st/xorg: lots of rendering and xv changes

extract all the rendering code to xorg_rendedrer, make both exa and xv
share that code. in the process cleanup the rendering code and implement
a lot more of the xv infrastructure.

11 files changed, 1170 insertions, 786 deletions

diff --git a/src/gallium/state_trackers/xorg/SConscript b/src/gallium/state_trackers/xorg/SConscript
index 65f55ea378..6165bae7a4 100644
--- a/src/gallium/state_trackers/xorg/SConscript
+++ b/src/gallium/state_trackers/xorg/SConscript
@@ -22,6 +22,7 @@ if 'xorg' in env['statetrackers']:
 		'xorg_exa.c',
 		'xorg_exa_tgsi.c',
 		'xorg_output.c',
+		'xorg_xv.c',
 		]
     )
     Export('st_xorg')
diff --git a/src/gallium/state_trackers/xorg/xorg_composite.c b/src/gallium/state_trackers/xorg/xorg_composite.c
index 98e9933f72..54b9ecaba1 100644
--- a/src/gallium/state_trackers/xorg/xorg_composite.c
+++ b/src/gallium/state_trackers/xorg/xorg_composite.c
@@ -1,5 +1,6 @@
 #include "xorg_composite.h"
 
+#include "xorg_renderer.h"
 #include "xorg_exa_tgsi.h"
 
 #include "cso_cache/cso_context.h"
@@ -111,173 +112,6 @@ render_repeat_to_gallium(int mode)
    return PIPE_TEX_WRAP_REPEAT;
 }
 
-
-static INLINE void
-setup_vertex0(float vertex[2][4], float x, float y,
-              float color[4])
-{
-   vertex[0][0] = x;
-   vertex[0][1] = y;
-   vertex[0][2] = 0.f; /*z*/
-   vertex[0][3] = 1.f; /*w*/
-
-   vertex[1][0] = color[0]; /*r*/
-   vertex[1][1] = color[1]; /*g*/
-   vertex[1][2] = color[2]; /*b*/
-   vertex[1][3] = color[3]; /*a*/
-}
-
-static struct pipe_buffer *
-setup_vertex_data0(struct exa_context *ctx,
-                   int srcX, int srcY, int maskX, int maskY,
-                   int dstX, int dstY, int width, int height)
-{
-   /* 1st vertex */
-   setup_vertex0(ctx->vertices2[0], dstX, dstY,
-                 ctx->solid_color);
-   /* 2nd vertex */
-   setup_vertex0(ctx->vertices2[1], dstX + width, dstY,
-                 ctx->solid_color);
-   /* 3rd vertex */
-   setup_vertex0(ctx->vertices2[2], dstX + width, dstY + height,
-                 ctx->solid_color);
-   /* 4th vertex */
-   setup_vertex0(ctx->vertices2[3], dstX, dstY + height,
-                 ctx->solid_color);
-
-   return pipe_user_buffer_create(ctx->pipe->screen,
-                                  ctx->vertices2,
-                                  sizeof(ctx->vertices2));
-}
-
-static INLINE void
-setup_vertex1(float vertex[2][4], float x, float y, float s, float t)
-{
-   vertex[0][0] = x;
-   vertex[0][1] = y;
-   vertex[0][2] = 0.f; /*z*/
-   vertex[0][3] = 1.f; /*w*/
-
-   vertex[1][0] = s;   /*s*/
-   vertex[1][1] = t;   /*t*/
-   vertex[1][2] = 0.f; /*r*/
-   vertex[1][3] = 1.f; /*q*/
-}
-
-static struct pipe_buffer *
-setup_vertex_data1(struct exa_context *ctx,
-                   int srcX, int srcY, int maskX, int maskY,
-                   int dstX, int dstY, int width, int height)
-{
-   float s0, t0, s1, t1;
-   struct pipe_texture *src = ctx->bound_textures[0];
-
-   s0 = srcX / src->width[0];
-   s1 = srcX + width / src->width[0];
-   t0 = srcY / src->height[0];
-   t1 = srcY + height / src->height[0];
-
-   /* 1st vertex */
-   setup_vertex1(ctx->vertices2[0], dstX, dstY,
-                 s0, t0);
-   /* 2nd vertex */
-   setup_vertex1(ctx->vertices2[1], dstX + width, dstY,
-                 s1, t0);
-   /* 3rd vertex */
-   setup_vertex1(ctx->vertices2[2], dstX + width, dstY + height,
-                 s1, t1);
-   /* 4th vertex */
-   setup_vertex1(ctx->vertices2[3], dstX, dstY + height,
-                 s0, t1);
-
-   return pipe_user_buffer_create(ctx->pipe->screen,
-                                  ctx->vertices2,
-                                  sizeof(ctx->vertices2));
-}
-
-static struct pipe_buffer *
-setup_vertex_data_tex(struct exa_context *ctx,
-                      float x0, float y0, float x1, float y1,
-                      float s0, float t0, float s1, float t1,
-                      float z)
-{
-   /* 1st vertex */
-   setup_vertex1(ctx->vertices2[0], x0, y0,
-                 s0, t0);
-   /* 2nd vertex */
-   setup_vertex1(ctx->vertices2[1], x1, y0,
-                 s1, t0);
-   /* 3rd vertex */
-   setup_vertex1(ctx->vertices2[2], x1, y1,
-                 s1, t1);
-   /* 4th vertex */
-   setup_vertex1(ctx->vertices2[3], x0, y1,
-                 s0, t1);
-
-   return pipe_user_buffer_create(ctx->pipe->screen,
-                                  ctx->vertices2,
-                                  sizeof(ctx->vertices2));
-}
-
-
-
-static INLINE void
-setup_vertex2(float vertex[3][4], float x, float y,
-              float s0, float t0, float s1, float t1)
-{
-   vertex[0][0] = x;
-   vertex[0][1] = y;
-   vertex[0][2] = 0.f; /*z*/
-   vertex[0][3] = 1.f; /*w*/
-
-   vertex[1][0] = s0;  /*s*/
-   vertex[1][1] = t0;  /*t*/
-   vertex[1][2] = 0.f; /*r*/
-   vertex[1][3] = 1.f; /*q*/
-
-   vertex[2][0] = s1;  /*s*/
-   vertex[2][1] = t1;  /*t*/
-   vertex[2][2] = 0.f; /*r*/
-   vertex[2][3] = 1.f; /*q*/
-}
-
-static struct pipe_buffer *
-setup_vertex_data2(struct exa_context *ctx,
-                   int srcX, int srcY, int maskX, int maskY,
-                   int dstX, int dstY, int width, int height)
-{
-   float st0[4], st1[4];
-   struct pipe_texture *src = ctx->bound_textures[0];
-   struct pipe_texture *mask = ctx->bound_textures[0];
-
-   st0[0] = srcX / src->width[0];
-   st0[1] = srcY / src->height[0];
-   st0[2] = srcX + width / src->width[0];
-   st0[3] = srcY + height / src->height[0];
-
-   st1[0] = maskX / mask->width[0];
-   st1[1] = maskY / mask->height[0];
-   st1[2] = maskX + width / mask->width[0];
-   st1[3] = maskY + height / mask->height[0];
-
-   /* 1st vertex */
-   setup_vertex2(ctx->vertices3[0], dstX, dstY,
-                 st0[0], st0[1], st1[0], st1[1]);
-   /* 2nd vertex */
-   setup_vertex2(ctx->vertices3[1], dstX + width, dstY,
-                 st0[2], st0[1], st1[2], st1[1]);
-   /* 3rd vertex */
-   setup_vertex2(ctx->vertices3[2], dstX + width, dstY + height,
-                 st0[2], st0[3], st1[2], st1[3]);
-   /* 4th vertex */
-   setup_vertex2(ctx->vertices3[3], dstX, dstY + height,
-                 st0[0], st0[3], st1[0], st1[3]);
-
-   return pipe_user_buffer_create(ctx->pipe->screen,
-                                  ctx->vertices3,
-                                  sizeof(ctx->vertices3));
-}
-
 boolean xorg_composite_accelerated(int op,
                                    PicturePtr pSrcPicture,
                                    PicturePtr pMaskPicture,
@@ -317,71 +151,6 @@ boolean xorg_composite_accelerated(int op,
 }
 
 static void
-bind_clip_state(struct exa_context *exa)
-{
-}
-
-static void
-bind_framebuffer_state(struct exa_context *exa, struct exa_pixmap_priv *pDst)
-{
-   unsigned i;
-   struct pipe_framebuffer_state state;
-   struct pipe_surface *surface = exa_gpu_surface(exa, pDst);
-   memset(&state, 0, sizeof(struct pipe_framebuffer_state));
-
-   state.width  = pDst->tex->width[0];
-   state.height = pDst->tex->height[0];
-
-   state.nr_cbufs = 1;
-   state.cbufs[0] = surface;
-   for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
-      state.cbufs[i] = 0;
-
-   /* currently we don't use depth/stencil */
-   state.zsbuf = 0;
-
-   cso_set_framebuffer(exa->cso, &state);
-
-   /* we do fire and forget for the framebuffer, this is the forget part */
-   pipe_surface_reference(&surface, NULL);
-}
-
-enum AxisOrientation {
-   Y0_BOTTOM,
-   Y0_TOP
-};
-
-static void
-set_viewport(struct exa_context *exa, int width, int height,
-             enum AxisOrientation orientation)
-{
-   struct pipe_viewport_state viewport;
-   float y_scale = (orientation == Y0_BOTTOM) ? -2.f : 2.f;
-
-   viewport.scale[0] =  width / 2.f;
-   viewport.scale[1] =  height / y_scale;
-   viewport.scale[2] =  1.0;
-   viewport.scale[3] =  1.0;
-   viewport.translate[0] = width / 2.f;
-   viewport.translate[1] = height / 2.f;
-   viewport.translate[2] = 0.0;
-   viewport.translate[3] = 0.0;
-
-   cso_set_viewport(exa->cso, &viewport);
-}
-
-static void
-bind_viewport_state(struct exa_context *exa, struct exa_pixmap_priv *pDst)
-{
-   int width = pDst->tex->width[0];
-   int height = pDst->tex->height[0];
-
-   /*debug_printf("Bind viewport (%d, %d)\n", width, height);*/
-
-   set_viewport(exa, width, height, Y0_TOP);
-}
-
-static void
 bind_blend_state(struct exa_context *exa, int op,
                  PicturePtr pSrcPicture, PicturePtr pMaskPicture)
 {
@@ -402,17 +171,9 @@ bind_blend_state(struct exa_context *exa, int op,
    blend.rgb_dst_factor   = blend_opt.rgb_dst_factor;
    blend.alpha_dst_factor = blend_opt.alpha_dst_factor;
 
-   cso_set_blend(exa->cso, &blend);
+   cso_set_blend(exa->renderer->cso, &blend);
 }
 
-static void
-bind_rasterizer_state(struct exa_context *exa)
-{
-   struct pipe_rasterizer_state raster;
-   memset(&raster, 0, sizeof(struct pipe_rasterizer_state));
-   raster.gl_rasterization_rules = 1;
-   cso_set_rasterizer(exa->cso, &raster);
-}
 
 static void
 bind_shaders(struct exa_context *exa, int op,
@@ -446,9 +207,9 @@ bind_shaders(struct exa_context *exa, int op,
       fs_traits |= FS_MASK;
    }
 
-   shader = xorg_shaders_get(exa->shaders, vs_traits, fs_traits);
-   cso_set_vertex_shader_handle(exa->cso, shader.vs);
-   cso_set_fragment_shader_handle(exa->cso, shader.fs);
+   shader = xorg_shaders_get(exa->renderer->shaders, vs_traits, fs_traits);
+   cso_set_vertex_shader_handle(exa->renderer->cso, shader.vs);
+   cso_set_fragment_shader_handle(exa->renderer->cso, shader.fs);
 }
 
 
@@ -500,9 +261,9 @@ bind_samplers(struct exa_context *exa, int op,
       ++exa->num_bound_samplers;
    }
 
-   cso_set_samplers(exa->cso, exa->num_bound_samplers,
+   cso_set_samplers(exa->renderer->cso, exa->num_bound_samplers,
                     (const struct pipe_sampler_state **)samplers);
-   cso_set_sampler_textures(exa->cso, exa->num_bound_samplers,
+   cso_set_sampler_textures(exa->renderer->cso, exa->num_bound_samplers,
                             exa->bound_textures);
 }
 
@@ -515,18 +276,8 @@ setup_vs_constant_buffer(struct exa_context *exa,
       2.f/width, 2.f/height, 1, 1,
       -1, -1, 0, 0
    };
-   struct pipe_constant_buffer *cbuf = &exa->vs_const_buffer;
-
-   pipe_buffer_reference(&cbuf->buffer, NULL);
-   cbuf->buffer = pipe_buffer_create(exa->pipe->screen, 16,
-                                     PIPE_BUFFER_USAGE_CONSTANT,
-                                     param_bytes);
-
-   if (cbuf->buffer) {
-      pipe_buffer_write(exa->pipe->screen, cbuf->buffer,
-                        0, param_bytes, vs_consts);
-   }
-   exa->pipe->set_constant_buffer(exa->pipe, PIPE_SHADER_VERTEX, 0, cbuf);
+   renderer_set_constants(exa->renderer, PIPE_SHADER_VERTEX,
+                          vs_consts, param_bytes);
 }
 
 
@@ -534,21 +285,11 @@ static void
 setup_fs_constant_buffer(struct exa_context *exa)
 {
    const int param_bytes = 4 * sizeof(float);
-   float fs_consts[8] = {
+   const float fs_consts[8] = {
       0, 0, 0, 1,
    };
-   struct pipe_constant_buffer *cbuf = &exa->fs_const_buffer;
-
-   pipe_buffer_reference(&cbuf->buffer, NULL);
-   cbuf->buffer = pipe_buffer_create(exa->pipe->screen, 16,
-                                     PIPE_BUFFER_USAGE_CONSTANT,
-                                     param_bytes);
-
-   if (cbuf->buffer) {
-      pipe_buffer_write(exa->pipe->screen, cbuf->buffer,
-                        0, param_bytes, fs_consts);
-   }
-   exa->pipe->set_constant_buffer(exa->pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+   renderer_set_constants(exa->renderer, PIPE_SHADER_FRAGMENT,
+                          fs_consts, param_bytes);
 }
 
 static void
@@ -570,14 +311,13 @@ boolean xorg_composite_bind_state(struct exa_context *exa,
                                   struct exa_pixmap_priv *pMask,
                                   struct exa_pixmap_priv *pDst)
 {
-   bind_framebuffer_state(exa, pDst);
-   bind_viewport_state(exa, pDst);
+   renderer_bind_framebuffer(exa->renderer, pDst);
+   renderer_bind_viewport(exa->renderer, pDst);
    bind_blend_state(exa, op, pSrcPicture, pMaskPicture);
-   bind_rasterizer_state(exa);
+   renderer_bind_rasterizer(exa->renderer);
    bind_shaders(exa, op, pSrcPicture, pMaskPicture);
    bind_samplers(exa, op, pSrcPicture, pMaskPicture,
                  pDstPicture, pSrc, pMask, pDst);
-   bind_clip_state(exa);
    setup_constant_buffers(exa, pDst);
 
    return FALSE;
@@ -588,42 +328,16 @@ void xorg_composite(struct exa_context *exa,
                     int srcX, int srcY, int maskX, int maskY,
                     int dstX, int dstY, int width, int height)
 {
-   struct pipe_context *pipe = exa->pipe;
-   struct pipe_buffer *buf = 0;
-
    if (exa->num_bound_samplers == 0 ) { /* solid fill */
-      buf = setup_vertex_data0(exa,
-                               srcX, srcY, maskX, maskY,
-                               dstX, dstY, width, height);
-   } else if (exa->num_bound_samplers == 1 ) /* src */
-      buf = setup_vertex_data1(exa,
-                               srcX, srcY, maskX, maskY,
-                               dstX, dstY, width, height);
-   else if (exa->num_bound_samplers == 2) /* src + mask */
-      buf = setup_vertex_data2(exa,
-                               srcX, srcY, maskX, maskY,
-                               dstX, dstY, width, height);
-   else if (exa->num_bound_samplers == 3) { /* src + mask + dst */
-      debug_assert(!"src/mask/dst not handled right now");
-#if 0
-      buf = setup_vertex_data2(exa,
-                               srcX, srcY, maskX, maskY,
-                               dstX, dstY, width, height);
-#endif
-   }
-
-   if (buf) {
-      int num_attribs = 1; /*pos*/
-      num_attribs += exa->num_bound_samplers;
-      if (exa->has_solid_color)
-         ++num_attribs;
-
-      util_draw_vertex_buffer(pipe, buf, 0,
-                              PIPE_PRIM_TRIANGLE_FAN,
-                              4,  /* verts */
-                              num_attribs); /* attribs/vert */
-
-      pipe_buffer_reference(&buf, NULL);
+      renderer_draw_solid_rect(exa->renderer,
+                               dstX, dstY, dstX + width, dstY + height,
+                               exa->solid_color);
+   } else {
+      int pos[6] = {srcX, srcY, maskX, maskY, dstX, dstY};
+      renderer_draw_textures(exa->renderer,
+                             pos, width, height,
+                             exa->bound_textures,
+                             exa->num_bound_samplers);
    }
 }
 
@@ -650,16 +364,15 @@ boolean xorg_solid_bind_state(struct exa_context *exa,
    vs_traits = VS_SOLID_FILL;
    fs_traits = FS_SOLID_FILL;
 
-   bind_framebuffer_state(exa, pixmap);
-   bind_viewport_state(exa, pixmap);
-   bind_rasterizer_state(exa);
+   renderer_bind_framebuffer(exa->renderer, pixmap);
+   renderer_bind_viewport(exa->renderer, pixmap);
+   renderer_bind_rasterizer(exa->renderer);
    bind_blend_state(exa, PictOpSrc, NULL, NULL);
    setup_constant_buffers(exa, pixmap);
-   bind_clip_state(exa);
 
-   shader = xorg_shaders_get(exa->shaders, vs_traits, fs_traits);
-   cso_set_vertex_shader_handle(exa->cso, shader.vs);
-   cso_set_fragment_shader_handle(exa->cso, shader.fs);
+   shader = xorg_shaders_get(exa->renderer->shaders, vs_traits, fs_traits);
+   cso_set_vertex_shader_handle(exa->renderer->cso, shader.vs);
+   cso_set_fragment_shader_handle(exa->renderer->cso, shader.fs);
 
    return TRUE;
 }
@@ -668,402 +381,7 @@ void xorg_solid(struct exa_context *exa,
                 struct exa_pixmap_priv *pixmap,
                 int x0, int y0, int x1, int y1)
 {
-   struct pipe_context *pipe = exa->pipe;
-   struct pipe_buffer *buf = 0;
-
-   /* 1st vertex */
-   setup_vertex0(exa->vertices2[0], x0, y0,
-                 exa->solid_color);
-   /* 2nd vertex */
-   setup_vertex0(exa->vertices2[1], x1, y0,
-                 exa->solid_color);
-   /* 3rd vertex */
-   setup_vertex0(exa->vertices2[2], x1, y1,
-                 exa->solid_color);
-   /* 4th vertex */
-   setup_vertex0(exa->vertices2[3], x0, y1,
-                 exa->solid_color);
-
-   buf = pipe_user_buffer_create(exa->pipe->screen,
-                                 exa->vertices2,
-                                 sizeof(exa->vertices2));
-
-
-   if (buf) {
-      util_draw_vertex_buffer(pipe, buf, 0,
-                              PIPE_PRIM_TRIANGLE_FAN,
-                              4,  /* verts */
-                              2); /* attribs/vert */
-
-      pipe_buffer_reference(&buf, NULL);
-   }
-}
-
-
-static INLINE void shift_rectx(float coords[4],
-                               const float *bounds,
-                               const float shift)
-{
-   coords[0] += shift;
-   coords[2] -= shift;
-   if (bounds) {
-      coords[2] = MIN2(coords[2], bounds[2]);
-      /* bound x/y + width/height */
-      if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
-         coords[2] = (bounds[0] + bounds[2]) - coords[0];
-      }
-   }
-}
-
-static INLINE void shift_recty(float coords[4],
-                               const float *bounds,
-                               const float shift)
-{
-   coords[1] += shift;
-   coords[3] -= shift;
-   if (bounds) {
-      coords[3] = MIN2(coords[3], bounds[3]);
-      if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
-         coords[3] = (bounds[1] + bounds[3]) - coords[1];
-      }
-   }
-}
-
-static INLINE void bound_rect(float coords[4],
-                              const float bounds[4],
-                              float shift[4])
-{
-   /* if outside the bounds */
-   if (coords[0] > (bounds[0] + bounds[2]) ||
-       coords[1] > (bounds[1] + bounds[3]) ||
-       (coords[0] + coords[2]) < bounds[0] ||
-       (coords[1] + coords[3]) < bounds[1]) {
-      coords[0] = 0.f;
-      coords[1] = 0.f;
-      coords[2] = 0.f;
-      coords[3] = 0.f;
-      shift[0] = 0.f;
-      shift[1] = 0.f;
-      return;
-   }
-
-   /* bound x */
-   if (coords[0] < bounds[0]) {
-      shift[0] = bounds[0] - coords[0];
-      coords[2] -= shift[0];
-      coords[0] = bounds[0];
-   } else
-      shift[0] = 0.f;
-
-   /* bound y */
-   if (coords[1] < bounds[1]) {
-      shift[1] = bounds[1] - coords[1];
-      coords[3] -= shift[1];
-      coords[1] = bounds[1];
-   } else
-      shift[1] = 0.f;
-
-   shift[2] = bounds[2] - coords[2];
-   shift[3] = bounds[3] - coords[3];
-   /* bound width/height */
-   coords[2] = MIN2(coords[2], bounds[2]);
-   coords[3] = MIN2(coords[3], bounds[3]);
-
-   /* bound x/y + width/height */
-   if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
-      coords[2] = (bounds[0] + bounds[2]) - coords[0];
-   }
-   if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
-      coords[3] = (bounds[1] + bounds[3]) - coords[1];
-   }
-
-   /* if outside the bounds */
-   if ((coords[0] + coords[2]) < bounds[0] ||
-       (coords[1] + coords[3]) < bounds[1]) {
-      coords[0] = 0.f;
-      coords[1] = 0.f;
-      coords[2] = 0.f;
-      coords[3] = 0.f;
-      return;
-   }
-}
-
-static INLINE void sync_size(float *src_loc, float *dst_loc)
-{
-   src_loc[2] = MIN2(src_loc[2], dst_loc[2]);
-   src_loc[3] = MIN2(src_loc[3], dst_loc[3]);
-   dst_loc[2] = src_loc[2];
-   dst_loc[3] = src_loc[3];
-}
-
-
-static void renderer_copy_texture(struct exa_context *exa,
-                                  struct pipe_texture *src,
-                                  float sx1, float sy1,
-                                  float sx2, float sy2,
-                                  struct pipe_texture *dst,
-                                  float dx1, float dy1,
-                                  float dx2, float dy2)
-{
-   struct pipe_context *pipe = exa->pipe;
-   struct pipe_screen *screen = pipe->screen;
-   struct pipe_buffer *buf;
-   struct pipe_surface *dst_surf = screen->get_tex_surface(
-      screen, dst, 0, 0, 0,
-      PIPE_BUFFER_USAGE_GPU_WRITE);
-   struct pipe_framebuffer_state fb;
-   float s0, t0, s1, t1;
-   struct xorg_shader shader;
-
-   assert(src->width[0] != 0);
-   assert(src->height[0] != 0);
-   assert(dst->width[0] != 0);
-   assert(dst->height[0] != 0);
-
-#if 1
-   s0 = sx1 / src->width[0];
-   s1 = sx2 / src->width[0];
-   t0 = sy1 / src->height[0];
-   t1 = sy2 / src->height[0];
-#else
-   s0 = 0;
-   s1 = 1;
-   t0 = 0;
-   t1 = 1;
-#endif
-
-#if 0
-   debug_printf("copy texture src=[%f, %f, %f, %f], dst=[%f, %f, %f, %f], tex=[%f, %f, %f, %f]\n",
-                sx1, sy1, sx2, sy2, dx1, dy1, dx2, dy2,
-                s0, t0, s1, t1);
-#endif
-
-   assert(screen->is_format_supported(screen, dst_surf->format,
-                                      PIPE_TEXTURE_2D,
-                                      PIPE_TEXTURE_USAGE_RENDER_TARGET,
-                                      0));
-
-   /* save state (restored below) */
-   cso_save_blend(exa->cso);
-   cso_save_samplers(exa->cso);
-   cso_save_sampler_textures(exa->cso);
-   cso_save_framebuffer(exa->cso);
-   cso_save_fragment_shader(exa->cso);
-   cso_save_vertex_shader(exa->cso);
-
-   cso_save_viewport(exa->cso);
-
-
-   /* set misc state we care about */
-   {
-      struct pipe_blend_state blend;
-      memset(&blend, 0, sizeof(blend));
-      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
-      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
-      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
-      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
-      blend.colormask = PIPE_MASK_RGBA;
-      cso_set_blend(exa->cso, &blend);
-   }
-
-   /* sampler */
-   {
-      struct pipe_sampler_state sampler;
-      memset(&sampler, 0, sizeof(sampler));
-      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
-      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
-      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
-      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
-      sampler.normalized_coords = 1;
-      cso_single_sampler(exa->cso, 0, &sampler);
-      cso_single_sampler_done(exa->cso);
-   }
-
-   set_viewport(exa, dst_surf->width, dst_surf->height, Y0_TOP);
-
-   /* texture */
-   cso_set_sampler_textures(exa->cso, 1, &src);
-
-   bind_rasterizer_state(exa);
-
-   /* shaders */
-   shader = xorg_shaders_get(exa->shaders,
-                             VS_COMPOSITE,
-                             FS_COMPOSITE);
-   cso_set_vertex_shader_handle(exa->cso, shader.vs);
-   cso_set_fragment_shader_handle(exa->cso, shader.fs);
-
-   /* drawing dest */
-   memset(&fb, 0, sizeof(fb));
-   fb.width = dst_surf->width;
-   fb.height = dst_surf->height;
-   fb.nr_cbufs = 1;
-   fb.cbufs[0] = dst_surf;
-   {
-      int i;
-      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
-         fb.cbufs[i] = 0;
-   }
-   cso_set_framebuffer(exa->cso, &fb);
-   setup_vs_constant_buffer(exa, fb.width, fb.height);
-   setup_fs_constant_buffer(exa);
-
-   /* draw quad */
-   buf = setup_vertex_data_tex(exa,
-                               dx1, dy1,
-                               dx2, dy2,
-                               s0, t0, s1, t1,
-                               0.0f);
-
-   if (buf) {
-      util_draw_vertex_buffer(exa->pipe, buf, 0,
-                              PIPE_PRIM_TRIANGLE_FAN,
-                              4,  /* verts */
-                              2); /* attribs/vert */
-
-      pipe_buffer_reference(&buf, NULL);
-   }
-
-   /* restore state we changed */
-   cso_restore_blend(exa->cso);
-   cso_restore_samplers(exa->cso);
-   cso_restore_sampler_textures(exa->cso);
-   cso_restore_framebuffer(exa->cso);
-   cso_restore_vertex_shader(exa->cso);
-   cso_restore_fragment_shader(exa->cso);
-   cso_restore_viewport(exa->cso);
-
-   pipe_surface_reference(&dst_surf, NULL);
-}
-
-
-static struct pipe_texture *
-create_sampler_texture(struct exa_context *ctx,
-                       struct pipe_texture *src)
-{
-   enum pipe_format format;
-   struct pipe_context *pipe = ctx->pipe;
-   struct pipe_screen *screen = pipe->screen;
-   struct pipe_texture *pt;
-   struct pipe_texture templ;
-
-   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
-
-   /* the coming in texture should already have that invariance */
-   debug_assert(screen->is_format_supported(screen, src->format,
-                                            PIPE_TEXTURE_2D,
-                                            PIPE_TEXTURE_USAGE_SAMPLER, 0));
-
-   format = src->format;
-
-   memset(&templ, 0, sizeof(templ));
-   templ.target = PIPE_TEXTURE_2D;
-   templ.format = format;
-   templ.last_level = 0;
-   templ.width[0] = src->width[0];
-   templ.height[0] = src->height[0];
-   templ.depth[0] = 1;
-   pf_get_block(format, &templ.block);
-   templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
-
-   pt = screen->texture_create(screen, &templ);
-
-   debug_assert(!pt || pipe_is_referenced(&pt->reference));
-
-   if (!pt)
-      return NULL;
-
-   {
-      /* copy source framebuffer surface into texture */
-      struct pipe_surface *ps_read = screen->get_tex_surface(
-         screen, src, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ);
-      struct pipe_surface *ps_tex = screen->get_tex_surface(
-         screen, pt, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_WRITE );
-      pipe->surface_copy(pipe,
-			 ps_tex, /* dest */
-			 0, 0, /* destx/y */
-			 ps_read,
-			 0, 0, src->width[0], src->height[0]);
-      pipe_surface_reference(&ps_read, NULL);
-      pipe_surface_reference(&ps_tex, NULL);
-   }
-
-   return pt;
-}
-
-void xorg_copy_pixmap(struct exa_context *ctx,
-                      struct exa_pixmap_priv *dst_priv, int dx, int dy,
-                      struct exa_pixmap_priv *src_priv, int sx, int sy,
-                      int width, int height)
-{
-   float dst_loc[4], src_loc[4];
-   float dst_bounds[4], src_bounds[4];
-   float src_shift[4], dst_shift[4], shift[4];
-   struct pipe_texture *dst = dst_priv->tex;
-   struct pipe_texture *src = src_priv->tex;
-
-   if (ctx->pipe->is_texture_referenced(ctx->pipe, src, 0, 0) &
-       PIPE_REFERENCED_FOR_WRITE)
-      ctx->pipe->flush(ctx->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
-
-   dst_loc[0] = dx;
-   dst_loc[1] = dy;
-   dst_loc[2] = width;
-   dst_loc[3] = height;
-   dst_bounds[0] = 0.f;
-   dst_bounds[1] = 0.f;
-   dst_bounds[2] = dst->width[0];
-   dst_bounds[3] = dst->height[0];
-
-   src_loc[0] = sx;
-   src_loc[1] = sy;
-   src_loc[2] = width;
-   src_loc[3] = height;
-   src_bounds[0] = 0.f;
-   src_bounds[1] = 0.f;
-   src_bounds[2] = src->width[0];
-   src_bounds[3] = src->height[0];
-
-   bound_rect(src_loc, src_bounds, src_shift);
-   bound_rect(dst_loc, dst_bounds, dst_shift);
-   shift[0] = src_shift[0] - dst_shift[0];
-   shift[1] = src_shift[1] - dst_shift[1];
-
-   if (shift[0] < 0)
-      shift_rectx(src_loc, src_bounds, -shift[0]);
-   else
-      shift_rectx(dst_loc, dst_bounds, shift[0]);
-
-   if (shift[1] < 0)
-      shift_recty(src_loc, src_bounds, -shift[1]);
-   else
-      shift_recty(dst_loc, dst_bounds, shift[1]);
-
-   sync_size(src_loc, dst_loc);
-
-   if (src_loc[2] >= 0 && src_loc[3] >= 0 &&
-       dst_loc[2] >= 0 && dst_loc[3] >= 0) {
-      struct pipe_texture *temp_src = src;
-
-      if (src == dst)
-         temp_src = create_sampler_texture(ctx, src);
-
-      renderer_copy_texture(ctx,
-                            temp_src,
-                            src_loc[0],
-                            src_loc[1],
-                            src_loc[0] + src_loc[2],
-                            src_loc[1] + src_loc[3],
-                            dst,
-                            dst_loc[0],
-                            dst_loc[1],
-                            dst_loc[0] + dst_loc[2],
-                            dst_loc[1] + dst_loc[3]);
-
-      if (src == dst)
-         pipe_texture_reference(&temp_src, NULL);
-   }
+   renderer_draw_solid_rect(exa->renderer,
+                            x0, y0, x1, y1, exa->solid_color);
 }
 
diff --git a/src/gallium/state_trackers/xorg/xorg_composite.h b/src/gallium/state_trackers/xorg/xorg_composite.h
index e73f1c704a..236addf1ce 100644
--- a/src/gallium/state_trackers/xorg/xorg_composite.h
+++ b/src/gallium/state_trackers/xorg/xorg_composite.h
@@ -29,9 +29,4 @@ void xorg_solid(struct exa_context *exa,
                 struct exa_pixmap_priv *pixmap,
                 int x0, int y0, int x1, int y1);
 
-void xorg_copy_pixmap(struct exa_context *ctx,
-                      struct exa_pixmap_priv *dst, int dx, int dy,
-                      struct exa_pixmap_priv *src, int sx, int sy,
-                      int width, int height);
-
 #endif
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c
index 2633e8caaf..29785c0040 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa.c
@@ -43,8 +43,6 @@
 #include "pipe/p_state.h"
 #include "pipe/p_inlines.h"
 
-#include "cso_cache/cso_context.h"
-
 #include "util/u_rect.h"
 
 #define DEBUG_PRINT 0
@@ -90,16 +88,6 @@ exa_get_pipe_format(int depth, enum pipe_format *format, int *bbp)
 }
 
 static void
-xorg_exa_init_state(struct exa_context *exa)
-{
-   struct pipe_depth_stencil_alpha_state dsa;
-
-   /* set common initial clip state */
-   memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
-   cso_set_depth_stencil_alpha(exa->cso, &dsa);
-}
-
-static void
 xorg_exa_common_done(struct exa_context *exa)
 {
    exa->copy.src = NULL;
@@ -444,9 +432,9 @@ ExaCopy(PixmapPtr pDstPixmap, int srcX, int srcY, int dstX, int dstY,
 
    debug_assert(priv == exa->copy.dst);
 
-   xorg_copy_pixmap(exa, exa->copy.dst, dstX, dstY,
-                    exa->copy.src, srcX, srcY,
-                    width, height);
+   renderer_copy_pixmap(exa->renderer, exa->copy.dst, dstX, dstY,
+                        exa->copy.src, srcX, srcY,
+                        width, height);
 }
 
 static Bool
@@ -701,7 +689,7 @@ ExaModifyPixmapHeader(PixmapPtr pPixmap, int width, int height,
 
 	    dst_surf = exa->scrn->get_tex_surface(
 		exa->scrn, texture, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_WRITE);
-	    src_surf = exa_gpu_surface(exa, priv);
+	    src_surf = xorg_gpu_surface(exa->pipe->screen, priv);
 	    exa->pipe->surface_copy(exa->pipe, dst_surf, 0, 0, src_surf,
 				    0, 0, min(width, texture->width[0]),
 				    min(height, texture->height[0]));
@@ -731,23 +719,8 @@ xorg_exa_close(ScrnInfoPtr pScrn)
 {
    modesettingPtr ms = modesettingPTR(pScrn);
    struct exa_context *exa = ms->exa;
-   struct pipe_constant_buffer *vsbuf = &exa->vs_const_buffer;
-   struct pipe_constant_buffer *fsbuf = &exa->fs_const_buffer;
 
-   if (exa->shaders) {
-      xorg_shaders_destroy(exa->shaders);
-   }
-
-   if (vsbuf && vsbuf->buffer)
-      pipe_buffer_reference(&vsbuf->buffer, NULL);
-
-   if (fsbuf && fsbuf->buffer)
-      pipe_buffer_reference(&fsbuf->buffer, NULL);
-
-   if (exa->cso) {
-      cso_release_all(exa->cso);
-      cso_destroy_context(exa->cso);
-   }
+   renderer_destroy(exa->renderer);
 
    if (exa->pipe)
       exa->pipe->destroy(exa->pipe);
@@ -822,10 +795,7 @@ xorg_exa_init(ScrnInfoPtr pScrn)
    /* Share context with DRI */
    ms->ctx = exa->pipe;
 
-   exa->cso = cso_create_context(exa->pipe);
-   exa->shaders = xorg_shaders_create(exa);
-
-   xorg_exa_init_state(exa);
+   exa->renderer = renderer_create(exa->pipe);
 
    return (void *)exa;
 
@@ -836,11 +806,11 @@ out_err:
 }
 
 struct pipe_surface *
-exa_gpu_surface(struct exa_context *exa, struct exa_pixmap_priv *priv)
+xorg_gpu_surface(struct pipe_screen *scrn, struct exa_pixmap_priv *priv)
 {
-   return exa->scrn->get_tex_surface(exa->scrn, priv->tex, 0, 0, 0,
-                                     PIPE_BUFFER_USAGE_GPU_READ |
-                                     PIPE_BUFFER_USAGE_GPU_WRITE);
+   return scrn->get_tex_surface(scrn, priv->tex, 0, 0, 0,
+                                PIPE_BUFFER_USAGE_GPU_READ |
+                                PIPE_BUFFER_USAGE_GPU_WRITE);
 
 }
 
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.h b/src/gallium/state_trackers/xorg/xorg_exa.h
index 28834e3ef5..292f964cec 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.h
+++ b/src/gallium/state_trackers/xorg/xorg_exa.h
@@ -16,11 +16,7 @@ struct exa_context
    ExaDriverPtr pExa;
    struct pipe_context *pipe;
    struct pipe_screen *scrn;
-   struct cso_context *cso;
-   struct xorg_shaders *shaders;
-
-   struct pipe_constant_buffer vs_const_buffer;
-   struct pipe_constant_buffer fs_const_buffer;
+   struct xorg_renderer *renderer;
 
    struct pipe_texture *bound_textures[MAX_EXA_SAMPLERS];
    int num_bound_samplers;
@@ -32,10 +28,6 @@ struct exa_context
       struct exa_pixmap_priv *src;
       struct exa_pixmap_priv *dst;
    } copy;
-
-   /* we should combine these two */
-   float vertices2[4][2][4];
-   float vertices3[4][3][4];
 };
 
 struct exa_pixmap_priv
@@ -60,7 +52,7 @@ do {                                                          \
 } while(0)
 
 struct pipe_surface *
-exa_gpu_surface(struct exa_context *exa, struct exa_pixmap_priv *priv);
+xorg_gpu_surface(struct pipe_screen *scrn, struct exa_pixmap_priv *priv);
 
 void xorg_exa_flush(struct exa_context *exa, uint pipeFlushFlags,
                     struct pipe_fence_handle **fence);
diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
index bb5a42af37..8ce06e374a 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
@@ -44,7 +44,7 @@
  */
 
 struct xorg_shaders {
-   struct exa_context *exa;
+   struct xorg_renderer *r;
 
    struct cso_hash *vs_hash;
    struct cso_hash *fs_hash;
@@ -400,11 +400,11 @@ create_fs(struct pipe_context *pipe,
    return ureg_create_shader_and_destroy(ureg, pipe);
 }
 
-struct xorg_shaders * xorg_shaders_create(struct exa_context *exa)
+struct xorg_shaders * xorg_shaders_create(struct xorg_renderer *r)
 {
    struct xorg_shaders *sc = CALLOC_STRUCT(xorg_shaders);
 
-   sc->exa = exa;
+   sc->r = r;
    sc->vs_hash = cso_hash_create();
    sc->fs_hash = cso_hash_create();
 
@@ -431,9 +431,9 @@ cache_destroy(struct cso_context *cso,
 
 void xorg_shaders_destroy(struct xorg_shaders *sc)
 {
-   cache_destroy(sc->exa->cso, sc->vs_hash,
+   cache_destroy(sc->r->cso, sc->vs_hash,
                  PIPE_SHADER_VERTEX);
-   cache_destroy(sc->exa->cso, sc->fs_hash,
+   cache_destroy(sc->r->cso, sc->fs_hash,
                  PIPE_SHADER_FRAGMENT);
 
    free(sc);
@@ -468,9 +468,9 @@ struct xorg_shader xorg_shaders_get(struct xorg_shaders *sc,
    struct xorg_shader shader = { NULL, NULL };
    void *vs, *fs;
 
-   vs = shader_from_cache(sc->exa->pipe, PIPE_SHADER_VERTEX,
+   vs = shader_from_cache(sc->r->pipe, PIPE_SHADER_VERTEX,
                           sc->vs_hash, vs_traits);
-   fs = shader_from_cache(sc->exa->pipe, PIPE_SHADER_FRAGMENT,
+   fs = shader_from_cache(sc->r->pipe, PIPE_SHADER_FRAGMENT,
                           sc->fs_hash, fs_traits);
 
    debug_assert(vs && fs);
diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h
index 1535a0c8c3..33c272070b 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h
+++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.h
@@ -1,7 +1,7 @@
 #ifndef XORG_EXA_TGSI_H
 #define XORG_EXA_TGSI_H
 
-#include "xorg_exa.h"
+#include "xorg_renderer.h"
 
 enum xorg_vs_traits {
    VS_COMPOSITE        = 1 << 0,
@@ -33,7 +33,7 @@ struct xorg_shader {
 
 struct xorg_shaders;
 
-struct xorg_shaders *xorg_shaders_create(struct exa_context *exa);
+struct xorg_shaders *xorg_shaders_create(struct xorg_renderer *renderer);
 void xorg_shaders_destroy(struct xorg_shaders *shaders);
 
 struct xorg_shader xorg_shaders_get(struct xorg_shaders *shaders,
diff --git a/src/gallium/state_trackers/xorg/xorg_renderer.c b/src/gallium/state_trackers/xorg/xorg_renderer.c
new file mode 100644
index 0000000000..362fb6f4f1
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_renderer.c
@@ -0,0 +1,748 @@
+#include "xorg_exa.h"
+#include "xorg_renderer.h"
+
+#include "xorg_exa_tgsi.h"
+
+#include "cso_cache/cso_context.h"
+#include "util/u_draw_quad.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+#include "pipe/p_inlines.h"
+
+enum AxisOrientation {
+   Y0_BOTTOM,
+   Y0_TOP
+};
+
+static void
+renderer_init_state(struct xorg_renderer *r)
+{
+   struct pipe_depth_stencil_alpha_state dsa;
+
+   /* set common initial clip state */
+   memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
+   cso_set_depth_stencil_alpha(r->cso, &dsa);
+}
+
+
+static INLINE void
+setup_vertex0(float vertex[2][4], float x, float y,
+              float color[4])
+{
+   vertex[0][0] = x;
+   vertex[0][1] = y;
+   vertex[0][2] = 0.f; /*z*/
+   vertex[0][3] = 1.f; /*w*/
+
+   vertex[1][0] = color[0]; /*r*/
+   vertex[1][1] = color[1]; /*g*/
+   vertex[1][2] = color[2]; /*b*/
+   vertex[1][3] = color[3]; /*a*/
+}
+
+static INLINE void
+setup_vertex1(float vertex[2][4], float x, float y, float s, float t)
+{
+   vertex[0][0] = x;
+   vertex[0][1] = y;
+   vertex[0][2] = 0.f; /*z*/
+   vertex[0][3] = 1.f; /*w*/
+
+   vertex[1][0] = s;   /*s*/
+   vertex[1][1] = t;   /*t*/
+   vertex[1][2] = 0.f; /*r*/
+   vertex[1][3] = 1.f; /*q*/
+}
+
+static struct pipe_buffer *
+setup_vertex_data1(struct xorg_renderer *r,
+                   int srcX, int srcY,  int dstX, int dstY,
+                   int width, int height,
+                   struct pipe_texture *src)
+{
+   float s0, t0, s1, t1;
+
+   s0 = srcX / src->width[0];
+   s1 = srcX + width / src->width[0];
+   t0 = srcY / src->height[0];
+   t1 = srcY + height / src->height[0];
+
+   /* 1st vertex */
+   setup_vertex1(r->vertices2[0], dstX, dstY, s0, t0);
+   /* 2nd vertex */
+   setup_vertex1(r->vertices2[1], dstX + width, dstY, s1, t0);
+   /* 3rd vertex */
+   setup_vertex1(r->vertices2[2], dstX + width, dstY + height, s1, t1);
+   /* 4th vertex */
+   setup_vertex1(r->vertices2[3], dstX, dstY + height, s0, t1);
+
+   return pipe_user_buffer_create(r->pipe->screen,
+                                  r->vertices2,
+                                  sizeof(r->vertices2));
+}
+
+static struct pipe_buffer *
+setup_vertex_data_tex(struct xorg_renderer *r,
+                      float x0, float y0, float x1, float y1,
+                      float s0, float t0, float s1, float t1,
+                      float z)
+{
+   /* 1st vertex */
+   setup_vertex1(r->vertices2[0], x0, y0, s0, t0);
+   /* 2nd vertex */
+   setup_vertex1(r->vertices2[1], x1, y0, s1, t0);
+   /* 3rd vertex */
+   setup_vertex1(r->vertices2[2], x1, y1, s1, t1);
+   /* 4th vertex */
+   setup_vertex1(r->vertices2[3], x0, y1, s0, t1);
+
+   return pipe_user_buffer_create(r->pipe->screen,
+                                  r->vertices2,
+                                  sizeof(r->vertices2));
+}
+
+static INLINE void
+setup_vertex2(float vertex[3][4], float x, float y,
+              float s0, float t0, float s1, float t1)
+{
+   vertex[0][0] = x;
+   vertex[0][1] = y;
+   vertex[0][2] = 0.f; /*z*/
+   vertex[0][3] = 1.f; /*w*/
+
+   vertex[1][0] = s0;  /*s*/
+   vertex[1][1] = t0;  /*t*/
+   vertex[1][2] = 0.f; /*r*/
+   vertex[1][3] = 1.f; /*q*/
+
+   vertex[2][0] = s1;  /*s*/
+   vertex[2][1] = t1;  /*t*/
+   vertex[2][2] = 0.f; /*r*/
+   vertex[2][3] = 1.f; /*q*/
+}
+
+static struct pipe_buffer *
+setup_vertex_data2(struct xorg_renderer *r,
+                   int srcX, int srcY, int maskX, int maskY,
+                   int dstX, int dstY, int width, int height,
+                   struct pipe_texture *src,
+                   struct pipe_texture *mask)
+{
+   float st0[4], st1[4];
+
+   st0[0] = srcX / src->width[0];
+   st0[1] = srcY / src->height[0];
+   st0[2] = srcX + width / src->width[0];
+   st0[3] = srcY + height / src->height[0];
+
+   st1[0] = maskX / mask->width[0];
+   st1[1] = maskY / mask->height[0];
+   st1[2] = maskX + width / mask->width[0];
+   st1[3] = maskY + height / mask->height[0];
+
+   /* 1st vertex */
+   setup_vertex2(r->vertices3[0], dstX, dstY,
+                 st0[0], st0[1], st1[0], st1[1]);
+   /* 2nd vertex */
+   setup_vertex2(r->vertices3[1], dstX + width, dstY,
+                 st0[2], st0[1], st1[2], st1[1]);
+   /* 3rd vertex */
+   setup_vertex2(r->vertices3[2], dstX + width, dstY + height,
+                 st0[2], st0[3], st1[2], st1[3]);
+   /* 4th vertex */
+   setup_vertex2(r->vertices3[3], dstX, dstY + height,
+                 st0[0], st0[3], st1[0], st1[3]);
+
+   return pipe_user_buffer_create(r->pipe->screen,
+                                  r->vertices3,
+                                  sizeof(r->vertices3));
+}
+
+
+
+static void
+set_viewport(struct xorg_renderer *r, int width, int height,
+             enum AxisOrientation orientation)
+{
+   struct pipe_viewport_state viewport;
+   float y_scale = (orientation == Y0_BOTTOM) ? -2.f : 2.f;
+
+   viewport.scale[0] =  width / 2.f;
+   viewport.scale[1] =  height / y_scale;
+   viewport.scale[2] =  1.0;
+   viewport.scale[3] =  1.0;
+   viewport.translate[0] = width / 2.f;
+   viewport.translate[1] = height / 2.f;
+   viewport.translate[2] = 0.0;
+   viewport.translate[3] = 0.0;
+
+   cso_set_viewport(r->cso, &viewport);
+}
+
+
+
+struct xorg_renderer * renderer_create(struct pipe_context *pipe)
+{
+   struct xorg_renderer *renderer = CALLOC_STRUCT(xorg_renderer);
+
+   renderer->cso = cso_create_context(pipe);
+   renderer->shaders = xorg_shaders_create(renderer);
+
+   renderer_init_state(renderer);
+
+   return renderer;
+}
+
+void renderer_destroy(struct xorg_renderer *r)
+{
+   struct pipe_constant_buffer *vsbuf = &r->vs_const_buffer;
+   struct pipe_constant_buffer *fsbuf = &r->fs_const_buffer;
+
+   if (vsbuf && vsbuf->buffer)
+      pipe_buffer_reference(&vsbuf->buffer, NULL);
+
+   if (fsbuf && fsbuf->buffer)
+      pipe_buffer_reference(&fsbuf->buffer, NULL);
+
+   if (r->cso) {
+      cso_release_all(r->cso);
+      cso_destroy_context(r->cso);
+   }
+
+   if (r->shaders) {
+      xorg_shaders_destroy(r->shaders);
+   }
+}
+
+void renderer_bind_framebuffer(struct xorg_renderer *r,
+                               struct exa_pixmap_priv *priv)
+{
+   unsigned i;
+   struct pipe_framebuffer_state state;
+   struct pipe_surface *surface = xorg_gpu_surface(r->pipe->screen, priv);
+   memset(&state, 0, sizeof(struct pipe_framebuffer_state));
+
+   state.width  = priv->tex->width[0];
+   state.height = priv->tex->height[0];
+
+   state.nr_cbufs = 1;
+   state.cbufs[0] = surface;
+   for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+      state.cbufs[i] = 0;
+
+   /* currently we don't use depth/stencil */
+   state.zsbuf = 0;
+
+   cso_set_framebuffer(r->cso, &state);
+
+   /* we do fire and forget for the framebuffer, this is the forget part */
+   pipe_surface_reference(&surface, NULL);
+}
+
+void renderer_bind_viewport(struct xorg_renderer *r,
+                            struct exa_pixmap_priv *dst)
+{
+   int width = dst->tex->width[0];
+   int height = dst->tex->height[0];
+
+   /*debug_printf("Bind viewport (%d, %d)\n", width, height);*/
+
+   set_viewport(r, width, height, Y0_TOP);
+}
+
+void renderer_bind_rasterizer(struct xorg_renderer *r)
+{
+   struct pipe_rasterizer_state raster;
+
+   /* XXX: move to renderer_init_state? */
+   memset(&raster, 0, sizeof(struct pipe_rasterizer_state));
+   raster.gl_rasterization_rules = 1;
+   cso_set_rasterizer(r->cso, &raster);
+}
+
+void renderer_set_constants(struct xorg_renderer *r,
+                            int shader_type,
+                            const float *params,
+                            int param_bytes)
+{
+   struct pipe_constant_buffer *cbuf =
+      (shader_type == PIPE_SHADER_VERTEX) ? &r->vs_const_buffer :
+      &r->fs_const_buffer;
+
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+   cbuf->buffer = pipe_buffer_create(r->pipe->screen, 16,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+
+   if (cbuf->buffer) {
+      pipe_buffer_write(r->pipe->screen, cbuf->buffer,
+                        0, param_bytes, params);
+   }
+   r->pipe->set_constant_buffer(r->pipe, shader_type, 0, cbuf);
+}
+
+static void
+setup_vs_constant_buffer(struct xorg_renderer *r,
+                         int width, int height)
+{
+   const int param_bytes = 8 * sizeof(float);
+   float vs_consts[8] = {
+      2.f/width, 2.f/height, 1, 1,
+      -1, -1, 0, 0
+   };
+   renderer_set_constants(r, PIPE_SHADER_VERTEX,
+                          vs_consts, param_bytes);
+}
+
+static void
+setup_fs_constant_buffer(struct xorg_renderer *r)
+{
+   const int param_bytes = 4 * sizeof(float);
+   const float fs_consts[8] = {
+      0, 0, 0, 1,
+   };
+   renderer_set_constants(r, PIPE_SHADER_FRAGMENT,
+                          fs_consts, param_bytes);
+}
+
+static INLINE void shift_rectx(float coords[4],
+                               const float *bounds,
+                               const float shift)
+{
+   coords[0] += shift;
+   coords[2] -= shift;
+   if (bounds) {
+      coords[2] = MIN2(coords[2], bounds[2]);
+      /* bound x/y + width/height */
+      if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
+         coords[2] = (bounds[0] + bounds[2]) - coords[0];
+      }
+   }
+}
+
+static INLINE void shift_recty(float coords[4],
+                               const float *bounds,
+                               const float shift)
+{
+   coords[1] += shift;
+   coords[3] -= shift;
+   if (bounds) {
+      coords[3] = MIN2(coords[3], bounds[3]);
+      if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
+         coords[3] = (bounds[1] + bounds[3]) - coords[1];
+      }
+   }
+}
+
+static INLINE void bound_rect(float coords[4],
+                              const float bounds[4],
+                              float shift[4])
+{
+   /* if outside the bounds */
+   if (coords[0] > (bounds[0] + bounds[2]) ||
+       coords[1] > (bounds[1] + bounds[3]) ||
+       (coords[0] + coords[2]) < bounds[0] ||
+       (coords[1] + coords[3]) < bounds[1]) {
+      coords[0] = 0.f;
+      coords[1] = 0.f;
+      coords[2] = 0.f;
+      coords[3] = 0.f;
+      shift[0] = 0.f;
+      shift[1] = 0.f;
+      return;
+   }
+
+   /* bound x */
+   if (coords[0] < bounds[0]) {
+      shift[0] = bounds[0] - coords[0];
+      coords[2] -= shift[0];
+      coords[0] = bounds[0];
+   } else
+      shift[0] = 0.f;
+
+   /* bound y */
+   if (coords[1] < bounds[1]) {
+      shift[1] = bounds[1] - coords[1];
+      coords[3] -= shift[1];
+      coords[1] = bounds[1];
+   } else
+      shift[1] = 0.f;
+
+   shift[2] = bounds[2] - coords[2];
+   shift[3] = bounds[3] - coords[3];
+   /* bound width/height */
+   coords[2] = MIN2(coords[2], bounds[2]);
+   coords[3] = MIN2(coords[3], bounds[3]);
+
+   /* bound x/y + width/height */
+   if ((coords[0] + coords[2]) > (bounds[0] + bounds[2])) {
+      coords[2] = (bounds[0] + bounds[2]) - coords[0];
+   }
+   if ((coords[1] + coords[3]) > (bounds[1] + bounds[3])) {
+      coords[3] = (bounds[1] + bounds[3]) - coords[1];
+   }
+
+   /* if outside the bounds */
+   if ((coords[0] + coords[2]) < bounds[0] ||
+       (coords[1] + coords[3]) < bounds[1]) {
+      coords[0] = 0.f;
+      coords[1] = 0.f;
+      coords[2] = 0.f;
+      coords[3] = 0.f;
+      return;
+   }
+}
+
+static INLINE void sync_size(float *src_loc, float *dst_loc)
+{
+   src_loc[2] = MIN2(src_loc[2], dst_loc[2]);
+   src_loc[3] = MIN2(src_loc[3], dst_loc[3]);
+   dst_loc[2] = src_loc[2];
+   dst_loc[3] = src_loc[3];
+}
+
+static void renderer_copy_texture(struct xorg_renderer *r,
+                                  struct pipe_texture *src,
+                                  float sx1, float sy1,
+                                  float sx2, float sy2,
+                                  struct pipe_texture *dst,
+                                  float dx1, float dy1,
+                                  float dx2, float dy2)
+{
+   struct pipe_context *pipe = r->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_buffer *buf;
+   struct pipe_surface *dst_surf = screen->get_tex_surface(
+      screen, dst, 0, 0, 0,
+      PIPE_BUFFER_USAGE_GPU_WRITE);
+   struct pipe_framebuffer_state fb;
+   float s0, t0, s1, t1;
+   struct xorg_shader shader;
+
+   assert(src->width[0] != 0);
+   assert(src->height[0] != 0);
+   assert(dst->width[0] != 0);
+   assert(dst->height[0] != 0);
+
+#if 1
+   s0 = sx1 / src->width[0];
+   s1 = sx2 / src->width[0];
+   t0 = sy1 / src->height[0];
+   t1 = sy2 / src->height[0];
+#else
+   s0 = 0;
+   s1 = 1;
+   t0 = 0;
+   t1 = 1;
+#endif
+
+#if 0
+   debug_printf("copy texture src=[%f, %f, %f, %f], dst=[%f, %f, %f, %f], tex=[%f, %f, %f, %f]\n",
+                sx1, sy1, sx2, sy2, dx1, dy1, dx2, dy2,
+                s0, t0, s1, t1);
+#endif
+
+   assert(screen->is_format_supported(screen, dst_surf->format,
+                                      PIPE_TEXTURE_2D,
+                                      PIPE_TEXTURE_USAGE_RENDER_TARGET,
+                                      0));
+
+   /* save state (restored below) */
+   cso_save_blend(r->cso);
+   cso_save_samplers(r->cso);
+   cso_save_sampler_textures(r->cso);
+   cso_save_framebuffer(r->cso);
+   cso_save_fragment_shader(r->cso);
+   cso_save_vertex_shader(r->cso);
+
+   cso_save_viewport(r->cso);
+
+
+   /* set misc state we care about */
+   {
+      struct pipe_blend_state blend;
+      memset(&blend, 0, sizeof(blend));
+      blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+      blend.rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+      blend.colormask = PIPE_MASK_RGBA;
+      cso_set_blend(r->cso, &blend);
+   }
+
+   /* sampler */
+   {
+      struct pipe_sampler_state sampler;
+      memset(&sampler, 0, sizeof(sampler));
+      sampler.wrap_s = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_t = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.wrap_r = PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+      sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+      sampler.min_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
+      sampler.normalized_coords = 1;
+      cso_single_sampler(r->cso, 0, &sampler);
+      cso_single_sampler_done(r->cso);
+   }
+
+   set_viewport(r, dst_surf->width, dst_surf->height, Y0_TOP);
+
+   /* texture */
+   cso_set_sampler_textures(r->cso, 1, &src);
+
+   renderer_bind_rasterizer(r);
+
+   /* shaders */
+   shader = xorg_shaders_get(r->shaders,
+                             VS_COMPOSITE,
+                             FS_COMPOSITE);
+   cso_set_vertex_shader_handle(r->cso, shader.vs);
+   cso_set_fragment_shader_handle(r->cso, shader.fs);
+
+   /* drawing dest */
+   memset(&fb, 0, sizeof(fb));
+   fb.width = dst_surf->width;
+   fb.height = dst_surf->height;
+   fb.nr_cbufs = 1;
+   fb.cbufs[0] = dst_surf;
+   {
+      int i;
+      for (i = 1; i < PIPE_MAX_COLOR_BUFS; ++i)
+         fb.cbufs[i] = 0;
+   }
+   cso_set_framebuffer(r->cso, &fb);
+   setup_vs_constant_buffer(r, fb.width, fb.height);
+   setup_fs_constant_buffer(r);
+
+   /* draw quad */
+   buf = setup_vertex_data_tex(r,
+                               dx1, dy1,
+                               dx2, dy2,
+                               s0, t0, s1, t1,
+                               0.0f);
+
+   if (buf) {
+      util_draw_vertex_buffer(r->pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference(&buf, NULL);
+   }
+
+   /* restore state we changed */
+   cso_restore_blend(r->cso);
+   cso_restore_samplers(r->cso);
+   cso_restore_sampler_textures(r->cso);
+   cso_restore_framebuffer(r->cso);
+   cso_restore_vertex_shader(r->cso);
+   cso_restore_fragment_shader(r->cso);
+   cso_restore_viewport(r->cso);
+
+   pipe_surface_reference(&dst_surf, NULL);
+}
+
+static struct pipe_texture *
+create_sampler_texture(struct xorg_renderer *r,
+                       struct pipe_texture *src)
+{
+   enum pipe_format format;
+   struct pipe_context *pipe = r->pipe;
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture *pt;
+   struct pipe_texture templ;
+
+   pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   /* the coming in texture should already have that invariance */
+   debug_assert(screen->is_format_supported(screen, src->format,
+                                            PIPE_TEXTURE_2D,
+                                            PIPE_TEXTURE_USAGE_SAMPLER, 0));
+
+   format = src->format;
+
+   memset(&templ, 0, sizeof(templ));
+   templ.target = PIPE_TEXTURE_2D;
+   templ.format = format;
+   templ.last_level = 0;
+   templ.width[0] = src->width[0];
+   templ.height[0] = src->height[0];
+   templ.depth[0] = 1;
+   pf_get_block(format, &templ.block);
+   templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+
+   pt = screen->texture_create(screen, &templ);
+
+   debug_assert(!pt || pipe_is_referenced(&pt->reference));
+
+   if (!pt)
+      return NULL;
+
+   {
+      /* copy source framebuffer surface into texture */
+      struct pipe_surface *ps_read = screen->get_tex_surface(
+         screen, src, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ);
+      struct pipe_surface *ps_tex = screen->get_tex_surface(
+         screen, pt, 0, 0, 0, PIPE_BUFFER_USAGE_GPU_WRITE );
+      pipe->surface_copy(pipe,
+			 ps_tex, /* dest */
+			 0, 0, /* destx/y */
+			 ps_read,
+			 0, 0, src->width[0], src->height[0]);
+      pipe_surface_reference(&ps_read, NULL);
+      pipe_surface_reference(&ps_tex, NULL);
+   }
+
+   return pt;
+}
+
+
+void renderer_copy_pixmap(struct xorg_renderer *r,
+                          struct exa_pixmap_priv *dst_priv, int dx, int dy,
+                          struct exa_pixmap_priv *src_priv, int sx, int sy,
+                          int width, int height)
+{
+   float dst_loc[4], src_loc[4];
+   float dst_bounds[4], src_bounds[4];
+   float src_shift[4], dst_shift[4], shift[4];
+   struct pipe_texture *dst = dst_priv->tex;
+   struct pipe_texture *src = src_priv->tex;
+
+   if (r->pipe->is_texture_referenced(r->pipe, src, 0, 0) &
+       PIPE_REFERENCED_FOR_WRITE)
+      r->pipe->flush(r->pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+
+   dst_loc[0] = dx;
+   dst_loc[1] = dy;
+   dst_loc[2] = width;
+   dst_loc[3] = height;
+   dst_bounds[0] = 0.f;
+   dst_bounds[1] = 0.f;
+   dst_bounds[2] = dst->width[0];
+   dst_bounds[3] = dst->height[0];
+
+   src_loc[0] = sx;
+   src_loc[1] = sy;
+   src_loc[2] = width;
+   src_loc[3] = height;
+   src_bounds[0] = 0.f;
+   src_bounds[1] = 0.f;
+   src_bounds[2] = src->width[0];
+   src_bounds[3] = src->height[0];
+
+   bound_rect(src_loc, src_bounds, src_shift);
+   bound_rect(dst_loc, dst_bounds, dst_shift);
+   shift[0] = src_shift[0] - dst_shift[0];
+   shift[1] = src_shift[1] - dst_shift[1];
+
+   if (shift[0] < 0)
+      shift_rectx(src_loc, src_bounds, -shift[0]);
+   else
+      shift_rectx(dst_loc, dst_bounds, shift[0]);
+
+   if (shift[1] < 0)
+      shift_recty(src_loc, src_bounds, -shift[1]);
+   else
+      shift_recty(dst_loc, dst_bounds, shift[1]);
+
+   sync_size(src_loc, dst_loc);
+
+   if (src_loc[2] >= 0 && src_loc[3] >= 0 &&
+       dst_loc[2] >= 0 && dst_loc[3] >= 0) {
+      struct pipe_texture *temp_src = src;
+
+      if (src == dst)
+         temp_src = create_sampler_texture(r, src);
+
+      renderer_copy_texture(r,
+                            temp_src,
+                            src_loc[0],
+                            src_loc[1],
+                            src_loc[0] + src_loc[2],
+                            src_loc[1] + src_loc[3],
+                            dst,
+                            dst_loc[0],
+                            dst_loc[1],
+                            dst_loc[0] + dst_loc[2],
+                            dst_loc[1] + dst_loc[3]);
+
+      if (src == dst)
+         pipe_texture_reference(&temp_src, NULL);
+   }
+}
+
+void renderer_draw_solid_rect(struct xorg_renderer *r,
+                              int x0, int y0,
+                              int x1, int y1,
+                              float *color)
+{
+   struct pipe_context *pipe = r->pipe;
+   struct pipe_buffer *buf = 0;
+
+   /* 1st vertex */
+   setup_vertex0(r->vertices2[0], x0, y0, color);
+   /* 2nd vertex */
+   setup_vertex0(r->vertices2[1], x1, y0, color);
+   /* 3rd vertex */
+   setup_vertex0(r->vertices2[2], x1, y1, color);
+   /* 4th vertex */
+   setup_vertex0(r->vertices2[3], x0, y1, color);
+
+   buf = pipe_user_buffer_create(pipe->screen,
+                                 r->vertices2,
+                                 sizeof(r->vertices2));
+
+
+   if (buf) {
+      util_draw_vertex_buffer(pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              2); /* attribs/vert */
+
+      pipe_buffer_reference(&buf, NULL);
+   }
+}
+
+void renderer_draw_textures(struct xorg_renderer *r,
+                            int *pos,
+                            int width, int height,
+                            struct pipe_texture **textures,
+                            int num_textures)
+{
+   struct pipe_context *pipe = r->pipe;
+   struct pipe_buffer *buf = 0;
+
+   switch(num_textures) {
+   case 1:
+      buf = setup_vertex_data1(r,
+                               pos[0], pos[1], /* src */
+                               pos[4], pos[5], /* dst */
+                               width, height,
+                               textures[0]);
+      break;
+   case 2:
+      buf = setup_vertex_data2(r,
+                               pos[0], pos[1], /* src */
+                               pos[2], pos[3], /* mask */
+                               pos[4], pos[5], /* dst */
+                               width, height,
+                               textures[0], textures[1]);
+      break;
+   default:
+      debug_assert(!"Unsupported number of textures");
+      break;
+   }
+
+   if (buf) {
+      int num_attribs = 1; /*pos*/
+      num_attribs += num_textures;
+
+      util_draw_vertex_buffer(pipe, buf, 0,
+                              PIPE_PRIM_TRIANGLE_FAN,
+                              4,  /* verts */
+                              num_attribs); /* attribs/vert */
+
+      pipe_buffer_reference(&buf, NULL);
+   }
+}
diff --git a/src/gallium/state_trackers/xorg/xorg_renderer.h b/src/gallium/state_trackers/xorg/xorg_renderer.h
new file mode 100644
index 0000000000..b6296d5fd6
--- /dev/null
+++ b/src/gallium/state_trackers/xorg/xorg_renderer.h
@@ -0,0 +1,53 @@
+#ifndef XORG_RENDERER_H
+#define XORG_RENDERER_H
+
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+
+struct xorg_shaders;
+struct exa_pixmap_priv;
+
+struct xorg_renderer {
+   struct pipe_context *pipe;
+
+   struct cso_context *cso;
+   struct xorg_shaders *shaders;
+
+   struct pipe_constant_buffer vs_const_buffer;
+   struct pipe_constant_buffer fs_const_buffer;
+
+   /* we should combine these two */
+   float vertices2[4][2][4];
+   float vertices3[4][3][4];
+};
+
+struct xorg_renderer *renderer_create(struct pipe_context *pipe);
+void renderer_destroy(struct xorg_renderer *renderer);
+
+void renderer_bind_framebuffer(struct xorg_renderer *r,
+                               struct exa_pixmap_priv *priv);
+void renderer_bind_viewport(struct xorg_renderer *r,
+                            struct exa_pixmap_priv *dst);
+void renderer_bind_rasterizer(struct xorg_renderer *r);
+void renderer_set_constants(struct xorg_renderer *r,
+                            int shader_type,
+                            const float *buffer,
+                            int size);
+void renderer_copy_pixmap(struct xorg_renderer *r,
+                          struct exa_pixmap_priv *dst_priv, int dx, int dy,
+                          struct exa_pixmap_priv *src_priv, int sx, int sy,
+                          int width, int height);
+
+void renderer_draw_solid_rect(struct xorg_renderer *r,
+                              int x0, int y0,
+                              int x1, int y1,
+                              float *color);
+
+void renderer_draw_textures(struct xorg_renderer *r,
+                            int *pos,
+                            int width, int height,
+                            struct pipe_texture **textures,
+                            int num_textures);
+
+
+#endif
diff --git a/src/gallium/state_trackers/xorg/xorg_tracker.h b/src/gallium/state_trackers/xorg/xorg_tracker.h
index db2f16f63e..24e1a4928e 100644
--- a/src/gallium/state_trackers/xorg/xorg_tracker.h
+++ b/src/gallium/state_trackers/xorg/xorg_tracker.h
@@ -51,6 +51,8 @@
 
 #define DRV_ERROR(msg)	xf86DrvMsg(pScrn->scrnIndex, X_ERROR, msg);
 
+struct exa_context;
+
 typedef struct
 {
     int lastInstance;
@@ -92,7 +94,7 @@ typedef struct _modesettingRec
     boolean ds_depth_bits_last;
 
     /* exa */
-    void *exa;
+    struct exa_context *exa;
     Bool noEvict;
     Bool debug_fallback;
 
diff --git a/src/gallium/state_trackers/xorg/xorg_xv.c b/src/gallium/state_trackers/xorg/xorg_xv.c
index 27d52700ec..efac9275b2 100644
--- a/src/gallium/state_trackers/xorg/xorg_xv.c
+++ b/src/gallium/state_trackers/xorg/xorg_xv.c
@@ -4,10 +4,34 @@
 #include <X11/extensions/Xv.h>
 #include <fourcc.h>
 
+#include "xorg_exa.h"
+#include "xorg_renderer.h"
+
+#include "pipe/p_screen.h"
+#include "pipe/p_inlines.h"
+
 /*XXX get these from pipe's texture limits */
 #define IMAGE_MAX_WIDTH		2048
 #define IMAGE_MAX_HEIGHT	2048
 
+#define RES_720P_X 1280
+#define RES_720P_Y 720
+
+
+/* The ITU-R BT.601 conversion matrix for SDTV. */
+static const float bt_601[] = {
+    1.0, 0.0, 1.4075,
+    1.0, -0.3455, -0.7169,
+    1.0, 1.7790, 0.
+};
+
+/* The ITU-R BT.709 conversion matrix for HDTV. */
+static const float bt_709[] = {
+    1.0, 0.0, 1.581,
+    1.0, -0.1881, -0.47,
+    1.0, 1.8629, 0.
+};
+
 #define MAKE_ATOM(a) MakeAtom(a, sizeof(a) - 1, TRUE)
 
 static Atom xvBrightness, xvContrast;
@@ -39,10 +63,17 @@ static XF86ImageRec Images[NUM_IMAGES] = {
 };
 
 struct xorg_xv_port_priv {
+   struct xorg_renderer *r;
+
    RegionRec clip;
 
    int brightness;
    int contrast;
+
+   int current_set;
+   /* juggle two sets of seperate Y, U and V
+    * textures */
+   struct pipe_texture *yuv[2][3];
 };
 
 
@@ -106,6 +137,225 @@ query_best_size(ScrnInfoPtr pScrn,
    *p_h = drw_h;
 }
 
+static INLINE struct pipe_texture *
+create_component_texture(struct pipe_context *pipe,
+                         int width, int height)
+{
+   struct pipe_screen *screen = pipe->screen;
+   struct pipe_texture *tex = 0;
+   struct pipe_texture templ;
+
+   memset(&templ, 0, sizeof(templ));
+   templ.target = PIPE_TEXTURE_2D;
+   templ.format = PIPE_FORMAT_L8_UNORM;
+   templ.last_level = 0;
+   templ.width[0] = width;
+   templ.height[0] = height;
+   templ.depth[0] = 1;
+   pf_get_block(PIPE_FORMAT_L8_UNORM, &templ.block);
+   templ.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER;
+
+   tex = screen->texture_create(screen, &templ);
+
+   return tex;
+}
+
+static int
+check_yuv_textures(struct xorg_xv_port_priv *priv,  int width, int height)
+{
+   struct pipe_texture **dst = priv->yuv[priv->current_set];
+   if (!dst[0] ||
+       dst[0]->width[0] != width ||
+       dst[0]->height[0] != height) {
+      pipe_texture_reference(&dst[0], NULL);
+   }
+   if (!dst[1] ||
+       dst[1]->width[0] != width ||
+       dst[1]->height[0] != height) {
+      pipe_texture_reference(&dst[1], NULL);
+   }
+   if (!dst[2] ||
+       dst[2]->width[0] != width ||
+       dst[2]->height[0] != height) {
+      pipe_texture_reference(&dst[2], NULL);
+   }
+
+   if (!dst[0])
+      dst[0] = create_component_texture(priv->r->pipe, width, height);
+
+   if (!dst[1])
+      dst[1] = create_component_texture(priv->r->pipe, width, height);
+
+   if (!dst[2])
+      dst[2] = create_component_texture(priv->r->pipe, width, height);
+
+   if (!dst[0] || !dst[1] || !dst[2])
+      return BadAlloc;
+
+   return Success;
+}
+
+static void
+copy_packed_data(ScrnInfoPtr pScrn,
+                 struct xorg_xv_port_priv *port,
+                 int id,
+                 unsigned char *buf,
+                 int srcPitch,
+                 int left,
+                 int top,
+                 int w, int h)
+{
+   unsigned char *src;
+   int i, j;
+   struct pipe_texture **dst = port->yuv[port->current_set];
+   struct pipe_transfer *ytrans, *utrans, *vtrans;
+   struct pipe_screen *screen = port->r->pipe->screen;
+   char *ymap, *vmap, *umap;
+   unsigned char y1, y2, u, v;
+   int yidx, uidx, vidx;
+
+   src = buf + (top * srcPitch) + (left << 1);
+
+   ytrans = screen->get_tex_transfer(screen, dst[0],
+                                     0, 0, 0,
+                                     PIPE_TRANSFER_WRITE,
+                                     left, top, w, h);
+   utrans = screen->get_tex_transfer(screen, dst[1],
+                                     0, 0, 0,
+                                     PIPE_TRANSFER_WRITE,
+                                     left, top, w, h);
+   vtrans = screen->get_tex_transfer(screen, dst[2],
+                                     0, 0, 0,
+                                     PIPE_TRANSFER_WRITE,
+                                     left, top, w, h);
+
+   ymap = screen->transfer_map(screen, ytrans);
+   umap = screen->transfer_map(screen, utrans);
+   vmap = screen->transfer_map(screen, vtrans);
+
+   switch (id) {
+   case FOURCC_YV12: {
+      int y_array_size = w * h;
+      for (i = 0; i < w; ++i) {
+         for (j = 0; i < h; ++j) {
+            /*XXX use src? */
+            y1  = buf[j*w + i];
+            u   = buf[(j/2) * (w/2) + i/2 + y_array_size];
+            v   = buf[(j/2) * (w/2) + i/2 + y_array_size + y_array_size/4];
+            ymap[yidx++] = y1;
+            umap[uidx++] = u;
+            vmap[vidx++] = v;
+         }
+      }
+   }
+      break;
+   case FOURCC_YUY2:
+      for (j = 0; j < h; ++j) {
+         for (i = 0; i < w; ++i) {
+            /* extracting two pixels */
+            y1 = buf[0];
+            u  = buf[1];
+            y2 = buf[2];
+            v  = buf[3];
+
+            ymap[yidx++] = y1;
+            ymap[yidx++] = y2;
+            umap[uidx++] = u;
+            umap[uidx++] = u;
+            vmap[vidx++] = v;
+            vmap[vidx++] = v;
+         }
+      }
+      break;
+   default:
+      debug_assert(!"Unsupported yuv format!");
+      break;
+   }
+
+   screen->transfer_unmap(screen, ytrans);
+   screen->transfer_unmap(screen, utrans);
+   screen->transfer_unmap(screen, vtrans);
+   screen->tex_transfer_destroy(ytrans);
+   screen->tex_transfer_destroy(utrans);
+   screen->tex_transfer_destroy(vtrans);
+}
+
+static void
+setup_video_constants(struct xorg_renderer *r, boolean hdtv)
+{
+   struct pipe_context *pipe = r->pipe;
+   const int param_bytes = 9 * sizeof(float);
+   struct pipe_constant_buffer *cbuf = &r->vs_const_buffer;
+
+   pipe_buffer_reference(&cbuf->buffer, NULL);
+   cbuf->buffer = pipe_buffer_create(pipe->screen, 16,
+                                     PIPE_BUFFER_USAGE_CONSTANT,
+                                     param_bytes);
+
+   if (cbuf->buffer) {
+      const float *video_constants = (hdtv) ? bt_709 : bt_601;
+
+      pipe_buffer_write(pipe->screen, cbuf->buffer,
+                        0, param_bytes, video_constants);
+   }
+   pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, cbuf);
+}
+
+static int
+display_video(ScrnInfoPtr pScrn, struct xorg_xv_port_priv *pPriv, int id,
+              RegionPtr dstRegion,
+              short width, short height,
+              int x1, int y1, int x2, int y2,
+              short src_w, short src_h, short drw_w, short drw_h,
+              PixmapPtr pPixmap)
+{
+   BoxPtr pbox;
+   int nbox;
+   int dxo, dyo;
+   Bool hdtv;
+   float tc0[2], tc1[2], tc2[2];
+
+   hdtv = ((src_w >= RES_720P_X) && (src_h >= RES_720P_Y));
+   setup_video_constants(pPriv->r, hdtv);
+
+   REGION_TRANSLATE(pScrn->pScreen, dstRegion, -pPixmap->screen_x,
+                    -pPixmap->screen_y);
+
+   dxo = dstRegion->extents.x1;
+   dyo = dstRegion->extents.y1;
+
+   pbox = REGION_RECTS(dstRegion);
+   nbox = REGION_NUM_RECTS(dstRegion);
+
+   while (nbox--) {
+      int box_x1 = pbox->x1;
+      int box_y1 = pbox->y1;
+      int box_x2 = pbox->x2;
+      int box_y2 = pbox->y2;
+
+      tc0[0] = (double) (box_x1 - dxo) / (double) drw_w; /* u0 */
+      tc0[1] = (double) (box_y1 - dyo) / (double) drw_h; /* v0 */
+      tc1[0] = (double) (box_x2 - dxo) / (double) drw_w; /* u1 */
+      tc1[1] = tc0[1];
+      tc2[0] = tc0[0];
+      tc2[1] = (double) (box_y2 - dyo) / (double) drw_h; /* v1 */
+
+#if 0
+      x = box_x1;
+      y = box_y1;
+      w = box_x2 - box_x1;
+      h = box_y2 - box_y1;
+
+      pbox++;
+      draw_yuv(pScrn, x, y, w, h, &src, 1, FALSE,
+               0, tc0, tc1, tc2, 1,
+               pPriv->conversionData);
+#endif
+   }
+   DamageDamageRegion(&pPixmap->drawable, dstRegion);
+   return TRUE;
+}
+
 static int
 put_image(ScrnInfoPtr pScrn,
           short src_x, short src_y,
@@ -117,7 +367,57 @@ put_image(ScrnInfoPtr pScrn,
           Bool sync, RegionPtr clipBoxes, pointer data,
           DrawablePtr pDraw)
 {
-   return 0;
+   struct xorg_xv_port_priv *pPriv = (struct xorg_xv_port_priv *) data;
+   ScreenPtr pScreen = screenInfo.screens[pScrn->scrnIndex];
+   PixmapPtr pPixmap;
+   INT32 x1, x2, y1, y2;
+   int srcPitch;
+   BoxRec dstBox;
+   int ret;
+
+   /* Clip */
+   x1 = src_x;
+   x2 = src_x + src_w;
+   y1 = src_y;
+   y2 = src_y + src_h;
+
+   dstBox.x1 = drw_x;
+   dstBox.x2 = drw_x + drw_w;
+   dstBox.y1 = drw_y;
+   dstBox.y2 = drw_y + drw_h;
+
+   if (!xf86XVClipVideoHelper(&dstBox, &x1, &x2, &y1, &y2, clipBoxes,
+			      width, height))
+      return Success;
+
+   switch (id) {
+   case FOURCC_UYVY:
+   case FOURCC_YUY2:
+   default:
+      srcPitch = width << 1;
+      break;
+   }
+
+   ret = check_yuv_textures(pPriv, width, height);
+
+   if (ret)
+      return ret;
+
+   copy_packed_data(pScrn, pPriv, id, buf, srcPitch,
+                    src_x, src_y, width, height);
+
+   if (pDraw->type == DRAWABLE_WINDOW) {
+      pPixmap = (*pScreen->GetWindowPixmap)((WindowPtr)pDraw);
+   } else {
+      pPixmap = (PixmapPtr)pDraw;
+   }
+
+   display_video(pScrn, pPriv, id, clipBoxes, width, height,
+                 x1, y1, x2, y2,
+                 src_w, src_h, drw_w, drw_h, pPixmap);
+
+   pPriv->current_set = (pPriv->current_set + 1) & 1;
+   return Success;
 }
 
 static int
@@ -152,10 +452,8 @@ query_image_attributes(ScrnInfoPtr pScrn,
 }
 
 static struct xorg_xv_port_priv *
-port_priv_create(ScreenPtr pScreen)
+port_priv_create(struct xorg_renderer *r)
 {
-   /*ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];*/
-   /*modesettingPtr ms = modesettingPTR(pScrn);*/
    struct xorg_xv_port_priv *priv = NULL;
 
    priv = calloc(1, sizeof(struct xorg_xv_port_priv));
@@ -163,16 +461,20 @@ port_priv_create(ScreenPtr pScreen)
    if (!priv)
       return NULL;
 
+   priv->r = r;
+
    REGION_NULL(pScreen, &priv->clip);
 
+   debug_assert(priv && priv->r);
+
    return priv;
 }
 
 static XF86VideoAdaptorPtr
 xorg_setup_textured_adapter(ScreenPtr pScreen)
 {
-   /*ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];*/
-   /*modesettingPtr ms = modesettingPTR(pScrn);*/
+   ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
+   modesettingPtr ms = modesettingPTR(pScrn);
    XF86VideoAdaptorPtr adapt;
    XF86AttributePtr attrs;
    DevUnion *dev_unions;
@@ -181,6 +483,9 @@ xorg_setup_textured_adapter(ScreenPtr pScreen)
 
    nattributes = NUM_TEXTURED_ATTRIBUTES;
 
+   debug_assert(ms->exa);
+   debug_assert(ms->exa->renderer);
+
    adapt = calloc(1, sizeof(XF86VideoAdaptorRec));
    dev_unions = calloc(nports, sizeof(DevUnion));
    attrs = calloc(nattributes, sizeof(XF86AttributeRec));
@@ -218,7 +523,7 @@ xorg_setup_textured_adapter(ScreenPtr pScreen)
 
    for (i = 0; i < nports; i++) {
       struct xorg_xv_port_priv *priv =
-         port_priv_create(pScreen);
+         port_priv_create(ms->exa->renderer);
 
       adapt->pPortPrivates[i].ptr = (pointer) (priv);
       adapt->nPorts++;