Merge branch 'glsl-to-tgsi'
[mesa.git] / src / mesa / state_tracker / st_cb_drawpixels.c
index 69a3dd45e80a49558e869f2c7f2ae9035211e9e5..390c518699f62651bdfce6f279397ee9ba160b72 100644 (file)
 #include "main/image.h"
 #include "main/bufferobj.h"
 #include "main/macros.h"
+#include "main/mfeatures.h"
+#include "main/mtypes.h"
+#include "main/pack.h"
+#include "main/pbo.h"
 #include "main/texformat.h"
+#include "main/teximage.h"
 #include "main/texstore.h"
 #include "program/program.h"
 #include "program/prog_print.h"
+#include "program/prog_instruction.h"
 
-#include "st_debug.h"
-#include "st_context.h"
 #include "st_atom.h"
 #include "st_atom_constbuf.h"
-#include "st_program.h"
 #include "st_cb_drawpixels.h"
 #include "st_cb_readpixels.h"
 #include "st_cb_fbo.h"
+#include "st_context.h"
+#include "st_debug.h"
 #include "st_format.h"
+#include "st_program.h"
 #include "st_texture.h"
 
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
-#include "util/u_inlines.h"
 #include "tgsi/tgsi_ureg.h"
-#include "util/u_tile.h"
 #include "util/u_draw_quad.h"
 #include "util/u_format.h"
+#include "util/u_inlines.h"
 #include "util/u_math.h"
-#include "program/prog_instruction.h"
+#include "util/u_tile.h"
 #include "cso_cache/cso_context.h"
 
 
@@ -89,93 +94,117 @@ is_passthrough_program(const struct gl_fragment_program *prog)
 }
 
 
+/**
+ * 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.
- * \return pointer to Gallium driver fragment shader
+ * \param fpIn  the current/incoming fragment program
+ * \param fpOut  returns the combined fragment program
  */
-static void *
-combined_drawpix_fragment_program(GLcontext *ctx)
+void
+st_make_drawpix_fragment_program(struct st_context *st,
+                                 struct gl_fragment_program *fpIn,
+                                 struct gl_fragment_program **fpOut)
 {
-   struct st_context *st = st_context(ctx);
-   struct st_fragment_program *stfp;
+   struct gl_program *newProg;
+   struct st_fragment_program *stfp = (struct st_fragment_program *) fpIn;
 
-   if (st->pixel_xfer.program->serialNo == st->pixel_xfer.xfer_prog_sn
-       && st->fp->serialNo == st->pixel_xfer.user_prog_sn) {
-      /* the pixel tranfer program has not changed and the user-defined
-       * program has not changed, so re-use the combined program.
-       */
-      stfp = st->pixel_xfer.combined_prog;
+   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 {
-      /* Concatenate the pixel transfer program with the current user-
-       * defined program.
-       */
-      if (is_passthrough_program(&st->fp->Base)) {
-         stfp = (struct st_fragment_program *)
-            _mesa_clone_fragment_program(ctx, &st->pixel_xfer.program->Base);
-      }
-      else {
 #if 0
-         printf("Base program:\n");
-         _mesa_print_program(&st->fp->Base.Base);
-         printf("DrawPix program:\n");
-         _mesa_print_program(&st->pixel_xfer.program->Base.Base);
+      /* debug */
+      printf("Base program:\n");
+      _mesa_print_program(&fpIn->Base);
+      printf("DrawPix program:\n");
+      _mesa_print_program(&st->pixel_xfer.program->Base.Base);
 #endif
-         stfp = (struct st_fragment_program *)
-            _mesa_combine_programs(ctx,
-                                   &st->pixel_xfer.program->Base.Base,
-                                   &st->fp->Base.Base);
-      }
+      newProg = _mesa_combine_programs(st->ctx,
+                                       &st->pixel_xfer.program->Base.Base,
+                                       &fpIn->Base);
+   }
 
 #if 0
-      {
-         struct gl_program *p = &stfp->Base.Base;
-         printf("Combined DrawPixels 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);
-      }
+   /* 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
 
-      /* translate to TGSI tokens */
-      st_translate_fragment_program(st, stfp);
-
-      /* save new program, update serial numbers */
-      st->pixel_xfer.xfer_prog_sn = st->pixel_xfer.program->serialNo;
-      st->pixel_xfer.user_prog_sn = st->fp->serialNo;
-      st->pixel_xfer.combined_prog_sn = stfp->serialNo;
-      /* can't reference new program directly, already have a reference on it */
-      st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
-      st->pixel_xfer.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 glDrawPixels.
-    */
-   st_upload_constants(st, stfp->Base.Base.Parameters, PIPE_SHADER_FRAGMENT);
-
-   return stfp->driver_shader;
+   *fpOut = (struct gl_fragment_program *) newProg;
 }
 
 
 /**
- * Create fragment shader that does a TEX() instruction to get a Z
- * value, then writes to FRAG_RESULT_DEPTH.
+ * 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 Gallium driver fragment shader
+ * \return pointer to the gl_fragment program
  */
-static void *
-make_fragment_shader_z(struct st_context *st)
+struct gl_fragment_program *
+st_make_drawpix_z_stencil_program(struct st_context *st,
+                                  GLboolean write_depth,
+                                  GLboolean write_stencil)
 {
-   GLcontext *ctx = st->ctx;
+   struct gl_context *ctx = st->ctx;
    struct gl_program *p;
+   struct gl_fragment_program *fp;
    GLuint ic = 0;
+   const GLuint shaderIndex = write_depth * 2 + write_stencil;
+
+   assert(shaderIndex < Elements(st->drawpix.shaders));
 
-   if (st->drawpix.z_shader) {
-      return st->drawpix.z_shader->driver_shader;
+   if (st->drawpix.shaders[shaderIndex]) {
+      /* already have the proper shader */
+      return st->drawpix.shaders[shaderIndex];
    }
 
    /*
@@ -185,7 +214,8 @@ make_fragment_shader_z(struct st_context *st)
    if (!p)
       return NULL;
 
-   p->NumInstructions = 3;
+   p->NumInstructions = write_depth ? 2 : 1;
+   p->NumInstructions += write_stencil ? 1 : 0;
 
    p->Instructions = _mesa_alloc_instructions(p->NumInstructions);
    if (!p->Instructions) {
@@ -194,24 +224,31 @@ make_fragment_shader_z(struct st_context *st)
    }
    _mesa_init_instructions(p->Instructions, p->NumInstructions);
 
-   /* 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++;
+   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++;
+   }
+
+   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++;
+   }
 
    /* END; */
    p->Instructions[ic++].Opcode = OPCODE_END;
@@ -219,15 +256,23 @@ make_fragment_shader_z(struct st_context *st)
    assert(ic == p->NumInstructions);
 
    p->InputsRead = FRAG_BIT_TEX0 | FRAG_BIT_COL0;
-   p->OutputsWritten = (1 << FRAG_RESULT_COLOR) | (1 << FRAG_RESULT_DEPTH);
-   p->SamplersUsed = 0x1;  /* sampler 0 (bit 0) is used */
+   p->OutputsWritten = 0;
+   if (write_depth)
+      p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_DEPTH);
+   if (write_stencil)
+      p->OutputsWritten |= BITFIELD64_BIT(FRAG_RESULT_STENCIL);
 
-   st->drawpix.z_shader = (struct st_fragment_program *) p;
-   st_translate_fragment_program(st, st->drawpix.z_shader);
+   p->SamplersUsed =  0x1;  /* sampler 0 (bit 0) is used */
+   if (write_stencil)
+      p->SamplersUsed |= 1 << 1;
 
-   return st->drawpix.z_shader->driver_shader;
-}
+   fp = (struct gl_fragment_program *) p;
 
+   /* save the new shader */
+   st->drawpix.shaders[shaderIndex] = fp;
+
+   return fp;
+}
 
 
 /**
@@ -239,8 +284,7 @@ make_passthrough_vertex_shader(struct st_context *st,
                                GLboolean passColor)
 {
    if (!st->drawpix.vert_shaders[passColor]) {
-      struct ureg_program *ureg = 
-         ureg_create( TGSI_PROCESSOR_VERTEX );
+      struct ureg_program *ureg = ureg_create( TGSI_PROCESSOR_VERTEX );
 
       if (ureg == NULL)
          return NULL;
@@ -274,20 +318,120 @@ make_passthrough_vertex_shader(struct st_context *st,
 
 /**
  * Return a texture internalFormat for drawing/copying an image
- * of the given type.
+ * of the given format and type.
  */
 static GLenum
-base_format(GLenum format)
+internal_format(struct gl_context *ctx, GLenum format, GLenum type)
 {
    switch (format) {
    case GL_DEPTH_COMPONENT:
-      return GL_DEPTH_COMPONENT;
+      switch (type) {
+      case GL_UNSIGNED_SHORT:
+         return GL_DEPTH_COMPONENT16;
+
+      case GL_UNSIGNED_INT:
+         return GL_DEPTH_COMPONENT32;
+
+      case GL_FLOAT:
+         if (ctx->Extensions.ARB_depth_buffer_float)
+            return GL_DEPTH_COMPONENT32F;
+         else
+            return GL_DEPTH_COMPONENT;
+
+      default:
+         return GL_DEPTH_COMPONENT;
+      }
+
    case GL_DEPTH_STENCIL:
-      return GL_DEPTH_STENCIL;
+      switch (type) {
+      case GL_FLOAT_32_UNSIGNED_INT_24_8_REV:
+         return GL_DEPTH32F_STENCIL8;
+
+      case GL_UNSIGNED_INT_24_8:
+      default:
+         return GL_DEPTH24_STENCIL8;
+      }
+
    case GL_STENCIL_INDEX:
       return GL_STENCIL_INDEX;
+
    default:
-      return GL_RGBA;
+      if (_mesa_is_integer_format(format)) {
+         switch (type) {
+         case GL_BYTE:
+            return GL_RGBA8I;
+         case GL_UNSIGNED_BYTE:
+            return GL_RGBA8UI;
+         case GL_SHORT:
+            return GL_RGBA16I;
+         case GL_UNSIGNED_SHORT:
+            return GL_RGBA16UI;
+         case GL_INT:
+            return GL_RGBA32I;
+         case GL_UNSIGNED_INT:
+            return GL_RGBA32UI;
+         default:
+            assert(0 && "Unexpected type in internal_format()");
+            return GL_RGBA_INTEGER;
+         }
+      }
+      else {
+         switch (type) {
+         case GL_UNSIGNED_BYTE:
+         case GL_UNSIGNED_INT_8_8_8_8:
+         case GL_UNSIGNED_INT_8_8_8_8_REV:
+         default:
+            return GL_RGBA8;
+
+         case GL_UNSIGNED_BYTE_3_3_2:
+         case GL_UNSIGNED_BYTE_2_3_3_REV:
+         case GL_UNSIGNED_SHORT_4_4_4_4:
+         case GL_UNSIGNED_SHORT_4_4_4_4_REV:
+            return GL_RGBA4;
+
+         case GL_UNSIGNED_SHORT_5_6_5:
+         case GL_UNSIGNED_SHORT_5_6_5_REV:
+         case GL_UNSIGNED_SHORT_5_5_5_1:
+         case GL_UNSIGNED_SHORT_1_5_5_5_REV:
+            return GL_RGB5_A1;
+
+         case GL_UNSIGNED_INT_10_10_10_2:
+         case GL_UNSIGNED_INT_2_10_10_10_REV:
+            return GL_RGB10_A2;
+
+         case GL_UNSIGNED_SHORT:
+         case GL_UNSIGNED_INT:
+            return GL_RGBA16;
+
+         case GL_BYTE:
+            return
+               ctx->Extensions.EXT_texture_snorm ? GL_RGBA8_SNORM : GL_RGBA8;
+
+         case GL_SHORT:
+         case GL_INT:
+            return
+               ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
+
+         case GL_HALF_FLOAT_ARB:
+            return
+               ctx->Extensions.ARB_texture_float ? GL_RGBA16F :
+               ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
+
+         case GL_FLOAT:
+         case GL_DOUBLE:
+            return
+               ctx->Extensions.ARB_texture_float ? GL_RGBA32F :
+               ctx->Extensions.EXT_texture_snorm ? GL_RGBA16_SNORM : GL_RGBA16;
+
+         case GL_UNSIGNED_INT_5_9_9_9_REV:
+            assert(ctx->Extensions.EXT_texture_shared_exponent);
+            return GL_RGB9_E5;
+
+         case GL_UNSIGNED_INT_10F_11F_11F_REV:
+            assert(ctx->Extensions.EXT_packed_float);
+            return GL_R11F_G11F_B10F;
+         }
+      }
    }
 }
 
@@ -301,37 +445,10 @@ static struct pipe_resource *
 alloc_texture(struct st_context *st, GLsizei width, GLsizei height,
               enum pipe_format texFormat)
 {
-   struct pipe_context *pipe = st->pipe;
-   struct pipe_screen *screen = pipe->screen;
    struct pipe_resource *pt;
-   int ptw, pth;
 
-   ptw = width;
-   pth = height;
-
-   /* Need to use POT texture? */
-   if (!screen->get_param(screen, PIPE_CAP_NPOT_TEXTURES)) {
-      int l2pt, maxSize;
-
-      l2pt = util_logbase2(width);
-      if (1 << l2pt != width) {
-         ptw = 1 << (l2pt + 1);
-      }
-
-      l2pt = util_logbase2(height);
-      if (1 << l2pt != height) {
-         pth = 1 << (l2pt + 1);
-      }
-
-      /* Check against maximum texture size */
-      maxSize = 1 << (pipe->screen->get_param(pipe->screen,
-                               PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
-      assert(ptw <= maxSize);
-      assert(pth <= maxSize);
-   }
-
-   pt = st_texture_create(st, PIPE_TEXTURE_2D, texFormat, 0,
-                          ptw, pth, 1, PIPE_BIND_SAMPLER_VIEW);
+   pt = st_texture_create(st, st->internal_target, texFormat, 0,
+                          width, height, 1, 1, PIPE_BIND_SAMPLER_VIEW);
 
    return pt;
 }
@@ -347,22 +464,22 @@ make_texture(struct st_context *st,
             const struct gl_pixelstore_attrib *unpack,
             const GLvoid *pixels)
 {
-   GLcontext *ctx = st->ctx;
+   struct gl_context *ctx = st->ctx;
    struct pipe_context *pipe = st->pipe;
    gl_format mformat;
    struct pipe_resource *pt;
    enum pipe_format pipeFormat;
-   GLuint cpp;
-   GLenum baseFormat;
+   GLenum baseInternalFormat, intFormat;
 
-   baseFormat = base_format(format);
+   intFormat = internal_format(ctx, format, type);
+   baseInternalFormat = _mesa_base_tex_format(ctx, intFormat);
 
-   mformat = st_ChooseTextureFormat(ctx, baseFormat, format, type);
+   mformat = st_ChooseTextureFormat_renderable(ctx, intFormat,
+                                               format, type, GL_FALSE);
    assert(mformat);
 
    pipeFormat = st_mesa_format_to_pipe_format(mformat);
    assert(pipeFormat);
-   cpp = util_format_get_blocksize(pipeFormat);
 
    pixels = _mesa_map_pbo_source(ctx, unpack, pixels);
    if (!pixels)
@@ -385,9 +502,9 @@ 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, 0,
-                                             PIPE_TRANSFER_WRITE, 0, 0,
-                                             width, height);
+      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);
@@ -398,7 +515,7 @@ make_texture(struct st_context *st,
        * the texture.  We deal with that with texcoords.
        */
       success = _mesa_texstore(ctx, 2,           /* dims */
-                               baseFormat,       /* baseInternalFormat */
+                               baseInternalFormat, /* baseInternalFormat */
                                mformat,          /* gl_format */
                                dest,             /* dest */
                                0, 0, 0,          /* dstX/Y/Zoffset */
@@ -432,7 +549,7 @@ make_texture(struct st_context *st,
  * \param invertTex  if true, flip texcoords vertically
  */
 static void
-draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z,
+draw_quad(struct gl_context *ctx, GLfloat x0, GLfloat y0, GLfloat z,
           GLfloat x1, GLfloat y1, const GLfloat *color,
           GLboolean invertTex, GLfloat maxXcoord, GLfloat maxYcoord)
 {
@@ -508,10 +625,11 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z,
       /* allocate/load buffer object with vertex data */
       buf = pipe_buffer_create(pipe->screen,
                               PIPE_BIND_VERTEX_BUFFER,
+                              PIPE_USAGE_STATIC,
                                sizeof(verts));
       pipe_buffer_write(st->pipe, buf, 0, sizeof(verts), verts);
 
-      util_draw_vertex_buffer(pipe, buf, 0,
+      util_draw_vertex_buffer(pipe, st->cso_context, buf, 0,
                               PIPE_PRIM_QUADS,
                               4,  /* verts */
                               3); /* attribs/vert */
@@ -522,26 +640,30 @@ draw_quad(GLcontext *ctx, GLfloat x0, GLfloat y0, GLfloat z,
 
 
 static void
-draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
+draw_textured_quad(struct gl_context *ctx, GLint x, GLint y, GLfloat z,
                    GLsizei width, GLsizei height,
                    GLfloat zoomX, GLfloat zoomY,
-                   struct pipe_sampler_view *sv,
+                   struct pipe_sampler_view **sv,
+                   int num_sampler_view,
                    void *driver_vp,
                    void *driver_fp,
                    const GLfloat *color,
-                   GLboolean invertTex)
+                   GLboolean invertTex,
+                   GLboolean write_depth, GLboolean write_stencil)
 {
    struct st_context *st = st_context(ctx);
    struct pipe_context *pipe = st->pipe;
    struct cso_context *cso = st->cso_context;
    GLfloat x0, y0, x1, y1;
    GLsizei maxSize;
+   boolean normalized = sv[0]->texture->target != PIPE_TEXTURE_RECT;
 
    /* 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);
+   maxSize = 1 << (pipe->screen->get_param(pipe->screen,
+                                        PIPE_CAP_MAX_TEXTURE_2D_LEVELS) - 1);
    assert(width <= maxSize);
    assert(height <= maxSize);
 
@@ -552,16 +674,48 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
    cso_save_fragment_shader(cso);
    cso_save_vertex_shader(cso);
    cso_save_vertex_elements(cso);
+   cso_save_vertex_buffers(cso);
+   if (write_stencil) {
+      cso_save_depth_stencil_alpha(cso);
+      cso_save_blend(cso);
+   }
 
    /* rasterizer state: just scissor */
    {
       struct pipe_rasterizer_state rasterizer;
       memset(&rasterizer, 0, sizeof(rasterizer));
+      rasterizer.clamp_fragment_color = ctx->Color._ClampFragmentColor;
       rasterizer.gl_rasterization_rules = 1;
       rasterizer.scissor = ctx->Scissor.Enabled;
       cso_set_rasterizer(cso, &rasterizer);
    }
 
+   if (write_stencil) {
+      /* Stencil writing bypasses the normal fragment pipeline to
+       * disable color writing and set stencil test to always pass.
+       */
+      struct pipe_depth_stencil_alpha_state dsa;
+      struct pipe_blend_state blend;
+
+      /* depth/stencil */
+      memset(&dsa, 0, sizeof(dsa));
+      dsa.stencil[0].enabled = 1;
+      dsa.stencil[0].func = PIPE_FUNC_ALWAYS;
+      dsa.stencil[0].writemask = ctx->Stencil.WriteMask[0] & 0xff;
+      dsa.stencil[0].zpass_op = PIPE_STENCIL_OP_REPLACE;
+      if (write_depth) {
+         /* writing depth+stencil: depth test always passes */
+         dsa.depth.enabled = 1;
+         dsa.depth.writemask = ctx->Depth.Mask;
+         dsa.depth.func = PIPE_FUNC_ALWAYS;
+      }
+      cso_set_depth_stencil_alpha(cso, &dsa);
+
+      /* blend (colormask) */
+      memset(&blend, 0, sizeof(blend));
+      cso_set_blend(cso, &blend);
+   }
+
    /* fragment shader state: TEX lookup program */
    cso_set_fragment_shader_handle(cso, driver_fp);
 
@@ -579,10 +733,10 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
       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;
+      sampler.normalized_coords = normalized;
 
       cso_single_sampler(cso, 0, &sampler);
-      if (st->pixel_xfer.pixelmap_enabled) {
+      if (num_sampler_view > 1) {
          cso_single_sampler(cso, 1, &sampler);
       }
       cso_single_sampler_done(cso);
@@ -607,15 +761,7 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
    cso_set_vertex_elements(cso, 3, st->velems_util_draw);
 
    /* texture state: */
-   if (st->pixel_xfer.pixelmap_enabled) {
-      struct pipe_sampler_view *sampler_views[2];
-      sampler_views[0] = sv;
-      sampler_views[1] = st->pixel_xfer.pixelmap_sampler_view;
-      cso_set_fragment_sampler_views(cso, 2, sampler_views);
-   }
-   else {
-      cso_set_fragment_sampler_views(cso, 1, &sv);
-   }
+   cso_set_fragment_sampler_views(cso, num_sampler_view, sv);
 
    /* Compute Gallium window coords (y=0=top) with pixel zoom.
     * Recall that these coords are transformed by the current
@@ -635,8 +781,8 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
    z = z * 2.0 - 1.0;
 
    draw_quad(ctx, x0, y0, z, x1, y1, color, invertTex,
-             (GLfloat) width / sv->texture->width0,
-             (GLfloat) height / sv->texture->height0);
+             normalized ? ((GLfloat) width / sv[0]->texture->width0) : (GLfloat)width,
+             normalized ? ((GLfloat) height / sv[0]->texture->height0) : (GLfloat)height);
 
    /* restore state */
    cso_restore_rasterizer(cso);
@@ -646,11 +792,20 @@ draw_textured_quad(GLcontext *ctx, GLint x, GLint y, GLfloat z,
    cso_restore_fragment_shader(cso);
    cso_restore_vertex_shader(cso);
    cso_restore_vertex_elements(cso);
+   cso_restore_vertex_buffers(cso);
+   if (write_stencil) {
+      cso_restore_depth_stencil_alpha(cso);
+      cso_restore_blend(cso);
+   }
 }
 
 
+/**
+ * Software fallback to do glDrawPixels(GL_STENCIL_INDEX) when we
+ * can't use a fragment shader to write stencil values.
+ */
 static void
-draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
+draw_stencil_pixels(struct gl_context *ctx, GLint x, GLint y,
                     GLsizei width, GLsizei height, GLenum format, GLenum type,
                     const struct gl_pixelstore_attrib *unpack,
                     const GLvoid *pixels)
@@ -681,14 +836,16 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
    }
 
    if(format != GL_DEPTH_STENCIL && 
-      util_format_get_component_bits(strb->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0)
+      util_format_get_component_bits(strb->format,
+                                     UTIL_FORMAT_COLORSPACE_ZS, 0) != 0)
       usage = PIPE_TRANSFER_READ_WRITE;
    else
       usage = PIPE_TRANSFER_WRITE;
 
-   pt = pipe_get_transfer(st_context(ctx)->pipe, strb->texture, 0, 0, 0,
-                                      usage, x, y,
-                                      width, height);
+   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);
 
@@ -704,6 +861,7 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
       for (row = 0; row < height; row++) {
          GLubyte sValues[MAX_WIDTH];
          GLuint zValues[MAX_WIDTH];
+         GLfloat *zValuesFloat = (GLfloat*)zValues;
          GLenum destType = GL_UNSIGNED_BYTE;
          const GLvoid *source = _mesa_image_address2d(&clippedUnpack, pixels,
                                                       width, height,
@@ -714,7 +872,11 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
                                    ctx->_ImageTransferState);
 
          if (format == GL_DEPTH_STENCIL) {
-            _mesa_unpack_depth_span(ctx, spanWidth, GL_UNSIGNED_INT, zValues,
+            GLenum ztype =
+               pt->resource->format == PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED ?
+               GL_FLOAT : GL_UNSIGNED_INT;
+
+            _mesa_unpack_depth_span(ctx, spanWidth, ztype, zValues,
                                     (1 << 24) - 1, type, source,
                                     &clippedUnpack);
          }
@@ -779,6 +941,26 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
                   }
                }
                break;
+            case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED:
+               if (format == GL_DEPTH_STENCIL) {
+                  uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
+                  GLfloat *destf = (GLfloat*)dest;
+                  GLint k;
+                  assert(usage == PIPE_TRANSFER_WRITE);
+                  for (k = 0; k < spanWidth; k++) {
+                     destf[k*2] = zValuesFloat[k];
+                     dest[k*2+1] = sValues[k] & 0xff;
+                  }
+               }
+               else {
+                  uint *dest = (uint *) (stmap + spanY * pt->stride + spanX*4);
+                  GLint k;
+                  assert(usage == PIPE_TRANSFER_READ_WRITE);
+                  for (k = 0; k < spanWidth; k++) {
+                     dest[k*2+1] = sValues[k] & 0xff;
+                  }
+               }
+               break;
             default:
                assert(0);
             }
@@ -795,20 +977,89 @@ draw_stencil_pixels(GLcontext *ctx, GLint x, GLint y,
 }
 
 
+/**
+ * Get fragment program variant for a glDrawPixels or glCopyPixels
+ * command for RGBA data.
+ */
+static struct st_fp_variant *
+get_color_fp_variant(struct st_context *st)
+{
+   struct gl_context *ctx = st->ctx;
+   struct st_fp_variant_key key;
+   struct st_fp_variant *fpv;
+
+   memset(&key, 0, sizeof(key));
+
+   key.st = st;
+   key.drawpixels = 1;
+   key.scaleAndBias = (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);
+   key.pixelMaps = ctx->Pixel.MapColorFlag;
+
+   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;
+
+   fpv = st_get_fp_variant(st, st->fp, &key);
+
+   return fpv;
+}
+
+
 /**
  * Called via ctx->Driver.DrawPixels()
  */
 static void
-st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
+st_DrawPixels(struct gl_context *ctx, GLint x, GLint y,
+              GLsizei width, GLsizei height,
               GLenum format, GLenum type,
               const struct gl_pixelstore_attrib *unpack, const GLvoid *pixels)
 {
    void *driver_vp, *driver_fp;
    struct st_context *st = st_context(ctx);
    const GLfloat *color;
-
-   if (format == GL_STENCIL_INDEX ||
-       format == GL_DEPTH_STENCIL) {
+   struct pipe_context *pipe = st->pipe;
+   GLboolean write_stencil = GL_FALSE, write_depth = GL_FALSE;
+   struct pipe_sampler_view *sv[2];
+   int num_sampler_view = 1;
+   struct st_fp_variant *fpv;
+
+   if (format == GL_DEPTH_STENCIL)
+      write_stencil = write_depth = GL_TRUE;
+   else if (format == GL_STENCIL_INDEX)
+      write_stencil = GL_TRUE;
+   else if (format == GL_DEPTH_COMPONENT)
+      write_depth = GL_TRUE;
+
+   if (write_stencil &&
+       !pipe->screen->get_param(pipe->screen, PIPE_CAP_SHADER_STENCIL_EXPORT)) {
+      /* software fallback */
       draw_stencil_pixels(ctx, x, y, width, height, format, type,
                           unpack, pixels);
       return;
@@ -819,32 +1070,85 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
 
    st_validate_state(st);
 
-   if (format == GL_DEPTH_COMPONENT) {
-      driver_fp = make_fragment_shader_z(st);
+   /*
+    * 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_vp = make_passthrough_vertex_shader(st, GL_TRUE);
+
       color = ctx->Current.RasterColor;
    }
    else {
-      driver_fp = combined_drawpix_fragment_program(ctx);
+      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++;
+      }
    }
 
+   /* update fragment program constants */
+   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
+
    /* draw with textured quad */
    {
       struct pipe_resource *pt
          = make_texture(st, width, height, format, type, unpack, pixels);
       if (pt) {
-         struct pipe_sampler_view *sv = st_create_texture_sampler_view(st->pipe, 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 = PIPE_FORMAT_NONE;
+
+               switch (pt->format) {
+               case PIPE_FORMAT_Z24_UNORM_S8_USCALED:
+               case PIPE_FORMAT_X24S8_USCALED:
+                  stencil_format = PIPE_FORMAT_X24S8_USCALED;
+                  break;
+               case PIPE_FORMAT_S8_USCALED_Z24_UNORM:
+               case PIPE_FORMAT_S8X24_USCALED:
+                  stencil_format = PIPE_FORMAT_S8X24_USCALED;
+                  break;
+               case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED:
+               case PIPE_FORMAT_X32_S8X24_USCALED:
+                  stencil_format = PIPE_FORMAT_X32_S8X24_USCALED;
+                  break;
+               case PIPE_FORMAT_S8_USCALED:
+                  stencil_format = PIPE_FORMAT_S8_USCALED;
+                  break;
+               default:
+                  assert(0);
+               }
+
+              sv[1] = st_create_texture_sampler_view_format(st->pipe, pt,
+                                                             stencil_format);
+              num_sampler_view++;
+           }
 
-         if (sv) {
             draw_textured_quad(ctx, x, y, ctx->Current.RasterPos[2],
-                               width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
+                               width, height,
+                               ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
                                sv,
-                               driver_vp, 
+                               num_sampler_view,
+                               driver_vp,
                                driver_fp,
-                               color, GL_FALSE);
-            pipe_sampler_view_reference(&sv, NULL);
+                               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);
       }
@@ -853,12 +1157,15 @@ st_DrawPixels(GLcontext *ctx, GLint x, GLint y, GLsizei width, GLsizei height,
 
 
 
+/**
+ * Software fallback for glCopyPixels(GL_STENCIL).
+ */
 static void
-copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
+copy_stencil_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
                     GLsizei width, GLsizei height,
                     GLint dstx, GLint dsty)
 {
-   struct st_renderbuffer *rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer);
+   struct st_renderbuffer *rbDraw;
    struct pipe_context *pipe = st_context(ctx)->pipe;
    enum pipe_transfer_usage usage;
    struct pipe_transfer *ptDraw;
@@ -872,24 +1179,46 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
       return;
    }
 
+   /* Get the dest renderbuffer.  If there's a wrapper, use the
+    * underlying renderbuffer.
+    */
+   rbDraw = st_renderbuffer(ctx->DrawBuffer->_StencilBuffer);
+   if (rbDraw->Base.Wrapped)
+      rbDraw = st_renderbuffer(rbDraw->Base.Wrapped);
+
    /* this will do stencil pixel transfer ops */
    st_read_stencil_pixels(ctx, srcx, srcy, width, height,
                           GL_STENCIL_INDEX, GL_UNSIGNED_BYTE,
                           &ctx->DefaultPacking, buffer);
 
-   if(util_format_get_component_bits(rbDraw->format, UTIL_FORMAT_COLORSPACE_ZS, 0) != 0)
+   if (0) {
+      /* debug code: dump stencil values */
+      GLint row, col;
+      for (row = 0; row < height; row++) {
+         printf("%3d: ", row);
+         for (col = 0; col < width; col++) {
+            printf("%02x ", buffer[col + row * width]);
+         }
+         printf("\n");
+      }
+   }
+
+   if (util_format_get_component_bits(rbDraw->format,
+                                     UTIL_FORMAT_COLORSPACE_ZS, 0) != 0)
       usage = PIPE_TRANSFER_READ_WRITE;
    else
       usage = PIPE_TRANSFER_WRITE;
-   
+
    if (st_fb_orientation(ctx->DrawBuffer) == Y_0_TOP) {
       dsty = rbDraw->Base.Height - dsty - height;
    }
 
-   ptDraw = pipe_get_transfer(st_context(ctx)->pipe,
-                                          rbDraw->texture, 0, 0, 0,
-                                          usage, dstx, dsty,
-                                          width, 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);
@@ -940,6 +1269,18 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
          assert(usage == PIPE_TRANSFER_WRITE);
          memcpy(dst, src, width);
          break;
+      case PIPE_FORMAT_Z32_FLOAT_S8X24_USCALED:
+         {
+            uint *dst4 = (uint *) dst;
+            int j;
+            dst4++;
+            assert(usage == PIPE_TRANSFER_READ_WRITE);
+            for (j = 0; j < width; j++) {
+               *dst4 = src[j] & 0xff;
+               dst4 += 2;
+            }
+         }
+         break;
       default:
          assert(0);
       }
@@ -953,8 +1294,115 @@ copy_stencil_pixels(GLcontext *ctx, GLint srcx, GLint srcy,
 }
 
 
+/** 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().
+ *
+ * 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.
+ */
+static GLboolean
+blit_copy_pixels(struct gl_context *ctx, GLint srcx, GLint srcy,
+                 GLsizei width, GLsizei height,
+                 GLint dstx, GLint dsty, GLenum type)
+{
+   struct st_context *st = st_context(ctx);
+   struct pipe_context *pipe = st->pipe;
+   struct gl_pixelstore_attrib pack, unpack;
+   GLint readX, readY, readW, readH;
+
+   if (type == GL_COLOR &&
+       ctx->Pixel.ZoomX == 1.0 &&
+       ctx->Pixel.ZoomY == 1.0 &&
+       ctx->_ImageTransferState == 0x0 &&
+       !ctx->Color.BlendEnabled &&
+       !ctx->Color.AlphaEnabled &&
+       !ctx->Depth.Test &&
+       !ctx->Fog.Enabled &&
+       !ctx->Stencil.Enabled &&
+       !ctx->FragmentProgram.Enabled &&
+       !ctx->VertexProgram.Enabled &&
+       !ctx->Shader.CurrentFragmentProgram &&
+       st_fb_orientation(ctx->ReadBuffer) == st_fb_orientation(ctx->DrawBuffer) &&
+       ctx->DrawBuffer->_NumColorDrawBuffers == 1 &&
+       !ctx->Query.CondRenderQuery) {
+      struct st_renderbuffer *rbRead, *rbDraw;
+      GLint drawX, drawY;
+
+      /*
+       * Clip the read region against the src buffer bounds.
+       * We'll still allocate a temporary buffer/texture for the original
+       * src region size but we'll only read the region which is on-screen.
+       * This may mean that we draw garbage pixels into the dest region, but
+       * that's expected.
+       */
+      readX = srcx;
+      readY = srcy;
+      readW = width;
+      readH = height;
+      pack = ctx->DefaultPacking;
+      if (!_mesa_clip_readpixels(ctx, &readX, &readY, &readW, &readH, &pack))
+         return GL_TRUE; /* all done */
+
+      /* clip against dest buffer bounds and scissor box */
+      drawX = dstx + pack.SkipPixels;
+      drawY = dsty + pack.SkipRows;
+      unpack = pack;
+      if (!_mesa_clip_drawpixels(ctx, &drawX, &drawY, &readW, &readH, &unpack))
+         return GL_TRUE; /* all done */
+
+      readX = readX - pack.SkipPixels + unpack.SkipPixels;
+      readY = readY - pack.SkipRows + unpack.SkipRows;
+
+      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;
+         }
+
+         u_box_2d(readX, readY, readW, readH, &srcBox);
+
+         pipe->resource_copy_region(pipe,
+                                    rbDraw->texture,
+                                    rbDraw->rtt_level, drawX, drawY, 0,
+                                    rbRead->texture,
+                                    rbRead->rtt_level, &srcBox);
+         return GL_TRUE;
+      }
+   }
+
+   return GL_FALSE;
+}
+
+
 static void
-st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
+st_CopyPixels(struct gl_context *ctx, GLint srcx, GLint srcy,
               GLsizei width, GLsizei height,
               GLint dstx, GLint dsty, GLenum type)
 {
@@ -964,13 +1412,15 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
    struct st_renderbuffer *rbRead;
    void *driver_vp, *driver_fp;
    struct pipe_resource *pt;
-   struct pipe_sampler_view *sv;
+   struct pipe_sampler_view *sv[2];
+   int num_sampler_view = 1;
    GLfloat *color;
    enum pipe_format srcFormat, texFormat;
    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);
 
@@ -980,20 +1430,52 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
       return;
    }
 
+   if (blit_copy_pixels(ctx, srcx, srcy, width, height, dstx, dsty, type))
+      return;
+
+   /*
+    * The subsequent code implements glCopyPixels by copying the source
+    * pixels into a temporary texture that's then applied to a textured quad.
+    * When we draw the textured quad, all the usual per-fragment operations
+    * are handled.
+    */
+
+
+   /*
+    * Get vertex/fragment shaders
+    */
    if (type == GL_COLOR) {
       rbRead = st_get_color_read_renderbuffer(ctx);
       color = NULL;
-      driver_fp = combined_drawpix_fragment_program(ctx);
+
+      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++;
+      }
    }
    else {
       assert(type == GL_DEPTH);
       rbRead = st_renderbuffer(ctx->ReadBuffer->_DepthBuffer);
       color = ctx->Current.Attrib[VERT_ATTRIB_COLOR0];
-      driver_fp = make_fragment_shader_z(st);
+
+      fpv = get_depth_stencil_fp_variant(st, GL_TRUE, GL_FALSE);
+      driver_fp = fpv->driver_shader;
+
       driver_vp = make_passthrough_vertex_shader(st, GL_TRUE);
    }
 
+   /* update fragment program constants */
+   st_upload_constants(st, fpv->parameters, PIPE_SHADER_FRAGMENT);
+
+
+   if (rbRead->Base.Wrapped)
+      rbRead = st_renderbuffer(rbRead->Base.Wrapped);
+
    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
@@ -1002,21 +1484,23 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
 
    srcFormat = rbRead->texture->format;
 
-   if (screen->is_format_supported(screen, srcFormat, PIPE_TEXTURE_2D, sample_count,
-                                   PIPE_BIND_SAMPLER_VIEW, 0)) {
+   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 (type == GL_DEPTH) {
          texFormat = st_choose_format(screen, GL_DEPTH_COMPONENT,
-                                      PIPE_TEXTURE_2D, sample_count,
-                                      PIPE_BIND_DEPTH_STENCIL);
+                                      GL_NONE, GL_NONE, st->internal_target,
+                                     sample_count, PIPE_BIND_DEPTH_STENCIL);
          assert(texFormat != PIPE_FORMAT_NONE);
       }
       else {
          /* default color format */
-         texFormat = st_choose_format(screen, GL_RGBA, PIPE_TEXTURE_2D, 
+         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);
       }
@@ -1047,8 +1531,8 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
    if (!pt)
       return;
 
-   sv = st_create_texture_sampler_view(st->pipe, pt);
-   if (!sv) {
+   sv[0] = st_create_texture_sampler_view(st->pipe, pt);
+   if (!sv[0]) {
       pipe_resource_reference(&pt, NULL);
       return;
    }
@@ -1056,27 +1540,26 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
    /* Make temporary texture which is a copy of the src region.
     */
    if (srcFormat == texFormat) {
-      struct pipe_subresource srcsub, dstsub;
-      srcsub.face = 0;
-      srcsub.level = 0;
-      dstsub.face = 0;
-      dstsub.level = 0;
+      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 */
-                                 dstsub,
+                                 0,                                 /* dest lvl */
                                  pack.SkipPixels, pack.SkipRows, 0, /* dest pos */
                                  rbRead->texture,                   /* src tex */
-                                 srcsub,
-                                 readX, readY, 0, readW, readH);    /* src region */
+                                 rbRead->rtt_level,                 /* src lvl */
+                                 &src_box);
 
    }
    else {
       /* CPU-based fallback/conversion */
       struct pipe_transfer *ptRead =
-         pipe_get_transfer(st->pipe, rbRead->texture, 0, 0, 0,
-                                        PIPE_TRANSFER_READ,
-                                        readX, readY, readW, readH);
+         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;
 
@@ -1088,22 +1571,26 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
       else
          transfer_usage = PIPE_TRANSFER_WRITE;
 
-      ptTex = pipe_get_transfer(st->pipe, pt, 0, 0, 0, transfer_usage,
-                                             0, 0, width, height);
+      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 = (GLfloat *) malloc(width * height * 4 * sizeof(GLfloat));
-         pipe_get_tile_rgba(pipe, ptRead, readX, readY, readW, readH, buf);
-         pipe_put_tile_rgba(pipe, ptTex, pack.SkipPixels, pack.SkipRows,
-                            readW, readH, buf);
+         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 = (GLuint *) malloc(width * height * sizeof(GLuint));
-         pipe_get_tile_z(pipe, ptRead, readX, readY, readW, readH, buf);
+         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);
@@ -1118,13 +1605,14 @@ st_CopyPixels(GLcontext *ctx, GLint srcx, GLint srcy,
     */
    draw_textured_quad(ctx, dstx, dsty, ctx->Current.RasterPos[2],
                       width, height, ctx->Pixel.ZoomX, ctx->Pixel.ZoomY,
-                      sv, 
+                      sv,
+                      num_sampler_view,
                       driver_vp, 
                       driver_fp,
-                      color, invertTex);
+                      color, invertTex, GL_FALSE, GL_FALSE);
 
    pipe_resource_reference(&pt, NULL);
-   pipe_sampler_view_reference(&sv, NULL);
+   pipe_sampler_view_reference(&sv[0], NULL);
 }
 
 
@@ -1139,7 +1627,13 @@ void st_init_drawpixels_functions(struct dd_function_table *functions)
 void
 st_destroy_drawpix(struct st_context *st)
 {
-   st_reference_fragprog(st, &st->drawpix.z_shader, NULL);
+   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);
+   }
+
    st_reference_fragprog(st, &st->pixel_xfer.combined_prog, NULL);
    if (st->drawpix.vert_shaders[0])
       ureg_free_tokens(st->drawpix.vert_shaders[0]);