0 +1,218 @@
+/**************************************************************************
+ *
+ * Copyright 2010 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, 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 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 NON-INFRINGEMENT. IN NO EVENT SHALL
+ * THE COPYRIGHT HOLDERS, AUTHORS AND/OR ITS SUPPLIERS 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.
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ **************************************************************************/
+
+
+#ifndef U_FORMAT_S3TC_H_
+#define U_FORMAT_S3TC_H_
+
+
+#include "pipe/p_compiler.h"
+
+
+enum util_format_dxtn {
+  UTIL_FORMAT_DXT1_RGB = 0x83F0,
+  UTIL_FORMAT_DXT1_RGBA = 0x83F1,
+  UTIL_FORMAT_DXT3_RGBA = 0x83F2,
+  UTIL_FORMAT_DXT5_RGBA = 0x83F3
+};
+
+
+typedef void
+(*util_format_dxtn_fetch_t)( int src_stride,
+                             const uint8_t *src,
+                             int col, int row,
+                             uint8_t *dst );
+
+typedef void
+(*util_format_dxtn_pack_t)( int src_comps,
+                            int width, int height,
+                            const uint8_t *src,
+                            enum util_format_dxtn dst_format,
+                            uint8_t *dst,
+                            int dst_stride);
+
+extern boolean util_format_s3tc_enabled;
+
+extern util_format_dxtn_fetch_t util_format_dxt1_rgb_fetch;
+extern util_format_dxtn_fetch_t util_format_dxt1_rgba_fetch;
+extern util_format_dxtn_fetch_t util_format_dxt3_rgba_fetch;
+extern util_format_dxtn_fetch_t util_format_dxt5_rgba_fetch;
+
+extern util_format_dxtn_pack_t util_format_dxtn_pack;
+
+
+void
+util_format_s3tc_init(void);
+
+
+void
+util_format_dxt1_rgb_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_rgb_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_rgb_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt1_rgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_rgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_rgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt3_rgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt3_rgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt3_rgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt5_rgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt5_rgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt5_rgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt1_srgb_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_srgb_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_srgb_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt1_srgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_srgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_srgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt3_srgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt3_srgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt3_srgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt5_srgba_unpack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt5_srgba_pack_8unorm(uint8_t *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt5_srgba_fetch_8unorm(uint8_t *dst, const uint8_t *src, unsigned i, unsigned j);
+
+
+void
+util_format_dxt1_rgb_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_rgb_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_rgb_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt1_rgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_rgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_rgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt3_rgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt3_rgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt3_rgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt5_rgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt5_rgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt5_rgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt1_srgb_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_srgb_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_srgb_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt1_srgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_srgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt1_srgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt3_srgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt3_srgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt3_srgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j);
+
+void
+util_format_dxt5_srgba_unpack_float(float *dst_row, unsigned dst_stride, const uint8_t *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt5_srgba_pack_float(uint8_t *dst_row, unsigned dst_stride, const float *src_row, unsigned src_stride, unsigned width, unsigned height);
+
+void
+util_format_dxt5_srgba_fetch_float(float *dst, const uint8_t *src, unsigned i, unsigned j);
+
+
+#endif /* U_FORMAT_S3TC_H_ */
diff --git a/src/gallium/auxiliary/util/u_format_table.py b/src/gallium/auxiliary/util/u_format_table.py
index fb68852a53..94a4331b15 100755
--- a/src/gallium/auxiliary/util/u_format_table.py
+++ b/src/gallium/auxiliary/util/u_format_table.py
@@ -3,7 +3,7 @@
 '''
 /**************************************************************************
  *
- * Copyright 2009 VMware, Inc.
+ * Copyright 2010 VMware, Inc.
  * All Rights Reserved.
  *
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -33,6 +33,7 @@
 import sys
 
 from u_format_parse import *
+import u_format_pack
 
 
 def layout_map(layout):
@@ -85,23 +86,11 @@ def write_format_table(formats):
     print __doc__.strip()
     print
     print '#include "u_format.h"'
+    print '#include "u_format_s3tc.h"'
     print
-    print 'const struct util_format_description'
-    print 'util_format_none_description = {'
-    print "   PIPE_FORMAT_NONE,"
-    print "   \"PIPE_FORMAT_NONE\","
-    print "   \"none\","
-    print "   {0, 0, 0},"
-    print "   0,"
-    print "   0,"
-    print "   0,"
-    print "   0,"
-    print "   0,"
-    print "   {{0, 0, 0}, {0, 0, 0}, {0, 0, 0}, {0, 0, 0}},"
-    print "   {0, 0, 0, 0},"
-    print "   0"
-    print "};"
-    print
+    
+    u_format_pack.generate(formats)
+    
     for format in formats:
         print 'const struct util_format_description'
         print 'util_format_%s_description = {' % (format.short_name(),)
@@ -140,8 +129,14 @@ def write_format_table(formats):
             print "      %s%s\t/* %s */" % (swizzle_map[swizzle], sep, comment)
         print "   },"
         print "   %s," % (colorspace_map(format.colorspace),)
+        print "   &util_format_%s_unpack_8unorm," % format.short_name() 
+        print "   &util_format_%s_pack_8unorm," % format.short_name() 
+        print "   &util_format_%s_unpack_float," % format.short_name() 
+        print "   &util_format_%s_pack_float," % format.short_name() 
+        print "   &util_format_%s_fetch_float" % format.short_name() 
         print "};"
         print
+        
     print "const struct util_format_description *"
     print "util_format_description(enum pipe_format format)"
     print "{"
@@ -150,8 +145,6 @@ def write_format_table(formats):
     print "   }"
     print
     print "   switch (format) {"
-    print "   case PIPE_FORMAT_NONE:"
-    print "      return &util_format_none_description;"
     for format in formats:
         print "   case %s:" % format.name
         print "      return &util_format_%s_description;" % (format.short_name(),)
diff --git a/src/gallium/auxiliary/util/u_format_tests.c b/src/gallium/auxiliary/util/u_format_tests.c
index 9d6debcd8c..4c4cc067a5 100644
--- a/src/gallium/auxiliary/util/u_format_tests.c
+++ b/src/gallium/auxiliary/util/u_format_tests.c
@@ -120,12 +120,11 @@ util_format_test_cases[] =
     * 16-bit rendertarget formats
     */
 
-   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x0000), {0.0, 0.0, 0.0, 0.0}},
-   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x001f), {0.0, 0.0, 1.0, 0.0}},
-   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x03e0), {0.0, 1.0, 0.0, 0.0}},
-   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x7c00), {1.0, 0.0, 0.0, 0.0}},
-   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x8000), {0.0, 0.0, 0.0, 1.0}},
-   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0xffff), {1.0, 1.0, 1.0, 1.0}},
+   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x0000), {0.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x001f), {0.0, 0.0, 1.0, 1.0}},
+   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x03e0), {0.0, 1.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x7c00), {1.0, 0.0, 0.0, 1.0}},
+   {PIPE_FORMAT_B5G5R5X1_UNORM, PACKED_1x16(0x7fff), PACKED_1x16(0x7fff), {1.0, 1.0, 1.0, 1.0}},
 
    {PIPE_FORMAT_B5G5R5A1_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x0000), {0.0, 0.0, 0.0, 0.0}},
    {PIPE_FORMAT_B5G5R5A1_UNORM, PACKED_1x16(0xffff), PACKED_1x16(0x001f), {0.0, 0.0, 1.0, 0.0}},
diff --git a/src/gallium/auxiliary/util/u_memory.h b/src/gallium/auxiliary/util/u_memory.h
index a2fc597356..53d56599fe 100644
--- a/src/gallium/auxiliary/util/u_memory.h
+++ b/src/gallium/auxiliary/util/u_memory.h
@@ -88,7 +88,7 @@ mem_dup(const void *src, uint size)
 /**
  * Offset of a field in a struct, in bytes.
  */
-#define Offset(TYPE, MEMBER) ((unsigned)&(((TYPE *)NULL)->MEMBER))
+#define Offset(TYPE, MEMBER) ((uintptr_t)&(((TYPE *)NULL)->MEMBER))
 
 
 
diff --git a/src/gallium/auxiliary/util/u_pack_color.h b/src/gallium/auxiliary/util/u_pack_color.h
index c5fd7a6783..ad4a06dc77 100644
--- a/src/gallium/auxiliary/util/u_pack_color.h
+++ b/src/gallium/auxiliary/util/u_pack_color.h
@@ -446,17 +446,17 @@ util_pack_z(enum pipe_format format, double z)
       return (uint) (z * 0xffffffff);
    case PIPE_FORMAT_Z32_FLOAT:
       return (uint)z;
-   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
    case PIPE_FORMAT_Z24X8_UNORM:
       if (z == 1.0)
          return 0xffffff;
       return (uint) (z * 0xffffff);
-   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
    case PIPE_FORMAT_X8Z24_UNORM:
       if (z == 1.0)
          return 0xffffff00;
       return ((uint) (z * 0xffffff)) << 8;
-   case PIPE_FORMAT_S8_UNORM:
+   case PIPE_FORMAT_S8_USCALED:
       /* this case can get it via util_pack_z_stencil() */
       return 0;
    default:
@@ -477,13 +477,13 @@ util_pack_z_stencil(enum pipe_format format, double z, uint s)
    unsigned packed = util_pack_z(format, z);
 
    switch (format) {
-   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
       packed |= s << 24;
       break;
-   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
       packed |= s;
       break;
-   case PIPE_FORMAT_S8_UNORM:
+   case PIPE_FORMAT_S8_USCALED:
       packed |= s;
       break;
    default:
diff --git a/src/gallium/auxiliary/util/u_tile.c b/src/gallium/auxiliary/util/u_tile.c
index 09b2382733..43b45eb926 100644
--- a/src/gallium/auxiliary/util/u_tile.c
+++ b/src/gallium/auxiliary/util/u_tile.c
@@ -108,436 +108,6 @@ pipe_put_tile_raw(struct pipe_context *pipe,
 
 
 
-/*** PIPE_FORMAT_B8G8R8A8_UNORM ***/
-
-static void
-a8r8g8b8_get_tile_rgba(const unsigned *src,
-                       unsigned w, unsigned h,
-                       float *p,
-                       unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         const unsigned pixel = *src++;
-         pRow[0] = ubyte_to_float((pixel >> 16) & 0xff);
-         pRow[1] = ubyte_to_float((pixel >>  8) & 0xff);
-         pRow[2] = ubyte_to_float((pixel >>  0) & 0xff);
-         pRow[3] = ubyte_to_float((pixel >> 24) & 0xff);
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-a8r8g8b8_put_tile_rgba(unsigned *dst,
-                       unsigned w, unsigned h,
-                       const float *p,
-                       unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r, g, b, a;
-         r = float_to_ubyte(pRow[0]);
-         g = float_to_ubyte(pRow[1]);
-         b = float_to_ubyte(pRow[2]);
-         a = float_to_ubyte(pRow[3]);
-         *dst++ = (a << 24) | (r << 16) | (g << 8) | b;
-      }
-      p += src_stride;
-   }
-}
-
-
-/*** PIPE_FORMAT_B8G8R8X8_UNORM ***/
-
-static void
-x8r8g8b8_get_tile_rgba(const unsigned *src,
-                       unsigned w, unsigned h,
-                       float *p,
-                       unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         const unsigned pixel = *src++;
-         pRow[0] = ubyte_to_float((pixel >> 16) & 0xff);
-         pRow[1] = ubyte_to_float((pixel >>  8) & 0xff);
-         pRow[2] = ubyte_to_float((pixel >>  0) & 0xff);
-         pRow[3] = 1.0F;
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-x8r8g8b8_put_tile_rgba(unsigned *dst,
-                       unsigned w, unsigned h,
-                       const float *p,
-                       unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r, g, b;
-         r = float_to_ubyte(pRow[0]);
-         g = float_to_ubyte(pRow[1]);
-         b = float_to_ubyte(pRow[2]);
-         *dst++ = (0xff << 24) | (r << 16) | (g << 8) | b;
-      }
-      p += src_stride;
-   }
-}
-
-
-/*** PIPE_FORMAT_A8R8G8B8_UNORM ***/
-
-static void
-b8g8r8a8_get_tile_rgba(const unsigned *src,
-                       unsigned w, unsigned h,
-                       float *p,
-                       unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         const unsigned pixel = *src++;
-         pRow[0] = ubyte_to_float((pixel >>  8) & 0xff);
-         pRow[1] = ubyte_to_float((pixel >> 16) & 0xff);
-         pRow[2] = ubyte_to_float((pixel >> 24) & 0xff);
-         pRow[3] = ubyte_to_float((pixel >>  0) & 0xff);
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-b8g8r8a8_put_tile_rgba(unsigned *dst,
-                       unsigned w, unsigned h,
-                       const float *p,
-                       unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r, g, b, a;
-         r = float_to_ubyte(pRow[0]);
-         g = float_to_ubyte(pRow[1]);
-         b = float_to_ubyte(pRow[2]);
-         a = float_to_ubyte(pRow[3]);
-         *dst++ = (b << 24) | (g << 16) | (r << 8) | a;
-      }
-      p += src_stride;
-   }
-}
-
-
-/*** PIPE_FORMAT_A8B8G8R8_UNORM ***/
-
-static void
-r8g8b8a8_get_tile_rgba(const unsigned *src,
-                       unsigned w, unsigned h,
-                       float *p,
-                       unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         const unsigned pixel = *src++;
-         pRow[0] = ubyte_to_float((pixel >> 24) & 0xff);
-         pRow[1] = ubyte_to_float((pixel >> 16) & 0xff);
-         pRow[2] = ubyte_to_float((pixel >>  8) & 0xff);
-         pRow[3] = ubyte_to_float((pixel >>  0) & 0xff);
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-r8g8b8a8_put_tile_rgba(unsigned *dst,
-                       unsigned w, unsigned h,
-                       const float *p,
-                       unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r, g, b, a;
-         r = float_to_ubyte(pRow[0]);
-         g = float_to_ubyte(pRow[1]);
-         b = float_to_ubyte(pRow[2]);
-         a = float_to_ubyte(pRow[3]);
-         *dst++ = (r << 24) | (g << 16) | (b << 8) | a;
-      }
-      p += src_stride;
-   }
-}
-
-
-/*** PIPE_FORMAT_B5G5R5X1_UNORM ***/
-
-static void
-x1r5g5b5_get_tile_rgba(const ushort *src,
-                       unsigned w, unsigned h,
-                       float *p,
-                       unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         const ushort pixel = *src++;
-         pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f);
-         pRow[1] = ((pixel >>  5) & 0x1f) * (1.0f / 31.0f);
-         pRow[2] = ((pixel      ) & 0x1f) * (1.0f / 31.0f);
-         pRow[3] = 1.0f;
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-x1r5g5b5_put_tile_rgba(ushort *dst,
-                       unsigned w, unsigned h,
-                       const float *p,
-                       unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r, g, b;
-         r = float_to_ubyte(pRow[0]);
-         g = float_to_ubyte(pRow[1]);
-         b = float_to_ubyte(pRow[2]);
-         r = r >> 3;  /* 5 bits */
-         g = g >> 3;  /* 5 bits */
-         b = b >> 3;  /* 5 bits */
-         *dst++ = (1 << 15) | (r << 10) | (g << 5) | b;
-      }
-      p += src_stride;
-   }
-}
-
-
-/*** PIPE_FORMAT_B5G5R5A1_UNORM ***/
-
-static void
-a1r5g5b5_get_tile_rgba(const ushort *src,
-                       unsigned w, unsigned h,
-                       float *p,
-                       unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         const ushort pixel = *src++;
-         pRow[0] = ((pixel >> 10) & 0x1f) * (1.0f / 31.0f);
-         pRow[1] = ((pixel >>  5) & 0x1f) * (1.0f / 31.0f);
-         pRow[2] = ((pixel      ) & 0x1f) * (1.0f / 31.0f);
-         pRow[3] = ((pixel >> 15)       ) * 1.0f;
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-a1r5g5b5_put_tile_rgba(ushort *dst,
-                       unsigned w, unsigned h,
-                       const float *p,
-                       unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r, g, b, a;
-         r = float_to_ubyte(pRow[0]);
-         g = float_to_ubyte(pRow[1]);
-         b = float_to_ubyte(pRow[2]);
-         a = float_to_ubyte(pRow[3]);
-         r = r >> 3;  /* 5 bits */
-         g = g >> 3;  /* 5 bits */
-         b = b >> 3;  /* 5 bits */
-         a = a >> 7;  /* 1 bit */
-         *dst++ = (a << 15) | (r << 10) | (g << 5) | b;
-      }
-      p += src_stride;
-   }
-}
-
-
-/*** PIPE_FORMAT_B4G4R4A4_UNORM ***/
-
-static void
-a4r4g4b4_get_tile_rgba(const ushort *src,
-                       unsigned w, unsigned h,
-                       float *p,
-                       unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         const ushort pixel = *src++;
-         pRow[0] = ((pixel >>  8) & 0xf) * (1.0f / 15.0f);
-         pRow[1] = ((pixel >>  4) & 0xf) * (1.0f / 15.0f);
-         pRow[2] = ((pixel      ) & 0xf) * (1.0f / 15.0f);
-         pRow[3] = ((pixel >> 12)      ) * (1.0f / 15.0f);
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-a4r4g4b4_put_tile_rgba(ushort *dst,
-                       unsigned w, unsigned h,
-                       const float *p,
-                       unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r, g, b, a;
-         r = float_to_ubyte(pRow[0]);
-         g = float_to_ubyte(pRow[1]);
-         b = float_to_ubyte(pRow[2]);
-         a = float_to_ubyte(pRow[3]);
-         r >>= 4;
-         g >>= 4;
-         b >>= 4;
-         a >>= 4;
-         *dst++ = (a << 12) | (r << 8) | (g << 4) | b;
-      }
-      p += src_stride;
-   }
-}
-
-
-/*** PIPE_FORMAT_B5G6R5_UNORM ***/
-
-static void
-r5g6b5_get_tile_rgba(const ushort *src,
-                     unsigned w, unsigned h,
-                     float *p,
-                     unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         const ushort pixel = *src++;
-         pRow[0] = ((pixel >> 11) & 0x1f) * (1.0f / 31.0f);
-         pRow[1] = ((pixel >>  5) & 0x3f) * (1.0f / 63.0f);
-         pRow[2] = ((pixel      ) & 0x1f) * (1.0f / 31.0f);
-         pRow[3] = 1.0f;
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-r5g6b5_put_tile_rgba(ushort *dst,
-                     unsigned w, unsigned h,
-                     const float *p,
-                     unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         uint r = (uint) (CLAMP(pRow[0], 0.0, 1.0) * 31.0);
-         uint g = (uint) (CLAMP(pRow[1], 0.0, 1.0) * 63.0);
-         uint b = (uint) (CLAMP(pRow[2], 0.0, 1.0) * 31.0);
-         *dst++ = (r << 11) | (g << 5) | (b);
-      }
-      p += src_stride;
-   }
-}
-
-
-
-/*** PIPE_FORMAT_R8G8B8_UNORM ***/
-
-static void
-r8g8b8_get_tile_rgba(const ubyte *src,
-                     unsigned w, unsigned h,
-                     float *p,
-                     unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         pRow[0] = ubyte_to_float(src[0]);
-         pRow[1] = ubyte_to_float(src[1]);
-         pRow[2] = ubyte_to_float(src[2]);
-         pRow[3] = 1.0f;
-         src += 3;
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-r8g8b8_put_tile_rgba(ubyte *dst,
-                     unsigned w, unsigned h,
-                     const float *p,
-                     unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         dst[0] = float_to_ubyte(pRow[0]);
-         dst[1] = float_to_ubyte(pRow[1]);
-         dst[2] = float_to_ubyte(pRow[2]);
-         dst += 3;
-      }
-      p += src_stride;
-   }
-}
-
-
-
 /*** PIPE_FORMAT_Z16_UNORM ***/
 
 /**
@@ -567,179 +137,6 @@ z16_get_tile_rgba(const ushort *src,
 
 
 
-/*** PIPE_FORMAT_L8_UNORM ***/
-
-static void
-l8_get_tile_rgba(const ubyte *src,
-                 unsigned w, unsigned h,
-                 float *p,
-                 unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, src++, pRow += 4) {
-         pRow[0] =
-         pRow[1] =
-         pRow[2] = ubyte_to_float(*src);
-         pRow[3] = 1.0;
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-l8_put_tile_rgba(ubyte *dst,
-                 unsigned w, unsigned h,
-                 const float *p,
-                 unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r;
-         r = float_to_ubyte(pRow[0]);
-         *dst++ = (ubyte) r;
-      }
-      p += src_stride;
-   }
-}
-
-
-
-/*** PIPE_FORMAT_A8_UNORM ***/
-
-static void
-a8_get_tile_rgba(const ubyte *src,
-                 unsigned w, unsigned h,
-                 float *p,
-                 unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, src++, pRow += 4) {
-         pRow[0] =
-         pRow[1] =
-         pRow[2] = 0.0;
-         pRow[3] = ubyte_to_float(*src);
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-a8_put_tile_rgba(ubyte *dst,
-                 unsigned w, unsigned h,
-                 const float *p,
-                 unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned a;
-         a = float_to_ubyte(pRow[3]);
-         *dst++ = (ubyte) a;
-      }
-      p += src_stride;
-   }
-}
-
-
-
-/*** PIPE_FORMAT_R16_SNORM ***/
-
-static void
-r16_get_tile_rgba(const short *src,
-                  unsigned w, unsigned h,
-                  float *p,
-                  unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, src++, pRow += 4) {
-         pRow[0] = SHORT_TO_FLOAT(src[0]);
-         pRow[1] =
-         pRow[2] = 0.0;
-         pRow[3] = 1.0;
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-r16_put_tile_rgba(short *dst,
-                  unsigned w, unsigned h,
-                  const float *p,
-                  unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, dst++, pRow += 4) {
-         UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]);
-      }
-      p += src_stride;
-   }
-}
-
-
-/*** PIPE_FORMAT_R16G16B16A16_SNORM ***/
-
-static void
-r16g16b16a16_get_tile_rgba(const short *src,
-                           unsigned w, unsigned h,
-                           float *p,
-                           unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, src += 4, pRow += 4) {
-         pRow[0] = SHORT_TO_FLOAT(src[0]);
-         pRow[1] = SHORT_TO_FLOAT(src[1]);
-         pRow[2] = SHORT_TO_FLOAT(src[2]);
-         pRow[3] = SHORT_TO_FLOAT(src[3]);
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-r16g16b16a16_put_tile_rgba(short *dst,
-                           unsigned w, unsigned h,
-                           const float *p,
-                           unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, dst += 4, pRow += 4) {
-         UNCLAMPED_FLOAT_TO_SHORT(dst[0], pRow[0]);
-         UNCLAMPED_FLOAT_TO_SHORT(dst[1], pRow[1]);
-         UNCLAMPED_FLOAT_TO_SHORT(dst[2], pRow[2]);
-         UNCLAMPED_FLOAT_TO_SHORT(dst[3], pRow[3]);
-      }
-      p += src_stride;
-   }
-}
-
-
 /*** PIPE_FORMAT_A8B8G8R8_SRGB ***/
 
 /**
@@ -919,96 +316,6 @@ l8_srgb_put_tile_rgba(ubyte *dst,
 }
 
 
-/*** PIPE_FORMAT_I8_UNORM ***/
-
-static void
-i8_get_tile_rgba(const ubyte *src,
-                 unsigned w, unsigned h,
-                 float *p,
-                 unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, src++, pRow += 4) {
-         pRow[0] =
-         pRow[1] =
-         pRow[2] =
-         pRow[3] = ubyte_to_float(*src);
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-i8_put_tile_rgba(ubyte *dst,
-                 unsigned w, unsigned h,
-                 const float *p,
-                 unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r;
-         r = float_to_ubyte(pRow[0]);
-         *dst++ = (ubyte) r;
-      }
-      p += src_stride;
-   }
-}
-
-
-/*** PIPE_FORMAT_L8A8_UNORM ***/
-
-static void
-a8l8_get_tile_rgba(const ushort *src,
-                   unsigned w, unsigned h,
-                   float *p,
-                   unsigned dst_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         ushort p = *src++;
-         pRow[0] =
-         pRow[1] =
-         pRow[2] = ubyte_to_float(p & 0xff);
-         pRow[3] = ubyte_to_float(p >> 8);
-      }
-      p += dst_stride;
-   }
-}
-
-
-static void
-a8l8_put_tile_rgba(ushort *dst,
-                   unsigned w, unsigned h,
-                   const float *p,
-                   unsigned src_stride)
-{
-   unsigned i, j;
-
-   for (i = 0; i < h; i++) {
-      const float *pRow = p;
-      for (j = 0; j < w; j++, pRow += 4) {
-         unsigned r, a;
-         r = float_to_ubyte(pRow[0]);
-         a = float_to_ubyte(pRow[3]);
-         *dst++ = (a << 8) | r;
-      }
-      p += src_stride;
-   }
-}
-
-
-
-
 /*** PIPE_FORMAT_Z32_UNORM ***/
 
 /**
@@ -1036,7 +343,7 @@ z32_get_tile_rgba(const unsigned *src,
 }
 
 
-/*** PIPE_FORMAT_Z24S8_UNORM ***/
+/*** PIPE_FORMAT_Z24_UNORM_S8_USCALED ***/
 
 /**
  * Return Z component as four float in [0,1].  Stencil part ignored.
@@ -1063,7 +370,7 @@ s8z24_get_tile_rgba(const unsigned *src,
 }
 
 
-/*** PIPE_FORMAT_S8Z24_UNORM ***/
+/*** PIPE_FORMAT_S8_USCALED_Z24_UNORM ***/
 
 /**
  * Return Z component as four float in [0,1].  Stencil part ignored.
@@ -1211,51 +518,6 @@ pipe_tile_raw_to_rgba(enum pipe_format format,
                       float *dst, unsigned dst_stride)
 {
    switch (format) {
-   case PIPE_FORMAT_B8G8R8A8_UNORM:
-      a8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_B8G8R8X8_UNORM:
-      x8r8g8b8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_A8R8G8B8_UNORM:
-      b8g8r8a8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_A8B8G8R8_UNORM:
-      r8g8b8a8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_B5G5R5X1_UNORM:
-      x1r5g5b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_B5G5R5A1_UNORM:
-      a1r5g5b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_B4G4R4A4_UNORM:
-      a4r4g4b4_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_B5G6R5_UNORM:
-      r5g6b5_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_R8G8B8_UNORM:
-      r8g8b8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_L8_UNORM:
-      l8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_A8_UNORM:
-      a8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_I8_UNORM:
-      i8_get_tile_rgba((ubyte *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_L8A8_UNORM:
-      a8l8_get_tile_rgba((ushort *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_R16_SNORM:
-      r16_get_tile_rgba((short *) src, w, h, dst, dst_stride);
-      break;
-   case PIPE_FORMAT_R16G16B16A16_SNORM:
-      r16g16b16a16_get_tile_rgba((short *) src, w, h, dst, dst_stride);
-      break;
    case PIPE_FORMAT_B8G8R8A8_SRGB:
       a8r8g8b8_srgb_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
       break;
@@ -1271,11 +533,11 @@ pipe_tile_raw_to_rgba(enum pipe_format format,
    case PIPE_FORMAT_Z32_UNORM:
       z32_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
       break;
-   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
    case PIPE_FORMAT_Z24X8_UNORM:
       s8z24_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
       break;
-   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
    case PIPE_FORMAT_X8Z24_UNORM:
       z24s8_get_tile_rgba((unsigned *) src, w, h, dst, dst_stride);
       break;
@@ -1408,51 +670,6 @@ pipe_put_tile_rgba(struct pipe_context *pipe,
       return;
 
    switch (format) {
-   case PIPE_FORMAT_B8G8R8A8_UNORM:
-      a8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_B8G8R8X8_UNORM:
-      x8r8g8b8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_A8R8G8B8_UNORM:
-      b8g8r8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_A8B8G8R8_UNORM:
-      r8g8b8a8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_B5G5R5X1_UNORM:
-      x1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_B5G5R5A1_UNORM:
-      a1r5g5b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_B5G6R5_UNORM:
-      r5g6b5_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_R8G8B8_UNORM:
-      r8g8b8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_B4G4R4A4_UNORM:
-      a4r4g4b4_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_L8_UNORM:
-      l8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_A8_UNORM:
-      a8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_I8_UNORM:
-      i8_put_tile_rgba((ubyte *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_L8A8_UNORM:
-      a8l8_put_tile_rgba((ushort *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_R16_SNORM:
-      r16_put_tile_rgba((short *) packed, w, h, p, src_stride);
-      break;
-   case PIPE_FORMAT_R16G16B16A16_SNORM:
-      r16g16b16a16_put_tile_rgba((short *) packed, w, h, p, src_stride);
-      break;
    case PIPE_FORMAT_B8G8R8A8_SRGB:
       a8r8g8b8_srgb_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);
       break;
@@ -1468,11 +685,11 @@ pipe_put_tile_rgba(struct pipe_context *pipe,
    case PIPE_FORMAT_Z32_UNORM:
       /*z32_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/
       break;
-   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
    case PIPE_FORMAT_Z24X8_UNORM:
       /*s8z24_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/
       break;
-   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
    case PIPE_FORMAT_X8Z24_UNORM:
       /*z24s8_put_tile_rgba((unsigned *) packed, w, h, p, src_stride);*/
       break;
@@ -1525,7 +742,7 @@ pipe_get_tile_z(struct pipe_context *pipe,
          }
       }
       break;
-   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
    case PIPE_FORMAT_Z24X8_UNORM:
       {
          const uint *ptrc
@@ -1540,7 +757,7 @@ pipe_get_tile_z(struct pipe_context *pipe,
          }
       }
       break;
-   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
    case PIPE_FORMAT_X8Z24_UNORM:
       {
          const uint *ptrc
@@ -1609,7 +826,7 @@ pipe_put_tile_z(struct pipe_context *pipe,
          }
       }
       break;
-   case PIPE_FORMAT_Z24S8_UNORM:
+   case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
       {
          uint *pDest = (uint *) (map + y * pt->stride + x*4);
          assert((pt->usage & PIPE_TRANSFER_READ_WRITE) == PIPE_TRANSFER_READ_WRITE);
@@ -1636,7 +853,7 @@ pipe_put_tile_z(struct pipe_context *pipe,
          }
       }
       break;
-   case PIPE_FORMAT_S8Z24_UNORM:
+   case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
       {
          uint *pDest = (uint *) (map + y * pt->stride + x*4);
          assert((pt->usage & PIPE_TRANSFER_READ_WRITE) == PIPE_TRANSFER_READ_WRITE);