Merge branch 'gallium-0.1' into gallium-draw-optigallium-draw-opti

80 files changed, 2357 insertions, 1245 deletions

diff --git a/progs/tests/copypixrate.c b/progs/tests/copypixrate.c
index e9a42a1c8c..142315364c 100644
--- a/progs/tests/copypixrate.c
+++ b/progs/tests/copypixrate.c
@@ -73,6 +73,32 @@ Rand(int max)
 }
 
 
+static void
+BlitOne(void)
+{
+   int x, y;
+
+   do {
+      x = Rand(WinWidth);
+      y = Rand(WinHeight);
+   } while (x <= ImgWidth && y <= ImgHeight);
+
+#ifdef GL_EXT_framebuffer_blit
+   if (UseBlit)
+   {
+      glBlitFramebufferEXT_func(0, 0, ImgWidth, ImgHeight,
+                                x, y, x + ImgWidth, y + ImgHeight,
+                                GL_COLOR_BUFFER_BIT, GL_LINEAR);
+   }
+   else
+#endif
+   {
+      glWindowPos2iARB(x, y);
+      glCopyPixels(0, 0, ImgWidth, ImgHeight, GL_COLOR);
+   }
+}
+
+
 /**
  * Measure glCopyPixels rate
  */
@@ -96,30 +122,14 @@ RunTest(void)
    bpp = (r + g + b + a) / 8;
 
    do {
-      int x, y;
-      x = Rand(WinWidth);
-      y = Rand(WinHeight);
+      BlitOne();
 
-      if (x > ImgWidth || y > ImgHeight) {
-#ifdef GL_EXT_framebuffer_blit
-         if (UseBlit)
-         {
-            glBlitFramebufferEXT_func(0, 0, ImgWidth, ImgHeight,
-                                      x, y, x + ImgWidth, y + ImgHeight,
-                                      GL_COLOR_BUFFER_BIT, GL_LINEAR);
-         }
-         else
-#endif
-         {
-            glWindowPos2iARB(x, y);
-            glCopyPixels(0, 0, ImgWidth, ImgHeight, GL_COLOR);
-         }
-         glFinish(); /* XXX OK? */
+      if (Buffer == GL_FRONT)
+         glFinish(); /* XXX to view progress */
 
-         iters++;
+      iters++;
 
-         t1 = glutGet(GLUT_ELAPSED_TIME) / 1000.0;
-      }
+      t1 = glutGet(GLUT_ELAPSED_TIME) / 1000.0;
    } while (t1 - t0 < 5.0);
 
    glDisable(GL_ALPHA_TEST);
@@ -151,8 +161,10 @@ Draw(void)
    else
       glutSwapBuffers();
 
+#if 1
    printf("exiting\n");
    exit(0);
+#endif
 }
 
 
@@ -175,9 +187,12 @@ Key(unsigned char key, int x, int y)
    (void) x;
    (void) y;
    switch (key) {
-      case 27:
-         exit(0);
-         break;
+   case 'b':
+      BlitOne();
+      break;
+   case 27:
+      exit(0);
+      break;
    }
    glutPostRedisplay();
 }
diff --git a/progs/tests/mipmap_view.c b/progs/tests/mipmap_view.c
index d821f432f0..117e201718 100644
--- a/progs/tests/mipmap_view.c
+++ b/progs/tests/mipmap_view.c
@@ -55,17 +55,10 @@ Display(void)
 
    for (bias = -1; bias < 11; bias++) {
 
-      glRasterPos2f(x, y + TexHeight + 5);
-      sprintf(str, "Texture LOD Bias = %d", bias);
-      PrintString(str);
-
-      glPushMatrix();
-      glTranslatef(x, y, 0);
-
-      glEnable(GL_TEXTURE_2D);
-
       if (ScaleQuads) {
          if (bias > 0) {
+            if (texWidth == 1 && texHeight == 1)
+               break;
             texWidth = TexWidth >> bias;
             texHeight = TexHeight >> bias;
             if (texWidth < 1)
@@ -79,6 +72,20 @@ Display(void)
          glTexEnvf(GL_TEXTURE_FILTER_CONTROL_EXT, GL_TEXTURE_LOD_BIAS_EXT, bias);
       }
 
+      glRasterPos2f(x, y + TexHeight + 5);
+      if (ScaleQuads)
+         sprintf(str, "Texture Level %d: %d x %d",
+                 (bias < 0 ? 0 : bias),
+                 texWidth, texHeight);
+      else
+         sprintf(str, "Texture LOD Bias = %d", bias);
+      PrintString(str);
+
+      glPushMatrix();
+      glTranslatef(x, y, 0);
+
+      glEnable(GL_TEXTURE_2D);
+
       glBegin(GL_POLYGON);
       glTexCoord2f(0, 0);  glVertex2f(0, 0);
       glTexCoord2f(1, 0);  glVertex2f(texWidth, 0);
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.c b/src/gallium/auxiliary/cso_cache/cso_context.c
index 294ac82281..4a1a6cb79c 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.c
+++ b/src/gallium/auxiliary/cso_cache/cso_context.c
@@ -52,11 +52,27 @@ struct cso_context {
    void *samplers[PIPE_MAX_SAMPLERS];
    unsigned nr_samplers;
 
-   void *blend;
-   void *depth_stencil;
-   void *rasterizer;
-   void *fragment_shader;
-   void *vertex_shader;
+   void *samplers_saved[PIPE_MAX_SAMPLERS];
+   unsigned nr_samplers_saved;
+
+   struct pipe_texture *textures[PIPE_MAX_SAMPLERS];
+   uint nr_textures;
+
+   struct pipe_texture *textures_saved[PIPE_MAX_SAMPLERS];
+   uint nr_textures_saved;
+
+   /** Current and saved state.
+    * The saved state is used as a 1-deep stack.
+    */
+   void *blend, *blend_saved;
+   void *depth_stencil, *depth_stencil_saved;
+   void *rasterizer, *rasterizer_saved;
+   void *fragment_shader, *fragment_shader_saved;
+   void *vertex_shader, *vertex_shader_saved;
+
+   struct pipe_framebuffer_state fb, fb_saved;
+   struct pipe_viewport_state vp, vp_saved;
+   struct pipe_blend_color blend_color;
 };
 
 
@@ -149,12 +165,23 @@ void cso_set_blend(struct cso_context *ctx,
    }
 }
 
-void cso_unset_blend(struct cso_context *ctx)
+void cso_save_blend(struct cso_context *ctx)
+{
+   assert(!ctx->blend_saved);
+   ctx->blend_saved = ctx->blend;
+}
+
+void cso_restore_blend(struct cso_context *ctx)
 {
-   ctx->blend = NULL;
+   if (ctx->blend != ctx->blend_saved) {
+      ctx->blend = ctx->blend_saved;
+      ctx->pipe->bind_blend_state(ctx->pipe, ctx->blend_saved);
+   }
+   ctx->blend_saved = NULL;
 }
 
 
+
 void cso_single_sampler(struct cso_context *ctx,
                         unsigned idx,
                         const struct pipe_sampler_state *templ)
@@ -190,9 +217,11 @@ void cso_single_sampler_done( struct cso_context *ctx )
 {
    unsigned i; 
 
-   for (i = 0; i < 8; i++)
-      if (ctx->samplers[i] == NULL)
+   /* find highest non-null sampler */
+   for (i = PIPE_MAX_SAMPLERS; i > 0; i--) {
+      if (ctx->samplers[i - 1] != NULL)
          break;
+   }
 
    ctx->nr_samplers = i;
 
@@ -226,15 +255,50 @@ void cso_set_samplers( struct cso_context *ctx,
    cso_single_sampler_done( ctx );
 }
 
-void cso_unset_samplers( struct cso_context *ctx )
+void cso_save_samplers(struct cso_context *ctx)
+{
+   ctx->nr_samplers_saved = ctx->nr_samplers;
+   memcpy(ctx->samplers_saved, ctx->samplers, sizeof(ctx->samplers));
+}
+
+void cso_restore_samplers(struct cso_context *ctx)
+{
+   cso_set_samplers(ctx, ctx->nr_samplers_saved,
+                    (const struct pipe_sampler_state **) ctx->samplers_saved);
+}
+
+
+void cso_set_sampler_textures( struct cso_context *ctx,
+                               uint count,
+                               struct pipe_texture **textures )
 {
    uint i;
-   for (i = 0; i < ctx->nr_samplers; i++)
-      ctx->samplers[i] = NULL;
+
+   ctx->nr_textures = count;
+
+   for (i = 0; i < count; i++)
+      ctx->textures[i] = textures[i];
+   for ( ; i < PIPE_MAX_SAMPLERS; i++)
+      ctx->textures[i] = NULL;
+
+   ctx->pipe->set_sampler_textures(ctx->pipe, count, textures);
+}
+
+void cso_save_sampler_textures( struct cso_context *ctx )
+{
+   ctx->nr_textures_saved = ctx->nr_textures;
+   memcpy(ctx->textures_saved, ctx->textures, sizeof(ctx->textures));
+}
+
+void cso_restore_sampler_textures( struct cso_context *ctx )
+{
+   cso_set_sampler_textures(ctx, ctx->nr_textures_saved, ctx->textures_saved);
+   ctx->nr_textures_saved = 0;
 }
 
 
 
+
 void cso_set_depth_stencil_alpha(struct cso_context *ctx,
                                  const struct pipe_depth_stencil_alpha_state *templ)
 {
@@ -267,15 +331,25 @@ void cso_set_depth_stencil_alpha(struct cso_context *ctx,
    }
 }
 
-void cso_unset_depth_stencil_alpha(struct cso_context *ctx)
+void cso_save_depth_stencil_alpha(struct cso_context *ctx)
 {
-   ctx->depth_stencil = NULL;
+   assert(!ctx->depth_stencil_saved);
+   ctx->depth_stencil_saved = ctx->depth_stencil;
+}
+
+void cso_restore_depth_stencil_alpha(struct cso_context *ctx)
+{
+   if (ctx->depth_stencil != ctx->depth_stencil_saved) {
+      ctx->depth_stencil = ctx->depth_stencil_saved;
+      ctx->pipe->bind_depth_stencil_alpha_state(ctx->pipe, ctx->depth_stencil_saved);
+   }
+   ctx->depth_stencil_saved = NULL;
 }
 
 
 
 void cso_set_rasterizer(struct cso_context *ctx,
-                              const struct pipe_rasterizer_state *templ)
+                        const struct pipe_rasterizer_state *templ)
 {
    unsigned hash_key = cso_construct_key((void*)templ,
                                          sizeof(struct pipe_rasterizer_state));
@@ -305,11 +379,20 @@ void cso_set_rasterizer(struct cso_context *ctx,
    }
 }
 
-void cso_unset_rasterizer(struct cso_context *ctx)
+void cso_save_rasterizer(struct cso_context *ctx)
 {
-   ctx->rasterizer = NULL;
+   assert(!ctx->rasterizer_saved);
+   ctx->rasterizer_saved = ctx->rasterizer;
 }
 
+void cso_restore_rasterizer(struct cso_context *ctx)
+{
+   if (ctx->rasterizer != ctx->rasterizer_saved) {
+      ctx->rasterizer = ctx->rasterizer_saved;
+      ctx->pipe->bind_rasterizer_state(ctx->pipe, ctx->rasterizer_saved);
+   }
+   ctx->rasterizer_saved = NULL;
+}
 
 
 void cso_set_fragment_shader(struct cso_context *ctx,
@@ -343,12 +426,24 @@ void cso_set_fragment_shader(struct cso_context *ctx,
    }
 }
 
-void cso_unset_fragment_shader(struct cso_context *ctx)
+void cso_save_fragment_shader(struct cso_context *ctx)
 {
-   ctx->fragment_shader = NULL;
+   assert(!ctx->fragment_shader_saved);
+   ctx->fragment_shader_saved = ctx->fragment_shader;
+}
+
+void cso_restore_fragment_shader(struct cso_context *ctx)
+{
+   assert(ctx->fragment_shader_saved);
+   if (ctx->fragment_shader_saved != ctx->fragment_shader) {
+      ctx->pipe->bind_fs_state(ctx->pipe, ctx->fragment_shader_saved);
+      ctx->fragment_shader = ctx->fragment_shader_saved;
+   }
+   ctx->fragment_shader_saved = NULL;
 }
 
 
+
 void cso_set_vertex_shader(struct cso_context *ctx,
                            const struct pipe_shader_state *templ)
 {
@@ -380,7 +475,79 @@ void cso_set_vertex_shader(struct cso_context *ctx,
    }
 }
 
-void cso_unset_vertex_shader(struct cso_context *ctx)
+void cso_save_vertex_shader(struct cso_context *ctx)
+{
+   assert(!ctx->vertex_shader_saved);
+   ctx->vertex_shader_saved = ctx->vertex_shader;
+}
+
+void cso_restore_vertex_shader(struct cso_context *ctx)
+{
+   assert(ctx->vertex_shader_saved);
+   if (ctx->vertex_shader_saved != ctx->vertex_shader) {
+      ctx->pipe->bind_fs_state(ctx->pipe, ctx->vertex_shader_saved);
+      ctx->vertex_shader = ctx->vertex_shader_saved;
+   }
+   ctx->vertex_shader_saved = NULL;
+}
+
+
+
+void cso_set_framebuffer(struct cso_context *ctx,
+                         const struct pipe_framebuffer_state *fb)
+{
+   /* XXX this memcmp() fails to detect buffer size changes */
+   if (1/*memcmp(&ctx->fb, fb, sizeof(*fb))*/) {
+      ctx->fb = *fb;
+      ctx->pipe->set_framebuffer_state(ctx->pipe, fb);
+   }
+}
+
+void cso_save_framebuffer(struct cso_context *ctx)
+{
+   ctx->fb_saved = ctx->fb;
+}
+
+void cso_restore_framebuffer(struct cso_context *ctx)
+{
+   if (memcmp(&ctx->fb, &ctx->fb_saved, sizeof(ctx->fb))) {
+      ctx->fb = ctx->fb_saved;
+      ctx->pipe->set_framebuffer_state(ctx->pipe, &ctx->fb);
+   }
+}
+
+
+void cso_set_viewport(struct cso_context *ctx,
+                      const struct pipe_viewport_state *vp)
+{
+   if (memcmp(&ctx->vp, vp, sizeof(*vp))) {
+      ctx->vp = *vp;
+      ctx->pipe->set_viewport_state(ctx->pipe, vp);
+   }
+}
+
+void cso_save_viewport(struct cso_context *ctx)
 {
-   ctx->vertex_shader = NULL;
+   ctx->vp_saved = ctx->vp;
+}
+
+
+void cso_restore_viewport(struct cso_context *ctx)
+{
+   if (memcmp(&ctx->vp, &ctx->vp_saved, sizeof(ctx->vp))) {
+      ctx->vp = ctx->vp_saved;
+      ctx->pipe->set_viewport_state(ctx->pipe, &ctx->vp);
+   }
+}
+
+
+
+
+void cso_set_blend_color(struct cso_context *ctx,
+                         const struct pipe_blend_color *bc)
+{
+   if (memcmp(&ctx->blend_color, bc, sizeof(ctx->blend_color))) {
+      ctx->blend_color = *bc;
+      ctx->pipe->set_blend_color(ctx->pipe, bc);
+   }
 }
diff --git a/src/gallium/auxiliary/cso_cache/cso_context.h b/src/gallium/auxiliary/cso_cache/cso_context.h
index 6aa619abf5..665e8d9911 100644
--- a/src/gallium/auxiliary/cso_cache/cso_context.h
+++ b/src/gallium/auxiliary/cso_cache/cso_context.h
@@ -41,27 +41,36 @@ struct cso_context;
 
 struct cso_context *cso_create_context( struct pipe_context *pipe );
 
+void cso_destroy_context( struct cso_context *cso );
+
+
+
 void cso_set_blend( struct cso_context *cso,
                     const struct pipe_blend_state *blend );
+void cso_save_blend(struct cso_context *cso);
+void cso_restore_blend(struct cso_context *cso);
+
 
-void cso_unset_blend(struct cso_context *cso);
 
 void cso_set_depth_stencil_alpha( struct cso_context *cso,
                                   const struct pipe_depth_stencil_alpha_state *dsa );
+void cso_save_depth_stencil_alpha(struct cso_context *cso);
+void cso_restore_depth_stencil_alpha(struct cso_context *cso);
+
 
-void cso_unset_depth_stencil_alpha( struct cso_context *cso );
 
 void cso_set_rasterizer( struct cso_context *cso,
                          const struct pipe_rasterizer_state *rasterizer );
+void cso_save_rasterizer(struct cso_context *cso);
+void cso_restore_rasterizer(struct cso_context *cso);
+
 
-void cso_unset_rasterizer( struct cso_context *cso );
 
 void cso_set_samplers( struct cso_context *cso,
                        unsigned count,
                        const struct pipe_sampler_state **states );
-
-void cso_unset_samplers( struct cso_context *cso );
-
+void cso_save_samplers(struct cso_context *cso);
+void cso_restore_samplers(struct cso_context *cso);
 
 /* Alternate interface to support state trackers that like to modify
  * samplers one at a time:
@@ -73,6 +82,15 @@ void cso_single_sampler( struct cso_context *cso,
 void cso_single_sampler_done( struct cso_context *cso );
 
 
+
+void cso_set_sampler_textures( struct cso_context *cso,
+                               uint count,
+                               struct pipe_texture **textures );
+void cso_save_sampler_textures( struct cso_context *cso );
+void cso_restore_sampler_textures( struct cso_context *cso );
+
+
+
 /* These aren't really sensible -- most of the time the api provides
  * object semantics for shaders anyway, and the cases where it doesn't
  * (eg mesa's internall-generated texenv programs), it will be up to
@@ -80,15 +98,32 @@ void cso_single_sampler_done( struct cso_context *cso );
  */
 void cso_set_fragment_shader( struct cso_context *cso,
                               const struct pipe_shader_state *shader );
+void cso_save_fragment_shader(struct cso_context *cso);
+void cso_restore_fragment_shader(struct cso_context *cso);
+
 
-void cso_unset_fragment_shader( struct cso_context *cso );
 
 void cso_set_vertex_shader( struct cso_context *cso,
                             const struct pipe_shader_state *shader );
+void cso_save_vertex_shader(struct cso_context *cso);
+void cso_restore_vertex_shader(struct cso_context *cso);
 
-void cso_unset_vertex_shader( struct cso_context *cso );
 
-void cso_destroy_context( struct cso_context *cso );
+
+void cso_set_framebuffer(struct cso_context *cso,
+                         const struct pipe_framebuffer_state *fb);
+void cso_save_framebuffer(struct cso_context *cso);
+void cso_restore_framebuffer(struct cso_context *cso);
+
+
+void cso_set_viewport(struct cso_context *cso,
+                      const struct pipe_viewport_state *vp);
+void cso_save_viewport(struct cso_context *cso);
+void cso_restore_viewport(struct cso_context *cso);
+
+
+void cso_set_blend_color(struct cso_context *cso,
+                         const struct pipe_blend_color *bc);
 
 
 #ifdef	__cplusplus
diff --git a/src/gallium/auxiliary/draw/draw_aaline.c b/src/gallium/auxiliary/draw/draw_aaline.c
index f2b983374e..6742f7f4b9 100644
--- a/src/gallium/auxiliary/draw/draw_aaline.c
+++ b/src/gallium/auxiliary/draw/draw_aaline.c
@@ -410,6 +410,7 @@ aaline_create_texture(struct aaline_stage *aaline)
       /* unmap */
       pipe_surface_unmap(surface);
       pipe_surface_reference(&surface, NULL);
+      pipe->texture_update(pipe, aaline->texture, 0, (1 << level));
    }
 }
 
@@ -664,7 +665,14 @@ aaline_reset_stipple_counter(struct draw_stage *stage)
 static void
 aaline_destroy(struct draw_stage *stage)
 {
+   struct aaline_stage *aaline = aaline_stage(stage);
+
+   aaline->pipe->delete_sampler_state(aaline->pipe, aaline->sampler_cso);
+
+   pipe_texture_release(&aaline->texture);
+
    draw_free_temp_verts( stage );
+
    FREE( stage );
 }
 
diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index 4cca965ac1..41da93cdf8 100644
--- a/src/gallium/auxiliary/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -113,6 +113,8 @@ void draw_destroy( struct draw_context *draw )
       draw->pipeline.aaline->destroy( draw->pipeline.aaline );
    if (draw->pipeline.aapoint)
       draw->pipeline.aapoint->destroy( draw->pipeline.aapoint );
+   if (draw->pipeline.pstipple)
+      draw->pipeline.pstipple->destroy( draw->pipeline.pstipple );
    if (draw->pipeline.rasterize)
       draw->pipeline.rasterize->destroy( draw->pipeline.rasterize );
    tgsi_exec_machine_free_data(&draw->machine);
diff --git a/src/gallium/auxiliary/draw/draw_pstipple.c b/src/gallium/auxiliary/draw/draw_pstipple.c
index 09d542002f..bd8d3a76ae 100644
--- a/src/gallium/auxiliary/draw/draw_pstipple.c
+++ b/src/gallium/auxiliary/draw/draw_pstipple.c
@@ -374,7 +374,7 @@ pstip_update_texture(struct pstip_stage *pstip)
    /* unmap */
    pipe_surface_unmap(surface);
    pipe_surface_reference(&surface, NULL);
-   pipe->texture_update(pipe, pstip->texture);
+   pipe->texture_update(pipe, pstip->texture, 0, 0x1);
 }
 
 
@@ -534,6 +534,12 @@ pstip_reset_stipple_counter(struct draw_stage *stage)
 static void
 pstip_destroy(struct draw_stage *stage)
 {
+   struct pstip_stage *pstip = pstip_stage(stage);
+
+   pstip->pipe->delete_sampler_state(pstip->pipe, pstip->sampler_cso);
+
+   pipe_texture_release(&pstip->texture);
+
    draw_free_temp_verts( stage );
    FREE( stage );
 }
diff --git a/src/gallium/auxiliary/util/p_debug.c b/src/gallium/auxiliary/util/p_debug.c
index bb84e8096d..c51e9e6a69 100644
--- a/src/gallium/auxiliary/util/p_debug.c
+++ b/src/gallium/auxiliary/util/p_debug.c
@@ -52,6 +52,9 @@ rpl_EngDebugPrint(const char *format, ...)
 }
 
 int rpl_vsnprintf(char *, size_t, const char *, va_list);
+int rpl_snprintf(char *str, size_t size, const char *format, ...);
+#define vsnprintf rpl_vsnprintf
+#define snprintf rpl_snprintf
 #endif
 
 
@@ -82,6 +85,22 @@ void debug_printf(const char *format, ...)
 }
 
 
+void debug_print_blob( const char *name,
+                       const void *blob,
+                       unsigned size )
+{
+   const unsigned *ublob = (const unsigned *)blob;
+   unsigned i;
+
+   debug_printf("%s (%d dwords%s)\n", name, size/4,
+                size%4 ? "... plus a few bytes" : "");
+
+   for (i = 0; i < size/4; i++) {
+      debug_printf("%d:\t%08x\n", i, ublob[i]);
+   }
+}
+
+
 /* TODO: implement a debug_abort that calls EngBugCheckEx on WIN32 */
 
 
@@ -181,3 +200,61 @@ void debug_mask_vprintf(uint32_t uuid, uint32_t what, const char *format, va_lis
    if(mask & what)
       debug_vprintf(format, ap);
 }
+
+
+const char *
+debug_dump_enum(const struct debug_named_value *names, 
+                unsigned long value)
+{
+   static char rest[256];
+   
+   while(names->name) {
+      if(names->value == value)
+	 return names->name;
+      ++names;
+   }
+
+   snprintf(rest, sizeof(rest), "0x%08x", value);
+   return rest;
+}
+
+
+const char *
+debug_dump_flags(const struct debug_named_value *names, 
+                 unsigned long value)
+{
+   static char output[4096];
+   static char rest[256];
+   int first = 1;
+
+   output[0] = '\0';
+
+   while(names->name) {
+      if((names->value & value) == names->value) {
+	 if (!first)
+	    strncat(output, "|", sizeof(output));
+	 else
+	    first = 0;
+	 strncat(output, names->name, sizeof(output));
+	 value &= ~names->value;
+      }
+      ++names;
+   }
+   
+   if (value) {
+      if (!first)
+	 strncat(output, "|", sizeof(output));
+      else
+	 first = 0;
+      
+      snprintf(rest, sizeof(rest), "0x%08x", value);
+      strncat(output, rest, sizeof(output));
+   }
+   
+   if(first)
+      return "0";
+   
+   return output;
+}
+
+
diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c
index 4b4ab8185f..28a404fd01 100644
--- a/src/gallium/auxiliary/util/u_blit.c
+++ b/src/gallium/auxiliary/util/u_blit.c
@@ -45,19 +45,25 @@
 #include "util/u_blit.h"
 #include "util/u_simple_shaders.h"
 
+#include "cso_cache/cso_context.h"
+
 
 struct blit_state
 {
    struct pipe_context *pipe;
+   struct cso_context *cso;
 
-   void *blend;
-   void *depthstencil;
-   void *rasterizer;
-   void *samplers[2];  /* one for linear, one for nearest sampling */
+   struct pipe_blend_state blend;
+   struct pipe_depth_stencil_alpha_state depthstencil;
+   struct pipe_rasterizer_state rasterizer;
+   struct pipe_sampler_state sampler;
 
    /*struct pipe_viewport_state viewport;*/
    struct pipe_sampler_state *vs;
    struct pipe_sampler_state *fs;
+
+   struct pipe_buffer *vbuf;  /**< quad vertices */
+   float vertices[4][2][4];   /**< vertex/texcoords for quad */
 };
 
 
@@ -66,56 +72,45 @@ struct blit_state
  * Intended to be created once and re-used for many blit() calls.
  */
 struct blit_state *
-util_create_blit(struct pipe_context *pipe)
+util_create_blit(struct pipe_context *pipe, struct cso_context *cso)
 {
-   struct pipe_blend_state blend;
-   struct pipe_depth_stencil_alpha_state depthstencil;
-   struct pipe_rasterizer_state rasterizer;
    struct blit_state *ctx;
-   struct pipe_sampler_state sampler;
+   uint i;
 
    ctx = CALLOC_STRUCT(blit_state);
    if (!ctx)
       return NULL;
 
    ctx->pipe = pipe;
+   ctx->cso = cso;
 
-   /* we don't use blending, but need to set valid values */
-   memset(&blend, 0, sizeof(blend));
-   blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
-   blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
-   blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
-   blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
-   blend.colormask = PIPE_MASK_RGBA;
-   ctx->blend = pipe->create_blend_state(pipe, &blend);
+   /* disabled blending/masking */
+   memset(&ctx->blend, 0, sizeof(ctx->blend));
+   ctx->blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   ctx->blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   ctx->blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   ctx->blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   ctx->blend.colormask = PIPE_MASK_RGBA;
 
-   /* depth/stencil/alpha */
-   memset(&depthstencil, 0, sizeof(depthstencil));
-   ctx->depthstencil = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil);
+   /* no-op depth/stencil/alpha */
+   memset(&ctx->depthstencil, 0, sizeof(ctx->depthstencil));
 
    /* rasterizer */
-   memset(&rasterizer, 0, sizeof(rasterizer));
-   rasterizer.front_winding = PIPE_WINDING_CW;
-   rasterizer.cull_mode = PIPE_WINDING_NONE;
-   rasterizer.bypass_clipping = 1;  /* bypasses viewport too */
-   /*rasterizer.bypass_vs = 1;*/
-   ctx->rasterizer = pipe->create_rasterizer_state(pipe, &rasterizer);
+   memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer));
+   ctx->rasterizer.front_winding = PIPE_WINDING_CW;
+   ctx->rasterizer.cull_mode = PIPE_WINDING_NONE;
+   ctx->rasterizer.bypass_clipping = 1;  /* bypasses viewport too */
+   /*ctx->rasterizer.bypass_vs = 1;*/
 
    /* samplers */
-   memset(&sampler, 0, sizeof(sampler));
-   sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-   sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-   sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-   sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
-   sampler.min_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
-   sampler.mag_img_filter = PIPE_TEX_MIPFILTER_NEAREST;
-   sampler.normalized_coords = 1;
-   ctx->samplers[0] = pipe->create_sampler_state(pipe, &sampler);
-
-   sampler.min_img_filter = PIPE_TEX_MIPFILTER_LINEAR;
-   sampler.mag_img_filter = PIPE_TEX_MIPFILTER_LINEAR;
-   ctx->samplers[1] = pipe->create_sampler_state(pipe, &sampler);
-
+   memset(&ctx->sampler, 0, sizeof(ctx->sampler));
+   ctx->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+   ctx->sampler.min_img_filter = 0; /* set later */
+   ctx->sampler.mag_img_filter = 0; /* set later */
+   ctx->sampler.normalized_coords = 1;
 
 #if 0
    /* viewport */
@@ -141,6 +136,24 @@ util_create_blit(struct pipe_context *pipe)
    /* fragment shader */
    ctx->fs = util_make_fragment_tex_shader(pipe);
 
+   ctx->vbuf = pipe->winsys->buffer_create(pipe->winsys,
+                                           32,
+                                           PIPE_BUFFER_USAGE_VERTEX,
+                                           sizeof(ctx->vertices));
+   if (!ctx->vbuf) {
+      FREE(ctx);
+      ctx->pipe->delete_fs_state(ctx->pipe, ctx->fs);
+      ctx->pipe->delete_vs_state(ctx->pipe, ctx->vs);
+      return NULL;
+   }
+
+   /* init vertex data that doesn't change */
+   for (i = 0; i < 4; i++) {
+      ctx->vertices[i][0][3] = 1.0f; /* w */
+      ctx->vertices[i][1][2] = 0.0f; /* r */
+      ctx->vertices[i][1][3] = 1.0f; /* q */
+   }
+
    return ctx;
 }
 
@@ -153,20 +166,59 @@ util_destroy_blit(struct blit_state *ctx)
 {
    struct pipe_context *pipe = ctx->pipe;
 
-   pipe->delete_blend_state(pipe, ctx->blend);
-   pipe->delete_depth_stencil_alpha_state(pipe, ctx->depthstencil);
-   pipe->delete_rasterizer_state(pipe, ctx->rasterizer);
-   pipe->delete_sampler_state(pipe, ctx->samplers[0]);
-   pipe->delete_sampler_state(pipe, ctx->samplers[1]);
-
    pipe->delete_vs_state(pipe, ctx->vs);
    pipe->delete_fs_state(pipe, ctx->fs);
 
+   pipe->winsys->buffer_destroy(pipe->winsys, ctx->vbuf);
+
    FREE(ctx);
 }
 
 
 /**
+ * Setup vertex data for the textured quad we'll draw.
+ * Note: y=0=top
+ */
+static void
+setup_vertex_data(struct blit_state *ctx,
+                  float x0, float y0, float x1, float y1, float z)
+{
+   void *buf;
+
+   ctx->vertices[0][0][0] = x0;
+   ctx->vertices[0][0][1] = y0;
+   ctx->vertices[0][0][2] = z;
+   ctx->vertices[0][1][0] = 0.0f; /*s*/
+   ctx->vertices[0][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[1][0][0] = x1;
+   ctx->vertices[1][0][1] = y0;
+   ctx->vertices[1][0][2] = z;
+   ctx->vertices[1][1][0] = 1.0f; /*s*/
+   ctx->vertices[1][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[2][0][0] = x1;
+   ctx->vertices[2][0][1] = y1;
+   ctx->vertices[2][0][2] = z;
+   ctx->vertices[2][1][0] = 1.0f;
+   ctx->vertices[2][1][1] = 1.0f;
+
+   ctx->vertices[3][0][0] = x0;
+   ctx->vertices[3][0][1] = y1;
+   ctx->vertices[3][0][2] = z;
+   ctx->vertices[3][1][0] = 0.0f;
+   ctx->vertices[3][1][1] = 1.0f;
+
+   buf = ctx->pipe->winsys->buffer_map(ctx->pipe->winsys, ctx->vbuf,
+                                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   memcpy(buf, ctx->vertices, sizeof(ctx->vertices));
+
+   ctx->pipe->winsys->buffer_unmap(ctx->pipe->winsys, ctx->vbuf);
+}
+
+
+/**
  * Copy pixel block from src surface to dst surface.
  * Overlapping regions are acceptable.
  * XXX need some control over blitting Z and/or stencil.
@@ -236,17 +288,24 @@ util_blit_pixels(struct blit_state *ctx,
                       src, srcLeft, srcTop, /* src */
                       srcW, srcH);     /* size */
 
-   /* drawing dest */
-   memset(&fb, 0, sizeof(fb));
-   fb.num_cbufs = 1;
-   fb.cbufs[0] = dst;
-   pipe->set_framebuffer_state(pipe, &fb);
+   /* save state (restored below) */
+   cso_save_blend(ctx->cso);
+   cso_save_depth_stencil_alpha(ctx->cso);
+   cso_save_rasterizer(ctx->cso);
+   cso_save_samplers(ctx->cso);
+   cso_save_sampler_textures(ctx->cso);
+   cso_save_framebuffer(ctx->cso);
+
+   /* set misc state we care about */
+   cso_set_blend(ctx->cso, &ctx->blend);
+   cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil);
+   cso_set_rasterizer(ctx->cso, &ctx->rasterizer);
 
    /* sampler */
-   if (filter == PIPE_TEX_MIPFILTER_NEAREST)
-      pipe->bind_sampler_states(pipe, 1, &ctx->samplers[0]);
-   else
-      pipe->bind_sampler_states(pipe, 1, &ctx->samplers[1]);
+   ctx->sampler.min_img_filter = filter;
+   ctx->sampler.mag_img_filter = filter;
+   cso_single_sampler(ctx->cso, 0, &ctx->sampler);
+   cso_single_sampler_done(ctx->cso);
 
    /* texture */
    pipe->set_sampler_textures(pipe, 1, &tex);
@@ -255,22 +314,34 @@ util_blit_pixels(struct blit_state *ctx,
    pipe->bind_fs_state(pipe, ctx->fs);
    pipe->bind_vs_state(pipe, ctx->vs);
 
-   /* misc state */
-   pipe->bind_blend_state(pipe, ctx->blend);
-   pipe->bind_depth_stencil_alpha_state(pipe, ctx->depthstencil);
-   pipe->bind_rasterizer_state(pipe, ctx->rasterizer);
+   /* drawing dest */
+   memset(&fb, 0, sizeof(fb));
+   fb.width = dst->width;
+   fb.height = dst->height;
+   fb.num_cbufs = 1;
+   fb.cbufs[0] = dst;
+   cso_set_framebuffer(ctx->cso, &fb);
 
    /* draw quad */
-   util_draw_texquad(pipe, dstX0, dstY0, dstX1, dstY1, z);
-
-   /* unbind */
-   pipe->set_sampler_textures(pipe, 0, NULL);
-   pipe->bind_sampler_states(pipe, 0, NULL);
-
-   /* free stuff */
+   setup_vertex_data(ctx,
+                     (float) dstX0, (float) dstY0, 
+                     (float) dstX1, (float) dstY1, z);
+
+   util_draw_vertex_buffer(ctx->pipe, ctx->vbuf,
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           2); /* attribs/vert */
+
+   /* restore state we changed */
+   cso_restore_blend(ctx->cso);
+   cso_restore_depth_stencil_alpha(ctx->cso);
+   cso_restore_rasterizer(ctx->cso);
+   cso_restore_samplers(ctx->cso);
+   cso_restore_sampler_textures(ctx->cso);
+   cso_restore_framebuffer(ctx->cso);
+
+   /* free the texture */
    pipe_surface_reference(&texSurf, NULL);
    screen->texture_release(screen, &tex);
-
-   /* Note: caller must restore pipe/gallium state at this time */
 }
 
diff --git a/src/gallium/auxiliary/util/u_blit.h b/src/gallium/auxiliary/util/u_blit.h
index a349be99ad..61f1d9bb32 100644
--- a/src/gallium/auxiliary/util/u_blit.h
+++ b/src/gallium/auxiliary/util/u_blit.h
@@ -30,15 +30,16 @@
 #define U_BLIT_H
 
 
+
 struct pipe_context;
 struct pipe_surface;
+struct cso_context;
 
 
 struct blit_state;
 
-
 extern struct blit_state *
-util_create_blit(struct pipe_context *pipe);
+util_create_blit(struct pipe_context *pipe, struct cso_context *cso);
 
 
 extern void
diff --git a/src/gallium/auxiliary/util/u_draw_quad.c b/src/gallium/auxiliary/util/u_draw_quad.c
index 79a69de633..37e8533609 100644
--- a/src/gallium/auxiliary/util/u_draw_quad.c
+++ b/src/gallium/auxiliary/util/u_draw_quad.c
@@ -34,15 +34,51 @@
 
 
 /**
+ * Draw a simple vertex buffer / primitive.
+ * Limited to float[4] vertex attribs, tightly packed.
+ */
+void 
+util_draw_vertex_buffer(struct pipe_context *pipe,
+                        struct pipe_buffer *vbuf,
+                        uint prim_type,
+                        uint num_verts,
+                        uint num_attribs)
+{
+   struct pipe_vertex_buffer vbuffer;
+   struct pipe_vertex_element velement;
+   uint i;
+
+   /* tell pipe about the vertex buffer */
+   vbuffer.buffer = vbuf;
+   vbuffer.pitch = num_attribs * 4 * sizeof(float);  /* vertex size */
+   vbuffer.buffer_offset = 0;
+   pipe->set_vertex_buffer(pipe, 0, &vbuffer);
+
+   /* tell pipe about the vertex attributes */
+   for (i = 0; i < num_attribs; i++) {
+      velement.src_offset = i * 4 * sizeof(float);
+      velement.vertex_buffer_index = 0;
+      velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+      velement.nr_components = 4;
+      pipe->set_vertex_element(pipe, i, &velement);
+   }
+
+   /* draw */
+   pipe->draw_arrays(pipe, prim_type, 0, num_verts);
+}
+
+
+
+/**
  * Draw screen-aligned textured quad.
+ * Note: this function allocs/destroys a vertex buffer and isn't especially
+ * efficient.
  */
 void 
 util_draw_texquad(struct pipe_context *pipe,
                   float x0, float y0, float x1, float y1, float z)
 {
    struct pipe_buffer *vbuf;
-   struct pipe_vertex_buffer vbuffer;
-   struct pipe_vertex_element velement;
    uint numAttribs = 2, vertexBytes, i, j;
    float *v;
 
@@ -89,24 +125,7 @@ util_draw_texquad(struct pipe_context *pipe,
 
    pipe->winsys->buffer_unmap(pipe->winsys, vbuf);
 
-   /* tell pipe about the vertex buffer */
-   vbuffer.buffer = vbuf;
-   vbuffer.pitch = numAttribs * 4 * sizeof(float);  /* vertex size */
-   vbuffer.buffer_offset = 0;
-   pipe->set_vertex_buffer(pipe, 0, &vbuffer);
-
-   /* tell pipe about the vertex attributes */
-   for (i = 0; i < numAttribs; i++) {
-      velement.src_offset = i * 4 * sizeof(float);
-      velement.vertex_buffer_index = 0;
-      velement.src_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
-      velement.nr_components = 4;
-      pipe->set_vertex_element(pipe, i, &velement);
-   }
-
-   /* draw */
-   pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_FAN, 0, 4);
+   util_draw_vertex_buffer(pipe, vbuf, PIPE_PRIM_TRIANGLE_FAN, 4, 2);
 
-   /* XXX: do one-time */
    pipe_buffer_reference(pipe->winsys, &vbuf, NULL);
 }
diff --git a/src/gallium/auxiliary/util/u_draw_quad.h b/src/gallium/auxiliary/util/u_draw_quad.h
index a97f55d2ef..5b6539a99c 100644
--- a/src/gallium/auxiliary/util/u_draw_quad.h
+++ b/src/gallium/auxiliary/util/u_draw_quad.h
@@ -29,9 +29,25 @@
 #define U_DRAWQUAD_H
 
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
+extern void 
+util_draw_vertex_buffer(struct pipe_context *pipe,
+                        struct pipe_buffer *vbuf,
+                        uint num_attribs, uint num_verts, uint prim_type);
+
+
 extern void 
 util_draw_texquad(struct pipe_context *pipe,
                   float x0, float y0, float x1, float y1, float z);
 
 
+#ifdef __cplusplus
+}
+#endif
+
+
 #endif
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index e18f8ab72a..cf02f00b1b 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -49,17 +49,24 @@
 #include "tgsi/util/tgsi_dump.h"
 #include "tgsi/util/tgsi_parse.h"
 
+#include "cso_cache/cso_context.h"
+
 
 struct gen_mipmap_state
 {
    struct pipe_context *pipe;
+   struct cso_context *cso;
 
-   void *blend;
-   void *depthstencil;
-   void *rasterizer;
+   struct pipe_blend_state blend;
+   struct pipe_depth_stencil_alpha_state depthstencil;
+   struct pipe_rasterizer_state rasterizer;
+   struct pipe_sampler_state sampler;
    /*struct pipe_viewport_state viewport;*/
    struct pipe_sampler_state *vs;
    struct pipe_sampler_state *fs;
+
+   struct pipe_buffer *vbuf;  /**< quad vertices */
+   float vertices[4][2][4];   /**< vertex/texcoords for quad */
 };
 
 
@@ -675,39 +682,47 @@ fallback_gen_mipmap(struct gen_mipmap_state *ctx,
  * generate a mipmap.
  */
 struct gen_mipmap_state *
-util_create_gen_mipmap(struct pipe_context *pipe)
+util_create_gen_mipmap(struct pipe_context *pipe,
+                       struct cso_context *cso)
 {
-   struct pipe_blend_state blend;
-   struct pipe_depth_stencil_alpha_state depthstencil;
-   struct pipe_rasterizer_state rasterizer;
    struct gen_mipmap_state *ctx;
+   uint i;
 
    ctx = CALLOC_STRUCT(gen_mipmap_state);
    if (!ctx)
       return NULL;
 
    ctx->pipe = pipe;
+   ctx->cso = cso;
 
-   /* we don't use blending, but need to set valid values */
-   memset(&blend, 0, sizeof(blend));
-   blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
-   blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
-   blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
-   blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
-   blend.colormask = PIPE_MASK_RGBA;
-   ctx->blend = pipe->create_blend_state(pipe, &blend);
+   /* disabled blending/masking */
+   memset(&ctx->blend, 0, sizeof(ctx->blend));
+   ctx->blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+   ctx->blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+   ctx->blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   ctx->blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+   ctx->blend.colormask = PIPE_MASK_RGBA;
 
-   /* depth/stencil/alpha */
-   memset(&depthstencil, 0, sizeof(depthstencil));
-   ctx->depthstencil = pipe->create_depth_stencil_alpha_state(pipe, &depthstencil);
+   /* no-op depth/stencil/alpha */
+   memset(&ctx->depthstencil, 0, sizeof(ctx->depthstencil));
 
    /* rasterizer */
-   memset(&rasterizer, 0, sizeof(rasterizer));
-   rasterizer.front_winding = PIPE_WINDING_CW;
-   rasterizer.cull_mode = PIPE_WINDING_NONE;
-   rasterizer.bypass_clipping = 1;  /* bypasses viewport too */
-   //rasterizer.bypass_vs = 1;
-   ctx->rasterizer = pipe->create_rasterizer_state(pipe, &rasterizer);
+   memset(&ctx->rasterizer, 0, sizeof(ctx->rasterizer));
+   ctx->rasterizer.front_winding = PIPE_WINDING_CW;
+   ctx->rasterizer.cull_mode = PIPE_WINDING_NONE;
+   ctx->rasterizer.bypass_clipping = 1;  /* bypasses viewport too */
+   /*ctx->rasterizer.bypass_vs = 1;*/
+
+   /* sampler state */
+   memset(&ctx->sampler, 0, sizeof(ctx->sampler));
+   ctx->sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+   ctx->sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST;
+   ctx->sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR;
+   ctx->sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR;
+   ctx->sampler.normalized_coords = 1;
+
 
 #if 0
    /* viewport */
@@ -733,10 +748,62 @@ util_create_gen_mipmap(struct pipe_context *pipe)
    /* fragment shader */
    ctx->fs = util_make_fragment_tex_shader(pipe);
 
+   ctx->vbuf = pipe->winsys->buffer_create(pipe->winsys,
+                                           32,
+                                           PIPE_BUFFER_USAGE_VERTEX,
+                                           sizeof(ctx->vertices));
+   if (!ctx->vbuf) {
+      FREE(ctx);
+      return NULL;
+   }
+
+   /* vertex data that doesn't change */
+   for (i = 0; i < 4; i++) {
+      ctx->vertices[i][0][2] = 0.0f; /* z */
+      ctx->vertices[i][0][3] = 1.0f; /* w */
+      ctx->vertices[i][1][2] = 0.0f; /* r */
+      ctx->vertices[i][1][3] = 1.0f; /* q */
+   }
+
    return ctx;
 }
 
 
+static void
+set_vertex_data(struct gen_mipmap_state *ctx, float width, float height)
+{
+   void *buf;
+
+   ctx->vertices[0][0][0] = 0.0f; /*x*/
+   ctx->vertices[0][0][1] = 0.0f; /*y*/
+   ctx->vertices[0][1][0] = 0.0f; /*s*/
+   ctx->vertices[0][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[1][0][0] = width; /*x*/
+   ctx->vertices[1][0][1] = 0.0f;  /*y*/
+   ctx->vertices[1][1][0] = 1.0f; /*s*/
+   ctx->vertices[1][1][1] = 0.0f; /*t*/
+
+   ctx->vertices[2][0][0] = width;
+   ctx->vertices[2][0][1] = height;
+   ctx->vertices[2][1][0] = 1.0f;
+   ctx->vertices[2][1][1] = 1.0f;
+
+   ctx->vertices[3][0][0] = 0.0f;
+   ctx->vertices[3][0][1] = height;
+   ctx->vertices[3][1][0] = 0.0f;
+   ctx->vertices[3][1][1] = 1.0f;
+
+   buf = ctx->pipe->winsys->buffer_map(ctx->pipe->winsys, ctx->vbuf,
+                                       PIPE_BUFFER_USAGE_CPU_WRITE);
+
+   memcpy(buf, ctx->vertices, sizeof(ctx->vertices));
+
+   ctx->pipe->winsys->buffer_unmap(ctx->pipe->winsys, ctx->vbuf);
+}
+
+
+
 /**
  * Destroy a mipmap generation context
  */
@@ -745,12 +812,11 @@ util_destroy_gen_mipmap(struct gen_mipmap_state *ctx)
 {
    struct pipe_context *pipe = ctx->pipe;
 
-   pipe->delete_blend_state(pipe, ctx->blend);
-   pipe->delete_depth_stencil_alpha_state(pipe, ctx->depthstencil);
-   pipe->delete_rasterizer_state(pipe, ctx->rasterizer);
    pipe->delete_vs_state(pipe, ctx->vs);
    pipe->delete_fs_state(pipe, ctx->fs);
 
+   pipe->winsys->buffer_destroy(pipe->winsys, ctx->vbuf);
+
    FREE(ctx);
 }
 
@@ -792,8 +858,6 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
    struct pipe_context *pipe = ctx->pipe;
    struct pipe_screen *screen = pipe->screen;
    struct pipe_framebuffer_state fb;
-   struct pipe_sampler_state sampler;
-   void *sampler_cso;
    uint dstLevel;
    uint zslice = 0;
 
@@ -803,30 +867,29 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
       return;
    }
 
-   /* init framebuffer state */
-   memset(&fb, 0, sizeof(fb));
-   fb.num_cbufs = 1;
-
-   /* sampler state */
-   memset(&sampler, 0, sizeof(sampler));
-   sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-   sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-   sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-   sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST;
-   sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR;
-   sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR;
-   sampler.normalized_coords = 1;
+   /* save state (restored below) */
+   cso_save_blend(ctx->cso);
+   cso_save_depth_stencil_alpha(ctx->cso);
+   cso_save_rasterizer(ctx->cso);
+   cso_save_samplers(ctx->cso);
+   cso_save_sampler_textures(ctx->cso);
+   cso_save_framebuffer(ctx->cso);
 
    /* bind our state */
-   pipe->bind_blend_state(pipe, ctx->blend);
-   pipe->bind_depth_stencil_alpha_state(pipe, ctx->depthstencil);
-   pipe->bind_rasterizer_state(pipe, ctx->rasterizer);
+   cso_set_blend(ctx->cso, &ctx->blend);
+   cso_set_depth_stencil_alpha(ctx->cso, &ctx->depthstencil);
+   cso_set_rasterizer(ctx->cso, &ctx->rasterizer);
+
    pipe->bind_vs_state(pipe, ctx->vs);
    pipe->bind_fs_state(pipe, ctx->fs);
 #if 0
    pipe->set_viewport_state(pipe, &ctx->viewport);
 #endif
 
+   /* init framebuffer state */
+   memset(&fb, 0, sizeof(fb));
+   fb.num_cbufs = 1;
+
    /*
     * XXX for small mipmap levels, it may be faster to use the software
     * fallback path...
@@ -838,7 +901,9 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
        * Setup framebuffer / dest surface
        */
       fb.cbufs[0] = screen->get_tex_surface(screen, pt, face, dstLevel, zslice);
-      pipe->set_framebuffer_state(pipe, &fb);
+      fb.width = pt->width[dstLevel];
+      fb.height = pt->height[dstLevel];
+      cso_set_framebuffer(ctx->cso, &fb);
 
       /*
        * Setup sampler state
@@ -847,11 +912,10 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
        * has trouble with min clamping so we also set the lod_bias to
        * try to work around that.
        */
-      sampler.min_lod = sampler.max_lod = (float) srcLevel;
-      sampler.lod_bias = (float) srcLevel;
-      sampler_cso = pipe->create_sampler_state(pipe, &sampler);
-      pipe->bind_sampler_states(pipe, 1, &sampler_cso);
-
+      ctx->sampler.min_lod = ctx->sampler.max_lod = (float) srcLevel;
+      ctx->sampler.lod_bias = (float) srcLevel;
+      cso_single_sampler(ctx->cso, 0, &ctx->sampler);
+      cso_single_sampler_done(ctx->cso);
 #if 0
       simple_viewport(pipe, pt->width[dstLevel], pt->height[dstLevel]);
 #endif
@@ -859,19 +923,25 @@ util_gen_mipmap(struct gen_mipmap_state *ctx,
       pipe->set_sampler_textures(pipe, 1, &pt);
 
       /* quad coords in window coords (bypassing clipping, viewport mapping) */
-      util_draw_texquad(pipe,
-                        0.0F, 0.0F, /* x0, y0 */
-                        (float) pt->width[dstLevel], /* x1 */
-                        (float) pt->height[dstLevel], /* y1 */
-                        0.0F);  /* z */
-
+      set_vertex_data(ctx,
+                      (float) pt->width[dstLevel],
+                      (float) pt->height[dstLevel]);
+      util_draw_vertex_buffer(ctx->pipe, ctx->vbuf,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
 
       pipe->flush(pipe, PIPE_FLUSH_WAIT);
 
-      /*pipe->texture_update(pipe, pt);  not really needed */
-
-      pipe->delete_sampler_state(pipe, sampler_cso);
+      /* need to signal that the texture has changed _after_ rendering to it */
+      pipe->texture_update(pipe, pt, face, (1 << dstLevel));
    }
 
-   /* Note: caller must restore pipe/gallium state at this time */
+   /* restore state we changed */
+   cso_restore_blend(ctx->cso);
+   cso_restore_depth_stencil_alpha(ctx->cso);
+   cso_restore_rasterizer(ctx->cso);
+   cso_restore_samplers(ctx->cso);
+   cso_restore_sampler_textures(ctx->cso);
+   cso_restore_framebuffer(ctx->cso);
 }
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.h b/src/gallium/auxiliary/util/u_gen_mipmap.h
index 80496140a2..eeabf3bf07 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.h
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.h
@@ -31,11 +31,15 @@
 #include "pipe/p_state.h"
 
 
+struct pipe_context;
+struct pipe_texture;
+struct cso_context;
+
 struct gen_mipmap_state;
 
 
 extern struct gen_mipmap_state *
-util_create_gen_mipmap(struct pipe_context *pipe);
+util_create_gen_mipmap(struct pipe_context *pipe, struct cso_context *cso);
 
 
 extern void
diff --git a/src/gallium/auxiliary/util/u_handle_table.c b/src/gallium/auxiliary/util/u_handle_table.c
index 0bfb9e1b4a..2176a00959 100644
--- a/src/gallium/auxiliary/util/u_handle_table.c
+++ b/src/gallium/auxiliary/util/u_handle_table.c
@@ -170,9 +170,8 @@ handle_table_set(struct handle_table *ht,
    unsigned index;
    
    assert(ht);
-   assert(handle > 0);
-   assert(handle <= ht->size);
-   if(!handle || handle > ht->size)
+   assert(handle);
+   if(!handle)
       return 0;
 
    assert(object);
@@ -185,7 +184,9 @@ handle_table_set(struct handle_table *ht,
    if(!handle_table_resize(ht, index))
       return 0;
 
-   assert(!ht->objects[index]);
+   if(ht->objects[index] && ht->destroy)
+      ht->destroy(ht->objects[index]);
+      
    ht->objects[index] = object;
    
    return handle;
@@ -199,13 +200,11 @@ handle_table_get(struct handle_table *ht,
    void *object;
    
    assert(ht);
-   assert(handle > 0);
-   assert(handle <= ht->size);
+   assert(handle);
    if(!handle || handle > ht->size)
       return NULL;
 
    object = ht->objects[handle - 1];
-   assert(object);
    
    return object;
 }
@@ -219,18 +218,14 @@ handle_table_remove(struct handle_table *ht,
    unsigned index;
    
    assert(ht);
-   assert(handle > 0);
-   assert(handle <= ht->size);
+   assert(handle);
    if(!handle || handle > ht->size)
       return;
 
    index = handle - 1;
    object = ht->objects[index];
-   if(!object) {
-      /* XXX: this warning may be noisy for legitimate use -- remove later */
-      debug_warning("removing empty handle");
+   if(!object)
       return;
-   }
    
    if(ht->destroy)
       ht->destroy(object);
@@ -248,7 +243,7 @@ handle_table_get_next_handle(struct handle_table *ht,
    unsigned index;
    
    for(index = handle; index < ht->size; ++index) {
-      if(!ht->objects[index])
+      if(ht->objects[index])
 	 return index + 1;
    }
 
diff --git a/src/gallium/auxiliary/util/u_simple_shaders.c b/src/gallium/auxiliary/util/u_simple_shaders.c
index 88e2ab05bd..d230ddfbeb 100644
--- a/src/gallium/auxiliary/util/u_simple_shaders.c
+++ b/src/gallium/auxiliary/util/u_simple_shaders.c
@@ -68,7 +68,7 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe,
    uint ti, i;
    struct pipe_shader_state shader;
 
-   tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0]));
+   tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0]));
 
    /* shader header
     */
@@ -84,16 +84,15 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe,
 
    /* declare inputs */
    for (i = 0; i < num_attribs; i++) {
-
       decl = tgsi_default_full_declaration();
       decl.Declaration.File = TGSI_FILE_INPUT;
-      /*
+
       decl.Declaration.Semantic = 1;
-      decl.Semantic.SemanticName = TGSI_SEMANTIC_POSITION;
-      decl.Semantic.SemanticIndex = 0;
-      */
+      decl.Semantic.SemanticName = semantic_names[i];
+      decl.Semantic.SemanticIndex = semantic_indexes[i];
+
       decl.u.DeclarationRange.First = 
-      decl.u.DeclarationRange.Last = 0;
+      decl.u.DeclarationRange.Last = i;
       ti += tgsi_build_full_declaration(&decl,
                                         &tokens[ti],
                                         header,
@@ -102,19 +101,17 @@ util_make_vertex_passthrough_shader(struct pipe_context *pipe,
 
    /* declare outputs */
    for (i = 0; i < num_attribs; i++) {
-
       decl = tgsi_default_full_declaration();
       decl.Declaration.File = TGSI_FILE_OUTPUT;
       decl.Declaration.Semantic = 1;
       decl.Semantic.SemanticName = semantic_names[i];
       decl.Semantic.SemanticIndex = semantic_indexes[i];
       decl.u.DeclarationRange.First = 
-         decl.u.DeclarationRange.Last = 0;
+         decl.u.DeclarationRange.Last = i;
       ti += tgsi_build_full_declaration(&decl,
                                         &tokens[ti],
                                         header,
                                         maxTokens - ti);
-
    }
 
    /* emit MOV instructions */
@@ -173,7 +170,7 @@ util_make_fragment_tex_shader(struct pipe_context *pipe)
    uint ti;
    struct pipe_shader_state shader;
 
-   tokens = (struct tgsi_token *) malloc(maxTokens * sizeof(tokens[0]));
+   tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0]));
 
    /* shader header
     */
@@ -261,3 +258,98 @@ util_make_fragment_tex_shader(struct pipe_context *pipe)
    return pipe->create_fs_state(pipe, &shader);
 }
 
+
+
+
+
+/**
+ * Make simple fragment color pass-through shader.
+ */
+void *
+util_make_fragment_passthrough_shader(struct pipe_context *pipe)
+{
+   uint maxTokens = 40;
+   struct tgsi_token *tokens;
+   struct tgsi_header *header;
+   struct tgsi_processor *processor;
+   struct tgsi_full_declaration decl;
+   struct tgsi_full_instruction inst;
+   const uint procType = TGSI_PROCESSOR_FRAGMENT;
+   uint ti;
+   struct pipe_shader_state shader;
+
+   tokens = (struct tgsi_token *) MALLOC(maxTokens * sizeof(tokens[0]));
+
+   /* shader header
+    */
+   *(struct tgsi_version *) &tokens[0] = tgsi_build_version();
+
+   header = (struct tgsi_header *) &tokens[1];
+   *header = tgsi_build_header();
+
+   processor = (struct tgsi_processor *) &tokens[2];
+   *processor = tgsi_build_processor( procType, header );
+
+   ti = 3;
+
+   /* declare input */
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_INPUT;
+   decl.Declaration.Semantic = 1;
+   decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
+   decl.Semantic.SemanticIndex = 0;
+   decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = 0;
+   ti += tgsi_build_full_declaration(&decl,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti);
+
+   /* declare output */
+   decl = tgsi_default_full_declaration();
+   decl.Declaration.File = TGSI_FILE_OUTPUT;
+   decl.Declaration.Semantic = 1;
+   decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
+   decl.Semantic.SemanticIndex = 0;
+   decl.u.DeclarationRange.First = 
+      decl.u.DeclarationRange.Last = 0;
+   ti += tgsi_build_full_declaration(&decl,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti);
+
+
+   /* MOVE out[0], in[0]; */
+   inst = tgsi_default_full_instruction();
+   inst.Instruction.Opcode = TGSI_OPCODE_MOV;
+   inst.Instruction.NumDstRegs = 1;
+   inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
+   inst.FullDstRegisters[0].DstRegister.Index = 0;
+   inst.Instruction.NumSrcRegs = 1;
+   inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
+   inst.FullSrcRegisters[0].SrcRegister.Index = 0;
+   ti += tgsi_build_full_instruction(&inst,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti );
+
+   /* END instruction */
+   inst = tgsi_default_full_instruction();
+   inst.Instruction.Opcode = TGSI_OPCODE_END;
+   inst.Instruction.NumDstRegs = 0;
+   inst.Instruction.NumSrcRegs = 0;
+   ti += tgsi_build_full_instruction(&inst,
+                                     &tokens[ti],
+                                     header,
+                                     maxTokens - ti );
+
+   assert(ti < maxTokens);
+
+#if 0 /*debug*/
+   tgsi_dump(tokens, 0);
+#endif
+
+   shader.tokens = tokens;
+   return pipe->create_fs_state(pipe, &shader);
+}
+
diff --git a/src/gallium/auxiliary/util/u_simple_shaders.h b/src/gallium/auxiliary/util/u_simple_shaders.h
index 3ef4f28801..ca219a092c 100644
--- a/src/gallium/auxiliary/util/u_simple_shaders.h
+++ b/src/gallium/auxiliary/util/u_simple_shaders.h
@@ -36,6 +36,11 @@
 struct pipe_context;
 
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
 extern void *
 util_make_vertex_passthrough_shader(struct pipe_context *pipe,
                                     uint num_attribs,
@@ -47,6 +52,13 @@ extern void *
 util_make_fragment_tex_shader(struct pipe_context *pipe);
 
 
+extern void *
+util_make_fragment_passthrough_shader(struct pipe_context *pipe);
+
+
+#ifdef __cplusplus
+}
 #endif
 
 
+#endif
diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h
index fe93fd8e1a..d59e4f7036 100644
--- a/src/gallium/drivers/cell/common.h
+++ b/src/gallium/drivers/cell/common.h
@@ -115,6 +115,16 @@ struct cell_command_depth_stencil_alpha_test {
 
 
 /**
+ * Upload code to perform framebuffer blend operation
+ */
+struct cell_command_blend {
+   uint64_t base;               /**< Effective address of code start. */
+   unsigned size;               /**< Size in bytes of test code. */
+   unsigned read_fb;            /**< Flag: should framebuffer be read? */
+};
+
+
+/**
  * Tell SPUs about the framebuffer size, location
  */
 struct cell_command_framebuffer
diff --git a/src/gallium/drivers/cell/ppu/cell_pipe_state.c b/src/gallium/drivers/cell/ppu/cell_pipe_state.c
index c880760e4b..00f4be7401 100644
--- a/src/gallium/drivers/cell/ppu/cell_pipe_state.c
+++ b/src/gallium/drivers/cell/ppu/cell_pipe_state.c
@@ -1,5 +1,5 @@
 /**************************************************************************
- * 
+ *
  * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
  * All Rights Reserved.
  *
@@ -10,11 +10,11 @@
  * 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.
@@ -22,7 +22,7 @@
  * 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.
- * 
+ *
  **************************************************************************/
 
 /* Authors:
@@ -47,21 +47,20 @@ cell_create_blend_state(struct pipe_context *pipe,
    struct cell_blend_state *cb = MALLOC(sizeof(struct cell_blend_state));
 
    (void) memcpy(cb, blend, sizeof(*blend));
-   cb->code.store = NULL;
+   cell_generate_alpha_blend(cb);
 
    return cb;
 }
 
 
 static void
-cell_bind_blend_state(struct pipe_context *pipe, void *blend)
+cell_bind_blend_state(struct pipe_context *pipe, void *state)
 {
    struct cell_context *cell = cell_context(pipe);
 
    draw_flush(cell->draw);
 
-   cell->blend = (const struct cell_blend_state *)blend;
-
+   cell->blend = (struct cell_blend_state *) state;
    cell->dirty |= CELL_NEW_BLEND;
 }
 
@@ -70,7 +69,7 @@ static void
 cell_delete_blend_state(struct pipe_context *pipe, void *blend)
 {
    struct cell_blend_state *cb = (struct cell_blend_state *) blend;
-   
+
    spe_release_func(& cb->code);
    FREE(cb);
 }
@@ -100,7 +99,7 @@ cell_create_depth_stencil_alpha_state(struct pipe_context *pipe,
        MALLOC(sizeof(struct cell_depth_stencil_alpha_state));
 
    (void) memcpy(cdsa, depth_stencil, sizeof(*depth_stencil));
-   cdsa->code.store = NULL;
+   cell_generate_depth_stencil_test(cdsa);
 
    return cdsa;
 }
@@ -111,16 +110,11 @@ cell_bind_depth_stencil_alpha_state(struct pipe_context *pipe,
                                     void *depth_stencil)
 {
    struct cell_context *cell = cell_context(pipe);
-   struct cell_depth_stencil_alpha_state *cdsa =
-       (struct cell_depth_stencil_alpha_state *) depth_stencil;
 
    draw_flush(cell->draw);
 
-   if ((cdsa != NULL) && (cdsa->code.store == NULL)) {
-      cell_generate_depth_stencil_test(cdsa);
-   }
-
-   cell->depth_stencil = cdsa;
+   cell->depth_stencil =
+       (struct cell_depth_stencil_alpha_state *) depth_stencil;
    cell->dirty |= CELL_NEW_DEPTH_STENCIL;
 }
 
@@ -357,4 +351,3 @@ cell_init_state_functions(struct cell_context *cell)
    cell->pipe.set_scissor_state = cell_set_scissor_state;
    cell->pipe.set_viewport_state = cell_set_viewport_state;
 }
-
diff --git a/src/gallium/drivers/cell/ppu/cell_state_derived.c b/src/gallium/drivers/cell/ppu/cell_state_derived.c
index 5c240a55c0..5480534ad9 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_derived.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_derived.c
@@ -141,17 +141,8 @@ calculate_vertex_layout( struct cell_context *cell )
 static void
 compute_cliprect(struct cell_context *sp)
 {
-   unsigned surfWidth, surfHeight;
-
-   if (sp->framebuffer.num_cbufs > 0) {
-      surfWidth = sp->framebuffer.cbufs[0]->width;
-      surfHeight = sp->framebuffer.cbufs[0]->height;
-   }
-   else {
-      /* no surface? */
-      surfWidth = sp->scissor.maxx;
-      surfHeight = sp->scissor.maxy;
-   }
+   uint surfWidth = sp->framebuffer.width;
+   uint surfHeight = sp->framebuffer.height;
 
    if (sp->rasterizer->scissor) {
       /* clip to scissor rect */
diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c
index 4d589bcdbf..5709b48f12 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_emit.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c
@@ -60,14 +60,25 @@ cell_emit_state(struct cell_context *cell)
       fb->color_format = cbuf->format;
       fb->depth_start = cell->zsbuf_map;
       fb->depth_format = zbuf ? zbuf->format : PIPE_FORMAT_NONE;
-      fb->width = cell->framebuffer.cbufs[0]->width;
-      fb->height = cell->framebuffer.cbufs[0]->height;
+      fb->width = cell->framebuffer.width;
+      fb->height = cell->framebuffer.height;
    }
 
    if (cell->dirty & CELL_NEW_BLEND) {
-      emit_state_cmd(cell, CELL_CMD_STATE_BLEND,
-                     cell->blend,
-                     sizeof(struct pipe_blend_state));
+      struct cell_command_blend blend;
+
+      if (cell->blend != NULL) {
+         blend.base = (intptr_t) cell->blend->code.store;
+         blend.size = (char *) cell->blend->code.csr
+             - (char *) cell->blend->code.store;
+         blend.read_fb = TRUE;
+      } else {
+         blend.base = 0;
+         blend.size = 0;
+         blend.read_fb = FALSE;
+      }
+
+      emit_state_cmd(cell, CELL_CMD_STATE_BLEND, &blend, sizeof(blend));
    }
 
    if (cell->dirty & CELL_NEW_DEPTH_STENCIL) {
diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c
index 9c47968459..c750b1d89d 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c
@@ -72,8 +72,8 @@ emit_alpha_test(struct pipe_depth_stencil_alpha_state *dsa,
       int tmp_a = spe_allocate_available_register(f);
       int tmp_b = spe_allocate_available_register(f);
       union {
-	 float f;
-	 unsigned u;
+         float f;
+         unsigned u;
       } ref_val;
       boolean complement = FALSE;
 
@@ -84,42 +84,42 @@ emit_alpha_test(struct pipe_depth_stencil_alpha_state *dsa,
 
       switch (dsa->alpha.func) {
       case PIPE_FUNC_NOTEQUAL:
-	 complement = TRUE;
-	 /* FALLTHROUGH */
+         complement = TRUE;
+         /* FALLTHROUGH */
 
       case PIPE_FUNC_EQUAL:
-	 spe_fceq(f, tmp_a, ref, alphas);
-	 break;
+         spe_fceq(f, tmp_a, ref, alphas);
+         break;
 
       case PIPE_FUNC_LEQUAL:
-	 complement = TRUE;
-	 /* FALLTHROUGH */
+         complement = TRUE;
+         /* FALLTHROUGH */
 
       case PIPE_FUNC_GREATER:
-	 spe_fcgt(f, tmp_a, ref, alphas);
-	 break;
+         spe_fcgt(f, tmp_a, ref, alphas);
+         break;
 
       case PIPE_FUNC_LESS:
-	 complement = TRUE;
-	 /* FALLTHROUGH */
+         complement = TRUE;
+         /* FALLTHROUGH */
 
       case PIPE_FUNC_GEQUAL:
-	 spe_fcgt(f, tmp_a, ref, alphas);
-	 spe_fceq(f, tmp_b, ref, alphas);
-	 spe_or(f, tmp_a, tmp_b, tmp_a);
-	 break;
+         spe_fcgt(f, tmp_a, ref, alphas);
+         spe_fceq(f, tmp_b, ref, alphas);
+         spe_or(f, tmp_a, tmp_b, tmp_a);
+         break;
 
       case PIPE_FUNC_ALWAYS:
       case PIPE_FUNC_NEVER:
       default:
-	 assert(0);
-	 break;
+         assert(0);
+         break;
       }
 
       if (complement) {
-	 spe_andc(f, mask, mask, tmp_a);
+         spe_andc(f, mask, mask, tmp_a);
       } else {
-	 spe_and(f, mask, mask, tmp_a);
+         spe_and(f, mask, mask, tmp_a);
       }
 
       spe_release_register(f, ref);
@@ -588,19 +588,13 @@ cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa)
  */
 static int
 emit_alpha_factor_calculation(struct spe_function *f,
-                              unsigned factor, float const_alpha,
-                              int src_alpha, int dst_alpha)
+                              unsigned factor,
+                              int src_alpha, int dst_alpha, int const_alpha)
 {
-   union {
-      float f;
-      unsigned u;
-   } alpha;
    int factor_reg;
    int tmp;
 
 
-   alpha.f = const_alpha;
-
    switch (factor) {
    case PIPE_BLENDFACTOR_ONE:
       factor_reg = -1;
@@ -621,13 +615,17 @@ emit_alpha_factor_calculation(struct spe_function *f,
       break;
 
    case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
-      const_alpha = 1.0 - const_alpha;
-      /* FALLTHROUGH */
-   case PIPE_BLENDFACTOR_CONST_ALPHA:
       factor_reg = spe_allocate_available_register(f);
 
-      spe_il(f, factor_reg, alpha.u & 0x0ffff);
-      spe_ilh(f, factor_reg, alpha.u >> 16);
+      tmp = spe_allocate_available_register(f);
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+      spe_fs(f, factor_reg, tmp, const_alpha);
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      factor_reg = const_alpha;
       break;
 
    case PIPE_BLENDFACTOR_ZERO:
@@ -669,29 +667,20 @@ emit_alpha_factor_calculation(struct spe_function *f,
 
 
 /**
- * \note Emits a maximum of 5 instructions
+ * \note Emits a maximum of 6 instructions
  */
 static void
 emit_color_factor_calculation(struct spe_function *f,
                               unsigned sF, unsigned mask,
-			      const struct pipe_blend_color *blend_color,
                               const int *src,
                               const int *dst,
+                              const int *const_color,
                               int *factor)
 {
-   union {
-      float f[4];
-      unsigned u[4];
-   } color;
    int tmp;
    unsigned i;
 
 
-   color.f[0] = blend_color->color[0];
-   color.f[1] = blend_color->color[1];
-   color.f[2] = blend_color->color[2];
-   color.f[3] = blend_color->color[3];
-
    factor[0] = -1;
    factor[1] = -1;
    factor[2] = -1;
@@ -748,29 +737,40 @@ emit_color_factor_calculation(struct spe_function *f,
       break;
 
    case PIPE_BLENDFACTOR_INV_CONST_COLOR:
-      color.f[0] = 1.0 - color.f[0];
-      color.f[1] = 1.0 - color.f[1];
-      color.f[2] = 1.0 - color.f[2];
-      /* FALLTHROUGH */
-   case PIPE_BLENDFACTOR_CONST_COLOR:
+      tmp = spe_allocate_available_register(f);
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+
       for (i = 0; i < 3; i++) {
-	 factor[i] = spe_allocate_available_register(f);
+         factor[i] = spe_allocate_available_register(f);
+
+         spe_fs(f, factor[i], tmp, const_color[i]);
+      }
+      spe_release_register(f, tmp);
+      break;
 
-	 spe_il(f, factor[i], color.u[i] & 0x0ffff);
-	 spe_ilh(f, factor[i], color.u[i] >> 16);
+   case PIPE_BLENDFACTOR_CONST_COLOR:
+      for (i = 0; i < 3; i++) {
+         factor[i] = const_color[i];
       }
       break;
 
    case PIPE_BLENDFACTOR_INV_CONST_ALPHA:
-      color.f[3] = 1.0 - color.f[3];
-      /* FALLTHROUGH */
-   case PIPE_BLENDFACTOR_CONST_ALPHA:
       factor[0] = spe_allocate_available_register(f);
       factor[1] = factor[0];
       factor[2] = factor[0];
 
-      spe_il(f, factor[0], color.u[3] & 0x0ffff);
-      spe_ilh(f, factor[0], color.u[3] >> 16);
+      tmp = spe_allocate_available_register(f);
+      spe_il(f, tmp, 1);
+      spe_cuflt(f, tmp, tmp, 0);
+      spe_fs(f, factor[0], tmp, const_color[3]);
+      spe_release_register(f, tmp);
+      break;
+
+   case PIPE_BLENDFACTOR_CONST_ALPHA:
+      factor[0] = const_color[3];
+      factor[1] = factor[0];
+      factor[2] = factor[0];
       break;
 
    case PIPE_BLENDFACTOR_ZERO:
@@ -864,7 +864,11 @@ emit_blend_calculation(struct spe_function *f,
             spe_il(f, src, 0);
          } else if (dF == PIPE_BLENDFACTOR_ONE) {
             spe_or(f, src, dst, dst);
+         } else {
+            spe_fm(f, src, dst, dst_factor);
          }
+      } else if (dF == PIPE_BLENDFACTOR_ZERO) {
+         spe_fm(f, src, src, src_factor);
       } else {
          spe_fm(f, tmp, dst, dst_factor);
          spe_fma(f, src, src, src_factor, tmp);
@@ -884,7 +888,11 @@ emit_blend_calculation(struct spe_function *f,
          } else if (dF == PIPE_BLENDFACTOR_ONE) {
             spe_il(f, tmp, 0);
             spe_fs(f, src, tmp, dst);
+         } else {
+            spe_fm(f, src, dst, dst_factor);
          }
+      } else if (dF == PIPE_BLENDFACTOR_ZERO) {
+         spe_fm(f, src, src, src_factor);
       } else {
          spe_fm(f, tmp, dst, dst_factor);
          spe_fms(f, src, src, src_factor, tmp);
@@ -904,7 +912,11 @@ emit_blend_calculation(struct spe_function *f,
             spe_il(f, src, 0);
          } else if (dF == PIPE_BLENDFACTOR_ONE) {
             spe_or(f, src, dst, dst);
+         } else {
+            spe_fm(f, src, dst, dst_factor);
          }
+      } else if (dF == PIPE_BLENDFACTOR_ZERO) {
+         spe_fm(f, src, src, src_factor);
       } else {
          spe_fm(f, tmp, src, src_factor);
          spe_fms(f, src, src, dst_factor, tmp);
@@ -913,12 +925,12 @@ emit_blend_calculation(struct spe_function *f,
 
    case PIPE_BLEND_MIN:
       spe_cgt(f, tmp, src, dst);
-      spe_selb(f, src, dst, src, tmp);
+      spe_selb(f, src, src, dst, tmp);
       break;
 
    case PIPE_BLEND_MAX:
       spe_cgt(f, tmp, src, dst);
-      spe_selb(f, src, src, dst, tmp);
+      spe_selb(f, src, dst, src, tmp);
       break;
 
    default:
@@ -933,16 +945,15 @@ emit_blend_calculation(struct spe_function *f,
  * Generate code to perform alpha blending on the SPE
  */
 void
-cell_generate_alpha_blend(struct cell_blend_state *cb,
-			  const struct pipe_blend_color *blend_color)
+cell_generate_alpha_blend(struct cell_blend_state *cb)
 {
    struct pipe_blend_state *const b = &cb->base;
    struct spe_function *const f = &cb->code;
 
    /* This code generates a maximum of 3 (source alpha factor)
-    * + 3 (destination alpha factor) + (3 * 5) (source color factor)
-    * + (3 * 5) (destination color factor) + (4 * 2) (blend equation)
-    * + 4 (fragment mask) + 1 (return) = 49 instlructions.  Round up to 64 to
+    * + 3 (destination alpha factor) + (3 * 6) (source color factor)
+    * + (3 * 6) (destination color factor) + (4 * 2) (blend equation)
+    * + 4 (fragment mask) + 1 (return) = 55 instlructions.  Round up to 64 to
     * make it a happy power-of-two.
     */
    spe_init_func(f, 4 * 64);
@@ -960,7 +971,13 @@ cell_generate_alpha_blend(struct cell_blend_state *cb,
       spe_allocate_register(f, 9),
       spe_allocate_register(f, 10),
    };
-   const int mask = spe_allocate_register(f, 11);
+   const int const_color[4] = {
+      spe_allocate_register(f, 11),
+      spe_allocate_register(f, 12),
+      spe_allocate_register(f, 13),
+      spe_allocate_register(f, 14),
+   };
+   const int mask = spe_allocate_register(f, 15);
    unsigned func[4];
    unsigned sF[4];
    unsigned dF[4];
@@ -984,24 +1001,64 @@ cell_generate_alpha_blend(struct cell_blend_state *cb,
        && (b->alpha_func != PIPE_BLEND_MAX);
 
 
-   sF[0] = b->rgb_src_factor;
-   sF[1] = sF[0];
-   sF[2] = sF[0];
-   sF[3] = (b->alpha_src_factor == PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE)
-       ? PIPE_BLENDFACTOR_ONE : b->alpha_src_factor;
+   if (b->blend_enable) {
+      sF[0] = b->rgb_src_factor;
+      sF[1] = sF[0];
+      sF[2] = sF[0];
+      switch (b->alpha_src_factor & 0x0f) {
+      case PIPE_BLENDFACTOR_SRC_ALPHA_SATURATE:
+         sF[3] = PIPE_BLENDFACTOR_ONE;
+         break;
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+      case PIPE_BLENDFACTOR_DST_COLOR:
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+      case PIPE_BLENDFACTOR_SRC1_COLOR:
+         sF[3] = b->alpha_src_factor + 1;
+         break;
+      default:
+         sF[3] = b->alpha_src_factor;
+      }
 
-   dF[0] = b->rgb_dst_factor;
-   dF[1] = dF[0];
-   dF[2] = dF[0];
-   dF[3] = b->rgb_dst_factor;
+      dF[0] = b->rgb_dst_factor;
+      dF[1] = dF[0];
+      dF[2] = dF[0];
+      switch (b->alpha_dst_factor & 0x0f) {
+      case PIPE_BLENDFACTOR_SRC_COLOR:
+      case PIPE_BLENDFACTOR_DST_COLOR:
+      case PIPE_BLENDFACTOR_CONST_COLOR:
+      case PIPE_BLENDFACTOR_SRC1_COLOR:
+         dF[3] = b->alpha_dst_factor + 1;
+         break;
+      default:
+         dF[3] = b->alpha_dst_factor;
+      }
+
+      func[0] = b->rgb_func;
+      func[1] = func[0];
+      func[2] = func[0];
+      func[3] = b->alpha_func;
+   } else {
+      sF[0] = PIPE_BLENDFACTOR_ONE;
+      sF[1] = PIPE_BLENDFACTOR_ONE;
+      sF[2] = PIPE_BLENDFACTOR_ONE;
+      sF[3] = PIPE_BLENDFACTOR_ONE;
+      dF[0] = PIPE_BLENDFACTOR_ZERO;
+      dF[1] = PIPE_BLENDFACTOR_ZERO;
+      dF[2] = PIPE_BLENDFACTOR_ZERO;
+      dF[3] = PIPE_BLENDFACTOR_ZERO;
+
+      func[0] = PIPE_BLEND_ADD;
+      func[1] = PIPE_BLEND_ADD;
+      func[2] = PIPE_BLEND_ADD;
+      func[3] = PIPE_BLEND_ADD;
+   }
 
 
    /* If alpha writing is enabled and the alpha blend mode requires use of
     * the alpha factor, calculate the alpha factor.
     */
    if (((b->colormask & 8) != 0) && need_alpha_factor) {
-      src_factor[3] = emit_alpha_factor_calculation(f, sF[3],
-						    blend_color->color[3],
+      src_factor[3] = emit_alpha_factor_calculation(f, sF[3], const_color[3],
                                                     frag[3], pixel[3]);
 
       /* If the alpha destination blend factor is the same as the alpha source
@@ -1009,8 +1066,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb,
        */
       dst_factor[3] = (dF[3] == sF[3])
           ? src_factor[3]
-          : emit_alpha_factor_calculation(f, dF[3],
-					  blend_color->color[3],
+          : emit_alpha_factor_calculation(f, dF[3], const_color[3],
                                           frag[3], pixel[3]);
    }
 
@@ -1027,8 +1083,7 @@ cell_generate_alpha_blend(struct cell_blend_state *cb,
       emit_color_factor_calculation(f,
                                     b->rgb_src_factor,
                                     b->colormask,
-                                    blend_color,
-                                    frag, pixel, src_factor);
+                                    frag, pixel, const_color, src_factor);
    }
 
 
@@ -1048,17 +1103,11 @@ cell_generate_alpha_blend(struct cell_blend_state *cb,
       emit_color_factor_calculation(f,
                                     b->rgb_dst_factor,
                                     b->colormask,
-                                    blend_color,
-                                    frag, pixel, dst_factor);
+                                    frag, pixel, const_color, dst_factor);
    }
 
 
 
-   func[0] = b->rgb_func;
-   func[1] = func[0];
-   func[2] = func[0];
-   func[3] = b->alpha_func;
-
    for (i = 0; i < 4; ++i) {
       if ((b->colormask & (1U << i)) != 0) {
          emit_blend_calculation(f,
@@ -1072,4 +1121,28 @@ cell_generate_alpha_blend(struct cell_blend_state *cb,
    }
 
    spe_bi(f, 0, 0, 0);
+
+#if 0
+   {
+      const uint32_t *p = f->store;
+
+      printf("# %u instructions\n", f->csr - f->store);
+      printf("# blend (%sabled)\n",
+             (cb->base.blend_enable) ? "en" : "dis");
+      printf("#    RGB func / sf / df: %u %u %u\n",
+             cb->base.rgb_func,
+             cb->base.rgb_src_factor,
+             cb->base.rgb_dst_factor);
+      printf("#    ALP func / sf / df: %u %u %u\n",
+             cb->base.alpha_func,
+             cb->base.alpha_src_factor,
+             cb->base.alpha_dst_factor);
+
+      printf("\t.text\n");
+      for (/* empty */; p < f->csr; p++) {
+         printf("\t.long\t0x%04x\n", *p);
+      }
+      fflush(stdout);
+   }
+#endif
 }
diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h
index 541c3b3be0..f699247f9e 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h
+++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h
@@ -29,7 +29,6 @@ extern void
 cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa);
 
 extern void
-cell_generate_alpha_blend(struct cell_blend_state *cb,
-			  const struct pipe_blend_color *blend_color);
+cell_generate_alpha_blend(struct cell_blend_state *cb);
 
 #endif /* CELL_STATE_PER_FRAGMENT_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_texture.c b/src/gallium/drivers/cell/ppu/cell_texture.c
index e235421107..9c694e136d 100644
--- a/src/gallium/drivers/cell/ppu/cell_texture.c
+++ b/src/gallium/drivers/cell/ppu/cell_texture.c
@@ -134,7 +134,8 @@ cell_texture_release_screen(struct pipe_screen *screen,
 
 
 static void
-cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture)
+cell_texture_update(struct pipe_context *pipe, struct pipe_texture *texture,
+                    uint face, uint levelsMask)
 {
    /* XXX TO DO:  re-tile the texture data ... */
 
diff --git a/src/gallium/drivers/cell/spu/Makefile b/src/gallium/drivers/cell/spu/Makefile
index 115ca8cd90..8e83610790 100644
--- a/src/gallium/drivers/cell/spu/Makefile
+++ b/src/gallium/drivers/cell/spu/Makefile
@@ -17,7 +17,6 @@ PROG_SPU_EMBED_O = $(PROG)_spu-embed.o
 
 SOURCES = \
 	spu_main.c \
-	spu_blend.c \
 	spu_dcache.c \
 	spu_per_fragment_op.c \
 	spu_render.c \
diff --git a/src/gallium/drivers/cell/spu/spu_blend.c b/src/gallium/drivers/cell/spu/spu_blend.c
deleted file mode 100644
index 23ec0eeb45..0000000000
--- a/src/gallium/drivers/cell/spu/spu_blend.c
+++ /dev/null
@@ -1,62 +0,0 @@
-/**************************************************************************
- * 
- * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
- * 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 TUNGSTEN GRAPHICS 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.
- * 
- **************************************************************************/
-
-
-#include "spu_main.h"
-#include "spu_blend.h"
-#include "spu_colorpack.h"
-
-
-void
-blend_quad(uint itx, uint ity, vector float colors[4])
-{
-   /* simple SRC_ALPHA, ONE_MINUS_SRC_ALPHA blending */
-   vector float fbc00 = spu_unpack_color(spu.ctile.ui[ity][itx]);
-   vector float fbc01 = spu_unpack_color(spu.ctile.ui[ity][itx+1]);
-   vector float fbc10 = spu_unpack_color(spu.ctile.ui[ity+1][itx]);
-   vector float fbc11 = spu_unpack_color(spu.ctile.ui[ity+1][itx+1]);
-
-   vector float alpha00 = spu_splats(spu_extract(colors[0], 3));
-   vector float alpha01 = spu_splats(spu_extract(colors[1], 3));
-   vector float alpha10 = spu_splats(spu_extract(colors[2], 3));
-   vector float alpha11 = spu_splats(spu_extract(colors[3], 3));
-
-   vector float one_minus_alpha00 = spu_sub(spu_splats(1.0f), alpha00);
-   vector float one_minus_alpha01 = spu_sub(spu_splats(1.0f), alpha01);
-   vector float one_minus_alpha10 = spu_sub(spu_splats(1.0f), alpha10);
-   vector float one_minus_alpha11 = spu_sub(spu_splats(1.0f), alpha11);
-
-   colors[0] = spu_add(spu_mul(colors[0], alpha00),
-                       spu_mul(fbc00, one_minus_alpha00));
-   colors[1] = spu_add(spu_mul(colors[1], alpha01),
-                       spu_mul(fbc01, one_minus_alpha01));
-   colors[2] = spu_add(spu_mul(colors[2], alpha10),
-                       spu_mul(fbc10, one_minus_alpha10));
-   colors[3] = spu_add(spu_mul(colors[3], alpha11),
-                       spu_mul(fbc11, one_minus_alpha11));
-}
-
diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c
index 937962285d..0a490ab277 100644
--- a/src/gallium/drivers/cell/spu/spu_main.c
+++ b/src/gallium/drivers/cell/spu/spu_main.c
@@ -61,6 +61,10 @@ static unsigned char attribute_fetch_code_buffer[136 * PIPE_ATTRIB_MAX]
 static unsigned char depth_stencil_code_buffer[4 * 64]
     ALIGN16_ATTRIB;
 
+static unsigned char fb_blend_code_buffer[4 * 64]
+    ALIGN16_ATTRIB;
+
+
 /**
  * Tell the PPU that this SPU has finished copying a buffer to
  * local store and that it may be reused by the PPU.
@@ -246,14 +250,28 @@ cmd_state_framebuffer(const struct cell_command_framebuffer *cmd)
 
 
 static void
-cmd_state_blend(const struct pipe_blend_state *state)
+cmd_state_blend(const struct cell_command_blend *state)
 {
    if (Debug)
       printf("SPU %u: BLEND: enabled %d\n",
              spu.init.id,
-             state->blend_enable);
+             (state->size != 0));
 
-   memcpy(&spu.blend, state, sizeof(*state));
+   ASSERT_ALIGN16(state->base);
+
+   if (state->size != 0) {
+      mfc_get(fb_blend_code_buffer,
+              (unsigned int) state->base,  /* src */
+              ROUNDUP16(state->size),
+              TAG_BATCH_BUFFER,
+              0, /* tid */
+              0  /* rid */);
+      wait_on_mask(1 << TAG_BATCH_BUFFER);
+      spu.blend = (blend_func) fb_blend_code_buffer;
+      spu.read_fb = state->read_fb;
+   } else {
+      spu.read_fb = FALSE;
+   }
 }
 
 
@@ -441,9 +459,8 @@ cmd_batch(uint opcode)
          pos += 1;
          break;
       case CELL_CMD_STATE_BLEND:
-         cmd_state_blend((struct pipe_blend_state *)
-                                 &buffer[pos+1]);
-         pos += (1 + ROUNDUP8(sizeof(struct pipe_blend_state)) / 8);
+         cmd_state_blend((struct cell_command_blend *) &buffer[pos+1]);
+         pos += (1 + ROUNDUP8(sizeof(struct cell_command_blend)) / 8);
          break;
       case CELL_CMD_STATE_DEPTH_STENCIL:
          cmd_state_depth_stencil((struct cell_command_depth_stencil_alpha_test *)
diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h
index 444e218645..49f5d99674 100644
--- a/src/gallium/drivers/cell/spu/spu_main.h
+++ b/src/gallium/drivers/cell/spu/spu_main.h
@@ -67,6 +67,19 @@ typedef struct spu_frag_test_results (*frag_test_func)(qword frag_mask,
     qword frag_alpha, qword facing);
 
 
+struct spu_blend_results {
+   qword r;
+   qword g;
+   qword b;
+   qword a;
+};
+
+typedef struct spu_blend_results (*blend_func)(
+    qword frag_r, qword frag_g, qword frag_b, qword frag_a,
+    qword pixel_r, qword pixel_g, qword pixel_b, qword pixel_a,
+    qword const_r, qword const_g, qword const_b, qword const_a,
+    qword frag_mask);
+
 struct spu_framebuffer {
    void *color_start;              /**< addr of color surface in main memory */
    void *depth_start;              /**< addr of depth surface in main memory */
@@ -93,7 +106,11 @@ struct spu_global
    boolean read_depth;
    boolean read_stencil;
    frag_test_func frag_test;
-   struct pipe_blend_state blend;
+   
+   boolean read_fb;
+   blend_func blend;
+   qword const_blend_color[4] ALIGN16_ATTRIB;
+
    struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS];
    struct cell_command_texture texture;
 
diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c
index 81823f2463..e6a1ce01df 100644
--- a/src/gallium/drivers/cell/spu/spu_tri.c
+++ b/src/gallium/drivers/cell/spu/spu_tri.c
@@ -29,10 +29,10 @@
  * Triangle rendering within a tile.
  */
 
+#include <transpose_matrix4x4.h>
 #include "pipe/p_compiler.h"
 #include "pipe/p_format.h"
 #include "pipe/p_util.h"
-#include "spu_blend.h"
 #include "spu_colorpack.h"
 #include "spu_main.h"
 #include "spu_texture.h"
@@ -261,6 +261,9 @@ do_depth_test(int x, int y, mask_t quadmask)
    float4 zvals;
    mask_t mask;
 
+   if (spu.fb.depth_format == PIPE_FORMAT_NONE)
+      return quadmask;
+
    zvals.v = eval_z((float) x, (float) y);
 
    mask = (mask_t) spu_do_depth_stencil(x - setup.cliprect_minx,
@@ -326,27 +329,47 @@ emit_quad( int x, int y, mask_t mask )
          eval_coeff(1, (float) x, (float) y, colors);
       }
 
-#if 1
-      if (spu.blend.blend_enable)
-         blend_quad(ix % TILE_SIZE, iy % TILE_SIZE, colors);
-#endif
 
-      if (spu_extract(mask, 0))
-         spu.ctile.ui[iy][ix] = spu_pack_color_shuffle(colors[0], shuffle);
-      if (spu_extract(mask, 1))
-         spu.ctile.ui[iy][ix+1] = spu_pack_color_shuffle(colors[1], shuffle);
-      if (spu_extract(mask, 2))
-         spu.ctile.ui[iy+1][ix] = spu_pack_color_shuffle(colors[2], shuffle);
-      if (spu_extract(mask, 3))
-         spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(colors[3], shuffle);
+      /* Read the current framebuffer values.
+       *
+       * Ignore read_fb for now.  In the future we can use this to avoid
+       * reading the framebuffer if read_fb is false and the fragment mask is
+       * all 0xffffffff.  This is the common case, so it is probably worth
+       * the effort.  We'll have to profile to determine whether or not the
+       * extra conditional branches hurt overall performance.
+       */
+      vec_float4 aos_pix[4] = {
+         spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+0]),
+         spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+1]),
+         spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+0]),
+         spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+1]),
+      };
 
-#if 0
-      /* SIMD_Z with swizzled color buffer (someday) */
-      vector unsigned int uicolors = *((vector unsigned int *) &colors);
-      spu.ctile.ui4[iy/2][ix/2] = spu_sel(spu.ctile.ui4[iy/2][ix/2], uicolors, mask);
-#endif
-   }
+      qword soa_pix[4];
+      qword soa_frag[4];
 
+      /* Convert pixel and fragment data from AoS to SoA format.
+       */
+      _transpose_matrix4x4((vec_float4 *) soa_pix, aos_pix);
+      _transpose_matrix4x4((vec_float4 *) soa_frag, colors);
+
+      const struct spu_blend_results result =
+          (*spu.blend)(soa_frag[0], soa_frag[1], soa_frag[2], soa_frag[3],
+                       soa_pix[0], soa_pix[1], soa_pix[2], soa_pix[3],
+                       spu.const_blend_color[0], spu.const_blend_color[1],
+                       spu.const_blend_color[2], spu.const_blend_color[3],
+                       (qword) mask);
+
+
+      /* Convert final pixel data from SoA to AoS format.
+       */
+      _transpose_matrix4x4(aos_pix, (const vec_float4 *) &result);
+
+      spu.ctile.ui[iy+0][ix+0] = spu_pack_color_shuffle(aos_pix[0], shuffle);
+      spu.ctile.ui[iy+0][ix+1] = spu_pack_color_shuffle(aos_pix[1], shuffle);
+      spu.ctile.ui[iy+1][ix+0] = spu_pack_color_shuffle(aos_pix[2], shuffle);
+      spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(aos_pix[3], shuffle);
+   }
 #endif
 }
 
diff --git a/src/gallium/drivers/i915simple/i915_state_sampler.c b/src/gallium/drivers/i915simple/i915_state_sampler.c
index 9dbb1b1b23..84f6529a3a 100644
--- a/src/gallium/drivers/i915simple/i915_state_sampler.c
+++ b/src/gallium/drivers/i915simple/i915_state_sampler.c
@@ -206,6 +206,12 @@ i915_update_texture(struct i915_context *i915, uint unit,
        | format
        | MS3_USE_FENCE_REGS);
 
+   /*
+    * XXX sampler->max_lod should be used to program the MAX_LOD field below.
+    * Also, when min_filter != mag_filter and there's just one mipmap level,
+    * set max_lod = 1 to make sure i915 chooses between min/mag filtering.
+    */
+
    /* MS4 state */
    state[1] =
       ((((pitch / 4) - 1) << MS4_PITCH_SHIFT)
diff --git a/src/gallium/drivers/i915simple/i915_texture.c b/src/gallium/drivers/i915simple/i915_texture.c
index ef5adff550..c39e747705 100644
--- a/src/gallium/drivers/i915simple/i915_texture.c
+++ b/src/gallium/drivers/i915simple/i915_texture.c
@@ -542,7 +542,8 @@ i915_texture_release_screen(struct pipe_screen *screen,
 
 
 static void
-i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture)
+i915_texture_update(struct pipe_context *pipe, struct pipe_texture *texture,
+                    uint face, uint levelsMask)
 {
    /* no-op? */
 }
diff --git a/src/gallium/drivers/i965simple/brw_tex_layout.c b/src/gallium/drivers/i965simple/brw_tex_layout.c
index b24ac87c37..b580f98204 100644
--- a/src/gallium/drivers/i965simple/brw_tex_layout.c
+++ b/src/gallium/drivers/i965simple/brw_tex_layout.c
@@ -358,7 +358,8 @@ brw_texture_release_screen(struct pipe_screen *screen,
 
 
 static void
-brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture)
+brw_texture_update(struct pipe_context *pipe, struct pipe_texture *texture,
+                   uint face, uint levelsMask)
 {
    /* no-op? */
 }
diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c
index 7aa9cf8e85..6a81e4d8cc 100644
--- a/src/gallium/drivers/softpipe/sp_prim_setup.c
+++ b/src/gallium/drivers/softpipe/sp_prim_setup.c
@@ -485,7 +485,7 @@ setup_fragcoord_coeff(struct setup_stage *setup, uint slot)
    /*Y*/
    if (setup->softpipe->rasterizer->origin_lower_left) {
       /* y=0=bottom */
-      const int winHeight = setup->softpipe->framebuffer.cbufs[0]->height;
+      const int winHeight = setup->softpipe->framebuffer.height;
       setup->coef[slot].a0[1] = (float) (winHeight - 1);
       setup->coef[slot].dady[1] = -1.0;
    }
diff --git a/src/gallium/drivers/softpipe/sp_quad_stipple.c b/src/gallium/drivers/softpipe/sp_quad_stipple.c
index 8660432259..f1e9b80e09 100644
--- a/src/gallium/drivers/softpipe/sp_quad_stipple.c
+++ b/src/gallium/drivers/softpipe/sp_quad_stipple.c
@@ -25,7 +25,7 @@ stipple_quad(struct quad_stage *qs, struct quad_header *quad)
       int y0, y1;
       uint stipple0, stipple1;
       if (softpipe->rasterizer->origin_lower_left) {
-         y0 = softpipe->framebuffer.cbufs[0]->height - 1 - quad->y0;
+         y0 = softpipe->framebuffer.height - 1 - quad->y0;
          y1 = y0 - 1;
       }
       else {
diff --git a/src/gallium/drivers/softpipe/sp_state_derived.c b/src/gallium/drivers/softpipe/sp_state_derived.c
index 82cb31ece7..14abb20eeb 100644
--- a/src/gallium/drivers/softpipe/sp_state_derived.c
+++ b/src/gallium/drivers/softpipe/sp_state_derived.c
@@ -171,17 +171,8 @@ softpipe_get_vbuf_vertex_info(struct softpipe_context *softpipe)
 static void
 compute_cliprect(struct softpipe_context *sp)
 {
-   unsigned surfWidth, surfHeight;
-
-   if (sp->framebuffer.num_cbufs > 0) {
-      surfWidth = sp->framebuffer.cbufs[0]->width;
-      surfHeight = sp->framebuffer.cbufs[0]->height;
-   }
-   else {
-      /* no surface? */
-      surfWidth = sp->scissor.maxx;
-      surfHeight = sp->scissor.maxy;
-   }
+   uint surfWidth = sp->framebuffer.width;
+   uint surfHeight = sp->framebuffer.height;
 
    if (sp->rasterizer->scissor) {
       /* clip to scissor rect */
diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c
index 124b18b708..ba8c9eece7 100644
--- a/src/gallium/drivers/softpipe/sp_state_surface.c
+++ b/src/gallium/drivers/softpipe/sp_state_surface.c
@@ -101,9 +101,8 @@ softpipe_set_framebuffer_state(struct pipe_context *pipe,
    }
 #endif
 
+   sp->framebuffer.width = fb->width;
+   sp->framebuffer.height = fb->height;
+
    sp->dirty |= SP_NEW_FRAMEBUFFER;
 }
-
-
-
-
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index 64bf353aa4..256586ec88 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -81,8 +81,8 @@ softpipe_texture_layout(struct softpipe_texture * spt)
 
 
 static struct pipe_texture *
-softpipe_texture_create_screen(struct pipe_screen *screen,
-                               const struct pipe_texture *templat)
+softpipe_texture_create(struct pipe_screen *screen,
+                        const struct pipe_texture *templat)
 {
    struct pipe_winsys *ws = screen->winsys;
    struct softpipe_texture *spt = CALLOC_STRUCT(softpipe_texture);
@@ -110,8 +110,8 @@ softpipe_texture_create_screen(struct pipe_screen *screen,
 
 
 static void
-softpipe_texture_release_screen(struct pipe_screen *screen,
-                                struct pipe_texture **pt)
+softpipe_texture_release(struct pipe_screen *screen,
+                         struct pipe_texture **pt)
 {
    if (!*pt)
       return;
@@ -136,9 +136,9 @@ softpipe_texture_release_screen(struct pipe_screen *screen,
 
 
 static struct pipe_surface *
-softpipe_get_tex_surface_screen(struct pipe_screen *screen,
-                                struct pipe_texture *pt,
-                                unsigned face, unsigned level, unsigned zslice)
+softpipe_get_tex_surface(struct pipe_screen *screen,
+                         struct pipe_texture *pt,
+                         unsigned face, unsigned level, unsigned zslice)
 {
    struct pipe_winsys *ws = screen->winsys;
    struct softpipe_texture *spt = softpipe_texture(pt);
@@ -174,7 +174,8 @@ softpipe_get_tex_surface_screen(struct pipe_screen *screen,
 
 static void
 softpipe_texture_update(struct pipe_context *pipe,
-                        struct pipe_texture *texture)
+                        struct pipe_texture *texture,
+                        uint face, uint levelsMask)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
    uint unit;
@@ -196,7 +197,7 @@ softpipe_init_texture_funcs( struct softpipe_context *softpipe )
 void
 softpipe_init_screen_texture_funcs(struct pipe_screen *screen)
 {
-   screen->texture_create = softpipe_texture_create_screen;
-   screen->texture_release = softpipe_texture_release_screen;
-   screen->get_tex_surface = softpipe_get_tex_surface_screen;
+   screen->texture_create = softpipe_texture_create;
+   screen->texture_release = softpipe_texture_release;
+   screen->get_tex_surface = softpipe_get_tex_surface;
 }
diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c
index 4caf2dd3fc..19f71887e7 100644
--- a/src/gallium/drivers/softpipe/sp_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.c
@@ -156,6 +156,7 @@ sp_tile_cache_set_surface(struct softpipe_tile_cache *tc,
    if (tc->surface_map) {
       /*assert(tc->surface != ps);*/
       pipe_surface_unmap(tc->surface);
+      tc->surface_map = NULL;
    }
 
    pipe_surface_reference(&tc->surface, ps);
diff --git a/src/gallium/include/pipe/p_context.h b/src/gallium/include/pipe/p_context.h
index b64948f0f3..a3824601be 100644
--- a/src/gallium/include/pipe/p_context.h
+++ b/src/gallium/include/pipe/p_context.h
@@ -192,12 +192,10 @@ struct pipe_context {
 
    /**
     * Called when texture data is changed.
-    * Note: we could pass some hints about which mip levels or cube faces
-    * have changed...
-    * XXX this may go away - could pass a 'write' flag to get_tex_surface()
     */
    void (*texture_update)(struct pipe_context *pipe,
-                          struct pipe_texture *texture);
+                          struct pipe_texture *texture,
+                          uint face, uint dirtyLevelsMask);
 
 
 
diff --git a/src/gallium/include/pipe/p_debug.h b/src/gallium/include/pipe/p_debug.h
index f3dfa06216..14f8056924 100644
--- a/src/gallium/include/pipe/p_debug.h
+++ b/src/gallium/include/pipe/p_debug.h
@@ -73,6 +73,14 @@ extern "C" {
  */
 void debug_printf(const char *format, ...);
 
+
+/* Dump a blob in hex to the same place that debug_printf sends its
+ * messages:
+ */
+void debug_print_blob( const char *name,
+                       const void *blob,
+                       unsigned size );
+
 /**
  * @sa debug_printf 
  */
@@ -168,6 +176,55 @@ void debug_mask_vprintf(uint32_t uuid,
 #endif
 
 
+/**
+ * Used by debug_dump_enum and debug_dump_flags to describe symbols.
+ */
+struct debug_named_value
+{
+   const char *name;
+   unsigned long value;
+};
+
+
+/**
+ * Some C pre-processor magic to simplify creating named values.
+ * 
+ * Example:
+ * @code
+ * static const debug_named_value my_names[] = {
+ *    DEBUG_NAMED_VALUE(MY_ENUM_VALUE_X),
+ *    DEBUG_NAMED_VALUE(MY_ENUM_VALUE_Y),
+ *    DEBUG_NAMED_VALUE(MY_ENUM_VALUE_Z),
+ *    DEBUG_NAMED_VALUE_END
+ * };
+ * 
+ *    ...
+ *    debug_printf("%s = %s\n", 
+ *                 name,
+ *                 debug_dump_enum(my_names, my_value));
+ *    ...
+ * @endcode
+ */
+#define DEBUG_NAMED_VALUE(__symbol) {#__symbol, (unsigned long)__symbol} 
+#define DEBUG_NAMED_VALUE_END {NULL, 0} 
+
+
+/**
+ * Convert a enum value to a string.
+ */
+const char *
+debug_dump_enum(const struct debug_named_value *names, 
+                unsigned long value);
+
+
+/**
+ * Convert binary flags value to a string.
+ */
+const char *
+debug_dump_flags(const struct debug_named_value *names, 
+                 unsigned long value);
+
+
 #ifdef	__cplusplus
 }
 #endif
diff --git a/src/gallium/include/pipe/p_format.h b/src/gallium/include/pipe/p_format.h
index f90087b3c9..3238b152b6 100644
--- a/src/gallium/include/pipe/p_format.h
+++ b/src/gallium/include/pipe/p_format.h
@@ -38,16 +38,19 @@ extern "C" {
 #endif
 
 /**
- * The PIPE_FORMAT is a 32-bit wide bitfield that encodes all the information
- * needed to uniquely describe a pixel format.
+ * The pipe_format enum is a 32-bit wide bitfield that encodes all the
+ * information needed to uniquely describe a pixel format.
  */
 
 /**
- * Possible format layouts -- occupy first 2 bits. The interpretation of
- * the remaining 30 bits depends on a particual format layout.
+ * Possible format layouts are encoded in the first 2 bits.
+ * The interpretation of the remaining 30 bits depends on a particular
+ * format layout.
  */
 #define PIPE_FORMAT_LAYOUT_RGBAZS   0
 #define PIPE_FORMAT_LAYOUT_YCBCR    1
+#define PIPE_FORMAT_LAYOUT_DXT      2  /**< XXX temporary? */
+
 
 static INLINE uint pf_layout(uint f)  /**< PIPE_FORMAT_LAYOUT_ */
 {
@@ -74,11 +77,13 @@ static INLINE uint pf_layout(uint f)  /**< PIPE_FORMAT_LAYOUT_ */
  * Format types for RGBAZS layout.
  */
 #define PIPE_FORMAT_TYPE_UNKNOWN 0
-#define PIPE_FORMAT_TYPE_FLOAT   1
-#define PIPE_FORMAT_TYPE_UNORM   2
-#define PIPE_FORMAT_TYPE_SNORM   3
-#define PIPE_FORMAT_TYPE_USCALED 4
-#define PIPE_FORMAT_TYPE_SSCALED 5
+#define PIPE_FORMAT_TYPE_FLOAT   1  /**< 16/32/64-bit/channel formats */
+#define PIPE_FORMAT_TYPE_UNORM   2  /**< uints, normalized to [0,1] */
+#define PIPE_FORMAT_TYPE_SNORM   3  /**< ints, normalized to [-1,1] */
+#define PIPE_FORMAT_TYPE_USCALED 4  /**< uints, not normalized */
+#define PIPE_FORMAT_TYPE_SSCALED 5  /**< ints, not normalized */
+#define PIPE_FORMAT_TYPE_SRGB    6  /**< sRGB colorspace */
+
 
 /**
  * Because the destination vector is assumed to be RGBA FLOAT, we
@@ -196,6 +201,20 @@ static INLINE uint pf_rev(pipe_format_ycbcr_t f)
    return (f >> 2) & 0x1;
 }
 
+
+/**
+  * Compresssed format layouts (this will probably change)
+  */
+#define _PIPE_FORMAT_DXT( LEVEL, RSIZE, GSIZE, BSIZE, ASIZE ) \
+   ((PIPE_FORMAT_LAYOUT_DXT << 0) | \
+    ((LEVEL) << 2) | \
+    ((RSIZE) << 5) | \
+    ((GSIZE) << 8) | \
+    ((BSIZE) << 11) | \
+    ((ASIZE) << 14) )
+
+
+
 /**
  * Texture/surface image formats (preliminary)
  */
@@ -281,11 +300,23 @@ enum pipe_format {
    PIPE_FORMAT_R8_SSCALED            = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_R000, 1, 0, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
    PIPE_FORMAT_R8G8_SSCALED          = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RG00, 1, 1, 0, 0, PIPE_FORMAT_TYPE_SSCALED ),
    PIPE_FORMAT_R8G8B8_SSCALED        = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SSCALED ),
-   PIPE_FORMAT_R8G8B8A8_SSCALED      = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED )
+   PIPE_FORMAT_R8G8B8A8_SSCALED      = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SSCALED ),
+   /* sRGB formats */
+   PIPE_FORMAT_L8_SRGB               = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRR1, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_A8_L8_SRGB            = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RRRG, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_R8G8B8_SRGB           = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGB0, 1, 1, 1, 0, PIPE_FORMAT_TYPE_SRGB ),
+   PIPE_FORMAT_R8G8B8A8_SRGB         = _PIPE_FORMAT_RGBAZS_8 ( _PIPE_FORMAT_RGBA, 1, 1, 1, 1, PIPE_FORMAT_TYPE_SRGB ),
+
+   /* compressed formats */
+   PIPE_FORMAT_DXT1_RGB              = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 0 ),
+   PIPE_FORMAT_DXT1_RGBA             = _PIPE_FORMAT_DXT( 1, 8, 8, 8, 8 ),
+   PIPE_FORMAT_DXT3_RGBA             = _PIPE_FORMAT_DXT( 3, 8, 8, 8, 8 ),
+   PIPE_FORMAT_DXT5_RGBA             = _PIPE_FORMAT_DXT( 5, 8, 8, 8, 8 )
 };
 
 
 /**
+ * Unsigned 8-bit stencil format.
  * XXX should remove this, but S8_UNORM is a poor name
  */
 #define PIPE_FORMAT_U_S8                  PIPE_FORMAT_S8_UNORM
@@ -294,11 +325,12 @@ enum pipe_format {
 /**
  * Builds pipe format name from format token.
  */
-static INLINE char *pf_sprint_name( char *str, uint format )
+static INLINE char *pf_sprint_name( char *str, enum pipe_format format )
 {
    strcpy( str, "PIPE_FORMAT_" );
    switch (pf_layout( format )) {
-   case PIPE_FORMAT_LAYOUT_RGBAZS: {
+   case PIPE_FORMAT_LAYOUT_RGBAZS:
+      {
          pipe_format_rgbazs_t rgbazs = (pipe_format_rgbazs_t) format;
          uint                 i;
          uint                 scale = 1 << (pf_exp8( rgbazs ) * 3);
@@ -362,7 +394,8 @@ static INLINE char *pf_sprint_name( char *str, uint format )
          }
       }
       break;
-   case PIPE_FORMAT_LAYOUT_YCBCR: {
+   case PIPE_FORMAT_LAYOUT_YCBCR:
+      {
          pipe_format_ycbcr_t  ycbcr = (pipe_format_ycbcr_t) format;
 
          strcat( str, "YCBCR" );
@@ -375,6 +408,10 @@ static INLINE char *pf_sprint_name( char *str, uint format )
    return str;
 }
 
+/**
+ * Return bits for a particular component.
+ * \param comp  component index, starting at 0
+ */
 static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp )
 {
    uint size;
@@ -397,6 +434,9 @@ static INLINE uint pf_get_component_bits( enum pipe_format format, uint comp )
    return size << (pf_exp8(format) * 3);
 }
 
+/**
+ * Return total bits needed for the pixel format.
+ */
 static INLINE uint pf_get_bits( enum pipe_format format )
 {
    if (pf_layout(format) == PIPE_FORMAT_LAYOUT_RGBAZS) {
@@ -417,7 +457,11 @@ static INLINE uint pf_get_bits( enum pipe_format format )
    }
 }
 
-static INLINE uint pf_get_size( enum pipe_format format ) {
+/**
+ * Return bytes per pixel for the given format.
+ */
+static INLINE uint pf_get_size( enum pipe_format format )
+{
    assert(pf_get_bits(format) % 8 == 0);
    return pf_get_bits(format) / 8;
 }
diff --git a/src/gallium/include/pipe/p_screen.h b/src/gallium/include/pipe/p_screen.h
index 6be9a82b68..35bd46d691 100644
--- a/src/gallium/include/pipe/p_screen.h
+++ b/src/gallium/include/pipe/p_screen.h
@@ -69,6 +69,7 @@ struct pipe_screen {
 
    float (*get_paramf)( struct pipe_screen *, int param );
 
+   /**< type is one of PIPE_TEXTURE, PIPE_SURFACE */
    boolean (*is_format_supported)( struct pipe_screen *,
                                    enum pipe_format format, 
                                    uint type );
diff --git a/src/gallium/include/pipe/p_state.h b/src/gallium/include/pipe/p_state.h
index 5791a10119..3e531c4da4 100644
--- a/src/gallium/include/pipe/p_state.h
+++ b/src/gallium/include/pipe/p_state.h
@@ -220,6 +220,8 @@ struct pipe_blend_color
 
 struct pipe_framebuffer_state
 {
+   unsigned width, height;
+
    /** multiple colorbuffers for multiple render targets */
    unsigned num_cbufs;
    struct pipe_surface *cbufs[PIPE_MAX_COLOR_BUFS];
diff --git a/src/mesa/SConscript b/src/mesa/SConscript
index 678e4ad0e8..b816a21b38 100644
--- a/src/mesa/SConscript
+++ b/src/mesa/SConscript
@@ -142,6 +142,8 @@ STATETRACKER_SOURCES = [
 	'state_tracker/st_atom_texture.c',
 	'state_tracker/st_atom_viewport.c',
 	'state_tracker/st_cb_accum.c',
+	'state_tracker/st_cb_bitmap.c',
+	'state_tracker/st_cb_blit.c',
 	'state_tracker/st_cb_bufferobjects.c',
 	'state_tracker/st_cb_clear.c',
 	'state_tracker/st_cb_flush.c',
@@ -154,7 +156,6 @@ STATETRACKER_SOURCES = [
 	'state_tracker/st_cb_readpixels.c',
 	'state_tracker/st_cb_strings.c',
 	'state_tracker/st_cb_texture.c',
-	'state_tracker/st_cache.c',
 	'state_tracker/st_context.c',
 	'state_tracker/st_debug.c',
 	'state_tracker/st_draw.c',
diff --git a/src/mesa/main/bufferobj.c b/src/mesa/main/bufferobj.c
index 009055a6ab..e762eb3b63 100644
--- a/src/mesa/main/bufferobj.c
+++ b/src/mesa/main/bufferobj.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.1
+ * Version:  7.1
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   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"),
@@ -470,6 +470,143 @@ _mesa_validate_pbo_access(GLuint dimensions,
 
 
 /**
+ * If the source of glBitmap data is a PBO, check that we won't read out
+ * of buffer bounds, then map the buffer.
+ * If not sourcing from a PBO, just return the bitmap pointer.
+ * This is a helper function for (some) drivers.
+ * Return NULL if error.
+ * If non-null return, must call _mesa_unmap_bitmap_pbo() when done.
+ */
+const GLubyte *
+_mesa_map_bitmap_pbo(GLcontext *ctx,
+                     const struct gl_pixelstore_attrib *unpack,
+                     const GLubyte *bitmap)
+{
+   const GLubyte *buf;
+
+   if (unpack->BufferObj->Name) {
+      /* unpack from PBO */
+      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                                              GL_READ_ONLY_ARB,
+                                              unpack->BufferObj);
+      if (!buf)
+         return NULL;
+
+      buf = ADD_POINTERS(buf, bitmap);
+   }
+   else {
+      /* unpack from normal memory */
+      buf = bitmap;
+   }
+
+   return buf;
+}
+
+
+/**
+ * Counterpart to _mesa_map_bitmap_pbo()
+ * This is a helper function for (some) drivers.
+ */
+void
+_mesa_unmap_bitmap_pbo(GLcontext *ctx,
+                       const struct gl_pixelstore_attrib *unpack)
+{
+   if (unpack->BufferObj->Name) {
+      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                              unpack->BufferObj);
+   }
+}
+
+
+/**
+ * \sa _mesa_map_bitmap_pbo
+ */
+const GLvoid *
+_mesa_map_drawpix_pbo(GLcontext *ctx,
+                      const struct gl_pixelstore_attrib *unpack,
+                      const GLvoid *pixels)
+{
+   const GLvoid *buf;
+
+   if (unpack->BufferObj->Name) {
+      /* unpack from PBO */
+      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                                              GL_READ_ONLY_ARB,
+                                              unpack->BufferObj);
+      if (!buf)
+         return NULL;
+
+      buf = ADD_POINTERS(buf, pixels);
+   }
+   else {
+      /* unpack from normal memory */
+      buf = pixels;
+   }
+
+   return buf;
+}
+
+
+/**
+ * \sa _mesa_unmap_bitmap_pbo
+ */
+void
+_mesa_unmap_drapix_pbo(GLcontext *ctx,
+                       const struct gl_pixelstore_attrib *unpack)
+{
+   if (unpack->BufferObj->Name) {
+      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
+                              unpack->BufferObj);
+   }
+}
+
+
+/**
+ * If PBO is bound, map the buffer, return dest pointer in mapped buffer.
+ * Call _mesa_unmap_readpix_pbo() when finished
+ * \return NULL if error
+ */
+void *
+_mesa_map_readpix_pbo(GLcontext *ctx,
+                      const struct gl_pixelstore_attrib *pack,
+                      GLvoid *dest)
+{
+   void *buf;
+
+   if (pack->BufferObj->Name) {
+      /* pack into PBO */
+      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
+                                              GL_WRITE_ONLY_ARB,
+                                              pack->BufferObj);
+      if (!buf)
+         return NULL;
+
+      buf = ADD_POINTERS(buf, dest);
+   }
+   else {
+      /* pack to normal memory */
+      buf = dest;
+   }
+
+   return buf;
+}
+
+
+/**
+ * Counterpart to _mesa_map_readpix_pbo()
+ */
+void
+_mesa_unmap_readpix_pbo(GLcontext *ctx,
+                        const struct gl_pixelstore_attrib *pack)
+{
+   if (pack->BufferObj->Name) {
+      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT, pack->BufferObj);
+   }
+}
+
+
+
+/**
  * Return the gl_buffer_object for the given ID.
  * Always return NULL for ID 0.
  */
diff --git a/src/mesa/main/bufferobj.h b/src/mesa/main/bufferobj.h
index f54f9e9ff0..46525f08ae 100644
--- a/src/mesa/main/bufferobj.h
+++ b/src/mesa/main/bufferobj.h
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.3
+ * Version:  7.1
  *
- * Copyright (C) 1999-2004  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   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"),
@@ -86,6 +86,35 @@ _mesa_validate_pbo_access(GLuint dimensions,
                           GLsizei width, GLsizei height, GLsizei depth,
                           GLenum format, GLenum type, const GLvoid *ptr);
 
+extern const GLubyte *
+_mesa_map_bitmap_pbo(GLcontext *ctx,
+                     const struct gl_pixelstore_attrib *unpack,
+                     const GLubyte *bitmap);
+
+extern void
+_mesa_unmap_bitmap_pbo(GLcontext *ctx,
+                       const struct gl_pixelstore_attrib *unpack);
+
+extern const GLvoid *
+_mesa_map_drawpix_pbo(GLcontext *ctx,
+                      const struct gl_pixelstore_attrib *unpack,
+                      const GLvoid *pixels);
+
+extern void
+_mesa_unmap_drapix_pbo(GLcontext *ctx,
+                       const struct gl_pixelstore_attrib *unpack);
+
+
+extern void *
+_mesa_map_readpix_pbo(GLcontext *ctx,
+                      const struct gl_pixelstore_attrib *pack,
+                      GLvoid *dest);
+
+extern void
+_mesa_unmap_readpix_pbo(GLcontext *ctx,
+                        const struct gl_pixelstore_attrib *pack);
+
+
 extern void
 _mesa_unbind_buffer_object( GLcontext *ctx, struct gl_buffer_object *bufObj );
 
diff --git a/src/mesa/main/context.c b/src/mesa/main/context.c
index a9f9bd9da4..d06644f65d 100644
--- a/src/mesa/main/context.c
+++ b/src/mesa/main/context.c
@@ -695,10 +695,10 @@ free_shared_state( GLcontext *ctx, struct gl_shared_state *ss )
    _mesa_DeleteHashTable(ss->Programs);
 #endif
 #if FEATURE_ARB_vertex_program
-   _mesa_delete_program(ctx, ss->DefaultVertexProgram);
+   ctx->Driver.DeleteProgram(ctx, ss->DefaultVertexProgram);
 #endif
 #if FEATURE_ARB_fragment_program
-   _mesa_delete_program(ctx, ss->DefaultFragmentProgram);
+   ctx->Driver.DeleteProgram(ctx, ss->DefaultFragmentProgram);
 #endif
 
 #if FEATURE_ATI_fragment_shader
diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c
index ae9c7e29a1..0f64f1c1c0 100644
--- a/src/mesa/main/drawpix.c
+++ b/src/mesa/main/drawpix.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5
+ * Version:  7.1
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   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"),
@@ -24,6 +24,7 @@
 
 #include "glheader.h"
 #include "imports.h"
+#include "bufferobj.h"
 #include "context.h"
 #include "drawpix.h"
 #include "feedback.h"
@@ -183,6 +184,23 @@ _mesa_DrawPixels( GLsizei width, GLsizei height,
       /* Round, to satisfy conformance tests (matches SGI's OpenGL) */
       GLint x = IROUND(ctx->Current.RasterPos[0]);
       GLint y = IROUND(ctx->Current.RasterPos[1]);
+
+      if (ctx->Unpack.BufferObj->Name) {
+         /* unpack from PBO */
+         if (!_mesa_validate_pbo_access(2, &ctx->Unpack, width, height, 1,
+                                        format, type, pixels)) {
+            _mesa_error(ctx, GL_INVALID_OPERATION,
+                        "glDrawPixels(invalid PBO access)");
+            return;
+         }
+         if (ctx->Unpack.BufferObj->Pointer) {
+            /* buffer is mapped - that's an error */
+            _mesa_error(ctx, GL_INVALID_OPERATION,
+                        "glDrawPixels(PBO is mapped)");
+            return;
+         }
+      }
+
       ctx->Driver.DrawPixels(ctx, x, y, width, height, format, type,
 			     &ctx->Unpack, pixels);
    }
@@ -303,6 +321,21 @@ _mesa_ReadPixels( GLint x, GLint y, GLsizei width, GLsizei height,
       return;
    }
 
+   if (ctx->Pack.BufferObj->Name) {
+      if (!_mesa_validate_pbo_access(2, &ctx->Pack, width, height, 1,
+                                     format, type, pixels)) {
+         _mesa_error(ctx, GL_INVALID_OPERATION,
+                     "glReadPixels(invalid PBO access)");
+         return;
+      }
+
+      if (ctx->Pack.BufferObj->Pointer) {
+         /* buffer is mapped - that's an error */
+         _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(PBO is mapped)");
+         return;
+      }
+   }
+
    ctx->Driver.ReadPixels(ctx, x, y, width, height,
 			  format, type, &ctx->Pack, pixels);
 }
@@ -343,12 +376,27 @@ _mesa_Bitmap( GLsizei width, GLsizei height,
    }
 
    if (ctx->RenderMode == GL_RENDER) {
-      if (bitmap && width && height) {
-         /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */
-         GLint x = IFLOOR(ctx->Current.RasterPos[0] - xorig);
-         GLint y = IFLOOR(ctx->Current.RasterPos[1] - yorig);
-         ctx->Driver.Bitmap( ctx, x, y, width, height, &ctx->Unpack, bitmap );
+      /* Truncate, to satisfy conformance tests (matches SGI's OpenGL). */
+      GLint x = IFLOOR(ctx->Current.RasterPos[0] - xorig);
+      GLint y = IFLOOR(ctx->Current.RasterPos[1] - yorig);
+
+      if (ctx->Unpack.BufferObj->Name) {
+         /* unpack from PBO */
+         if (!_mesa_validate_pbo_access(2, &ctx->Unpack, width, height, 1,
+                                        GL_COLOR_INDEX, GL_BITMAP,
+                                        (GLvoid *) bitmap)) {
+            _mesa_error(ctx, GL_INVALID_OPERATION,
+                        "glBitmap(invalid PBO access)");
+            return;
+         }
+         if (ctx->Unpack.BufferObj->Pointer) {
+            /* buffer is mapped - that's an error */
+            _mesa_error(ctx, GL_INVALID_OPERATION, "glBitmap(PBO is mapped)");
+            return;
+         }
       }
+
+      ctx->Driver.Bitmap( ctx, x, y, width, height, &ctx->Unpack, bitmap );
    }
 #if _HAVE_FULL_GL
    else if (ctx->RenderMode == GL_FEEDBACK) {
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index f15e404527..384e145520 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -2323,9 +2323,6 @@ _mesa_GetTexImage( GLenum target, GLint level, GLenum format,
       return;
    }
 
-   if (!pixels)
-      return;
-
    _mesa_lock_texture(ctx, texObj);
    {
       texImage = _mesa_select_tex_image(ctx, texObj, target, level);
diff --git a/src/mesa/shader/program.c b/src/mesa/shader/program.c
index 3869e31714..39c4ca1dd5 100644
--- a/src/mesa/shader/program.c
+++ b/src/mesa/shader/program.c
@@ -354,7 +354,7 @@ _mesa_clone_program(GLcontext *ctx, const struct gl_program *prog)
    clone->Format = prog->Format;
    clone->Instructions = _mesa_alloc_instructions(prog->NumInstructions);
    if (!clone->Instructions) {
-      _mesa_delete_program(ctx, clone);
+      ctx->Driver.DeleteProgram(ctx, clone);
       return NULL;
    }
    _mesa_copy_instructions(clone->Instructions, prog->Instructions,
diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index 4470a87af0..4cb8bb8ed1 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -79,7 +79,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx,
          /* to prevent a double-free in the next call */
          shProg->VertexProgram->Base.Parameters = NULL;
       }
-      _mesa_delete_program(ctx, &shProg->VertexProgram->Base);
+      ctx->Driver.DeleteProgram(ctx, &shProg->VertexProgram->Base);
       shProg->VertexProgram = NULL;
    }
 
@@ -88,7 +88,7 @@ _mesa_clear_shader_program_data(GLcontext *ctx,
          /* to prevent a double-free in the next call */
          shProg->FragmentProgram->Base.Parameters = NULL;
       }
-      _mesa_delete_program(ctx, &shProg->FragmentProgram->Base);
+      ctx->Driver.DeleteProgram(ctx, &shProg->FragmentProgram->Base);
       shProg->FragmentProgram = NULL;
    }
 
@@ -246,7 +246,7 @@ _mesa_free_shader(GLcontext *ctx, struct gl_shader *sh)
       _mesa_free(sh->InfoLog);
    for (i = 0; i < sh->NumPrograms; i++) {
       assert(sh->Programs[i]);
-      _mesa_delete_program(ctx, sh->Programs[i]);
+      ctx->Driver.DeleteProgram(ctx, sh->Programs[i]);
    }
    if (sh->Programs)
       _mesa_free(sh->Programs);
diff --git a/src/mesa/sources b/src/mesa/sources
index 287af7121a..d109dce5bc 100644
--- a/src/mesa/sources
+++ b/src/mesa/sources
@@ -172,6 +172,7 @@ STATETRACKER_SOURCES = \
 	state_tracker/st_atom_texture.c \
 	state_tracker/st_atom_viewport.c \
 	state_tracker/st_cb_accum.c \
+	state_tracker/st_cb_bitmap.c \
 	state_tracker/st_cb_blit.c \
 	state_tracker/st_cb_bufferobjects.c \
 	state_tracker/st_cb_clear.c \
diff --git a/src/mesa/state_tracker/st_atom_blend.c b/src/mesa/state_tracker/st_atom_blend.c
index 6c13fc8141..2a0e92245c 100644
--- a/src/mesa/state_tracker/st_atom_blend.c
+++ b/src/mesa/state_tracker/st_atom_blend.c
@@ -39,6 +39,7 @@
 #include "pipe/p_defines.h"
 #include "cso_cache/cso_context.h"
 
+#include "main/macros.h"
 
 /**
  * Convert GLenum blend tokens to pipe tokens.
@@ -213,13 +214,10 @@ update_blend( struct st_context *st )
 
    cso_set_blend(st->cso_context, blend);
 
-   if (memcmp(st->ctx->Color.BlendColor, &st->state.blend_color, 4 * sizeof(GLfloat)) != 0) {
-      /* state has changed */
-      st->state.blend_color.color[0] = st->ctx->Color.BlendColor[0];
-      st->state.blend_color.color[1] = st->ctx->Color.BlendColor[1];
-      st->state.blend_color.color[2] = st->ctx->Color.BlendColor[2];
-      st->state.blend_color.color[3] = st->ctx->Color.BlendColor[3];
-      st->pipe->set_blend_color(st->pipe, (struct pipe_blend_color *) st->ctx->Color.BlendColor);
+   {
+      struct pipe_blend_color bc;
+      COPY_4FV(bc.color, st->ctx->Color.BlendColor);
+      cso_set_blend_color(st->cso_context, &bc);
    }
 }
 
diff --git a/src/mesa/state_tracker/st_atom_constbuf.c b/src/mesa/state_tracker/st_atom_constbuf.c
index 21416da2e0..cd7372902b 100644
--- a/src/mesa/state_tracker/st_atom_constbuf.c
+++ b/src/mesa/state_tracker/st_atom_constbuf.c
@@ -66,7 +66,8 @@ void st_upload_constants( struct st_context *st,
       /* Update our own dependency flags.  This works because this
        * function will also be called whenever the program changes.
        */
-      st->constants.tracked_state[id].dirty.mesa = params->StateFlags;
+      st->constants.tracked_state[id].dirty.mesa =
+         (params->StateFlags | _NEW_PROGRAM);
 
       _mesa_load_state_parameters(st->ctx, params);
 
@@ -115,7 +116,7 @@ static void update_vs_constants(struct st_context *st )
 const struct st_tracked_state st_update_vs_constants = {
    .name = "st_update_vs_constants",
    .dirty = {
-      .mesa  = 0, 
+      .mesa  = 0,  /* set dynamically above */
       .st   = ST_NEW_VERTEX_PROGRAM,
    },
    .update = update_vs_constants
@@ -134,7 +135,7 @@ static void update_fs_constants(struct st_context *st )
 const struct st_tracked_state st_update_fs_constants = {
    .name = "st_update_fs_constants",
    .dirty = {
-      .mesa  = 0, 
+      .mesa  = 0,  /* set dynamically above */
       .st   = ST_NEW_FRAGMENT_PROGRAM,
    },
    .update = update_fs_constants
diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c
index 3e58d49f1f..02573af8f0 100644
--- a/src/mesa/state_tracker/st_atom_framebuffer.c
+++ b/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -35,6 +35,7 @@
 #include "st_atom.h"
 #include "st_cb_fbo.h"
 #include "pipe/p_context.h"
+#include "cso_cache/cso_context.h"
 
 
 /**
@@ -44,45 +45,42 @@
 static void
 update_framebuffer_state( struct st_context *st )
 {
-   struct pipe_framebuffer_state framebuffer;
+   struct pipe_framebuffer_state *framebuffer = &st->state.framebuffer;
    struct gl_framebuffer *fb = st->ctx->DrawBuffer;
    struct st_renderbuffer *strb;
    GLuint i;
 
-   memset(&framebuffer, 0, sizeof(framebuffer));
+   memset(framebuffer, 0, sizeof(*framebuffer));
+
+   framebuffer->width = fb->Width;
+   framebuffer->height = fb->Height;
 
    /* Examine Mesa's ctx->DrawBuffer->_ColorDrawBuffers state
     * to determine which surfaces to draw to
     */
-   framebuffer.num_cbufs = fb->_NumColorDrawBuffers[0];
-   for (i = 0; i < framebuffer.num_cbufs; i++) {
+   framebuffer->num_cbufs = fb->_NumColorDrawBuffers[0];
+   for (i = 0; i < framebuffer->num_cbufs; i++) {
       strb = st_renderbuffer(fb->_ColorDrawBuffers[0][i]);
       assert(strb->surface);
-      framebuffer.cbufs[i] = strb->surface;
+      framebuffer->cbufs[i] = strb->surface;
    }
 
    strb = st_renderbuffer(fb->Attachment[BUFFER_DEPTH].Renderbuffer);
    if (strb) {
       strb = st_renderbuffer(strb->Base.Wrapped);
       assert(strb->surface);
-      framebuffer.zsbuf = strb->surface;
+      framebuffer->zsbuf = strb->surface;
    }
    else {
       strb = st_renderbuffer(fb->Attachment[BUFFER_STENCIL].Renderbuffer);
       if (strb) {
          strb = st_renderbuffer(strb->Base.Wrapped);
          assert(strb->surface);
-         framebuffer.zsbuf = strb->surface;
+         framebuffer->zsbuf = strb->surface;
       }
    }
 
-   /* XXX: The memcmp is insufficient for eliminating redundant state changes,
-    * but we should probably do more work here to that end.
-    */
-   if (1 /*memcmp(&framebuffer, &st->state.framebuffer, sizeof(framebuffer)) != 0*/) {
-      st->state.framebuffer = framebuffer;
-      st->pipe->set_framebuffer_state( st->pipe, &framebuffer );
-   }
+   cso_set_framebuffer(st->cso_context, framebuffer);
 }
 
 
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index 0726688493..ec39026eb3 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -78,34 +78,6 @@ struct translated_vertex_program
 
 
 /**
- * Free data hanging off the st vert prog.
- */
-void
-st_remove_vertex_program(struct st_context *st, struct st_vertex_program *stvp)
-{
-   /* no-op, for now? */
-}
-
-
-/**
- * Free data hanging off the st frag prog.
- */
-void
-st_remove_fragment_program(struct st_context *st,
-                          struct st_fragment_program *stfp)
-{
-   struct translated_vertex_program *xvp, *next;
-
-   for (xvp = stfp->vertex_programs; xvp; xvp = next) {
-      next = xvp->next;
-      /* XXX free xvp->vs */
-      free(xvp);
-   }
-}
-
-
-
-/**
  * Given a vertex program output attribute, return the corresponding
  * fragment program input attribute.
  * \return -1 for vertex outputs that have no corresponding fragment input
@@ -265,6 +237,21 @@ find_translated_vp(struct st_context *st,
 }
 
 
+void
+st_free_translated_vertex_programs(struct st_context *st,
+                                   struct translated_vertex_program *xvp)
+{
+   struct translated_vertex_program *next;
+
+   while (xvp) {
+      next = xvp->next;
+      free(xvp);
+      xvp = next;
+   }
+}
+
+
+
 static void
 update_linkage( struct st_context *st )
 {
diff --git a/src/mesa/state_tracker/st_atom_shader.h b/src/mesa/state_tracker/st_atom_shader.h
index f3124d87c0..8403bc66c9 100644
--- a/src/mesa/state_tracker/st_atom_shader.h
+++ b/src/mesa/state_tracker/st_atom_shader.h
@@ -31,10 +31,8 @@
 
 
 extern void
-st_remove_vertex_program(struct st_context *, struct st_vertex_program *);
-
-extern void
-st_remove_fragment_program(struct st_context *, struct st_fragment_program *);
+st_free_translated_vertex_programs(struct st_context *st,
+                                   struct translated_vertex_program *xvp);
 
 
 #endif /* ST_ATOM_SHADER_H */
diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c
index e53a897637..9aef30f456 100644
--- a/src/mesa/state_tracker/st_atom_texture.c
+++ b/src/mesa/state_tracker/st_atom_texture.c
@@ -64,7 +64,10 @@ update_textures(struct st_context *st)
          GLboolean flush, retval;
 
          retval = st_finalize_texture(st->ctx, st->pipe, texObj, &flush);
-         /* XXX retval indicates whether there's a texture border */
+         if (!retval) {
+            /* out of mem */
+            continue;
+         }
 
          st->state.num_textures = unit + 1;
       }
@@ -76,11 +79,6 @@ update_textures(struct st_context *st)
 
       pt = st_get_stobj_texture(stObj);
       pipe_texture_reference(&st->state.sampler_texture[unit], pt);
-
-      if (stObj && stObj->dirtyData) {
-         st->pipe->texture_update(st->pipe, pt);
-         stObj->dirtyData = GL_FALSE;
-      }
    }
 
    st->pipe->set_sampler_textures(st->pipe, st->state.num_textures,
diff --git a/src/mesa/state_tracker/st_atom_viewport.c b/src/mesa/state_tracker/st_atom_viewport.c
index 147aa3c51a..eb3f62cfbe 100644
--- a/src/mesa/state_tracker/st_atom_viewport.c
+++ b/src/mesa/state_tracker/st_atom_viewport.c
@@ -29,9 +29,9 @@
 #include "context.h"
 #include "colormac.h"
 #include "st_context.h"
-#include "pipe/p_context.h"
 #include "st_atom.h"
-
+#include "pipe/p_context.h"
+#include "cso_cache/cso_context.h"
 
 /**
  * Update the viewport transformation matrix.  Depends on:
@@ -65,22 +65,18 @@ update_viewport( struct st_context *st )
       GLfloat half_width = ctx->Viewport.Width / 2.0;
       GLfloat half_height = ctx->Viewport.Height / 2.0;
       GLfloat half_depth = (ctx->Viewport.Far - ctx->Viewport.Near) / 2.0;
-      struct pipe_viewport_state vp;
       
-      vp.scale[0] = half_width;
-      vp.scale[1] = half_height * yScale;
-      vp.scale[2] = half_depth;
-      vp.scale[3] = 1.0;
+      st->state.viewport.scale[0] = half_width;
+      st->state.viewport.scale[1] = half_height * yScale;
+      st->state.viewport.scale[2] = half_depth;
+      st->state.viewport.scale[3] = 1.0;
 
-      vp.translate[0] = half_width + x;
-      vp.translate[1] = (half_height + y) * yScale + yBias;
-      vp.translate[2] = half_depth + z;
-      vp.translate[3] = 0.0;
+      st->state.viewport.translate[0] = half_width + x;
+      st->state.viewport.translate[1] = (half_height + y) * yScale + yBias;
+      st->state.viewport.translate[2] = half_depth + z;
+      st->state.viewport.translate[3] = 0.0;
 
-      if (memcmp(&vp, &st->state.viewport, sizeof(vp)) != 0) {
-	 st->state.viewport = vp;
-	 st->pipe->set_viewport_state(st->pipe, &vp);
-      }
+      cso_set_viewport(st->cso_context, &st->state.viewport);
    }
 }
 
diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c
new file mode 100644
index 0000000000..acc22d4323
--- /dev/null
+++ b/src/mesa/state_tracker/st_cb_bitmap.c
@@ -0,0 +1,527 @@
+/**************************************************************************
+ * 
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * 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 TUNGSTEN GRAPHICS 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.
+ * 
+ **************************************************************************/
+
+ /*
+  * Authors:
+  *   Brian Paul
+  */
+
+#include "main/imports.h"
+#include "main/image.h"
+#include "main/bufferobj.h"
+#include "main/macros.h"
+#include "main/texformat.h"
+#include "shader/program.h"
+#include "shader/prog_parameter.h"
+#include "shader/prog_print.h"
+
+#include "st_context.h"
+#include "st_atom.h"
+#include "st_atom_constbuf.h"
+#include "st_program.h"
+#include "st_cb_bitmap.h"
+#include "st_mesa_to_tgsi.h"
+#include "st_texture.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_inlines.h"
+#include "pipe/p_winsys.h"
+#include "util/p_tile.h"
+#include "util/u_draw_quad.h"
+#include "util/u_simple_shaders.h"
+#include "shader/prog_instruction.h"
+#include "cso_cache/cso_context.h"
+
+
+
+/**
+ * Make fragment program for glBitmap:
+ *   Sample the texture and kill the fragment if the bit is 0.
+ * This program will be combined with the user's fragment program.
+ */
+static struct st_fragment_program *
+make_bitmap_fragment_program(GLcontext *ctx)
+{
+   struct st_fragment_program *stfp;
+   struct gl_program *p;
+   GLuint ic = 0;
+
+   p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
+   if (!p)
+      return NULL;
+
+   p->NumInstructions = 5;
+
+   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
+   if (!p->Instructions) {
+      ctx->Driver.DeleteProgram(ctx, p);
+      return NULL;
+   }
+   _mesa_init_instructions(p->Instructions, p->NumInstructions);
+
+   /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */
+   p->Instructions[ic].Opcode = OPCODE_TEX;
+   p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].DstReg.Index = 0;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
+   p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
+   p->Instructions[ic].TexSrcUnit = 0;
+   p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
+   ic++;
+
+   /* SWZ tmp0.x, tmp0.x, 1111; # tmp0.x = 1.0 */
+   p->Instructions[ic].Opcode = OPCODE_SWZ;
+   p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].DstReg.Index = 0;
+   p->Instructions[ic].DstReg.WriteMask = WRITEMASK_X;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].SrcReg[0].Index = 0;
+   p->Instructions[ic].SrcReg[0].Swizzle
+      = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE );
+   ic++;
+
+   /* SUB tmp0, tmp0.wwww, tmp0.xxxx;  #  tmp0.w -= 1 */
+   p->Instructions[ic].Opcode = OPCODE_SUB;
+   p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].DstReg.Index = 0;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].SrcReg[0].Index = 0;
+   p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW;
+   p->Instructions[ic].SrcReg[1].File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].SrcReg[1].Index = 0;
+   p->Instructions[ic].SrcReg[1].Swizzle = SWIZZLE_XXXX; /* 1.0 */
+   ic++;
+
+   /* KIL if tmp0 < 0 */
+   p->Instructions[ic].Opcode = OPCODE_KIL;
+   p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
+   p->Instructions[ic].SrcReg[0].Index = 0;
+   ic++;
+
+   /* END; */
+   p->Instructions[ic++].Opcode = OPCODE_END;
+
+   assert(ic == p->NumInstructions);
+
+   p->InputsRead = FRAG_BIT_TEX0;
+   p->OutputsWritten = 0x0;
+
+   stfp = (struct st_fragment_program *) p;
+   stfp->Base.UsesKill = GL_TRUE;
+   st_translate_fragment_program(ctx->st, stfp, NULL);
+
+   return stfp;
+}
+
+
+/**
+ * Combine basic bitmap fragment program with the user-defined program.
+ */
+static struct st_fragment_program *
+combined_bitmap_fragment_program(GLcontext *ctx)
+{
+   struct st_context *st = ctx->st;
+   struct st_fragment_program *stfp;
+
+   if (!st->bitmap.program) {
+      /* create the basic bitmap fragment program */
+      st->bitmap.program = make_bitmap_fragment_program(ctx);
+   }
+
+   if (st->bitmap.user_prog_sn == st->fp->serialNo) {
+      /* re-use */
+      stfp = st->bitmap.combined_prog;
+   }
+   else {
+      /* Concatenate the bitmap program with the current user-defined program.
+       */
+      stfp = (struct st_fragment_program *)
+         _mesa_combine_programs(ctx,
+                                &st->bitmap.program->Base.Base,
+                                &st->fp->Base.Base);
+
+#if 0
+      {
+         struct gl_program *p = &stfp->Base.Base;
+         printf("Combined bitmap program:\n");
+         _mesa_print_program(p);
+         printf("InputsRead: 0x%x\n", p->InputsRead);
+         printf("OutputsWritten: 0x%x\n", p->OutputsWritten);
+         _mesa_print_parameter_list(p->Parameters);
+      }
+#endif
+
+      /* translate to TGSI tokens */
+      st_translate_fragment_program(st, stfp, NULL);
+
+      /* save new program, update serial numbers */
+      st->bitmap.user_prog_sn = st->fp->serialNo;
+      st->bitmap.combined_prog = stfp;
+   }
+
+   /* Ideally we'd have updated the pipe constants during the normal
+    * st/atom mechanism.  But we can't since this is specific to glBitmap.
+    */
+   st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT);
+
+   return stfp;
+}
+
+
+/**
+ * Create a texture which represents a bitmap image.
+ */
+static struct pipe_texture *
+make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height,
+                    const struct gl_pixelstore_attrib *unpack,
+                    const GLubyte *bitmap)
+{
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_surface *surface;
+   uint format = 0, cpp, comp;
+   ubyte *dest;
+   struct pipe_texture *pt;
+   int row, col;
+
+   /* find a texture format we know */
+   if (screen->is_format_supported( screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE )) {
+      format = PIPE_FORMAT_U_I8;
+      cpp = 1;
+      comp = 0;
+   }
+   else if (screen->is_format_supported( screen, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_TEXTURE )) {
+      format = PIPE_FORMAT_A8R8G8B8_UNORM;
+      cpp = 4;
+      comp = 3; /* alpha channel */ /*XXX little-endian dependency */
+   }
+   else {
+      /* XXX support more formats */
+      assert( 0 );
+   }
+
+   /* PBO source... */
+   bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap);
+   if (!bitmap) {
+      return NULL;
+   }
+
+   /**
+    * Create texture to hold bitmap pattern.
+    */
+   pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, width, height,
+			  1, 0);
+   if (!pt) {
+      _mesa_unmap_bitmap_pbo(ctx, unpack);
+      return NULL;
+   }
+
+   surface = screen->get_tex_surface(screen, pt, 0, 0, 0);
+
+   /* map texture surface */
+   dest = pipe_surface_map(surface);
+
+   /* Put image into texture surface.
+    * Note that the image is actually going to be upside down in
+    * the texture.  We deal with that with texcoords.
+    */
+
+   for (row = 0; row < height; row++) {
+      const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack,
+                 bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0);
+      ubyte *destRow = dest + row * surface->pitch * cpp;
+
+      if (unpack->LsbFirst) {
+         /* Lsb first */
+         GLubyte mask = 1U << (unpack->SkipPixels & 0x7);
+         for (col = 0; col < width; col++) {
+
+            /* set texel to 255 if bit is set */
+            destRow[comp] = (*src & mask) ? 255 : 0;
+            destRow += cpp;
+
+            if (mask == 128U) {
+               src++;
+               mask = 1U;
+            }
+            else {
+               mask = mask << 1;
+            }
+         }
+
+         /* get ready for next row */
+         if (mask != 1)
+            src++;
+      }
+      else {
+         /* Msb first */
+         GLubyte mask = 128U >> (unpack->SkipPixels & 0x7);
+         for (col = 0; col < width; col++) {
+
+            /* set texel to 255 if bit is set */
+            destRow[comp] =(*src & mask) ? 255 : 0;
+            destRow += cpp;
+
+            if (mask == 1U) {
+               src++;
+               mask = 128U;
+            }
+            else {
+               mask = mask >> 1;
+            }
+         }
+
+         /* get ready for next row */
+         if (mask != 128)
+            src++;
+      }
+
+   } /* row */
+
+   _mesa_unmap_bitmap_pbo(ctx, unpack);
+
+   /* Release surface */
+   pipe_surface_unmap(surface);
+   pipe_surface_reference(&surface, NULL);
+   pipe->texture_update(pipe, pt, 0, 0x1);
+
+   pt->format = format;
+
+   return pt;
+}
+
+
+static void
+setup_bitmap_vertex_data(struct st_context *st,
+                         int x, int y, int width, int height,
+                         float z, const float color[4])
+{
+   struct pipe_context *pipe = st->pipe;
+   const struct gl_framebuffer *fb = st->ctx->DrawBuffer;
+   const GLboolean invert = (st_fb_orientation(fb) == Y_0_TOP);
+   const GLfloat x0 = x;
+   const GLfloat x1 = x + width;
+   const GLfloat y0 = invert ? (fb->Height - y - height) : y;
+   const GLfloat y1 = invert ? (y0 + height) : y + height;
+   const GLfloat bias = st->bitmap_texcoord_bias;
+   const GLfloat xBias = bias / (x1-x0);
+   const GLfloat yBias = bias / (y1-y0);
+   const GLfloat sLeft = 0.0 + xBias, sRight = 1.0 + xBias;
+   const GLfloat tTop = 1.0 - yBias, tBot = 1.0 - tTop - yBias;
+   GLuint i;
+   void *buf;
+
+   if (!st->bitmap.vbuf) {
+      st->bitmap.vbuf = pipe->winsys->buffer_create(pipe->winsys, 32,
+                                                   PIPE_BUFFER_USAGE_VERTEX,
+                                                   sizeof(st->bitmap.vertices));
+   }
+
+   /* positions, texcoords */
+   st->bitmap.vertices[0][0][0] = x0;
+   st->bitmap.vertices[0][0][1] = y0;
+   st->bitmap.vertices[0][2][0] = sLeft;
+   st->bitmap.vertices[0][2][1] = tTop;
+
+   st->bitmap.vertices[1][0][0] = x1;
+   st->bitmap.vertices[1][0][1] = y0;
+   st->bitmap.vertices[1][2][0] = sRight;
+   st->bitmap.vertices[1][2][1] = tTop;
+   
+   st->bitmap.vertices[2][0][0] = x1;
+   st->bitmap.vertices[2][0][1] = y1;
+   st->bitmap.vertices[2][2][0] = sRight;
+   st->bitmap.vertices[2][2][1] = tBot;
+   
+   st->bitmap.vertices[3][0][0] = x0;
+   st->bitmap.vertices[3][0][1] = y1;
+   st->bitmap.vertices[3][2][0] = sLeft;
+   st->bitmap.vertices[3][2][1] = tBot;
+   
+   /* same for all verts: */
+   for (i = 0; i < 4; i++) {
+      st->bitmap.vertices[i][0][2] = z;
+      st->bitmap.vertices[i][0][3] = 1.0;
+      st->bitmap.vertices[i][1][0] = color[0];
+      st->bitmap.vertices[i][1][1] = color[1];
+      st->bitmap.vertices[i][1][2] = color[2];
+      st->bitmap.vertices[i][1][3] = color[3];
+      st->bitmap.vertices[i][2][2] = 0.0; /*R*/
+      st->bitmap.vertices[i][2][3] = 1.0; /*Q*/
+   }
+
+   /* put vertex data into vbuf */
+   buf = pipe->winsys->buffer_map(pipe->winsys, st->bitmap.vbuf,
+                                  PIPE_BUFFER_USAGE_CPU_WRITE);
+   memcpy(buf, st->bitmap.vertices, sizeof(st->bitmap.vertices));
+   pipe->winsys->buffer_unmap(pipe->winsys, st->bitmap.vbuf);
+}
+
+
+
+/**
+ * Render a glBitmap by drawing a textured quad
+ */
+static void
+draw_bitmap_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
+                 GLsizei width, GLsizei height,
+                 struct pipe_texture *pt,
+                 struct st_fragment_program *stfp)
+{
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = ctx->st->pipe;
+   struct cso_context *cso = ctx->st->cso_context;
+   GLuint maxSize;
+
+   /* limit checks */
+   /* XXX if DrawPixels image is larger than max texture size, break
+    * it up into chunks.
+    */
+   maxSize = 1 << (pipe->screen->get_param(pipe->screen, PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
+   assert(width <= maxSize);
+   assert(height <= maxSize);
+
+   cso_save_rasterizer(cso);
+   //cso_save_viewport(cso);
+
+   /* rasterizer state: just scissor */
+   {
+      struct pipe_rasterizer_state rasterizer;
+      memset(&rasterizer, 0, sizeof(rasterizer));
+      if (ctx->Scissor.Enabled)
+         rasterizer.scissor = 1;
+      rasterizer.bypass_clipping = 1;
+
+      cso_set_rasterizer(cso, &rasterizer);
+   }
+
+   /* fragment shader state: TEX lookup program */
+   pipe->bind_fs_state(pipe, stfp->driver_shader);
+
+   /* vertex shader state: position + texcoord pass-through */
+   pipe->bind_vs_state(pipe, st->bitmap.vs);
+
+   /* sampler / texture state */
+   {
+      struct pipe_sampler_state sampler;
+      memset(&sampler, 0, sizeof(sampler));
+      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP;
+      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP;
+      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP;
+      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.normalized_coords = 1;
+
+      cso_single_sampler(cso, 0, &sampler);
+      cso_single_sampler_done(cso);
+
+      pipe->set_sampler_textures(pipe, 1, &pt);
+   }
+
+   /* draw textured quad */
+   setup_bitmap_vertex_data(st, x, y, width, height,
+                            ctx->Current.RasterPos[2],
+                            ctx->Current.RasterColor);
+
+   util_draw_vertex_buffer(pipe, st->bitmap.vbuf,
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           3); /* attribs/vert */
+
+
+   /* restore state */
+   cso_restore_rasterizer(cso);
+   cso_restore_samplers(cso);
+   //cso_restore_viewport(cso);
+   /* shaders don't go through cso yet */
+   pipe->bind_fs_state(pipe, st->fp->driver_shader);
+   pipe->bind_vs_state(pipe, st->vp->driver_shader);
+
+   pipe->set_sampler_textures(pipe, ctx->st->state.num_textures,
+                              ctx->st->state.sampler_texture);
+}
+
+
+
+static void
+st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
+          const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap )
+{
+   struct st_fragment_program *stfp;
+   struct st_context *st = ctx->st;
+   struct pipe_texture *pt;
+
+   stfp = combined_bitmap_fragment_program(ctx);
+
+   if (!st->bitmap.vs) {
+      /* create pass-through vertex shader now */
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_COLOR,
+                                      TGSI_SEMANTIC_GENERIC };
+      const uint semantic_indexes[] = { 0, 0, 0 };
+      st->bitmap.vs = util_make_vertex_passthrough_shader(st->pipe, 3,
+                                                          semantic_names,
+                                                          semantic_indexes);
+   }
+
+   st_validate_state(st);
+
+   pt = make_bitmap_texture(ctx, width, height, unpack, bitmap);
+   if (pt) {
+      draw_bitmap_quad(ctx, x, y, ctx->Current.RasterPos[2],
+                       width, height,
+                       pt, stfp);
+      pipe_texture_reference(&pt, NULL);
+   }
+}
+
+
+
+void st_init_bitmap_functions(struct dd_function_table *functions)
+{
+   functions->Bitmap = st_Bitmap;
+}
+
+
+void
+st_destroy_bitmap(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   /* XXX free frag shader state */
+
+   if (st->bitmap.vs) {
+      pipe->delete_vs_state(pipe, st->bitmap.vs);
+      st->bitmap.vs = NULL;
+   }
+   if (st->bitmap.vbuf) {
+      pipe->winsys->buffer_destroy(pipe->winsys, st->bitmap.vbuf);
+      st->bitmap.vbuf = NULL;
+   }
+}
+
diff --git a/src/gallium/drivers/cell/spu/spu_blend.h b/src/mesa/state_tracker/st_cb_bitmap.h
index 2b594b578b..ac19e0ebb1 100644
--- a/src/gallium/drivers/cell/spu/spu_blend.h
+++ b/src/mesa/state_tracker/st_cb_bitmap.h
@@ -1,8 +1,8 @@
 /**************************************************************************
  * 
- * Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
  * 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
@@ -26,12 +26,16 @@
  **************************************************************************/
 
 
-#ifndef SPU_BLEND_H
-#define SPU_BLEND_H
+#ifndef ST_CB_BITMAP_H
+#define ST_CB_BITMAP_H
+
+
+extern void
+st_init_bitmap_functions(struct dd_function_table *functions);
 
 
 extern void
-blend_quad(uint itx, uint ity, vector float colors[4]);
+st_destroy_bitmap(struct st_context *st);
 
 
-#endif /* SPU_BLEND_H */
+#endif /* ST_CB_BITMAP_H */
diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c
index dfa79c975c..63211d8b66 100644
--- a/src/mesa/state_tracker/st_cb_blit.c
+++ b/src/mesa/state_tracker/st_cb_blit.c
@@ -52,7 +52,7 @@
 void
 st_init_blit(struct st_context *st)
 {
-   st->blit = util_create_blit(st->pipe);
+   st->blit = util_create_blit(st->pipe, st->cso_context);
 }
 
 
@@ -85,11 +85,14 @@ st_BlitFramebuffer(GLcontext *ctx,
       struct pipe_surface *srcSurf = srcRb->surface;
       struct pipe_surface *dstSurf = dstRb->surface;
 
-      srcY0 = srcRb->Base.Height - srcY0;
-      srcY1 = srcRb->Base.Height - srcY1;
+      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+         /* invert Y */
+         srcY0 = srcRb->Base.Height - srcY0;
+         srcY1 = srcRb->Base.Height - srcY1;
 
-      dstY0 = dstRb->Base.Height - dstY0;
-      dstY1 = dstRb->Base.Height - dstY1;
+         dstY0 = dstRb->Base.Height - dstY0;
+         dstY1 = dstRb->Base.Height - dstY1;
+      }
 
       util_blit_pixels(st->blit,
                        srcSurf, srcX0, srcY0, srcX1, srcY1,
@@ -98,22 +101,9 @@ st_BlitFramebuffer(GLcontext *ctx,
 
    }
 
-#if 0
-   /* XXX is this sufficient? */
-   st_invalidate_state(ctx, _NEW_COLOR | _NEW_TEXTURE);
-#else
-   /* need to "unset" cso state because we went behind the back of the cso
-    * tracker.  Without unset, the _set_ calls would be no-ops.
-    */
-   cso_unset_blend(st->cso_context);
-   cso_unset_depth_stencil_alpha(st->cso_context);
-   cso_unset_rasterizer(st->cso_context);
-   cso_set_blend(st->cso_context, &st->state.blend);
-   cso_set_depth_stencil_alpha(st->cso_context, &st->state.depth_stencil);
-   cso_set_rasterizer(st->cso_context, &st->state.rasterizer);
+   /* shaders don't go through CSO yet */
    pipe->bind_fs_state(pipe, st->fp->driver_shader);
    pipe->bind_vs_state(pipe, st->vp->driver_shader);
-#endif
 }
 
 
diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c
index 693cddedf7..5ca15df602 100644
--- a/src/mesa/state_tracker/st_cb_clear.c
+++ b/src/mesa/state_tracker/st_cb_clear.c
@@ -49,14 +49,37 @@
 #include "pipe/p_defines.h"
 #include "pipe/p_winsys.h"
 #include "util/u_pack_color.h"
+#include "util/u_simple_shaders.h"
+#include "util/u_draw_quad.h"
 
 #include "cso_cache/cso_context.h"
 
 
 /* XXX for testing draw module vertex passthrough: */
+/* XXX this hack is broken now */
 #define TEST_DRAW_PASSTHROUGH 0
 
 
+void
+st_destroy_clear(struct st_context *st)
+{
+   struct pipe_context *pipe = st->pipe;
+
+   if (st->clear.fs) {
+      pipe->delete_fs_state(pipe, st->clear.fs);
+      st->clear.fs = NULL;
+   }
+   if (st->clear.vs) {
+      pipe->delete_vs_state(pipe, st->clear.vs);
+      st->clear.vs = NULL;
+   }
+   if (st->clear.vbuf) {
+      pipe->winsys->buffer_destroy(pipe->winsys, st->clear.vbuf);
+      st->clear.vbuf = NULL;
+   }
+}
+
+
 static GLboolean
 is_depth_stencil_format(enum pipe_format pipeFormat)
 {
@@ -72,104 +95,6 @@ is_depth_stencil_format(enum pipe_format pipeFormat)
 
 
 /**
- * Create a simple fragment shader that just passes through the fragment color.
- */
-static struct st_fragment_program *
-make_frag_shader(struct st_context *st)
-{
-   GLcontext *ctx = st->ctx;
-   struct st_fragment_program *stfp;
-   struct gl_program *p;
-   GLuint interpMode[16];
-   GLuint i;
-
-   /* XXX temporary */
-   for (i = 0; i < 16; i++)
-      interpMode[i] = TGSI_INTERPOLATE_LINEAR;
-
-   p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
-   if (!p)
-      return NULL;
-
-   p->NumInstructions = 2;
-   p->Instructions = _mesa_alloc_instructions(2);
-   if (!p->Instructions) {
-      ctx->Driver.DeleteProgram(ctx, p);
-      return NULL;
-   }
-   _mesa_init_instructions(p->Instructions, 2);
-   /* MOV result.color, fragment.color; */
-   p->Instructions[0].Opcode = OPCODE_MOV;
-   p->Instructions[0].DstReg.File = PROGRAM_OUTPUT;
-   p->Instructions[0].DstReg.Index = FRAG_RESULT_COLR;
-   p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT;
-   p->Instructions[0].SrcReg[0].Index = FRAG_ATTRIB_COL0;
-   /* END; */
-   p->Instructions[1].Opcode = OPCODE_END;
-
-   p->InputsRead = FRAG_BIT_COL0;
-   p->OutputsWritten = (1 << FRAG_RESULT_COLR);
-
-   stfp = (struct st_fragment_program *) p;
-   st_translate_fragment_program(st, stfp, NULL);
-
-   return stfp;
-}
-
-
-/**
- * Create a simple vertex shader that just passes through the
- * vertex position and color.
- */
-static struct st_vertex_program *
-make_vertex_shader(struct st_context *st)
-{
-   GLcontext *ctx = st->ctx;
-   struct st_vertex_program *stvp;
-   struct gl_program *p;
-
-   p = ctx->Driver.NewProgram(ctx, GL_VERTEX_PROGRAM_ARB, 0);
-   if (!p)
-      return NULL;
-
-   p->NumInstructions = 3;
-   p->Instructions = _mesa_alloc_instructions(3);
-   if (!p->Instructions) {
-      ctx->Driver.DeleteProgram(ctx, p);
-      return NULL;
-   }
-   _mesa_init_instructions(p->Instructions, 3);
-   /* MOV result.pos, vertex.pos; */
-   p->Instructions[0].Opcode = OPCODE_MOV;
-   p->Instructions[0].DstReg.File = PROGRAM_OUTPUT;
-   p->Instructions[0].DstReg.Index = VERT_RESULT_HPOS;
-   p->Instructions[0].SrcReg[0].File = PROGRAM_INPUT;
-   p->Instructions[0].SrcReg[0].Index = VERT_ATTRIB_POS;
-   /* MOV result.color, vertex.color; */
-   p->Instructions[1].Opcode = OPCODE_MOV;
-   p->Instructions[1].DstReg.File = PROGRAM_OUTPUT;
-   p->Instructions[1].DstReg.Index = VERT_RESULT_COL0;
-   p->Instructions[1].SrcReg[0].File = PROGRAM_INPUT;
-   p->Instructions[1].SrcReg[0].Index = VERT_ATTRIB_COLOR0;
-   /* END; */
-   p->Instructions[2].Opcode = OPCODE_END;
-
-   p->InputsRead = VERT_BIT_POS | VERT_BIT_COLOR0;
-   p->OutputsWritten = ((1 << VERT_RESULT_COL0) |
-                        (1 << VERT_RESULT_HPOS));
-
-   stvp = (struct st_vertex_program *) p;
-   st_translate_vertex_program(st, stvp, NULL);
-#if 0
-   assert(stvp->cso);
-#endif
-
-   return stvp;
-}
-
-
-
-/**
  * Draw a screen-aligned quadrilateral.
  * Coords are window coords with y=0=bottom.  These coords will be transformed
  * by the vertex shader and viewport transform (which will flip Y if needed).
@@ -179,45 +104,51 @@ draw_quad(GLcontext *ctx,
           float x0, float y0, float x1, float y1, GLfloat z,
           const GLfloat color[4])
 {
-   GLfloat verts[4][2][4]; /* four verts, two attribs, XYZW */
+   struct st_context *st = ctx->st;
+   struct pipe_context *pipe = st->pipe;
    GLuint i;
+   void *buf;
 
-#if TEST_DRAW_PASSTHROUGH
-   /* invert Y coords (may be off by one pixel) */
-   y0 = ctx->DrawBuffer->Height - y0;
-   y1 = ctx->DrawBuffer->Height - y1;
-#endif
+   if (!st->clear.vbuf) {
+      st->clear.vbuf = pipe->winsys->buffer_create(pipe->winsys, 32,
+                                                   PIPE_BUFFER_USAGE_VERTEX,
+                                                   sizeof(st->clear.vertices));
+   }
 
    /* positions */
-   verts[0][0][0] = x0;
-   verts[0][0][1] = y0;
+   st->clear.vertices[0][0][0] = x0;
+   st->clear.vertices[0][0][1] = y0;
 
-   verts[1][0][0] = x1;
-   verts[1][0][1] = y0;
+   st->clear.vertices[1][0][0] = x1;
+   st->clear.vertices[1][0][1] = y0;
 
-   verts[2][0][0] = x1;
-   verts[2][0][1] = y1;
+   st->clear.vertices[2][0][0] = x1;
+   st->clear.vertices[2][0][1] = y1;
 
-   verts[3][0][0] = x0;
-   verts[3][0][1] = y1;
+   st->clear.vertices[3][0][0] = x0;
+   st->clear.vertices[3][0][1] = y1;
 
    /* same for all verts: */
    for (i = 0; i < 4; i++) {
-      verts[i][0][2] = z;
-      verts[i][0][3] = 1.0;
-      verts[i][1][0] = color[0];
-      verts[i][1][1] = color[1];
-      verts[i][1][2] = color[2];
-      verts[i][1][3] = color[3];
+      st->clear.vertices[i][0][2] = z;
+      st->clear.vertices[i][0][3] = 1.0;
+      st->clear.vertices[i][1][0] = color[0];
+      st->clear.vertices[i][1][1] = color[1];
+      st->clear.vertices[i][1][2] = color[2];
+      st->clear.vertices[i][1][3] = color[3];
    }
 
-   st_draw_vertices(ctx, PIPE_PRIM_POLYGON, 4, (float *) verts, 2,
-#if TEST_DRAW_PASSTHROUGH
-                    GL_TRUE
-#else
-                    GL_FALSE
-#endif
-                    );
+   /* put vertex data into vbuf */
+   buf = pipe->winsys->buffer_map(pipe->winsys, st->clear.vbuf,
+                                  PIPE_BUFFER_USAGE_CPU_WRITE);
+   memcpy(buf, st->clear.vertices, sizeof(st->clear.vertices));
+   pipe->winsys->buffer_unmap(pipe->winsys, st->clear.vbuf);
+
+   /* draw */
+   util_draw_vertex_buffer(pipe, st->clear.vbuf,
+                           PIPE_PRIM_TRIANGLE_FAN,
+                           4,  /* verts */
+                           2); /* attribs/vert */
 }
 
 
@@ -234,9 +165,17 @@ clear_with_quad(GLcontext *ctx,
    struct st_context *st = ctx->st;
    struct pipe_context *pipe = st->pipe;
    const GLfloat x0 = ctx->DrawBuffer->_Xmin;
-   const GLfloat y0 = ctx->DrawBuffer->_Ymin;
    const GLfloat x1 = ctx->DrawBuffer->_Xmax;
-   const GLfloat y1 = ctx->DrawBuffer->_Ymax;
+   GLfloat y0, y1;
+
+   if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+      y0 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymax;
+      y1 = ctx->DrawBuffer->Height - ctx->DrawBuffer->_Ymin;
+   }
+   else {
+      y0 = ctx->DrawBuffer->_Ymin;
+      y1 = ctx->DrawBuffer->_Ymax;
+   }
 
    /*
    printf("%s %s%s%s %f,%f %f,%f\n", __FUNCTION__, 
@@ -247,10 +186,19 @@ clear_with_quad(GLcontext *ctx,
 	  x1, y1);
    */
 
+   cso_save_blend(st->cso_context);
+   cso_save_depth_stencil_alpha(st->cso_context);
+   cso_save_rasterizer(st->cso_context);
+   cso_save_viewport(st->cso_context);
+
    /* blend state: RGBA masking */
    {
       struct pipe_blend_state blend;
       memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
       if (color) {
          if (ctx->Color.ColorMask[0])
             blend.colormask |= PIPE_MASK_R;
@@ -290,42 +238,35 @@ clear_with_quad(GLcontext *ctx,
       cso_set_depth_stencil_alpha(st->cso_context, &depth_stencil);
    }
 
-   /* rasterizer state: nothing */
+   /* rasterizer state: bypass clipping */
    {
       struct pipe_rasterizer_state raster;
       memset(&raster, 0, sizeof(raster));
-#if 0
-      /* don't do per-pixel scissor; we'll just draw a PIPE_PRIM_QUAD
-       * that matches the scissor bounds.
-       */
-      if (ctx->Scissor.Enabled)
-         raster.scissor = 1;
-#endif
-#if TEST_DRAW_PASSTHROUGH
       raster.bypass_clipping = 1;
+#if TEST_DRAW_PASSTHROUGH
       raster.bypass_vs = 1;
 #endif
       cso_set_rasterizer(st->cso_context, &raster);
    }
 
    /* fragment shader state: color pass-through program */
-   {
-      static struct st_fragment_program *stfp = NULL;
-      if (!stfp) {
-         stfp = make_frag_shader(st);
-      }
-      pipe->bind_fs_state(pipe, stfp->driver_shader);
+   if (!st->clear.fs) {
+      st->clear.fs = util_make_fragment_passthrough_shader(pipe);
    }
+   pipe->bind_fs_state(pipe, st->clear.fs);
+
 
 #if !TEST_DRAW_PASSTHROUGH
    /* vertex shader state: color/position pass-through */
-   {
-      static struct st_vertex_program *stvp = NULL;
-      if (!stvp) {
-         stvp = make_vertex_shader(st);
-      }
-      pipe->bind_vs_state(pipe, stvp->driver_shader);
+   if (!st->clear.vs) {
+      const uint semantic_names[] = { TGSI_SEMANTIC_POSITION,
+                                      TGSI_SEMANTIC_COLOR };
+      const uint semantic_indexes[] = { 0, 0 };
+      st->clear.vs = util_make_vertex_passthrough_shader(pipe, 2,
+                                                         semantic_names,
+                                                         semantic_indexes);
    }
+   pipe->bind_vs_state(pipe, st->clear.vs);
 #endif
 
 #if !TEST_DRAW_PASSTHROUGH
@@ -342,28 +283,21 @@ clear_with_quad(GLcontext *ctx,
       vp.translate[1] = 0.5 * height;
       vp.translate[2] = 0.0;
       vp.translate[3] = 0.0;
-      pipe->set_viewport_state(pipe, &vp);
+      cso_set_viewport(st->cso_context, &vp);
    }
 #endif
 
    /* draw quad matching scissor rect (XXX verify coord round-off) */
    draw_quad(ctx, x0, y0, x1, y1, ctx->Depth.Clear, ctx->Color.ClearColor);
 
-#if 0
-   /* Can't depend on old state objects still existing -- may have
-    * been deleted to make room in the hash, etc.  (Should get
-    * fixed...)
-    */
-   st_invalidate_state(ctx, _NEW_COLOR | _NEW_DEPTH | _NEW_STENCIL);
-#else
    /* Restore pipe state */
-   cso_set_blend(st->cso_context, &st->state.blend);
-   cso_set_depth_stencil_alpha(st->cso_context, &st->state.depth_stencil);
-   cso_set_rasterizer(st->cso_context, &st->state.rasterizer);
+   cso_restore_blend(st->cso_context);
+   cso_restore_depth_stencil_alpha(st->cso_context);
+   cso_restore_rasterizer(st->cso_context);
+   cso_restore_viewport(st->cso_context);
+   /* these don't go through cso yet */
    pipe->bind_fs_state(pipe, st->fp->driver_shader);
    pipe->bind_vs_state(pipe, st->vp->driver_shader);
-#endif
-   pipe->set_viewport_state(pipe, &st->state.viewport);
 }
 
 
diff --git a/src/mesa/state_tracker/st_cb_clear.h b/src/mesa/state_tracker/st_cb_clear.h
index c715e56bd5..dfa4033faa 100644
--- a/src/mesa/state_tracker/st_cb_clear.h
+++ b/src/mesa/state_tracker/st_cb_clear.h
@@ -31,6 +31,10 @@
 
 
 extern void
+st_destroy_clear(struct st_context *st);
+
+
+extern void
 st_init_clear_functions(struct dd_function_table *functions);
 
 
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index 18ec9645c4..2ebaf8a2c3 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -32,6 +32,7 @@
 
 #include "main/imports.h"
 #include "main/image.h"
+#include "main/bufferobj.h"
 #include "main/macros.h"
 #include "main/texformat.h"
 #include "shader/program.h"
@@ -56,6 +57,7 @@
 #include "pipe/p_inlines.h"
 #include "pipe/p_winsys.h"
 #include "util/p_tile.h"
+#include "util/u_draw_quad.h"
 #include "shader/prog_instruction.h"
 #include "cso_cache/cso_context.h"
 
@@ -85,140 +87,6 @@ is_passthrough_program(const struct gl_fragment_program *prog)
 }
 
 
-/**
- * Make fragment program for glBitmap:
- *   Sample the texture and kill the fragment if the bit is 0.
- * This program will be combined with the user's fragment program.
- */
-static struct st_fragment_program *
-make_bitmap_fragment_program(GLcontext *ctx)
-{
-   struct st_fragment_program *stfp;
-   struct gl_program *p;
-   GLuint ic = 0;
-
-   p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
-   if (!p)
-      return NULL;
-
-   p->NumInstructions = 5;
-
-   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
-   if (!p->Instructions) {
-      ctx->Driver.DeleteProgram(ctx, p);
-      return NULL;
-   }
-   _mesa_init_instructions(p->Instructions, p->NumInstructions);
-
-   /* TEX tmp0, fragment.texcoord[0], texture[0], 2D; */
-   p->Instructions[ic].Opcode = OPCODE_TEX;
-   p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY;
-   p->Instructions[ic].DstReg.Index = 0;
-   p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
-   p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
-   p->Instructions[ic].TexSrcUnit = 0;
-   p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
-   ic++;
-
-   /* SWZ tmp0.x, tmp0.x, 1111; # tmp0.x = 1.0 */
-   p->Instructions[ic].Opcode = OPCODE_SWZ;
-   p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY;
-   p->Instructions[ic].DstReg.Index = 0;
-   p->Instructions[ic].DstReg.WriteMask = WRITEMASK_X;
-   p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
-   p->Instructions[ic].SrcReg[0].Index = 0;
-   p->Instructions[ic].SrcReg[0].Swizzle
-      = MAKE_SWIZZLE4(SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE, SWIZZLE_ONE );
-   ic++;
-
-   /* SUB tmp0, tmp0.wwww, tmp0.xxxx;  #  tmp0.w -= 1 */
-   p->Instructions[ic].Opcode = OPCODE_SUB;
-   p->Instructions[ic].DstReg.File = PROGRAM_TEMPORARY;
-   p->Instructions[ic].DstReg.Index = 0;
-   p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
-   p->Instructions[ic].SrcReg[0].Index = 0;
-   p->Instructions[ic].SrcReg[0].Swizzle = SWIZZLE_WWWW;
-   p->Instructions[ic].SrcReg[1].File = PROGRAM_TEMPORARY;
-   p->Instructions[ic].SrcReg[1].Index = 0;
-   p->Instructions[ic].SrcReg[1].Swizzle = SWIZZLE_XXXX; /* 1.0 */
-   ic++;
-
-   /* KIL if tmp0 < 0 */
-   p->Instructions[ic].Opcode = OPCODE_KIL;
-   p->Instructions[ic].SrcReg[0].File = PROGRAM_TEMPORARY;
-   p->Instructions[ic].SrcReg[0].Index = 0;
-   ic++;
-
-   /* END; */
-   p->Instructions[ic++].Opcode = OPCODE_END;
-
-   assert(ic == p->NumInstructions);
-
-   p->InputsRead = FRAG_BIT_TEX0;
-   p->OutputsWritten = 0x0;
-
-   stfp = (struct st_fragment_program *) p;
-   stfp->Base.UsesKill = GL_TRUE;
-   st_translate_fragment_program(ctx->st, stfp, NULL);
-
-   return stfp;
-}
-
-
-/**
- * Combine basic bitmap fragment program with the user-defined program.
- */
-static struct st_fragment_program *
-combined_bitmap_fragment_program(GLcontext *ctx)
-{
-   struct st_context *st = ctx->st;
-   struct st_fragment_program *stfp;
-
-   if (!st->bitmap.program) {
-      /* create the basic bitmap fragment program */
-      st->bitmap.program = make_bitmap_fragment_program(ctx);
-   }
-
-   if (st->bitmap.user_prog_sn == st->fp->serialNo) {
-      /* re-use */
-      stfp = st->bitmap.combined_prog;
-   }
-   else {
-      /* Concatenate the bitmap program with the current user-defined program.
-       */
-      stfp = (struct st_fragment_program *)
-         _mesa_combine_programs(ctx,
-                                &st->bitmap.program->Base.Base,
-                                &st->fp->Base.Base);
-
-#if 0
-      {
-         struct gl_program *p = &stfp->Base.Base;
-         printf("Combined bitmap program:\n");
-         _mesa_print_program(p);
-         printf("InputsRead: 0x%x\n", p->InputsRead);
-         printf("OutputsWritten: 0x%x\n", p->OutputsWritten);
-         _mesa_print_parameter_list(p->Parameters);
-      }
-#endif
-
-      /* translate to TGSI tokens */
-      st_translate_fragment_program(st, stfp, NULL);
-
-      /* save new program, update serial numbers */
-      st->bitmap.user_prog_sn = st->fp->serialNo;
-      st->bitmap.combined_prog = stfp;
-   }
-
-   /* Ideally we'd have updated the pipe constants during the normal
-    * st/atom mechanism.  But we can't since this is specific to glBitmap.
-    */
-   st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT);
-
-   return stfp;
-}
-
-
 
 /**
  * Make fragment shader for glDraw/CopyPixels.  This shader is made
@@ -351,7 +219,7 @@ make_fragment_shader_z(struct st_context *st)
  * Create a simple vertex shader that just passes through the
  * vertex position and texcoord (and optionally, color).
  */
-struct st_vertex_program *
+static struct st_vertex_program *
 st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor)
 {
    /* only make programs once and re-use */
@@ -467,16 +335,15 @@ make_texture(struct st_context *st,
    assert(pipeFormat);
    cpp = st_sizeof_format(pipeFormat);
 
+   pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels);
+   if (!pixels)
+      return NULL;
+
    pt = st_texture_create(st, PIPE_TEXTURE_2D, pipeFormat, 0, width, height,
 			  1, 0);
-   if (!pt)
+   if (!pt) {
+      _mesa_unmap_drapix_pbo(ctx, unpack);
       return NULL;
-
-   if (unpack->BufferObj && unpack->BufferObj->Name) {
-      /*
-      pt->region = buffer_object_region(unpack->BufferObj);
-      */
-      printf("st_DrawPixels (sourcing from PBO not implemented yet)\n");
    }
 
    {
@@ -513,12 +380,16 @@ make_texture(struct st_context *st,
       /* unmap */
       pipe_surface_unmap(surface);
       pipe_surface_reference(&surface, NULL);
+      pipe->texture_update(pipe, pt, 0, 0x1);
+
       assert(success);
 
       /* restore */
       ctx->_ImageTransferState = imageTransferStateSave;
    }
 
+   _mesa_unmap_drapix_pbo(ctx, unpack);
+
    return pt;
 }
 
@@ -650,14 +521,17 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
    assert(width <= maxSize);
    assert(height <= maxSize);
 
-   /* setup state: just scissor */
+   cso_save_rasterizer(cso);
+   cso_save_viewport(cso);
+
+   /* rasterizer state: just scissor */
    {
-      struct pipe_rasterizer_state  setup;
-      memset(&setup, 0, sizeof(setup));
+      struct pipe_rasterizer_state rasterizer;
+      memset(&rasterizer, 0, sizeof(rasterizer));
       if (ctx->Scissor.Enabled)
-         setup.scissor = 1;
+         rasterizer.scissor = 1;
 
-      cso_set_rasterizer(cso, &setup);
+      cso_set_rasterizer(cso, &rasterizer);
    }
 
    /* fragment shader state: TEX lookup program */
@@ -696,7 +570,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
       vp.translate[1] = 0.5 * height;
       vp.translate[2] = 0.0;
       vp.translate[3] = 0.0;
-      pipe->set_viewport_state(pipe, &vp);
+      cso_set_viewport(cso, &vp);
    }
 
    /* texture state: */
@@ -719,26 +593,18 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
    else
       draw_quad(ctx, x0, y0, z, x1, y1, invertTex);
 
-   /* restore GL state */
-   pipe->set_sampler_textures(pipe, ctx->st->state.num_textures,
-                              ctx->st->state.sampler_texture);
-
-   pipe->set_viewport_state(pipe, &ctx->st->state.viewport);
-
-#if 0
-   /* Can't depend on old state objects still existing -- may have
-    * been deleted to make room in the hash, etc.  (Should get
-    * fixed...)
-    */
-   st_invalidate_state(ctx, _NEW_COLOR | _NEW_TEXTURE);
-#else
    /* restore state */
+   cso_restore_rasterizer(cso);
+   cso_restore_viewport(cso);
+   /* shaders don't go through cso yet */
    pipe->bind_fs_state(pipe, st->fp->driver_shader);
    pipe->bind_vs_state(pipe, st->vp->driver_shader);
+
    cso_set_rasterizer(cso, &st->state.rasterizer);
    cso_set_samplers(cso, PIPE_MAX_SAMPLERS,
                  (const struct pipe_sampler_state **) st->state.sampler_list);
-#endif
+   pipe->set_sampler_textures(pipe, ctx->st->state.num_textures,
+                              ctx->st->state.sampler_texture);
 }
 
 
@@ -972,9 +838,9 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
 
    bufferFormat = ps->format;
 
-   if (any_fragment_ops(st) ||
+   if (1/*any_fragment_ops(st) ||
        any_pixel_transfer_ops(st) ||
-       !compatible_formats(format, type, ps->format)) {
+       !compatible_formats(format, type, ps->format)*/) {
       /* textured quad */
       struct pipe_texture *pt
          = make_texture(ctx->st, width, height, format, type, unpack, pixels);
@@ -993,152 +859,6 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
 
 
 
-/**
- * Create a texture which represents a bitmap image.
- */
-static struct pipe_texture *
-make_bitmap_texture(GLcontext *ctx, GLsizei width, GLsizei height,
-                    const struct gl_pixelstore_attrib *unpack,
-                    const GLubyte *bitmap)
-{
-   struct pipe_context *pipe = ctx->st->pipe;
-   struct pipe_screen *screen = pipe->screen;
-   struct pipe_surface *surface;
-   uint format = 0, cpp, comp;
-   ubyte *dest;
-   struct pipe_texture *pt;
-   int row, col;
-
-   /* find a texture format we know */
-   if (screen->is_format_supported( screen, PIPE_FORMAT_U_I8, PIPE_TEXTURE )) {
-      format = PIPE_FORMAT_U_I8;
-      cpp = 1;
-      comp = 0;
-   }
-   else if (screen->is_format_supported( screen, PIPE_FORMAT_A8R8G8B8_UNORM, PIPE_TEXTURE )) {
-      format = PIPE_FORMAT_A8R8G8B8_UNORM;
-      cpp = 4;
-      comp = 3; /* alpha channel */ /*XXX little-endian dependency */
-   }
-   else {
-      /* XXX support more formats */
-      assert( 0 );
-   }
-
-   /**
-    * Create a texture.
-    */
-   pt = st_texture_create(ctx->st, PIPE_TEXTURE_2D, format, 0, width, height,
-			  1, 0);
-   if (!pt)
-      return NULL;
-
-   if (unpack->BufferObj && unpack->BufferObj->Name) {
-      /*
-      pt->region = buffer_object_region(unpack->BufferObj);
-      */
-      printf("st_Bitmap (sourcing from PBO not implemented yet)\n");
-   }
-
-   surface = screen->get_tex_surface(screen, pt, 0, 0, 0);
-
-   /* map texture surface */
-   dest = pipe_surface_map(surface);
-
-   /* Put image into texture surface.
-    * Note that the image is actually going to be upside down in
-    * the texture.  We deal with that with texcoords.
-    */
-
-   for (row = 0; row < height; row++) {
-      const GLubyte *src = (const GLubyte *) _mesa_image_address2d(unpack,
-                 bitmap, width, height, GL_COLOR_INDEX, GL_BITMAP, row, 0);
-      ubyte *destRow = dest + row * surface->pitch * cpp;
-
-      if (unpack->LsbFirst) {
-         /* Lsb first */
-         GLubyte mask = 1U << (unpack->SkipPixels & 0x7);
-         for (col = 0; col < width; col++) {
-
-            /* set texel to 255 if bit is set */
-            destRow[comp] = (*src & mask) ? 255 : 0;
-            destRow += cpp;
-
-            if (mask == 128U) {
-               src++;
-               mask = 1U;
-            }
-            else {
-               mask = mask << 1;
-            }
-         }
-
-         /* get ready for next row */
-         if (mask != 1)
-            src++;
-      }
-      else {
-         /* Msb first */
-         GLubyte mask = 128U >> (unpack->SkipPixels & 0x7);
-         for (col = 0; col < width; col++) {
-
-            /* set texel to 255 if bit is set */
-            destRow[comp] =(*src & mask) ? 255 : 0;
-            destRow += cpp;
-
-            if (mask == 1U) {
-               src++;
-               mask = 128U;
-            }
-            else {
-               mask = mask >> 1;
-            }
-         }
-
-         /* get ready for next row */
-         if (mask != 128)
-            src++;
-      }
-
-   } /* row */
-
-   /* Release surface */
-   pipe_surface_unmap(surface);
-   pipe_surface_reference(&surface, NULL);
-
-   pt->format = format;
-
-   return pt;
-}
-
-
-
-static void
-st_Bitmap(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
-          const struct gl_pixelstore_attrib *unpack, const GLubyte *bitmap )
-{
-   struct st_fragment_program *stfp;
-   struct st_vertex_program *stvp;
-   struct st_context *st = ctx->st;
-   struct pipe_texture *pt;
-
-   stvp = st_make_passthrough_vertex_shader(ctx->st, GL_TRUE);
-   stfp = combined_bitmap_fragment_program(ctx);
-
-   st_validate_state(st);
-
-   pt = make_bitmap_texture(ctx, width, height, unpack, bitmap);
-   if (pt) {
-      draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
-                         width, height, 1.0, 1.0,
-                         pt, stvp, stfp,
-                         ctx->Current.RasterColor, GL_FALSE);
-
-      pipe_texture_reference(&pt, NULL);
-   }
-}
-
-
 static void
 copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
                     GLsizei width, GLsizei height,
@@ -1339,5 +1059,4 @@ void st_init_drawpixels_functions(struct dd_function_table *functions)
 {
    functions->DrawPixels = st_DrawPixels;
    functions->CopyPixels = st_CopyPixels;
-   functions->Bitmap = st_Bitmap;
 }
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.h b/src/mesa/state_tracker/st_cb_drawpixels.h
index b8b906f06b..71ba487020 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.h
+++ b/src/mesa/state_tracker/st_cb_drawpixels.h
@@ -30,10 +30,6 @@
 #define ST_CB_DRAWPIXELS_H
 
 
-extern struct st_vertex_program *
-st_make_passthrough_vertex_shader(struct st_context *st, GLboolean passColor);
-
-
 extern void st_init_drawpixels_functions(struct dd_function_table *functions);
 
 
diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c
index 0a62e20e0c..a739fcd336 100644
--- a/src/mesa/state_tracker/st_cb_program.c
+++ b/src/mesa/state_tracker/st_cb_program.c
@@ -126,19 +126,39 @@ static void st_delete_program( GLcontext *ctx,
 			       struct gl_program *prog )
 {
    struct st_context *st = st_context(ctx);
+   struct pipe_context *pipe = st->pipe;
 
    switch( prog->Target ) {
    case GL_VERTEX_PROGRAM_ARB:
       {
          struct st_vertex_program *stvp = (struct st_vertex_program *) prog;
-         st_remove_vertex_program(st, stvp);
+
+         if (stvp->driver_shader) {
+            pipe->delete_vs_state(pipe, stvp->driver_shader);
+            stvp->driver_shader = NULL;
+         }
+
+         if (stvp->state.tokens) {
+            FREE((void *) stvp->state.tokens);
+            stvp->state.tokens = NULL;
+         }
       }
       break;
    case GL_FRAGMENT_PROGRAM_ARB:
       {
-         struct st_fragment_program *stfp
-            = (struct st_fragment_program *) prog;
-         st_remove_fragment_program(st, stfp);
+         struct st_fragment_program *stfp = (struct st_fragment_program *) prog;
+
+         if (stfp->driver_shader) {
+            pipe->delete_fs_state(pipe, stfp->driver_shader);
+            stfp->driver_shader = NULL;
+         }
+         
+         if (stfp->state.tokens) {
+            FREE((void *) stfp->state.tokens);
+            stfp->state.tokens = NULL;
+         }
+
+         st_free_translated_vertex_programs(st, stfp->vertex_programs);
       }
       break;
    default:
@@ -162,27 +182,25 @@ static void st_program_string_notify( GLcontext *ctx,
 				      struct gl_program *prog )
 {
    struct st_context *st = st_context(ctx);
+   struct pipe_context *pipe = st->pipe;
 
    if (target == GL_FRAGMENT_PROGRAM_ARB) {
       struct st_fragment_program *stfp = (struct st_fragment_program *) prog;
 
       stfp->serialNo++;
 
-#if 0
-      if (stfp->cso) {
-         /* free the TGSI code */
-         // cso_delete(stfp->vs);
-         stfp->cso = NULL;
+      if (stfp->driver_shader) {
+         pipe->delete_fs_state(pipe, stfp->driver_shader);
+         stfp->driver_shader = NULL;
       }
-#endif
-
-      stfp->param_state = stfp->Base.Base.Parameters->StateFlags;
 
       if (stfp->state.tokens) {
          FREE((void *) stfp->state.tokens);
          stfp->state.tokens = NULL;
       }
 
+      stfp->param_state = stfp->Base.Base.Parameters->StateFlags;
+
       if (st->fp == stfp)
 	 st->dirty.st |= ST_NEW_FRAGMENT_PROGRAM;
    }
@@ -191,26 +209,23 @@ static void st_program_string_notify( GLcontext *ctx,
 
       stvp->serialNo++;
 
-#if 0
-      if (stvp->cso) {
-         /* free the CSO data */
-         st->pipe->delete_vs_state(st->pipe, stvp->cso->data);
-         FREE((void *) stvp->cso);
-         stvp->cso = NULL;
+      if (stvp->driver_shader) {
+         pipe->delete_vs_state(pipe, stvp->driver_shader);
+         stvp->driver_shader = NULL;
       }
-#endif
+
       if (stvp->draw_shader) {
          draw_delete_vertex_shader(st->draw, stvp->draw_shader);
          stvp->draw_shader = NULL;
       }
 
-      stvp->param_state = stvp->Base.Base.Parameters->StateFlags;
-
       if (stvp->state.tokens) {
          FREE((void *) stvp->state.tokens);
          stvp->state.tokens = NULL;
       }
 
+      stvp->param_state = stvp->Base.Base.Parameters->StateFlags;
+
       if (st->vp == stvp)
 	 st->dirty.st |= ST_NEW_VERTEX_PROGRAM;
    }
diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c
index c89c74229e..4cf9adcd28 100644
--- a/src/mesa/state_tracker/st_cb_readpixels.c
+++ b/src/mesa/state_tracker/st_cb_readpixels.c
@@ -34,6 +34,7 @@
 
 
 #include "main/imports.h"
+#include "main/bufferobj.h"
 #include "main/context.h"
 #include "main/image.h"
 
@@ -126,7 +127,6 @@ st_read_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
 }
 
 
-
 /**
  * Do glReadPixels by getting rows from the framebuffer surface with
  * get_tile().  Convert to requested format/type with Mesa image routines.
@@ -155,18 +155,13 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
       return;
    }
 
+   dest = _mesa_map_readpix_pbo(ctx, &clippedPacking, dest);
+   if (!dest)
+      return;
+
    /* make sure rendering has completed */
    pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE);
 
-   if (pack->BufferObj && pack->BufferObj->Name) {
-      /* reading into a PBO */
-
-   }
-   else {
-      /* reading into user memory/buffer */
-
-   }
-
    if (format == GL_STENCIL_INDEX) {
       st_read_stencil_pixels(ctx, x, y, width, height, type, pack, dest);
       return;
@@ -285,6 +280,8 @@ st_readpixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
          }
       }
    }
+
+   _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
 }
 
 
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index d4731c7737..a6c1a35355 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -476,6 +476,7 @@ st_TexImage(GLcontext * ctx,
             struct gl_texture_image *texImage,
             GLsizei imageSize, int compressed)
 {
+   struct pipe_context *pipe = ctx->st->pipe;
    struct st_texture_object *stObj = st_texture_object(texObj);
    struct st_texture_image *stImage = st_texture_image(texImage);
    GLint postConvWidth, postConvHeight;
@@ -563,7 +564,8 @@ st_TexImage(GLcontext * ctx,
    if (!stObj->pt) {
       guess_and_alloc_texture(ctx->st, stObj, stImage);
       if (!stObj->pt) {
-	 DBG("guess_and_alloc_texture: failed\n");
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+         return;
       }
    }
 
@@ -690,8 +692,8 @@ st_TexImage(GLcontext * ctx,
       texImage->Data = NULL;
    }
 
-   /* flag data as dirty */
-   stObj->dirtyData = GL_TRUE;
+   if (stObj->pt)
+      pipe->texture_update(pipe, stObj->pt, stImage->face, (1 << level));
 
    if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
       ctx->Driver.GenerateMipmap(ctx, target, texObj);
@@ -866,6 +868,7 @@ st_TexSubimage(GLcontext * ctx,
                  struct gl_texture_object *texObj,
                  struct gl_texture_image *texImage)
 {
+   struct pipe_context *pipe = ctx->st->pipe;
    struct st_texture_object *stObj = st_texture_object(texObj);
    struct st_texture_image *stImage = st_texture_image(texImage);
    GLuint dstRowStride;
@@ -924,8 +927,7 @@ st_TexSubimage(GLcontext * ctx,
       texImage->Data = NULL;
    }
 
-   /* flag data as dirty */
-   stObj->dirtyData = GL_TRUE;
+   pipe->texture_update(pipe, stObj->pt, stImage->face, (1 << level));
 }
 
 
@@ -1179,8 +1181,7 @@ do_copy_texsubimage(GLcontext *ctx,
 
    pipe_surface_reference(&dest_surface, NULL);
 
-   /* flag data as dirty */
-   stObj->dirtyData = GL_TRUE;
+   pipe->texture_update(pipe, stObj->pt, stImage->face, (1 << level));
 
    if (level == texObj->BaseLevel && texObj->GenerateMipmap) {
       ctx->Driver.GenerateMipmap(ctx, target, texObj);
@@ -1379,7 +1380,7 @@ copy_image_data_to_texture(struct st_context *st,
 /**
  * Called during state validation.  When this function is finished,
  * the texture object should be ready for rendering.
- * \return GL_FALSE if a texture border is present, GL_TRUE otherwise
+ * \return GL_TRUE for success, GL_FALSE for failure (out of mem)
  */
 GLboolean
 st_finalize_texture(GLcontext *ctx,
@@ -1405,6 +1406,7 @@ st_finalize_texture(GLcontext *ctx,
    calculate_first_last_level(stObj);
    firstImage = st_texture_image(stObj->base.Image[0][stObj->base.BaseLevel]);
 
+#if 0
    /* Fallback case:
     */
    if (firstImage->base.Border) {
@@ -1413,7 +1415,7 @@ st_finalize_texture(GLcontext *ctx,
       }
       return GL_FALSE;
    }
-
+#endif
 
    /* If both firstImage and stObj point to a texture which can contain
     * all active images, favour firstImage.  Note that because of the
@@ -1466,6 +1468,10 @@ st_finalize_texture(GLcontext *ctx,
                                     firstImage->base.Height,
                                     firstImage->base.Depth,
                                     comp_byte);
+      if (!stObj->pt) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glTexImage");
+         return GL_FALSE;
+      }
    }
 
    /* Pull in any images not in the object's texture:
@@ -1481,11 +1487,11 @@ st_finalize_texture(GLcontext *ctx,
          if (stImage && stObj->pt != stImage->pt) {
             copy_image_data_to_texture(ctx->st, stObj, level, stImage);
 	    *needFlush = GL_TRUE;
+            pipe->texture_update(pipe, stObj->pt, face, (1 << level));
          }
       }
    }
 
-
    return GL_TRUE;
 }
 
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
index e1fc885e0e..d9e8722976 100644
--- a/src/mesa/state_tracker/st_context.c
+++ b/src/mesa/state_tracker/st_context.c
@@ -35,8 +35,9 @@
 #include "st_public.h"
 #include "st_context.h"
 #include "st_cb_accum.h"
-#include "st_cb_bufferobjects.h"
+#include "st_cb_bitmap.h"
 #include "st_cb_blit.h"
+#include "st_cb_bufferobjects.h"
 #include "st_cb_clear.h"
 #include "st_cb_drawpixels.h"
 #include "st_cb_fbo.h"
@@ -154,7 +155,9 @@ static void st_destroy_context_priv( struct st_context *st )
    st_destroy_atoms( st );
    st_destroy_draw( st );
    st_destroy_generate_mipmap(st);
+   st_destroy_bitmap(st);
    st_destroy_blit(st);
+   st_destroy_clear(st);
 
    _vbo_DestroyContext(st->ctx);
 
@@ -220,8 +223,9 @@ void st_init_driver_functions(struct dd_function_table *functions)
    _mesa_init_glsl_driver_functions(functions);
 
    st_init_accum_functions(functions);
-   st_init_bufferobject_functions(functions);
+   st_init_bitmap_functions(functions);
    st_init_blit_functions(functions);
+   st_init_bufferobject_functions(functions);
    st_init_clear_functions(functions);
    st_init_drawpixels_functions(functions);
    st_init_fbo_functions(functions);
diff --git a/src/mesa/state_tracker/st_context.h b/src/mesa/state_tracker/st_context.h
index 63150dbeaf..2d37086799 100644
--- a/src/mesa/state_tracker/st_context.h
+++ b/src/mesa/state_tracker/st_context.h
@@ -82,7 +82,6 @@ struct st_context
       struct pipe_rasterizer_state          rasterizer;
       struct pipe_sampler_state             samplers[PIPE_MAX_SAMPLERS];
       struct pipe_sampler_state             *sampler_list[PIPE_MAX_SAMPLERS];
-      struct pipe_blend_color  blend_color;
       struct pipe_clip_state clip;
       struct pipe_constant_buffer constants[2];
       struct pipe_framebuffer_state framebuffer;
@@ -143,12 +142,24 @@ struct st_context
       GLuint combined_prog_sn;
    } pixel_xfer;
 
+   /** for glBitmap */
    struct {
       struct st_fragment_program *program;  /**< bitmap tex/kil program */
       GLuint user_prog_sn;  /**< user fragment program serial no. */
       struct st_fragment_program *combined_prog;
+      void *vs;
+      float vertices[4][3][4];  /**< vertex pos + color + texcoord */
+      struct pipe_buffer *vbuf;
    } bitmap;
 
+   /** for glClear */
+   struct {
+      void *vs;
+      void *fs;
+      float vertices[4][2][4];  /**< vertex pos + color */
+      struct pipe_buffer *vbuf;
+   } clear;
+
    struct gen_mipmap_state *gen_mipmap;
    struct blit_state *blit;
 
diff --git a/src/mesa/state_tracker/st_debug.c b/src/mesa/state_tracker/st_debug.c
index 8b9b91aa34..23ecfff0aa 100644
--- a/src/mesa/state_tracker/st_debug.c
+++ b/src/mesa/state_tracker/st_debug.c
@@ -50,9 +50,10 @@ st_print_current(void)
 {
    GET_CURRENT_CONTEXT(ctx);
    struct st_context *st = ctx->st;
-   int i;
 
 #if 0
+   int i;
+
    printf("Vertex Transform Inputs:\n");
    for (i = 0; i < st->vp->state.num_inputs; i++) {
       printf("  Slot %d:  VERT_ATTRIB_%d\n", i, st->vp->index_to_input[i]);
diff --git a/src/mesa/state_tracker/st_draw.c b/src/mesa/state_tracker/st_draw.c
index 5ebd0bbcce..20af90df7d 100644
--- a/src/mesa/state_tracker/st_draw.c
+++ b/src/mesa/state_tracker/st_draw.c
@@ -223,7 +223,7 @@ st_draw_vbo(GLcontext *ctx,
    /* loop over TGSI shader inputs to determine vertex buffer
     * and attribute info
     */
-   for (attr = 0; attr < /*vs*/vp->num_inputs; attr++) {
+   for (attr = 0; attr < vp->num_inputs; attr++) {
       const GLuint mesaAttr = vp->index_to_input[attr];
       struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
       struct pipe_vertex_element velement;
@@ -333,7 +333,7 @@ st_draw_vbo(GLcontext *ctx,
    }
 
    /* unreference buffers (frees wrapped user-space buffer objects) */
-   for (attr = 0; attr < /*vs*/vp->num_inputs; attr++) {
+   for (attr = 0; attr < vp->num_inputs; attr++) {
       pipe_buffer_reference(winsys, &vbuffer[attr].buffer, NULL);
       assert(!vbuffer[attr].buffer);
       pipe->set_vertex_buffer(pipe, attr, &vbuffer[attr]);
@@ -502,7 +502,7 @@ st_feedback_draw_vbo(GLcontext *ctx,
    /* loop over TGSI shader inputs to determine vertex buffer
     * and attribute info
     */
-   for (attr = 0; attr < /*vs*/vp->num_inputs; attr++) {
+   for (attr = 0; attr < vp->num_inputs; attr++) {
       const GLuint mesaAttr = vp->index_to_input[attr];
       struct gl_buffer_object *bufobj = arrays[mesaAttr]->BufferObj;
       struct pipe_vertex_element velement;
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
index 0962b5f74c..7fab4d041d 100644
--- a/src/mesa/state_tracker/st_extensions.c
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -155,7 +155,6 @@ void st_init_extensions(struct st_context *st)
    ctx->Extensions.EXT_texture_env_combine = GL_TRUE;
    ctx->Extensions.EXT_texture_env_dot3 = GL_TRUE;
    ctx->Extensions.EXT_texture_lod_bias = GL_TRUE;
-   ctx->Extensions.EXT_texture_sRGB = GL_TRUE; /* XXX temp */
 
    ctx->Extensions.NV_blend_square = GL_TRUE;
    ctx->Extensions.NV_texgen_reflection = GL_TRUE;
@@ -190,10 +189,6 @@ void st_init_extensions(struct st_context *st)
       ctx->Extensions.ATI_separate_stencil = GL_TRUE;
    }
 
-   if (screen->get_param(screen, PIPE_CAP_S3TC)) {
-      ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE;
-   }
-
    if (screen->get_param(screen, PIPE_CAP_ANISOTROPIC_FILTER)) {
       ctx->Extensions.EXT_texture_filter_anisotropic = GL_TRUE;
    }
@@ -214,4 +209,14 @@ void st_init_extensions(struct st_context *st)
       /*ctx->Extensions.ARB_shadow_ambient = GL_TRUE;*/
    }
 
+   if (screen->is_format_supported(screen, PIPE_FORMAT_R8G8B8A8_SRGB,
+                                   PIPE_TEXTURE)) {
+      ctx->Extensions.EXT_texture_sRGB = GL_TRUE;
+   }
+
+   if (screen->is_format_supported(screen, PIPE_FORMAT_DXT5_RGBA,
+                                   PIPE_TEXTURE)) {
+      ctx->Extensions.EXT_texture_compression_s3tc = GL_TRUE;
+   }
+
 }
diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c
index 6c3afca1ba..61e1d9621c 100644
--- a/src/mesa/state_tracker/st_gen_mipmap.c
+++ b/src/mesa/state_tracker/st_gen_mipmap.c
@@ -59,7 +59,7 @@
 void
 st_init_generate_mipmap(struct st_context *st)
 {
-   st->gen_mipmap = util_create_gen_mipmap(st->pipe);
+   st->gen_mipmap = util_create_gen_mipmap(st->pipe, st->cso_context);
 }
 
 
@@ -94,19 +94,11 @@ st_render_mipmap(struct st_context *st,
 
    util_gen_mipmap(st->gen_mipmap, pt, face, baseLevel, lastLevel);
 
-   /* restore pipe state */
-#if 0
-   cso_set_rasterizer(st->cso_context, &st->state.rasterizer);
-   cso_set_samplers(st->cso_context, st->state.samplers_list);
-   pipe->bind_fs_state(pipe, st->fp->shader_program);
-   pipe->bind_vs_state(pipe, st->vp->shader_program);
-   pipe->set_sampler_textures(pipe, st->state.num_textures,
-                              st->state.sampler_texture);
-   pipe->set_viewport_state(pipe, &st->state.viewport);
-#else
-   /* XXX is this sufficient? */
-   st_invalidate_state(st->ctx, _NEW_COLOR | _NEW_TEXTURE);
-#endif
+   /* shaders don't go through CSO yet */
+   if (st->fp)
+      pipe->bind_fs_state(pipe, st->fp->driver_shader);
+   if (st->vp)
+      pipe->bind_vs_state(pipe, st->vp->driver_shader);
 
    return TRUE;
 }
diff --git a/src/mesa/state_tracker/st_program.c b/src/mesa/state_tracker/st_program.c
index d9d11ee0e8..101a6a48af 100644
--- a/src/mesa/state_tracker/st_program.c
+++ b/src/mesa/state_tracker/st_program.c
@@ -260,6 +260,16 @@ st_translate_vertex_program(struct st_context *st,
       outputMapping = defaultOutputMapping;
    }
 
+   /* free old shader state, if any */
+   if (stvp->state.tokens) {
+      FREE((void *) stvp->state.tokens);
+      stvp->state.tokens = NULL;
+   }
+   if (stvp->driver_shader) {
+      pipe->delete_vs_state(pipe, stvp->driver_shader);
+      stvp->driver_shader = NULL;
+   }
+
    /* XXX: fix static allocation of tokens:
     */
    num_tokens = tgsi_translate_mesa_program( TGSI_PROCESSOR_VERTEX,
diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h
index 78f5f451ed..7abccb3a69 100644
--- a/src/mesa/state_tracker/st_texture.h
+++ b/src/mesa/state_tracker/st_texture.h
@@ -71,8 +71,6 @@ struct st_texture_object
    GLboolean imageOverride;
    GLint depthOverride;
    GLuint pitchOverride;
-
-   GLboolean dirtyData;
 };
 
 
diff --git a/src/mesa/swrast/s_bitmap.c b/src/mesa/swrast/s_bitmap.c
index 1e7f6c18e6..f3dda12e25 100644
--- a/src/mesa/swrast/s_bitmap.c
+++ b/src/mesa/swrast/s_bitmap.c
@@ -1,8 +1,8 @@
 /*
  * Mesa 3-D graphics library
- * Version:  6.5.2
+ * Version:  7.1
  *
- * Copyright (C) 1999-2006  Brian Paul   All Rights Reserved.
+ * Copyright (C) 1999-2008  Brian Paul   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"),
@@ -57,25 +57,9 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py,
 
    ASSERT(ctx->RenderMode == GL_RENDER);
 
-   if (unpack->BufferObj->Name) {
-      /* unpack from PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(2, unpack, width, height, 1,
-                                     GL_COLOR_INDEX, GL_BITMAP,
-                                     (GLvoid *) bitmap)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,"glBitmap(invalid PBO access)");
-         return;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              unpack->BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION, "glBitmap(PBO is mapped)");
-         return;
-      }
-      bitmap = ADD_POINTERS(buf, bitmap);
-   }
+   bitmap = _mesa_map_bitmap_pbo(ctx, unpack, bitmap);
+   if (!bitmap)
+      return;
 
    RENDER_START(swrast,ctx);
 
@@ -150,11 +134,7 @@ _swrast_Bitmap( GLcontext *ctx, GLint px, GLint py,
 
    RENDER_FINISH(swrast,ctx);
 
-   if (unpack->BufferObj->Name) {
-      /* done with PBO so unmap it now */
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              unpack->BufferObj);
-   }
+   _mesa_unmap_bitmap_pbo(ctx, unpack);
 }
 
 
diff --git a/src/mesa/swrast/s_drawpix.c b/src/mesa/swrast/s_drawpix.c
index 0cf425e1c6..fb04d9f746 100644
--- a/src/mesa/swrast/s_drawpix.c
+++ b/src/mesa/swrast/s_drawpix.c
@@ -812,7 +812,6 @@ draw_depth_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
 }
 
 
-
 /**
  * Execute software-based glDrawPixels.
  * By time we get here, all error checking will have been done.
@@ -835,25 +834,9 @@ _swrast_DrawPixels( GLcontext *ctx,
    if (swrast->NewState)
       _swrast_validate_derived( ctx );
 
-   if (unpack->BufferObj->Name) {
-      /* unpack from PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(2, unpack, width, height, 1,
-                                     format, type, pixels)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glDrawPixels(invalid PBO access)");
-         goto end;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                                              GL_READ_ONLY_ARB,
-                                              unpack->BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION, "glDrawPixels(PBO is mapped)");
-         goto end;
-      }
-      pixels = ADD_POINTERS(buf, pixels);
-   }
+   pixels = _mesa_map_drawpix_pbo(ctx, unpack, pixels);
+   if (!pixels)
+      return;
 
    switch (format) {
    case GL_STENCIL_INDEX:
@@ -894,11 +877,7 @@ end:
 
    RENDER_FINISH(swrast,ctx);
 
-   if (unpack->BufferObj->Name) {
-      /* done with PBO so unmap it now */
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_UNPACK_BUFFER_EXT,
-                              unpack->BufferObj);
-   }
+   _mesa_unmap_drapix_pbo(ctx, unpack);
 }
 
 
diff --git a/src/mesa/swrast/s_readpix.c b/src/mesa/swrast/s_readpix.c
index fe9a70f4ea..2f155d0b70 100644
--- a/src/mesa/swrast/s_readpix.c
+++ b/src/mesa/swrast/s_readpix.c
@@ -572,25 +572,9 @@ _swrast_ReadPixels( GLcontext *ctx,
       goto end;
    }
 
-   if (clippedPacking.BufferObj->Name) {
-      /* pack into PBO */
-      GLubyte *buf;
-      if (!_mesa_validate_pbo_access(2, &clippedPacking, width, height, 1,
-                                     format, type, pixels)) {
-         _mesa_error(ctx, GL_INVALID_OPERATION,
-                     "glReadPixels(invalid PBO access)");
-         goto end;
-      }
-      buf = (GLubyte *) ctx->Driver.MapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                                              GL_WRITE_ONLY_ARB,
-                                              clippedPacking.BufferObj);
-      if (!buf) {
-         /* buffer is already mapped - that's an error */
-         _mesa_error(ctx, GL_INVALID_OPERATION, "glReadPixels(PBO is mapped)");
-         goto end;
-      }
-      pixels = ADD_POINTERS(buf, pixels);
-   }
+   pixels = _mesa_map_readpix_pbo(ctx, &clippedPacking, pixels);
+   if (!pixels)
+      return;
 
    switch (format) {
       case GL_COLOR_INDEX:
@@ -632,9 +616,5 @@ _swrast_ReadPixels( GLcontext *ctx,
 end:
    RENDER_FINISH(swrast, ctx);
 
-   if (clippedPacking.BufferObj->Name) {
-      /* done with PBO so unmap it now */
-      ctx->Driver.UnmapBuffer(ctx, GL_PIXEL_PACK_BUFFER_EXT,
-                              clippedPacking.BufferObj);
-   }
+   _mesa_unmap_readpix_pbo(ctx, &clippedPacking);
 }