Merge commit 'origin/master' into gallium-msaa

93 files changed, 1478 insertions, 1407 deletions

diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile
index 2daed382cf..e0d9b31354 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 a0673df8a8..29a1bd0ecb 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 55e0ded1b3..20a8612dca 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 3b2df054c3..ea9b7c90a5 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 d2a492f2b4..6d00b0fbd4 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 c2832eefa2..b9db886a24 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 20ae958714..f372a48846 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 0000000000..e095a0abe3
--- /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 8ceb20473d..955b1d21ee 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 74dc9e1d81..3f0ea05b79 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 <jfonseca@vmware.com>
+ * @author Brian Paul <brianp@vmware.com>
  */
 
 #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 2eac5da6c6..5ce1385c48 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 d3c769e28b..0b1a28a7ab 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 8300020018..1ad183b8dd 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 82eac05dc4..9b1ca7fa85 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 a22873e4c2..11bbaf6722 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 76b7564cc3..ce0a92f7fb 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 0b468a9184..71a0d96331 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 f725405ade..49b854b123 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 0130a77aad..f32420dd87 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 a9272fbb49..f8dbd2b36a 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 956aedc8a1..1abe31db46 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 048bd5c34d..c7c1e830e0 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 2c32db6175..40da2d75a7 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 fb6ade5c06..8ba076949b 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 5b279b8fe2..bb989c29d8 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 0000000000..bdac12dbca
--- /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 0000000000..adf1254e1b
--- /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 098cdfd58b..9bbcf1c8c4 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 b44d821904..40d57e662d 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 42440d0d67..b47c92ca28 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 119fcd4ce8..c43169b527 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 */
diff --git a/src/gallium/drivers/cell/ppu/cell_surface.c b/src/gallium/drivers/cell/ppu/cell_surface.c
index ffb8595d82..6696a4591c 100644
--- a/src/gallium/drivers/cell/ppu/cell_surface.c
+++ b/src/gallium/drivers/cell/ppu/cell_surface.c
@@ -25,7 +25,7 @@
  * 
  **************************************************************************/
 
-#include "util/u_rect.h"
+#include "util/u_surface.h"
 #include "cell_context.h"
 #include "cell_surface.h"
 
diff --git a/src/gallium/drivers/i965/brw_wm_debug.c b/src/gallium/drivers/i965/brw_wm_debug.c
index 3d11fa074c..e2767264e7 100644
--- a/src/gallium/drivers/i965/brw_wm_debug.c
+++ b/src/gallium/drivers/i965/brw_wm_debug.c
@@ -202,7 +202,8 @@ static const char *file_strings[TGSI_FILE_COUNT+1] = {
    "SAMPLER",
    "ADDR",
    "IMM",
-   "LOOP",
+   "PRED",
+   "SV",
    "PAYLOAD"
 };
 
diff --git a/src/gallium/drivers/i965/intel_decode.c b/src/gallium/drivers/i965/intel_decode.c
index 6c47415cac..bd8b9174a8 100644
--- a/src/gallium/drivers/i965/intel_decode.c
+++ b/src/gallium/drivers/i965/intel_decode.c
@@ -40,6 +40,7 @@
 #include <stdint.h>
 #include <string.h>
 
+#include "util/u_string.h"
 #include "intel_decode.h"
 
 /*#include "intel_chipset.h"*/
@@ -478,7 +479,7 @@ i915_get_instruction_src0(const uint32_t *data, int i, char *srcname)
     char swizzle[100];
 
     i915_get_instruction_src_name((a0 >> 7) & 0x7, src_nr, srcname);
-    sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
+    util_snprintf(swizzle, sizeof(swizzle), ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
     if (strcmp(swizzle, ".xyzw") != 0)
 	strcat(srcname, swizzle);
 }
@@ -496,7 +497,7 @@ i915_get_instruction_src1(const uint32_t *data, int i, char *srcname)
     char swizzle[100];
 
     i915_get_instruction_src_name((a1 >> 13) & 0x7, src_nr, srcname);
-    sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
+    util_snprintf(swizzle, sizeof(swizzle), ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
     if (strcmp(swizzle, ".xyzw") != 0)
 	strcat(srcname, swizzle);
 }
@@ -513,7 +514,7 @@ i915_get_instruction_src2(const uint32_t *data, int i, char *srcname)
     char swizzle[100];
 
     i915_get_instruction_src_name((a2 >> 21) & 0x7, src_nr, srcname);
-    sprintf(swizzle, ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
+    util_snprintf(swizzle, sizeof(swizzle), ".%s%s%s%s", swizzle_x, swizzle_y, swizzle_z, swizzle_w);
     if (strcmp(swizzle, ".xyzw") != 0)
 	strcat(srcname, swizzle);
 }
@@ -642,7 +643,7 @@ i915_decode_dcl(const uint32_t *data, uint32_t hw_offset, int i, char *instr_pre
 
     switch ((d0 >> 19) & 0x3) {
     case 1:
-	sprintf(dcl_mask, ".%s%s%s%s", dcl_x, dcl_y, dcl_z, dcl_w);
+	util_snprintf(dcl_mask, sizeof(dcl_mask), ".%s%s%s%s", dcl_x, dcl_y, dcl_z, dcl_w);
 	if (strcmp(dcl_mask, ".") == 0)
 	    fprintf(out, "bad (empty) dcl mask\n");
 
@@ -976,7 +977,7 @@ decode_3d_1d(const uint32_t *data, int count, uint32_t hw_offset, int *failures,
 
 	    if (i + 3 >= count)
 		BUFFER_FAIL(count, len, "3DSTATE_PIXEL_SHADER_PROGRAM");
-	    sprintf(instr_prefix, "PS%03d", instr);
+	    util_snprintf(instr_prefix, sizeof(instr_prefix), "PS%03d", instr);
 	    i915_decode_instruction(data, hw_offset, i, instr_prefix);
 	    i += 3;
 	}
diff --git a/src/gallium/drivers/identity/id_context.c b/src/gallium/drivers/identity/id_context.c
index 0bc8bf2196..7e62213597 100644
--- a/src/gallium/drivers/identity/id_context.c
+++ b/src/gallium/drivers/identity/id_context.c
@@ -684,79 +684,67 @@ identity_flush(struct pipe_context *_pipe,
 
 static unsigned int
 identity_is_resource_referenced(struct pipe_context *_pipe,
-                               struct pipe_resource *_resource,
-                               unsigned face,
-                               unsigned level)
+                                struct pipe_resource *_resource,
+                                unsigned face,
+                                unsigned level)
 {
    struct identity_context *id_pipe = identity_context(_pipe);
    struct identity_resource *id_resource = identity_resource(_resource);
    struct pipe_context *pipe = id_pipe->pipe;
-   struct pipe_resource *texture = id_resource->resource;
+   struct pipe_resource *resource = id_resource->resource;
 
    return pipe->is_resource_referenced(pipe,
-                                      texture,
-                                      face,
-                                      level);
+                                       resource,
+                                       face,
+                                       level);
 }
 
 static struct pipe_sampler_view *
-identity_create_sampler_view(struct pipe_context *pipe,
-                             struct pipe_resource *texture,
-                             const struct pipe_sampler_view *templ)
+identity_context_create_sampler_view(struct pipe_context *_pipe,
+                                     struct pipe_resource *_resource,
+                                     const struct pipe_sampler_view *templ)
 {
-   struct identity_context *id_pipe = identity_context(pipe);
-   struct identity_resource *id_resource = identity_resource(texture);
-   struct pipe_context *pipe_unwrapped = id_pipe->pipe;
-   struct pipe_resource *texture_unwrapped = id_resource->resource;
-   struct identity_sampler_view *view = MALLOC(sizeof(struct identity_sampler_view));
-
-   view->sampler_view = pipe_unwrapped->create_sampler_view(pipe_unwrapped,
-                                                            texture_unwrapped,
-                                                            templ);
+   struct identity_context *id_context = identity_context(_pipe);
+   struct identity_resource *id_resource = identity_resource(_resource);
+   struct pipe_context *pipe = id_context->pipe;
+   struct pipe_resource *resource = id_resource->resource;
+   struct pipe_sampler_view *result;
 
-   view->base = *templ;
-   view->base.reference.count = 1;
-   view->base.texture = NULL;
-   pipe_resource_reference(&view->base.texture, texture);
-   view->base.context = pipe;
+   result = pipe->create_sampler_view(pipe,
+                                      resource,
+                                      templ);
 
-   return &view->base;
+   if (result)
+      return identity_sampler_view_create(id_context, id_resource, result);
+   return NULL;
 }
 
 static void
-identity_sampler_view_destroy(struct pipe_context *pipe,
-                              struct pipe_sampler_view *view)
+identity_context_sampler_view_destroy(struct pipe_context *_pipe,
+                                      struct pipe_sampler_view *_view)
 {
-   struct identity_context *id_pipe = identity_context(pipe);
-   struct identity_sampler_view *id_view = identity_sampler_view(view);
-   struct pipe_context *pipe_unwrapped = id_pipe->pipe;
-   struct pipe_sampler_view *view_unwrapped = id_view->sampler_view;
-
-   pipe_unwrapped->sampler_view_destroy(pipe_unwrapped,
-                                        view_unwrapped);
-
-   pipe_resource_reference(&view->texture, NULL);
-   FREE(view);
+   identity_sampler_view_destroy(identity_context(_pipe),
+                                 identity_sampler_view(_view));
 }
 
 static struct pipe_transfer *
 identity_context_get_transfer(struct pipe_context *_context,
-			      struct pipe_resource *_resource,
-			      struct pipe_subresource sr,
-			      unsigned usage,
-			      const struct pipe_box *box)
+                              struct pipe_resource *_resource,
+                              struct pipe_subresource sr,
+                              unsigned usage,
+                              const struct pipe_box *box)
 {
    struct identity_context *id_context = identity_context(_context);
    struct identity_resource *id_resource = identity_resource(_resource);
    struct pipe_context *context = id_context->pipe;
-   struct pipe_resource *texture = id_resource->resource;
+   struct pipe_resource *resource = id_resource->resource;
    struct pipe_transfer *result;
 
    result = context->get_transfer(context,
-				  texture,
-				  sr,
-				  usage,
-				  box);
+                                  resource,
+                                  sr,
+                                  usage,
+                                  box);
 
    if (result)
       return identity_transfer_create(id_context, id_resource, result);
@@ -765,7 +753,7 @@ identity_context_get_transfer(struct pipe_context *_context,
 
 static void
 identity_context_transfer_destroy(struct pipe_context *_pipe,
-                                      struct pipe_transfer *_transfer)
+                                  struct pipe_transfer *_transfer)
 {
    identity_transfer_destroy(identity_context(_pipe),
                              identity_transfer(_transfer));
@@ -773,7 +761,7 @@ identity_context_transfer_destroy(struct pipe_context *_pipe,
 
 static void *
 identity_context_transfer_map(struct pipe_context *_context,
-                             struct pipe_transfer *_transfer)
+                              struct pipe_transfer *_transfer)
 {
    struct identity_context *id_context = identity_context(_context);
    struct identity_transfer *id_transfer = identity_transfer(_transfer);
@@ -781,15 +769,15 @@ identity_context_transfer_map(struct pipe_context *_context,
    struct pipe_transfer *transfer = id_transfer->transfer;
 
    return context->transfer_map(context,
-				transfer);
+                                transfer);
 }
 
 
 
 static void
-identity_context_transfer_flush_region( struct pipe_context *_context,
-					struct pipe_transfer *_transfer,
-					const struct pipe_box *box)
+identity_context_transfer_flush_region(struct pipe_context *_context,
+                                       struct pipe_transfer *_transfer,
+                                       const struct pipe_box *box)
 {
    struct identity_context *id_context = identity_context(_context);
    struct identity_transfer *id_transfer = identity_transfer(_transfer);
@@ -797,14 +785,14 @@ identity_context_transfer_flush_region( struct pipe_context *_context,
    struct pipe_transfer *transfer = id_transfer->transfer;
 
    context->transfer_flush_region(context,
-				  transfer,
-				  box);
+                                  transfer,
+                                  box);
 }
 
 
 static void
 identity_context_transfer_unmap(struct pipe_context *_context,
-                               struct pipe_transfer *_transfer)
+                                struct pipe_transfer *_transfer)
 {
    struct identity_context *id_context = identity_context(_context);
    struct identity_transfer *id_transfer = identity_transfer(_transfer);
@@ -812,33 +800,33 @@ identity_context_transfer_unmap(struct pipe_context *_context,
    struct pipe_transfer *transfer = id_transfer->transfer;
 
    context->transfer_unmap(context,
-                          transfer);
+                           transfer);
 }
 
 
 static void 
-identity_context_transfer_inline_write( struct pipe_context *_context,
-					struct pipe_resource *_resource,
-					struct pipe_subresource sr,
-					unsigned usage,
-					const struct pipe_box *box,
-					const void *data,
-					unsigned stride,
-					unsigned slice_stride)
+identity_context_transfer_inline_write(struct pipe_context *_context,
+                                       struct pipe_resource *_resource,
+                                       struct pipe_subresource sr,
+                                       unsigned usage,
+                                       const struct pipe_box *box,
+                                       const void *data,
+                                       unsigned stride,
+                                       unsigned slice_stride)
 {
    struct identity_context *id_context = identity_context(_context);
    struct identity_resource *id_resource = identity_resource(_resource);
    struct pipe_context *context = id_context->pipe;
-   struct pipe_resource *texture = id_resource->resource;
+   struct pipe_resource *resource = id_resource->resource;
 
    context->transfer_inline_write(context,
-				  texture,
-				  sr,
-				  usage,
-				  box,
-				  data,
-				  stride,
-				  slice_stride);
+                                  resource,
+                                  sr,
+                                  usage,
+                                  box,
+                                  data,
+                                  stride,
+                                  slice_stride);
 }
 
 
@@ -905,8 +893,8 @@ identity_context_create(struct pipe_screen *_screen, struct pipe_context *pipe)
    id_pipe->base.clear = identity_clear;
    id_pipe->base.flush = identity_flush;
    id_pipe->base.is_resource_referenced = identity_is_resource_referenced;
-   id_pipe->base.create_sampler_view = identity_create_sampler_view;
-   id_pipe->base.sampler_view_destroy = identity_sampler_view_destroy;
+   id_pipe->base.create_sampler_view = identity_context_create_sampler_view;
+   id_pipe->base.sampler_view_destroy = identity_context_sampler_view_destroy;
    id_pipe->base.get_transfer = identity_context_get_transfer;
    id_pipe->base.transfer_destroy = identity_context_transfer_destroy;
    id_pipe->base.transfer_map = identity_context_transfer_map;
diff --git a/src/gallium/drivers/identity/id_objects.c b/src/gallium/drivers/identity/id_objects.c
index d50914e7d5..ca4743f9ef 100644
--- a/src/gallium/drivers/identity/id_objects.c
+++ b/src/gallium/drivers/identity/id_objects.c
@@ -108,9 +108,45 @@ identity_surface_destroy(struct identity_surface *id_surface)
 }
 
 
+struct pipe_sampler_view *
+identity_sampler_view_create(struct identity_context *id_context,
+                             struct identity_resource *id_resource,
+                             struct pipe_sampler_view *view)
+{
+   struct identity_sampler_view *id_view;
+
+   if (!view)
+      goto error;
+
+   assert(view->texture == id_resource->resource);
+
+   id_view = MALLOC(sizeof(struct identity_sampler_view));
+
+   id_view->base = *view;
+   id_view->base.reference.count = 1;
+   id_view->base.texture = NULL;
+   pipe_resource_reference(&id_view->base.texture, id_resource->resource);
+   id_view->base.context = id_context->pipe;
+
+   return &id_view->base;
+error:
+   return NULL;
+}
+
+void
+identity_sampler_view_destroy(struct identity_context *id_context,
+                              struct identity_sampler_view *id_view)
+{
+   pipe_resource_reference(&id_view->base.texture, NULL);
+   id_context->pipe->sampler_view_destroy(id_context->pipe,
+                                          id_view->sampler_view);
+   FREE(id_view);
+}
+
+
 struct pipe_transfer *
 identity_transfer_create(struct identity_context *id_context,
-			 struct identity_resource *id_resource,
+                         struct identity_resource *id_resource,
                          struct pipe_transfer *transfer)
 {
    struct identity_transfer *id_transfer;
@@ -144,8 +180,8 @@ identity_transfer_destroy(struct identity_context *id_context,
                           struct identity_transfer *id_transfer)
 {
    pipe_resource_reference(&id_transfer->base.resource, NULL);
-   id_context->pipe->transfer_destroy(id_context->pipe,
-				      id_transfer->transfer);
+   id_transfer->pipe->transfer_destroy(id_context->pipe,
+                                       id_transfer->transfer);
    FREE(id_transfer);
 }
 
diff --git a/src/gallium/drivers/identity/id_objects.h b/src/gallium/drivers/identity/id_objects.h
index 058cf3009d..5eea10b0b5 100644
--- a/src/gallium/drivers/identity/id_objects.h
+++ b/src/gallium/drivers/identity/id_objects.h
@@ -142,7 +142,7 @@ identity_transfer_unwrap(struct pipe_transfer *_transfer)
 
 struct pipe_resource *
 identity_resource_create(struct identity_screen *id_screen,
-                        struct pipe_resource *resource);
+                         struct pipe_resource *resource);
 
 void
 identity_resource_destroy(struct identity_resource *id_resource);
@@ -154,9 +154,18 @@ identity_surface_create(struct identity_resource *id_resource,
 void
 identity_surface_destroy(struct identity_surface *id_surface);
 
+struct pipe_sampler_view *
+identity_sampler_view_create(struct identity_context *id_context,
+                             struct identity_resource *id_resource,
+                             struct pipe_sampler_view *view);
+
+void
+identity_sampler_view_destroy(struct identity_context *id_context,
+                              struct identity_sampler_view *id_sampler_view);
+
 struct pipe_transfer *
 identity_transfer_create(struct identity_context *id_context,
-			 struct identity_resource *id_resource,
+                         struct identity_resource *id_resource,
                          struct pipe_transfer *transfer);
 
 void
diff --git a/src/gallium/drivers/identity/id_screen.c b/src/gallium/drivers/identity/id_screen.c
index 52573b211f..7671bded98 100644
--- a/src/gallium/drivers/identity/id_screen.c
+++ b/src/gallium/drivers/identity/id_screen.c
@@ -120,14 +120,14 @@ identity_screen_context_create(struct pipe_screen *_screen,
 
 static struct pipe_resource *
 identity_screen_resource_create(struct pipe_screen *_screen,
-                               const struct pipe_resource *templat)
+                                const struct pipe_resource *templat)
 {
    struct identity_screen *id_screen = identity_screen(_screen);
    struct pipe_screen *screen = id_screen->screen;
    struct pipe_resource *result;
 
    result = screen->resource_create(screen,
-                                   templat);
+                                    templat);
 
    if (result)
       return identity_resource_create(id_screen, result);
@@ -136,8 +136,8 @@ identity_screen_resource_create(struct pipe_screen *_screen,
 
 static struct pipe_resource *
 identity_screen_resource_from_handle(struct pipe_screen *_screen,
-                                    const struct pipe_resource *templ,
-                                    struct winsys_handle *handle)
+                                     const struct pipe_resource *templ,
+                                     struct winsys_handle *handle)
 {
    struct identity_screen *id_screen = identity_screen(_screen);
    struct pipe_screen *screen = id_screen->screen;
@@ -154,44 +154,44 @@ identity_screen_resource_from_handle(struct pipe_screen *_screen,
 
 static boolean
 identity_screen_resource_get_handle(struct pipe_screen *_screen,
-                                   struct pipe_resource *_texture,
-                                   struct winsys_handle *handle)
+                                    struct pipe_resource *_resource,
+                                    struct winsys_handle *handle)
 {
    struct identity_screen *id_screen = identity_screen(_screen);
-   struct identity_resource *id_resource = identity_resource(_texture);
+   struct identity_resource *id_resource = identity_resource(_resource);
    struct pipe_screen *screen = id_screen->screen;
-   struct pipe_resource *texture = id_resource->resource;
+   struct pipe_resource *resource = id_resource->resource;
 
    /* TODO trace call */
 
-   return screen->resource_get_handle(screen, texture, handle);
+   return screen->resource_get_handle(screen, resource, handle);
 }
 
 
 
 static void
 identity_screen_resource_destroy(struct pipe_screen *screen,
-				 struct pipe_resource *_texture)
+                                 struct pipe_resource *_resource)
 {
-   identity_resource_destroy(identity_resource(_texture));
+   identity_resource_destroy(identity_resource(_resource));
 }
 
 static struct pipe_surface *
 identity_screen_get_tex_surface(struct pipe_screen *_screen,
-                                struct pipe_resource *_texture,
+                                struct pipe_resource *_resource,
                                 unsigned face,
                                 unsigned level,
                                 unsigned zslice,
                                 unsigned usage)
 {
    struct identity_screen *id_screen = identity_screen(_screen);
-   struct identity_resource *id_resource = identity_resource(_texture);
+   struct identity_resource *id_resource = identity_resource(_resource);
    struct pipe_screen *screen = id_screen->screen;
-   struct pipe_resource *texture = id_resource->resource;
+   struct pipe_resource *resource = id_resource->resource;
    struct pipe_surface *result;
 
    result = screen->get_tex_surface(screen,
-                                    texture,
+                                    resource,
                                     face,
                                     level,
                                     zslice,
@@ -214,7 +214,7 @@ static struct pipe_resource *
 identity_screen_user_buffer_create(struct pipe_screen *_screen,
                                    void *ptr,
                                    unsigned bytes,
-				   unsigned usage)
+                                   unsigned usage)
 {
    struct identity_screen *id_screen = identity_screen(_screen);
    struct pipe_screen *screen = id_screen->screen;
@@ -223,7 +223,7 @@ identity_screen_user_buffer_create(struct pipe_screen *_screen,
    result = screen->user_buffer_create(screen,
                                        ptr,
                                        bytes,
-				       usage);
+                                       usage);
 
    if (result)
       return identity_resource_create(id_screen, result);
diff --git a/src/gallium/drivers/llvmpipe/.gitignore b/src/gallium/drivers/llvmpipe/.gitignore
index 257b72d7b2..a1b6f56e0d 100644
--- a/src/gallium/drivers/llvmpipe/.gitignore
+++ b/src/gallium/drivers/llvmpipe/.gitignore
@@ -1 +1,5 @@
 lp_tile_soa.c
+lp_test_blend
+lp_test_conv
+lp_test_format
+lp_test_printf
diff --git a/src/gallium/drivers/llvmpipe/lp_jit.c b/src/gallium/drivers/llvmpipe/lp_jit.c
index 466a2f54fb..30e206a2b4 100644
--- a/src/gallium/drivers/llvmpipe/lp_jit.c
+++ b/src/gallium/drivers/llvmpipe/lp_jit.c
@@ -185,6 +185,8 @@ lp_jit_screen_init(struct llvmpipe_screen *screen)
       /* These are the passes currently listed in llvm-c/Transforms/Scalar.h,
        * but there are more on SVN. */
       /* TODO: Add more passes */
+      LLVMAddCFGSimplificationPass(screen->pass);
+      LLVMAddPromoteMemoryToRegisterPass(screen->pass);
       LLVMAddConstantPropagationPass(screen->pass);
       if(util_cpu_caps.has_sse4_1) {
          /* FIXME: There is a bug in this pass, whereby the combination of fptosi
@@ -193,9 +195,7 @@ lp_jit_screen_init(struct llvmpipe_screen *screen)
           */
          LLVMAddInstructionCombiningPass(screen->pass);
       }
-      LLVMAddPromoteMemoryToRegisterPass(screen->pass);
       LLVMAddGVNPass(screen->pass);
-      LLVMAddCFGSimplificationPass(screen->pass);
    }
 
    lp_jit_init_globals(screen);
diff --git a/src/gallium/drivers/llvmpipe/lp_scene.c b/src/gallium/drivers/llvmpipe/lp_scene.c
index 1482a777ff..887f2dbad9 100644
--- a/src/gallium/drivers/llvmpipe/lp_scene.c
+++ b/src/gallium/drivers/llvmpipe/lp_scene.c
@@ -25,6 +25,7 @@
  *
  **************************************************************************/
 
+#include "util/u_framebuffer.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
 #include "util/u_inlines.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_setup.c b/src/gallium/drivers/llvmpipe/lp_setup.c
index 2150956008..1a2cd55b16 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup.c
@@ -33,6 +33,7 @@
  */
 
 #include "pipe/p_defines.h"
+#include "util/u_framebuffer.h"
 #include "util/u_inlines.h"
 #include "util/u_memory.h"
 #include "util/u_pack_color.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
index 965777b641..5f861d6ca4 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -618,7 +618,6 @@ generate_fragment(struct llvmpipe_context *lp,
    struct lp_type fs_type;
    struct lp_type blend_type;
    LLVMTypeRef fs_elem_type;
-   LLVMTypeRef fs_vec_type;
    LLVMTypeRef fs_int_vec_type;
    LLVMTypeRef blend_vec_type;
    LLVMTypeRef arg_types[15];
@@ -675,7 +674,6 @@ generate_fragment(struct llvmpipe_context *lp,
     */
 
    fs_elem_type = lp_build_elem_type(fs_type);
-   fs_vec_type = lp_build_vec_type(fs_type);
    fs_int_vec_type = lp_build_int_vec_type(fs_type);
 
    blend_vec_type = lp_build_vec_type(blend_type);
@@ -1064,6 +1062,34 @@ llvmpipe_set_constant_buffer(struct pipe_context *pipe,
 
 
 /**
+ * Return the blend factor equivalent to a destination alpha of one.
+ */
+static INLINE unsigned
+force_dst_alpha_one(unsigned factor, boolean alpha)
+{
+   switch(factor) {
+   case PIPE_BLENDFACTOR_DST_ALPHA:
+      return PIPE_BLENDFACTOR_ONE;
+   case PIPE_BLENDFACTOR_INV_DST_ALPHA:
+      return PIPE_BLENDFACTOR_ZERO;
+   case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+      return PIPE_BLENDFACTOR_ZERO;
+   }
+
+   if (alpha) {
+      switch(factor) {
+      case PIPE_BLENDFACTOR_DST_COLOR:
+         return PIPE_BLENDFACTOR_ONE;
+      case PIPE_BLENDFACTOR_INV_DST_COLOR:
+         return PIPE_BLENDFACTOR_ZERO;
+      }
+   }
+
+   return factor;
+}
+
+
+/**
  * We need to generate several variants of the fragment pipeline to match
  * all the combinations of the contributing state atoms.
  *
@@ -1104,6 +1130,7 @@ make_variant_key(struct llvmpipe_context *lp,
 
    key->nr_cbufs = lp->framebuffer.nr_cbufs;
    for (i = 0; i < lp->framebuffer.nr_cbufs; i++) {
+      struct pipe_rt_blend_state *blend_rt = &key->blend.rt[i];
       const struct util_format_description *format_desc;
       unsigned chan;
 
@@ -1111,7 +1138,7 @@ make_variant_key(struct llvmpipe_context *lp,
       assert(format_desc->colorspace == UTIL_FORMAT_COLORSPACE_RGB ||
              format_desc->colorspace == UTIL_FORMAT_COLORSPACE_SRGB);
 
-      key->blend.rt[i].colormask = lp->blend->rt[i].colormask;
+      blend_rt->colormask = lp->blend->rt[i].colormask;
 
       /* mask out color channels not present in the color buffer.
        * Should be simple to incorporate per-cbuf writemasks:
@@ -1120,7 +1147,27 @@ make_variant_key(struct llvmpipe_context *lp,
          enum util_format_swizzle swizzle = format_desc->swizzle[chan];
 
          if(swizzle > UTIL_FORMAT_SWIZZLE_W)
-            key->blend.rt[i].colormask &= ~(1 << chan);
+            blend_rt->colormask &= ~(1 << chan);
+      }
+
+      /*
+       * Our swizzled render tiles always have an alpha channel, but the linear
+       * render target format often does not, so force here the dst alpha to be
+       * one.
+       *
+       * This is not a mere optimization. Wrong results will be produced if the
+       * dst alpha is used, the dst format does not have alpha, and the previous
+       * rendering was not flushed from the swizzled to linear buffer. For
+       * example, NonPowTwo DCT.
+       *
+       * TODO: This should be generalized to all channels for better
+       * performance, but only alpha causes correctness issues.
+       */
+      if (format_desc->swizzle[3] > UTIL_FORMAT_SWIZZLE_W) {
+         blend_rt->rgb_src_factor = force_dst_alpha_one(blend_rt->rgb_src_factor, FALSE);
+         blend_rt->rgb_dst_factor = force_dst_alpha_one(blend_rt->rgb_dst_factor, FALSE);
+         blend_rt->alpha_src_factor = force_dst_alpha_one(blend_rt->alpha_src_factor, TRUE);
+         blend_rt->alpha_dst_factor = force_dst_alpha_one(blend_rt->alpha_dst_factor, TRUE);
       }
    }
 
diff --git a/src/gallium/drivers/llvmpipe/lp_state_surface.c b/src/gallium/drivers/llvmpipe/lp_state_surface.c
index 63b8f27b39..4b135aaf8b 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_surface.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_surface.c
@@ -30,6 +30,7 @@
 
 #include "pipe/p_state.h"
 #include "util/u_inlines.h"
+#include "util/u_framebuffer.h"
 #include "util/u_surface.h"
 #include "lp_context.h"
 #include "lp_scene.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_surface.c b/src/gallium/drivers/llvmpipe/lp_surface.c
index 8bd83f576f..245171120d 100644
--- a/src/gallium/drivers/llvmpipe/lp_surface.c
+++ b/src/gallium/drivers/llvmpipe/lp_surface.c
@@ -26,6 +26,7 @@
  **************************************************************************/
 
 #include "util/u_rect.h"
+#include "util/u_surface.h"
 #include "lp_context.h"
 #include "lp_flush.h"
 #include "lp_limits.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample.c b/src/gallium/drivers/llvmpipe/lp_tex_sample.c
index 74b7393e4e..d3a9d39f61 100644
--- a/src/gallium/drivers/llvmpipe/lp_tex_sample.c
+++ b/src/gallium/drivers/llvmpipe/lp_tex_sample.c
@@ -170,7 +170,10 @@ lp_llvm_sampler_soa_emit_fetch_texel(struct lp_build_sampler_soa *base,
                                      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_llvm_sampler_soa *sampler = (struct lp_llvm_sampler_soa *)base;
@@ -182,9 +185,9 @@ lp_llvm_sampler_soa_emit_fetch_texel(struct lp_build_sampler_soa *base,
                        &sampler->dynamic_state.base,
                        type,
                        unit,
-                       num_coords,
-                       coords,
-                       lodbias,
+                       num_coords, coords,
+                       ddx, ddy,
+                       lod_bias, explicit_lod,
                        texel);
 }
 
diff --git a/src/gallium/drivers/nv50/nv50_program.c b/src/gallium/drivers/nv50/nv50_program.c
index b8b6b12120..0156ff95ff 100644
--- a/src/gallium/drivers/nv50/nv50_program.c
+++ b/src/gallium/drivers/nv50/nv50_program.c
@@ -3169,15 +3169,16 @@ nv50_program_tx_insn(struct nv50_pc *pc,
 		if (pc->p->type == PIPE_SHADER_FRAGMENT)
 			nv50_fp_move_results(pc);
 
-		/* last insn must be long so it can have the exit bit set */
-		if (!is_long(pc->p->exec_tail))
-			convert_to_long(pc, pc->p->exec_tail);
-		else
-		if (is_immd(pc->p->exec_tail) ||
+		if (!pc->p->exec_tail ||
+		    is_immd(pc->p->exec_tail) ||
 		    is_join(pc->p->exec_tail) ||
 		    is_control_flow(pc->p->exec_tail))
 			emit_nop(pc);
 
+		/* last insn must be long so it can have the exit bit set */
+		if (!is_long(pc->p->exec_tail))
+			convert_to_long(pc, pc->p->exec_tail);
+
 		pc->p->exec_tail->inst[1] |= 1; /* set exit bit */
 
 		terminate_mbb(pc);
@@ -4162,7 +4163,7 @@ nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p)
 	struct pipe_transfer *transfer;
 
 	if (!p->data[0] && p->immd_nr) {
-		struct nouveau_resource *heap = nv50->screen->immd_heap[0];
+		struct nouveau_resource *heap = nv50->screen->immd_heap;
 
 		if (nouveau_resource_alloc(heap, p->immd_nr, p, &p->data[0])) {
 			while (heap->next && heap->size < p->immd_nr) {
@@ -4180,7 +4181,7 @@ nv50_program_validate_data(struct nv50_context *nv50, struct nv50_program *p)
 					 p->immd_nr, NV50_CB_PMISC);
 	}
 
-	assert(p->param_nr <= 512);
+	assert(p->param_nr <= 16384);
 
 	if (p->param_nr) {
 		unsigned cb;
diff --git a/src/gallium/drivers/nv50/nv50_screen.c b/src/gallium/drivers/nv50/nv50_screen.c
index ad17991be9..2dd1042424 100644
--- a/src/gallium/drivers/nv50/nv50_screen.c
+++ b/src/gallium/drivers/nv50/nv50_screen.c
@@ -190,9 +190,7 @@ nv50_screen_destroy(struct pipe_screen *pscreen)
 	nouveau_grobj_free(&screen->tesla);
 	nouveau_grobj_free(&screen->eng2d);
 	nouveau_grobj_free(&screen->m2mf);
-	nouveau_resource_destroy(&screen->immd_heap[0]);
-	nouveau_resource_destroy(&screen->parm_heap[0]);
-	nouveau_resource_destroy(&screen->parm_heap[1]);
+	nouveau_resource_destroy(&screen->immd_heap);
 	nouveau_screen_fini(&screen->base);
 	FREE(screen);
 }
@@ -242,7 +240,7 @@ nv50_screen_relocs(struct nv50_screen *screen)
 		OUT_RELOCh(chan, screen->constbuf_parm[i], 0, rl);
 		OUT_RELOCl(chan, screen->constbuf_parm[i], 0, rl);
 		OUT_RELOC (chan, screen->constbuf_parm[i],
-			   ((NV50_CB_PVP + i) << 16) | 0x0800, rl, 0, 0);
+			   ((NV50_CB_PVP + i) << 16) | 0x0000, rl, 0, 0);
 	}
 }
 
@@ -411,7 +409,7 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 	OUT_RING  (chan, (NV50_CB_AUX << 16) | 0x0200);
 
 	for (i = 0; i < 3; i++) {
-		ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, (256 * 4) * 4,
+		ret = nouveau_bo_new(dev, NOUVEAU_BO_VRAM, 0, (4096 * 4) * 4,
 				     &screen->constbuf_parm[i]);
 		if (ret) {
 			nv50_screen_destroy(pscreen);
@@ -420,14 +418,12 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
 		BEGIN_RING(chan, screen->tesla, NV50TCL_CB_DEF_ADDRESS_HIGH, 3);
 		OUT_RELOCh(chan, screen->constbuf_parm[i], 0, rl);
 		OUT_RELOCl(chan, screen->constbuf_parm[i], 0, rl);
-		OUT_RING  (chan, ((NV50_CB_PVP + i) << 16) | 0x0800);
+		/* CB_DEF_SET_SIZE value of 0x0000 means 65536 */
+		OUT_RING  (chan, ((NV50_CB_PVP + i) << 16) | 0x0000);
 	}
 
-	if (nouveau_resource_init(&screen->immd_heap[0], 0, 128) ||
-	    nouveau_resource_init(&screen->parm_heap[0], 0, 512) ||
-	    nouveau_resource_init(&screen->parm_heap[1], 0, 512))
-	{
-		NOUVEAU_ERR("Error initialising constant buffers.\n");
+	if (nouveau_resource_init(&screen->immd_heap, 0, 128)) {
+		NOUVEAU_ERR("Error initialising shader immediates heap.\n");
 		nv50_screen_destroy(pscreen);
 		return NULL;
 	}
diff --git a/src/gallium/drivers/nv50/nv50_screen.h b/src/gallium/drivers/nv50/nv50_screen.h
index 40ebbee72e..fbf15a7596 100644
--- a/src/gallium/drivers/nv50/nv50_screen.h
+++ b/src/gallium/drivers/nv50/nv50_screen.h
@@ -20,8 +20,7 @@ struct nv50_screen {
 	struct nouveau_bo *constbuf_misc[1];
 	struct nouveau_bo *constbuf_parm[PIPE_SHADER_TYPES];
 
-	struct nouveau_resource *immd_heap[1];
-	struct nouveau_resource *parm_heap[PIPE_SHADER_TYPES];
+	struct nouveau_resource *immd_heap;
 
 	struct pipe_resource *strm_vbuf[16];
 
diff --git a/src/gallium/drivers/nv50/nv50_surface.c b/src/gallium/drivers/nv50/nv50_surface.c
index c2d9e83526..d905d95354 100644
--- a/src/gallium/drivers/nv50/nv50_surface.c
+++ b/src/gallium/drivers/nv50/nv50_surface.c
@@ -31,6 +31,27 @@
 #include "util/u_tile.h"
 #include "util/u_format.h"
 
+/* return TRUE for formats that can be converted among each other by NV50_2D */
+static INLINE boolean
+nv50_2d_format_faithful(enum pipe_format format)
+{
+	switch (format) {
+	case PIPE_FORMAT_B8G8R8A8_UNORM:
+	case PIPE_FORMAT_B8G8R8X8_UNORM:
+	case PIPE_FORMAT_B8G8R8A8_SRGB:
+	case PIPE_FORMAT_B8G8R8X8_SRGB:
+	case PIPE_FORMAT_B5G6R5_UNORM:
+	case PIPE_FORMAT_B5G5R5A1_UNORM:
+	case PIPE_FORMAT_B10G10R10A2_UNORM:
+	case PIPE_FORMAT_R8_UNORM:
+	case PIPE_FORMAT_R32G32B32A32_FLOAT:
+	case PIPE_FORMAT_R32G32B32_FLOAT:
+		return TRUE;
+	default:
+		return FALSE;
+	}
+}
+
 static INLINE int
 nv50_format(enum pipe_format format)
 {
@@ -47,9 +68,12 @@ nv50_format(enum pipe_format format)
 		return NV50_2D_DST_FORMAT_R5G6B5_UNORM;
 	case PIPE_FORMAT_B5G5R5A1_UNORM:
 		return NV50_2D_DST_FORMAT_A1R5G5B5_UNORM;
+	case PIPE_FORMAT_B10G10R10A2_UNORM:
+		return NV50_2D_DST_FORMAT_A2R10G10B10_UNORM;
 	case PIPE_FORMAT_A8_UNORM:
 	case PIPE_FORMAT_I8_UNORM:
 	case PIPE_FORMAT_L8_UNORM:
+	case PIPE_FORMAT_R8_UNORM:
 		return NV50_2D_DST_FORMAT_R8_UNORM;
 	case PIPE_FORMAT_R32G32B32A32_FLOAT:
 		return NV50_2D_DST_FORMAT_R32G32B32A32_FLOAT;
@@ -178,7 +202,9 @@ nv50_surface_copy(struct pipe_context *pipe,
 	struct nv50_context *nv50 = nv50_context(pipe);
 	struct nv50_screen *screen = nv50->screen;
 
-	assert(src->format == dest->format);
+	assert((src->format == dest->format) ||
+	       (nv50_2d_format_faithful(src->format) &&
+		nv50_2d_format_faithful(dest->format)));
 
 	nv50_surface_do_copy(screen, dest, destx, desty, src, srcx,
 				     srcy, width, height);
diff --git a/src/gallium/drivers/r300/r300_blit.c b/src/gallium/drivers/r300/r300_blit.c
index 928ad300ee..85c2c14901 100644
--- a/src/gallium/drivers/r300/r300_blit.c
+++ b/src/gallium/drivers/r300/r300_blit.c
@@ -129,15 +129,16 @@ void r300_surface_copy(struct pipe_context* pipe,
     if (dst->texture->format != src->texture->format) {
         debug_printf("r300: Implementation error: Format mismatch in %s\n"
             "    : src: %s dst: %s\n", __FUNCTION__,
-            util_format_name(src->texture->format),
-            util_format_name(dst->texture->format));
+            util_format_short_name(src->texture->format),
+            util_format_short_name(dst->texture->format));
         debug_assert(0);
     }
 
     if (!pipe->screen->is_format_supported(pipe->screen,
                                            old_format, src->texture->target,
                                            PIPE_BIND_RENDER_TARGET |
-                                           PIPE_BIND_SAMPLER_VIEW, 0)) {
+                                           PIPE_BIND_SAMPLER_VIEW, 0) &&
+        util_format_is_plain(old_format)) {
         switch (util_format_get_blocksize(old_format)) {
             case 1:
                 new_format = PIPE_FORMAT_I8_UNORM;
@@ -154,7 +155,7 @@ void r300_surface_copy(struct pipe_context* pipe,
             default:
                 debug_printf("r300: surface_copy: Unhandled format: %s. Falling back to software.\n"
                              "r300: surface_copy: Software fallback doesn't work for tiled textures.\n",
-                             util_format_name(old_format));
+                             util_format_short_name(old_format));
         }
     }
 
diff --git a/src/gallium/drivers/r300/r300_context.c b/src/gallium/drivers/r300/r300_context.c
index 4f721ebb59..e84bce0010 100644
--- a/src/gallium/drivers/r300/r300_context.c
+++ b/src/gallium/drivers/r300/r300_context.c
@@ -76,7 +76,6 @@ static void r300_destroy_context(struct pipe_context* context)
     FREE(r300->rs_block_state.state);
     FREE(r300->scissor_state.state);
     FREE(r300->textures_state.state);
-    FREE(r300->vap_output_state.state);
     FREE(r300->viewport_state.state);
     FREE(r300->ztop_state.state);
     FREE(r300->fs_constants.state);
@@ -125,7 +124,6 @@ static void r300_setup_atoms(struct r300_context* r300)
     R300_INIT_ATOM(viewport_state, 9);
     R300_INIT_ATOM(rs_block_state, 0);
     R300_INIT_ATOM(vertex_stream_state, 0);
-    R300_INIT_ATOM(vap_output_state, 6);
     R300_INIT_ATOM(pvs_flush, 2);
     R300_INIT_ATOM(vs_state, 0);
     R300_INIT_ATOM(vs_constants, 0);
@@ -149,7 +147,6 @@ static void r300_setup_atoms(struct r300_context* r300)
     r300->rs_block_state.state = CALLOC_STRUCT(r300_rs_block);
     r300->scissor_state.state = CALLOC_STRUCT(pipe_scissor_state);
     r300->textures_state.state = CALLOC_STRUCT(r300_textures_state);
-    r300->vap_output_state.state = CALLOC_STRUCT(r300_vap_output_state);
     r300->viewport_state.state = CALLOC_STRUCT(r300_viewport_state);
     r300->ztop_state.state = CALLOC_STRUCT(r300_ztop_state);
     r300->fs_constants.state = CALLOC_STRUCT(r300_constant_buffer);
diff --git a/src/gallium/drivers/r300/r300_context.h b/src/gallium/drivers/r300/r300_context.h
index a05bf3ce09..e9c8fcdc15 100644
--- a/src/gallium/drivers/r300/r300_context.h
+++ b/src/gallium/drivers/r300/r300_context.h
@@ -119,6 +119,10 @@ struct r300_rs_state {
 };
 
 struct r300_rs_block {
+    uint32_t vap_vtx_state_cntl;  /* R300_VAP_VTX_STATE_CNTL: 0x2180 */
+    uint32_t vap_vsm_vtx_assm;    /* R300_VAP_VSM_VTX_ASSM: 0x2184 */
+    uint32_t vap_out_vtx_fmt[2];  /* R300_VAP_OUTPUT_VTX_FMT_[0-1]: 0x2090 */
+
     uint32_t ip[8]; /* R300_RS_IP_[0-7], R500_RS_IP_[0-7] */
     uint32_t count; /* R300_RS_COUNT */
     uint32_t inst_count; /* R300_RS_INST_COUNT */
@@ -190,12 +194,6 @@ struct r300_vertex_stream_state {
     unsigned count;
 };
 
-struct r300_vap_output_state {
-    uint32_t vap_vtx_state_cntl;  /* R300_VAP_VTX_STATE_CNTL: 0x2180 */
-    uint32_t vap_vsm_vtx_assm;    /* R300_VAP_VSM_VTX_ASSM: 0x2184 */
-    uint32_t vap_out_vtx_fmt[2];  /* R300_VAP_OUTPUT_VTX_FMT_[0-1]: 0x2090 */
-};
-
 struct r300_viewport_state {
     float xscale;         /* R300_VAP_VPORT_XSCALE:  0x2098 */
     float xoffset;        /* R300_VAP_VPORT_XOFFSET: 0x209c */
@@ -381,7 +379,7 @@ struct r300_context {
     struct r300_atom query_start;
     /* Rasterizer state. */
     struct r300_atom rs_state;
-    /* RS block state. */
+    /* RS block state + VAP (vertex shader) output mapping state. */
     struct r300_atom rs_block_state;
     /* Scissor state. */
     struct r300_atom scissor_state;
@@ -389,8 +387,6 @@ struct r300_context {
     struct r300_atom textures_state;
     /* Vertex stream formatting state. */
     struct r300_atom vertex_stream_state;
-    /* VAP (vertex shader) output mapping state. */
-    struct r300_atom vap_output_state;
     /* Vertex shader. */
     struct r300_atom vs_state;
     /* Vertex shader constant buffer. */
@@ -424,6 +420,9 @@ struct r300_context {
 
     struct pipe_viewport_state viewport;
 
+    /* Stream locations for SWTCL. */
+    int stream_loc_notcl[16];
+
     /* Flag indicating whether or not the HW is dirty. */
     uint32_t dirty_hw;
     /* Whether polygon offset is enabled. */
diff --git a/src/gallium/drivers/r300/r300_debug.c b/src/gallium/drivers/r300/r300_debug.c
index 4c2836f36a..8eb321fa08 100644
--- a/src/gallium/drivers/r300/r300_debug.c
+++ b/src/gallium/drivers/r300/r300_debug.c
@@ -37,6 +37,7 @@ static struct debug_option debug_options[] = {
     { "cs", DBG_CS, "Command submissions (for debugging)" },
     { "draw", DBG_DRAW, "Draw and emit (for debugging)" },
     { "tex", DBG_TEX, "Textures (for debugging)" },
+    { "texalloc", DBG_TEXALLOC, "Texture allocation (for debugging)" },
     { "fall", DBG_FALL, "Fallbacks (for debugging)" },
     { "rs", DBG_RS, "Rasterizer (for debugging)" },
     { "anisohq", DBG_ANISOHQ, "High quality anisotropic filtering (for benchmarking)" },
diff --git a/src/gallium/drivers/r300/r300_emit.c b/src/gallium/drivers/r300/r300_emit.c
index 92b7517b8d..23bbc6a99c 100644
--- a/src/gallium/drivers/r300/r300_emit.c
+++ b/src/gallium/drivers/r300/r300_emit.c
@@ -717,6 +717,13 @@ void r300_emit_rs_block_state(struct r300_context* r300,
     DBG(r300, DBG_DRAW, "r300: RS emit:\n");
 
     BEGIN_CS(size);
+    OUT_CS_REG_SEQ(R300_VAP_VTX_STATE_CNTL, 2);
+    OUT_CS(rs->vap_vtx_state_cntl);
+    OUT_CS(rs->vap_vsm_vtx_assm);
+    OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2);
+    OUT_CS(rs->vap_out_vtx_fmt[0]);
+    OUT_CS(rs->vap_out_vtx_fmt[1]);
+
     if (r300->screen->caps.is_r500) {
         OUT_CS_REG_SEQ(R500_RS_IP_0, count);
     } else {
@@ -901,25 +908,6 @@ void r300_emit_vertex_stream_state(struct r300_context* r300,
     END_CS;
 }
 
-void r300_emit_vap_output_state(struct r300_context* r300,
-                               unsigned size, void* state)
-{
-    struct r300_vap_output_state *vap_out_state =
-        (struct r300_vap_output_state*)state;
-    CS_LOCALS(r300);
-
-    DBG(r300, DBG_DRAW, "r300: VAP emit:\n");
-
-    BEGIN_CS(size);
-    OUT_CS_REG_SEQ(R300_VAP_VTX_STATE_CNTL, 2);
-    OUT_CS(vap_out_state->vap_vtx_state_cntl);
-    OUT_CS(vap_out_state->vap_vsm_vtx_assm);
-    OUT_CS_REG_SEQ(R300_VAP_OUTPUT_VTX_FMT_0, 2);
-    OUT_CS(vap_out_state->vap_out_vtx_fmt[0]);
-    OUT_CS(vap_out_state->vap_out_vtx_fmt[1]);
-    END_CS;
-}
-
 void r300_emit_pvs_flush(struct r300_context* r300, unsigned size, void* state)
 {
     CS_LOCALS(r300);
diff --git a/src/gallium/drivers/r300/r300_emit.h b/src/gallium/drivers/r300/r300_emit.h
index c3eb195d4e..3c0edf6fdc 100644
--- a/src/gallium/drivers/r300/r300_emit.h
+++ b/src/gallium/drivers/r300/r300_emit.h
@@ -81,9 +81,6 @@ void r300_emit_vertex_buffer(struct r300_context* r300);
 void r300_emit_vertex_stream_state(struct r300_context* r300,
                                    unsigned size, void* state);
 
-void r300_emit_vap_output_state(struct r300_context* r300,
-                               unsigned size, void* state);
-
 void r300_emit_vs_constants(struct r300_context* r300,
                             unsigned size, void *state);
 
diff --git a/src/gallium/drivers/r300/r300_screen.h b/src/gallium/drivers/r300/r300_screen.h
index 735c233c9e..d58aa138a7 100644
--- a/src/gallium/drivers/r300/r300_screen.h
+++ b/src/gallium/drivers/r300/r300_screen.h
@@ -73,6 +73,7 @@ static INLINE struct r300_screen* r300_screen(struct pipe_screen* screen) {
 #define DBG_NO_IMMD 0x0000200
 #define DBG_STATS   0x0000400
 #define DBG_RS      0x0000800
+#define DBG_TEXALLOC 0x0001000
 /*@}*/
 
 static INLINE boolean SCREEN_DBG_ON(struct r300_screen * screen, unsigned flags)
diff --git a/src/gallium/drivers/r300/r300_state.c b/src/gallium/drivers/r300/r300_state.c
index 9319dadfd1..446422ca0f 100644
--- a/src/gallium/drivers/r300/r300_state.c
+++ b/src/gallium/drivers/r300/r300_state.c
@@ -538,46 +538,12 @@ static void r300_set_stencil_ref(struct pipe_context* pipe,
 }
 
 /* This switcheroo is needed just because of goddamned MACRO_SWITCH. */
-static void r300_fb_update_tiling_flags(struct r300_context *r300,
+static void r300_fb_set_tiling_flags(struct r300_context *r300,
                                const struct pipe_framebuffer_state *old_state,
                                const struct pipe_framebuffer_state *new_state)
 {
     struct r300_texture *tex;
-    unsigned i, j, level;
-
-    /* Reset tiling flags for old surfaces to default values. */
-    for (i = 0; i < old_state->nr_cbufs; i++) {
-        for (j = 0; j < new_state->nr_cbufs; j++) {
-            if (old_state->cbufs[i]->texture == new_state->cbufs[j]->texture) {
-                break;
-            }
-        }
-        /* If not binding the surface again... */
-        if (j != new_state->nr_cbufs) {
-            continue;
-        }
-
-        tex = r300_texture(old_state->cbufs[i]->texture);
-
-        if (tex) {
-            r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
-                                            tex->pitch[0],
-                                            tex->microtile,
-                                            tex->macrotile);
-        }
-    }
-    if (old_state->zsbuf &&
-        (!new_state->zsbuf ||
-         old_state->zsbuf->texture != new_state->zsbuf->texture)) {
-        tex = r300_texture(old_state->zsbuf->texture);
-
-        if (tex) {
-            r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
-                                            tex->pitch[0],
-                                            tex->microtile,
-                                            tex->macrotile);
-        }
-    }
+    unsigned i, level;
 
     /* Set tiling flags for new surfaces. */
     for (i = 0; i < new_state->nr_cbufs; i++) {
@@ -585,7 +551,7 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
         level = new_state->cbufs[i]->level;
 
         r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
-                                        tex->pitch[level],
+                                        tex->pitch[0],
                                         tex->microtile,
                                         tex->mip_macrotile[level]);
     }
@@ -594,7 +560,7 @@ static void r300_fb_update_tiling_flags(struct r300_context *r300,
         level = new_state->zsbuf->level;
 
         r300->rws->buffer_set_tiling(r300->rws, tex->buffer,
-                                        tex->pitch[level],
+                                        tex->pitch[0],
                                         tex->microtile,
                                         tex->mip_macrotile[level]);
     }
@@ -644,7 +610,8 @@ static void
         r300->dsa_state.dirty = TRUE;
     }
 
-    r300_fb_update_tiling_flags(r300, r300->fb_state.state, state);
+    /* The tiling flags are dependent on the surface miplevel, unfortunately. */
+    r300_fb_set_tiling_flags(r300, r300->fb_state.state, state);
 
     memcpy(r300->fb_state.state, state, sizeof(struct pipe_framebuffer_state));
 
@@ -719,10 +686,6 @@ static void r300_bind_fs_state(struct pipe_context* pipe, void* shader)
     r300_mark_fs_code_dirty(r300);
 
     r300->rs_block_state.dirty = TRUE; /* Will be updated before the emission. */
-
-    if (r300->vs_state.state && r300_vertex_shader_setup_wpos(r300)) {
-        r300->vap_output_state.dirty = TRUE;
-    }
 }
 
 /* Delete fragment shader state. */
@@ -853,11 +816,8 @@ static void* r300_create_rs_state(struct pipe_context* pipe,
 
     rs->clip_rule = state->scissor ? 0xAAAA : 0xFFFF;
 
-    /* XXX Disable point sprites until we know what's wrong with them. */
-    rs->rs.sprite_coord_enable = 0;
-
     /* Point sprites */
-    if (rs->rs.sprite_coord_enable) {
+    if (state->sprite_coord_enable) {
         rs->stuffing_enable = R300_GB_POINT_STUFF_ENABLE;
 	for (i = 0; i < 8; i++) {
 	    if (state->sprite_coord_enable & (1 << i))
@@ -1357,7 +1317,7 @@ static void* r300_create_vertex_elements_state(struct pipe_context* pipe,
                     /* XXX Shouldn't we align the format? */
                     fprintf(stderr, "r300_create_vertex_elements_state: "
                             "Unaligned format %s:%i isn't supported\n",
-                            util_format_name(*format), size);
+                            util_format_short_name(*format), size);
                     assert(0);
                     abort();
                 }
@@ -1428,14 +1388,6 @@ static void r300_bind_vs_state(struct pipe_context* pipe, void* shader)
     }
     r300->vs_state.state = vs;
 
-    // VS output mapping for HWTCL or stream mapping for SWTCL to the RS block
-    if (r300->fs.state) {
-        r300_vertex_shader_setup_wpos(r300);
-    }
-    memcpy(r300->vap_output_state.state, &vs->vap_out,
-           sizeof(struct r300_vap_output_state));
-    r300->vap_output_state.dirty = TRUE;
-
     /* The majority of the RS block bits is dependent on the vertex shader. */
     r300->rs_block_state.dirty = TRUE; /* Will be updated before the emission. */
 
diff --git a/src/gallium/drivers/r300/r300_state_derived.c b/src/gallium/drivers/r300/r300_state_derived.c
index ccc4b583a6..e3adace0fa 100644
--- a/src/gallium/drivers/r300/r300_state_derived.c
+++ b/src/gallium/drivers/r300/r300_state_derived.c
@@ -115,12 +115,11 @@ static void r300_draw_emit_all_attribs(struct r300_context* r300)
 static void r300_swtcl_vertex_psc(struct r300_context *r300)
 {
     struct r300_vertex_stream_state *vstream = r300->vertex_stream_state.state;
-    struct r300_vertex_shader* vs = r300->vs_state.state;
-    struct vertex_info* vinfo = &r300->vertex_info;
+        struct vertex_info* vinfo = &r300->vertex_info;
     uint16_t type, swizzle;
     enum pipe_format format;
     unsigned i, attrib_count;
-    int* vs_output_tab = vs->stream_loc_notcl;
+    int* vs_output_tab = r300->stream_loc_notcl;
 
     /* XXX hax */
     memset(vstream, 0, sizeof(struct r300_vertex_stream_state));
@@ -269,21 +268,29 @@ static void r500_rs_tex_write(struct r300_rs_block* rs, int id, int fp_offset)
 
 /* Set up the RS block.
  *
- * This is the part of the chipset that actually does the rasterization
- * of vertices into fragments. This is also the part of the chipset that
- * locks up if any part of it is even slightly wrong. */
-static void r300_update_rs_block(struct r300_context* r300,
-                                 struct r300_shader_semantics* vs_outputs,
-                                 struct r300_shader_semantics* fs_inputs)
+ * This is the part of the chipset that is responsible for linking vertex
+ * and fragment shaders and stuffed texture coordinates.
+ *
+ * The rasterizer reads data from VAP, which produces vertex shader outputs,
+ * and GA, which produces stuffed texture coordinates. VAP outputs have
+ * precedence over GA. All outputs must be rasterized otherwise it locks up.
+ * If there are more outputs rasterized than is set in VAP/GA, it locks up
+ * too. The funky part is that this info has been pretty much obtained by trial
+ * and error. */
+static void r300_update_rs_block(struct r300_context *r300)
 {
-    struct r300_rs_block rs = { { 0 } };
-    int i, col_count = 0, tex_count = 0, fp_offset = 0, count;
+    struct r300_vertex_shader *vs = r300->vs_state.state;
+    struct r300_shader_semantics *vs_outputs = &vs->outputs;
+    struct r300_shader_semantics *fs_inputs = &r300_fs(r300)->shader->inputs;
+    struct r300_rs_block rs = {0};
+    int i, col_count = 0, tex_count = 0, fp_offset = 0, count, loc = 0;
     void (*rX00_rs_col)(struct r300_rs_block*, int, int, enum r300_rs_swizzle);
     void (*rX00_rs_col_write)(struct r300_rs_block*, int, int);
     void (*rX00_rs_tex)(struct r300_rs_block*, int, int, enum r300_rs_swizzle);
     void (*rX00_rs_tex_write)(struct r300_rs_block*, int, int);
     boolean any_bcolor_used = vs_outputs->bcolor[0] != ATTR_UNUSED ||
                               vs_outputs->bcolor[1] != ATTR_UNUSED;
+    int *stream_loc_notcl = r300->stream_loc_notcl;
 
     if (r300->screen->caps.is_r500) {
         rX00_rs_col       = r500_rs_col;
@@ -297,15 +304,31 @@ static void r300_update_rs_block(struct r300_context* r300,
         rX00_rs_tex_write = r300_rs_tex_write;
     }
 
-    /* Rasterize colors. */
+    /* The position is always present in VAP. */
+    rs.vap_vsm_vtx_assm |= R300_INPUT_CNTL_POS;
+    rs.vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
+    stream_loc_notcl[loc++] = 0;
+
+    /* Set up the point size in VAP. */
+    if (vs_outputs->psize != ATTR_UNUSED) {
+        rs.vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT;
+        stream_loc_notcl[loc++] = 1;
+    }
+
+    /* Set up and rasterize colors. */
     for (i = 0; i < ATTR_COLOR_COUNT; i++) {
         if (vs_outputs->color[i] != ATTR_UNUSED || any_bcolor_used ||
             vs_outputs->color[1] != ATTR_UNUSED) {
-            /* Always rasterize if it's written by the VS,
-             * otherwise it locks up. */
+            /* Set up the color in VAP. */
+            rs.vap_vsm_vtx_assm |= R300_INPUT_CNTL_COLOR;
+            rs.vap_out_vtx_fmt[0] |=
+                    R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i;
+            stream_loc_notcl[loc++] = 2 + i;
+
+            /* Rasterize it. */
             rX00_rs_col(&rs, col_count, col_count, SWIZ_XYZW);
 
-            /* Write it to the FS input register if it's used by the FS. */
+            /* Write it to the FS input register if it's needed by the FS. */
             if (fs_inputs->color[i] != ATTR_UNUSED) {
                 rX00_rs_col_write(&rs, col_count, fp_offset);
                 fp_offset++;
@@ -328,17 +351,33 @@ static void r300_update_rs_block(struct r300_context* r300,
         }
     }
 
+    /* Set up back-face colors. The rasterizer will do the color selection
+     * automatically. */
+    if (any_bcolor_used) {
+        for (i = 0; i < ATTR_COLOR_COUNT; i++) {
+            rs.vap_vsm_vtx_assm |= R300_INPUT_CNTL_COLOR;
+            rs.vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << (2+i);
+            stream_loc_notcl[loc++] = 4 + i;
+        }
+    }
+
     /* Rasterize texture coordinates. */
-    for (i = 0; i < ATTR_GENERIC_COUNT; i++) {
+    for (i = 0; i < ATTR_GENERIC_COUNT && tex_count < 8; i++) {
 	bool sprite_coord = !!(r300->sprite_coord_enable & (1 << i));
 
         if (vs_outputs->generic[i] != ATTR_UNUSED || sprite_coord) {
-            /* Always rasterize if it's written by the VS,
-             * otherwise it locks up. */
+            if (!sprite_coord) {
+                /* Set up the texture coordinates in VAP. */
+                rs.vap_vsm_vtx_assm |= (R300_INPUT_CNTL_TC0 << tex_count);
+                rs.vap_out_vtx_fmt[1] |= (4 << (3 * tex_count));
+                stream_loc_notcl[loc++] = 6 + tex_count;
+            }
+
+            /* Rasterize it. */
             rX00_rs_tex(&rs, tex_count, tex_count,
 			sprite_coord ? SWIZ_XY01 : SWIZ_XYZW);
 
-            /* Write it to the FS input register if it's used by the FS. */
+            /* Write it to the FS input register if it's needed by the FS. */
             if (fs_inputs->generic[i] != ATTR_UNUSED) {
                 rX00_rs_tex_write(&rs, tex_count, fp_offset);
                 fp_offset++;
@@ -365,12 +404,16 @@ static void r300_update_rs_block(struct r300_context* r300,
     }
 
     /* Rasterize fog coordinates. */
-    if (vs_outputs->fog != ATTR_UNUSED) {
-        /* Always rasterize if it's written by the VS,
-         * otherwise it locks up. */
+    if (vs_outputs->fog != ATTR_UNUSED && tex_count < 8) {
+        /* Set up the fog coordinates in VAP. */
+        rs.vap_vsm_vtx_assm |= (R300_INPUT_CNTL_TC0 << tex_count);
+        rs.vap_out_vtx_fmt[1] |= (4 << (3 * tex_count));
+        stream_loc_notcl[loc++] = 6 + tex_count;
+
+        /* Rasterize it. */
         rX00_rs_tex(&rs, tex_count, tex_count, SWIZ_X001);
 
-        /* Write it to the FS input register if it's used by the FS. */
+        /* Write it to the FS input register if it's needed by the FS. */
         if (fs_inputs->fog != ATTR_UNUSED) {
             rX00_rs_tex_write(&rs, tex_count, fp_offset);
             fp_offset++;
@@ -391,9 +434,17 @@ static void r300_update_rs_block(struct r300_context* r300,
     }
 
     /* Rasterize WPOS. */
-    /* If the FS doesn't need it, it's not written by the VS. */
-    if (vs_outputs->wpos != ATTR_UNUSED && fs_inputs->wpos != ATTR_UNUSED) {
+    /* Don't set it in VAP if the FS doesn't need it. */
+    if (fs_inputs->wpos != ATTR_UNUSED && tex_count < 8) {
+        /* Set up the WPOS coordinates in VAP. */
+        rs.vap_vsm_vtx_assm |= (R300_INPUT_CNTL_TC0 << tex_count);
+        rs.vap_out_vtx_fmt[1] |= (4 << (3 * tex_count));
+        stream_loc_notcl[loc++] = 6 + tex_count;
+
+        /* Rasterize it. */
         rX00_rs_tex(&rs, tex_count, tex_count, SWIZ_XYZW);
+
+        /* Write it to the FS input register. */
         rX00_rs_tex_write(&rs, tex_count, fp_offset);
 
         DBG(r300, DBG_RS, "r300: Rasterized WPOS written to FS.\n");
@@ -402,6 +453,11 @@ static void r300_update_rs_block(struct r300_context* r300,
         tex_count++;
     }
 
+    /* Invalidate the rest of the no-TCL (GA) stream locations. */
+    for (; loc < 16;) {
+        stream_loc_notcl[loc++] = -1;
+    }
+
     /* Rasterize at least one color, or bad things happen. */
     if (col_count == 0 && tex_count == 0) {
         rX00_rs_col(&rs, 0, 0, SWIZ_0001);
@@ -422,18 +478,10 @@ static void r300_update_rs_block(struct r300_context* r300,
     /* Now, after all that, see if we actually need to update the state. */
     if (memcmp(r300->rs_block_state.state, &rs, sizeof(struct r300_rs_block))) {
         memcpy(r300->rs_block_state.state, &rs, sizeof(struct r300_rs_block));
-        r300->rs_block_state.size = 5 + count*2;
+        r300->rs_block_state.size = 11 + count*2;
     }
 }
 
-/* Update the shader-dependant states. */
-static void r300_update_derived_shader_state(struct r300_context* r300)
-{
-    struct r300_vertex_shader* vs = r300->vs_state.state;
-
-    r300_update_rs_block(r300, &vs->outputs, &r300_fs(r300)->shader->inputs);
-}
-
 static void r300_merge_textures_and_samplers(struct r300_context* r300)
 {
     struct r300_textures_state *state =
@@ -529,7 +577,7 @@ void r300_update_derived_state(struct r300_context* r300)
     }
 
     if (r300->rs_block_state.dirty) {
-        r300_update_derived_shader_state(r300);
+        r300_update_rs_block(r300);
     }
 
     if (r300->draw) {
diff --git a/src/gallium/drivers/r300/r300_state_inlines.h b/src/gallium/drivers/r300/r300_state_inlines.h
index fcbdb91b67..715601042f 100644
--- a/src/gallium/drivers/r300/r300_state_inlines.h
+++ b/src/gallium/drivers/r300/r300_state_inlines.h
@@ -371,7 +371,7 @@ r300_translate_vertex_data_type(enum pipe_format format) {
     desc = util_format_description(format);
 
     if (desc->layout != UTIL_FORMAT_LAYOUT_PLAIN) {
-        fprintf(stderr, "r300: Bad format %s in %s:%d\n", util_format_name(format),
+        fprintf(stderr, "r300: Bad format %s in %s:%d\n", util_format_short_name(format),
             __FUNCTION__, __LINE__);
         assert(0);
         abort();
@@ -394,7 +394,7 @@ r300_translate_vertex_data_type(enum pipe_format format) {
                     break;
                 default:
                     fprintf(stderr, "r300: Bad format %s in %s:%d\n",
-                        util_format_name(format), __FUNCTION__, __LINE__);
+                        util_format_short_name(format), __FUNCTION__, __LINE__);
                     assert(0);
                     abort();
             }
@@ -416,7 +416,7 @@ r300_translate_vertex_data_type(enum pipe_format format) {
                     break;
                 default:
                     fprintf(stderr, "r300: Bad format %s in %s:%d\n",
-                        util_format_name(format), __FUNCTION__, __LINE__);
+                        util_format_short_name(format), __FUNCTION__, __LINE__);
                     fprintf(stderr, "r300: desc->channel[0].size == %d\n",
                         desc->channel[0].size);
                     assert(0);
@@ -425,7 +425,7 @@ r300_translate_vertex_data_type(enum pipe_format format) {
             break;
         default:
             fprintf(stderr, "r300: Bad format %s in %s:%d\n",
-                util_format_name(format), __FUNCTION__, __LINE__);
+                util_format_short_name(format), __FUNCTION__, __LINE__);
             assert(0);
             abort();
     }
@@ -449,7 +449,7 @@ r300_translate_vertex_data_swizzle(enum pipe_format format) {
 
     if (desc->layout != UTIL_FORMAT_LAYOUT_PLAIN) {
         fprintf(stderr, "r300: Bad format %s in %s:%d\n",
-            util_format_name(format), __FUNCTION__, __LINE__);
+            util_format_short_name(format), __FUNCTION__, __LINE__);
         return 0;
     }
 
diff --git a/src/gallium/drivers/r300/r300_texture.c b/src/gallium/drivers/r300/r300_texture.c
index 69e6a12445..d6f629cf9c 100644
--- a/src/gallium/drivers/r300/r300_texture.c
+++ b/src/gallium/drivers/r300/r300_texture.c
@@ -46,18 +46,6 @@ static const unsigned microblock_table[5][3][2] = {
     {{ 2, 1}, {0, 0}, {0, 0}}  /* 128 bits per pixel */
 };
 
-/* Return true for non-compressed and non-YUV formats. */
-static boolean r300_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;
-}
-
 /* Translate a pipe_format into a useful texture format for sampling.
  *
  * Some special formats are translated directly using R300_EASY_TX_FORMAT,
@@ -585,9 +573,6 @@ static void r300_texture_setup_immutable_state(struct r300_screen* screen,
 
     f->tile_config = R300_TXO_MACRO_TILE(tex->macrotile) |
                      R300_TXO_MICRO_TILE(tex->microtile);
-
-    SCREEN_DBG(screen, DBG_TEX, "r300: Set texture state (%dx%d, %d levels)\n",
-               pt->width0, pt->height0, pt->last_level);
 }
 
 static void r300_texture_setup_fb_state(struct r300_screen* screen,
@@ -622,8 +607,8 @@ void r300_texture_reinterpret_format(struct pipe_screen *screen,
 {
     struct r300_screen *r300screen = r300_screen(screen);
 
-    SCREEN_DBG(r300screen, DBG_TEX, "r300: Reinterpreting format: %s -> %s\n",
-               util_format_name(tex->format), util_format_name(new_format));
+    SCREEN_DBG(r300screen, DBG_TEX, "r300: texture_reinterpret_format: %s -> %s\n",
+               util_format_short_name(tex->format), util_format_short_name(new_format));
 
     tex->format = new_format;
 
@@ -714,7 +699,7 @@ unsigned r300_texture_get_stride(struct r300_screen* screen,
 
     width = u_minify(tex->b.b.width0, level);
 
-    if (r300_format_is_plain(tex->b.b.format)) {
+    if (util_format_is_plain(tex->b.b.format)) {
         tile_width = r300_texture_get_tile_size(tex, TILE_WIDTH,
                                                 tex->mip_macrotile[level]);
         width = align(width, tile_width);
@@ -732,7 +717,7 @@ static unsigned r300_texture_get_nblocksy(struct r300_texture* tex,
 
     height = u_minify(tex->b.b.height0, level);
 
-    if (r300_format_is_plain(tex->b.b.format)) {
+    if (util_format_is_plain(tex->b.b.format)) {
         tile_height = r300_texture_get_tile_size(tex, TILE_HEIGHT,
                                                  tex->mip_macrotile[level]);
         height = align(height, tile_height);
@@ -774,8 +759,8 @@ static void r300_setup_miptree(struct r300_screen* screen,
     unsigned stride, size, layer_size, nblocksy, i;
     boolean rv350_mode = screen->caps.is_rv350;
 
-    SCREEN_DBG(screen, DBG_TEX, "r300: Making miptree for texture, format %s\n",
-               util_format_name(base->format));
+    SCREEN_DBG(screen, DBG_TEXALLOC, "r300: Making miptree for texture, format %s\n",
+               util_format_short_name(base->format));
 
     for (i = 0; i <= base->last_level; i++) {
         /* Let's see if this miplevel can be macrotiled. */
@@ -801,7 +786,7 @@ static void r300_setup_miptree(struct r300_screen* screen,
         tex->hwpitch[i] =
                 tex->pitch[i] * util_format_get_blockwidth(base->format);
 
-        SCREEN_DBG(screen, DBG_TEX, "r300: Texture miptree: Level %d "
+        SCREEN_DBG(screen, DBG_TEXALLOC, "r300: Texture miptree: Level %d "
                 "(%dx%dx%d px, pitch %d bytes) %d bytes total, macrotiled %s\n",
                 i, u_minify(base->width0, i), u_minify(base->height0, i),
                 u_minify(base->depth0, i), stride, tex->size,
@@ -825,7 +810,7 @@ static void r300_setup_tiling(struct pipe_screen *screen,
     boolean is_zb = util_format_is_depth_or_stencil(format);
     boolean dbg_no_tiling = SCREEN_DBG_ON(r300_screen(screen), DBG_NO_TILING);
 
-    if (!r300_format_is_plain(format)) {
+    if (!util_format_is_plain(format)) {
         return;
     }
 
@@ -927,6 +912,17 @@ struct pipe_resource* r300_texture_create(struct pipe_screen* screen,
         return NULL;
     }
 
+    /* Refuse to create a texture with size 0. */
+    if (!base->width0 ||
+        (!base->height0 && (base->target == PIPE_TEXTURE_2D ||
+                            base->target == PIPE_TEXTURE_CUBE)) ||
+        (!base->depth0 && base->target == PIPE_TEXTURE_3D)) {
+        fprintf(stderr, "r300: texture_create: "
+                "Got invalid texture dimensions: %ix%ix%i\n",
+                base->width0, base->height0, base->depth0);
+        return NULL;
+    }
+
     tex->b.b = *base;
     tex->b.vtbl = &r300_texture_vtbl;
     pipe_reference_init(&tex->b.b.reference, 1);
@@ -942,6 +938,15 @@ struct pipe_resource* r300_texture_create(struct pipe_screen* screen,
     r300_texture_setup_immutable_state(rscreen, tex);
     r300_texture_setup_fb_state(rscreen, tex);
 
+    SCREEN_DBG(rscreen, DBG_TEX,
+               "r300: texture_create: Macro: %s, Micro: %s, Pitch: %i, "
+               "Dim: %ix%ix%i, LastLevel: %i, Format: %s\n",
+               tex->macrotile ? "YES" : " NO",
+               tex->microtile ? "YES" : " NO",
+               tex->hwpitch[0],
+               base->width0, base->height0, base->depth0, base->last_level,
+               util_format_short_name(base->format));
+
     tex->buffer = rws->buffer_create(rws, 2048,
 				     PIPE_BIND_SAMPLER_VIEW, /* XXX */
 				     tex->size);
@@ -1039,6 +1044,14 @@ r300_texture_from_handle(struct pipe_screen* screen,
 
     rws->buffer_get_tiling(rws, buffer, &tex->microtile, &tex->macrotile);
     r300_setup_flags(tex);
+    SCREEN_DBG(rscreen, DBG_TEX,
+               "r300: texture_from_handle: Macro: %s, Micro: %s, "
+               "Pitch: % 4i, Dim: %ix%i, Format: %s\n",
+               tex->macrotile ? "YES" : " NO",
+               tex->microtile ? "YES" : " NO",
+               stride / util_format_get_blocksize(base->format),
+               base->width0, base->height0,
+               util_format_short_name(base->format));
 
     /* Enforce microtiled zbuffer. */
     override_zb_flags = util_format_is_depth_or_stencil(base->format) &&
diff --git a/src/gallium/drivers/r300/r300_vs.c b/src/gallium/drivers/r300/r300_vs.c
index bfab9c3b01..f3186431e1 100644
--- a/src/gallium/drivers/r300/r300_vs.c
+++ b/src/gallium/drivers/r300/r300_vs.c
@@ -22,7 +22,6 @@
  * USE OR OTHER DEALINGS IN THE SOFTWARE. */
 
 #include "r300_vs.h"
-#include "r300_fs.h"
 
 #include "r300_context.h"
 #include "r300_screen.h"
@@ -94,94 +93,6 @@ static void r300_shader_read_vs_outputs(
     vs_outputs->wpos = i;
 }
 
-/* This function sets up:
- * - VAP mapping, which maps VS registers to output semantics and
- *   at the same time it indicates which attributes are enabled and should
- *   be rasterized.
- * - Stream mapping to VS outputs if TCL is not present. */
-static void r300_init_vs_output_mapping(struct r300_vertex_shader* vs)
-{
-    struct r300_shader_semantics* vs_outputs = &vs->outputs;
-    struct r300_vap_output_state *vap_out = &vs->vap_out;
-    int *stream_loc = vs->stream_loc_notcl;
-    int i, gen_count, tabi = 0;
-    boolean any_bcolor_used = vs_outputs->bcolor[0] != ATTR_UNUSED ||
-                              vs_outputs->bcolor[1] != ATTR_UNUSED;
-
-    vap_out->vap_vtx_state_cntl = 0x5555; /* XXX this is classic Mesa bonghits */
-
-    /* Position. */
-    if (vs_outputs->pos != ATTR_UNUSED) {
-        vap_out->vap_vsm_vtx_assm |= R300_INPUT_CNTL_POS;
-        vap_out->vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT;
-
-        stream_loc[tabi++] = 0;
-    } else {
-        assert(0);
-    }
-
-    /* Point size. */
-    if (vs_outputs->psize != ATTR_UNUSED) {
-        vap_out->vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__PT_SIZE_PRESENT;
-
-        stream_loc[tabi++] = 1;
-    }
-
-    /* Colors. */
-    for (i = 0; i < ATTR_COLOR_COUNT; i++) {
-        if (vs_outputs->color[i] != ATTR_UNUSED || any_bcolor_used ||
-            vs_outputs->color[1] != ATTR_UNUSED) {
-            vap_out->vap_vsm_vtx_assm |= R300_INPUT_CNTL_COLOR;
-            vap_out->vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << i;
-
-            stream_loc[tabi++] = 2 + i;
-        }
-    }
-
-    /* Back-face colors. */
-    if (any_bcolor_used) {
-        for (i = 0; i < ATTR_COLOR_COUNT; i++) {
-            vap_out->vap_vsm_vtx_assm |= R300_INPUT_CNTL_COLOR;
-            vap_out->vap_out_vtx_fmt[0] |= R300_VAP_OUTPUT_VTX_FMT_0__COLOR_0_PRESENT << (2+i);
-
-            stream_loc[tabi++] = 4 + i;
-        }
-    }
-
-    /* Texture coordinates. */
-    gen_count = 0;
-    for (i = 0; i < ATTR_GENERIC_COUNT && gen_count < 8; i++) {
-        if (vs_outputs->generic[i] != ATTR_UNUSED) {
-            vap_out->vap_vsm_vtx_assm |= (R300_INPUT_CNTL_TC0 << gen_count);
-            vap_out->vap_out_vtx_fmt[1] |= (4 << (3 * gen_count));
-
-            stream_loc[tabi++] = 6 + gen_count;
-            gen_count++;
-        }
-    }
-
-    /* Fog coordinates. */
-    if (gen_count < 8 && vs_outputs->fog != ATTR_UNUSED) {
-        vap_out->vap_vsm_vtx_assm |= (R300_INPUT_CNTL_TC0 << gen_count);
-        vap_out->vap_out_vtx_fmt[1] |= (4 << (3 * gen_count));
-
-        stream_loc[tabi++] = 6 + gen_count;
-        gen_count++;
-    }
-
-    /* WPOS. */
-    if (gen_count < 8) {
-        vs->wpos_tex_output = gen_count;
-        stream_loc[tabi++] = 6 + gen_count;
-    } else {
-        vs_outputs->wpos = ATTR_UNUSED;
-    }
-
-    for (; tabi < 16;) {
-        stream_loc[tabi++] = -1;
-    }
-}
-
 static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c)
 {
     struct r300_vertex_shader * vs = c->UserData;
@@ -246,9 +157,7 @@ static void set_vertex_inputs_outputs(struct r300_vertex_program_compiler * c)
     }
 
     /* WPOS. */
-    if (outputs->wpos != ATTR_UNUSED) {
-        c->code->outputs[outputs->wpos] = reg++;
-    }
+    c->code->outputs[outputs->wpos] = reg++;
 }
 
 static void r300_dummy_vertex_shader(
@@ -286,7 +195,6 @@ void r300_translate_vertex_shader(struct r300_context* r300,
 
     tgsi_scan_shader(tokens, &vs->info);
     r300_shader_read_vs_outputs(&vs->info, &vs->outputs);
-    r300_init_vs_output_mapping(vs);
 
     /* Setup the compiler */
     rc_init(&compiler.Base);
@@ -307,16 +215,11 @@ void r300_translate_vertex_shader(struct r300_context* r300,
 
     r300_tgsi_to_rc(&ttr, tokens);
 
-    compiler.RequiredOutputs =
-        ~(~0 << (vs->info.num_outputs +
-                 (vs->outputs.wpos != ATTR_UNUSED ? 1 : 0)));
-
+    compiler.RequiredOutputs = ~(~0 << (vs->info.num_outputs + 1));
     compiler.SetHwInputOutput = &set_vertex_inputs_outputs;
 
     /* Insert the WPOS output. */
-    if (vs->outputs.wpos != ATTR_UNUSED) {
-        rc_copy_output(&compiler.Base, 0, vs->outputs.wpos);
-    }
+    rc_copy_output(&compiler.Base, 0, vs->outputs.wpos);
 
     /* Invoke the compiler */
     r3xx_compile_vertex_program(&compiler);
@@ -343,32 +246,3 @@ void r300_translate_vertex_shader(struct r300_context* r300,
     /* And, finally... */
     rc_destroy(&compiler.Base);
 }
-
-boolean r300_vertex_shader_setup_wpos(struct r300_context* r300)
-{
-    struct r300_vertex_shader* vs = r300->vs_state.state;
-    struct r300_vap_output_state *vap_out = &vs->vap_out;
-    int tex_output = vs->wpos_tex_output;
-    uint32_t tex_fmt = R300_INPUT_CNTL_TC0 << tex_output;
-
-    if (vs->outputs.wpos == ATTR_UNUSED) {
-        return FALSE;
-    }
-
-    if (r300_fs(r300)->shader->inputs.wpos != ATTR_UNUSED) {
-        /* Enable WPOS in VAP. */
-        if (!(vap_out->vap_vsm_vtx_assm & tex_fmt)) {
-            vap_out->vap_vsm_vtx_assm |= tex_fmt;
-            vap_out->vap_out_vtx_fmt[1] |= (4 << (3 * tex_output));
-            return TRUE;
-        }
-    } else {
-        /* Disable WPOS in VAP. */
-        if (vap_out->vap_vsm_vtx_assm & tex_fmt) {
-            vap_out->vap_vsm_vtx_assm &= ~tex_fmt;
-            vap_out->vap_out_vtx_fmt[1] &= ~(4 << (3 * tex_output));
-            return TRUE;
-        }
-    }
-    return FALSE;
-}
diff --git a/src/gallium/drivers/r300/r300_vs.h b/src/gallium/drivers/r300/r300_vs.h
index 56bcc3b70b..57b3fbca0b 100644
--- a/src/gallium/drivers/r300/r300_vs.h
+++ b/src/gallium/drivers/r300/r300_vs.h
@@ -39,7 +39,6 @@ struct r300_vertex_shader {
 
     struct tgsi_shader_info info;
     struct r300_shader_semantics outputs;
-    struct r300_vap_output_state vap_out;
 
     /* Whether the shader was replaced by a dummy one due to a shader
      * compilation failure. */
@@ -49,12 +48,6 @@ struct r300_vertex_shader {
     unsigned externals_count;
     unsigned immediates_count;
 
-    /* Stream locations for SWTCL or if TCL is bypassed. */
-    int stream_loc_notcl[16];
-
-    /* Output stream location for WPOS. */
-    int wpos_tex_output;
-
     /* HWTCL-specific.  */
     /* Machine code (if translated) */
     struct r300_vertex_program_code code;
@@ -67,7 +60,4 @@ void r300_translate_vertex_shader(struct r300_context* r300,
                                   struct r300_vertex_shader* vs,
                                   const struct tgsi_token *tokens);
 
-/* Return TRUE if VAP (hwfmt) needs to be re-emitted. */
-boolean r300_vertex_shader_setup_wpos(struct r300_context* r300);
-
 #endif /* R300_VS_H */
diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c
index ead489393e..00187febf0 100644
--- a/src/gallium/drivers/softpipe/sp_quad_blend.c
+++ b/src/gallium/drivers/softpipe/sp_quad_blend.c
@@ -33,6 +33,7 @@
 #include "pipe/p_defines.h"
 #include "util/u_math.h"
 #include "util/u_memory.h"
+#include "util/u_format.h"
 #include "sp_context.h"
 #include "sp_quad.h"
 #include "sp_tile_cache.h"
@@ -224,7 +225,8 @@ static void
 blend_quad(struct quad_stage *qs, 
            float (*quadColor)[4],
            float (*dest)[4],
-           unsigned cbuf)
+           unsigned cbuf,
+           boolean has_dst_alpha)
 {
    static const float zero[4] = { 0, 0, 0, 0 };
    static const float one[4] = { 1, 1, 1, 1 };
@@ -259,24 +261,34 @@ blend_quad(struct quad_stage *qs,
       VEC4_MUL(source[2], quadColor[2], dest[2]); /* B */
       break;
    case PIPE_BLENDFACTOR_DST_ALPHA:
-   {
-      const float *alpha = dest[3];
-      VEC4_MUL(source[0], quadColor[0], alpha); /* R */
-      VEC4_MUL(source[1], quadColor[1], alpha); /* G */
-      VEC4_MUL(source[2], quadColor[2], alpha); /* B */
-   }
-   break;
+      if (has_dst_alpha) {
+         const float *alpha = dest[3];
+         VEC4_MUL(source[0], quadColor[0], alpha); /* R */
+         VEC4_MUL(source[1], quadColor[1], alpha); /* G */
+         VEC4_MUL(source[2], quadColor[2], alpha); /* B */
+      } 
+      else {
+         VEC4_COPY(source[0], quadColor[0]); /* R */
+         VEC4_COPY(source[1], quadColor[1]); /* G */
+         VEC4_COPY(source[2], quadColor[2]); /* B */
+      }
+      break;
    case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
-   {
-      const float *alpha = quadColor[3];
-      float diff[4], temp[4];
-      VEC4_SUB(diff, one, dest[3]);
-      VEC4_MIN(temp, alpha, diff);
-      VEC4_MUL(source[0], quadColor[0], temp); /* R */
-      VEC4_MUL(source[1], quadColor[1], temp); /* G */
-      VEC4_MUL(source[2], quadColor[2], temp); /* B */
-   }
-   break;
+      if (has_dst_alpha) {
+         const float *alpha = quadColor[3];
+         float diff[4], temp[4];
+         VEC4_SUB(diff, one, dest[3]);
+         VEC4_MIN(temp, alpha, diff);
+         VEC4_MUL(source[0], quadColor[0], temp); /* R */
+         VEC4_MUL(source[1], quadColor[1], temp); /* G */
+         VEC4_MUL(source[2], quadColor[2], temp); /* B */
+      }
+      else {
+         VEC4_COPY(source[0], zero); /* R */
+         VEC4_COPY(source[1], zero); /* G */
+         VEC4_COPY(source[2], zero); /* B */
+      }
+      break;
    case PIPE_BLENDFACTOR_CONST_COLOR:
    {
       float comp[4];
@@ -329,14 +341,19 @@ blend_quad(struct quad_stage *qs,
    }
    break;
    case PIPE_BLENDFACTOR_INV_DST_ALPHA:
-   {
-      float inv_alpha[4];
-      VEC4_SUB(inv_alpha, one, dest[3]);
-      VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */
-      VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */
-      VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */
-   }
-   break;
+      if (has_dst_alpha) {
+         float inv_alpha[4];
+         VEC4_SUB(inv_alpha, one, dest[3]);
+         VEC4_MUL(source[0], quadColor[0], inv_alpha); /* R */
+         VEC4_MUL(source[1], quadColor[1], inv_alpha); /* G */
+         VEC4_MUL(source[2], quadColor[2], inv_alpha); /* B */
+      }
+      else {
+         VEC4_COPY(source[0], zero); /* R */
+         VEC4_COPY(source[1], zero); /* G */
+         VEC4_COPY(source[2], zero); /* B */
+      }
+      break;
    case PIPE_BLENDFACTOR_INV_DST_COLOR:
    {
       float inv_comp[4];
@@ -399,7 +416,10 @@ blend_quad(struct quad_stage *qs,
    case PIPE_BLENDFACTOR_DST_COLOR:
       /* fall-through */
    case PIPE_BLENDFACTOR_DST_ALPHA:
-      VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */
+      if (has_dst_alpha)
+         VEC4_MUL(source[3], quadColor[3], dest[3]); /* A */
+      else
+         VEC4_COPY(source[3], quadColor[3]); /* A */
       break;
    case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
       /* multiply alpha by 1.0 */
@@ -429,12 +449,15 @@ blend_quad(struct quad_stage *qs,
    case PIPE_BLENDFACTOR_INV_DST_COLOR:
       /* fall-through */
    case PIPE_BLENDFACTOR_INV_DST_ALPHA:
-   {
-      float inv_alpha[4];
-      VEC4_SUB(inv_alpha, one, dest[3]);
-      VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */
-   }
-   break;
+      if (has_dst_alpha) {
+         float inv_alpha[4];
+         VEC4_SUB(inv_alpha, one, dest[3]);
+         VEC4_MUL(source[3], quadColor[3], inv_alpha); /* A */
+      }
+      else {
+         VEC4_COPY(source[3], zero); /* A */
+      }
+      break;
    case PIPE_BLENDFACTOR_INV_CONST_COLOR:
       /* fall-through */
    case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
@@ -468,9 +491,14 @@ blend_quad(struct quad_stage *qs,
       VEC4_MUL(dest[2], dest[2], quadColor[3]); /* B * A */
       break;
    case PIPE_BLENDFACTOR_DST_ALPHA:
-      VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */
-      VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */
-      VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */
+      if (has_dst_alpha) {
+         VEC4_MUL(dest[0], dest[0], dest[3]); /* R * A */
+         VEC4_MUL(dest[1], dest[1], dest[3]); /* G * A */
+         VEC4_MUL(dest[2], dest[2], dest[3]); /* B * A */
+      }
+      else {
+         /* dest = dest * 1   NO-OP, leave dest as-is */
+      }
       break;
    case PIPE_BLENDFACTOR_DST_COLOR:
       VEC4_MUL(dest[0], dest[0], dest[0]); /* R */
@@ -478,15 +506,20 @@ blend_quad(struct quad_stage *qs,
       VEC4_MUL(dest[2], dest[2], dest[2]); /* B */
       break;
    case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
-   {
-      const float *alpha = quadColor[3];
-      float diff[4], temp[4];
-      VEC4_SUB(diff, one, dest[3]);
-      VEC4_MIN(temp, alpha, diff);
-      VEC4_MUL(dest[0], quadColor[0], temp); /* R */
-      VEC4_MUL(dest[1], quadColor[1], temp); /* G */
-      VEC4_MUL(dest[2], quadColor[2], temp); /* B */
-   }
+      if (has_dst_alpha) {
+         const float *alpha = quadColor[3];
+         float diff[4], temp[4];
+         VEC4_SUB(diff, one, dest[3]);
+         VEC4_MIN(temp, alpha, diff);
+         VEC4_MUL(dest[0], quadColor[0], temp); /* R */
+         VEC4_MUL(dest[1], quadColor[1], temp); /* G */
+         VEC4_MUL(dest[2], quadColor[2], temp); /* B */
+      }
+      else {
+         VEC4_COPY(dest[0], zero); /* R */
+         VEC4_COPY(dest[1], zero); /* G */
+         VEC4_COPY(dest[2], zero); /* B */
+      }
       break;
    case PIPE_BLENDFACTOR_CONST_COLOR:
    {
@@ -539,13 +572,18 @@ blend_quad(struct quad_stage *qs,
    }
    break;
    case PIPE_BLENDFACTOR_INV_DST_ALPHA:
-   {
-      float inv_comp[4];
-      VEC4_SUB(inv_comp, one, dest[3]); /* A */
-      VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */
-      VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */
-      VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */
-   }
+      if (has_dst_alpha) {
+         float inv_comp[4];
+         VEC4_SUB(inv_comp, one, dest[3]); /* A */
+         VEC4_MUL(dest[0], inv_comp, dest[0]); /* R */
+         VEC4_MUL(dest[1], inv_comp, dest[1]); /* G */
+         VEC4_MUL(dest[2], inv_comp, dest[2]); /* B */
+      }
+      else {
+         VEC4_COPY(dest[0], zero); /* R */
+         VEC4_COPY(dest[1], zero); /* G */
+         VEC4_COPY(dest[2], zero); /* B */
+      }
    break;
    case PIPE_BLENDFACTOR_INV_DST_COLOR:
    {
@@ -605,7 +643,12 @@ blend_quad(struct quad_stage *qs,
    case PIPE_BLENDFACTOR_DST_COLOR:
       /* fall-through */
    case PIPE_BLENDFACTOR_DST_ALPHA:
-      VEC4_MUL(dest[3], dest[3], dest[3]); /* A */
+      if (has_dst_alpha) {
+         VEC4_MUL(dest[3], dest[3], dest[3]); /* A */
+      }
+      else {
+         /* dest = dest * 1   NO-OP, leave dest as-is */
+      }
       break;
    case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
       /* dest = dest * 1   NO-OP, leave dest as-is */
@@ -634,12 +677,15 @@ blend_quad(struct quad_stage *qs,
    case PIPE_BLENDFACTOR_INV_DST_COLOR:
       /* fall-through */
    case PIPE_BLENDFACTOR_INV_DST_ALPHA:
-   {
-      float inv_comp[4];
-      VEC4_SUB(inv_comp, one, dest[3]); /* A */
-      VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */
-   }
-   break;
+      if (has_dst_alpha) {
+         float inv_comp[4];
+         VEC4_SUB(inv_comp, one, dest[3]); /* A */
+         VEC4_MUL(dest[3], inv_comp, dest[3]); /* A */
+      }
+      else {
+         VEC4_COPY(dest[3], zero); /* A */
+      }
+      break;
    case PIPE_BLENDFACTOR_INV_CONST_COLOR:
       /* fall-through */
    case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
@@ -751,6 +797,8 @@ blend_fallback(struct quad_stage *qs,
          = sp_get_cached_tile(softpipe->cbuf_cache[cbuf],
                               quads[0]->input.x0, 
                               quads[0]->input.y0);
+      boolean has_dst_alpha
+         = util_format_has_alpha(softpipe->framebuffer.cbufs[cbuf]->format);
       uint q, i, j;
 
       for (q = 0; q < nr; q++) {
@@ -774,7 +822,7 @@ blend_fallback(struct quad_stage *qs,
             logicop_quad( qs, quadColor, dest );
          }
          else if (blend->rt[blend_buf].blend_enable) {
-            blend_quad( qs, quadColor, dest, cbuf );
+            blend_quad( qs, quadColor, dest, cbuf, has_dst_alpha );
          }
 
          if (blend->rt[blend_buf].colormask != 0xf)
diff --git a/src/gallium/drivers/softpipe/sp_surface.c b/src/gallium/drivers/softpipe/sp_surface.c
index b04c2a63ad..32cab06004 100644
--- a/src/gallium/drivers/softpipe/sp_surface.c
+++ b/src/gallium/drivers/softpipe/sp_surface.c
@@ -25,7 +25,7 @@
  * 
  **************************************************************************/
 
-#include "util/u_rect.h"
+#include "util/u_surface.h"
 #include "sp_context.h"
 #include "sp_surface.h"
 
diff --git a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c
index fbce9e042b..b3e1c49406 100644
--- a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c
@@ -115,6 +115,20 @@ sp_tex_tile_cache_validate_texture(struct softpipe_tex_tile_cache *tc)
    }
 }
 
+static boolean
+sp_tex_tile_is_compat_view(struct softpipe_tex_tile_cache *tc,
+                           struct pipe_sampler_view *view)
+{
+   if (!view)
+      return FALSE;
+   return (tc->texture == view->texture &&
+           tc->format == view->format &&
+           tc->swizzle_r == view->swizzle_r &&
+           tc->swizzle_g == view->swizzle_g &&
+           tc->swizzle_b == view->swizzle_b &&
+           tc->swizzle_a == view->swizzle_a);
+}
+
 /**
  * Specify the sampler view to cache.
  */
@@ -127,7 +141,7 @@ sp_tex_tile_cache_set_sampler_view(struct softpipe_tex_tile_cache *tc,
 
    assert(!tc->transfer);
 
-   if (tc->texture != texture) {
+   if (!sp_tex_tile_is_compat_view(tc, view)) {
       pipe_resource_reference(&tc->texture, texture);
 
       if (tc->tex_trans) {
diff --git a/src/gallium/drivers/svga/svga_context.c b/src/gallium/drivers/svga/svga_context.c
index 3228a6d3d7..75d8afb2ea 100644
--- a/src/gallium/drivers/svga/svga_context.c
+++ b/src/gallium/drivers/svga/svga_context.c
@@ -194,6 +194,7 @@ void svga_context_flush( struct svga_context *svga,
                          struct pipe_fence_handle **pfence )
 {
    struct svga_screen *svgascreen = svga_screen(svga->pipe.screen);
+   struct pipe_fence_handle *fence = NULL;
 
    svga->curr.nr_fbs = 0;
 
@@ -202,21 +203,26 @@ void svga_context_flush( struct svga_context *svga,
    u_upload_flush(svga->upload_vb);
    u_upload_flush(svga->upload_ib);
 
-   /* Flush screen, to ensure that texture dma uploads are processed
+   /* Ensure that texture dma uploads are processed
     * before submitting commands.
     */
-   svga_screen_flush(svgascreen, NULL);
-   
    svga_context_flush_buffers(svga);
 
    /* Flush pending commands to hardware:
     */
-   svga->swc->flush(svga->swc, pfence);
+   svga->swc->flush(svga->swc, &fence);
+
+   svga_screen_cache_flush(svgascreen, fence);
 
    if (SVGA_DEBUG & DEBUG_SYNC) {
-      if (pfence && *pfence)
-         svga->pipe.screen->fence_finish( svga->pipe.screen, *pfence, 0);
+      if (fence)
+         svga->pipe.screen->fence_finish( svga->pipe.screen, fence, 0);
    }
+
+   if(pfence)
+      *pfence = fence;
+   else
+      svgascreen->sws->fence_reference(svgascreen->sws, &fence, NULL);
 }
 
 
diff --git a/src/gallium/drivers/svga/svga_draw_arrays.c b/src/gallium/drivers/svga/svga_draw_arrays.c
index 005996d05d..da33fae62f 100644
--- a/src/gallium/drivers/svga/svga_draw_arrays.c
+++ b/src/gallium/drivers/svga/svga_draw_arrays.c
@@ -277,12 +277,13 @@ svga_hwtnl_draw_arrays( struct svga_hwtnl *hwtnl,
       ret = svga_hwtnl_simple_draw_range_elements( hwtnl,
                                                    gen_buf,
                                                    gen_size,
+                                                   start,
                                                    0,
                                                    count - 1,
                                                    gen_prim,
                                                    0,
-                                                   gen_nr,
-                                                   start );
+                                                   gen_nr );
+
       if (ret)
          goto done;
 
diff --git a/src/gallium/drivers/svga/svga_pipe_query.c b/src/gallium/drivers/svga/svga_pipe_query.c
index 9c6f5858ba..96fb4b8e53 100644
--- a/src/gallium/drivers/svga/svga_pipe_query.c
+++ b/src/gallium/drivers/svga/svga_pipe_query.c
@@ -68,6 +68,7 @@ static boolean svga_get_query_result(struct pipe_context *pipe,
 static struct pipe_query *svga_create_query( struct pipe_context *pipe,
                                              unsigned query_type )
 {
+   struct svga_context *svga = svga_context( pipe );
    struct svga_screen *svgascreen = svga_screen(pipe->screen);
    struct svga_winsys_screen *sws = svgascreen->sws;
    struct svga_query *sq;
@@ -80,7 +81,7 @@ static struct pipe_query *svga_create_query( struct pipe_context *pipe,
 
    sq->type = SVGA3D_QUERYTYPE_OCCLUSION;
 
-   sq->hwbuf = svga_winsys_buffer_create(svgascreen, 
+   sq->hwbuf = svga_winsys_buffer_create(svga,
                                          1,
                                          SVGA_BUFFER_USAGE_PINNED,
                                          sizeof *sq->queryResult);
diff --git a/src/gallium/drivers/svga/svga_resource_buffer.c b/src/gallium/drivers/svga/svga_resource_buffer.c
index 18eeff9c67..198d401332 100644
--- a/src/gallium/drivers/svga/svga_resource_buffer.c
+++ b/src/gallium/drivers/svga/svga_resource_buffer.c
@@ -125,8 +125,6 @@ svga_buffer_map_range( struct pipe_screen *screen,
    }
 
    if(map) {
-      pipe_mutex_lock(ss->swc_mutex);
-
       ++sbuf->map.count;
 
       if (usage & PIPE_TRANSFER_WRITE) {
@@ -135,8 +133,6 @@ svga_buffer_map_range( struct pipe_screen *screen,
          if (usage & PIPE_TRANSFER_FLUSH_EXPLICIT)
             sbuf->map.flush_explicit = TRUE;
       }
-      
-      pipe_mutex_unlock(ss->swc_mutex);
    }
    
    return map;
diff --git a/src/gallium/drivers/svga/svga_resource_buffer.h b/src/gallium/drivers/svga/svga_resource_buffer.h
index 55e7321184..d3ec11bfd5 100644
--- a/src/gallium/drivers/svga/svga_resource_buffer.h
+++ b/src/gallium/drivers/svga/svga_resource_buffer.h
@@ -238,7 +238,7 @@ void
 svga_context_flush_buffers(struct svga_context *svga);
 
 struct svga_winsys_buffer *
-svga_winsys_buffer_create(struct svga_screen *ss,
+svga_winsys_buffer_create(struct svga_context *svga,
                           unsigned alignment, 
                           unsigned usage,
                           unsigned size);
diff --git a/src/gallium/drivers/svga/svga_resource_buffer_host.c b/src/gallium/drivers/svga/svga_resource_buffer_host.c
deleted file mode 100644
index 139597f9cb..0000000000
--- a/src/gallium/drivers/svga/svga_resource_buffer_host.c
+++ /dev/null
@@ -1,2 +0,0 @@
-
-
diff --git a/src/gallium/drivers/svga/svga_resource_buffer_upload.c b/src/gallium/drivers/svga/svga_resource_buffer_upload.c
index acef60f464..3de5216a94 100644
--- a/src/gallium/drivers/svga/svga_resource_buffer_upload.c
+++ b/src/gallium/drivers/svga/svga_resource_buffer_upload.c
@@ -40,15 +40,21 @@
 #include "svga_debug.h"
 
 
-/* Allocate a winsys_buffer (ie. DMA, aka GMR memory).
+/**
+ * Allocate a winsys_buffer (ie. DMA, aka GMR memory).
+ *
+ * It will flush and retry in case the first attempt to create a DMA buffer
+ * fails, so it should not be called from any function involved in flushing
+ * to avoid recursion.
  */
 struct svga_winsys_buffer *
-svga_winsys_buffer_create( struct svga_screen *ss,
+svga_winsys_buffer_create( struct svga_context *svga,
                            unsigned alignment, 
                            unsigned usage,
                            unsigned size )
 {
-   struct svga_winsys_screen *sws = ss->sws;
+   struct svga_screen *svgascreen = svga_screen(svga->pipe.screen);
+   struct svga_winsys_screen *sws = svgascreen->sws;
    struct svga_winsys_buffer *buf;
    
    /* Just try */
@@ -59,9 +65,8 @@ svga_winsys_buffer_create( struct svga_screen *ss,
                size); 
       
       /* Try flushing all pending DMAs */
-      svga_screen_flush(ss, NULL);
+      svga_context_flush(svga, NULL);
       buf = sws->buffer_create(sws, alignment, usage, size);
-
    }
    
    return buf;
@@ -95,11 +100,12 @@ svga_buffer_create_hw_storage(struct svga_screen *ss,
    assert(!sbuf->user);
 
    if(!sbuf->hwbuf) {
+      struct svga_winsys_screen *sws = ss->sws;
       unsigned alignment = 16;
       unsigned usage = 0;
       unsigned size = sbuf->b.b.width0;
       
-      sbuf->hwbuf = svga_winsys_buffer_create(ss, alignment, usage, size);
+      sbuf->hwbuf = sws->buffer_create(sws, alignment, usage, size);
       if(!sbuf->hwbuf)
          return PIPE_ERROR_OUT_OF_MEMORY;
       
@@ -476,12 +482,12 @@ svga_buffer_upload_piecewise(struct svga_screen *ss,
          if (offset + size > range->end)
             size = range->end - offset;
 
-         hwbuf = svga_winsys_buffer_create(ss, alignment, usage, size);
+         hwbuf = sws->buffer_create(sws, alignment, usage, size);
          while (!hwbuf) {
             size /= 2;
             if (!size)
                return PIPE_ERROR_OUT_OF_MEMORY;
-            hwbuf = svga_winsys_buffer_create(ss, alignment, usage, size);
+            hwbuf = sws->buffer_create(sws, alignment, usage, size);
          }
 
          SVGA_DBG(DEBUG_DMA, "  bytes %u - %u\n",
diff --git a/src/gallium/drivers/svga/svga_resource_texture.c b/src/gallium/drivers/svga/svga_resource_texture.c
index f06b0323d8..ff83c750aa 100644
--- a/src/gallium/drivers/svga/svga_resource_texture.c
+++ b/src/gallium/drivers/svga/svga_resource_texture.c
@@ -163,12 +163,12 @@ svga_translate_format_render(enum pipe_format format)
 
 
 static INLINE void
-svga_transfer_dma_band(struct svga_transfer *st,
+svga_transfer_dma_band(struct svga_context *svga,
+                       struct svga_transfer *st,
                        SVGA3dTransferType transfer,
                        unsigned y, unsigned h, unsigned srcy)
 {
    struct svga_texture *texture = svga_texture(st->base.resource); 
-   struct svga_screen *screen = svga_screen(texture->b.b.screen);
    SVGA3dCopyBox box;
    enum pipe_error ret;
    
@@ -195,20 +195,19 @@ svga_transfer_dma_band(struct svga_transfer *st,
    box.srcy = srcy;
    box.srcz = 0;
 
-   pipe_mutex_lock(screen->swc_mutex);
-   ret = SVGA3D_SurfaceDMA(screen->swc, st, transfer, &box, 1);
+   ret = SVGA3D_SurfaceDMA(svga->swc, st, transfer, &box, 1);
    if(ret != PIPE_OK) {
-      screen->swc->flush(screen->swc, NULL);
-      ret = SVGA3D_SurfaceDMA(screen->swc, st, transfer, &box, 1);
+      svga->swc->flush(svga->swc, NULL);
+      ret = SVGA3D_SurfaceDMA(svga->swc, st, transfer, &box, 1);
       assert(ret == PIPE_OK);
    }
-   pipe_mutex_unlock(screen->swc_mutex);
 }
 
 
 static INLINE void
-svga_transfer_dma(struct svga_transfer *st,
-                 SVGA3dTransferType transfer)
+svga_transfer_dma(struct svga_context *svga,
+                  struct svga_transfer *st,
+                  SVGA3dTransferType transfer)
 {
    struct svga_texture *texture = svga_texture(st->base.resource); 
    struct svga_screen *screen = svga_screen(texture->b.b.screen);
@@ -223,10 +222,10 @@ svga_transfer_dma(struct svga_transfer *st,
    if(!st->swbuf) {
       /* Do the DMA transfer in a single go */
       
-      svga_transfer_dma_band(st, transfer, st->base.box.y, st->base.box.height, 0);
+      svga_transfer_dma_band(svga, st, transfer, st->base.box.y, st->base.box.height, 0);
 
       if(transfer == SVGA3D_READ_HOST_VRAM) {
-         svga_screen_flush(screen, &fence);
+         svga_context_flush(svga, &fence);
          sws->fence_finish(sws, fence, 0);
          sws->fence_reference(sws, &fence, NULL);
       }
@@ -256,7 +255,7 @@ svga_transfer_dma(struct svga_transfer *st,
             /* Wait for the previous DMAs to complete */
             /* TODO: keep one DMA (at half the size) in the background */
             if(y) {
-               svga_screen_flush(screen, &fence);
+               svga_context_flush(svga, &fence);
                sws->fence_finish(sws, fence, 0);
                sws->fence_reference(sws, &fence, NULL);
             }
@@ -269,10 +268,10 @@ svga_transfer_dma(struct svga_transfer *st,
             }
          }
          
-         svga_transfer_dma_band(st, transfer, y, h, srcy);
+         svga_transfer_dma_band(svga, st, transfer, y, h, srcy);
          
          if(transfer == SVGA3D_READ_HOST_VRAM) {
-            svga_screen_flush(screen, &fence);
+            svga_context_flush(svga, &fence);
             sws->fence_finish(sws, fence, 0);
 
             hw = sws->buffer_map(sws, st->hwbuf, PIPE_TRANSFER_READ);
@@ -342,6 +341,7 @@ svga_texture_get_transfer(struct pipe_context *pipe,
 			  unsigned usage,
 			  const struct pipe_box *box)
 {
+   struct svga_context *svga = svga_context(pipe);
    struct svga_screen *ss = svga_screen(pipe->screen);
    struct svga_winsys_screen *sws = ss->sws;
    struct svga_transfer *st;
@@ -365,12 +365,12 @@ svga_texture_get_transfer(struct pipe_context *pipe,
 
    st->hw_nblocksy = nblocksy;
    
-   st->hwbuf = svga_winsys_buffer_create(ss, 
+   st->hwbuf = svga_winsys_buffer_create(svga,
                                          1, 
                                          0,
                                          st->hw_nblocksy*st->base.stride);
    while(!st->hwbuf && (st->hw_nblocksy /= 2)) {
-      st->hwbuf = svga_winsys_buffer_create(ss, 
+      st->hwbuf = svga_winsys_buffer_create(svga,
                                             1, 
                                             0,
                                             st->hw_nblocksy*st->base.stride);
@@ -393,7 +393,7 @@ svga_texture_get_transfer(struct pipe_context *pipe,
    }
    
    if (usage & PIPE_TRANSFER_READ)
-      svga_transfer_dma(st, SVGA3D_READ_HOST_VRAM);
+      svga_transfer_dma(svga, st, SVGA3D_READ_HOST_VRAM);
 
    return &st->base;
 
@@ -445,13 +445,14 @@ static void
 svga_texture_transfer_destroy(struct pipe_context *pipe,
 			      struct pipe_transfer *transfer)
 {
+   struct svga_context *svga = svga_context(pipe);
    struct svga_texture *tex = svga_texture(transfer->resource);
    struct svga_screen *ss = svga_screen(pipe->screen);
    struct svga_winsys_screen *sws = ss->sws;
    struct svga_transfer *st = svga_transfer(transfer);
 
    if (st->base.usage & PIPE_TRANSFER_WRITE) {
-      svga_transfer_dma(st, SVGA3D_WRITE_HOST_VRAM);
+      svga_transfer_dma(svga, st, SVGA3D_WRITE_HOST_VRAM);
       ss->texture_timestamp++;
       tex->view_age[transfer->sr.level] = ++(tex->age);
       tex->defined[transfer->sr.face][transfer->sr.level] = TRUE;
diff --git a/src/gallium/drivers/svga/svga_sampler_view.c b/src/gallium/drivers/svga/svga_sampler_view.c
index fbae552f78..6911f13f77 100644
--- a/src/gallium/drivers/svga/svga_sampler_view.c
+++ b/src/gallium/drivers/svga/svga_sampler_view.c
@@ -169,7 +169,7 @@ svga_validate_sampler_view(struct svga_context *svga, struct svga_sampler_view *
    for (i = v->min_lod; i <= v->max_lod; i++) {
       for (k = 0; k < numFaces; k++) {
          if (v->age < tex->view_age[i])
-            svga_texture_copy_handle(svga, NULL,
+            svga_texture_copy_handle(svga,
                                      tex->handle, 0, 0, 0, i, k,
                                      v->handle, 0, 0, 0, i - v->min_lod, k,
                                      u_minify(tex->b.b.width0, i),
diff --git a/src/gallium/drivers/svga/svga_screen.c b/src/gallium/drivers/svga/svga_screen.c
index aeda3dcad5..27ac09e275 100644
--- a/src/gallium/drivers/svga/svga_screen.c
+++ b/src/gallium/drivers/svga/svga_screen.c
@@ -34,6 +34,7 @@
 #include "svga_resource_texture.h"
 #include "svga_resource.h"
 #include "svga_debug.h"
+#include "svga_surface.h"
 
 #include "svga3d_shaderdefs.h"
 
@@ -346,8 +347,6 @@ svga_destroy_screen( struct pipe_screen *screen )
    pipe_mutex_destroy(svgascreen->swc_mutex);
    pipe_mutex_destroy(svgascreen->tex_mutex);
 
-   svgascreen->swc->destroy(svgascreen->swc);
-   
    svgascreen->sws->destroy(svgascreen->sws);
    
    FREE(svgascreen);
@@ -397,6 +396,7 @@ svga_screen_create(struct svga_winsys_screen *sws)
    screen->fence_finish = svga_fence_finish;
    svgascreen->sws = sws;
 
+   svga_screen_init_surface_functions(screen);
    svga_init_screen_resource_functions(svgascreen);
 
    svgascreen->use_ps30 =
@@ -416,10 +416,6 @@ svga_screen_create(struct svga_winsys_screen *sws)
       svgascreen->use_vs30 = svgascreen->use_ps30 = FALSE;
 #endif
 
-   svgascreen->swc = sws->context_create(sws);
-   if(!svgascreen->swc)
-      goto error2;
-
    pipe_mutex_init(svgascreen->tex_mutex);
    pipe_mutex_init(svgascreen->swc_mutex);
 
@@ -432,25 +428,6 @@ error1:
    return NULL;
 }
 
-void svga_screen_flush( struct svga_screen *svgascreen, 
-                        struct pipe_fence_handle **pfence )
-{
-   struct pipe_fence_handle *fence = NULL;
-
-   SVGA_DBG(DEBUG_PERF, "%s\n", __FUNCTION__);
-   
-   pipe_mutex_lock(svgascreen->swc_mutex);
-   svgascreen->swc->flush(svgascreen->swc, &fence);
-   pipe_mutex_unlock(svgascreen->swc_mutex);
-   
-   svga_screen_cache_flush(svgascreen, fence);
-   
-   if(pfence)
-      *pfence = fence;
-   else
-      svgascreen->sws->fence_reference(svgascreen->sws, &fence, NULL);
-}
-
 struct svga_winsys_screen *
 svga_winsys_screen(struct pipe_screen *screen)
 {
diff --git a/src/gallium/drivers/svga/svga_screen.h b/src/gallium/drivers/svga/svga_screen.h
index 9dc229b0a8..86ec89d88c 100644
--- a/src/gallium/drivers/svga/svga_screen.h
+++ b/src/gallium/drivers/svga/svga_screen.h
@@ -60,14 +60,11 @@ struct svga_screen
       boolean no_sampler_view;
    } debug;
 
-   /* The screen needs its own context */
-   struct svga_winsys_context *swc;
-   struct SVGACmdMemory *fifo;
-
    unsigned texture_timestamp;
    pipe_mutex tex_mutex; 
-   pipe_mutex swc_mutex; /* Protects the use of swc and dirty_buffers */
-   
+
+   pipe_mutex swc_mutex; /* Used for buffer uploads */
+
    struct svga_host_surface_cache cache;
 };
 
@@ -83,7 +80,4 @@ struct svga_screen *
 svga_screen(struct pipe_screen *screen);
 #endif
 
-void svga_screen_flush( struct svga_screen *svga_screen, 
-                        struct pipe_fence_handle **pfence );
-
 #endif /* SVGA_SCREEN_H */
diff --git a/src/gallium/drivers/svga/svga_surface.c b/src/gallium/drivers/svga/svga_surface.c
index dc1d9a850d..bd2cc38004 100644
--- a/src/gallium/drivers/svga/svga_surface.c
+++ b/src/gallium/drivers/svga/svga_surface.c
@@ -43,7 +43,6 @@
 
 void
 svga_texture_copy_handle(struct svga_context *svga,
-                         struct svga_screen *ss,
                          struct svga_winsys_surface *src_handle,
                          unsigned src_x, unsigned src_y, unsigned src_z,
                          unsigned src_level, unsigned src_face,
@@ -56,7 +55,7 @@ svga_texture_copy_handle(struct svga_context *svga,
    enum pipe_error ret;
    SVGA3dCopyBox box, *boxes;
 
-   assert(svga || ss);
+   assert(svga);
 
    src.handle = src_handle;
    src.real_level = src_level;
@@ -84,39 +83,20 @@ svga_texture_copy_handle(struct svga_context *svga,
             dst_handle, dst_level, dst_x, dst_y, dst_z);
 */
 
-   if (svga) {
+   ret = SVGA3D_BeginSurfaceCopy(svga->swc,
+                                 &src.base,
+                                 &dst.base,
+                                 &boxes, 1);
+   if(ret != PIPE_OK) {
+      svga_context_flush(svga, NULL);
       ret = SVGA3D_BeginSurfaceCopy(svga->swc,
                                     &src.base,
                                     &dst.base,
                                     &boxes, 1);
-      if(ret != PIPE_OK) {
-         svga_context_flush(svga, NULL);
-         ret = SVGA3D_BeginSurfaceCopy(svga->swc,
-                                       &src.base,
-                                       &dst.base,
-                                       &boxes, 1);
-         assert(ret == PIPE_OK);
-      }
-      *boxes = box;
-      SVGA_FIFOCommitAll(svga->swc);
-   } else {
-      pipe_mutex_lock(ss->swc_mutex);
-      ret = SVGA3D_BeginSurfaceCopy(ss->swc,
-                                    &src.base,
-                                    &dst.base,
-                                    &boxes, 1);
-      if(ret != PIPE_OK) {
-         ss->swc->flush(ss->swc, NULL);
-         ret = SVGA3D_BeginSurfaceCopy(ss->swc,
-                                       &src.base,
-                                       &dst.base,
-                                       &boxes, 1);
-         assert(ret == PIPE_OK);
-      }
-      *boxes = box;
-      SVGA_FIFOCommitAll(ss->swc);
-      pipe_mutex_unlock(ss->swc_mutex);
+      assert(ret == PIPE_OK);
    }
+   *boxes = box;
+   SVGA_FIFOCommitAll(svga->swc);
 }
 
 
@@ -183,7 +163,6 @@ svga_texture_view_surface(struct pipe_context *pipe,
                               1);
 
             svga_texture_copy_handle(svga_context(pipe),
-                                     ss,
                                      tex->handle, 
                                      0, 0, z_offset, 
                                      i + start_mip, 
@@ -346,7 +325,7 @@ svga_propagate_surface(struct pipe_context *pipe, struct pipe_surface *surf)
 
    if (s->handle != tex->handle) {
       SVGA_DBG(DEBUG_VIEWS, "svga: Surface propagate: tex %p, level %u, from %p\n", tex, surf->level, surf);
-      svga_texture_copy_handle(svga_context(pipe), ss,
+      svga_texture_copy_handle(svga_context(pipe),
                                s->handle, 0, 0, 0, s->real_level, s->real_face,
                                tex->handle, 0, 0, surf->zslice, surf->level, surf->face,
                                u_minify(tex->b.b.width0, surf->level),
diff --git a/src/gallium/drivers/svga/svga_surface.h b/src/gallium/drivers/svga/svga_surface.h
index b50ecdc994..13bd5b19b6 100644
--- a/src/gallium/drivers/svga/svga_surface.h
+++ b/src/gallium/drivers/svga/svga_surface.h
@@ -74,7 +74,6 @@ svga_texture_view_surface(struct pipe_context *pipe,
 
 void
 svga_texture_copy_handle(struct svga_context *svga,
-                         struct svga_screen *ss,
                          struct svga_winsys_surface *src_handle,
                          unsigned src_x, unsigned src_y, unsigned src_z,
                          unsigned src_level, unsigned src_face,
diff --git a/src/gallium/drivers/svga/svga_winsys.h b/src/gallium/drivers/svga/svga_winsys.h
index 3892addafd..c155f5dae2 100644
--- a/src/gallium/drivers/svga/svga_winsys.h
+++ b/src/gallium/drivers/svga/svga_winsys.h
@@ -85,7 +85,7 @@ struct svga_winsys_context
    /**
     * Emit a relocation for a host surface.
     * 
-    * @param flags PIPE_BUFFER_USAGE_GPU_READ/WRITE
+    * @param flags bitmask of SVGA_RELOC_* flags
     * 
     * NOTE: Order of this call does matter. It should be the same order
     * as relocations appear in the command buffer.
@@ -99,7 +99,7 @@ struct svga_winsys_context
    /**
     * Emit a relocation for a guest memory region.
     * 
-    * @param flags PIPE_BUFFER_USAGE_GPU_READ/WRITE
+    * @param flags bitmask of SVGA_RELOC_* flags
     * 
     * NOTE: Order of this call does matter. It should be the same order
     * as relocations appear in the command buffer.
@@ -230,7 +230,7 @@ struct svga_winsys_screen
    /**
     * Buffer management. Buffer attributes are mostly fixed over its lifetime.
     *
-    * XXX usage seems to be a bitmask of SVGA_BUFFER_USAGE_* flags.
+    * @param usage bitmask of SVGA_BUFFER_USAGE_* flags.
     *
     * alignment indicates the client's alignment requirements, eg for
     * SSE instructions.
@@ -243,15 +243,12 @@ struct svga_winsys_screen
 
    /** 
     * Map the entire data store of a buffer object into the client's address.
-    * flags is a bitmask of:
-    * - PB_USAGE_CPU_READ/WRITE
-    * - PB_USAGE_DONTBLOCK
-    * - PB_USAGE_UNSYNCHRONIZED
+    * flags is a bitmaks of PIPE_TRANSFER_*
     */
    void *
    (*buffer_map)( struct svga_winsys_screen *sws, 
 	          struct svga_winsys_buffer *buf,
-		  unsigned usage );
+		  unsigned flags );
    
    void 
    (*buffer_unmap)( struct svga_winsys_screen *sws, 
diff --git a/src/gallium/drivers/trace/tr_context.c b/src/gallium/drivers/trace/tr_context.c
index 71ba1e909d..3d1fb96b91 100644
--- a/src/gallium/drivers/trace/tr_context.c
+++ b/src/gallium/drivers/trace/tr_context.c
@@ -812,7 +812,7 @@ trace_context_delete_vertex_elements_state(struct pipe_context *_pipe,
    struct trace_context *tr_ctx = trace_context(_pipe);
    struct pipe_context *pipe = tr_ctx->pipe;
 
-   trace_dump_call_begin("pipe_context", "delete_verte_elements_state");
+   trace_dump_call_begin("pipe_context", "delete_vertex_elements_state");
 
    trace_dump_arg(ptr, pipe);
    trace_dump_arg(ptr, state);
diff --git a/src/gallium/include/pipe/p_shader_tokens.h b/src/gallium/include/pipe/p_shader_tokens.h
index e21aaacc18..0a70237f26 100644
--- a/src/gallium/include/pipe/p_shader_tokens.h
+++ b/src/gallium/include/pipe/p_shader_tokens.h
@@ -73,9 +73,8 @@ enum tgsi_file_type {
    TGSI_FILE_SAMPLER      =5,
    TGSI_FILE_ADDRESS      =6,
    TGSI_FILE_IMMEDIATE    =7,
-   TGSI_FILE_LOOP         =8,
-   TGSI_FILE_PREDICATE    =9,
-   TGSI_FILE_SYSTEM_VALUE =10,
+   TGSI_FILE_PREDICATE    =8,
+   TGSI_FILE_SYSTEM_VALUE =9,
    TGSI_FILE_COUNT      /**< how many TGSI_FILE_ types */
 };
 
diff --git a/src/gallium/state_trackers/dri/common/dri_context.c b/src/gallium/state_trackers/dri/common/dri_context.c
index a808d2d9dd..203682ef33 100644
--- a/src/gallium/state_trackers/dri/common/dri_context.c
+++ b/src/gallium/state_trackers/dri/common/dri_context.c
@@ -49,7 +49,7 @@ dri_init_extensions(struct dri_context *ctx)
 }
 
 GLboolean
-dri_create_context(const __GLcontextModes * visual,
+dri_create_context(gl_api api, const __GLcontextModes * visual,
 		   __DRIcontext * cPriv, void *sharedContextPrivate)
 {
    __DRIscreen *sPriv = cPriv->driScreenPriv;
diff --git a/src/gallium/state_trackers/dri/common/dri_context.h b/src/gallium/state_trackers/dri/common/dri_context.h
index 9fe6b58101..54e56c6499 100644
--- a/src/gallium/state_trackers/dri/common/dri_context.h
+++ b/src/gallium/state_trackers/dri/common/dri_context.h
@@ -34,6 +34,7 @@
 
 #include "pipe/p_compiler.h"
 #include "dri_wrapper.h"
+#include "main/mtypes.h"
 
 struct pipe_context;
 struct pipe_fence;
@@ -83,7 +84,8 @@ struct dri_context *
 dri_get_current(__DRIscreen * driScreenPriv);
 
 boolean
-dri_create_context(const __GLcontextModes * visual,
+dri_create_context(gl_api api,
+		   const __GLcontextModes * visual,
 		   __DRIcontext * driContextPriv,
 		   void *sharedContextPrivate);
 
diff --git a/src/gallium/state_trackers/dri/drm/dri1.c b/src/gallium/state_trackers/dri/drm/dri1.c
index 23c21ed839..326ff8bcad 100644
--- a/src/gallium/state_trackers/dri/drm/dri1.c
+++ b/src/gallium/state_trackers/dri/drm/dri1.c
@@ -33,6 +33,7 @@
 
 #include "util/u_memory.h"
 #include "util/u_rect.h"
+#include "util/u_surface.h"
 #include "util/u_inlines.h"
 #include "pipe/p_context.h"
 #include "state_tracker/dri1_api.h"
diff --git a/src/gallium/state_trackers/vega/renderer.c b/src/gallium/state_trackers/vega/renderer.c
index 48fbc3b330..e6aea482a7 100644
--- a/src/gallium/state_trackers/vega/renderer.c
+++ b/src/gallium/state_trackers/vega/renderer.c
@@ -40,6 +40,7 @@
 #include "util/u_memory.h"
 #include "util/u_rect.h"
 #include "util/u_sampler.h"
+#include "util/u_surface.h"
 
 #include "cso_cache/cso_context.h"
 
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c
index d5a1be8174..65be8c332a 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa.c
@@ -46,6 +46,7 @@
 #include "util/u_math.h"
 #include "util/u_debug.h"
 #include "util/u_format.h"
+#include "util/u_surface.h"
 
 #define DEBUG_PRINT 0
 #define ROUND_UP_TEXTURES 1
diff --git a/src/gallium/state_trackers/xorg/xorg_renderer.c b/src/gallium/state_trackers/xorg/xorg_renderer.c
index 13fa561390..583493116d 100644
--- a/src/gallium/state_trackers/xorg/xorg_renderer.c
+++ b/src/gallium/state_trackers/xorg/xorg_renderer.c
@@ -9,6 +9,7 @@
 #include "util/u_memory.h"
 #include "util/u_rect.h"
 #include "util/u_sampler.h"
+#include "util/u_surface.h"
 
 #include "util/u_inlines.h"
 
diff --git a/src/gallium/targets/dri-swrast/Makefile b/src/gallium/targets/dri-swrast/Makefile
index fcfd690e43..3db9781c20 100644
--- a/src/gallium/targets/dri-swrast/Makefile
+++ b/src/gallium/targets/dri-swrast/Makefile
@@ -8,7 +8,8 @@ DRIVER_DEFINES = -D__NOT_HAVE_DRM_H -DGALLIUM_SOFTPIPE
 PIPE_DRIVERS = \
 	$(TOP)/src/gallium/state_trackers/dri/sw/libdrisw.a \
 	$(TOP)/src/gallium/winsys/sw/dri/libswdri.a \
-	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a
+	$(TOP)/src/gallium/drivers/softpipe/libsoftpipe.a \
+	$(TOP)/src/gallium/drivers/trace/libtrace.a
 
 SWRAST_COMMON_GALLIUM_SOURCES = \
 	$(TOP)/src/mesa/drivers/dri/common/utils.c \
diff --git a/src/gallium/targets/dri-swrast/swrast_drm_api.c b/src/gallium/targets/dri-swrast/swrast_drm_api.c
index e8d6d8069c..84142be80c 100644
--- a/src/gallium/targets/dri-swrast/swrast_drm_api.c
+++ b/src/gallium/targets/dri-swrast/swrast_drm_api.c
@@ -31,6 +31,7 @@
 #include "state_tracker/drm_api.h"
 #include "state_tracker/sw_winsys.h"
 #include "dri_sw_winsys.h"
+#include "trace/tr_public.h"
 
 /* Copied from targets/libgl-xlib */
 
@@ -80,7 +81,7 @@ swrast_create_screen(struct sw_winsys *winsys)
       screen = softpipe_create_screen( winsys );
 #endif
 
-   return screen;
+   return trace_screen_create(screen);;
 }
 
 struct pipe_screen *
diff --git a/src/gallium/tests/python/retrace/interpreter.py b/src/gallium/tests/python/retrace/interpreter.py
index 7118ff85ed..e58a69322e 100755
--- a/src/gallium/tests/python/retrace/interpreter.py
+++ b/src/gallium/tests/python/retrace/interpreter.py
@@ -432,6 +432,9 @@ class Context(Object):
                        swizzle_b = templ.swizzle_g,
                        swizzle_a = templ.swizzle_a)
 
+    def sampler_view_destroy(self, view):
+        pass
+
     def set_fragment_sampler_views(self, num, views):
         for i in range(num):
             self.real.set_fragment_sampler_view(i, views[i])
@@ -456,6 +459,10 @@ class Context(Object):
         return elements[0:num_elements]
 
     def bind_vertex_elements_state(self, state):
+        if state is None:
+            self.real.set_vertex_elements(0)
+            return
+
         elements = state
         num_elements = len(elements)
         self.velems = elements
diff --git a/src/gallium/winsys/radeon/drm/radeon_drm_buffer.c b/src/gallium/winsys/radeon/drm/radeon_drm_buffer.c
index 9824ada5b3..b836649892 100644
--- a/src/gallium/winsys/radeon/drm/radeon_drm_buffer.c
+++ b/src/gallium/winsys/radeon/drm/radeon_drm_buffer.c
@@ -22,6 +22,8 @@ struct radeon_drm_buffer {
 
     boolean flinked;
     uint32_t flink;
+    uint32_t tileflags;
+    uint32_t pitch;
 
     struct radeon_drm_buffer *next, *prev;
 };
@@ -318,6 +320,9 @@ void radeon_drm_bufmgr_get_tiling(struct pb_buffer *_buf,
 
     radeon_bo_get_tiling(buf->bo, &flags, &pitch);
 
+    buf->tileflags = flags;
+    buf->pitch = pitch;
+
     *microtiled = R300_BUFFER_LINEAR;
     *macrotiled = R300_BUFFER_LINEAR;
     if (flags & RADEON_BO_FLAGS_MICRO_TILE)
@@ -333,7 +338,7 @@ void radeon_drm_bufmgr_set_tiling(struct pb_buffer *_buf,
                                   uint32_t pitch)
 {
     struct radeon_drm_buffer *buf = get_drm_buffer(_buf);
-    uint32_t flags = 0, old_flags, old_pitch;
+    uint32_t flags = 0;
     if (microtiled == R300_BUFFER_TILED)
         flags |= RADEON_BO_FLAGS_MICRO_TILE;
 /* XXX Remove this ifdef when libdrm version 2.4.19 becomes mandatory. */
@@ -344,17 +349,15 @@ void radeon_drm_bufmgr_set_tiling(struct pb_buffer *_buf,
     if (macrotiled == R300_BUFFER_TILED)
         flags |= RADEON_BO_FLAGS_MACRO_TILE;
 
-    radeon_bo_get_tiling(buf->bo, &old_flags, &old_pitch);
-
-    if (flags != old_flags || pitch != old_pitch) {
+    if (flags != buf->tileflags || pitch != buf->pitch) {
         /* Tiling determines how DRM treats the buffer data.
          * We must flush CS when changing it if the buffer is referenced. */
         if (radeon_bo_is_referenced_by_cs(buf->bo,  buf->mgr->rws->cs)) {
 	    buf->mgr->rws->flush_cb(buf->mgr->rws->flush_data);
         }
-    }
-    radeon_bo_set_tiling(buf->bo, flags, pitch);
 
+        radeon_bo_set_tiling(buf->bo, flags, pitch);
+    }
 }
 
 boolean radeon_drm_bufmgr_add_buffer(struct pb_buffer *_buf,
diff --git a/src/gallium/winsys/radeon/drm/radeon_r300.c b/src/gallium/winsys/radeon/drm/radeon_r300.c
index 80923de937..fb779e4033 100644
--- a/src/gallium/winsys/radeon/drm/radeon_r300.c
+++ b/src/gallium/winsys/radeon/drm/radeon_r300.c
@@ -181,7 +181,7 @@ static boolean radeon_check_cs(struct r300_winsys_screen *rws, int size)
     struct radeon_libdrm_winsys *ws = radeon_winsys_screen(rws);
     struct radeon_cs *cs = ws->cs;
 
-    return radeon_validate(rws) && cs->cdw + size <= cs->ndw;
+    return cs->cdw + size <= cs->ndw;
 }
 
 static void radeon_begin_cs(struct r300_winsys_screen *rws,