Merge commit 'origin/master' into gallium-msaa

31 files changed, 843 insertions, 782 deletions

diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile
index 2daed382cf3..e0d9b313544 100644
--- a/src/gallium/auxiliary/Makefile
+++ b/src/gallium/auxiliary/Makefile
@@ -113,6 +113,7 @@ C_SOURCES = \
 	util/u_format_tests.c \
 	util/u_format_yuv.c \
 	util/u_format_zs.c \
+	util/u_framebuffer.c \
 	util/u_gen_mipmap.c \
 	util/u_half.c \
 	util/u_handle_table.c \
diff --git a/src/gallium/auxiliary/SConscript b/src/gallium/auxiliary/SConscript
index a0673df8a8e..29a1bd0ecba 100644
--- a/src/gallium/auxiliary/SConscript
+++ b/src/gallium/auxiliary/SConscript
@@ -162,6 +162,7 @@ source = [
     'util/u_format_tests.c',
     'util/u_format_yuv.c',
     'util/u_format_zs.c',
+    'util/u_framebuffer.c',
     'util/u_gen_mipmap.c',
     'util/u_half.c',
     'util/u_handle_table.c',
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c
index 55e0ded1b30..20a8612dcae 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.c
+++ b/src/gallium/auxiliary/cso_cache/cso_context.c
@@ -36,6 +36,7 @@
   */
 
 #include "pipe/p_state.h"
+#include "util/u_framebuffer.h"
 #include "util/u_inlines.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
@@ -103,10 +104,6 @@ struct cso_context {
 };
 
 
-static void
-free_framebuffer_state(struct pipe_framebuffer_state *fb);
-
-
 static boolean delete_blend_state(struct cso_context *ctx, void *state)
 {
    struct cso_blend *cso = (struct cso_blend *)state;
@@ -304,8 +301,8 @@ void cso_release_all( struct cso_context *ctx )
       pipe_sampler_view_reference(&ctx->vertex_sampler_views_saved[i], NULL);
    }
 
-   free_framebuffer_state(&ctx->fb);
-   free_framebuffer_state(&ctx->fb_saved);
+   util_unreference_framebuffer_state(&ctx->fb);
+   util_unreference_framebuffer_state(&ctx->fb_saved);
 
    if (ctx->cache) {
       cso_cache_delete( ctx->cache );
@@ -897,42 +894,11 @@ void cso_restore_vertex_shader(struct cso_context *ctx)
 }
 
 
-/**
- * Copy framebuffer state from src to dst with refcounting of surfaces.
- */
-static void
-copy_framebuffer_state(struct pipe_framebuffer_state *dst,
-                       const struct pipe_framebuffer_state *src)
-{
-   uint i;
-
-   dst->width = src->width;
-   dst->height = src->height;
-   dst->nr_cbufs = src->nr_cbufs;
-   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
-      pipe_surface_reference(&dst->cbufs[i], src->cbufs[i]);
-   }
-   pipe_surface_reference(&dst->zsbuf, src->zsbuf);
-}
-
-
-static void
-free_framebuffer_state(struct pipe_framebuffer_state *fb)
-{
-   uint i;
-
-   for (i = 0; i < PIPE_MAX_COLOR_BUFS; i++) {
-      pipe_surface_reference(&fb->cbufs[i], NULL);
-   }
-   pipe_surface_reference(&fb->zsbuf, NULL);
-}
-
-
 enum pipe_error cso_set_framebuffer(struct cso_context *ctx,
                                     const struct pipe_framebuffer_state *fb)
 {
    if (memcmp(&ctx->fb, fb, sizeof(*fb)) != 0) {
-      copy_framebuffer_state(&ctx->fb, fb);
+      util_copy_framebuffer_state(&ctx->fb, fb);
       ctx->pipe->set_framebuffer_state(ctx->pipe, fb);
    }
    return PIPE_OK;
@@ -940,15 +906,15 @@ enum pipe_error cso_set_framebuffer(struct cso_context *ctx,
 
 void cso_save_framebuffer(struct cso_context *ctx)
 {
-   copy_framebuffer_state(&ctx->fb_saved, &ctx->fb);
+   util_copy_framebuffer_state(&ctx->fb_saved, &ctx->fb);
 }
 
 void cso_restore_framebuffer(struct cso_context *ctx)
 {
    if (memcmp(&ctx->fb, &ctx->fb_saved, sizeof(ctx->fb))) {
-      copy_framebuffer_state(&ctx->fb, &ctx->fb_saved);
+      util_copy_framebuffer_state(&ctx->fb, &ctx->fb_saved);
       ctx->pipe->set_framebuffer_state(ctx->pipe, &ctx->fb);
-      free_framebuffer_state(&ctx->fb_saved);
+      util_unreference_framebuffer_state(&ctx->fb_saved);
    }
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_llvm.c b/src/gallium/auxiliary/draw/draw_llvm.c
index 3b2df054c3c..ea9b7c90a51 100644
--- a/src/gallium/auxiliary/draw/draw_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_llvm.c
@@ -182,6 +182,8 @@ draw_llvm_create(struct draw_context *draw)
    /* These are the passes currently listed in llvm-c/Transforms/Scalar.h,
     * but there are more on SVN. */
    /* TODO: Add more passes */
+   LLVMAddCFGSimplificationPass(llvm->pass);
+   LLVMAddPromoteMemoryToRegisterPass(llvm->pass);
    LLVMAddConstantPropagationPass(llvm->pass);
    if(util_cpu_caps.has_sse4_1) {
       /* FIXME: There is a bug in this pass, whereby the combination of fptosi
@@ -190,9 +192,7 @@ draw_llvm_create(struct draw_context *draw)
        */
       LLVMAddInstructionCombiningPass(llvm->pass);
    }
-   LLVMAddPromoteMemoryToRegisterPass(llvm->pass);
    LLVMAddGVNPass(llvm->pass);
-   LLVMAddCFGSimplificationPass(llvm->pass);
 
    init_globals(llvm);
 
@@ -207,6 +207,8 @@ draw_llvm_create(struct draw_context *draw)
 void
 draw_llvm_destroy(struct draw_llvm *llvm)
 {
+   LLVMDisposePassManager(llvm->pass);
+
    FREE(llvm);
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
index d2a492f2b4c..6d00b0fbd44 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline_llvm.c
@@ -66,7 +66,6 @@ llvm_middle_end_prepare( struct draw_pt_middle_end *middle,
    struct llvm_middle_end *fpme = (struct llvm_middle_end *)middle;
    struct draw_context *draw = fpme->draw;
    struct draw_vertex_shader *vs = draw->vs.vertex_shader;
-   struct draw_geometry_shader *gs = draw->gs.geometry_shader;
    struct draw_llvm_variant_key key;
    struct draw_llvm_variant *variant = NULL;
    unsigned i;
@@ -97,17 +96,6 @@ llvm_middle_end_prepare( struct draw_pt_middle_end *middle,
    fpme->vertex_size = sizeof(struct vertex_header) + nr * 4 * sizeof(float);
 
 
-
-   draw_pt_fetch_prepare( fpme->fetch,
-                          vs->info.num_inputs,
-                          fpme->vertex_size,
-                          instance_id_index );
-   if (opt & PT_SHADE) {
-      vs->prepare(vs, draw);
-      draw_geometry_shader_prepare(gs, draw);
-   }
-
-
    /* XXX: it's not really gl rasterization rules we care about here,
     * but gl vs dx9 clip spaces.
     */
@@ -349,7 +337,31 @@ static void llvm_middle_end_finish( struct draw_pt_middle_end *middle )
 static void llvm_middle_end_destroy( struct draw_pt_middle_end *middle )
 {
    struct llvm_middle_end *fpme = (struct llvm_middle_end *)middle;
+   struct draw_context *draw = fpme->draw;
+   struct draw_llvm_variant *variant = NULL;
 
+   variant = fpme->variants;
+   while(variant) {
+      struct draw_llvm_variant *next = variant->next;
+
+      if (variant->function_elts) {
+         if (variant->function_elts)
+            LLVMFreeMachineCodeForFunction(draw->engine,
+                                           variant->function_elts);
+         LLVMDeleteFunction(variant->function_elts);
+      }
+
+      if (variant->function) {
+         if (variant->function)
+            LLVMFreeMachineCodeForFunction(draw->engine,
+                                           variant->function);
+         LLVMDeleteFunction(variant->function);
+      }
+
+      FREE(variant);
+
+      variant = next;
+   }
    if (fpme->fetch)
       draw_pt_fetch_destroy( fpme->fetch );
 
diff --git a/src/gallium/auxiliary/draw/draw_vs.c b/src/gallium/auxiliary/draw/draw_vs.c
index c2832eefa2a..b9db886a24d 100644
--- a/src/gallium/auxiliary/draw/draw_vs.c
+++ b/src/gallium/auxiliary/draw/draw_vs.c
@@ -85,18 +85,22 @@ struct draw_vertex_shader *
 draw_create_vertex_shader(struct draw_context *draw,
                           const struct pipe_shader_state *shader)
 {
-   struct draw_vertex_shader *vs;
+   struct draw_vertex_shader *vs = NULL;
 
    if (draw->dump_vs) {
       tgsi_dump(shader->tokens, 0);
    }
 
-   vs = draw_create_vs_sse( draw, shader );
-   if (!vs) {
+   if (!draw->pt.middle.llvm) {
+#if defined(PIPE_ARCH_X86)
+      vs = draw_create_vs_sse( draw, shader );
+#elif defined(PIPE_ARCH_PPC)
       vs = draw_create_vs_ppc( draw, shader );
-      if (!vs) {
-         vs = draw_create_vs_exec( draw, shader );
-      }
+#endif
+   }
+
+   if (!vs) {
+      vs = draw_create_vs_exec( draw, shader );
    }
 
    if (vs)
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_arit.c b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
index 20ae958714b..f372a48846f 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_arit.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_arit.c
@@ -1177,9 +1177,34 @@ lp_build_rcp(struct lp_build_context *bld,
    if(LLVMIsConstant(a))
       return LLVMConstFDiv(bld->one, a);
 
-   if(util_cpu_caps.has_sse && type.width == 32 && type.length == 4)
-      /* FIXME: improve precision */
+   if(util_cpu_caps.has_sse && type.width == 32 && type.length == 4) {
+      /*
+       * XXX: Added precision is not always necessary, so only enable this
+       * when we have a better system in place to track minimum precision.
+       */
+
+#if 0
+      /*
+       * Do one Newton-Raphson step to improve precision:
+       *
+       *   x1 = (2 - a * rcp(a)) * rcp(a)
+       */
+
+      LLVMValueRef two = lp_build_const_vec(bld->type, 2.0);
+      LLVMValueRef rcp_a;
+      LLVMValueRef res;
+
+      rcp_a = lp_build_intrinsic_unary(bld->builder, "llvm.x86.sse.rcp.ps", lp_build_vec_type(type), a);
+
+      res = LLVMBuildMul(bld->builder, a, rcp_a, "");
+      res = LLVMBuildSub(bld->builder, two, res, "");
+      res = LLVMBuildMul(bld->builder, res, rcp_a, "");
+
+      return rcp_a;
+#else
       return lp_build_intrinsic_unary(bld->builder, "llvm.x86.sse.rcp.ps", lp_build_vec_type(type), a);
+#endif
+   }
 
    return LLVMBuildFDiv(bld->builder, bld->one, a, "");
 }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_limits.h b/src/gallium/auxiliary/gallivm/lp_bld_limits.h
new file mode 100644
index 00000000000..e095a0abe3a
--- /dev/null
+++ b/src/gallium/auxiliary/gallivm/lp_bld_limits.h
@@ -0,0 +1,53 @@
+/**************************************************************************
+ *
+ * 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 LP_BLD_LIMITS_H_
+#define LP_BLD_LIMITS_H_
+
+/*
+ * TGSI translation limits.
+ *
+ * Some are slightly above SM 3.0 requirements to give some wiggle room to
+ * the state trackers.
+ */
+
+#define LP_MAX_TGSI_TEMPS 256
+
+#define LP_MAX_TGSI_ADDRS 16
+
+#define LP_MAX_TGSI_IMMEDIATES 256
+
+/**
+ * Maximum control flow nesting
+ *
+ * SM3.0 requires 24
+ */
+#define LP_MAX_TGSI_NESTING 32
+
+
+#endif /* LP_BLD_LIMITS_H_ */
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample.h b/src/gallium/auxiliary/gallivm/lp_bld_sample.h
index 8ceb20473d5..955b1d21ee5 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample.h
@@ -173,7 +173,10 @@ lp_build_sample_soa(LLVMBuilderRef builder,
                     unsigned unit,
                     unsigned num_coords,
                     const LLVMValueRef *coords,
-                    LLVMValueRef lodbias,
+                    const LLVMValueRef *ddx,
+                    const LLVMValueRef *ddy,
+                    LLVMValueRef lod_bias,
+                    LLVMValueRef explicit_lod,
                     LLVMValueRef *texel);
 
 
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
index 74dc9e1d812..3f0ea05b795 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c
@@ -30,6 +30,7 @@
  * Texture sampling -- SoA.
  *
  * @author Jose Fonseca <[email protected]>
+ * @author Brian Paul <[email protected]>
  */
 
 #include "pipe/p_defines.h"
@@ -325,6 +326,18 @@ lp_build_sample_texel_soa(struct lp_build_sample_context *bld,
                                    bld->format_desc,
                                    x, y, z, y_stride, z_stride);
 
+   if (use_border) {
+      /* If we can sample the border color, it means that texcoords may
+       * lie outside the bounds of the texture image.  We need to do
+       * something to prevent reading out of bounds and causing a segfault.
+       *
+       * Simply AND the texture coords with !use_border.  This will cause
+       * coords which are out of bounds to become zero.  Zero's guaranteed
+       * to be inside the texture image.
+       */
+      offset = lp_build_andc(&bld->uint_coord_bld, offset, use_border);
+   }
+
    lp_build_fetch_rgba_soa(bld->builder,
                            bld->format_desc,
                            bld->texel_type,
@@ -891,17 +904,24 @@ lp_build_minify(struct lp_build_sample_context *bld,
  * \param s  vector of texcoord s values
  * \param t  vector of texcoord t values
  * \param r  vector of texcoord r values
- * \param shader_lod_bias  vector float with the shader lod bias,
+ * \param lod_bias  optional float vector with the shader lod bias
+ * \param explicit_lod  optional float vector with the explicit lod
  * \param width  scalar int texture width
  * \param height  scalar int texture height
  * \param depth  scalar int texture depth
+ *
+ * XXX: The resulting lod is scalar, so ignore all but the first element of
+ * derivatives, lod_bias, etc that are passed by the shader.
  */
 static LLVMValueRef
 lp_build_lod_selector(struct lp_build_sample_context *bld,
                       LLVMValueRef s,
                       LLVMValueRef t,
                       LLVMValueRef r,
-                      LLVMValueRef shader_lod_bias,
+                      const LLVMValueRef *ddx,
+                      const LLVMValueRef *ddy,
+                      LLVMValueRef lod_bias, /* optional */
+                      LLVMValueRef explicit_lod, /* optional */
                       LLVMValueRef width,
                       LLVMValueRef height,
                       LLVMValueRef depth)
@@ -914,7 +934,6 @@ lp_build_lod_selector(struct lp_build_sample_context *bld,
       return LLVMConstReal(LLVMFloatType(), bld->static_state->min_lod);
    }
    else {
-      const int dims = texture_dims(bld->static_state->target);
       struct lp_build_context *float_bld = &bld->float_bld;
       LLVMValueRef sampler_lod_bias = LLVMConstReal(LLVMFloatType(),
                                                     bld->static_state->lod_bias);
@@ -922,83 +941,76 @@ lp_build_lod_selector(struct lp_build_sample_context *bld,
                                            bld->static_state->min_lod);
       LLVMValueRef max_lod = LLVMConstReal(LLVMFloatType(),
                                            bld->static_state->max_lod);
-
       LLVMValueRef index0 = LLVMConstInt(LLVMInt32Type(), 0, 0);
-      LLVMValueRef index1 = LLVMConstInt(LLVMInt32Type(), 1, 0);
-      LLVMValueRef index2 = LLVMConstInt(LLVMInt32Type(), 2, 0);
-
-      LLVMValueRef s0, s1, s2;
-      LLVMValueRef t0, t1, t2;
-      LLVMValueRef r0, r1, r2;
-      LLVMValueRef dsdx, dsdy, dtdx, dtdy, drdx, drdy;
-      LLVMValueRef rho, lod;
-
-      /*
-       * dsdx = abs(s[1] - s[0]);
-       * dsdy = abs(s[2] - s[0]);
-       * dtdx = abs(t[1] - t[0]);
-       * dtdy = abs(t[2] - t[0]);
-       * drdx = abs(r[1] - r[0]);
-       * drdy = abs(r[2] - r[0]);
-       * XXX we're assuming a four-element quad in 2x2 layout here.
-       */
-      s0 = LLVMBuildExtractElement(bld->builder, s, index0, "s0");
-      s1 = LLVMBuildExtractElement(bld->builder, s, index1, "s1");
-      s2 = LLVMBuildExtractElement(bld->builder, s, index2, "s2");
-      dsdx = LLVMBuildSub(bld->builder, s1, s0, "");
-      dsdx = lp_build_abs(float_bld, dsdx);
-      dsdy = LLVMBuildSub(bld->builder, s2, s0, "");
-      dsdy = lp_build_abs(float_bld, dsdy);
-      if (dims > 1) {
-         t0 = LLVMBuildExtractElement(bld->builder, t, index0, "t0");
-         t1 = LLVMBuildExtractElement(bld->builder, t, index1, "t1");
-         t2 = LLVMBuildExtractElement(bld->builder, t, index2, "t2");
-         dtdx = LLVMBuildSub(bld->builder, t1, t0, "");
-         dtdx = lp_build_abs(float_bld, dtdx);
-         dtdy = LLVMBuildSub(bld->builder, t2, t0, "");
-         dtdy = lp_build_abs(float_bld, dtdy);
-         if (dims > 2) {
-            r0 = LLVMBuildExtractElement(bld->builder, r, index0, "r0");
-            r1 = LLVMBuildExtractElement(bld->builder, r, index1, "r1");
-            r2 = LLVMBuildExtractElement(bld->builder, r, index2, "r2");
-            drdx = LLVMBuildSub(bld->builder, r1, r0, "");
-            drdx = lp_build_abs(float_bld, drdx);
-            drdy = LLVMBuildSub(bld->builder, r2, r0, "");
-            drdy = lp_build_abs(float_bld, drdy);
-         }
+      LLVMValueRef lod;
+
+      if (explicit_lod) {
+         lod = LLVMBuildExtractElement(bld->builder, explicit_lod,
+                                       index0, "");
       }
+      else {
+         const int dims = texture_dims(bld->static_state->target);
+         LLVMValueRef dsdx, dsdy, dtdx, dtdy, drdx, drdy;
+         LLVMValueRef rho;
+
+         /*
+          * dsdx = abs(s[1] - s[0]);
+          * dsdy = abs(s[2] - s[0]);
+          * dtdx = abs(t[1] - t[0]);
+          * dtdy = abs(t[2] - t[0]);
+          * drdx = abs(r[1] - r[0]);
+          * drdy = abs(r[2] - r[0]);
+          */
+         dsdx = LLVMBuildExtractElement(bld->builder, ddx[0], index0, "dsdx");
+         dsdx = lp_build_abs(float_bld, dsdx);
+         dsdy = LLVMBuildExtractElement(bld->builder, ddy[0], index0, "dsdy");
+         dsdy = lp_build_abs(float_bld, dsdy);
+         if (dims > 1) {
+            dtdx = LLVMBuildExtractElement(bld->builder, ddx[1], index0, "dtdx");
+            dtdx = lp_build_abs(float_bld, dtdx);
+            dtdy = LLVMBuildExtractElement(bld->builder, ddy[1], index0, "dtdy");
+            dtdy = lp_build_abs(float_bld, dtdy);
+            if (dims > 2) {
+               drdx = LLVMBuildExtractElement(bld->builder, ddx[2], index0, "drdx");
+               drdx = lp_build_abs(float_bld, drdx);
+               drdy = LLVMBuildExtractElement(bld->builder, ddy[2], index0, "drdy");
+               drdy = lp_build_abs(float_bld, drdy);
+            }
+         }
 
-      /* Compute rho = max of all partial derivatives scaled by texture size.
-       * XXX this could be vectorized somewhat
-       */
-      rho = LLVMBuildMul(bld->builder,
-                         lp_build_max(float_bld, dsdx, dsdy),
-                         lp_build_int_to_float(float_bld, width), "");
-      if (dims > 1) {
-         LLVMValueRef max;
-         max = LLVMBuildMul(bld->builder,
-                            lp_build_max(float_bld, dtdx, dtdy),
-                            lp_build_int_to_float(float_bld, height), "");
-         rho = lp_build_max(float_bld, rho, max);
-         if (dims > 2) {
+         /* Compute rho = max of all partial derivatives scaled by texture size.
+          * XXX this could be vectorized somewhat
+          */
+         rho = LLVMBuildMul(bld->builder,
+                            lp_build_max(float_bld, dsdx, dsdy),
+                            lp_build_int_to_float(float_bld, width), "");
+         if (dims > 1) {
+            LLVMValueRef max;
             max = LLVMBuildMul(bld->builder,
-                               lp_build_max(float_bld, drdx, drdy),
-                               lp_build_int_to_float(float_bld, depth), "");
+                               lp_build_max(float_bld, dtdx, dtdy),
+                               lp_build_int_to_float(float_bld, height), "");
             rho = lp_build_max(float_bld, rho, max);
+            if (dims > 2) {
+               max = LLVMBuildMul(bld->builder,
+                                  lp_build_max(float_bld, drdx, drdy),
+                                  lp_build_int_to_float(float_bld, depth), "");
+               rho = lp_build_max(float_bld, rho, max);
+            }
          }
-      }
 
-      /* compute lod = log2(rho) */
-      lod = lp_build_log2(float_bld, rho);
+         /* compute lod = log2(rho) */
+         lod = lp_build_log2(float_bld, rho);
 
-      /* add sampler lod bias */
-      lod = LLVMBuildAdd(bld->builder, lod, sampler_lod_bias, "sampler LOD bias");
+         /* add shader lod bias */
+         if (lod_bias) {
+            lod_bias = LLVMBuildExtractElement(bld->builder, lod_bias,
+                                               index0, "");
+            lod = LLVMBuildAdd(bld->builder, lod, lod_bias, "shader_lod_bias");
+         }
+      }
 
-      /* add shader lod bias */
-      /* XXX for now we take only the first element since our lod is scalar */
-      shader_lod_bias = LLVMBuildExtractElement(bld->builder, shader_lod_bias,
-                                                LLVMConstInt(LLVMInt32Type(), 0, 0), "");
-      lod = LLVMBuildAdd(bld->builder, lod, shader_lod_bias, "shader LOD bias");
+      /* add sampler lod bias */
+      lod = LLVMBuildAdd(bld->builder, lod, sampler_lod_bias, "sampler_lod_bias");
 
       /* clamp lod */
       lod = lp_build_clamp(float_bld, lod, min_lod, max_lod);
@@ -1584,7 +1596,10 @@ lp_build_sample_general(struct lp_build_sample_context *bld,
                         LLVMValueRef s,
                         LLVMValueRef t,
                         LLVMValueRef r,
-                        LLVMValueRef lodbias,
+                        const LLVMValueRef *ddx,
+                        const LLVMValueRef *ddy,
+                        LLVMValueRef lod_bias, /* optional */
+                        LLVMValueRef explicit_lod, /* optional */
                         LLVMValueRef width,
                         LLVMValueRef height,
                         LLVMValueRef depth,
@@ -1622,7 +1637,9 @@ lp_build_sample_general(struct lp_build_sample_context *bld,
       /* Need to compute lod either to choose mipmap levels or to
        * distinguish between minification/magnification with one mipmap level.
        */
-      lod = lp_build_lod_selector(bld, s, t, r, lodbias, width, height, depth);
+      lod = lp_build_lod_selector(bld, s, t, r, ddx, ddy,
+                                  lod_bias, explicit_lod,
+                                  width, height, depth);
    }
 
    /*
@@ -2049,6 +2066,24 @@ lp_build_sample_compare(struct lp_build_sample_context *bld,
 
 
 /**
+ * Just set texels to white instead of actually sampling the texture.
+ * For debugging.
+ */
+static void
+lp_build_sample_nop(struct lp_build_sample_context *bld,
+                    LLVMValueRef *texel)
+{
+   struct lp_build_context *texel_bld = &bld->texel_bld;
+   unsigned chan;
+
+   for (chan = 0; chan < 4; chan++) {
+      /*lp_bld_mov(texel_bld, texel, texel_bld->one);*/
+      texel[chan] = texel_bld->one;
+   }  
+}
+
+
+/**
  * Build texture sampling code.
  * 'texel' will return a vector of four LLVMValueRefs corresponding to
  * R, G, B, A.
@@ -2062,7 +2097,10 @@ lp_build_sample_soa(LLVMBuilderRef builder,
                     unsigned unit,
                     unsigned num_coords,
                     const LLVMValueRef *coords,
-                    LLVMValueRef lodbias,
+                    const LLVMValueRef *ddx,
+                    const LLVMValueRef *ddy,
+                    LLVMValueRef lod_bias, /* optional */
+                    LLVMValueRef explicit_lod, /* optional */
                     LLVMValueRef *texel)
 {
    struct lp_build_sample_context bld;
@@ -2113,19 +2151,24 @@ 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 (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 &&
-       static_state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE &&
-       is_simple_wrap_mode(static_state->wrap_s) &&
-       is_simple_wrap_mode(static_state->wrap_t)) {
+   if (0) {
+      /* For debug: no-op texture sampling */
+      lp_build_sample_nop(&bld, texel);
+   }
+   else 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 &&
+            static_state->min_mip_filter == PIPE_TEX_MIPFILTER_NONE &&
+            is_simple_wrap_mode(static_state->wrap_s) &&
+            is_simple_wrap_mode(static_state->wrap_t)) {
       /* special case */
       lp_build_sample_2d_linear_aos(&bld, s, t, width_vec, height_vec,
                                     row_stride_array, data_array, texel);
    }
    else {
-      lp_build_sample_general(&bld, unit, s, t, r, lodbias,
+      lp_build_sample_general(&bld, unit, s, t, r, ddx, ddy,
+                              lod_bias, explicit_lod,
                               width, height, depth,
                               width_vec, height_vec, depth_vec,
                               row_stride_array, img_stride_array,
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
index 2eac5da6c69..5ce1385c480 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi.h
@@ -65,7 +65,10 @@ struct lp_build_sampler_soa
                         unsigned unit,
                         unsigned num_coords,
                         const LLVMValueRef *coords,
-                        LLVMValueRef lodbias,
+                        const LLVMValueRef *ddx,
+                        const LLVMValueRef *ddy,
+                        LLVMValueRef lod_bias, /* optional */
+                        LLVMValueRef explicit_lod, /* optional */
                         LLVMValueRef *texel);
 };
 
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
index d3c769e28b8..0b1a28a7ab7 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_tgsi_soa.c
@@ -54,13 +54,10 @@
 #include "lp_bld_swizzle.h"
 #include "lp_bld_flow.h"
 #include "lp_bld_tgsi.h"
+#include "lp_bld_limits.h"
 #include "lp_bld_debug.h"
 
 
-#define LP_MAX_TEMPS 256
-#define LP_MAX_IMMEDIATES 256
-
-
 #define FOR_EACH_CHANNEL( CHAN )\
    for (CHAN = 0; CHAN < NUM_CHANNELS; CHAN++)
 
@@ -84,7 +81,6 @@
 #define QUAD_BOTTOM_LEFT  2
 #define QUAD_BOTTOM_RIGHT 3
 
-#define LP_TGSI_MAX_NESTING 16
 
 struct lp_exec_mask {
    struct lp_build_context *bld;
@@ -93,19 +89,19 @@ struct lp_exec_mask {
 
    LLVMTypeRef int_vec_type;
 
-   LLVMValueRef cond_stack[LP_TGSI_MAX_NESTING];
+   LLVMValueRef cond_stack[LP_MAX_TGSI_NESTING];
    int cond_stack_size;
    LLVMValueRef cond_mask;
 
-   LLVMValueRef break_stack[LP_TGSI_MAX_NESTING];
+   LLVMValueRef break_stack[LP_MAX_TGSI_NESTING];
    int break_stack_size;
    LLVMValueRef break_mask;
 
-   LLVMValueRef cont_stack[LP_TGSI_MAX_NESTING];
+   LLVMValueRef cont_stack[LP_MAX_TGSI_NESTING];
    int cont_stack_size;
    LLVMValueRef cont_mask;
 
-   LLVMBasicBlockRef loop_stack[LP_TGSI_MAX_NESTING];
+   LLVMBasicBlockRef loop_stack[LP_MAX_TGSI_NESTING];
    int loop_stack_size;
    LLVMBasicBlockRef loop_block;
 
@@ -124,9 +120,9 @@ struct lp_build_tgsi_soa_context
 
    struct lp_build_sampler_soa *sampler;
 
-   LLVMValueRef immediates[LP_MAX_IMMEDIATES][NUM_CHANNELS];
-   LLVMValueRef temps[LP_MAX_TEMPS][NUM_CHANNELS];
-   LLVMValueRef addr[LP_MAX_TEMPS][NUM_CHANNELS];
+   LLVMValueRef immediates[LP_MAX_TGSI_IMMEDIATES][NUM_CHANNELS];
+   LLVMValueRef temps[LP_MAX_TGSI_TEMPS][NUM_CHANNELS];
+   LLVMValueRef addr[LP_MAX_TGSI_ADDRS][NUM_CHANNELS];
 
    /* we allocate an array of temps if we have indirect
     * addressing and then the temps above is unused */
@@ -198,6 +194,7 @@ static void lp_exec_mask_update(struct lp_exec_mask *mask)
 static void lp_exec_mask_cond_push(struct lp_exec_mask *mask,
                                    LLVMValueRef val)
 {
+   assert(mask->cond_stack_size < LP_MAX_TGSI_NESTING);
    mask->cond_stack[mask->cond_stack_size++] = mask->cond_mask;
    mask->cond_mask = LLVMBuildBitCast(mask->bld->builder, val,
                                       mask->int_vec_type, "");
@@ -239,6 +236,10 @@ static void lp_exec_bgnloop(struct lp_exec_mask *mask)
    if (mask->cond_stack_size == 0)
       mask->cond_mask = LLVMConstAllOnes(mask->int_vec_type);
 
+   assert(mask->break_stack_size < LP_MAX_TGSI_NESTING);
+   assert(mask->cont_stack_size < LP_MAX_TGSI_NESTING);
+   assert(mask->break_stack_size < LP_MAX_TGSI_NESTING);
+
    mask->break_stack[mask->break_stack_size++] = mask->break_mask;
    mask->cont_stack[mask->cont_stack_size++] = mask->cont_mask;
    mask->loop_stack[mask->loop_stack_size++] = mask->loop_block;
@@ -255,16 +256,9 @@ static void lp_exec_break(struct lp_exec_mask *mask)
                                          mask->exec_mask,
                                          "break");
 
-   /* mask->break_stack_size > 1 implies that we encountered a break
-    * statemant already and if that's the case we want to make sure
-    * our mask is a combination of the previous break and the current
-    * execution mask */
-   if (mask->break_stack_size > 1) {
-      mask->break_mask = LLVMBuildAnd(mask->bld->builder,
-                                      mask->break_mask,
-                                      exec_mask, "break_full");
-   } else
-      mask->break_mask = exec_mask;
+   mask->break_mask = LLVMBuildAnd(mask->bld->builder,
+                                   mask->break_mask,
+                                   exec_mask, "break_full");
 
    lp_exec_mask_update(mask);
 }
@@ -275,12 +269,9 @@ static void lp_exec_continue(struct lp_exec_mask *mask)
                                          mask->exec_mask,
                                          "");
 
-   if (mask->cont_stack_size > 1) {
-      mask->cont_mask = LLVMBuildAnd(mask->bld->builder,
-                                     mask->cont_mask,
-                                     exec_mask, "");
-   } else
-      mask->cont_mask = exec_mask;
+   mask->cont_mask = LLVMBuildAnd(mask->bld->builder,
+                                  mask->cont_mask,
+                                  exec_mask, "");
 
    lp_exec_mask_update(mask);
 }
@@ -589,7 +580,6 @@ emit_store(
 
    case TGSI_FILE_PREDICATE:
       /* FIXME */
-      assert(0);
       break;
 
    default:
@@ -602,21 +592,37 @@ emit_store(
  * High-level instruction translators.
  */
 
+enum tex_modifier {
+   TEX_MODIFIER_NONE = 0,
+   TEX_MODIFIER_PROJECTED,
+   TEX_MODIFIER_LOD_BIAS,
+   TEX_MODIFIER_EXPLICIT_LOD,
+   TEX_MODIFIER_EXPLICIT_DERIV
+};
 
 static void
 emit_tex( struct lp_build_tgsi_soa_context *bld,
           const struct tgsi_full_instruction *inst,
-          boolean apply_lodbias,
-          boolean projected,
+          enum tex_modifier modifier,
           LLVMValueRef *texel)
 {
-   const uint unit = inst->Src[1].Register.Index;
-   LLVMValueRef lodbias;
+   unsigned unit;
+   LLVMValueRef lod_bias, explicit_lod;
    LLVMValueRef oow = NULL;
    LLVMValueRef coords[3];
+   LLVMValueRef ddx[3];
+   LLVMValueRef ddy[3];
    unsigned num_coords;
    unsigned i;
 
+   if (!bld->sampler) {
+      _debug_printf("warning: found texture instruction but no sampler generator supplied\n");
+      for (i = 0; i < 4; i++) {
+         texel[i] = bld->base.undef;
+      }
+      return;
+   }
+
    switch (inst->Texture.Texture) {
    case TGSI_TEXTURE_1D:
       num_coords = 1;
@@ -637,29 +643,57 @@ emit_tex( struct lp_build_tgsi_soa_context *bld,
       return;
    }
 
-   if(apply_lodbias)
-      lodbias = emit_fetch( bld, inst, 0, 3 );
-   else
-      lodbias = bld->base.zero;
+   if (modifier == TEX_MODIFIER_LOD_BIAS) {
+      lod_bias = emit_fetch( bld, inst, 0, 3 );
+      explicit_lod = NULL;
+   }
+   else if (modifier == TEX_MODIFIER_EXPLICIT_LOD) {
+      lod_bias = NULL;
+      explicit_lod = emit_fetch( bld, inst, 0, 3 );
+   }
+   else {
+      lod_bias = NULL;
+      explicit_lod = NULL;
+   }
 
-   if (projected) {
+   if (modifier == TEX_MODIFIER_PROJECTED) {
       oow = emit_fetch( bld, inst, 0, 3 );
       oow = lp_build_rcp(&bld->base, oow);
    }
 
    for (i = 0; i < num_coords; i++) {
       coords[i] = emit_fetch( bld, inst, 0, i );
-      if (projected)
+      if (modifier == TEX_MODIFIER_PROJECTED)
          coords[i] = lp_build_mul(&bld->base, coords[i], oow);
    }
    for (i = num_coords; i < 3; i++) {
       coords[i] = bld->base.undef;
    }
 
+   if (modifier == TEX_MODIFIER_EXPLICIT_DERIV) {
+      for (i = 0; i < num_coords; i++) {
+         ddx[i] = emit_fetch( bld, inst, 1, i );
+         ddy[i] = emit_fetch( bld, inst, 2, i );
+      }
+      unit = inst->Src[3].Register.Index;
+   }  else {
+      for (i = 0; i < num_coords; i++) {
+         ddx[i] = emit_ddx( bld, coords[i] );
+         ddy[i] = emit_ddy( bld, coords[i] );
+      }
+      unit = inst->Src[1].Register.Index;
+   }
+   for (i = num_coords; i < 3; i++) {
+      ddx[i] = bld->base.undef;
+      ddy[i] = bld->base.undef;
+   }
+
    bld->sampler->emit_fetch_texel(bld->sampler,
                                   bld->base.builder,
                                   bld->base.type,
-                                  unit, num_coords, coords, lodbias,
+                                  unit, num_coords, coords,
+                                  ddx, ddy,
+                                  lod_bias, explicit_lod,
                                   texel);
 }
 
@@ -739,7 +773,7 @@ emit_kilp(struct lp_build_tgsi_soa_context *bld,
    lp_build_mask_update(bld->mask, mask);
 }
 
-static int
+static void
 emit_declaration(
    struct lp_build_tgsi_soa_context *bld,
    const struct tgsi_full_declaration *decl)
@@ -753,6 +787,7 @@ emit_declaration(
    for (idx = first; idx <= last; ++idx) {
       switch (decl->Declaration.File) {
       case TGSI_FILE_TEMPORARY:
+         assert(idx < LP_MAX_TGSI_TEMPS);
          if (bld->has_indirect_addressing) {
             LLVMValueRef val = LLVMConstInt(LLVMInt32Type(),
                                             last*4 + 4, 0);
@@ -772,18 +807,21 @@ emit_declaration(
          break;
 
       case TGSI_FILE_ADDRESS:
+         assert(idx < LP_MAX_TGSI_ADDRS);
          for (i = 0; i < NUM_CHANNELS; i++)
             bld->addr[idx][i] = lp_build_alloca(bld->base.builder,
                                                 vec_type, "");
          break;
 
+      case TGSI_FILE_PREDICATE:
+         _debug_printf("warning: predicate registers not yet implemented\n");
+         break;
+
       default:
          /* don't need to declare other vars */
          break;
       }
    }
-
-   return TRUE;
 }
 
 
@@ -1359,12 +1397,11 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TEX:
-      emit_tex( bld, inst, FALSE, FALSE, dst0 );
+      emit_tex( bld, inst, TEX_MODIFIER_NONE, dst0 );
       break;
 
    case TGSI_OPCODE_TXD:
-      /* FIXME */
-      return FALSE;
+      emit_tex( bld, inst, TEX_MODIFIER_EXPLICIT_DERIV, dst0 );
       break;
 
    case TGSI_OPCODE_UP2H:
@@ -1466,7 +1503,7 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TXB:
-      emit_tex( bld, inst, TRUE, FALSE, dst0 );
+      emit_tex( bld, inst, TEX_MODIFIER_LOD_BIAS, dst0 );
       break;
 
    case TGSI_OPCODE_NRM:
@@ -1571,11 +1608,11 @@ emit_instruction(
       break;
 
    case TGSI_OPCODE_TXL:
-      emit_tex( bld, inst, TRUE, FALSE, dst0 );
+      emit_tex( bld, inst, TEX_MODIFIER_EXPLICIT_LOD, dst0 );
       break;
 
    case TGSI_OPCODE_TXP:
-      emit_tex( bld, inst, FALSE, TRUE, dst0 );
+      emit_tex( bld, inst, TEX_MODIFIER_PROJECTED, dst0 );
       break;
 
    case TGSI_OPCODE_BRK:
@@ -1765,10 +1802,7 @@ lp_build_tgsi_soa(LLVMBuilderRef builder,
       switch( parse.FullToken.Token.Type ) {
       case TGSI_TOKEN_TYPE_DECLARATION:
          /* Inputs already interpolated */
-         {
-            if (!emit_declaration( &bld, &parse.FullToken.FullDeclaration ))
-               _debug_printf("warning: failed to define LLVM variable\n");
-         }
+         emit_declaration( &bld, &parse.FullToken.FullDeclaration );
          break;
 
       case TGSI_TOKEN_TYPE_INSTRUCTION:
@@ -1787,7 +1821,7 @@ lp_build_tgsi_soa(LLVMBuilderRef builder,
          {
             const uint size = parse.FullToken.FullImmediate.Immediate.NrTokens - 1;
             assert(size <= 4);
-            assert(num_immediates < LP_MAX_IMMEDIATES);
+            assert(num_immediates < LP_MAX_TGSI_IMMEDIATES);
             for( i = 0; i < size; ++i )
                bld.immediates[num_immediates][i] =
                   lp_build_const_vec(type, parse.FullToken.FullImmediate.u[i].Float);
diff --git a/src/gallium/auxiliary/tgsi/tgsi_dump.c b/src/gallium/auxiliary/tgsi/tgsi_dump.c
index 83000200189..1ad183b8dd4 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_dump.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_dump.c
@@ -100,7 +100,6 @@ static const char *file_names[TGSI_FILE_COUNT] =
    "SAMP",
    "ADDR",
    "IMM",
-   "LOOP",
    "PRED",
    "SV"
 };
@@ -494,6 +493,30 @@ iter_instruction(
       TXT( "  " );
    ctx->indent += info->post_indent;
    
+   if (inst->Instruction.Predicate) {
+      CHR( '(' );
+
+      if (inst->Predicate.Negate)
+         CHR( '!' );
+
+      TXT( "PRED[" );
+      SID( inst->Predicate.Index );
+      CHR( ']' );
+
+      if (inst->Predicate.SwizzleX != TGSI_SWIZZLE_X ||
+          inst->Predicate.SwizzleY != TGSI_SWIZZLE_Y ||
+          inst->Predicate.SwizzleZ != TGSI_SWIZZLE_Z ||
+          inst->Predicate.SwizzleW != TGSI_SWIZZLE_W) {
+         CHR( '.' );
+         ENM( inst->Predicate.SwizzleX, swizzle_names );
+         ENM( inst->Predicate.SwizzleY, swizzle_names );
+         ENM( inst->Predicate.SwizzleZ, swizzle_names );
+         ENM( inst->Predicate.SwizzleW, swizzle_names );
+      }
+
+      TXT( ") " );
+   }
+
    TXT( info->mnemonic );
 
    switch (inst->Instruction.Saturate) {
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.c b/src/gallium/auxiliary/tgsi/tgsi_exec.c
index 82eac05dc4d..9b1ca7fa851 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.c
@@ -1346,13 +1346,6 @@ store_dest(struct tgsi_exec_machine *mach,
       dst = &mach->Addrs[index].xyzw[chan_index];
       break;
 
-   case TGSI_FILE_LOOP:
-      assert(reg->Register.Index == 0);
-      assert(mach->LoopCounterStackTop > 0);
-      assert(chan_index == CHAN_X);
-      dst = &mach->LoopCounterStack[mach->LoopCounterStackTop - 1].xyzw[chan_index];
-      break;
-
    case TGSI_FILE_PREDICATE:
       index = reg->Register.Index;
       assert(index < TGSI_EXEC_NUM_PREDS);
diff --git a/src/gallium/auxiliary/tgsi/tgsi_exec.h b/src/gallium/auxiliary/tgsi/tgsi_exec.h
index a22873e4c2b..11bbaf6722e 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_exec.h
@@ -299,10 +299,6 @@ struct tgsi_exec_machine
    uint LoopLabelStack[TGSI_EXEC_MAX_LOOP_NESTING];
    int LoopLabelStackTop;
 
-   /** Loop counter stack (x = index, y = counter, z = step) */
-   struct tgsi_exec_vector LoopCounterStack[TGSI_EXEC_MAX_LOOP_NESTING];
-   int LoopCounterStackTop;
-   
    /** Loop continue mask stack (see comments in tgsi_exec.c) */
    uint ContStack[TGSI_EXEC_MAX_LOOP_NESTING];
    int ContStackTop;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_sanity.c b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
index 76b7564cc36..ce0a92f7fb3 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_sanity.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_sanity.c
@@ -235,8 +235,8 @@ static const char *file_names[TGSI_FILE_COUNT] =
    "SAMP",
    "ADDR",
    "IMM",
-   "LOOP",
-   "PRED"
+   "PRED",
+   "SV"
 };
 
 static boolean
diff --git a/src/gallium/auxiliary/tgsi/tgsi_text.c b/src/gallium/auxiliary/tgsi/tgsi_text.c
index 0b468a9184e..71a0d96331a 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_text.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_text.c
@@ -278,7 +278,6 @@ static const char *file_names[TGSI_FILE_COUNT] =
    "SAMP",
    "ADDR",
    "IMM",
-   "LOOP",
    "PRED",
    "SV"
 };
@@ -816,6 +815,45 @@ parse_instruction(
    struct tgsi_full_instruction inst;
    uint advance;
 
+   inst = tgsi_default_full_instruction();
+
+   /* Parse predicate.
+    */
+   eat_opt_white( &ctx->cur );
+   if (*ctx->cur == '(') {
+      uint file;
+      int index;
+      uint swizzle[4];
+      boolean parsed_swizzle;
+
+      inst.Instruction.Predicate = 1;
+
+      ctx->cur++;
+      if (*ctx->cur == '!') {
+         ctx->cur++;
+         inst.Predicate.Negate = 1;
+      }
+
+      if (!parse_register_dst( ctx, &file, &index ))
+         return FALSE;
+
+      if (parse_optional_swizzle( ctx, swizzle, &parsed_swizzle )) {
+         if (parsed_swizzle) {
+            inst.Predicate.SwizzleX = swizzle[0];
+            inst.Predicate.SwizzleY = swizzle[1];
+            inst.Predicate.SwizzleZ = swizzle[2];
+            inst.Predicate.SwizzleW = swizzle[3];
+         }
+      }
+
+      if (*ctx->cur != ')') {
+         report_error( ctx, "Expected `)'" );
+         return FALSE;
+      }
+
+      ctx->cur++;
+   }
+
    /* Parse instruction name.
     */
    eat_opt_white( &ctx->cur );
@@ -849,7 +887,6 @@ parse_instruction(
       return FALSE;
    }
 
-   inst = tgsi_default_full_instruction();
    inst.Instruction.Opcode = i;
    inst.Instruction.Saturate = saturate;
    inst.Instruction.NumDstRegs = info->num_dst;
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.c b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
index f725405ade1..49b854b1234 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ureg.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.c
@@ -74,7 +74,6 @@ struct ureg_tokens {
 #define UREG_MAX_IMMEDIATE 32
 #define UREG_MAX_TEMP 256
 #define UREG_MAX_ADDR 2
-#define UREG_MAX_LOOP 1
 #define UREG_MAX_PRED 1
 
 struct const_decl {
@@ -151,7 +150,6 @@ struct ureg_program
 
    unsigned nr_addrs;
    unsigned nr_preds;
-   unsigned nr_loops;
    unsigned nr_instructions;
 
    struct ureg_tokens domain[2];
@@ -537,19 +535,6 @@ struct ureg_dst ureg_DECL_address( struct ureg_program *ureg )
    return ureg_dst_register( TGSI_FILE_ADDRESS, 0 );
 }
 
-/* Allocate a new loop register.
- */
-struct ureg_dst
-ureg_DECL_loop(struct ureg_program *ureg)
-{
-   if (ureg->nr_loops < UREG_MAX_LOOP) {
-      return ureg_dst_register(TGSI_FILE_LOOP, ureg->nr_loops++);
-   }
-
-   assert(0);
-   return ureg_dst_register(TGSI_FILE_LOOP, 0);
-}
-
 /* Allocate a new predicate register.
  */
 struct ureg_dst
@@ -1356,13 +1341,6 @@ static void emit_decls( struct ureg_program *ureg )
                        0, ureg->nr_addrs );
    }
 
-   if (ureg->nr_loops) {
-      emit_decl_range(ureg,
-                      TGSI_FILE_LOOP,
-                      0,
-                      ureg->nr_loops);
-   }
-
    if (ureg->nr_preds) {
       emit_decl_range(ureg,
                       TGSI_FILE_PREDICATE,
diff --git a/src/gallium/auxiliary/tgsi/tgsi_ureg.h b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
index 0130a77aadb..f32420dd872 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_ureg.h
+++ b/src/gallium/auxiliary/tgsi/tgsi_ureg.h
@@ -231,9 +231,6 @@ struct ureg_dst
 ureg_DECL_address( struct ureg_program * );
 
 struct ureg_dst
-ureg_DECL_loop( struct ureg_program * );
-
-struct ureg_dst
 ureg_DECL_predicate(struct ureg_program *);
 
 /* Supply an index to the sampler declaration as this is the hook to
diff --git a/src/gallium/auxiliary/translate/translate_generic.c b/src/gallium/auxiliary/translate/translate_generic.c
index a9272fbb491..f8dbd2b36a1 100644
--- a/src/gallium/auxiliary/translate/translate_generic.c
+++ b/src/gallium/auxiliary/translate/translate_generic.c
@@ -31,6 +31,7 @@
   */
 
 #include "util/u_memory.h"
+#include "util/u_format.h"
 #include "util/u_math.h"
 #include "pipe/p_state.h"
 #include "translate.h"
@@ -38,7 +39,9 @@
 
 #define DRAW_DBG 0
 
-typedef void (*fetch_func)(const void *ptr, float *attrib);
+typedef void (*fetch_func)(float *dst,
+                           const uint8_t *src,
+                           unsigned i, unsigned j);
 typedef void (*emit_func)(const float *attrib, void *ptr);
 
 
@@ -57,7 +60,7 @@ struct translate_generic {
       emit_func emit;
       unsigned output_offset;
       
-      char *input_ptr;
+      const uint8_t *input_ptr;
       unsigned input_stride;
       unsigned max_index;
 
@@ -79,22 +82,7 @@ static struct translate_generic *translate_generic( struct translate *translate
  * This is probably needed/dupliocated elsewhere, eg format
  * conversion, texture sampling etc.
  */
-#define ATTRIB( NAME, SZ, TYPE, FROM, TO )		\
-static void						\
-fetch_##NAME(const void *ptr, float *attrib)		\
-{							\
-   const float defaults[4] = { 0.0f,0.0f,0.0f,1.0f };	\
-   unsigned i;						\
-							\
-   for (i = 0; i < SZ; i++) {				\
-      attrib[i] = FROM(i);				\
-   }							\
-							\
-   for (; i < 4; i++) {					\
-      attrib[i] = defaults[i];				\
-   }							\
-}							\
-							\
+#define ATTRIB( NAME, SZ, TYPE, TO )	    	        \
 static void						\
 emit_##NAME(const float *attrib, void *ptr)		\
 {  \
@@ -107,27 +95,6 @@ emit_##NAME(const float *attrib, void *ptr)		\
 }
 
 
-#define FROM_64_FLOAT(i)   ((float) ((double *) ptr)[i])
-#define FROM_32_FLOAT(i)   (((float *) ptr)[i])
-
-#define FROM_8_USCALED(i)  ((float) ((unsigned char *) ptr)[i])
-#define FROM_16_USCALED(i) ((float) ((unsigned short *) ptr)[i])
-#define FROM_32_USCALED(i) ((float) ((unsigned int *) ptr)[i])
-
-#define FROM_8_SSCALED(i)  ((float) ((char *) ptr)[i])
-#define FROM_16_SSCALED(i) ((float) ((short *) ptr)[i])
-#define FROM_32_SSCALED(i) ((float) ((int *) ptr)[i])
-
-#define FROM_8_UNORM(i)    ((float) ((unsigned char *) ptr)[i] / 255.0f)
-#define FROM_16_UNORM(i)   ((float) ((unsigned short *) ptr)[i] / 65535.0f)
-#define FROM_32_UNORM(i)   ((float) ((unsigned int *) ptr)[i] / 4294967295.0f)
-
-#define FROM_8_SNORM(i)    ((float) ((char *) ptr)[i] / 127.0f)
-#define FROM_16_SNORM(i)   ((float) ((short *) ptr)[i] / 32767.0f)
-#define FROM_32_SNORM(i)   ((float) ((int *) ptr)[i] / 2147483647.0f)
-
-#define FROM_32_FIXED(i)   (((int *) ptr)[i] / 65536.0f)
-
 #define TO_64_FLOAT(x)   ((double) x)
 #define TO_32_FLOAT(x)   (x)
 
@@ -150,95 +117,79 @@ emit_##NAME(const float *attrib, void *ptr)		\
 #define TO_32_FIXED(x)   ((int) (x * 65536.0f))
 
 
+ATTRIB( R64G64B64A64_FLOAT,   4, double, TO_64_FLOAT )
+ATTRIB( R64G64B64_FLOAT,      3, double, TO_64_FLOAT )
+ATTRIB( R64G64_FLOAT,         2, double, TO_64_FLOAT )
+ATTRIB( R64_FLOAT,            1, double, TO_64_FLOAT )
+
+ATTRIB( R32G32B32A32_FLOAT,   4, float, TO_32_FLOAT )
+ATTRIB( R32G32B32_FLOAT,      3, float, TO_32_FLOAT )
+ATTRIB( R32G32_FLOAT,         2, float, TO_32_FLOAT )
+ATTRIB( R32_FLOAT,            1, float, TO_32_FLOAT )
+
+ATTRIB( R32G32B32A32_USCALED, 4, unsigned, TO_32_USCALED )
+ATTRIB( R32G32B32_USCALED,    3, unsigned, TO_32_USCALED )
+ATTRIB( R32G32_USCALED,       2, unsigned, TO_32_USCALED )
+ATTRIB( R32_USCALED,          1, unsigned, TO_32_USCALED )
+
+ATTRIB( R32G32B32A32_SSCALED, 4, int, TO_32_SSCALED )
+ATTRIB( R32G32B32_SSCALED,    3, int, TO_32_SSCALED )
+ATTRIB( R32G32_SSCALED,       2, int, TO_32_SSCALED )
+ATTRIB( R32_SSCALED,          1, int, TO_32_SSCALED )
+
+ATTRIB( R32G32B32A32_UNORM, 4, unsigned, TO_32_UNORM )
+ATTRIB( R32G32B32_UNORM,    3, unsigned, TO_32_UNORM )
+ATTRIB( R32G32_UNORM,       2, unsigned, TO_32_UNORM )
+ATTRIB( R32_UNORM,          1, unsigned, TO_32_UNORM )
+
+ATTRIB( R32G32B32A32_SNORM, 4, int, TO_32_SNORM )
+ATTRIB( R32G32B32_SNORM,    3, int, TO_32_SNORM )
+ATTRIB( R32G32_SNORM,       2, int, TO_32_SNORM )
+ATTRIB( R32_SNORM,          1, int, TO_32_SNORM )
+
+ATTRIB( R16G16B16A16_USCALED, 4, ushort, TO_16_USCALED )
+ATTRIB( R16G16B16_USCALED,    3, ushort, TO_16_USCALED )
+ATTRIB( R16G16_USCALED,       2, ushort, TO_16_USCALED )
+ATTRIB( R16_USCALED,          1, ushort, TO_16_USCALED )
+
+ATTRIB( R16G16B16A16_SSCALED, 4, short, TO_16_SSCALED )
+ATTRIB( R16G16B16_SSCALED,    3, short, TO_16_SSCALED )
+ATTRIB( R16G16_SSCALED,       2, short, TO_16_SSCALED )
+ATTRIB( R16_SSCALED,          1, short, TO_16_SSCALED )
+
+ATTRIB( R16G16B16A16_UNORM, 4, ushort, TO_16_UNORM )
+ATTRIB( R16G16B16_UNORM,    3, ushort, TO_16_UNORM )
+ATTRIB( R16G16_UNORM,       2, ushort, TO_16_UNORM )
+ATTRIB( R16_UNORM,          1, ushort, TO_16_UNORM )
+
+ATTRIB( R16G16B16A16_SNORM, 4, short, TO_16_SNORM )
+ATTRIB( R16G16B16_SNORM,    3, short, TO_16_SNORM )
+ATTRIB( R16G16_SNORM,       2, short, TO_16_SNORM )
+ATTRIB( R16_SNORM,          1, short, TO_16_SNORM )
+
+ATTRIB( R8G8B8A8_USCALED,   4, ubyte, TO_8_USCALED )
+ATTRIB( R8G8B8_USCALED,     3, ubyte, TO_8_USCALED )
+ATTRIB( R8G8_USCALED,       2, ubyte, TO_8_USCALED )
+ATTRIB( R8_USCALED,         1, ubyte, TO_8_USCALED )
+
+ATTRIB( R8G8B8A8_SSCALED,  4, char, TO_8_SSCALED )
+ATTRIB( R8G8B8_SSCALED,    3, char, TO_8_SSCALED )
+ATTRIB( R8G8_SSCALED,      2, char, TO_8_SSCALED )
+ATTRIB( R8_SSCALED,        1, char, TO_8_SSCALED )
+
+ATTRIB( R8G8B8A8_UNORM,  4, ubyte, TO_8_UNORM )
+ATTRIB( R8G8B8_UNORM,    3, ubyte, TO_8_UNORM )
+ATTRIB( R8G8_UNORM,      2, ubyte, TO_8_UNORM )
+ATTRIB( R8_UNORM,        1, ubyte, TO_8_UNORM )
+
+ATTRIB( R8G8B8A8_SNORM,  4, char, TO_8_SNORM )
+ATTRIB( R8G8B8_SNORM,    3, char, TO_8_SNORM )
+ATTRIB( R8G8_SNORM,      2, char, TO_8_SNORM )
+ATTRIB( R8_SNORM,        1, char, TO_8_SNORM )
+
+ATTRIB( A8R8G8B8_UNORM,       4, ubyte, TO_8_UNORM )
+/*ATTRIB( R8G8B8A8_UNORM,       4, ubyte, TO_8_UNORM )*/
 
-ATTRIB( R64G64B64A64_FLOAT,   4, double, FROM_64_FLOAT, TO_64_FLOAT )
-ATTRIB( R64G64B64_FLOAT,      3, double, FROM_64_FLOAT, TO_64_FLOAT )
-ATTRIB( R64G64_FLOAT,         2, double, FROM_64_FLOAT, TO_64_FLOAT )
-ATTRIB( R64_FLOAT,            1, double, FROM_64_FLOAT, TO_64_FLOAT )
-
-ATTRIB( R32G32B32A32_FLOAT,   4, float, FROM_32_FLOAT, TO_32_FLOAT )
-ATTRIB( R32G32B32_FLOAT,      3, float, FROM_32_FLOAT, TO_32_FLOAT )
-ATTRIB( R32G32_FLOAT,         2, float, FROM_32_FLOAT, TO_32_FLOAT )
-ATTRIB( R32_FLOAT,            1, float, FROM_32_FLOAT, TO_32_FLOAT )
-
-ATTRIB( R32G32B32A32_USCALED, 4, unsigned, FROM_32_USCALED, TO_32_USCALED )
-ATTRIB( R32G32B32_USCALED,    3, unsigned, FROM_32_USCALED, TO_32_USCALED )
-ATTRIB( R32G32_USCALED,       2, unsigned, FROM_32_USCALED, TO_32_USCALED )
-ATTRIB( R32_USCALED,          1, unsigned, FROM_32_USCALED, TO_32_USCALED )
-
-ATTRIB( R32G32B32A32_SSCALED, 4, int, FROM_32_SSCALED, TO_32_SSCALED )
-ATTRIB( R32G32B32_SSCALED,    3, int, FROM_32_SSCALED, TO_32_SSCALED )
-ATTRIB( R32G32_SSCALED,       2, int, FROM_32_SSCALED, TO_32_SSCALED )
-ATTRIB( R32_SSCALED,          1, int, FROM_32_SSCALED, TO_32_SSCALED )
-
-ATTRIB( R32G32B32A32_UNORM, 4, unsigned, FROM_32_UNORM, TO_32_UNORM )
-ATTRIB( R32G32B32_UNORM,    3, unsigned, FROM_32_UNORM, TO_32_UNORM )
-ATTRIB( R32G32_UNORM,       2, unsigned, FROM_32_UNORM, TO_32_UNORM )
-ATTRIB( R32_UNORM,          1, unsigned, FROM_32_UNORM, TO_32_UNORM )
-
-ATTRIB( R32G32B32A32_SNORM, 4, int, FROM_32_SNORM, TO_32_SNORM )
-ATTRIB( R32G32B32_SNORM,    3, int, FROM_32_SNORM, TO_32_SNORM )
-ATTRIB( R32G32_SNORM,       2, int, FROM_32_SNORM, TO_32_SNORM )
-ATTRIB( R32_SNORM,          1, int, FROM_32_SNORM, TO_32_SNORM )
-
-ATTRIB( R16G16B16A16_USCALED, 4, ushort, FROM_16_USCALED, TO_16_USCALED )
-ATTRIB( R16G16B16_USCALED,    3, ushort, FROM_16_USCALED, TO_16_USCALED )
-ATTRIB( R16G16_USCALED,       2, ushort, FROM_16_USCALED, TO_16_USCALED )
-ATTRIB( R16_USCALED,          1, ushort, FROM_16_USCALED, TO_16_USCALED )
-
-ATTRIB( R16G16B16A16_SSCALED, 4, short, FROM_16_SSCALED, TO_16_SSCALED )
-ATTRIB( R16G16B16_SSCALED,    3, short, FROM_16_SSCALED, TO_16_SSCALED )
-ATTRIB( R16G16_SSCALED,       2, short, FROM_16_SSCALED, TO_16_SSCALED )
-ATTRIB( R16_SSCALED,          1, short, FROM_16_SSCALED, TO_16_SSCALED )
-
-ATTRIB( R16G16B16A16_UNORM, 4, ushort, FROM_16_UNORM, TO_16_UNORM )
-ATTRIB( R16G16B16_UNORM,    3, ushort, FROM_16_UNORM, TO_16_UNORM )
-ATTRIB( R16G16_UNORM,       2, ushort, FROM_16_UNORM, TO_16_UNORM )
-ATTRIB( R16_UNORM,          1, ushort, FROM_16_UNORM, TO_16_UNORM )
-
-ATTRIB( R16G16B16A16_SNORM, 4, short, FROM_16_SNORM, TO_16_SNORM )
-ATTRIB( R16G16B16_SNORM,    3, short, FROM_16_SNORM, TO_16_SNORM )
-ATTRIB( R16G16_SNORM,       2, short, FROM_16_SNORM, TO_16_SNORM )
-ATTRIB( R16_SNORM,          1, short, FROM_16_SNORM, TO_16_SNORM )
-
-ATTRIB( R8G8B8A8_USCALED,   4, ubyte, FROM_8_USCALED, TO_8_USCALED )
-ATTRIB( R8G8B8_USCALED,     3, ubyte, FROM_8_USCALED, TO_8_USCALED )
-ATTRIB( R8G8_USCALED,       2, ubyte, FROM_8_USCALED, TO_8_USCALED )
-ATTRIB( R8_USCALED,         1, ubyte, FROM_8_USCALED, TO_8_USCALED )
-
-ATTRIB( R8G8B8A8_SSCALED,  4, char, FROM_8_SSCALED, TO_8_SSCALED )
-ATTRIB( R8G8B8_SSCALED,    3, char, FROM_8_SSCALED, TO_8_SSCALED )
-ATTRIB( R8G8_SSCALED,      2, char, FROM_8_SSCALED, TO_8_SSCALED )
-ATTRIB( R8_SSCALED,        1, char, FROM_8_SSCALED, TO_8_SSCALED )
-
-ATTRIB( R8G8B8A8_UNORM,  4, ubyte, FROM_8_UNORM, TO_8_UNORM )
-ATTRIB( R8G8B8_UNORM,    3, ubyte, FROM_8_UNORM, TO_8_UNORM )
-ATTRIB( R8G8_UNORM,      2, ubyte, FROM_8_UNORM, TO_8_UNORM )
-ATTRIB( R8_UNORM,        1, ubyte, FROM_8_UNORM, TO_8_UNORM )
-
-ATTRIB( R8G8B8A8_SNORM,  4, char, FROM_8_SNORM, TO_8_SNORM )
-ATTRIB( R8G8B8_SNORM,    3, char, FROM_8_SNORM, TO_8_SNORM )
-ATTRIB( R8G8_SNORM,      2, char, FROM_8_SNORM, TO_8_SNORM )
-ATTRIB( R8_SNORM,        1, char, FROM_8_SNORM, TO_8_SNORM )
-
-ATTRIB( A8R8G8B8_UNORM,       4, ubyte, FROM_8_UNORM, TO_8_UNORM )
-/*ATTRIB( R8G8B8A8_UNORM,       4, ubyte, FROM_8_UNORM, TO_8_UNORM )*/
-
-ATTRIB( R32G32B32A32_FIXED,   4, int, FROM_32_FIXED, TO_32_FIXED )
-ATTRIB( R32G32B32_FIXED,      3, int, FROM_32_FIXED, TO_32_FIXED )
-ATTRIB( R32G32_FIXED,         2, int, FROM_32_FIXED, TO_32_FIXED )
-ATTRIB( R32_FIXED,            1, int, FROM_32_FIXED, TO_32_FIXED )
-
-
-
-static void
-fetch_B8G8R8A8_UNORM(const void *ptr, float *attrib)
-{
-   attrib[2] = FROM_8_UNORM(0);
-   attrib[1] = FROM_8_UNORM(1);
-   attrib[0] = FROM_8_UNORM(2);
-   attrib[3] = FROM_8_UNORM(3);
-}
 
 static void
 emit_B8G8R8A8_UNORM( const float *attrib, void *ptr)
@@ -251,181 +202,13 @@ emit_B8G8R8A8_UNORM( const float *attrib, void *ptr)
 }
 
 static void 
-fetch_NULL( const void *ptr, float *attrib )
-{
-   attrib[0] = 0;
-   attrib[1] = 0;
-   attrib[2] = 0;
-   attrib[3] = 1;
-}
-
-static void 
 emit_NULL( const float *attrib, void *ptr )
 {
    /* do nothing is the only sensible option */
 }
 
-static fetch_func get_fetch_func( enum pipe_format format )
-{
-   switch (format) {
-   case PIPE_FORMAT_R64_FLOAT:
-      return &fetch_R64_FLOAT;
-   case PIPE_FORMAT_R64G64_FLOAT:
-      return &fetch_R64G64_FLOAT;
-   case PIPE_FORMAT_R64G64B64_FLOAT:
-      return &fetch_R64G64B64_FLOAT;
-   case PIPE_FORMAT_R64G64B64A64_FLOAT:
-      return &fetch_R64G64B64A64_FLOAT;
-
-   case PIPE_FORMAT_R32_FLOAT:
-      return &fetch_R32_FLOAT;
-   case PIPE_FORMAT_R32G32_FLOAT:
-      return &fetch_R32G32_FLOAT;
-   case PIPE_FORMAT_R32G32B32_FLOAT:
-      return &fetch_R32G32B32_FLOAT;
-   case PIPE_FORMAT_R32G32B32A32_FLOAT:
-      return &fetch_R32G32B32A32_FLOAT;
-
-   case PIPE_FORMAT_R32_UNORM:
-      return &fetch_R32_UNORM;
-   case PIPE_FORMAT_R32G32_UNORM:
-      return &fetch_R32G32_UNORM;
-   case PIPE_FORMAT_R32G32B32_UNORM:
-      return &fetch_R32G32B32_UNORM;
-   case PIPE_FORMAT_R32G32B32A32_UNORM:
-      return &fetch_R32G32B32A32_UNORM;
-
-   case PIPE_FORMAT_R32_USCALED:
-      return &fetch_R32_USCALED;
-   case PIPE_FORMAT_R32G32_USCALED:
-      return &fetch_R32G32_USCALED;
-   case PIPE_FORMAT_R32G32B32_USCALED:
-      return &fetch_R32G32B32_USCALED;
-   case PIPE_FORMAT_R32G32B32A32_USCALED:
-      return &fetch_R32G32B32A32_USCALED;
-
-   case PIPE_FORMAT_R32_SNORM:
-      return &fetch_R32_SNORM;
-   case PIPE_FORMAT_R32G32_SNORM:
-      return &fetch_R32G32_SNORM;
-   case PIPE_FORMAT_R32G32B32_SNORM:
-      return &fetch_R32G32B32_SNORM;
-   case PIPE_FORMAT_R32G32B32A32_SNORM:
-      return &fetch_R32G32B32A32_SNORM;
-
-   case PIPE_FORMAT_R32_SSCALED:
-      return &fetch_R32_SSCALED;
-   case PIPE_FORMAT_R32G32_SSCALED:
-      return &fetch_R32G32_SSCALED;
-   case PIPE_FORMAT_R32G32B32_SSCALED:
-      return &fetch_R32G32B32_SSCALED;
-   case PIPE_FORMAT_R32G32B32A32_SSCALED:
-      return &fetch_R32G32B32A32_SSCALED;
-
-   case PIPE_FORMAT_R16_UNORM:
-      return &fetch_R16_UNORM;
-   case PIPE_FORMAT_R16G16_UNORM:
-      return &fetch_R16G16_UNORM;
-   case PIPE_FORMAT_R16G16B16_UNORM:
-      return &fetch_R16G16B16_UNORM;
-   case PIPE_FORMAT_R16G16B16A16_UNORM:
-      return &fetch_R16G16B16A16_UNORM;
-
-   case PIPE_FORMAT_R16_USCALED:
-      return &fetch_R16_USCALED;
-   case PIPE_FORMAT_R16G16_USCALED:
-      return &fetch_R16G16_USCALED;
-   case PIPE_FORMAT_R16G16B16_USCALED:
-      return &fetch_R16G16B16_USCALED;
-   case PIPE_FORMAT_R16G16B16A16_USCALED:
-      return &fetch_R16G16B16A16_USCALED;
-
-   case PIPE_FORMAT_R16_SNORM:
-      return &fetch_R16_SNORM;
-   case PIPE_FORMAT_R16G16_SNORM:
-      return &fetch_R16G16_SNORM;
-   case PIPE_FORMAT_R16G16B16_SNORM:
-      return &fetch_R16G16B16_SNORM;
-   case PIPE_FORMAT_R16G16B16A16_SNORM:
-      return &fetch_R16G16B16A16_SNORM;
-
-   case PIPE_FORMAT_R16_SSCALED:
-      return &fetch_R16_SSCALED;
-   case PIPE_FORMAT_R16G16_SSCALED:
-      return &fetch_R16G16_SSCALED;
-   case PIPE_FORMAT_R16G16B16_SSCALED:
-      return &fetch_R16G16B16_SSCALED;
-   case PIPE_FORMAT_R16G16B16A16_SSCALED:
-      return &fetch_R16G16B16A16_SSCALED;
-
-   case PIPE_FORMAT_R8_UNORM:
-      return &fetch_R8_UNORM;
-   case PIPE_FORMAT_R8G8_UNORM:
-      return &fetch_R8G8_UNORM;
-   case PIPE_FORMAT_R8G8B8_UNORM:
-      return &fetch_R8G8B8_UNORM;
-   case PIPE_FORMAT_R8G8B8A8_UNORM:
-      return &fetch_R8G8B8A8_UNORM;
-
-   case PIPE_FORMAT_R8_USCALED:
-      return &fetch_R8_USCALED;
-   case PIPE_FORMAT_R8G8_USCALED:
-      return &fetch_R8G8_USCALED;
-   case PIPE_FORMAT_R8G8B8_USCALED:
-      return &fetch_R8G8B8_USCALED;
-   case PIPE_FORMAT_R8G8B8A8_USCALED:
-      return &fetch_R8G8B8A8_USCALED;
-
-   case PIPE_FORMAT_R8_SNORM:
-      return &fetch_R8_SNORM;
-   case PIPE_FORMAT_R8G8_SNORM:
-      return &fetch_R8G8_SNORM;
-   case PIPE_FORMAT_R8G8B8_SNORM:
-      return &fetch_R8G8B8_SNORM;
-   case PIPE_FORMAT_R8G8B8A8_SNORM:
-      return &fetch_R8G8B8A8_SNORM;
-
-   case PIPE_FORMAT_R8_SSCALED:
-      return &fetch_R8_SSCALED;
-   case PIPE_FORMAT_R8G8_SSCALED:
-      return &fetch_R8G8_SSCALED;
-   case PIPE_FORMAT_R8G8B8_SSCALED:
-      return &fetch_R8G8B8_SSCALED;
-   case PIPE_FORMAT_R8G8B8A8_SSCALED:
-      return &fetch_R8G8B8A8_SSCALED;
-
-   case PIPE_FORMAT_B8G8R8A8_UNORM:
-      return &fetch_B8G8R8A8_UNORM;
-
-   case PIPE_FORMAT_A8R8G8B8_UNORM:
-      return &fetch_A8R8G8B8_UNORM;
-
-   case PIPE_FORMAT_R32_FIXED:
-      return &fetch_R32_FIXED;
-   case PIPE_FORMAT_R32G32_FIXED:
-      return &fetch_R32G32_FIXED;
-   case PIPE_FORMAT_R32G32B32_FIXED:
-      return &fetch_R32G32B32_FIXED;
-   case PIPE_FORMAT_R32G32B32A32_FIXED:
-      return &fetch_R32G32B32A32_FIXED;
-
-   default:
-      assert(0); 
-      return &fetch_NULL;
-   }
-}
-
-
-
-
 static emit_func get_emit_func( enum pipe_format format )
 {
-   /* silence warnings */
-   (void) emit_R32G32B32A32_FIXED;
-   (void) emit_R32G32B32_FIXED;
-   (void) emit_R32G32_FIXED;
-   (void) emit_R32_FIXED;
-
    switch (format) {
    case PIPE_FORMAT_R64_FLOAT:
       return &emit_R64_FLOAT;
@@ -589,7 +372,7 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate,
 
       for (attr = 0; attr < nr_attrs; attr++) {
 	 float data[4];
-         const char *src;
+         const uint8_t *src;
          unsigned index;
 
 	 char *dst = (vert + 
@@ -606,7 +389,7 @@ static void PIPE_CDECL generic_run_elts( struct translate *translate,
          src = tg->attrib[attr].input_ptr +
                tg->attrib[attr].input_stride * index;
 
-	 tg->attrib[attr].fetch( src, data );
+	 tg->attrib[attr].fetch( data, src, 0, 0 );
 
          if (0) debug_printf("vert %d/%d attr %d: %f %f %f %f\n",
                              i, elt, attr, data[0], data[1], data[2], data[3]);
@@ -644,7 +427,7 @@ static void PIPE_CDECL generic_run( struct translate *translate,
 		      tg->attrib[attr].output_offset);
 
          if (tg->attrib[attr].type == TRANSLATE_ELEMENT_NORMAL) {
-            const char *src;
+            const uint8_t *src;
 
             if (tg->attrib[attr].instance_divisor) {
                src = tg->attrib[attr].input_ptr +
@@ -655,7 +438,7 @@ static void PIPE_CDECL generic_run( struct translate *translate,
                      tg->attrib[attr].input_stride * elt;
             }
 
-            tg->attrib[attr].fetch( src, data );
+            tg->attrib[attr].fetch( data, src, 0, 0 );
          } else {
             data[0] = (float)instance_id;
          }
@@ -683,7 +466,7 @@ static void generic_set_buffer( struct translate *translate,
 
    for (i = 0; i < tg->nr_attrib; i++) {
       if (tg->attrib[i].buffer == buf) {
-	 tg->attrib[i].input_ptr = ((char *)ptr +
+	 tg->attrib[i].input_ptr = ((const uint8_t *)ptr +
 				    tg->attrib[i].input_offset);
 	 tg->attrib[i].input_stride = stride;
          tg->attrib[i].max_index = max_index;
@@ -714,9 +497,15 @@ struct translate *translate_generic_create( const struct translate_key *key )
    tg->translate.run = generic_run;
 
    for (i = 0; i < key->nr_elements; i++) {
+      const struct util_format_description *format_desc =
+            util_format_description(key->element[i].input_format);
+
+      assert(format_desc);
+      assert(format_desc->fetch_rgba_float);
+
       tg->attrib[i].type = key->element[i].type;
 
-      tg->attrib[i].fetch = get_fetch_func(key->element[i].input_format);
+      tg->attrib[i].fetch = format_desc->fetch_rgba_float;
       tg->attrib[i].buffer = key->element[i].input_buffer;
       tg->attrib[i].input_offset = key->element[i].input_offset;
       tg->attrib[i].instance_divisor = key->element[i].instance_divisor;
diff --git a/src/gallium/auxiliary/util/u_blitter.c b/src/gallium/auxiliary/util/u_blitter.c
index 956aedc8a15..1abe31db466 100644
--- a/src/gallium/auxiliary/util/u_blitter.c
+++ b/src/gallium/auxiliary/util/u_blitter.c
@@ -47,6 +47,7 @@
 #include "util/u_rect.h"
 #include "util/u_sampler.h"
 #include "util/u_simple_shaders.h"
+#include "util/u_surface.h"
 #include "util/u_texture.h"
 
 #define INVALID_PTR ((void*)~0)
diff --git a/src/gallium/auxiliary/util/u_caps.c b/src/gallium/auxiliary/util/u_caps.c
index 048bd5c34da..c7c1e830e01 100644
--- a/src/gallium/auxiliary/util/u_caps.c
+++ b/src/gallium/auxiliary/util/u_caps.c
@@ -73,6 +73,7 @@ util_check_caps_out(struct pipe_screen *screen, const unsigned *list, int *out)
             *out = i - 2;
             return FALSE;
          }
+         break;
       case UTIL_CAPS_CHECK_UNIMPLEMENTED:
          *out = i - 1;
          return FALSE;
diff --git a/src/gallium/auxiliary/util/u_clear.h b/src/gallium/auxiliary/util/u_clear.h
index 2c32db61756..40da2d75a72 100644
--- a/src/gallium/auxiliary/util/u_clear.h
+++ b/src/gallium/auxiliary/util/u_clear.h
@@ -33,6 +33,7 @@
 #include "pipe/p_state.h"
 #include "util/u_pack_color.h"
 #include "util/u_rect.h"
+#include "util/u_surface.h"
 
 
 /**
diff --git a/src/gallium/auxiliary/util/u_format.h b/src/gallium/auxiliary/util/u_format.h
index fb6ade5c06b..8ba076949b6 100644
--- a/src/gallium/auxiliary/util/u_format.h
+++ b/src/gallium/auxiliary/util/u_format.h
@@ -338,6 +338,34 @@ util_format_name(enum pipe_format format)
    return desc->name;
 }
 
+static INLINE const char *
+util_format_short_name(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   assert(desc);
+   if (!desc) {
+      return "???";
+   }
+
+   return desc->short_name;
+}
+
+/**
+ * Whether this format is plain, see UTIL_FORMAT_LAYOUT_PLAIN for more info.
+ */
+static INLINE boolean
+util_format_is_plain(enum pipe_format format)
+{
+   const struct util_format_description *desc = util_format_description(format);
+
+   if (!format) {
+      return FALSE;
+   }
+
+   return desc->layout == UTIL_FORMAT_LAYOUT_PLAIN ? TRUE : FALSE;
+}
+
 static INLINE boolean 
 util_format_is_s3tc(enum pipe_format format)
 {
diff --git a/src/gallium/auxiliary/util/u_format_s3tc.c b/src/gallium/auxiliary/util/u_format_s3tc.c
index 5b279b8fe26..bb989c29d81 100644
--- a/src/gallium/auxiliary/util/u_format_s3tc.c
+++ b/src/gallium/auxiliary/util/u_format_s3tc.c
@@ -120,7 +120,7 @@ util_format_s3tc_init(void)
    library = util_dl_open(DXTN_LIBNAME);
    if (!library) {
       debug_printf("couldn't open " DXTN_LIBNAME ", software DXTn "
-         "compression/decompression unavailable");
+         "compression/decompression unavailable\n");
       return;
    }
 
@@ -142,7 +142,7 @@ util_format_s3tc_init(void)
        !util_format_dxtn_pack) {
       debug_printf("couldn't reference all symbols in " DXTN_LIBNAME
                    ", software DXTn compression/decompression "
-                   "unavailable");
+                   "unavailable\n");
       util_dl_close(library);
       return;
    }
diff --git a/src/gallium/auxiliary/util/u_framebuffer.c b/src/gallium/auxiliary/util/u_framebuffer.c
new file mode 100644
index 00000000000..bdac12dbca2
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_framebuffer.c
@@ -0,0 +1,111 @@
+/**************************************************************************
+ *
+ * Copyright 2009-2010 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
+ * Framebuffer utility functions.
+ *  
+ * @author Brian Paul
+ */
+
+
+#include "pipe/p_screen.h"
+#include "pipe/p_state.h"
+#include "pipe/p_defines.h"
+#include "util/u_inlines.h"
+
+#include "util/u_memory.h"
+#include "util/u_framebuffer.h"
+
+
+/**
+ * Compare pipe_framebuffer_state objects.
+ * \return TRUE if same, FALSE if different
+ */
+boolean
+util_framebuffer_state_equal(const struct pipe_framebuffer_state *dst,
+                             const struct pipe_framebuffer_state *src)
+{
+   unsigned i;
+
+   if (dst->width != src->width ||
+       dst->height != src->height)
+      return FALSE;
+
+   for (i = 0; i < Elements(src->cbufs); i++) {
+      if (dst->cbufs[i] != src->cbufs[i]) {
+         return FALSE;
+      }
+   }
+
+   if (dst->nr_cbufs != src->nr_cbufs) {
+      return FALSE;
+   }
+
+   if (dst->zsbuf != src->zsbuf) {
+      return FALSE;
+   }
+
+   return TRUE;
+}
+
+
+/**
+ * Copy framebuffer state from src to dst, updating refcounts.
+ */
+void
+util_copy_framebuffer_state(struct pipe_framebuffer_state *dst,
+                            const struct pipe_framebuffer_state *src)
+{
+   unsigned i;
+
+   dst->width = src->width;
+   dst->height = src->height;
+
+   for (i = 0; i < Elements(src->cbufs); i++) {
+      pipe_surface_reference(&dst->cbufs[i], src->cbufs[i]);
+   }
+
+   dst->nr_cbufs = src->nr_cbufs;
+
+   pipe_surface_reference(&dst->zsbuf, src->zsbuf);
+}
+
+
+void
+util_unreference_framebuffer_state(struct pipe_framebuffer_state *fb)
+{
+   unsigned i;
+
+   for (i = 0; i < fb->nr_cbufs; i++) {
+      pipe_surface_reference(&fb->cbufs[i], NULL);
+   }
+
+   pipe_surface_reference(&fb->zsbuf, NULL);
+
+   fb->width = fb->height = 0;
+   fb->nr_cbufs = 0;
+}
diff --git a/src/gallium/auxiliary/util/u_framebuffer.h b/src/gallium/auxiliary/util/u_framebuffer.h
new file mode 100644
index 00000000000..adf1254e1b6
--- /dev/null
+++ b/src/gallium/auxiliary/util/u_framebuffer.h
@@ -0,0 +1,49 @@
+/**************************************************************************
+ *
+ * Copyright 2009-2010 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.
+ *
+ **************************************************************************/
+
+
+#ifndef U_FRAMEBUFFER_H
+#define U_FRAMEBUFFER_H
+
+
+#include "pipe/p_compiler.h"
+#include "pipe/p_state.h"
+
+
+extern boolean
+util_framebuffer_state_equal(const struct pipe_framebuffer_state *dst,
+                             const struct pipe_framebuffer_state *src);
+
+extern void
+util_copy_framebuffer_state(struct pipe_framebuffer_state *dst,
+                            const struct pipe_framebuffer_state *src);
+
+
+extern void
+util_unreference_framebuffer_state(struct pipe_framebuffer_state *fb);
+
+
+#endif /* U_FRAMEBUFFER_H */
diff --git a/src/gallium/auxiliary/util/u_rect.c b/src/gallium/auxiliary/util/u_rect.c
index 098cdfd58b1..9bbcf1c8c49 100644
--- a/src/gallium/auxiliary/util/u_rect.c
+++ b/src/gallium/auxiliary/util/u_rect.c
@@ -30,12 +30,7 @@
  */
 
 
-#include "pipe/p_defines.h"
-#include "pipe/p_format.h"
-#include "pipe/p_context.h"
-#include "pipe/p_screen.h"
 #include "util/u_format.h"
-#include "util/u_inlines.h"
 #include "util/u_rect.h"
 
 
@@ -152,156 +147,3 @@ util_fill_rect(ubyte * dst,
 	 break;
    }
 }
-
-
-
-/**
- * Fallback function for pipe->surface_copy().
- * Note: (X,Y)=(0,0) is always the upper-left corner.
- * if do_flip, flip the image vertically on its way from src rect to dst rect.
- * XXX should probably put this in new u_surface.c file...
- */
-void
-util_surface_copy(struct pipe_context *pipe,
-                  boolean do_flip,
-                  struct pipe_surface *dst,
-                  unsigned dst_x, unsigned dst_y,
-                  struct pipe_surface *src,
-                  unsigned src_x, unsigned src_y, 
-                  unsigned w, unsigned h)
-{
-   struct pipe_transfer *src_trans, *dst_trans;
-   void *dst_map;
-   const void *src_map;
-   enum pipe_format src_format, dst_format;
-
-   assert(src->texture && dst->texture);
-   if (!src->texture || !dst->texture)
-      return;
-
-   src_format = src->texture->format;
-   dst_format = dst->texture->format;
-
-   src_trans = pipe_get_transfer(pipe,
-				 src->texture,
-				 src->face,
-				 src->level,
-				 src->zslice,
-				 PIPE_TRANSFER_READ,
-				 src_x, src_y, w, h);
-
-   dst_trans = pipe_get_transfer(pipe,
-				 dst->texture,
-				 dst->face,
-				 dst->level,
-				 dst->zslice,
-				 PIPE_TRANSFER_WRITE,
-				 dst_x, dst_y, w, h);
-
-   assert(util_format_get_blocksize(dst_format) == util_format_get_blocksize(src_format));
-   assert(util_format_get_blockwidth(dst_format) == util_format_get_blockwidth(src_format));
-   assert(util_format_get_blockheight(dst_format) == util_format_get_blockheight(src_format));
-
-   src_map = pipe->transfer_map(pipe, src_trans);
-   dst_map = pipe->transfer_map(pipe, dst_trans);
-
-   assert(src_map);
-   assert(dst_map);
-
-   if (src_map && dst_map) {
-      /* If do_flip, invert src_y position and pass negative src stride */
-      util_copy_rect(dst_map,
-                     dst_format,
-                     dst_trans->stride,
-                     0, 0,
-                     w, h,
-                     src_map,
-                     do_flip ? -(int) src_trans->stride : src_trans->stride,
-                     0,
-                     do_flip ? h - 1 : 0);
-   }
-
-   pipe->transfer_unmap(pipe, src_trans);
-   pipe->transfer_unmap(pipe, dst_trans);
-
-   pipe->transfer_destroy(pipe, src_trans);
-   pipe->transfer_destroy(pipe, dst_trans);
-}
-
-
-
-#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8))
-
-
-/**
- * Fallback for pipe->surface_fill() function.
- * XXX should probably put this in new u_surface.c file...
- */
-void
-util_surface_fill(struct pipe_context *pipe,
-                  struct pipe_surface *dst,
-                  unsigned dstx, unsigned dsty,
-                  unsigned width, unsigned height, unsigned value)
-{
-   struct pipe_transfer *dst_trans;
-   void *dst_map;
-
-   assert(dst->texture);
-   if (!dst->texture)
-      return;
-   dst_trans = pipe_get_transfer(pipe,
-				 dst->texture,
-				 dst->face,
-				 dst->level,
-				 dst->zslice,
-				 PIPE_TRANSFER_WRITE,
-				 dstx, dsty, width, height);
-
-   dst_map = pipe->transfer_map(pipe, dst_trans);
-
-   assert(dst_map);
-
-   if (dst_map) {
-      assert(dst_trans->stride > 0);
-
-      switch (util_format_get_blocksize(dst->texture->format)) {
-      case 1:
-      case 2:
-      case 4:
-         util_fill_rect(dst_map, dst->texture->format,
-			dst_trans->stride,
-                        0, 0, width, height, value);
-         break;
-      case 8:
-      {
-	 /* expand the 4-byte clear value to an 8-byte value */
-	 ushort *row = (ushort *) dst_map;
-	 ushort val0 = UBYTE_TO_USHORT((value >>  0) & 0xff);
-	 ushort val1 = UBYTE_TO_USHORT((value >>  8) & 0xff);
-	 ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff);
-	 ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff);
-	 unsigned i, j;
-	 val0 = (val0 << 8) | val0;
-	 val1 = (val1 << 8) | val1;
-	 val2 = (val2 << 8) | val2;
-	 val3 = (val3 << 8) | val3;
-	 for (i = 0; i < height; i++) {
-	    for (j = 0; j < width; j++) {
-	       row[j*4+0] = val0;
-	       row[j*4+1] = val1;
-	       row[j*4+2] = val2;
-	       row[j*4+3] = val3;
-	    }
-	    row += dst_trans->stride/2;
-	 }
-      }
-      break;
-      default:
-         assert(0);
-         break;
-      }
-   }
-
-   pipe->transfer_unmap(pipe, dst_trans);
-   pipe->transfer_destroy(pipe, dst_trans);
-}
diff --git a/src/gallium/auxiliary/util/u_rect.h b/src/gallium/auxiliary/util/u_rect.h
index b44d821904b..40d57e662d7 100644
--- a/src/gallium/auxiliary/util/u_rect.h
+++ b/src/gallium/auxiliary/util/u_rect.h
@@ -37,9 +37,6 @@
 
 #include "pipe/p_format.h"
 
-struct pipe_context;
-struct pipe_surface;
-
 
 extern void
 util_copy_rect(ubyte * dst, enum pipe_format format,
@@ -53,20 +50,4 @@ util_fill_rect(ubyte * dst, enum pipe_format format,
                unsigned width, unsigned height, uint32_t value);
 
 
-extern void
-util_surface_copy(struct pipe_context *pipe,
-                  boolean do_flip,
-                  struct pipe_surface *dst,
-                  unsigned dst_x, unsigned dst_y,
-                  struct pipe_surface *src,
-                  unsigned src_x, unsigned src_y, 
-                  unsigned w, unsigned h);
-
-extern void
-util_surface_fill(struct pipe_context *pipe,
-                  struct pipe_surface *dst,
-                  unsigned dstx, unsigned dsty,
-                  unsigned width, unsigned height, unsigned value);
-
-
 #endif /* U_RECT_H */
diff --git a/src/gallium/auxiliary/util/u_surface.c b/src/gallium/auxiliary/util/u_surface.c
index 42440d0d673..b47c92ca28f 100644
--- a/src/gallium/auxiliary/util/u_surface.c
+++ b/src/gallium/auxiliary/util/u_surface.c
@@ -32,12 +32,14 @@
  */
 
 
+#include "pipe/p_defines.h"
 #include "pipe/p_screen.h"
 #include "pipe/p_state.h"
-#include "pipe/p_defines.h"
-#include "util/u_inlines.h"
 
+#include "util/u_format.h"
+#include "util/u_inlines.h"
 #include "util/u_memory.h"
+#include "util/u_rect.h"
 #include "util/u_surface.h"
 
 
@@ -118,70 +120,150 @@ util_destroy_rgba_surface(struct pipe_resource *texture,
 
 
 /**
- * Compare pipe_framebuffer_state objects.
- * \return TRUE if same, FALSE if different
+ * Fallback function for pipe->surface_copy().
+ * Note: (X,Y)=(0,0) is always the upper-left corner.
+ * if do_flip, flip the image vertically on its way from src rect to dst rect.
  */
-boolean
-util_framebuffer_state_equal(const struct pipe_framebuffer_state *dst,
-                             const struct pipe_framebuffer_state *src)
+void
+util_surface_copy(struct pipe_context *pipe,
+                  boolean do_flip,
+                  struct pipe_surface *dst,
+                  unsigned dst_x, unsigned dst_y,
+                  struct pipe_surface *src,
+                  unsigned src_x, unsigned src_y, 
+                  unsigned w, unsigned h)
 {
-   unsigned i;
-
-   if (dst->width != src->width ||
-       dst->height != src->height)
-      return FALSE;
-
-   for (i = 0; i < Elements(src->cbufs); i++) {
-      if (dst->cbufs[i] != src->cbufs[i]) {
-         return FALSE;
-      }
-   }
-
-   if (dst->nr_cbufs != src->nr_cbufs) {
-      return FALSE;
+   struct pipe_transfer *src_trans, *dst_trans;
+   void *dst_map;
+   const void *src_map;
+   enum pipe_format src_format, dst_format;
+
+   assert(src->texture && dst->texture);
+   if (!src->texture || !dst->texture)
+      return;
+
+   src_format = src->texture->format;
+   dst_format = dst->texture->format;
+
+   src_trans = pipe_get_transfer(pipe,
+				 src->texture,
+				 src->face,
+				 src->level,
+				 src->zslice,
+				 PIPE_TRANSFER_READ,
+				 src_x, src_y, w, h);
+
+   dst_trans = pipe_get_transfer(pipe,
+				 dst->texture,
+				 dst->face,
+				 dst->level,
+				 dst->zslice,
+				 PIPE_TRANSFER_WRITE,
+				 dst_x, dst_y, w, h);
+
+   assert(util_format_get_blocksize(dst_format) == util_format_get_blocksize(src_format));
+   assert(util_format_get_blockwidth(dst_format) == util_format_get_blockwidth(src_format));
+   assert(util_format_get_blockheight(dst_format) == util_format_get_blockheight(src_format));
+
+   src_map = pipe->transfer_map(pipe, src_trans);
+   dst_map = pipe->transfer_map(pipe, dst_trans);
+
+   assert(src_map);
+   assert(dst_map);
+
+   if (src_map && dst_map) {
+      /* If do_flip, invert src_y position and pass negative src stride */
+      util_copy_rect(dst_map,
+                     dst_format,
+                     dst_trans->stride,
+                     0, 0,
+                     w, h,
+                     src_map,
+                     do_flip ? -(int) src_trans->stride : src_trans->stride,
+                     0,
+                     do_flip ? h - 1 : 0);
    }
 
-   if (dst->zsbuf != src->zsbuf) {
-      return FALSE;
-   }
+   pipe->transfer_unmap(pipe, src_trans);
+   pipe->transfer_unmap(pipe, dst_trans);
 
-   return TRUE;
+   pipe->transfer_destroy(pipe, src_trans);
+   pipe->transfer_destroy(pipe, dst_trans);
 }
 
 
-/**
- * Copy framebuffer state from src to dst, updating refcounts.
- */
-void
-util_copy_framebuffer_state(struct pipe_framebuffer_state *dst,
-                            const struct pipe_framebuffer_state *src)
-{
-   unsigned i;
-
-   dst->width = src->width;
-   dst->height = src->height;
 
-   for (i = 0; i < Elements(src->cbufs); i++) {
-      pipe_surface_reference(&dst->cbufs[i], src->cbufs[i]);
-   }
-
-   dst->nr_cbufs = src->nr_cbufs;
-
-   pipe_surface_reference(&dst->zsbuf, src->zsbuf);
-}
+#define UBYTE_TO_USHORT(B) ((B) | ((B) << 8))
 
 
+/**
+ * Fallback for pipe->surface_fill() function.
+ */
 void
-util_unreference_framebuffer_state(struct pipe_framebuffer_state *fb)
+util_surface_fill(struct pipe_context *pipe,
+                  struct pipe_surface *dst,
+                  unsigned dstx, unsigned dsty,
+                  unsigned width, unsigned height, unsigned value)
 {
-   unsigned i;
-
-   for (i = 0; i < fb->nr_cbufs; i++) {
-      pipe_surface_reference(&fb->cbufs[i], NULL);
+   struct pipe_transfer *dst_trans;
+   void *dst_map;
+
+   assert(dst->texture);
+   if (!dst->texture)
+      return;
+   dst_trans = pipe_get_transfer(pipe,
+				 dst->texture,
+				 dst->face,
+				 dst->level,
+				 dst->zslice,
+				 PIPE_TRANSFER_WRITE,
+				 dstx, dsty, width, height);
+
+   dst_map = pipe->transfer_map(pipe, dst_trans);
+
+   assert(dst_map);
+
+   if (dst_map) {
+      assert(dst_trans->stride > 0);
+
+      switch (util_format_get_blocksize(dst->texture->format)) {
+      case 1:
+      case 2:
+      case 4:
+         util_fill_rect(dst_map, dst->texture->format,
+			dst_trans->stride,
+                        0, 0, width, height, value);
+         break;
+      case 8:
+      {
+	 /* expand the 4-byte clear value to an 8-byte value */
+	 ushort *row = (ushort *) dst_map;
+	 ushort val0 = UBYTE_TO_USHORT((value >>  0) & 0xff);
+	 ushort val1 = UBYTE_TO_USHORT((value >>  8) & 0xff);
+	 ushort val2 = UBYTE_TO_USHORT((value >> 16) & 0xff);
+	 ushort val3 = UBYTE_TO_USHORT((value >> 24) & 0xff);
+	 unsigned i, j;
+	 val0 = (val0 << 8) | val0;
+	 val1 = (val1 << 8) | val1;
+	 val2 = (val2 << 8) | val2;
+	 val3 = (val3 << 8) | val3;
+	 for (i = 0; i < height; i++) {
+	    for (j = 0; j < width; j++) {
+	       row[j*4+0] = val0;
+	       row[j*4+1] = val1;
+	       row[j*4+2] = val2;
+	       row[j*4+3] = val3;
+	    }
+	    row += dst_trans->stride/2;
+	 }
+      }
+      break;
+      default:
+         assert(0);
+         break;
+      }
    }
 
-   pipe_surface_reference(&fb->zsbuf, NULL);
-
-   fb->width = fb->height = 0;
-   fb->nr_cbufs = 0;
+   pipe->transfer_unmap(pipe, dst_trans);
+   pipe->transfer_destroy(pipe, dst_trans);
 }
diff --git a/src/gallium/auxiliary/util/u_surface.h b/src/gallium/auxiliary/util/u_surface.h
index 119fcd4ce8e..c43169b5278 100644
--- a/src/gallium/auxiliary/util/u_surface.h
+++ b/src/gallium/auxiliary/util/u_surface.h
@@ -62,17 +62,22 @@ util_destroy_rgba_surface(struct pipe_resource *texture,
                           struct pipe_surface *surface);
 
 
-extern boolean
-util_framebuffer_state_equal(const struct pipe_framebuffer_state *dst,
-                             const struct pipe_framebuffer_state *src);
 
 extern void
-util_copy_framebuffer_state(struct pipe_framebuffer_state *dst,
-                            const struct pipe_framebuffer_state *src);
-
+util_surface_copy(struct pipe_context *pipe,
+                  boolean do_flip,
+                  struct pipe_surface *dst,
+                  unsigned dst_x, unsigned dst_y,
+                  struct pipe_surface *src,
+                  unsigned src_x, unsigned src_y, 
+                  unsigned w, unsigned h);
 
 extern void
-util_unreference_framebuffer_state(struct pipe_framebuffer_state *fb);
+util_surface_fill(struct pipe_context *pipe,
+                  struct pipe_surface *dst,
+                  unsigned dstx, unsigned dsty,
+                  unsigned width, unsigned height, unsigned value);
+
 
 
 #endif /* U_SURFACE_H */