u_upload_mgr: pass alignment to u_upload_alloc manually
[mesa.git] / src / mesa / state_tracker / st_cb_drawpixels.c
index a8a01db969d3b573047cb6863391b7aa57b66f04..86e8a55e25e7dd233fb12f00a838da731b595b80 100644 (file)
@@ -1,6 +1,6 @@
 /**************************************************************************
  * 
- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * Copyright 2007 VMware, Inc.
  * All Rights Reserved.
  * 
  * Permission is hereby granted, free of charge, to any person obtaining a
@@ -18,7 +18,7 @@
  * 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
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
  * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
  * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
  * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
@@ -33,9 +33,9 @@
 #include "main/imports.h"
 #include "main/image.h"
 #include "main/bufferobj.h"
+#include "main/blit.h"
 #include "main/format_pack.h"
 #include "main/macros.h"
-#include "main/mfeatures.h"
 #include "main/mtypes.h"
 #include "main/pack.h"
 #include "main/pbo.h"
 #include "cso_cache/cso_context.h"
 
 
-/**
- * Check if the given program is:
- * 0: MOVE result.color, fragment.color;
- * 1: END;
- */
-static GLboolean
-is_passthrough_program(const struct gl_fragment_program *prog)
-{
-   if (prog->Base.NumInstructions == 2) {
-      const struct prog_instruction *inst = prog->Base.Instructions;
-      if (inst[0].Opcode == OPCODE_MOV &&
-          inst[1].Opcode == OPCODE_END &&
-          inst[0].DstReg.File == PROGRAM_OUTPUT &&
-          inst[0].DstReg.Index == FRAG_RESULT_COLOR &&
-          inst[0].DstReg.WriteMask == WRITEMASK_XYZW &&
-          inst[0].SrcReg[0].File == PROGRAM_INPUT &&
-          inst[0].SrcReg[0].Index == FRAG_ATTRIB_COL0 &&
-          inst[0].SrcReg[0].Swizzle == SWIZZLE_XYZW) {
-         return GL_TRUE;
-      }
-   }
-   return GL_FALSE;
-}
-
-
-/**
- * Returns a fragment program which implements the current pixel transfer ops.
- */
-static struct gl_fragment_program *
-get_glsl_pixel_transfer_program(struct st_context *st,
-                                struct st_fragment_program *orig)
-{
-   int pixelMaps = 0, scaleAndBias = 0;
-   struct gl_context *ctx = st->ctx;
-   struct st_fragment_program *fp = (struct st_fragment_program *)
-      ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
-
-   if (!fp)
-      return NULL;
-
-   if (ctx->Pixel.RedBias != 0.0 || ctx->Pixel.RedScale != 1.0 ||
-       ctx->Pixel.GreenBias != 0.0 || ctx->Pixel.GreenScale != 1.0 ||
-       ctx->Pixel.BlueBias != 0.0 || ctx->Pixel.BlueScale != 1.0 ||
-       ctx->Pixel.AlphaBias != 0.0 || ctx->Pixel.AlphaScale != 1.0) {
-      scaleAndBias = 1;
-   }
-
-   pixelMaps = ctx->Pixel.MapColorFlag;
-
-   if (pixelMaps) {
-      /* create the colormap/texture now if not already done */
-      if (!st->pixel_xfer.pixelmap_texture) {
-         st->pixel_xfer.pixelmap_texture = st_create_color_map_texture(ctx);
-         st->pixel_xfer.pixelmap_sampler_view =
-            st_create_texture_sampler_view(st->pipe,
-                                           st->pixel_xfer.pixelmap_texture);
-      }
-   }
-
-   get_pixel_transfer_visitor(fp, orig->glsl_to_tgsi,
-                              scaleAndBias, pixelMaps);
-
-   return &fp->Base;
-}
-
-
-/**
- * Make fragment shader for glDraw/CopyPixels.  This shader is made
- * by combining the pixel transfer shader with the user-defined shader.
- * \param fpIn  the current/incoming fragment program
- * \param fpOut  returns the combined fragment program
- */
-void
-st_make_drawpix_fragment_program(struct st_context *st,
-                                 struct gl_fragment_program *fpIn,
-                                 struct gl_fragment_program **fpOut)
-{
-   struct gl_program *newProg;
-   struct st_fragment_program *stfp = (struct st_fragment_program *) fpIn;
-
-   if (is_passthrough_program(fpIn)) {
-      newProg = (struct gl_program *) _mesa_clone_fragment_program(st->ctx,
-                                             &st->pixel_xfer.program->Base);
-   }
-   else if (stfp->glsl_to_tgsi != NULL) {
-      newProg = (struct gl_program *) get_glsl_pixel_transfer_program(st, stfp);
-   }
-   else {
-#if 0
-      /* debug */
-      printf("Base program:\n");
-      _mesa_print_program(&fpIn->Base);
-      printf("DrawPix program:\n");
-      _mesa_print_program(&st->pixel_xfer.program->Base.Base);
-#endif
-      newProg = _mesa_combine_programs(st->ctx,
-                                       &st->pixel_xfer.program->Base.Base,
-                                       &fpIn->Base);
-   }
-
-#if 0
-   /* debug */
-   printf("Combined DrawPixels program:\n");
-   _mesa_print_program(newProg);
-   printf("InputsRead: 0x%x\n", newProg->InputsRead);
-   printf("OutputsWritten: 0x%x\n", newProg->OutputsWritten);
-   _mesa_print_parameter_list(newProg->Parameters);
-#endif
-
-   *fpOut = (struct gl_fragment_program *) newProg;
-}
-
-
 /**
  * Create fragment program that does a TEX() instruction to get a Z and/or
  * stencil value value, then writes to FRAG_RESULT_DEPTH/FRAG_RESULT_STENCIL.
  * Used for glDrawPixels(GL_DEPTH_COMPONENT / GL_STENCIL_INDEX).
  * Pass fragment color through as-is.
- * \return pointer to the gl_fragment program
+ *
+ * \return CSO of the fragment shader.
  */
-struct gl_fragment_program *
-st_make_drawpix_z_stencil_program(struct st_context *st,
-                                  GLboolean write_depth,
-                                  GLboolean write_stencil)
+static void *
+get_drawpix_z_stencil_program(struct st_context *st,
+                              GLboolean write_depth,
+                              GLboolean write_stencil)
 {
-   struct gl_context *ctx = st->ctx;
-   struct gl_program *p;
-   struct gl_fragment_program *fp;
-   GLuint ic = 0;
+   struct ureg_program *ureg;
+   struct ureg_src depth_sampler, stencil_sampler;
+   struct ureg_src texcoord, color;
+   struct ureg_dst out_color, out_depth, out_stencil;
    const GLuint shaderIndex = write_depth * 2 + write_stencil;
+   void *cso;
 
-   assert(shaderIndex < Elements(st->drawpix.shaders));
+   assert(shaderIndex < ARRAY_SIZE(st->drawpix.zs_shaders));
 
-   if (st->drawpix.shaders[shaderIndex]) {
+   if (st->drawpix.zs_shaders[shaderIndex]) {
       /* already have the proper shader */
-      return st->drawpix.shaders[shaderIndex];
+      return st->drawpix.zs_shaders[shaderIndex];
    }
 
-   /*
-    * Create shader now
-    */
-   p = ctx->Driver.NewProgram(ctx, GL_FRAGMENT_PROGRAM_ARB, 0);
-   if (!p)
+   ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
+   if (ureg == NULL)
       return NULL;
 
-   p->NumInstructions = write_depth ? 3 : 1;
-   p->NumInstructions += write_stencil ? 1 : 0;
-
-   p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
-   if (!p->Instructions) {
-      ctx->Driver.DeleteProgram(ctx, p);
-      return NULL;
-   }
-   _mesa_init_instructions(p->Instructions, p->NumInstructions);
+   ureg_property(ureg, TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS, TRUE);
 
    if (write_depth) {
-      /* TEX result.depth, fragment.texcoord[0], texture[0], 2D; */
-      p->Instructions[ic].Opcode = OPCODE_TEX;
-      p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
-      p->Instructions[ic].DstReg.Index = FRAG_RESULT_DEPTH;
-      p->Instructions[ic].DstReg.WriteMask = WRITEMASK_Z;
-      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++;
-      /* MOV result.color, fragment.color; */
-      p->Instructions[ic].Opcode = OPCODE_MOV;
-      p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
-      p->Instructions[ic].DstReg.Index = FRAG_RESULT_COLOR;
-      p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
-      p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_COL0;
-      ic++;
+      color = ureg_DECL_fs_input(ureg, TGSI_SEMANTIC_COLOR, 0,
+                                 TGSI_INTERPOLATE_COLOR);
+      out_color = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
+
+      depth_sampler = ureg_DECL_sampler(ureg, 0);
+      out_depth = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
    }
 
    if (write_stencil) {
-      /* TEX result.stencil, fragment.texcoord[0], texture[0], 2D; */
-      p->Instructions[ic].Opcode = OPCODE_TEX;
-      p->Instructions[ic].DstReg.File = PROGRAM_OUTPUT;
-      p->Instructions[ic].DstReg.Index = FRAG_RESULT_STENCIL;
-      p->Instructions[ic].DstReg.WriteMask = WRITEMASK_Y;
-      p->Instructions[ic].SrcReg[0].File = PROGRAM_INPUT;
-      p->Instructions[ic].SrcReg[0].Index = FRAG_ATTRIB_TEX0;
-      p->Instructions[ic].TexSrcUnit = 1;
-      p->Instructions[ic].TexSrcTarget = TEXTURE_2D_INDEX;
-      ic++;
+      stencil_sampler = ureg_DECL_sampler(ureg, 1);
+      out_stencil = ureg_DECL_output(ureg, TGSI_SEMANTIC_STENCIL, 0);
    }
 
-   /* END; */
-   p->Instructions[ic++].Opcode = OPCODE_END;
+   texcoord = ureg_DECL_fs_input(ureg,
+                                 st->needs_texcoord_semantic ?
+                                    TGSI_SEMANTIC_TEXCOORD :
+                                    TGSI_SEMANTIC_GENERIC,
+                                 0, TGSI_INTERPOLATE_LINEAR);
 
-   assert(ic == p->NumInstructions);
-
-   p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0;
-   p->OutputsWritten = 0;
    if (write_depth) {
-      p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_DEPTH);
-      p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_COLOR);
+      ureg_TEX(ureg, ureg_writemask(out_depth, TGSI_WRITEMASK_Z),
+               TGSI_TEXTURE_2D, texcoord, depth_sampler);
+      ureg_MOV(ureg, out_color, color);
    }
-   if (write_stencil)
-      p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_STENCIL);
 
-   p->SamplersUsed =  0x1;  /* sampler 0 (bit 0) is used */
    if (write_stencil)
-      p->SamplersUsed |= 1 << 1;
+      ureg_TEX(ureg, ureg_writemask(out_stencil, TGSI_WRITEMASK_Y),
+               TGSI_TEXTURE_2D, texcoord, stencil_sampler);
 
-   fp = (struct gl_fragment_program *) p;
+   ureg_END(ureg);
+   cso = ureg_create_shader_and_destroy(ureg, st->pipe);
 
    /* save the new shader */
-   st->drawpix.shaders[shaderIndex] = fp;
-
-   return fp;
+   st->drawpix.zs_shaders[shaderIndex] = cso;
+   return cso;
 }
 
 
@@ -294,6 +151,9 @@ static void *
 make_passthrough_vertex_shader(struct st_context *st, 
                                GLboolean passColor)
 {
+   const unsigned texcoord_semantic = st->needs_texcoord_semantic ?
+      TGSI_SEMANTIC_TEXCOORD : TGSI_SEMANTIC_GENERIC;
+
    if (!st->drawpix.vert_shaders[passColor]) {
       struct ureg_program *ureg = ureg_create( TGSI_PROCESSOR_VERTEX );
 
@@ -307,7 +167,7 @@ make_passthrough_vertex_shader(struct st_context *st,
       
       /* MOV result.texcoord0, vertex.attr[1]; */
       ureg_MOV(ureg, 
-               ureg_DECL_output( ureg, TGSI_SEMANTIC_GENERIC, 0 ),
+               ureg_DECL_output( ureg, texcoord_semantic, 0 ),
                ureg_DECL_vs_input( ureg, 1 ));
       
       if (passColor) {
@@ -396,6 +256,8 @@ internal_format(struct gl_context *ctx, GLenum format, GLenum type)
 
          case GL_UNSIGNED_BYTE_3_3_2:
          case GL_UNSIGNED_BYTE_2_3_3_REV:
+            return GL_R3_G3_B2;
+
          case GL_UNSIGNED_SHORT_4_4_4_4:
          case GL_UNSIGNED_SHORT_4_4_4_4_REV:
             return GL_RGBA4;
@@ -456,12 +318,12 @@ internal_format(struct gl_context *ctx, GLenum format, GLenum type)
  */
 static struct pipe_resource *
 alloc_texture(struct st_context *st, GLsizei width, GLsizei height,
-              enum pipe_format texFormat)
+              enum pipe_format texFormat, unsigned bind)
 {
    struct pipe_resource *pt;
 
    pt = st_texture_create(st, st->internal_target, texFormat, 0,
-                          width, height, 1, 1, PIPE_BIND_SAMPLER_VIEW);
+                          width, height, 1, 1, 0, bind);
 
    return pt;
 }
@@ -479,27 +341,36 @@ make_texture(struct st_context *st,
 {
    struct gl_context *ctx = st->ctx;
    struct pipe_context *pipe = st->pipe;
-   gl_format mformat;
+   mesa_format mformat;
    struct pipe_resource *pt;
    enum pipe_format pipeFormat;
-   GLenum baseInternalFormat, intFormat;
+   GLenum baseInternalFormat;
+
+   /* Choose a pixel format for the temp texture which will hold the
+    * image to draw.
+    */
+   pipeFormat = st_choose_matching_format(st, PIPE_BIND_SAMPLER_VIEW,
+                                          format, type, unpack->SwapBytes);
 
-   intFormat = internal_format(ctx, format, type);
-   baseInternalFormat = _mesa_base_tex_format(ctx, intFormat);
+   if (pipeFormat == PIPE_FORMAT_NONE) {
+      /* Use the generic approach. */
+      GLenum intFormat = internal_format(ctx, format, type);
 
-   mformat = st_ChooseTextureFormat_renderable(ctx, intFormat,
-                                               format, type, GL_FALSE);
-   assert(mformat);
+      pipeFormat = st_choose_format(st, intFormat, format, type,
+                                    PIPE_TEXTURE_2D, 0, PIPE_BIND_SAMPLER_VIEW,
+                                    FALSE);
+      assert(pipeFormat != PIPE_FORMAT_NONE);
+   }
 
-   pipeFormat = st_mesa_format_to_pipe_format(mformat);
-   assert(pipeFormat);
+   mformat = st_pipe_format_to_mesa_format(pipeFormat);
+   baseInternalFormat = _mesa_get_format_base_format(mformat);
 
    pixels = _mesa_map_pbo_source(ctx, unpack, pixels);
    if (!pixels)
       return NULL;
 
    /* alloc temporary texture */
-   pt = alloc_texture(st, width, height, pipeFormat);
+   pt = alloc_texture(st, width, height, pipeFormat, PIPE_BIND_SAMPLER_VIEW);
    if (!pt) {
       _mesa_unmap_pbo_source(ctx, unpack);
       return NULL;
@@ -514,31 +385,48 @@ make_texture(struct st_context *st,
       /* we'll do pixel transfer in a fragment shader */
       ctx->_ImageTransferState = 0x0;
 
-      transfer = pipe_get_transfer(st->pipe, pt, 0, 0,
-                                   PIPE_TRANSFER_WRITE, 0, 0,
-                                   width, height);
-
       /* map texture transfer */
-      dest = pipe_transfer_map(pipe, transfer);
+      dest = pipe_transfer_map(pipe, pt, 0, 0,
+                               PIPE_TRANSFER_WRITE, 0, 0,
+                               width, height, &transfer);
 
 
       /* Put image into texture transfer.
        * Note that the image is actually going to be upside down in
        * the texture.  We deal with that with texcoords.
        */
-      success = _mesa_texstore(ctx, 2,           /* dims */
-                               baseInternalFormat, /* baseInternalFormat */
-                               mformat,          /* gl_format */
-                               transfer->stride, /* dstRowStride, bytes */
-                               &dest,            /* destSlices */
-                               width, height, 1, /* size */
-                               format, type,     /* src format/type */
-                               pixels,           /* data source */
-                               unpack);
+      if ((format == GL_RGBA || format == GL_BGRA)
+          && type == GL_UNSIGNED_BYTE) {
+         /* Use a memcpy-based texstore to avoid software pixel swizzling.
+          * We'll do the necessary swizzling with the pipe_sampler_view to
+          * give much better performance.
+          * XXX in the future, expand this to accomodate more format and
+          * type combinations.
+          */
+         _mesa_memcpy_texture(ctx, 2,
+                              mformat,          /* mesa_format */
+                              transfer->stride, /* dstRowStride, bytes */
+                              &dest,            /* destSlices */
+                              width, height, 1, /* size */
+                              format, type,     /* src format/type */
+                              pixels,           /* data source */
+                              unpack);
+         success = GL_TRUE;
+      }
+      else {
+         success = _mesa_texstore(ctx, 2,           /* dims */
+                                  baseInternalFormat, /* baseInternalFormat */
+                                  mformat,          /* mesa_format */
+                                  transfer->stride, /* dstRowStride, bytes */
+                                  &dest,            /* destSlices */
+                                  width, height, 1, /* size */
+                                  format, type,     /* src format/type */
+                                  pixels,           /* data source */
+                                  unpack);
+      }
 
       /* unmap */
       pipe_transfer_unmap(pipe, transfer);
-      pipe->transfer_destroy(pipe, transfer);
 
       assert(success);
 
@@ -569,8 +457,8 @@ draw_quad(struct gl_context *ctx, GLfloat x0, GLfloat y0, GLfloat z,
    struct pipe_resource *buf = NULL;
    unsigned offset;
 
-   u_upload_alloc(st->uploader, 0, 4 * sizeof(verts[0]), &offset, &buf,
-                 (void**)&verts);
+   u_upload_alloc(st->uploader, 0, 4 * sizeof(verts[0]), 4, &offset,
+                  &buf, (void **) &verts);
    if (!buf) {
       return;
    }
@@ -638,7 +526,9 @@ draw_quad(struct gl_context *ctx, GLfloat x0, GLfloat y0, GLfloat z,
    }
 
    u_upload_unmap(st->uploader);
-   util_draw_vertex_buffer(pipe, st->cso_context, buf, offset,
+   util_draw_vertex_buffer(pipe, st->cso_context, buf,
+                           cso_get_aux_vertex_buffer_slot(st->cso_context),
+                           offset,
                           PIPE_PRIM_QUADS,
                           4,  /* verts */
                           3); /* attribs/vert */
@@ -655,6 +545,7 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
                    int num_sampler_view,
                    void *driver_vp,
                    void *driver_fp,
+                   struct st_fp_variant *fpv,
                    const GLfloat *color,
                    GLboolean invertTex,
                    GLboolean write_depth, GLboolean write_stencil)
@@ -677,14 +568,16 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
 
    cso_save_rasterizer(cso);
    cso_save_viewport(cso);
-   cso_save_samplers(cso, PIPE_SHADER_FRAGMENT);
-   cso_save_sampler_views(cso, PIPE_SHADER_FRAGMENT);
+   cso_save_fragment_samplers(cso);
+   cso_save_fragment_sampler_views(cso);
    cso_save_fragment_shader(cso);
    cso_save_stream_outputs(cso);
    cso_save_vertex_shader(cso);
+   cso_save_tessctrl_shader(cso);
+   cso_save_tesseval_shader(cso);
    cso_save_geometry_shader(cso);
    cso_save_vertex_elements(cso);
-   cso_save_vertex_buffers(cso);
+   cso_save_aux_vertex_buffer_slot(cso);
    if (write_stencil) {
       cso_save_depth_stencil_alpha(cso);
       cso_save_blend(cso);
@@ -695,11 +588,11 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
       struct pipe_rasterizer_state rasterizer;
       memset(&rasterizer, 0, sizeof(rasterizer));
       rasterizer.clamp_fragment_color = !st->clamp_frag_color_in_shader &&
-                                        ctx->Color._ClampFragmentColor &&
-                                        !ctx->DrawBuffer->_IntegerColor;
-      rasterizer.gl_rasterization_rules = 1;
+                                        ctx->Color._ClampFragmentColor;
+      rasterizer.half_pixel_center = 1;
+      rasterizer.bottom_edge_rule = 1;
       rasterizer.depth_clip = !ctx->Transform.DepthClamp;
-      rasterizer.scissor = ctx->Scissor.Enabled;
+      rasterizer.scissor = ctx->Scissor.EnableFlags;
       cso_set_rasterizer(cso, &rasterizer);
    }
 
@@ -735,12 +628,15 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    /* vertex shader state: position + texcoord pass-through */
    cso_set_vertex_shader_handle(cso, driver_vp);
 
-   /* geometry shader state: disabled */
+   /* disable other shaders */
+   cso_set_tessctrl_shader_handle(cso, NULL);
+   cso_set_tesseval_shader_handle(cso, NULL);
    cso_set_geometry_shader_handle(cso, NULL);
 
-   /* texture sampling state: */
+   /* user samplers, plus the drawpix samplers */
    {
       struct pipe_sampler_state sampler;
+
       memset(&sampler, 0, sizeof(sampler));
       sampler.wrap_s = PIPE_TEX_WRAP_CLAMP;
       sampler.wrap_t = PIPE_TEX_WRAP_CLAMP;
@@ -750,11 +646,25 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
       sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
       sampler.normalized_coords = normalized;
 
-      cso_single_sampler(cso, PIPE_SHADER_FRAGMENT, 0, &sampler);
-      if (num_sampler_view > 1) {
-         cso_single_sampler(cso, PIPE_SHADER_FRAGMENT, 1, &sampler);
+      if (fpv) {
+         const struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+         uint num = MAX2(MAX2(fpv->drawpix_sampler, fpv->pixelmap_sampler) + 1,
+                         st->state.num_samplers[PIPE_SHADER_FRAGMENT]);
+         uint i;
+
+         for (i = 0; i < st->state.num_samplers[PIPE_SHADER_FRAGMENT]; i++)
+            samplers[i] = &st->state.samplers[PIPE_SHADER_FRAGMENT][i];
+
+         samplers[fpv->drawpix_sampler] = &sampler;
+         if (sv[1])
+            samplers[fpv->pixelmap_sampler] = &sampler;
+
+         cso_set_samplers(cso, PIPE_SHADER_FRAGMENT, num, samplers);
+      } else {
+         const struct pipe_sampler_state *samplers[2] = {&sampler, &sampler};
+
+         cso_set_samplers(cso, PIPE_SHADER_FRAGMENT, num_sampler_view, samplers);
       }
-      cso_single_sampler_done(cso, PIPE_SHADER_FRAGMENT);
    }
 
    /* viewport state: viewport matching window dims */
@@ -765,19 +675,31 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
       vp.scale[0] =  0.5f * w;
       vp.scale[1] = -0.5f * h;
       vp.scale[2] = 0.5f;
-      vp.scale[3] = 1.0f;
       vp.translate[0] = 0.5f * w;
       vp.translate[1] = 0.5f * h;
       vp.translate[2] = 0.5f;
-      vp.translate[3] = 0.0f;
       cso_set_viewport(cso, &vp);
    }
 
    cso_set_vertex_elements(cso, 3, st->velems_util_draw);
-   cso_set_stream_outputs(st->cso_context, 0, NULL, 0);
+   cso_set_stream_outputs(st->cso_context, 0, NULL, NULL);
+
+   /* user textures, plus the drawpix textures */
+   if (fpv) {
+      struct pipe_sampler_view *sampler_views[PIPE_MAX_SAMPLERS];
+      uint num = MAX3(fpv->drawpix_sampler + 1,
+                      fpv->pixelmap_sampler + 1,
+                      st->state.num_sampler_views[PIPE_SHADER_FRAGMENT]);
 
-   /* texture state: */
-   cso_set_sampler_views(cso, PIPE_SHADER_FRAGMENT, num_sampler_view, sv);
+      memcpy(sampler_views, st->state.sampler_views[PIPE_SHADER_FRAGMENT],
+             sizeof(sampler_views));
+
+      sampler_views[fpv->drawpix_sampler] = sv[0];
+      if (sv[1])
+         sampler_views[fpv->pixelmap_sampler] = sv[1];
+      cso_set_sampler_views(cso, PIPE_SHADER_FRAGMENT, num, sampler_views);
+   } else
+      cso_set_sampler_views(cso, PIPE_SHADER_FRAGMENT, num_sampler_view, sv);
 
    /* Compute Gallium window coords (y=0=top) with pixel zoom.
     * Recall that these coords are transformed by the current
@@ -794,7 +716,7 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    y1 = y + height * ctx->Pixel.ZoomY;
 
    /* convert Z from [0,1] to [-1,-1] to match viewport Z scale/bias */
-   z = z * 2.0 - 1.0;
+   z = z * 2.0f - 1.0f;
 
    draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex,
              normalized ? ((GLfloat) width / sv[0]->texture->width0) : (GLfloat)width,
@@ -803,13 +725,15 @@ draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
    /* restore state */
    cso_restore_rasterizer(cso);
    cso_restore_viewport(cso);
-   cso_restore_samplers(cso, PIPE_SHADER_FRAGMENT);
-   cso_restore_sampler_views(cso, PIPE_SHADER_FRAGMENT);
+   cso_restore_fragment_samplers(cso);
+   cso_restore_fragment_sampler_views(cso);
    cso_restore_fragment_shader(cso);
    cso_restore_vertex_shader(cso);
+   cso_restore_tessctrl_shader(cso);
+   cso_restore_tesseval_shader(cso);
    cso_restore_geometry_shader(cso);
    cso_restore_vertex_elements(cso);
-   cso_restore_vertex_buffers(cso);
+   cso_restore_aux_vertex_buffer_slot(cso);
    cso_restore_stream_outputs(cso);
    if (write_stencil) {
       cso_restore_depth_stencil_alpha(cso);
@@ -863,12 +787,11 @@ draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
       usage = PIPE_TRANSFER_WRITE;
    }
 
-   pt = pipe_get_transfer(pipe, strb->texture,
-                          strb->rtt_level, strb->rtt_face + strb->rtt_slice,
-                          usage, x, y,
-                          width, height);
-
-   stmap = pipe_transfer_map(pipe, pt);
+   stmap = pipe_transfer_map(pipe, strb->texture,
+                             strb->surface->u.tex.level,
+                             strb->surface->u.tex.first_layer,
+                             usage, x, y,
+                             width, height, &pt);
 
    pixels = _mesa_map_pbo_source(ctx, &clippedUnpack, pixels);
    assert(pixels);
@@ -996,7 +919,6 @@ draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
 
    /* unmap the stencil buffer */
    pipe_transfer_unmap(pipe, pt);
-   pipe->transfer_destroy(pipe, pt);
 }
 
 
@@ -1013,7 +935,7 @@ get_color_fp_variant(struct st_context *st)
 
    memset(&key, 0, sizeof(key));
 
-   key.st = st;
+   key.st = st->has_shareable_shaders ? NULL : st;
    key.drawpixels = 1;
    key.scaleAndBias = (ctx->Pixel.RedBias != 0.0 ||
                        ctx->Pixel.RedScale != 1.0 ||
@@ -1025,32 +947,7 @@ get_color_fp_variant(struct st_context *st)
                        ctx->Pixel.AlphaScale != 1.0);
    key.pixelMaps = ctx->Pixel.MapColorFlag;
    key.clamp_color = st->clamp_frag_color_in_shader &&
-                     st->ctx->Color._ClampFragmentColor &&
-                     !st->ctx->DrawBuffer->_IntegerColor;
-
-   fpv = st_get_fp_variant(st, st->fp, &key);
-
-   return fpv;
-}
-
-
-/**
- * Get fragment program variant for a glDrawPixels or glCopyPixels
- * command for depth/stencil data.
- */
-static struct st_fp_variant *
-get_depth_stencil_fp_variant(struct st_context *st, GLboolean write_depth,
-                             GLboolean write_stencil)
-{
-   struct st_fp_variant_key key;
-   struct st_fp_variant *fpv;
-
-   memset(&key, 0, sizeof(key));
-
-   key.st = st;
-   key.drawpixels = 1;
-   key.drawpixels_z = write_depth;
-   key.drawpixels_stencil = write_stencil;
+                     st->ctx->Color._ClampFragmentColor;
 
    fpv = st_get_fp_variant(st, st->fp, &key);
 
@@ -1065,7 +962,7 @@ static void
 clamp_size(struct pipe_context *pipe, GLsizei *width, GLsizei *height,
            struct gl_pixelstore_attrib *unpack)
 {
-   const unsigned maxSize = 
+   const int maxSize =
       1 << (pipe->screen->get_param(pipe->screen,
                                     PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
 
@@ -1080,6 +977,69 @@ clamp_size(struct pipe_context *pipe, GLsizei *width, GLsizei *height,
 }
 
 
+/**
+ * Search the array of 4 swizzle components for the named component and return
+ * its position.
+ */
+static unsigned
+search_swizzle(const unsigned char swizzle[4], unsigned component)
+{
+   unsigned i;
+   for (i = 0; i < 4; i++) {
+      if (swizzle[i] == component)
+         return i;
+   }
+   assert(!"search_swizzle() failed");
+   return 0;
+}
+
+
+/**
+ * Set the sampler view's swizzle terms.  This is used to handle RGBA
+ * swizzling when the incoming image format isn't an exact match for
+ * the actual texture format.  For example, if we have glDrawPixels(
+ * GL_RGBA, GL_UNSIGNED_BYTE) and we chose the texture format
+ * PIPE_FORMAT_B8G8R8A8 then we can do use the sampler view swizzle to
+ * avoid swizzling all the pixels in software in the texstore code.
+ */
+static void
+setup_sampler_swizzle(struct pipe_sampler_view *sv, GLenum format, GLenum type)
+{
+   if ((format == GL_RGBA || format == GL_BGRA) && type == GL_UNSIGNED_BYTE) {
+      const struct util_format_description *desc =
+         util_format_description(sv->texture->format);
+      unsigned c0, c1, c2, c3;
+
+      /* Every gallium driver supports at least one 32-bit packed RGBA format.
+       * We must have chosen one for (GL_RGBA, GL_UNSIGNED_BYTE).
+       */
+      assert(desc->block.bits == 32);
+
+      /* invert the format's swizzle to setup the sampler's swizzle */
+      if (format == GL_RGBA) {
+         c0 = UTIL_FORMAT_SWIZZLE_X;
+         c1 = UTIL_FORMAT_SWIZZLE_Y;
+         c2 = UTIL_FORMAT_SWIZZLE_Z;
+         c3 = UTIL_FORMAT_SWIZZLE_W;
+      }
+      else {
+         assert(format == GL_BGRA);
+         c0 = UTIL_FORMAT_SWIZZLE_Z;
+         c1 = UTIL_FORMAT_SWIZZLE_Y;
+         c2 = UTIL_FORMAT_SWIZZLE_X;
+         c3 = UTIL_FORMAT_SWIZZLE_W;
+      }
+      sv->swizzle_r = search_swizzle(desc->swizzle, c0);
+      sv->swizzle_g = search_swizzle(desc->swizzle, c1);
+      sv->swizzle_b = search_swizzle(desc->swizzle, c2);
+      sv->swizzle_a = search_swizzle(desc->swizzle, c3);
+   }
+   else {
+      /* use the default sampler swizzle */
+   }
+}
+
+
 /**
  * Called via ctx->Driver.DrawPixels()
  */
@@ -1094,10 +1054,11 @@ st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
    const GLfloat *color;
    struct pipe_context *pipe = st->pipe;
    GLboolean write_stencil = GL_FALSE, write_depth = GL_FALSE;
-   struct pipe_sampler_view *sv[2];
+   struct pipe_sampler_view *sv[2] = { NULL };
    int num_sampler_view = 1;
-   struct st_fp_variant *fpv;
    struct gl_pixelstore_attrib clippedUnpack;
+   struct st_fp_variant *fpv = NULL;
+   struct pipe_resource *pt;
 
    /* Mesa state should be up to date by now */
    assert(ctx->NewState == 0x0);
@@ -1131,66 +1092,81 @@ st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
     * Get vertex/fragment shaders
     */
    if (write_depth || write_stencil) {
-      fpv = get_depth_stencil_fp_variant(st, write_depth, write_stencil);
-
-      driver_fp = fpv->driver_shader;
-
+      driver_fp = get_drawpix_z_stencil_program(st, write_depth,
+                                                write_stencil);
       driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
-
       color = ctx->Current.RasterColor;
    }
    else {
       fpv = get_color_fp_variant(st);
 
       driver_fp = fpv->driver_shader;
-
       driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
 
       color = NULL;
-      if (st->pixel_xfer.pixelmap_enabled) {
-         sv[1] = st->pixel_xfer.pixelmap_sampler_view;
-         num_sampler_view++;
+      if (ctx->Pixel.MapColorFlag) {
+         pipe_sampler_view_reference(&sv[1],
+                                     st->pixel_xfer.pixelmap_sampler_view);
+         num_sampler_view++;
       }
+
+      /* compiling a new fragment shader variant added new state constants
+       * into the constant buffer, we need to update them
+       */
+      st_upload_constants(st, st->fp->Base.Base.Parameters,
+                          PIPE_SHADER_FRAGMENT);
    }
 
-   /* update fragment program constants */
-   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
+   /* Put glDrawPixels image into a texture */
+   pt = make_texture(st, width, height, format, type, unpack, pixels);
+   if (!pt) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels");
+      return;
+   }
 
-   /* draw with textured quad */
-   {
-      struct pipe_resource *pt
-         = make_texture(st, width, height, format, type, unpack, pixels);
-      if (pt) {
-         sv[0] = st_create_texture_sampler_view(st->pipe, pt);
-
-         if (sv[0]) {
-            /* Create a second sampler view to read stencil.
-             * The stencil is written using the shader stencil export
-             * functionality. */
-            if (write_stencil) {
-               enum pipe_format stencil_format =
-                     util_format_stencil_only(pt->format);
-
-               sv[1] = st_create_texture_sampler_view_format(st->pipe, pt,
-                                                             stencil_format);
-               num_sampler_view++;
-            }
+   /* create sampler view for the image */
+   sv[0] = st_create_texture_sampler_view(st->pipe, pt);
+   if (!sv[0]) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels");
+      pipe_resource_reference(&pt, NULL);
+      return;
+   }
 
-            draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
-                               width, height,
-                               ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
-                               sv,
-                               num_sampler_view,
-                               driver_vp,
-                               driver_fp,
-                               color, GL_FALSE, write_depth, write_stencil);
-            pipe_sampler_view_reference(&sv[0], NULL);
-            if (num_sampler_view > 1)
-               pipe_sampler_view_reference(&sv[1], NULL);
-         }
+   /* Set up the sampler view's swizzle */
+   setup_sampler_swizzle(sv[0], format, type);
+
+   /* Create a second sampler view to read stencil.  The stencil is
+    * written using the shader stencil export functionality.
+    */
+   if (write_stencil) {
+      enum pipe_format stencil_format =
+         util_format_stencil_only(pt->format);
+      /* we should not be doing pixel map/transfer (see above) */
+      assert(num_sampler_view == 1);
+      sv[1] = st_create_texture_sampler_view_format(st->pipe, pt,
+                                                    stencil_format);
+      if (!sv[1]) {
+         _mesa_error(ctx, GL_OUT_OF_MEMORY, "glDrawPixels");
          pipe_resource_reference(&pt, NULL);
+         pipe_sampler_view_reference(&sv[0], NULL);
+         return;
       }
+      num_sampler_view++;
    }
+
+   draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
+                      width, height,
+                      ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
+                      sv,
+                      num_sampler_view,
+                      driver_vp,
+                      driver_fp, fpv,
+                      color, GL_FALSE, write_depth, write_stencil);
+   pipe_sampler_view_reference(&sv[0], NULL);
+   if (num_sampler_view > 1)
+      pipe_sampler_view_reference(&sv[1], NULL);
+
+   pipe_resource_reference(&pt, NULL);
 }
 
 
@@ -1247,18 +1223,16 @@ copy_stencil_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
       dsty = rbDraw->Base.Height - dsty - height;
    }
 
-   ptDraw = pipe_get_transfer(pipe,
-                              rbDraw->texture,
-                              rbDraw->rtt_level,
-                              rbDraw->rtt_face + rbDraw->rtt_slice,
-                              usage, dstx, dsty,
-                              width, height);
-
-   assert(util_format_get_blockwidth(ptDraw->resource->format) == 1);
-   assert(util_format_get_blockheight(ptDraw->resource->format) == 1);
+   assert(util_format_get_blockwidth(rbDraw->texture->format) == 1);
+   assert(util_format_get_blockheight(rbDraw->texture->format) == 1);
 
    /* map the stencil buffer */
-   drawMap = pipe_transfer_map(pipe, ptDraw);
+   drawMap = pipe_transfer_map(pipe,
+                               rbDraw->texture,
+                               rbDraw->surface->u.tex.level,
+                               rbDraw->surface->u.tex.first_layer,
+                               usage, dstx, dsty,
+                               width, height, &ptDraw);
 
    /* draw */
    /* XXX PixelZoom not handled yet */
@@ -1283,7 +1257,6 @@ copy_stencil_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
 
    /* unmap the stencil buffer */
    pipe_transfer_unmap(pipe, ptDraw);
-   pipe->transfer_destroy(pipe, ptDraw);
 }
 
 
@@ -1301,29 +1274,13 @@ st_get_color_read_renderbuffer(struct gl_context *ctx)
 }
 
 
-/** Do the src/dest regions overlap? */
-static GLboolean
-regions_overlap(GLint srcX, GLint srcY, GLint dstX, GLint dstY,
-                GLsizei width, GLsizei height)
-{
-   if (srcX + width <= dstX ||
-       dstX + width <= srcX ||
-       srcY + height <= dstY ||
-       dstY + height <= srcY)
-      return GL_FALSE;
-   else
-      return GL_TRUE;
-}
-
-
 /**
  * Try to do a glCopyPixels for simple cases with a blit by calling
- * pipe->resource_copy_region().
+ * pipe->blit().
  *
  * We can do this when we're copying color pixels (depth/stencil
  * eventually) with no pixel zoom, no pixel transfer ops, no
- * per-fragment ops, the src/dest regions don't overlap and the
- * src/dest pixel formats are the same.
+ * per-fragment ops, and the src/dest regions don't overlap.
  */
 static GLboolean
 blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
@@ -1332,8 +1289,9 @@ blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
 {
    struct st_context *st = st_context(ctx);
    struct pipe_context *pipe = st->pipe;
+   struct pipe_screen *screen = pipe->screen;
    struct gl_pixelstore_attrib pack, unpack;
-   GLint readX, readY, readW, readH;
+   GLint readX, readY, readW, readH, drawX, drawY, drawW, drawH;
 
    if (type == GL_COLOR &&
        ctx->Pixel.ZoomX == 1.0 &&
@@ -1346,12 +1304,11 @@ blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
        !ctx->Stencil.Enabled &&
        !ctx->FragmentProgram.Enabled &&
        !ctx->VertexProgram.Enabled &&
-       !ctx->Shader.CurrentFragmentProgram &&
-       st_fb_orientation(ctx->ReadBuffer) == st_fb_orientation(ctx->DrawBuffer) &&
+       !ctx->_Shader->CurrentProgram[MESA_SHADER_FRAGMENT] &&
        ctx->DrawBuffer->_NumColorDrawBuffers == 1 &&
-       !ctx->Query.CondRenderQuery) {
+       !ctx->Query.CondRenderQuery &&
+       !ctx->Query.CurrentOcclusionObject) {
       struct st_renderbuffer *rbRead, *rbDraw;
-      GLint drawX, drawY;
 
       /*
        * Clip the read region against the src buffer bounds.
@@ -1378,29 +1335,65 @@ blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
       readX = readX - pack.SkipPixels + unpack.SkipPixels;
       readY = readY - pack.SkipRows + unpack.SkipRows;
 
+      drawW = readW;
+      drawH = readH;
+
       rbRead = st_get_color_read_renderbuffer(ctx);
       rbDraw = st_renderbuffer(ctx->DrawBuffer->_ColorDrawBuffers[0]);
 
-      if ((rbRead != rbDraw ||
-           !regions_overlap(readX, readY, drawX, drawY, readW, readH)) &&
-          rbRead->Base.Format == rbDraw->Base.Format) {
-         struct pipe_box srcBox;
-
-         /* flip src/dst position if needed */
-         if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
-            /* both buffers will have the same orientation */
-            readY = ctx->ReadBuffer->Height - readY - readH;
-            drawY = ctx->DrawBuffer->Height - drawY - readH;
-         }
+      /* Flip src/dst position depending on the orientation of buffers. */
+      if (st_fb_orientation(ctx->ReadBuffer) == Y_0_TOP) {
+         readY = rbRead->Base.Height - readY;
+         readH = -readH;
+      }
 
-         u_box_2d(readX, readY, readW, readH, &srcBox);
+      if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
+         /* We can't flip the destination for pipe->blit, so we only adjust
+          * its position and flip the source.
+          */
+         drawY = rbDraw->Base.Height - drawY - drawH;
+         readY += readH;
+         readH = -readH;
+      }
 
-         pipe->resource_copy_region(pipe,
-                                    rbDraw->texture,
-                                    rbDraw->rtt_level, drawX, drawY, 0,
-                                    rbRead->texture,
-                                    rbRead->rtt_level, &srcBox);
-         return GL_TRUE;
+      if (rbRead != rbDraw ||
+          !_mesa_regions_overlap(readX, readY, readX + readW, readY + readH,
+                                 drawX, drawY, drawX + drawW, drawY + drawH)) {
+         struct pipe_blit_info blit;
+
+         memset(&blit, 0, sizeof(blit));
+         blit.src.resource = rbRead->texture;
+         blit.src.level = rbRead->surface->u.tex.level;
+         blit.src.format = rbRead->texture->format;
+         blit.src.box.x = readX;
+         blit.src.box.y = readY;
+         blit.src.box.z = rbRead->surface->u.tex.first_layer;
+         blit.src.box.width = readW;
+         blit.src.box.height = readH;
+         blit.src.box.depth = 1;
+         blit.dst.resource = rbDraw->texture;
+         blit.dst.level = rbDraw->surface->u.tex.level;
+         blit.dst.format = rbDraw->texture->format;
+         blit.dst.box.x = drawX;
+         blit.dst.box.y = drawY;
+         blit.dst.box.z = rbDraw->surface->u.tex.first_layer;
+         blit.dst.box.width = drawW;
+         blit.dst.box.height = drawH;
+         blit.dst.box.depth = 1;
+         blit.mask = PIPE_MASK_RGBA;
+         blit.filter = PIPE_TEX_FILTER_NEAREST;
+
+         if (screen->is_format_supported(screen, blit.src.format,
+                                         blit.src.resource->target,
+                                         blit.src.resource->nr_samples,
+                                         PIPE_BIND_SAMPLER_VIEW) &&
+             screen->is_format_supported(screen, blit.dst.format,
+                                         blit.dst.resource->target,
+                                         blit.dst.resource->nr_samples,
+                                         PIPE_BIND_RENDER_TARGET)) {
+            pipe->blit(pipe, &blit);
+            return GL_TRUE;
+         }
       }
    }
 
@@ -1419,15 +1412,15 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
    struct st_renderbuffer *rbRead;
    void *driver_vp, *driver_fp;
    struct pipe_resource *pt;
-   struct pipe_sampler_view *sv[2];
+   struct pipe_sampler_view *sv[2] = { NULL };
+   struct st_fp_variant *fpv = NULL;
    int num_sampler_view = 1;
    GLfloat *color;
-   enum pipe_format srcFormat, texFormat;
+   enum pipe_format srcFormat;
+   unsigned srcBind;
    GLboolean invertTex = GL_FALSE;
    GLint readX, readY, readW, readH;
-   GLuint sample_count;
    struct gl_pixelstore_attrib pack = ctx->DefaultPacking;
-   struct st_fp_variant *fpv;
 
    st_validate_state(st);
 
@@ -1459,18 +1452,25 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
     * Get vertex/fragment shaders
     */
    if (type == GL_COLOR) {
+      fpv = get_color_fp_variant(st);
+
       rbRead = st_get_color_read_renderbuffer(ctx);
       color = NULL;
 
-      fpv = get_color_fp_variant(st);
       driver_fp = fpv->driver_shader;
-
       driver_vp = make_passthrough_vertex_shader(st, GL_FALSE);
 
-      if (st->pixel_xfer.pixelmap_enabled) {
-         sv[1] = st->pixel_xfer.pixelmap_sampler_view;
-         num_sampler_view++;
+      if (ctx->Pixel.MapColorFlag) {
+         pipe_sampler_view_reference(&sv[1],
+                                     st->pixel_xfer.pixelmap_sampler_view);
+         num_sampler_view++;
       }
+
+      /* compiling a new fragment shader variant added new state constants
+       * into the constant buffer, we need to update them
+       */
+      st_upload_constants(st, st->fp->Base.Base.Parameters,
+                          PIPE_SHADER_FRAGMENT);
    }
    else {
       assert(type == GL_DEPTH);
@@ -1478,42 +1478,56 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
                                Attachment[BUFFER_DEPTH].Renderbuffer);
       color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
 
-      fpv = get_depth_stencil_fp_variant(st, GL_TRUE, GL_FALSE);
-      driver_fp = fpv->driver_shader;
-
+      driver_fp = get_drawpix_z_stencil_program(st, GL_TRUE, GL_FALSE);
       driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
    }
 
-   /* update fragment program constants */
-   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
-
-   sample_count = rbRead->texture->nr_samples;
-   /* I believe this would be legal, presumably would need to do a resolve
-      for color, and for depth/stencil spec says to just use one of the
-      depth/stencil samples per pixel? Need some transfer clarifications. */
-   assert(sample_count < 2);
-
+   /* Choose the format for the temporary texture. */
    srcFormat = rbRead->texture->format;
+   srcBind = PIPE_BIND_SAMPLER_VIEW |
+      (type == GL_COLOR ? PIPE_BIND_RENDER_TARGET : PIPE_BIND_DEPTH_STENCIL);
 
-   if (screen->is_format_supported(screen, srcFormat, st->internal_target,
-                                   sample_count,
-                                   PIPE_BIND_SAMPLER_VIEW)) {
-      texFormat = srcFormat;
-   }
-   else {
-      /* srcFormat can't be used as a texture format */
+   if (!screen->is_format_supported(screen, srcFormat, st->internal_target, 0,
+                                    srcBind)) {
+      /* srcFormat is non-renderable. Find a compatible renderable format. */
       if (type == GL_DEPTH) {
-         texFormat = st_choose_format(screen, GL_DEPTH_COMPONENT,
-                                      GL_NONE, GL_NONE, st->internal_target,
-                                     sample_count, PIPE_BIND_DEPTH_STENCIL);
-         assert(texFormat != PIPE_FORMAT_NONE);
+         srcFormat = st_choose_format(st, GL_DEPTH_COMPONENT, GL_NONE,
+                                      GL_NONE, st->internal_target, 0,
+                                      srcBind, FALSE);
       }
       else {
-         /* default color format */
-         texFormat = st_choose_format(screen, GL_RGBA,
-                                      GL_NONE, GL_NONE, st->internal_target,
-                                      sample_count, PIPE_BIND_SAMPLER_VIEW);
-         assert(texFormat != PIPE_FORMAT_NONE);
+         assert(type == GL_COLOR);
+
+         if (util_format_is_float(srcFormat)) {
+            srcFormat = st_choose_format(st, GL_RGBA32F, GL_NONE,
+                                         GL_NONE, st->internal_target, 0,
+                                         srcBind, FALSE);
+         }
+         else if (util_format_is_pure_sint(srcFormat)) {
+            srcFormat = st_choose_format(st, GL_RGBA32I, GL_NONE,
+                                         GL_NONE, st->internal_target, 0,
+                                         srcBind, FALSE);
+         }
+         else if (util_format_is_pure_uint(srcFormat)) {
+            srcFormat = st_choose_format(st, GL_RGBA32UI, GL_NONE,
+                                         GL_NONE, st->internal_target, 0,
+                                         srcBind, FALSE);
+         }
+         else if (util_format_is_snorm(srcFormat)) {
+            srcFormat = st_choose_format(st, GL_RGBA16_SNORM, GL_NONE,
+                                         GL_NONE, st->internal_target, 0,
+                                         srcBind, FALSE);
+         }
+         else {
+            srcFormat = st_choose_format(st, GL_RGBA, GL_NONE,
+                                         GL_NONE, st->internal_target, 0,
+                                         srcBind, FALSE);
+         }
+      }
+
+      if (srcFormat == PIPE_FORMAT_NONE) {
+         assert(0 && "cannot choose a format for src of CopyPixels");
+         return;
       }
    }
 
@@ -1545,8 +1559,8 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
    readW = MAX2(0, readW);
    readH = MAX2(0, readH);
 
-   /* alloc temporary texture */
-   pt = alloc_texture(st, width, height, texFormat);
+   /* Allocate the temporary texture. */
+   pt = alloc_texture(st, width, height, srcFormat, srcBind);
    if (!pt)
       return;
 
@@ -1556,67 +1570,33 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
       return;
    }
 
-   /* Make temporary texture which is a copy of the src region.
-    */
-   if (srcFormat == texFormat) {
-      struct pipe_box src_box;
-      u_box_2d(readX, readY, readW, readH, &src_box);
-      /* copy source framebuffer surface into mipmap/texture */
-      pipe->resource_copy_region(pipe,
-                                 pt,                                /* dest tex */
-                                 0,                                 /* dest lvl */
-                                 pack.SkipPixels, pack.SkipRows, 0, /* dest pos */
-                                 rbRead->texture,                   /* src tex */
-                                 rbRead->rtt_level,                 /* src lvl */
-                                 &src_box);
-
-   }
-   else {
-      /* CPU-based fallback/conversion */
-      struct pipe_transfer *ptRead =
-         pipe_get_transfer(st->pipe, rbRead->texture,
-                           rbRead->rtt_level,
-                           rbRead->rtt_face + rbRead->rtt_slice,
-                           PIPE_TRANSFER_READ,
-                           readX, readY, readW, readH);
-      struct pipe_transfer *ptTex;
-      enum pipe_transfer_usage transfer_usage;
-
-      if (ST_DEBUG & DEBUG_FALLBACK)
-         debug_printf("%s: fallback processing\n", __FUNCTION__);
-
-      if (type == GL_DEPTH && util_format_is_depth_and_stencil(pt->format))
-         transfer_usage = PIPE_TRANSFER_READ_WRITE;
-      else
-         transfer_usage = PIPE_TRANSFER_WRITE;
-
-      ptTex = pipe_get_transfer(st->pipe, pt, 0, 0, transfer_usage,
-                                0, 0, width, height);
-
-      /* copy image from ptRead surface to ptTex surface */
-      if (type == GL_COLOR) {
-         /* alternate path using get/put_tile() */
-         GLfloat *buf = malloc(width * height * 4 * sizeof(GLfloat));
-         enum pipe_format readFormat, drawFormat;
-         readFormat = util_format_linear(rbRead->texture->format);
-         drawFormat = util_format_linear(pt->format);
-         pipe_get_tile_rgba_format(pipe, ptRead, 0, 0, readW, readH,
-                                   readFormat, buf);
-         pipe_put_tile_rgba_format(pipe, ptTex, pack.SkipPixels, pack.SkipRows,
-                                   readW, readH, drawFormat, buf);
-         free(buf);
-      }
-      else {
-         /* GL_DEPTH */
-         GLuint *buf = malloc(width * height * sizeof(GLuint));
-         pipe_get_tile_z(pipe, ptRead, 0, 0, readW, readH, buf);
-         pipe_put_tile_z(pipe, ptTex, pack.SkipPixels, pack.SkipRows,
-                         readW, readH, buf);
-         free(buf);
-      }
-
-      pipe->transfer_destroy(pipe, ptRead);
-      pipe->transfer_destroy(pipe, ptTex);
+   /* Copy the src region to the temporary texture. */
+   {
+      struct pipe_blit_info blit;
+
+      memset(&blit, 0, sizeof(blit));
+      blit.src.resource = rbRead->texture;
+      blit.src.level = rbRead->surface->u.tex.level;
+      blit.src.format = rbRead->texture->format;
+      blit.src.box.x = readX;
+      blit.src.box.y = readY;
+      blit.src.box.z = rbRead->surface->u.tex.first_layer;
+      blit.src.box.width = readW;
+      blit.src.box.height = readH;
+      blit.src.box.depth = 1;
+      blit.dst.resource = pt;
+      blit.dst.level = 0;
+      blit.dst.format = pt->format;
+      blit.dst.box.x = pack.SkipPixels;
+      blit.dst.box.y = pack.SkipRows;
+      blit.dst.box.z = 0;
+      blit.dst.box.width = readW;
+      blit.dst.box.height = readH;
+      blit.dst.box.depth = 1;
+      blit.mask = util_format_get_mask(pt->format) & ~PIPE_MASK_S;
+      blit.filter = PIPE_TEX_FILTER_NEAREST;
+
+      pipe->blit(pipe, &blit);
    }
 
    /* OK, the texture 'pt' contains the src image/pixels.  Now draw a
@@ -1627,7 +1607,7 @@ st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
                       sv,
                       num_sampler_view,
                       driver_vp, 
-                      driver_fp,
+                      driver_fp, fpv,
                       color, invertTex, GL_FALSE, GL_FALSE);
 
    pipe_resource_reference(&pt, NULL);
@@ -1648,12 +1628,12 @@ st_destroy_drawpix(struct st_context *st)
 {
    GLuint i;
 
-   for (i = 0; i < Elements(st->drawpix.shaders); i++) {
-      if (st->drawpix.shaders[i])
-         _mesa_reference_fragprog(st->ctx, &st->drawpix.shaders[i], NULL);
+   for (i = 0; i < ARRAY_SIZE(st->drawpix.zs_shaders); i++) {
+      if (st->drawpix.zs_shaders[i])
+         cso_delete_fragment_shader(st->cso_context,
+                                    st->drawpix.zs_shaders[i]);
    }
 
-   st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
    if (st->drawpix.vert_shaders[0])
       cso_delete_vertex_shader(st->cso_context, st->drawpix.vert_shaders[0]);
    if (st->drawpix.vert_shaders[1])