Merge branch 'xa_branch'

Conflicts:
	configure.ac

Signed-off-by: Thomas Hellstrom <[email protected]>

14 files changed, 3606 insertions, 0 deletions

diff --git a/src/gallium/state_trackers/xa/Makefile b/src/gallium/state_trackers/xa/Makefile
new file mode 100644
index 00000000000..d95f9382630
--- /dev/null
+++ b/src/gallium/state_trackers/xa/Makefile
@@ -0,0 +1,67 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+##### MACROS #####
+
+XA_MAJOR = 0
+XA_MINOR = 4
+XA_TINY = 0
+XA_CFLAGS = -g -fPIC -Wall
+
+XA_INCLUDES= -I$(TOP)/src/gallium/ \
+	-I$(TOP)/src/gallium/auxiliary \
+	-I$(TOP)/src/gallium/include \
+	-I$(TOP)/src/gallium/winsys \
+	-I$(TOP)/src/gallium/drivers
+
+XA_LIB = xatracker
+XA_LIB_NAME = lib$(XA_LIB).o
+XA_LIB_DEPS =
+
+COMMON_GALLIUM_SOURCES=
+
+SOURCES = \
+	xa_tracker.c \
+	xa_context.c \
+	xa_renderer.c \
+	xa_tgsi.c \
+	xa_yuv.c \
+	xa_composite.c
+OBJECTS = $(SOURCES:.c=.o)
+
+##### RULES #####
+
+.c.o:
+	$(CC) -c $(XA_CFLAGS) $(XA_INCLUDES) $<
+
+
+##### TARGETS #####
+
+default: $(XA_LIB_NAME)
+
+
+# Make the library
+$(XA_LIB_NAME): depend $(OBJECTS)
+	$(CC) -r -nostdlib -o $(XA_LIB_NAME) $(OBJECTS)
+
+install: FORCE
+
+clean:
+	-rm -f *.o *~
+	-rm -f *.lo
+	-rm -f *.la
+	-rm -f *.pc
+	-rm -rf .libs
+	-rm -f depend depend.bak
+
+
+depend: $(SOURCES)
+	@ echo "running $(MKDEP)"
+	@ rm -f depend
+	@ touch depend
+	@ $(MKDEP) $(MKDEP_OPTIONS) -I$(TOP)/include $(XA_INCLUDES) $(SOURCES) \
+		> /dev/null
+
+-include depend
+
+FORCE:
diff --git a/src/gallium/state_trackers/xa/README b/src/gallium/state_trackers/xa/README
new file mode 100644
index 00000000000..1f08861588c
--- /dev/null
+++ b/src/gallium/state_trackers/xa/README
@@ -0,0 +1,72 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ * Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+The XA state tracker is intended as a versioned interface to gallium for
+xorg driver writers. Initially it's mostly based on Zack Rusin's
+composite / video work for the Xorg state tracker.
+
+The motivation behind this state tracker is that the Xorg state tracker has
+a number of interfaces to work with:
+
+1) The Xorg sdk (versioned)
+2) Gallium3D (not versioned)
+3) KMS modesetting (versioned)
+4) Driver-private (hopefully versioned)
+
+Since Gallium3D is not versioned, the Xorg state tracker needs to be compiled
+with Gallium, but it's really beneficial to be able to compile xorg drivers
+standalone.
+
+Therefore the xa state tracker is intended to supply the following
+functionality:
+
+1) Versioning.
+2) Surface functionality (creation and copying for a basic dri2 implementation)
+3) YUV blits for textured Xv.
+4) Solid fills without ROP functionality.
+5) Copies with format conversion and - reinterpretation but without ROP
+6) Xrender- type compositing for general acceleration.
+
+
+The first user will be the vmwgfx xorg driver. When there are more users,
+we need to be able to load the appropriate gallium pipe driver, and we
+should investigate sharing the loadig mechanism with the EGL state tracker.
+
+IMPORTANT:
+Version compatibilities:
+While this library remains OUTSIDE any mesa release branch,
+and the major version number is still 0. Any minor bump should be viewed as
+an incompatibility event, and any user of this library should test for that
+and refuse to use the library if minor versions differ.
+As soon as the library enters a mesa release branch, if not earlier, major
+will be bumped to 1, and normal incompatibility rules (major bump)
+will be followed.
+It is allowed to add function interfaces while only bumping minor. Any
+user that uses these function interfaces must therefore use lazy symbol
+lookups and test minor for compatibility before using such a function.
diff --git a/src/gallium/state_trackers/xa/xa-indent b/src/gallium/state_trackers/xa/xa-indent
new file mode 100755
index 00000000000..1972e53226f
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa-indent
@@ -0,0 +1,3 @@
+#
+indent --linux-style -i4 -ip4 -bad -bap -psl $*
+
diff --git a/src/gallium/state_trackers/xa/xa_composite.c b/src/gallium/state_trackers/xa/xa_composite.c
new file mode 100644
index 00000000000..5389af6f363
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_composite.c
@@ -0,0 +1,499 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ * Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#include "xa_composite.h"
+#include "xa_context.h"
+#include "xa_priv.h"
+#include "cso_cache/cso_context.h"
+#include "util/u_sampler.h"
+#include "util/u_inlines.h"
+
+
+/*XXX also in Xrender.h but the including it here breaks compilition */
+#define XFixedToDouble(f)    (((double) (f)) / 65536.)
+
+struct xa_composite_blend {
+    enum xa_composite_op op : 8;
+
+    unsigned alpha_dst : 4;
+    unsigned alpha_src : 4;
+
+    unsigned rgb_src : 8;    /**< PIPE_BLENDFACTOR_x */
+    unsigned rgb_dst : 8;    /**< PIPE_BLENDFACTOR_x */
+};
+
+#define XA_BLEND_OP_OVER 3
+static const struct xa_composite_blend xa_blends[] = {
+    { xa_op_clear,
+      0, 0, PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_ZERO},
+    { xa_op_src,
+      0, 0, PIPE_BLENDFACTOR_ONE, PIPE_BLENDFACTOR_ZERO},
+    { xa_op_dst,
+      0, 0, PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_ONE},
+    { xa_op_over,
+      0, 1, PIPE_BLENDFACTOR_ONE, PIPE_BLENDFACTOR_INV_SRC_ALPHA},
+    { xa_op_over_reverse,
+      1, 0, PIPE_BLENDFACTOR_INV_DST_ALPHA, PIPE_BLENDFACTOR_ONE},
+    { xa_op_in,
+      1, 0, PIPE_BLENDFACTOR_DST_ALPHA, PIPE_BLENDFACTOR_ZERO},
+    { xa_op_in_reverse,
+      0, 1, PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_SRC_ALPHA},
+    { xa_op_out,
+      1, 0, PIPE_BLENDFACTOR_INV_DST_ALPHA, PIPE_BLENDFACTOR_ZERO},
+    { xa_op_out_reverse,
+      0, 1, PIPE_BLENDFACTOR_ZERO, PIPE_BLENDFACTOR_INV_SRC_ALPHA},
+    { xa_op_atop,
+      1, 1, PIPE_BLENDFACTOR_DST_ALPHA, PIPE_BLENDFACTOR_INV_SRC_ALPHA},
+    { xa_op_atop_reverse,
+      1, 1, PIPE_BLENDFACTOR_INV_DST_ALPHA, PIPE_BLENDFACTOR_SRC_ALPHA},
+    { xa_op_xor,
+      1, 1, PIPE_BLENDFACTOR_INV_DST_ALPHA, PIPE_BLENDFACTOR_INV_SRC_ALPHA},
+    { xa_op_add,
+      0, 0, PIPE_BLENDFACTOR_ONE, PIPE_BLENDFACTOR_ONE},
+};
+
+
+static boolean
+blend_for_op(struct xa_composite_blend *blend,
+	     enum xa_composite_op op,
+	     struct xa_picture *src_pic,
+	     struct xa_picture *mask_pic,
+	     struct xa_picture *dst_pic)
+{
+    const int num_blends =
+	sizeof(xa_blends)/sizeof(struct xa_composite_blend);
+    int i;
+    boolean supported = FALSE;
+
+    /*
+     * our default in case something goes wrong
+     */
+
+    *blend = xa_blends[XA_BLEND_OP_OVER];
+
+    for (i = 0; i < num_blends; ++i) {
+	if (xa_blends[i].op == op) {
+	    *blend = xa_blends[i];
+	    supported = TRUE;
+	}
+    }
+
+
+    /*
+     * If there's no dst alpha channel, adjust the blend op so that we'll treat
+     * it as always 1.
+     */
+
+    if (dst_pic &&
+	xa_format_a(dst_pic->pict_format) == 0 &&
+	blend->alpha_dst) {
+	if (blend->rgb_src == PIPE_BLENDFACTOR_DST_ALPHA)
+	    blend->rgb_src = PIPE_BLENDFACTOR_ONE;
+	else if (blend->rgb_src == PIPE_BLENDFACTOR_INV_DST_ALPHA)
+	    blend->rgb_src = PIPE_BLENDFACTOR_ZERO;
+    }
+
+    /*
+     * If the source alpha is being used, then we should only be in a case where
+     * the source blend factor is 0, and the source blend value is the mask
+     * channels multiplied by the source picture's alpha.
+     */
+    if (mask_pic && mask_pic->component_alpha &&
+	xa_format_rgb(mask_pic->pict_format) &&
+	blend->alpha_src) {
+	if (blend->rgb_dst == PIPE_BLENDFACTOR_SRC_ALPHA) {
+	    blend->rgb_dst = PIPE_BLENDFACTOR_SRC_COLOR;
+	} else if (blend->rgb_dst == PIPE_BLENDFACTOR_INV_SRC_ALPHA) {
+	    blend->rgb_dst = PIPE_BLENDFACTOR_INV_SRC_COLOR;
+	}
+    }
+
+    return supported;
+}
+
+
+static INLINE int
+xa_repeat_to_gallium(int mode)
+{
+    switch(mode) {
+    case xa_wrap_clamp_to_border:
+	return PIPE_TEX_WRAP_CLAMP_TO_BORDER;
+    case xa_wrap_repeat:
+	return PIPE_TEX_WRAP_REPEAT;
+    case xa_wrap_mirror_repeat:
+	return PIPE_TEX_WRAP_MIRROR_REPEAT;
+    case xa_wrap_clamp_to_edge:
+	return PIPE_TEX_WRAP_CLAMP_TO_EDGE;
+    default:
+	break;
+    }
+    return PIPE_TEX_WRAP_REPEAT;
+}
+
+static INLINE boolean
+xa_filter_to_gallium(int xrender_filter, int *out_filter)
+{
+
+    switch (xrender_filter) {
+    case xa_filter_nearest:
+	*out_filter = PIPE_TEX_FILTER_NEAREST;
+	break;
+    case xa_filter_linear:
+	*out_filter = PIPE_TEX_FILTER_LINEAR;
+	break;
+    default:
+	*out_filter = PIPE_TEX_FILTER_NEAREST;
+	return FALSE;
+    }
+    return TRUE;
+}
+
+static int
+xa_is_filter_accelerated(struct xa_picture *pic)
+{
+    int filter;
+    if (pic && !xa_filter_to_gallium(pic->filter, &filter))
+	return 0;
+    return 1;
+}
+
+int
+xa_composite_check_accelerated(const struct xa_composite *comp)
+{
+    struct xa_composite_blend blend;
+    struct xa_picture *src_pic = comp->src;
+
+    if (!xa_is_filter_accelerated(src_pic) ||
+	!xa_is_filter_accelerated(comp->mask)) {
+	return XA_ERR_INVAL;
+    }
+
+
+    if (src_pic->src_pict) {
+	if (src_pic->src_pict->type != xa_src_pict_solid_fill)
+	    return XA_ERR_INVAL;
+    }
+
+    if (blend_for_op(&blend, comp->op, comp->src, comp->mask, comp->dst)) {
+	struct xa_picture *mask = comp->mask;
+	if (mask && mask->component_alpha &&
+	    xa_format_rgb(mask->pict_format)) {
+	    if (blend.alpha_src && blend.rgb_src != PIPE_BLENDFACTOR_ZERO) {
+		return XA_ERR_INVAL;
+	    }
+	}
+
+	return XA_ERR_NONE;
+    }
+    return XA_ERR_INVAL;
+}
+
+static void
+bind_composite_blend_state(struct xa_context *ctx,
+			   const struct xa_composite *comp)
+{
+    struct xa_composite_blend blend_opt;
+    struct pipe_blend_state blend;
+
+    blend_for_op(&blend_opt, comp->op, comp->src, comp->mask, comp->dst);
+
+    memset(&blend, 0, sizeof(struct pipe_blend_state));
+    blend.rt[0].blend_enable = 1;
+    blend.rt[0].colormask = PIPE_MASK_RGBA;
+
+    blend.rt[0].rgb_src_factor   = blend_opt.rgb_src;
+    blend.rt[0].alpha_src_factor = blend_opt.rgb_src;
+    blend.rt[0].rgb_dst_factor   = blend_opt.rgb_dst;
+    blend.rt[0].alpha_dst_factor = blend_opt.rgb_dst;
+
+    cso_set_blend(ctx->cso, &blend);
+}
+
+static unsigned int
+picture_format_fixups(struct xa_picture *src_pic,
+		      struct xa_picture *dst_pic,
+		      int mask)
+{
+    boolean set_alpha = FALSE;
+    boolean swizzle = FALSE;
+    unsigned ret = 0;
+    struct xa_surface *src = src_pic->srf;
+    enum xa_formats src_hw_format, src_pic_format;
+    enum xa_surface_type src_hw_type, src_pic_type;
+
+    if (!src)
+	return 0;
+
+    src_hw_format = xa_surface_format(src);
+    src_pic_format = src_pic->pict_format;
+
+    if (!src || src_hw_format == src_pic_format) {
+	if (src_pic_format == xa_format_a8) {
+	    if (mask)
+		return FS_MASK_LUMINANCE;
+	    else if (dst_pic->pict_format != xa_format_a8) {
+
+		/*
+		 * if both dst and src are luminance then
+		 * we don't want to swizzle the alpha (X) of the
+		 * source into W component of the dst because
+		 * it will break our destination
+		 */
+		return FS_SRC_LUMINANCE;
+	    }
+	}
+	return 0;
+    }
+
+    src_hw_type = xa_format_type(src_hw_format);
+    src_pic_type = xa_format_type(src_pic_format);
+
+    swizzle = ((src_hw_type == xa_type_argb &&
+		src_pic_type == xa_type_abgr) ||
+	       ((src_hw_type == xa_type_abgr &&
+		 src_pic_type == xa_type_argb)));
+
+    if (!swizzle && (src_hw_type != src_pic_type))
+	return 0;
+
+    set_alpha = (xa_format_type_is_color(src_pic_format) &&
+		 xa_format_a(src_pic_type) == 0);
+
+    if (set_alpha)
+	ret |= mask ? FS_MASK_SET_ALPHA : FS_SRC_SET_ALPHA;
+    if (swizzle)
+	ret |= mask ? FS_MASK_SWIZZLE_RGB : FS_SRC_SWIZZLE_RGB;
+
+    return ret;
+}
+
+static void
+bind_shaders(struct xa_context *ctx, const struct xa_composite *comp)
+{
+    unsigned vs_traits = 0, fs_traits = 0;
+    struct xa_shader shader;
+    struct xa_picture *src_pic = comp->src;
+    struct xa_picture *mask_pic = comp->mask;
+    struct xa_picture *dst_pic = comp->dst;
+
+    ctx->has_solid_color = FALSE;
+
+    if (src_pic) {
+	if (src_pic->wrap == xa_wrap_clamp_to_border && src_pic->has_transform)
+	    fs_traits |= FS_SRC_REPEAT_NONE;
+
+	if (src_pic->src_pict) {
+	    if (src_pic->src_pict->type == xa_src_pict_solid_fill) {
+		fs_traits |= FS_SOLID_FILL;
+		vs_traits |= VS_SOLID_FILL;
+		xa_pixel_to_float4(src_pic->src_pict->solid_fill.color,
+				   ctx->solid_color);
+		ctx->has_solid_color = TRUE;
+	    }
+	} else {
+	    fs_traits |= FS_COMPOSITE;
+	    vs_traits |= VS_COMPOSITE;
+	}
+
+	fs_traits |= picture_format_fixups(src_pic, dst_pic, 0);
+    }
+
+    if (mask_pic) {
+	vs_traits |= VS_MASK;
+	fs_traits |= FS_MASK;
+	if (mask_pic->wrap == xa_wrap_clamp_to_border &&
+	    mask_pic->has_transform)
+	    fs_traits |= FS_MASK_REPEAT_NONE;
+
+	if (mask_pic->component_alpha) {
+	    struct xa_composite_blend blend;
+	    blend_for_op(&blend, comp->op, src_pic, mask_pic, NULL);
+	    if (blend.alpha_src) {
+		fs_traits |= FS_CA_SRCALPHA;
+	    } else
+		fs_traits |= FS_CA_FULL;
+	}
+
+	fs_traits |= picture_format_fixups(mask_pic, dst_pic, 1);
+    }
+
+    shader = xa_shaders_get(ctx->shaders, vs_traits, fs_traits);
+    cso_set_vertex_shader_handle(ctx->cso, shader.vs);
+    cso_set_fragment_shader_handle(ctx->cso, shader.fs);
+}
+
+static void
+bind_samplers(struct xa_context *ctx,
+	      const struct xa_composite *comp)
+{
+    struct pipe_sampler_state *samplers[PIPE_MAX_SAMPLERS];
+    struct pipe_sampler_state src_sampler, mask_sampler;
+    struct pipe_sampler_view view_templ;
+    struct pipe_sampler_view *src_view;
+    struct pipe_context *pipe = ctx->pipe;
+    struct xa_picture *src_pic = comp->src;
+    struct xa_picture *mask_pic = comp->mask;
+
+    ctx->num_bound_samplers = 0;
+
+    memset(&src_sampler, 0, sizeof(struct pipe_sampler_state));
+    memset(&mask_sampler, 0, sizeof(struct pipe_sampler_state));
+
+    if (src_pic) {
+	if (ctx->has_solid_color) {
+	    samplers[0] = NULL;
+	    pipe_sampler_view_reference(&ctx->bound_sampler_views[0], NULL);
+	} else {
+	    unsigned src_wrap = xa_repeat_to_gallium(src_pic->wrap);
+	    int filter;
+
+	    (void) xa_filter_to_gallium(src_pic->filter, &filter);
+
+	    src_sampler.wrap_s = src_wrap;
+	    src_sampler.wrap_t = src_wrap;
+	    src_sampler.min_img_filter = filter;
+	    src_sampler.mag_img_filter = filter;
+	    src_sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST;
+	    src_sampler.normalized_coords = 1;
+	    samplers[0] = &src_sampler;
+	    ctx->num_bound_samplers = 1;
+	    u_sampler_view_default_template(&view_templ,
+					    src_pic->srf->tex,
+					    src_pic->srf->tex->format);
+	    src_view = pipe->create_sampler_view(pipe, src_pic->srf->tex,
+						 &view_templ);
+	    pipe_sampler_view_reference(&ctx->bound_sampler_views[0], NULL);
+	    ctx->bound_sampler_views[0] = src_view;
+	}
+    }
+
+    if (mask_pic) {
+	unsigned mask_wrap = xa_repeat_to_gallium(mask_pic->wrap);
+	int filter;
+
+	(void) xa_filter_to_gallium(mask_pic->filter, &filter);
+
+	mask_sampler.wrap_s = mask_wrap;
+	mask_sampler.wrap_t = mask_wrap;
+	mask_sampler.min_img_filter = filter;
+	mask_sampler.mag_img_filter = filter;
+	src_sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST;
+	mask_sampler.normalized_coords = 1;
+	samplers[1] = &mask_sampler;
+	ctx->num_bound_samplers = 2;
+	u_sampler_view_default_template(&view_templ,
+					mask_pic->srf->tex,
+					mask_pic->srf->tex->format);
+	src_view = pipe->create_sampler_view(pipe, mask_pic->srf->tex,
+					     &view_templ);
+	pipe_sampler_view_reference(&ctx->bound_sampler_views[1], NULL);
+	ctx->bound_sampler_views[1] = src_view;
+    }
+
+    cso_set_samplers(ctx->cso, ctx->num_bound_samplers,
+		     (const struct pipe_sampler_state **)samplers);
+    cso_set_fragment_sampler_views(ctx->cso, ctx->num_bound_samplers,
+				   ctx->bound_sampler_views);
+}
+
+int
+xa_composite_prepare(struct xa_context *ctx,
+		     const struct xa_composite *comp)
+{
+    struct xa_surface *dst_srf = comp->dst->srf;
+    int ret;
+
+    ret = xa_surface_psurf_create(ctx, dst_srf);
+    if (ret != XA_ERR_NONE)
+	return ret;
+
+    renderer_bind_destination(ctx, dst_srf->srf,
+			      dst_srf->srf->width,
+			      dst_srf->srf->height);
+
+    bind_composite_blend_state(ctx, comp);
+    bind_shaders(ctx, comp);
+    bind_samplers(ctx, comp);
+
+    if (ctx->num_bound_samplers == 0 ) { /* solid fill */
+	renderer_begin_solid(ctx);
+    } else {
+	renderer_begin_textures(ctx);
+	ctx->comp = comp;
+    }
+
+    xa_surface_psurf_destroy(dst_srf);
+    return XA_ERR_NONE;
+}
+
+void xa_composite_rect(struct xa_context *ctx,
+		       int srcX, int srcY, int maskX, int maskY,
+		       int dstX, int dstY, int width, int height)
+{
+    if (ctx->num_bound_samplers == 0 ) { /* solid fill */
+	renderer_solid(ctx, dstX, dstY, dstX + width, dstY + height,
+		       ctx->solid_color);
+    } else {
+	const struct xa_composite *comp = ctx->comp;
+	int pos[6] = {srcX, srcY, maskX, maskY, dstX, dstY};
+	const float *src_matrix = NULL;
+	const float *mask_matrix = NULL;
+
+	if (comp->src->has_transform)
+	    src_matrix = comp->src->transform;
+	if (comp->mask && comp->mask->has_transform)
+	    mask_matrix = comp->mask->transform;
+
+	renderer_texture(ctx, pos, width, height,
+			 src_matrix, mask_matrix);
+    }
+}
+
+void
+xa_composite_done(struct xa_context *ctx)
+{
+    renderer_draw_flush(ctx);
+    ctx->pipe->flush(ctx->pipe, &ctx->last_fence);
+
+    ctx->comp = NULL;
+    ctx->has_solid_color = FALSE;
+    ctx->num_bound_samplers = 0;
+}
+
+static const struct xa_composite_allocation a = {
+    .xa_composite_size = sizeof(struct xa_composite),
+    .xa_picture_size = sizeof(struct xa_picture),
+    .xa_source_pict_size = sizeof(union xa_source_pict),
+};
+
+const struct xa_composite_allocation *
+xa_composite_allocation(void)
+{
+    return &a;
+}
diff --git a/src/gallium/state_trackers/xa/xa_composite.h b/src/gallium/state_trackers/xa/xa_composite.h
new file mode 100644
index 00000000000..d16ef89ebd8
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_composite.h
@@ -0,0 +1,140 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ * Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#ifndef _XA_COMPOSITE_H_
+#define _XA_COMPOSITE_H_
+
+#include "xa_tracker.h"
+#include "xa_context.h"
+
+/*
+ * Supported composite ops.
+ */
+enum xa_composite_op {
+    xa_op_clear,
+    xa_op_src,
+    xa_op_dst,
+    xa_op_over,
+    xa_op_over_reverse,
+    xa_op_in,
+    xa_op_in_reverse,
+    xa_op_out,
+    xa_op_out_reverse,
+    xa_op_atop,
+    xa_op_atop_reverse,
+    xa_op_xor,
+    xa_op_add
+};
+
+/*
+ * Supported filters.
+ */
+enum xa_composite_filter {
+    xa_filter_nearest,
+    xa_filter_linear
+};
+
+/*
+ * Supported clamp methods.
+ */
+enum xa_composite_wrap {
+    xa_wrap_clamp_to_border,
+    xa_wrap_repeat,
+    xa_wrap_mirror_repeat,
+    xa_wrap_clamp_to_edge
+};
+
+/*
+ * Src picture types.
+ */
+enum xa_composite_src_pict_type {
+    xa_src_pict_solid_fill
+};
+
+struct xa_pict_solid_fill {
+    enum xa_composite_src_pict_type type;
+    unsigned int class;
+    uint32_t color;
+};
+
+union xa_source_pict {
+    unsigned int type;
+    struct xa_pict_solid_fill solid_fill;
+};
+
+struct xa_picture {
+    enum xa_formats pict_format;
+    struct xa_surface *srf;
+    struct xa_surface *alpha_map;
+    float transform[9];
+    int has_transform;
+    int component_alpha;
+    enum xa_composite_wrap wrap;
+    enum xa_composite_filter filter;
+    union xa_source_pict *src_pict;
+};
+
+struct xa_composite {
+    struct xa_picture *src, *mask, *dst;
+    int op;
+    int no_solid;
+};
+
+struct xa_composite_allocation {
+    unsigned int xa_composite_size;
+    unsigned int xa_picture_size;
+    unsigned int xa_source_pict_size;
+};
+
+/*
+ * Get allocation sizes for minor bump compatibility.
+ */
+
+extern const struct xa_composite_allocation *
+xa_composite_allocation(void);
+
+/*
+ * This function checks most things except the format of the hardware
+ * surfaces, since they are generally not available at the time this
+ * function is called. Returns usual XA error codes.
+ */
+extern int
+xa_composite_check_accelerated(const struct xa_composite *comp);
+
+extern int
+xa_composite_prepare(struct xa_context *ctx, const struct xa_composite *comp);
+
+extern void
+xa_composite_rect(struct xa_context *ctx,
+		  int srcX, int srcY, int maskX, int maskY,
+		  int dstX, int dstY, int width, int height);
+extern void
+xa_composite_done(struct xa_context *ctx);
+
+#endif
diff --git a/src/gallium/state_trackers/xa/xa_context.c b/src/gallium/state_trackers/xa/xa_context.c
new file mode 100644
index 00000000000..3cc25ed2071
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_context.c
@@ -0,0 +1,386 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ * Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+#include "xa_context.h"
+#include "xa_priv.h"
+#include "cso_cache/cso_context.h"
+#include "util/u_inlines.h"
+#include "util/u_rect.h"
+#include "util/u_surface.h"
+#include "pipe/p_context.h"
+
+
+struct xa_context *
+xa_context_default(struct xa_tracker *xa)
+{
+    return xa->default_ctx;
+}
+
+struct xa_context *
+xa_context_create(struct xa_tracker *xa)
+{
+    struct xa_context *ctx = calloc(1, sizeof(*ctx));
+
+    ctx->xa = xa;
+    ctx->pipe = xa->screen->context_create(xa->screen, NULL);
+    ctx->cso = cso_create_context(ctx->pipe);
+    ctx->shaders = xa_shaders_create(ctx);
+    renderer_init_state(ctx);
+
+    return ctx;
+}
+
+void
+xa_context_destroy(struct xa_context *r)
+{
+    struct pipe_resource **vsbuf = &r->vs_const_buffer;
+    struct pipe_resource **fsbuf = &r->fs_const_buffer;
+
+    if (*vsbuf)
+	pipe_resource_reference(vsbuf, NULL);
+
+    if (*fsbuf)
+	pipe_resource_reference(fsbuf, NULL);
+
+    if (r->shaders) {
+	xa_shaders_destroy(r->shaders);
+	r->shaders = NULL;
+    }
+
+    if (r->cso) {
+	cso_release_all(r->cso);
+	cso_destroy_context(r->cso);
+	r->cso = NULL;
+    }
+}
+
+int
+xa_surface_dma(struct xa_context *ctx,
+	       struct xa_surface *srf,
+	       void *data,
+	       unsigned int pitch,
+	       int to_surface, struct xa_box *boxes, unsigned int num_boxes)
+{
+    struct pipe_transfer *transfer;
+    void *map;
+    int w, h, i;
+    enum pipe_transfer_usage transfer_direction;
+    struct pipe_context *pipe = ctx->pipe;
+
+    transfer_direction = (to_surface ? PIPE_TRANSFER_WRITE :
+			  PIPE_TRANSFER_READ);
+
+    for (i = 0; i < num_boxes; ++i, ++boxes) {
+	w = boxes->x2 - boxes->x1;
+	h = boxes->y2 - boxes->y1;
+
+	transfer = pipe_get_transfer(pipe, srf->tex, 0, 0,
+				     transfer_direction, boxes->x1, boxes->y1,
+				     w, h);
+	if (!transfer)
+	    return -XA_ERR_NORES;
+
+	map = pipe_transfer_map(ctx->pipe, transfer);
+	if (!map)
+	    goto out_no_map;
+
+	if (to_surface) {
+	    util_copy_rect(map, srf->tex->format, transfer->stride,
+			   0, 0, w, h, data, pitch, boxes->x1, boxes->y1);
+	} else {
+	    util_copy_rect(data, srf->tex->format, pitch,
+			   boxes->x1, boxes->y1, w, h, map, transfer->stride, 0,
+			   0);
+	}
+	pipe->transfer_unmap(pipe, transfer);
+	pipe->transfer_destroy(pipe, transfer);
+	if (to_surface)
+	    pipe->flush(pipe, &ctx->last_fence);
+    }
+    return XA_ERR_NONE;
+ out_no_map:
+    pipe->transfer_destroy(pipe, transfer);
+    return -XA_ERR_NORES;
+}
+
+void *
+xa_surface_map(struct xa_context *ctx,
+	       struct xa_surface *srf, unsigned int usage)
+{
+    void *map;
+    unsigned int transfer_direction = 0;
+    struct pipe_context *pipe = ctx->pipe;
+
+    if (srf->transfer)
+	return NULL;
+
+    if (usage & XA_MAP_READ)
+	transfer_direction = PIPE_TRANSFER_READ;
+    if (usage & XA_MAP_WRITE)
+	transfer_direction = PIPE_TRANSFER_WRITE;
+
+    if (!transfer_direction)
+	return NULL;
+
+    srf->transfer = pipe_get_transfer(pipe, srf->tex, 0, 0,
+				      transfer_direction, 0, 0,
+				      srf->tex->width0, srf->tex->height0);
+    if (!srf->transfer)
+	return NULL;
+
+    map = pipe_transfer_map(pipe, srf->transfer);
+    if (!map)
+	pipe->transfer_destroy(pipe, srf->transfer);
+
+    srf->mapping_pipe = pipe;
+    return map;
+}
+
+void
+xa_surface_unmap(struct xa_surface *srf)
+{
+    if (srf->transfer) {
+	struct pipe_context *pipe = srf->mapping_pipe;
+
+	pipe->transfer_unmap(pipe, srf->transfer);
+	pipe->transfer_destroy(pipe, srf->transfer);
+	srf->transfer = NULL;
+    }
+}
+
+int
+xa_surface_psurf_create(struct xa_context *ctx, struct xa_surface *dst)
+{
+    struct pipe_screen *screen = ctx->pipe->screen;
+    struct pipe_surface srf_templ;
+
+    if (dst->srf)
+	return -XA_ERR_INVAL;
+
+    if (!screen->is_format_supported(screen,  dst->tex->format,
+				     PIPE_TEXTURE_2D, 0,
+				     PIPE_BIND_RENDER_TARGET))
+	return -XA_ERR_INVAL;
+
+    u_surface_default_template(&srf_templ, dst->tex,
+			       PIPE_BIND_RENDER_TARGET);
+    dst->srf = ctx->pipe->create_surface(ctx->pipe, dst->tex, &srf_templ);
+    if (!dst->srf)
+	return -XA_ERR_NORES;
+
+    return XA_ERR_NONE;
+}
+
+void
+xa_surface_psurf_destroy(struct xa_surface *dst)
+{
+    pipe_surface_reference(&dst->srf, NULL);
+}
+
+int
+xa_copy_prepare(struct xa_context *ctx,
+		struct xa_surface *dst, struct xa_surface *src)
+{
+    if (src == dst || dst->srf != NULL)
+	return -XA_ERR_INVAL;
+
+    if (src->tex->format != dst->tex->format) {
+	int ret = xa_surface_psurf_create(ctx, dst);
+	if (ret != XA_ERR_NONE)
+	    return ret;
+	renderer_copy_prepare(ctx, dst->srf, src->tex);
+	ctx->simple_copy = 0;
+    } else
+	ctx->simple_copy = 1;
+
+    ctx->src = src;
+    ctx->dst = dst;
+
+    return 0;
+}
+
+void
+xa_copy(struct xa_context *ctx,
+	int dx, int dy, int sx, int sy, int width, int height)
+{
+    struct pipe_box src_box;
+
+    if (ctx->simple_copy) {
+	u_box_2d(sx, sy, width, height, &src_box);
+	ctx->pipe->resource_copy_region(ctx->pipe,
+					ctx->dst->tex, 0, dx, dy, 0,
+					ctx->src->tex,
+					0, &src_box);
+    } else
+	renderer_copy(ctx, dx, dy, sx, sy, width, height,
+		      (float) width, (float) height);
+}
+
+void
+xa_copy_done(struct xa_context *ctx)
+{
+    if (!ctx->simple_copy) {
+	   renderer_draw_flush(ctx);
+	   ctx->pipe->flush(ctx->pipe, &ctx->last_fence);
+	   xa_surface_psurf_destroy(ctx->dst);
+    } else
+	ctx->pipe->flush(ctx->pipe, &ctx->last_fence);
+}
+
+static void
+bind_solid_blend_state(struct xa_context *ctx)
+{
+    struct pipe_blend_state blend;
+
+    memset(&blend, 0, sizeof(struct pipe_blend_state));
+    blend.rt[0].blend_enable = 0;
+    blend.rt[0].colormask = PIPE_MASK_RGBA;
+
+    blend.rt[0].rgb_src_factor   = PIPE_BLENDFACTOR_ONE;
+    blend.rt[0].alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+    blend.rt[0].rgb_dst_factor   = PIPE_BLENDFACTOR_ZERO;
+    blend.rt[0].alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+
+    cso_set_blend(ctx->cso, &blend);
+}
+
+int
+xa_solid_prepare(struct xa_context *ctx, struct xa_surface *dst,
+		 uint32_t fg)
+{
+    unsigned vs_traits, fs_traits;
+    struct xa_shader shader;
+    int width, height;
+    int ret;
+
+    xa_pixel_to_float4(fg, ctx->solid_color);
+    ctx->has_solid_color = 1;
+
+    ret = xa_surface_psurf_create(ctx, dst);
+    if (ret != XA_ERR_NONE)
+	return ret;
+
+    ctx->dst = dst;
+    width = dst->srf->width;
+    height = dst->srf->height;
+
+#if 0
+    debug_printf("Color Pixel=(%d, %d, %d, %d), RGBA=(%f, %f, %f, %f)\n",
+		 (fg >> 24) & 0xff, (fg >> 16) & 0xff,
+		 (fg >> 8) & 0xff,  (fg >> 0) & 0xff,
+		 exa->solid_color[0], exa->solid_color[1],
+		 exa->solid_color[2], exa->solid_color[3]);
+#endif
+
+    vs_traits = VS_SOLID_FILL;
+    fs_traits = FS_SOLID_FILL;
+
+    renderer_bind_destination(ctx, dst->srf, width, height);
+    bind_solid_blend_state(ctx);
+    cso_set_samplers(ctx->cso, 0, NULL);
+    cso_set_fragment_sampler_views(ctx->cso, 0, NULL);
+
+    shader = xa_shaders_get(ctx->shaders, vs_traits, fs_traits);
+    cso_set_vertex_shader_handle(ctx->cso, shader.vs);
+    cso_set_fragment_shader_handle(ctx->cso, shader.fs);
+
+    renderer_begin_solid(ctx);
+
+    xa_surface_psurf_destroy(dst);
+    return XA_ERR_NONE;
+}
+
+void
+xa_solid(struct xa_context *ctx, int x, int y, int width, int height)
+{
+    renderer_solid(ctx, x, y, x + width, y + height, ctx->solid_color);
+}
+
+void
+xa_solid_done(struct xa_context *ctx)
+{
+    renderer_draw_flush(ctx);
+    ctx->pipe->flush(ctx->pipe, &ctx->last_fence);
+
+    ctx->comp = NULL;
+    ctx->has_solid_color = FALSE;
+    ctx->num_bound_samplers = 0;
+}
+
+struct xa_fence *
+xa_fence_get(struct xa_context *ctx)
+{
+    struct xa_fence *fence = malloc(sizeof(*fence));
+    struct pipe_screen *screen = ctx->xa->screen;
+
+    if (!fence)
+	return NULL;
+
+    fence->xa = ctx->xa;
+
+    if (ctx->last_fence == NULL)
+	fence->pipe_fence = NULL;
+    else
+	screen->fence_reference(screen, &fence->pipe_fence, ctx->last_fence);
+
+    return fence;
+}
+
+int
+xa_fence_wait(struct xa_fence *fence, uint64_t timeout)
+{
+    if (!fence)
+	return XA_ERR_NONE;
+
+    if (fence->pipe_fence) {
+	struct pipe_screen *screen = fence->xa->screen;
+	boolean timed_out;
+
+	timed_out = !screen->fence_finish(screen, fence->pipe_fence, timeout);
+	if (timed_out)
+	    return -XA_ERR_BUSY;
+
+	screen->fence_reference(screen, &fence->pipe_fence, NULL);
+    }
+    return XA_ERR_NONE;
+}
+
+void
+xa_fence_destroy(struct xa_fence *fence)
+{
+    if (!fence)
+	return;
+
+    if (fence->pipe_fence) {
+	struct pipe_screen *screen = fence->xa->screen;
+
+	screen->fence_reference(screen, &fence->pipe_fence, NULL);
+    }
+
+    free(fence);
+}
diff --git a/src/gallium/state_trackers/xa/xa_context.h b/src/gallium/state_trackers/xa/xa_context.h
new file mode 100644
index 00000000000..ea2b923a5a1
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_context.h
@@ -0,0 +1,86 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ * Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#ifndef _XA_CONTEXT_H_
+#define _XA_CONTEXT_H_
+#include "xa_tracker.h"
+#include <stdint.h>
+
+struct xa_context;
+
+extern struct xa_context *xa_context_default(struct xa_tracker *xa);
+
+extern struct xa_context *xa_context_create(struct xa_tracker *xa);
+
+extern void xa_context_destroy(struct xa_context *r);
+
+extern int xa_yuv_planar_blit(struct xa_context *r,
+			      int src_x,
+			      int src_y,
+			      int src_w,
+			      int src_h,
+			      int dst_x,
+			      int dst_y,
+			      int dst_w,
+			      int dst_h,
+			      struct xa_box *box,
+			      unsigned int num_boxes,
+			      const float conversion_matrix[],
+			      struct xa_surface *dst, struct xa_surface *yuv[]);
+
+extern int xa_copy_prepare(struct xa_context *ctx,
+			   struct xa_surface *dst, struct xa_surface *src);
+
+extern void xa_copy(struct xa_context *ctx,
+		    int dx, int dy, int sx, int sy, int width, int height);
+
+extern void xa_copy_done(struct xa_context *ctx);
+
+extern int xa_surface_dma(struct xa_context *ctx,
+			  struct xa_surface *srf,
+			  void *data,
+			  unsigned int byte_pitch,
+			  int to_surface, struct xa_box *boxes,
+			  unsigned int num_boxes);
+
+extern int
+xa_solid_prepare(struct xa_context *ctx, struct xa_surface *dst,
+		 uint32_t fg);
+extern void
+xa_solid(struct xa_context *ctx, int x, int y, int width, int height);
+
+extern void
+xa_solid_done(struct xa_context *ctx);
+
+extern struct xa_fence *xa_fence_get(struct xa_context *ctx);
+
+extern int xa_fence_wait(struct xa_fence *fence, uint64_t timeout);
+
+extern void xa_fence_destroy(struct xa_fence *fence);
+#endif
diff --git a/src/gallium/state_trackers/xa/xa_priv.h b/src/gallium/state_trackers/xa/xa_priv.h
new file mode 100644
index 00000000000..94627e1e9d0
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_priv.h
@@ -0,0 +1,241 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ * Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#ifndef _XA_PRIV_H_
+#define _XA_PRIV_H_
+
+#include "xa_tracker.h"
+#include "xa_context.h"
+#include "xa_composite.h"
+
+#include "pipe/p_screen.h"
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+
+#define XA_VB_SIZE (100 * 4 * 3 * 4)
+#define XA_LAST_SURFACE_TYPE (xa_type_yuv_component + 1)
+#define XA_MAX_SAMPLERS 3
+
+struct xa_fence {
+    struct pipe_fence_handle *pipe_fence;
+    struct xa_tracker *xa;
+};
+
+struct xa_format_descriptor {
+    enum pipe_format format;
+    enum xa_formats xa_format;
+};
+
+struct xa_surface {
+    struct pipe_resource template;
+    struct xa_tracker *xa;
+    struct pipe_resource *tex;
+    struct pipe_surface *srf;
+    struct pipe_sampler_view *view;
+    unsigned int flags;
+    struct xa_format_descriptor fdesc;
+    struct pipe_transfer *transfer;
+    struct pipe_context *mapping_pipe;
+};
+
+struct xa_tracker {
+    enum xa_formats *supported_formats;
+    unsigned int format_map[XA_LAST_SURFACE_TYPE][2];
+    int d_depth_bits_last;
+    int ds_depth_bits_last;
+    struct pipe_screen *screen;
+    struct xa_context *default_ctx;
+};
+
+struct xa_context {
+    struct xa_tracker *xa;
+    struct pipe_context *pipe;
+
+    struct cso_context *cso;
+    struct xa_shaders *shaders;
+
+    struct pipe_resource *vs_const_buffer;
+    struct pipe_resource *fs_const_buffer;
+
+    float buffer[XA_VB_SIZE];
+    unsigned int buffer_size;
+    struct pipe_vertex_element velems[3];
+
+    /* number of attributes per vertex for the current
+     * draw operation */
+    unsigned int attrs_per_vertex;
+
+    unsigned int fb_width;
+    unsigned int fb_height;
+
+    struct pipe_fence_handle *last_fence;
+    struct xa_surface *src;
+    struct xa_surface *dst;
+    int simple_copy;
+
+    int has_solid_color;
+    float solid_color[4];
+
+    unsigned int num_bound_samplers;
+    struct pipe_sampler_view *bound_sampler_views[XA_MAX_SAMPLERS];
+    const struct xa_composite *comp;
+};
+
+enum xa_vs_traits {
+    VS_COMPOSITE = 1 << 0,
+    VS_MASK = 1 << 1,
+    VS_SOLID_FILL = 1 << 2,
+    VS_LINGRAD_FILL = 1 << 3,
+    VS_RADGRAD_FILL = 1 << 4,
+    VS_YUV = 1 << 5,
+
+    VS_FILL = (VS_SOLID_FILL | VS_LINGRAD_FILL | VS_RADGRAD_FILL)
+};
+
+enum xa_fs_traits {
+    FS_COMPOSITE = 1 << 0,
+    FS_MASK = 1 << 1,
+    FS_SOLID_FILL = 1 << 2,
+    FS_LINGRAD_FILL = 1 << 3,
+    FS_RADGRAD_FILL = 1 << 4,
+    FS_CA_FULL = 1 << 5,	/* src.rgba * mask.rgba */
+    FS_CA_SRCALPHA = 1 << 6,	/* src.aaaa * mask.rgba */
+    FS_YUV = 1 << 7,
+    FS_SRC_REPEAT_NONE = 1 << 8,
+    FS_MASK_REPEAT_NONE = 1 << 9,
+    FS_SRC_SWIZZLE_RGB = 1 << 10,
+    FS_MASK_SWIZZLE_RGB = 1 << 11,
+    FS_SRC_SET_ALPHA = 1 << 12,
+    FS_MASK_SET_ALPHA = 1 << 13,
+    FS_SRC_LUMINANCE = 1 << 14,
+    FS_MASK_LUMINANCE = 1 << 15,
+
+    FS_FILL = (FS_SOLID_FILL | FS_LINGRAD_FILL | FS_RADGRAD_FILL),
+    FS_COMPONENT_ALPHA = (FS_CA_FULL | FS_CA_SRCALPHA)
+};
+
+struct xa_shader {
+    void *fs;
+    void *vs;
+};
+
+struct xa_shaders;
+
+/*
+ * Inline utilities
+ */
+
+static INLINE int
+xa_min(int a, int b)
+{
+    return ((a <= b) ? a : b);
+}
+
+static INLINE void
+xa_pixel_to_float4(uint32_t pixel, float *color)
+{
+    uint32_t	    r, g, b, a;
+
+    a = (pixel >> 24) & 0xff;
+    r = (pixel >> 16) & 0xff;
+    g = (pixel >>  8) & 0xff;
+    b = (pixel >>  0) & 0xff;
+    color[0] = ((float)r) / 255.;
+    color[1] = ((float)g) / 255.;
+    color[2] = ((float)b) / 255.;
+    color[3] = ((float)a) / 255.;
+}
+
+
+/*
+ * xa_tgsi.c
+ */
+
+extern struct xa_shaders *xa_shaders_create(struct xa_context *);
+
+void xa_shaders_destroy(struct xa_shaders *shaders);
+
+struct xa_shader xa_shaders_get(struct xa_shaders *shaders,
+				unsigned vs_traits, unsigned fs_traits);
+
+/*
+ * xa_context.c
+ */
+extern int
+xa_surface_psurf_create(struct xa_context *ctx, struct xa_surface *dst);
+
+extern void
+xa_surface_psurf_destroy(struct xa_surface *dst);
+
+/*
+ * xa_renderer.c
+ */
+void renderer_set_constants(struct xa_context *r,
+			    int shader_type, const float *params,
+			    int param_bytes);
+
+void renderer_draw_yuv(struct xa_context *r,
+		       float src_x,
+		       float src_y,
+		       float src_w,
+		       float src_h,
+		       int dst_x,
+		       int dst_y, int dst_w, int dst_h,
+		       struct xa_surface *srf[]);
+
+void renderer_bind_destination(struct xa_context *r,
+			       struct pipe_surface *surface, int width,
+			       int height);
+
+void renderer_init_state(struct xa_context *r);
+void renderer_copy_prepare(struct xa_context *r,
+			   struct pipe_surface *dst_surface,
+			   struct pipe_resource *src_texture);
+void renderer_copy(struct xa_context *r, int dx,
+		   int dy,
+		   int sx,
+		   int sy,
+		   int width, int height, float src_width, float src_height);
+
+void renderer_draw_flush(struct xa_context *r);
+
+void renderer_begin_solid(struct xa_context *r);
+void renderer_solid(struct xa_context *r,
+		    int x0, int y0, int x1, int y1, float *color);
+void
+renderer_begin_textures(struct xa_context *r);
+
+void
+renderer_texture(struct xa_context *r,
+		 int *pos,
+		 int width, int height,
+		 const float *src_matrix,
+		 const float *mask_matrix);
+
+#endif
diff --git a/src/gallium/state_trackers/xa/xa_renderer.c b/src/gallium/state_trackers/xa/xa_renderer.c
new file mode 100644
index 00000000000..559b2699da6
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_renderer.c
@@ -0,0 +1,626 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ */
+
+#include "xa_context.h"
+#include "xa_priv.h"
+#include <math.h>
+#include "cso_cache/cso_context.h"
+#include "util/u_inlines.h"
+#include "util/u_sampler.h"
+#include "util/u_draw_quad.h"
+
+#define floatsEqual(x, y) (fabs(x - y) <= 0.00001f * MIN2(fabs(x), fabs(y)))
+#define floatIsZero(x) (floatsEqual((x) + 1, 1))
+
+#define NUM_COMPONENTS 4
+
+void
+
+
+renderer_set_constants(struct xa_context *r,
+		       int shader_type, const float *params, int param_bytes);
+
+static INLINE boolean
+is_affine(float *matrix)
+{
+    return floatIsZero(matrix[2]) && floatIsZero(matrix[5])
+	&& floatsEqual(matrix[8], 1);
+}
+
+static INLINE void
+map_point(float *mat, float x, float y, float *out_x, float *out_y)
+{
+    if (!mat) {
+	*out_x = x;
+	*out_y = y;
+	return;
+    }
+
+    *out_x = mat[0] * x + mat[3] * y + mat[6];
+    *out_y = mat[1] * x + mat[4] * y + mat[7];
+    if (!is_affine(mat)) {
+	float w = 1 / (mat[2] * x + mat[5] * y + mat[8]);
+
+	*out_x *= w;
+	*out_y *= w;
+    }
+}
+
+static INLINE struct pipe_resource *
+renderer_buffer_create(struct xa_context *r)
+{
+    struct pipe_resource *buf = pipe_user_buffer_create(r->pipe->screen,
+							r->buffer,
+							sizeof(float) *
+							r->buffer_size,
+							PIPE_BIND_VERTEX_BUFFER);
+
+    r->buffer_size = 0;
+
+    return buf;
+}
+
+static INLINE void
+renderer_draw(struct xa_context *r)
+{
+    struct pipe_context *pipe = r->pipe;
+    struct pipe_resource *buf = 0;
+    int num_verts = r->buffer_size / (r->attrs_per_vertex * NUM_COMPONENTS);
+
+    if (!r->buffer_size)
+	return;
+
+    buf = renderer_buffer_create(r);
+
+    if (buf) {
+	cso_set_vertex_elements(r->cso, r->attrs_per_vertex, r->velems);
+
+	util_draw_vertex_buffer(pipe, r->cso, buf, 0, PIPE_PRIM_QUADS, num_verts,	/* verts */
+				r->attrs_per_vertex);	/* attribs/vert */
+
+	pipe_resource_reference(&buf, NULL);
+    }
+}
+
+static INLINE void
+renderer_draw_conditional(struct xa_context *r, int next_batch)
+{
+    if (r->buffer_size + next_batch >= XA_VB_SIZE ||
+	(next_batch == 0 && r->buffer_size)) {
+	renderer_draw(r);
+    }
+}
+
+void
+renderer_init_state(struct xa_context *r)
+{
+    struct pipe_depth_stencil_alpha_state dsa;
+    struct pipe_rasterizer_state raster;
+    unsigned i;
+
+    /* set common initial clip state */
+    memset(&dsa, 0, sizeof(struct pipe_depth_stencil_alpha_state));
+    cso_set_depth_stencil_alpha(r->cso, &dsa);
+
+    /* 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);
+
+    /* vertex elements state */
+    memset(&r->velems[0], 0, sizeof(r->velems[0]) * 3);
+    for (i = 0; i < 3; i++) {
+	r->velems[i].src_offset = i * 4 * sizeof(float);
+	r->velems[i].instance_divisor = 0;
+	r->velems[i].vertex_buffer_index = 0;
+	r->velems[i].src_format = PIPE_FORMAT_R32G32B32A32_FLOAT;
+    }
+}
+
+static INLINE void
+add_vertex_color(struct xa_context *r, float x, float y, float color[4])
+{
+    float *vertex = r->buffer + r->buffer_size;
+
+    vertex[0] = x;
+    vertex[1] = y;
+    vertex[2] = 0.f;		/*z */
+    vertex[3] = 1.f;		/*w */
+
+    vertex[4] = color[0];	/*r */
+    vertex[5] = color[1];	/*g */
+    vertex[6] = color[2];	/*b */
+    vertex[7] = color[3];	/*a */
+
+    r->buffer_size += 8;
+}
+
+static INLINE void
+add_vertex_1tex(struct xa_context *r, float x, float y, float s, float t)
+{
+    float *vertex = r->buffer + r->buffer_size;
+
+    vertex[0] = x;
+    vertex[1] = y;
+    vertex[2] = 0.f;		/*z */
+    vertex[3] = 1.f;		/*w */
+
+    vertex[4] = s;		/*s */
+    vertex[5] = t;		/*t */
+    vertex[6] = 0.f;		/*r */
+    vertex[7] = 1.f;		/*q */
+
+    r->buffer_size += 8;
+}
+
+static INLINE void
+add_vertex_2tex(struct xa_context *r,
+		float x, float y, float s0, float t0, float s1, float t1)
+{
+    float *vertex = r->buffer + r->buffer_size;
+
+    vertex[0] = x;
+    vertex[1] = y;
+    vertex[2] = 0.f;		/*z */
+    vertex[3] = 1.f;		/*w */
+
+    vertex[4] = s0;		/*s */
+    vertex[5] = t0;		/*t */
+    vertex[6] = 0.f;		/*r */
+    vertex[7] = 1.f;		/*q */
+
+    vertex[8] = s1;		/*s */
+    vertex[9] = t1;		/*t */
+    vertex[10] = 0.f;		/*r */
+    vertex[11] = 1.f;		/*q */
+
+    r->buffer_size += 12;
+}
+
+static void
+add_vertex_data1(struct xa_context *r,
+                 float srcX, float srcY,  float dstX, float dstY,
+                 float width, float height,
+                 struct pipe_resource *src, const float *src_matrix)
+{
+    float s0, t0, s1, t1, s2, t2, s3, t3;
+    float pt0[2], pt1[2], pt2[2], pt3[2];
+
+    pt0[0] = srcX;
+    pt0[1] = srcY;
+    pt1[0] = (srcX + width);
+    pt1[1] = srcY;
+    pt2[0] = (srcX + width);
+    pt2[1] = (srcY + height);
+    pt3[0] = srcX;
+    pt3[1] = (srcY + height);
+
+    if (src_matrix) {
+	map_point((float *)src_matrix, pt0[0], pt0[1], &pt0[0], &pt0[1]);
+	map_point((float *)src_matrix, pt1[0], pt1[1], &pt1[0], &pt1[1]);
+	map_point((float *)src_matrix, pt2[0], pt2[1], &pt2[0], &pt2[1]);
+	map_point((float *)src_matrix, pt3[0], pt3[1], &pt3[0], &pt3[1]);
+    }
+
+    s0 =  pt0[0] / src->width0;
+    s1 =  pt1[0] / src->width0;
+    s2 =  pt2[0] / src->width0;
+    s3 =  pt3[0] / src->width0;
+    t0 =  pt0[1] / src->height0;
+    t1 =  pt1[1] / src->height0;
+    t2 =  pt2[1] / src->height0;
+    t3 =  pt3[1] / src->height0;
+
+    /* 1st vertex */
+    add_vertex_1tex(r, dstX, dstY, s0, t0);
+    /* 2nd vertex */
+    add_vertex_1tex(r, dstX + width, dstY, s1, t1);
+    /* 3rd vertex */
+    add_vertex_1tex(r, dstX + width, dstY + height, s2, t2);
+    /* 4th vertex */
+    add_vertex_1tex(r, dstX, dstY + height, s3, t3);
+}
+
+static void
+add_vertex_data2(struct xa_context *r,
+                 float srcX, float srcY, float maskX, float maskY,
+                 float dstX, float dstY, float width, float height,
+                 struct pipe_resource *src,
+                 struct pipe_resource *mask,
+                 const float *src_matrix, const float *mask_matrix)
+{
+    float src_s0, src_t0, src_s1, src_t1;
+    float mask_s0, mask_t0, mask_s1, mask_t1;
+    float spt0[2], spt1[2];
+    float mpt0[2], mpt1[2];
+
+    spt0[0] = srcX;
+    spt0[1] = srcY;
+    spt1[0] = srcX + width;
+    spt1[1] = srcY + height;
+
+    mpt0[0] = maskX;
+    mpt0[1] = maskY;
+    mpt1[0] = maskX + width;
+    mpt1[1] = maskY + height;
+
+    if (src_matrix) {
+	map_point((float *)src_matrix, spt0[0], spt0[1], &spt0[0], &spt0[1]);
+	map_point((float *)src_matrix, spt1[0], spt1[1], &spt1[0], &spt1[1]);
+    }
+
+    if (mask_matrix) {
+	map_point((float *)mask_matrix, mpt0[0], mpt0[1], &mpt0[0], &mpt0[1]);
+	map_point((float *)mask_matrix, mpt1[0], mpt1[1], &mpt1[0], &mpt1[1]);
+    }
+
+    src_s0 = spt0[0] / src->width0;
+    src_t0 = spt0[1] / src->height0;
+    src_s1 = spt1[0] / src->width0;
+    src_t1 = spt1[1] / src->height0;
+
+    mask_s0 = mpt0[0] / mask->width0;
+    mask_t0 = mpt0[1] / mask->height0;
+    mask_s1 = mpt1[0] / mask->width0;
+    mask_t1 = mpt1[1] / mask->height0;
+
+    /* 1st vertex */
+    add_vertex_2tex(r, dstX, dstY,
+		    src_s0, src_t0, mask_s0, mask_t0);
+    /* 2nd vertex */
+    add_vertex_2tex(r, dstX + width, dstY,
+		    src_s1, src_t0, mask_s1, mask_t0);
+    /* 3rd vertex */
+    add_vertex_2tex(r, dstX + width, dstY + height,
+		    src_s1, src_t1, mask_s1, mask_t1);
+    /* 4th vertex */
+    add_vertex_2tex(r, dstX, dstY + height,
+		    src_s0, src_t1, mask_s0, mask_t1);
+}
+
+static struct pipe_resource *
+setup_vertex_data_yuv(struct xa_context *r,
+		      float srcX,
+		      float srcY,
+		      float srcW,
+		      float srcH,
+		      float dstX,
+		      float dstY,
+		      float dstW, float dstH, struct xa_surface *srf[])
+{
+    float s0, t0, s1, t1;
+    float spt0[2], spt1[2];
+    struct pipe_resource *tex;
+
+    spt0[0] = srcX;
+    spt0[1] = srcY;
+    spt1[0] = srcX + srcW;
+    spt1[1] = srcY + srcH;
+
+    tex = srf[0]->tex;
+    s0 = spt0[0] / tex->width0;
+    t0 = spt0[1] / tex->height0;
+    s1 = spt1[0] / tex->width0;
+    t1 = spt1[1] / tex->height0;
+
+    /* 1st vertex */
+    add_vertex_1tex(r, dstX, dstY, s0, t0);
+    /* 2nd vertex */
+    add_vertex_1tex(r, dstX + dstW, dstY, s1, t0);
+    /* 3rd vertex */
+    add_vertex_1tex(r, dstX + dstW, dstY + dstH, s1, t1);
+    /* 4th vertex */
+    add_vertex_1tex(r, dstX, dstY + dstH, s0, t1);
+
+    return renderer_buffer_create(r);
+}
+
+/* Set up framebuffer, viewport and vertex shader constant buffer
+ * state for a particular destinaton surface.  In all our rendering,
+ * these concepts are linked.
+ */
+void
+renderer_bind_destination(struct xa_context *r,
+			  struct pipe_surface *surface, int width, int height)
+{
+
+    struct pipe_framebuffer_state fb;
+    struct pipe_viewport_state viewport;
+
+    /* Framebuffer uses actual surface width/height
+     */
+    memset(&fb, 0, sizeof fb);
+    fb.width = surface->width;
+    fb.height = surface->height;
+    fb.nr_cbufs = 1;
+    fb.cbufs[0] = surface;
+    fb.zsbuf = 0;
+
+    /* Viewport just touches the bit we're interested in:
+     */
+    viewport.scale[0] = width / 2.f;
+    viewport.scale[1] = height / 2.f;
+    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;
+
+    /* Constant buffer set up to match viewport dimensions:
+     */
+    if (r->fb_width != width || r->fb_height != height) {
+	float vs_consts[8] = {
+	    2.f / width, 2.f / height, 1, 1,
+	    -1, -1, 0, 0
+	};
+
+	r->fb_width = width;
+	r->fb_height = height;
+
+	renderer_set_constants(r, PIPE_SHADER_VERTEX,
+			       vs_consts, sizeof vs_consts);
+    }
+
+    cso_set_framebuffer(r->cso, &fb);
+    cso_set_viewport(r->cso, &viewport);
+}
+
+void
+renderer_set_constants(struct xa_context *r,
+		       int shader_type, const float *params, int param_bytes)
+{
+    struct pipe_resource **cbuf =
+	(shader_type == PIPE_SHADER_VERTEX) ? &r->vs_const_buffer :
+	&r->fs_const_buffer;
+
+    pipe_resource_reference(cbuf, NULL);
+    *cbuf = pipe_buffer_create(r->pipe->screen,
+			       PIPE_BIND_CONSTANT_BUFFER, PIPE_USAGE_STATIC,
+			       param_bytes);
+
+    if (*cbuf) {
+	pipe_buffer_write(r->pipe, *cbuf, 0, param_bytes, params);
+    }
+    r->pipe->set_constant_buffer(r->pipe, shader_type, 0, *cbuf);
+}
+
+void
+renderer_copy_prepare(struct xa_context *r,
+		      struct pipe_surface *dst_surface,
+		      struct pipe_resource *src_texture)
+{
+    struct pipe_context *pipe = r->pipe;
+    struct pipe_screen *screen = pipe->screen;
+    struct xa_shader shader;
+
+    assert(screen->is_format_supported(screen, dst_surface->format,
+				       PIPE_TEXTURE_2D, 0,
+				       PIPE_BIND_RENDER_TARGET));
+    (void)screen;
+
+    /* set misc state we care about */
+    {
+	struct pipe_blend_state blend;
+
+	memset(&blend, 0, sizeof(blend));
+	blend.rt[0].rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+	blend.rt[0].alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+	blend.rt[0].rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+	blend.rt[0].alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+	blend.rt[0].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);
+    }
+
+    renderer_bind_destination(r, dst_surface,
+			      dst_surface->width, dst_surface->height);
+
+    /* texture/sampler view */
+    {
+	struct pipe_sampler_view templ;
+	struct pipe_sampler_view *src_view;
+
+	u_sampler_view_default_template(&templ,
+					src_texture, src_texture->format);
+	src_view = pipe->create_sampler_view(pipe, src_texture, &templ);
+	cso_set_fragment_sampler_views(r->cso, 1, &src_view);
+	pipe_sampler_view_reference(&src_view, NULL);
+    }
+
+    /* shaders */
+    shader = xa_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);
+
+    r->buffer_size = 0;
+    r->attrs_per_vertex = 2;
+}
+
+void
+renderer_copy(struct xa_context *r,
+	      int dx,
+	      int dy,
+	      int sx,
+	      int sy,
+	      int width, int height, float src_width, float src_height)
+{
+    float s0, t0, s1, t1;
+    float x0, y0, x1, y1;
+
+    /* XXX: could put the texcoord scaling calculation into the vertex
+     * shader.
+     */
+    s0 = sx / src_width;
+    s1 = (sx + width) / src_width;
+    t0 = sy / src_height;
+    t1 = (sy + height) / src_height;
+
+    x0 = dx;
+    x1 = dx + width;
+    y0 = dy;
+    y1 = dy + height;
+
+    /* draw quad */
+    renderer_draw_conditional(r, 4 * 8);
+    add_vertex_1tex(r, x0, y0, s0, t0);
+    add_vertex_1tex(r, x1, y0, s1, t0);
+    add_vertex_1tex(r, x1, y1, s1, t1);
+    add_vertex_1tex(r, x0, y1, s0, t1);
+}
+
+void
+renderer_draw_yuv(struct xa_context *r,
+		  float src_x,
+		  float src_y,
+		  float src_w,
+		  float src_h,
+		  int dst_x,
+		  int dst_y, int dst_w, int dst_h, struct xa_surface *srf[])
+{
+    struct pipe_context *pipe = r->pipe;
+    struct pipe_resource *buf = 0;
+
+    buf = setup_vertex_data_yuv(r,
+				src_x, src_y, src_w, src_h, dst_x, dst_y, dst_w,
+				dst_h, srf);
+
+    if (buf) {
+	const int num_attribs = 2;	/*pos + tex coord */
+
+	cso_set_vertex_elements(r->cso, num_attribs, r->velems);
+
+	util_draw_vertex_buffer(pipe, r->cso, buf, 0, PIPE_PRIM_QUADS, 4,	/* verts */
+				num_attribs);	/* attribs/vert */
+
+	pipe_resource_reference(&buf, NULL);
+    }
+}
+
+void
+renderer_begin_solid(struct xa_context *r)
+{
+    r->buffer_size = 0;
+    r->attrs_per_vertex = 2;
+}
+
+void
+renderer_solid(struct xa_context *r,
+	       int x0, int y0, int x1, int y1, float *color)
+{
+    /*
+     * debug_printf("solid rect[(%d, %d), (%d, %d)], rgba[%f, %f, %f, %f]\n",
+     * x0, y0, x1, y1, color[0], color[1], color[2], color[3]); */
+
+    renderer_draw_conditional(r, 4 * 8);
+
+    /* 1st vertex */
+    add_vertex_color(r, x0, y0, color);
+    /* 2nd vertex */
+    add_vertex_color(r, x1, y0, color);
+    /* 3rd vertex */
+    add_vertex_color(r, x1, y1, color);
+    /* 4th vertex */
+    add_vertex_color(r, x0, y1, color);
+}
+
+void
+renderer_draw_flush(struct xa_context *r)
+{
+    renderer_draw_conditional(r, 0);
+}
+
+void
+renderer_begin_textures(struct xa_context *r)
+{
+    r->attrs_per_vertex = 1 + r->num_bound_samplers;
+    r->buffer_size = 0;
+}
+
+void
+renderer_texture(struct xa_context *r,
+		 int *pos,
+		 int width, int height,
+		 const float *src_matrix,
+		 const float *mask_matrix)
+{
+    struct pipe_sampler_view **sampler_view = r->bound_sampler_views;
+
+#if 0
+    if (src_matrix) {
+	debug_printf("src_matrix = \n");
+	debug_printf("%f, %f, %f\n", src_matrix[0], src_matrix[1], src_matrix[2]);
+	debug_printf("%f, %f, %f\n", src_matrix[3], src_matrix[4], src_matrix[5]);
+	debug_printf("%f, %f, %f\n", src_matrix[6], src_matrix[7], src_matrix[8]);
+    }
+    if (mask_matrix) {
+	debug_printf("mask_matrix = \n");
+	debug_printf("%f, %f, %f\n", mask_matrix[0], mask_matrix[1], mask_matrix[2]);
+	debug_printf("%f, %f, %f\n", mask_matrix[3], mask_matrix[4], mask_matrix[5]);
+	debug_printf("%f, %f, %f\n", mask_matrix[6], mask_matrix[7], mask_matrix[8]);
+    }
+#endif
+
+    switch(r->attrs_per_vertex) {
+    case 2:
+	renderer_draw_conditional(r, 4 * 8);
+	add_vertex_data1(r,
+			 pos[0], pos[1], /* src */
+			 pos[4], pos[5], /* dst */
+			 width, height,
+			 sampler_view[0]->texture, src_matrix);
+	break;
+    case 3:
+	renderer_draw_conditional(r, 4 * 12);
+	add_vertex_data2(r,
+			 pos[0], pos[1], /* src */
+			 pos[2], pos[3], /* mask */
+			 pos[4], pos[5], /* dst */
+			 width, height,
+			 sampler_view[0]->texture, sampler_view[1]->texture,
+			 src_matrix, mask_matrix);
+	break;
+    default:
+	break;
+    }
+}
diff --git a/src/gallium/state_trackers/xa/xa_symbols b/src/gallium/state_trackers/xa/xa_symbols
new file mode 100644
index 00000000000..6da701f9702
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_symbols
@@ -0,0 +1,30 @@
+xa_tracker_version
+xa_tracker_create
+xa_tracker_destroy
+xa_surface_create
+xa_surface_destroy
+xa_surface_redefine
+xa_surface_dma
+xa_surface_map
+xa_surface_unmap
+xa_surface_format
+xa_surface_handle
+xa_format_check_supported
+xa_context_default
+xa_context_create
+xa_context_destroy
+xa_fence_get
+xa_fence_wait
+xa_fence_destroy
+xa_copy_prepare
+xa_copy
+xa_copy_done
+xa_solid_prepare
+xa_solid
+xa_solid_done
+xa_composite_allocation
+xa_composite_check_accelerated
+xa_composite_prepare
+xa_composite_rect
+xa_composite_done
+xa_yuv_planar_blit
diff --git a/src/gallium/state_trackers/xa/xa_tgsi.c b/src/gallium/state_trackers/xa/xa_tgsi.c
new file mode 100644
index 00000000000..fb6ffefd636
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_tgsi.c
@@ -0,0 +1,651 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ */
+#include "xa_priv.h"
+
+#include "pipe/p_format.h"
+#include "pipe/p_context.h"
+#include "pipe/p_state.h"
+#include "pipe/p_shader_tokens.h"
+
+#include "util/u_memory.h"
+
+#include "tgsi/tgsi_ureg.h"
+
+#include "cso_cache/cso_context.h"
+#include "cso_cache/cso_hash.h"
+
+/* Vertex shader:
+ * IN[0]    = vertex pos
+ * IN[1]    = src tex coord | solid fill color
+ * IN[2]    = mask tex coord
+ * IN[3]    = dst tex coord
+ * CONST[0] = (2/dst_width, 2/dst_height, 1, 1)
+ * CONST[1] = (-1, -1, 0, 0)
+ *
+ * OUT[0]   = vertex pos
+ * OUT[1]   = src tex coord | solid fill color
+ * OUT[2]   = mask tex coord
+ * OUT[3]   = dst tex coord
+ */
+
+/* Fragment shader:
+ * SAMP[0]  = src
+ * SAMP[1]  = mask
+ * SAMP[2]  = dst
+ * IN[0]    = pos src | solid fill color
+ * IN[1]    = pos mask
+ * IN[2]    = pos dst
+ * CONST[0] = (0, 0, 0, 1)
+ *
+ * OUT[0] = color
+ */
+
+static void
+print_fs_traits(int fs_traits)
+{
+    const char *strings[] = {
+	"FS_COMPOSITE",		/* = 1 << 0, */
+	"FS_MASK",		/* = 1 << 1, */
+	"FS_SOLID_FILL",	/* = 1 << 2, */
+	"FS_LINGRAD_FILL",	/* = 1 << 3, */
+	"FS_RADGRAD_FILL",	/* = 1 << 4, */
+	"FS_CA_FULL",		/* = 1 << 5, *//* src.rgba * mask.rgba */
+	"FS_CA_SRCALPHA",	/* = 1 << 6, *//* src.aaaa * mask.rgba */
+	"FS_YUV",		/* = 1 << 7, */
+	"FS_SRC_REPEAT_NONE",	/* = 1 << 8, */
+	"FS_MASK_REPEAT_NONE",	/* = 1 << 9, */
+	"FS_SRC_SWIZZLE_RGB",	/* = 1 << 10, */
+	"FS_MASK_SWIZZLE_RGB",	/* = 1 << 11, */
+	"FS_SRC_SET_ALPHA",	/* = 1 << 12, */
+	"FS_MASK_SET_ALPHA",	/* = 1 << 13, */
+	"FS_SRC_LUMINANCE",	/* = 1 << 14, */
+	"FS_MASK_LUMINANCE",	/* = 1 << 15, */
+    };
+    int i, k;
+
+    debug_printf("%s: ", __func__);
+
+    for (i = 0, k = 1; k < (1 << 16); i++, k <<= 1) {
+	if (fs_traits & k)
+	    debug_printf("%s, ", strings[i]);
+    }
+
+    debug_printf("\n");
+}
+
+struct xa_shaders {
+    struct xa_context *r;
+
+    struct cso_hash *vs_hash;
+    struct cso_hash *fs_hash;
+};
+
+static INLINE void
+src_in_mask(struct ureg_program *ureg,
+	    struct ureg_dst dst,
+	    struct ureg_src src,
+	    struct ureg_src mask,
+	    unsigned component_alpha, unsigned mask_luminance)
+{
+    if (component_alpha == FS_CA_FULL) {
+	ureg_MUL(ureg, dst, src, mask);
+    } else if (component_alpha == FS_CA_SRCALPHA) {
+	ureg_MUL(ureg, dst, ureg_scalar(src, TGSI_SWIZZLE_W), mask);
+    } else {
+	if (mask_luminance)
+	    ureg_MUL(ureg, dst, src, ureg_scalar(mask, TGSI_SWIZZLE_X));
+	else
+	    ureg_MUL(ureg, dst, src, ureg_scalar(mask, TGSI_SWIZZLE_W));
+    }
+}
+
+static struct ureg_src
+vs_normalize_coords(struct ureg_program *ureg,
+		    struct ureg_src coords,
+		    struct ureg_src const0, struct ureg_src const1)
+{
+    struct ureg_dst tmp = ureg_DECL_temporary(ureg);
+    struct ureg_src ret;
+
+    ureg_MAD(ureg, tmp, coords, const0, const1);
+    ret = ureg_src(tmp);
+    ureg_release_temporary(ureg, tmp);
+    return ret;
+}
+
+static void
+linear_gradient(struct ureg_program *ureg,
+		struct ureg_dst out,
+		struct ureg_src pos,
+		struct ureg_src sampler,
+		struct ureg_src coords,
+		struct ureg_src const0124,
+		struct ureg_src matrow0,
+		struct ureg_src matrow1, struct ureg_src matrow2)
+{
+    struct ureg_dst temp0 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp1 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp2 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp3 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp4 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp5 = ureg_DECL_temporary(ureg);
+
+    ureg_MOV(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_XY), pos);
+    ureg_MOV(ureg,
+	     ureg_writemask(temp0, TGSI_WRITEMASK_Z),
+	     ureg_scalar(const0124, TGSI_SWIZZLE_Y));
+
+    ureg_DP3(ureg, temp1, matrow0, ureg_src(temp0));
+    ureg_DP3(ureg, temp2, matrow1, ureg_src(temp0));
+    ureg_DP3(ureg, temp3, matrow2, ureg_src(temp0));
+    ureg_RCP(ureg, temp3, ureg_src(temp3));
+    ureg_MUL(ureg, temp1, ureg_src(temp1), ureg_src(temp3));
+    ureg_MUL(ureg, temp2, ureg_src(temp2), ureg_src(temp3));
+
+    ureg_MOV(ureg, ureg_writemask(temp4, TGSI_WRITEMASK_X), ureg_src(temp1));
+    ureg_MOV(ureg, ureg_writemask(temp4, TGSI_WRITEMASK_Y), ureg_src(temp2));
+
+    ureg_MUL(ureg, temp0,
+	     ureg_scalar(coords, TGSI_SWIZZLE_Y),
+	     ureg_scalar(ureg_src(temp4), TGSI_SWIZZLE_Y));
+    ureg_MAD(ureg, temp1,
+	     ureg_scalar(coords, TGSI_SWIZZLE_X),
+	     ureg_scalar(ureg_src(temp4), TGSI_SWIZZLE_X), ureg_src(temp0));
+
+    ureg_MUL(ureg, temp2, ureg_src(temp1), ureg_scalar(coords, TGSI_SWIZZLE_Z));
+
+    ureg_TEX(ureg, out, TGSI_TEXTURE_1D, ureg_src(temp2), sampler);
+
+    ureg_release_temporary(ureg, temp0);
+    ureg_release_temporary(ureg, temp1);
+    ureg_release_temporary(ureg, temp2);
+    ureg_release_temporary(ureg, temp3);
+    ureg_release_temporary(ureg, temp4);
+    ureg_release_temporary(ureg, temp5);
+}
+
+static void
+radial_gradient(struct ureg_program *ureg,
+		struct ureg_dst out,
+		struct ureg_src pos,
+		struct ureg_src sampler,
+		struct ureg_src coords,
+		struct ureg_src const0124,
+		struct ureg_src matrow0,
+		struct ureg_src matrow1, struct ureg_src matrow2)
+{
+    struct ureg_dst temp0 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp1 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp2 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp3 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp4 = ureg_DECL_temporary(ureg);
+    struct ureg_dst temp5 = ureg_DECL_temporary(ureg);
+
+    ureg_MOV(ureg, ureg_writemask(temp0, TGSI_WRITEMASK_XY), pos);
+    ureg_MOV(ureg,
+	     ureg_writemask(temp0, TGSI_WRITEMASK_Z),
+	     ureg_scalar(const0124, TGSI_SWIZZLE_Y));
+
+    ureg_DP3(ureg, temp1, matrow0, ureg_src(temp0));
+    ureg_DP3(ureg, temp2, matrow1, ureg_src(temp0));
+    ureg_DP3(ureg, temp3, matrow2, ureg_src(temp0));
+    ureg_RCP(ureg, temp3, ureg_src(temp3));
+    ureg_MUL(ureg, temp1, ureg_src(temp1), ureg_src(temp3));
+    ureg_MUL(ureg, temp2, ureg_src(temp2), ureg_src(temp3));
+
+    ureg_MOV(ureg, ureg_writemask(temp5, TGSI_WRITEMASK_X), ureg_src(temp1));
+    ureg_MOV(ureg, ureg_writemask(temp5, TGSI_WRITEMASK_Y), ureg_src(temp2));
+
+    ureg_MUL(ureg, temp0, ureg_scalar(coords, TGSI_SWIZZLE_Y),
+	     ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_Y));
+    ureg_MAD(ureg, temp1,
+	     ureg_scalar(coords, TGSI_SWIZZLE_X),
+	     ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_X), ureg_src(temp0));
+    ureg_ADD(ureg, temp1, ureg_src(temp1), ureg_src(temp1));
+    ureg_MUL(ureg, temp3,
+	     ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_Y),
+	     ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_Y));
+    ureg_MAD(ureg, temp4,
+	     ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_X),
+	     ureg_scalar(ureg_src(temp5), TGSI_SWIZZLE_X), ureg_src(temp3));
+    ureg_MOV(ureg, temp4, ureg_negate(ureg_src(temp4)));
+    ureg_MUL(ureg, temp2, ureg_scalar(coords, TGSI_SWIZZLE_Z), ureg_src(temp4));
+    ureg_MUL(ureg, temp0,
+	     ureg_scalar(const0124, TGSI_SWIZZLE_W), ureg_src(temp2));
+    ureg_MUL(ureg, temp3, ureg_src(temp1), ureg_src(temp1));
+    ureg_SUB(ureg, temp2, ureg_src(temp3), ureg_src(temp0));
+    ureg_RSQ(ureg, temp2, ureg_abs(ureg_src(temp2)));
+    ureg_RCP(ureg, temp2, ureg_src(temp2));
+    ureg_SUB(ureg, temp1, ureg_src(temp2), ureg_src(temp1));
+    ureg_ADD(ureg, temp0,
+	     ureg_scalar(coords, TGSI_SWIZZLE_Z),
+	     ureg_scalar(coords, TGSI_SWIZZLE_Z));
+    ureg_RCP(ureg, temp0, ureg_src(temp0));
+    ureg_MUL(ureg, temp2, ureg_src(temp1), ureg_src(temp0));
+    ureg_TEX(ureg, out, TGSI_TEXTURE_1D, ureg_src(temp2), sampler);
+
+    ureg_release_temporary(ureg, temp0);
+    ureg_release_temporary(ureg, temp1);
+    ureg_release_temporary(ureg, temp2);
+    ureg_release_temporary(ureg, temp3);
+    ureg_release_temporary(ureg, temp4);
+    ureg_release_temporary(ureg, temp5);
+}
+
+static void *
+create_vs(struct pipe_context *pipe, unsigned vs_traits)
+{
+    struct ureg_program *ureg;
+    struct ureg_src src;
+    struct ureg_dst dst;
+    struct ureg_src const0, const1;
+    boolean is_fill = (vs_traits & VS_FILL) != 0;
+    boolean is_composite = (vs_traits & VS_COMPOSITE) != 0;
+    boolean has_mask = (vs_traits & VS_MASK) != 0;
+    boolean is_yuv = (vs_traits & VS_YUV) != 0;
+    unsigned input_slot = 0;
+
+    ureg = ureg_create(TGSI_PROCESSOR_VERTEX);
+    if (ureg == NULL)
+	return 0;
+
+    const0 = ureg_DECL_constant(ureg, 0);
+    const1 = ureg_DECL_constant(ureg, 1);
+
+    /* it has to be either a fill or a composite op */
+    debug_assert((is_fill ^ is_composite) ^ is_yuv);
+
+    src = ureg_DECL_vs_input(ureg, input_slot++);
+    dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_POSITION, 0);
+    src = vs_normalize_coords(ureg, src, const0, const1);
+    ureg_MOV(ureg, dst, src);
+
+    if (is_yuv) {
+	src = ureg_DECL_vs_input(ureg, input_slot++);
+	dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 0);
+	ureg_MOV(ureg, dst, src);
+    }
+
+    if (is_composite) {
+	src = ureg_DECL_vs_input(ureg, input_slot++);
+	dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 0);
+	ureg_MOV(ureg, dst, src);
+    }
+
+    if (is_fill) {
+	src = ureg_DECL_vs_input(ureg, input_slot++);
+	dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
+	ureg_MOV(ureg, dst, src);
+    }
+
+    if (has_mask) {
+	src = ureg_DECL_vs_input(ureg, input_slot++);
+	dst = ureg_DECL_output(ureg, TGSI_SEMANTIC_GENERIC, 1);
+	ureg_MOV(ureg, dst, src);
+    }
+
+    ureg_END(ureg);
+
+    return ureg_create_shader_and_destroy(ureg, pipe);
+}
+
+static void *
+create_yuv_shader(struct pipe_context *pipe, struct ureg_program *ureg)
+{
+    struct ureg_src y_sampler, u_sampler, v_sampler;
+    struct ureg_src pos;
+    struct ureg_src matrow0, matrow1, matrow2;
+    struct ureg_dst y, u, v, rgb;
+    struct ureg_dst out = ureg_DECL_output(ureg,
+					   TGSI_SEMANTIC_COLOR,
+					   0);
+
+    pos = ureg_DECL_fs_input(ureg,
+			     TGSI_SEMANTIC_GENERIC, 0,
+			     TGSI_INTERPOLATE_PERSPECTIVE);
+
+    rgb = ureg_DECL_temporary(ureg);
+    y = ureg_DECL_temporary(ureg);
+    u = ureg_DECL_temporary(ureg);
+    v = ureg_DECL_temporary(ureg);
+
+    y_sampler = ureg_DECL_sampler(ureg, 0);
+    u_sampler = ureg_DECL_sampler(ureg, 1);
+    v_sampler = ureg_DECL_sampler(ureg, 2);
+
+    matrow0 = ureg_DECL_constant(ureg, 0);
+    matrow1 = ureg_DECL_constant(ureg, 1);
+    matrow2 = ureg_DECL_constant(ureg, 2);
+
+    ureg_TEX(ureg, y, TGSI_TEXTURE_2D, pos, y_sampler);
+    ureg_TEX(ureg, u, TGSI_TEXTURE_2D, pos, u_sampler);
+    ureg_TEX(ureg, v, TGSI_TEXTURE_2D, pos, v_sampler);
+
+    ureg_SUB(ureg, u, ureg_src(u), ureg_scalar(matrow0, TGSI_SWIZZLE_W));
+    ureg_SUB(ureg, v, ureg_src(v), ureg_scalar(matrow0, TGSI_SWIZZLE_W));
+
+    ureg_MUL(ureg, rgb, ureg_scalar(ureg_src(y), TGSI_SWIZZLE_X), matrow0);
+    ureg_MAD(ureg, rgb,
+	     ureg_scalar(ureg_src(u), TGSI_SWIZZLE_X), matrow1, ureg_src(rgb));
+    ureg_MAD(ureg, rgb,
+	     ureg_scalar(ureg_src(v), TGSI_SWIZZLE_X), matrow2, ureg_src(rgb));
+
+    /* rgb.a = 1; */
+    ureg_MOV(ureg, ureg_writemask(rgb, TGSI_WRITEMASK_W),
+	     ureg_scalar(matrow0, TGSI_SWIZZLE_X));
+
+    ureg_MOV(ureg, out, ureg_src(rgb));
+
+    ureg_release_temporary(ureg, rgb);
+    ureg_release_temporary(ureg, y);
+    ureg_release_temporary(ureg, u);
+    ureg_release_temporary(ureg, v);
+
+    ureg_END(ureg);
+
+    return ureg_create_shader_and_destroy(ureg, pipe);
+}
+
+static INLINE void
+xrender_tex(struct ureg_program *ureg,
+	    struct ureg_dst dst,
+	    struct ureg_src coords,
+	    struct ureg_src sampler,
+	    struct ureg_src imm0,
+	    boolean repeat_none, boolean swizzle, boolean set_alpha)
+{
+    if (repeat_none) {
+	struct ureg_dst tmp0 = ureg_DECL_temporary(ureg);
+	struct ureg_dst tmp1 = ureg_DECL_temporary(ureg);
+
+	ureg_SGT(ureg, tmp1, ureg_swizzle(coords,
+					  TGSI_SWIZZLE_X,
+					  TGSI_SWIZZLE_Y,
+					  TGSI_SWIZZLE_X,
+					  TGSI_SWIZZLE_Y), ureg_scalar(imm0,
+								       TGSI_SWIZZLE_X));
+	ureg_SLT(ureg, tmp0,
+		 ureg_swizzle(coords, TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y,
+			      TGSI_SWIZZLE_X, TGSI_SWIZZLE_Y), ureg_scalar(imm0,
+									   TGSI_SWIZZLE_W));
+	ureg_MIN(ureg, tmp0, ureg_src(tmp0), ureg_src(tmp1));
+	ureg_MIN(ureg, tmp0, ureg_scalar(ureg_src(tmp0), TGSI_SWIZZLE_X),
+		 ureg_scalar(ureg_src(tmp0), TGSI_SWIZZLE_Y));
+	ureg_TEX(ureg, tmp1, TGSI_TEXTURE_2D, coords, sampler);
+	if (swizzle)
+	    ureg_MOV(ureg, tmp1, ureg_swizzle(ureg_src(tmp1),
+					      TGSI_SWIZZLE_Z,
+					      TGSI_SWIZZLE_Y, TGSI_SWIZZLE_X,
+					      TGSI_SWIZZLE_W));
+	if (set_alpha)
+	    ureg_MOV(ureg,
+		     ureg_writemask(tmp1, TGSI_WRITEMASK_W),
+		     ureg_scalar(imm0, TGSI_SWIZZLE_W));
+	ureg_MUL(ureg, dst, ureg_src(tmp1), ureg_src(tmp0));
+	ureg_release_temporary(ureg, tmp0);
+	ureg_release_temporary(ureg, tmp1);
+    } else {
+	if (swizzle) {
+	    struct ureg_dst tmp = ureg_DECL_temporary(ureg);
+
+	    ureg_TEX(ureg, tmp, TGSI_TEXTURE_2D, coords, sampler);
+	    ureg_MOV(ureg, dst, ureg_swizzle(ureg_src(tmp),
+					     TGSI_SWIZZLE_Z,
+					     TGSI_SWIZZLE_Y, TGSI_SWIZZLE_X,
+					     TGSI_SWIZZLE_W));
+	    ureg_release_temporary(ureg, tmp);
+	} else {
+	    ureg_TEX(ureg, dst, TGSI_TEXTURE_2D, coords, sampler);
+	}
+	if (set_alpha)
+	    ureg_MOV(ureg,
+		     ureg_writemask(dst, TGSI_WRITEMASK_W),
+		     ureg_scalar(imm0, TGSI_SWIZZLE_W));
+    }
+}
+
+static void *
+create_fs(struct pipe_context *pipe, unsigned fs_traits)
+{
+    struct ureg_program *ureg;
+    struct ureg_src /*dst_sampler, */ src_sampler, mask_sampler;
+    struct ureg_src /*dst_pos, */ src_input, mask_pos;
+    struct ureg_dst src, mask;
+    struct ureg_dst out;
+    struct ureg_src imm0 = { 0 };
+    unsigned has_mask = (fs_traits & FS_MASK) != 0;
+    unsigned is_fill = (fs_traits & FS_FILL) != 0;
+    unsigned is_composite = (fs_traits & FS_COMPOSITE) != 0;
+    unsigned is_solid = (fs_traits & FS_SOLID_FILL) != 0;
+    unsigned is_lingrad = (fs_traits & FS_LINGRAD_FILL) != 0;
+    unsigned is_radgrad = (fs_traits & FS_RADGRAD_FILL) != 0;
+    unsigned comp_alpha_mask = fs_traits & FS_COMPONENT_ALPHA;
+    unsigned is_yuv = (fs_traits & FS_YUV) != 0;
+    unsigned src_repeat_none = (fs_traits & FS_SRC_REPEAT_NONE) != 0;
+    unsigned mask_repeat_none = (fs_traits & FS_MASK_REPEAT_NONE) != 0;
+    unsigned src_swizzle = (fs_traits & FS_SRC_SWIZZLE_RGB) != 0;
+    unsigned mask_swizzle = (fs_traits & FS_MASK_SWIZZLE_RGB) != 0;
+    unsigned src_set_alpha = (fs_traits & FS_SRC_SET_ALPHA) != 0;
+    unsigned mask_set_alpha = (fs_traits & FS_MASK_SET_ALPHA) != 0;
+    unsigned src_luminance = (fs_traits & FS_SRC_LUMINANCE) != 0;
+    unsigned mask_luminance = (fs_traits & FS_MASK_LUMINANCE) != 0;
+
+#if 0
+    print_fs_traits(fs_traits);
+#else
+    (void)print_fs_traits;
+#endif
+
+    ureg = ureg_create(TGSI_PROCESSOR_FRAGMENT);
+    if (ureg == NULL)
+	return 0;
+
+    /* it has to be either a fill, a composite op or a yuv conversion */
+    debug_assert((is_fill ^ is_composite) ^ is_yuv);
+    (void)is_yuv;
+
+    out = ureg_DECL_output(ureg, TGSI_SEMANTIC_COLOR, 0);
+
+    if (src_repeat_none || mask_repeat_none ||
+	src_set_alpha || mask_set_alpha || src_luminance) {
+	imm0 = ureg_imm4f(ureg, 0, 0, 0, 1);
+    }
+    if (is_composite) {
+	src_sampler = ureg_DECL_sampler(ureg, 0);
+	src_input = ureg_DECL_fs_input(ureg,
+				       TGSI_SEMANTIC_GENERIC, 0,
+				       TGSI_INTERPOLATE_PERSPECTIVE);
+    } else if (is_fill) {
+	if (is_solid)
+	    src_input = ureg_DECL_fs_input(ureg,
+					   TGSI_SEMANTIC_COLOR, 0,
+					   TGSI_INTERPOLATE_PERSPECTIVE);
+	else
+	    src_input = ureg_DECL_fs_input(ureg,
+					   TGSI_SEMANTIC_POSITION, 0,
+					   TGSI_INTERPOLATE_PERSPECTIVE);
+    } else {
+	debug_assert(is_yuv);
+	return create_yuv_shader(pipe, ureg);
+    }
+
+    if (has_mask) {
+	mask_sampler = ureg_DECL_sampler(ureg, 1);
+	mask_pos = ureg_DECL_fs_input(ureg,
+				      TGSI_SEMANTIC_GENERIC, 1,
+				      TGSI_INTERPOLATE_PERSPECTIVE);
+    }
+#if 0				/* unused right now */
+    dst_sampler = ureg_DECL_sampler(ureg, 2);
+    dst_pos = ureg_DECL_fs_input(ureg,
+				 TGSI_SEMANTIC_POSITION, 2,
+				 TGSI_INTERPOLATE_PERSPECTIVE);
+#endif
+
+    if (is_composite) {
+	if (has_mask || src_luminance)
+	    src = ureg_DECL_temporary(ureg);
+	else
+	    src = out;
+	xrender_tex(ureg, src, src_input, src_sampler, imm0,
+		    src_repeat_none, src_swizzle, src_set_alpha);
+    } else if (is_fill) {
+	if (is_solid) {
+	    if (has_mask || src_luminance)
+		src = ureg_dst(src_input);
+	    else
+		ureg_MOV(ureg, out, src_input);
+	} else if (is_lingrad || is_radgrad) {
+	    struct ureg_src coords, const0124, matrow0, matrow1, matrow2;
+
+	    if (has_mask || src_luminance)
+		src = ureg_DECL_temporary(ureg);
+	    else
+		src = out;
+
+	    coords = ureg_DECL_constant(ureg, 0);
+	    const0124 = ureg_DECL_constant(ureg, 1);
+	    matrow0 = ureg_DECL_constant(ureg, 2);
+	    matrow1 = ureg_DECL_constant(ureg, 3);
+	    matrow2 = ureg_DECL_constant(ureg, 4);
+
+	    if (is_lingrad) {
+		linear_gradient(ureg, src,
+				src_input, src_sampler,
+				coords, const0124, matrow0, matrow1, matrow2);
+	    } else if (is_radgrad) {
+		radial_gradient(ureg, src,
+				src_input, src_sampler,
+				coords, const0124, matrow0, matrow1, matrow2);
+	    }
+	} else
+	    debug_assert(!"Unknown fill type!");
+    }
+    if (src_luminance) {
+	ureg_MOV(ureg, src, ureg_scalar(ureg_src(src), TGSI_SWIZZLE_X));
+	ureg_MOV(ureg, ureg_writemask(src, TGSI_WRITEMASK_XYZ),
+		 ureg_scalar(imm0, TGSI_SWIZZLE_X));
+	if (!has_mask)
+	    ureg_MOV(ureg, out, ureg_src(src));
+    }
+
+    if (has_mask) {
+	mask = ureg_DECL_temporary(ureg);
+	xrender_tex(ureg, mask, mask_pos, mask_sampler, imm0,
+		    mask_repeat_none, mask_swizzle, mask_set_alpha);
+	/* src IN mask */
+	src_in_mask(ureg, out, ureg_src(src), ureg_src(mask),
+		    comp_alpha_mask, mask_luminance);
+	ureg_release_temporary(ureg, mask);
+    }
+
+    ureg_END(ureg);
+
+    return ureg_create_shader_and_destroy(ureg, pipe);
+}
+
+struct xa_shaders *
+xa_shaders_create(struct xa_context *r)
+{
+    struct xa_shaders *sc = CALLOC_STRUCT(xa_shaders);
+
+    sc->r = r;
+    sc->vs_hash = cso_hash_create();
+    sc->fs_hash = cso_hash_create();
+
+    return sc;
+}
+
+static void
+cache_destroy(struct cso_context *cso,
+	      struct cso_hash *hash, unsigned processor)
+{
+    struct cso_hash_iter iter = cso_hash_first_node(hash);
+
+    while (!cso_hash_iter_is_null(iter)) {
+	void *shader = (void *)cso_hash_iter_data(iter);
+
+	if (processor == PIPE_SHADER_FRAGMENT) {
+	    cso_delete_fragment_shader(cso, shader);
+	} else if (processor == PIPE_SHADER_VERTEX) {
+	    cso_delete_vertex_shader(cso, shader);
+	}
+	iter = cso_hash_erase(hash, iter);
+    }
+    cso_hash_delete(hash);
+}
+
+void
+xa_shaders_destroy(struct xa_shaders *sc)
+{
+    cache_destroy(sc->r->cso, sc->vs_hash, PIPE_SHADER_VERTEX);
+    cache_destroy(sc->r->cso, sc->fs_hash, PIPE_SHADER_FRAGMENT);
+
+    FREE(sc);
+}
+
+static INLINE void *
+shader_from_cache(struct pipe_context *pipe,
+		  unsigned type, struct cso_hash *hash, unsigned key)
+{
+    void *shader = 0;
+
+    struct cso_hash_iter iter = cso_hash_find(hash, key);
+
+    if (cso_hash_iter_is_null(iter)) {
+	if (type == PIPE_SHADER_VERTEX)
+	    shader = create_vs(pipe, key);
+	else
+	    shader = create_fs(pipe, key);
+	cso_hash_insert(hash, key, shader);
+    } else
+	shader = (void *)cso_hash_iter_data(iter);
+
+    return shader;
+}
+
+struct xa_shader
+xa_shaders_get(struct xa_shaders *sc, unsigned vs_traits, unsigned fs_traits)
+{
+    struct xa_shader shader = { NULL, NULL };
+    void *vs, *fs;
+
+    vs = shader_from_cache(sc->r->pipe, PIPE_SHADER_VERTEX,
+			   sc->vs_hash, vs_traits);
+    fs = shader_from_cache(sc->r->pipe, PIPE_SHADER_FRAGMENT,
+			   sc->fs_hash, fs_traits);
+
+    debug_assert(vs && fs);
+    if (!vs || !fs)
+	return shader;
+
+    shader.vs = vs;
+    shader.fs = fs;
+
+    return shader;
+}
diff --git a/src/gallium/state_trackers/xa/xa_tracker.c b/src/gallium/state_trackers/xa/xa_tracker.c
new file mode 100644
index 00000000000..50922d38378
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_tracker.c
@@ -0,0 +1,448 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#include "xa_tracker.h"
+#include "xa_priv.h"
+#include "pipe/p_state.h"
+#include "pipe/p_format.h"
+#include "state_tracker/drm_driver.h"
+#include "util/u_inlines.h"
+
+/*
+ * format_map [xa_surface_type][first..last in list].
+ * Needs to be updated when enum xa_formats is updated.
+ */
+
+static const enum xa_formats preferred_a[] = { xa_format_a8 };
+
+static const enum xa_formats preferred_argb[] =
+    { xa_format_a8r8g8b8, xa_format_x8r8g8b8, xa_format_r5g6b5,
+    xa_format_x1r5g5b5
+};
+static const enum xa_formats preferred_z[] =
+    { xa_format_z32, xa_format_z24, xa_format_z16 };
+static const enum xa_formats preferred_sz[] =
+    { xa_format_x8z24, xa_format_s8z24 };
+static const enum xa_formats preferred_zs[] =
+    { xa_format_z24x8, xa_format_z24s8 };
+static const enum xa_formats preferred_yuv[] = { xa_format_yuv8 };
+
+static const enum xa_formats *preferred[] =
+    { NULL, preferred_a, preferred_argb, NULL, NULL,
+    preferred_z, preferred_zs, preferred_sz, preferred_yuv
+};
+
+static const unsigned int num_preferred[] = { 0,
+    sizeof(preferred_a) / sizeof(enum xa_formats),
+    sizeof(preferred_argb) / sizeof(enum xa_formats),
+    0,
+    0,
+    sizeof(preferred_z) / sizeof(enum xa_formats),
+    sizeof(preferred_zs) / sizeof(enum xa_formats),
+    sizeof(preferred_sz) / sizeof(enum xa_formats),
+    sizeof(preferred_yuv) / sizeof(enum xa_formats)
+};
+
+static const unsigned int stype_bind[XA_LAST_SURFACE_TYPE] = { 0,
+    PIPE_BIND_SAMPLER_VIEW,
+    PIPE_BIND_SAMPLER_VIEW,
+    PIPE_BIND_SAMPLER_VIEW,
+    PIPE_BIND_SAMPLER_VIEW,
+    PIPE_BIND_DEPTH_STENCIL,
+    PIPE_BIND_DEPTH_STENCIL,
+    PIPE_BIND_DEPTH_STENCIL,
+    PIPE_BIND_SAMPLER_VIEW
+};
+
+static struct xa_format_descriptor
+xa_get_pipe_format(enum xa_formats xa_format)
+{
+    struct xa_format_descriptor fdesc;
+
+    fdesc.xa_format = xa_format;
+
+    switch (xa_format) {
+    case xa_format_a8r8g8b8:
+	fdesc.format = PIPE_FORMAT_B8G8R8A8_UNORM;
+	break;
+    case xa_format_x8r8g8b8:
+	fdesc.format = PIPE_FORMAT_B8G8R8X8_UNORM;
+	break;
+    case xa_format_r5g6b5:
+	fdesc.format = PIPE_FORMAT_B5G6R5_UNORM;
+	break;
+    case xa_format_x1r5g5b5:
+	fdesc.format = PIPE_FORMAT_B5G5R5A1_UNORM;
+	break;
+    case xa_format_a8:
+	fdesc.format = PIPE_FORMAT_L8_UNORM;
+	break;
+    case xa_format_z24:
+	fdesc.format = PIPE_FORMAT_Z24X8_UNORM;
+	break;
+    case xa_format_z16:
+	fdesc.format = PIPE_FORMAT_Z16_UNORM;
+	break;
+    case xa_format_z32:
+	fdesc.format = PIPE_FORMAT_Z32_UNORM;
+	break;
+    case xa_format_x8z24:
+	fdesc.format = PIPE_FORMAT_Z24X8_UNORM;
+	break;
+    case xa_format_z24x8:
+	fdesc.format = PIPE_FORMAT_X8Z24_UNORM;
+	break;
+    case xa_format_s8z24:
+	fdesc.format = PIPE_FORMAT_Z24_UNORM_S8_USCALED;
+	break;
+    case xa_format_z24s8:
+	fdesc.format = PIPE_FORMAT_S8_USCALED_Z24_UNORM;
+	break;
+    case xa_format_yuv8:
+	fdesc.format = PIPE_FORMAT_L8_UNORM;
+	break;
+    default:
+	fdesc.xa_format = xa_format_unknown;
+	break;
+    }
+    return fdesc;
+}
+
+struct xa_tracker *
+xa_tracker_create(int drm_fd)
+{
+    struct xa_tracker *xa = calloc(1, sizeof(struct xa_tracker));
+    enum xa_surface_type stype;
+    unsigned int num_formats;
+
+    if (!xa)
+	return NULL;
+
+    xa->screen = driver_descriptor.create_screen(drm_fd);
+    if (!xa->screen)
+	goto out_no_screen;
+
+    xa->default_ctx = xa_context_create(xa);
+    if (!xa->default_ctx)
+	goto out_no_pipe;
+
+    num_formats = 0;
+    for (stype = 0; stype < XA_LAST_SURFACE_TYPE; ++stype)
+	num_formats += num_preferred[stype];
+
+    num_formats += 1;
+    xa->supported_formats = calloc(num_formats, sizeof(*xa->supported_formats));
+    if (!xa->supported_formats)
+	goto out_sf_alloc_fail;
+
+    xa->supported_formats[0] = xa_format_unknown;
+    num_formats = 1;
+    memset(xa->format_map, 0, sizeof(xa->format_map));
+
+    for (stype = 0; stype < XA_LAST_SURFACE_TYPE; ++stype) {
+	unsigned int bind = stype_bind[stype];
+	enum xa_formats xa_format;
+	int i;
+
+	for (i = 0; i < num_preferred[stype]; ++i) {
+	    xa_format = preferred[stype][i];
+
+	    struct xa_format_descriptor fdesc = xa_get_pipe_format(xa_format);
+
+	    if (xa->screen->is_format_supported(xa->screen, fdesc.format,
+						PIPE_TEXTURE_2D, 0, bind)) {
+		if (xa->format_map[stype][0] == 0)
+		    xa->format_map[stype][0] = num_formats;
+		xa->format_map[stype][1] = num_formats;
+		xa->supported_formats[num_formats++] = xa_format;
+	    }
+	}
+    }
+    return xa;
+
+ out_sf_alloc_fail:
+    xa_context_destroy(xa->default_ctx);
+ out_no_pipe:
+    xa->screen->destroy(xa->screen);
+ out_no_screen:
+    free(xa);
+    return NULL;
+}
+
+void
+xa_tracker_destroy(struct xa_tracker *xa)
+{
+    free(xa->supported_formats);
+    xa_context_destroy(xa->default_ctx);
+    xa->screen->destroy(xa->screen);
+    free(xa);
+}
+
+static int
+xa_flags_compat(unsigned int old_flags, unsigned int new_flags)
+{
+    unsigned int flag_diff = (old_flags ^ new_flags);
+
+    if (flag_diff == 0)
+	return 1;
+
+    if (flag_diff & XA_FLAG_SHARED)
+	return 0;
+    /*
+     * Don't recreate if we're dropping the render target flag.
+     */
+    if (flag_diff & XA_FLAG_RENDER_TARGET)
+	return ((new_flags & XA_FLAG_RENDER_TARGET) == 0);
+
+    /*
+     * Always recreate for unknown / unimplemented flags.
+     */
+    return 0;
+}
+
+static struct xa_format_descriptor
+xa_get_format_stype_depth(struct xa_tracker *xa,
+			  enum xa_surface_type stype, unsigned int depth)
+{
+    unsigned int i;
+    struct xa_format_descriptor fdesc;
+    int found = 0;
+
+    for (i = xa->format_map[stype][0]; i <= xa->format_map[stype][1]; ++i) {
+	fdesc = xa_get_pipe_format(xa->supported_formats[i]);
+	if (fdesc.xa_format != xa_format_unknown &&
+	    xa_format_depth(fdesc.xa_format) == depth) {
+	    found = 1;
+	    break;
+	}
+    }
+
+    if (!found)
+	fdesc.xa_format = xa_format_unknown;
+
+    return fdesc;
+}
+
+int
+xa_format_check_supported(struct xa_tracker *xa,
+			  enum xa_formats xa_format, unsigned int flags)
+{
+    struct xa_format_descriptor fdesc = xa_get_pipe_format(xa_format);
+    unsigned int bind;
+
+    if (fdesc.xa_format == xa_format_unknown)
+	return -XA_ERR_INVAL;
+
+    bind = stype_bind[xa_format_type(fdesc.xa_format)];
+    if (flags & XA_FLAG_SHARED)
+	bind |= PIPE_BIND_SHARED;
+    if (flags & XA_FLAG_RENDER_TARGET)
+	bind |= PIPE_BIND_RENDER_TARGET;
+
+    if (!xa->screen->is_format_supported(xa->screen, fdesc.format,
+					 PIPE_TEXTURE_2D, 0, bind))
+	return -XA_ERR_INVAL;
+
+    return XA_ERR_NONE;
+}
+
+struct xa_surface *
+xa_surface_create(struct xa_tracker *xa,
+		  int width,
+		  int height,
+		  int depth,
+		  enum xa_surface_type stype,
+		  enum xa_formats xa_format, unsigned int flags)
+{
+    struct pipe_resource *template;
+    struct xa_surface *srf;
+    struct xa_format_descriptor fdesc;
+
+    if (xa_format == xa_format_unknown)
+	fdesc = xa_get_format_stype_depth(xa, stype, depth);
+    else
+	fdesc = xa_get_pipe_format(xa_format);
+
+    if (fdesc.xa_format == xa_format_unknown)
+	return NULL;
+
+    srf = calloc(1, sizeof(*srf));
+    if (!srf)
+	return NULL;
+
+    template = &srf->template;
+    template->format = fdesc.format;
+    template->target = PIPE_TEXTURE_2D;
+    template->width0 = width;
+    template->height0 = height;
+    template->depth0 = 1;
+    template->array_size = 1;
+    template->last_level = 0;
+    template->bind = stype_bind[xa_format_type(fdesc.xa_format)];
+
+    if (flags & XA_FLAG_SHARED)
+	template->bind |= PIPE_BIND_SHARED;
+    if (flags & XA_FLAG_RENDER_TARGET)
+	template->bind |= PIPE_BIND_RENDER_TARGET;
+
+    srf->tex = xa->screen->resource_create(xa->screen, template);
+    if (!srf->tex)
+	goto out_no_tex;
+
+    srf->srf = NULL;
+    srf->xa = xa;
+    srf->flags = flags;
+    srf->fdesc = fdesc;
+
+    return srf;
+ out_no_tex:
+    free(srf);
+    return NULL;
+}
+
+int
+xa_surface_redefine(struct xa_surface *srf,
+		    int width,
+		    int height,
+		    int depth,
+		    enum xa_surface_type stype,
+		    enum xa_formats xa_format,
+		    unsigned int new_flags,
+		    int copy_contents)
+{
+    struct pipe_resource *template = &srf->template;
+    struct pipe_resource *texture;
+    struct pipe_box src_box;
+    struct xa_tracker *xa = srf->xa;
+    int save_width;
+    int save_height;
+    struct xa_format_descriptor fdesc;
+
+    if (xa_format == xa_format_unknown)
+	fdesc = xa_get_format_stype_depth(xa, stype, depth);
+    else
+	fdesc = xa_get_pipe_format(xa_format);
+
+    if (width == template->width0 && height == template->height0 &&
+	template->format == fdesc.format &&
+	xa_flags_compat(srf->flags, new_flags))
+	return XA_ERR_NONE;
+
+    template->bind = stype_bind[xa_format_type(fdesc.xa_format)];
+    if (new_flags & XA_FLAG_SHARED)
+	template->bind |= PIPE_BIND_SHARED;
+    if (new_flags & XA_FLAG_RENDER_TARGET)
+	template->bind |= PIPE_BIND_RENDER_TARGET;
+
+    if (copy_contents) {
+	if (!xa_format_type_is_color(fdesc.xa_format) ||
+	    xa_format_type(fdesc.xa_format) == xa_type_a)
+	    return -XA_ERR_INVAL;
+
+	if (!xa->screen->is_format_supported(xa->screen, fdesc.format,
+					     PIPE_TEXTURE_2D, 0,
+					     template->bind |
+					     PIPE_BIND_RENDER_TARGET))
+	    return -XA_ERR_INVAL;
+    }
+
+    save_width = template->width0;
+    save_height = template->height0;
+
+    template->width0 = width;
+    template->height0 = height;
+
+    texture = xa->screen->resource_create(xa->screen, template);
+    if (!texture) {
+	template->width0 = save_width;
+	template->height0 = save_height;
+	return -XA_ERR_NORES;
+    }
+
+    pipe_surface_reference(&srf->srf, NULL);
+
+    if (copy_contents) {
+	struct pipe_context *pipe = xa->default_ctx->pipe;
+
+	u_box_origin_2d(xa_min(save_width, template->width0),
+			xa_min(save_height, template->height0), &src_box);
+	pipe->resource_copy_region(pipe, texture,
+				   0, 0, 0, 0, srf->tex, 0, &src_box);
+	pipe->flush(pipe, &xa->default_ctx->last_fence);
+    }
+
+    pipe_resource_reference(&srf->tex, texture);
+    pipe_resource_reference(&texture, NULL);
+    srf->fdesc = fdesc;
+    srf->flags = new_flags;
+
+    return XA_ERR_NONE;
+}
+
+void
+xa_surface_destroy(struct xa_surface *srf)
+{
+    pipe_surface_reference(&srf->srf, NULL);
+    pipe_resource_reference(&srf->tex, NULL);
+    free(srf);
+}
+
+extern void
+xa_tracker_version(int *major, int *minor, int *patch)
+{
+    *major = XA_TRACKER_VERSION_MAJOR;
+    *minor = XA_TRACKER_VERSION_MINOR;
+    *patch = XA_TRACKER_VERSION_PATCH;
+}
+
+extern int
+xa_surface_handle(struct xa_surface *srf,
+		  uint32_t * handle, unsigned int *stride)
+{
+    struct winsys_handle whandle;
+
+    struct pipe_screen *screen = srf->xa->screen;
+    boolean res;
+
+    memset(&whandle, 0, sizeof(whandle));
+    whandle.type = DRM_API_HANDLE_TYPE_SHARED;
+    res = screen->resource_get_handle(screen, srf->tex, &whandle);
+    if (!res)
+	return -XA_ERR_INVAL;
+
+    *handle = whandle.handle;
+    *stride = whandle.stride;
+
+    return XA_ERR_NONE;
+}
+
+enum xa_formats
+xa_surface_format(const struct xa_surface *srf)
+{
+    return srf->fdesc.xa_format;
+}
diff --git a/src/gallium/state_trackers/xa/xa_tracker.h b/src/gallium/state_trackers/xa/xa_tracker.h
new file mode 100644
index 00000000000..62f8a210fb6
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_tracker.h
@@ -0,0 +1,178 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ * The format encoding idea is partially borrowed from libpixman, but it is not
+ * considered a "substantial part of the software", so the pixman copyright
+ * is left out for simplicity, and acknowledgment is instead given in this way.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ * Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#ifndef _XA_TRACKER_H_
+#define _XA_TRACKER_H_
+
+#include <stdint.h>
+
+#define XA_TRACKER_VERSION_MAJOR 0
+#define XA_TRACKER_VERSION_MINOR 4
+#define XA_TRACKER_VERSION_PATCH 0
+
+#define XA_FLAG_SHARED         (1 << 0)
+#define XA_FLAG_RENDER_TARGET  (1 << 1)
+
+#define XA_MAP_READ            (1 << 0)
+#define XA_MAP_WRITE           (1 << 1)
+
+#define XA_ERR_NONE            0
+#define XA_ERR_NORES           1
+#define XA_ERR_INVAL           2
+#define XA_ERR_BUSY            3
+
+enum xa_surface_type {
+    xa_type_other,
+    xa_type_a,
+    xa_type_argb,
+    xa_type_abgr,
+    xa_type_bgra,
+    xa_type_z,
+    xa_type_zs,
+    xa_type_sz,
+    xa_type_yuv_component
+};
+
+/*
+ * Note that these formats should not be assumed to be binary compatible with
+ * pixman formats, but with the below macros and a format type map,
+ * conversion should be simple. Macros for now. We might replace with
+ * inline functions.
+ */
+
+#define xa_format(bpp,type,a,r,g,b)	(((bpp) << 24) |  \
+					 ((type) << 16) | \
+					 ((a) << 12) |	  \
+					 ((r) << 8) |	  \
+					 ((g) << 4) |	  \
+					 ((b)))
+/*
+ *  Non-RGBA one- and two component formats.
+ */
+
+#define xa_format_c(bpp,type,c1,c2) (((bpp) << 24) |	  \
+				     ((type) << 16) |	  \
+				     ((c1) << 8) |	  \
+				     ((c2)))
+#define xa_format_bpp(f)	(((f) >> 24)       )
+#define xa_format_type(f)	(((f) >> 16) & 0xff)
+#define xa_format_a(f)	(((f) >> 12) & 0x0f)
+#define xa_format_r(f)	(((f) >>  8) & 0x0f)
+#define xa_format_g(f)	(((f) >>  4) & 0x0f)
+#define xa_format_b(f)	(((f)      ) & 0x0f)
+#define xa_format_rgb(f)	(((f)      ) & 0xfff)
+#define xa_format_c1(f)          (((f) >> 8 ) & 0xff)
+#define xa_format_c2(f)          (((f)      ) & 0xff)
+#define xa_format_argb_depth(f)	(xa_format_a(f) +	\
+				 xa_format_r(f) +	\
+				 xa_format_g(f) +	\
+				 xa_format_b(f))
+#define xa_format_c_depth(f)    (xa_format_c1(f) + \
+				 xa_format_c2(f))
+
+static inline int
+xa_format_type_is_color(uint32_t xa_format)
+{
+    return (xa_format_type(xa_format) < xa_type_z);
+}
+
+static inline unsigned int
+xa_format_depth(uint32_t xa_format)
+{
+    return ((xa_format_type_is_color(xa_format)) ?
+	    xa_format_argb_depth(xa_format) : xa_format_c_depth(xa_format));
+}
+
+enum xa_formats {
+    xa_format_unknown = 0,
+    xa_format_a8 = xa_format(8, xa_type_a, 8, 0, 0, 0),
+
+    xa_format_a8r8g8b8 = xa_format(32, xa_type_argb, 8, 8, 8, 8),
+    xa_format_x8r8g8b8 = xa_format(32, xa_type_argb, 0, 8, 8, 8),
+    xa_format_r5g6b5 = xa_format(16, xa_type_argb, 0, 5, 6, 5),
+    xa_format_x1r5g5b5 = xa_format(16, xa_type_argb, 0, 5, 5, 5),
+
+    xa_format_z16 = xa_format_c(16, xa_type_z, 16, 0),
+    xa_format_z32 = xa_format_c(32, xa_type_z, 32, 0),
+    xa_format_z24 = xa_format_c(32, xa_type_z, 24, 0),
+
+    xa_format_x8z24 = xa_format_c(32, xa_type_sz, 24, 0),
+    xa_format_s8z24 = xa_format_c(32, xa_type_sz, 24, 8),
+    xa_format_z24x8 = xa_format_c(32, xa_type_zs, 24, 0),
+    xa_format_z24s8 = xa_format_c(32, xa_type_zs, 24, 8),
+
+    xa_format_yuv8 = xa_format_c(8, xa_type_yuv_component, 8, 0)
+};
+
+struct xa_tracker;
+struct xa_surface;
+
+struct xa_box {
+    uint16_t x1, y1, x2, y2;
+};
+
+extern void xa_tracker_version(int *major, int *minor, int *patch);
+
+extern struct xa_tracker *xa_tracker_create(int drm_fd);
+
+extern void xa_tracker_destroy(struct xa_tracker *xa);
+
+extern int xa_format_check_supported(struct xa_tracker *xa,
+				     enum xa_formats xa_format,
+				     unsigned int flags);
+
+extern struct xa_surface *xa_surface_create(struct xa_tracker *xa,
+					    int width,
+					    int height,
+					    int depth,
+					    enum xa_surface_type stype,
+					    enum xa_formats pform,
+					    unsigned int flags);
+
+enum xa_formats xa_surface_format(const struct xa_surface *srf);
+
+extern void xa_surface_destroy(struct xa_surface *srf);
+
+extern int xa_surface_redefine(struct xa_surface *srf,
+			       int width,
+			       int height,
+			       int depth,
+			       enum xa_surface_type stype,
+			       enum xa_formats rgb_format,
+			       unsigned int new_flags,
+			       int copy_contents);
+
+extern int xa_surface_handle(struct xa_surface *srf,
+			     uint32_t * handle, unsigned int *byte_stride);
+
+#endif
diff --git a/src/gallium/state_trackers/xa/xa_yuv.c b/src/gallium/state_trackers/xa/xa_yuv.c
new file mode 100644
index 00000000000..66cbc5393b5
--- /dev/null
+++ b/src/gallium/state_trackers/xa/xa_yuv.c
@@ -0,0 +1,179 @@
+/**********************************************************
+ * Copyright 2009-2011 VMware, Inc. All rights reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person
+ * obtaining a copy of this software and associated documentation
+ * files (the "Software"), to deal in the Software without
+ * restriction, including without limitation the rights to use, copy,
+ * modify, merge, publish, distribute, sublicense, and/or sell copies
+ * of the Software, and to permit persons to whom the Software is
+ * furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be
+ * included in all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ *
+ *********************************************************
+ * Authors:
+ * Zack Rusin <zackr-at-vmware-dot-com>
+ * Thomas Hellstrom <thellstrom-at-vmware-dot-com>
+ */
+
+#include "xa_context.h"
+#include "xa_priv.h"
+#include "util/u_inlines.h"
+#include "util/u_sampler.h"
+#include "util/u_surface.h"
+#include "cso_cache/cso_context.h"
+
+static void
+xa_yuv_bind_blend_state(struct xa_context *r)
+{
+    struct pipe_blend_state blend;
+
+    memset(&blend, 0, sizeof(struct pipe_blend_state));
+    blend.rt[0].blend_enable = 0;
+    blend.rt[0].colormask = PIPE_MASK_RGBA;
+
+    /* porter&duff src */
+    blend.rt[0].rgb_src_factor = PIPE_BLENDFACTOR_ONE;
+    blend.rt[0].alpha_src_factor = PIPE_BLENDFACTOR_ONE;
+    blend.rt[0].rgb_dst_factor = PIPE_BLENDFACTOR_ZERO;
+    blend.rt[0].alpha_dst_factor = PIPE_BLENDFACTOR_ZERO;
+
+    cso_set_blend(r->cso, &blend);
+}
+
+static void
+xa_yuv_bind_shaders(struct xa_context *r)
+{
+    unsigned vs_traits = 0, fs_traits = 0;
+    struct xa_shader shader;
+
+    vs_traits |= VS_YUV;
+    fs_traits |= FS_YUV;
+
+    shader = xa_shaders_get(r->shaders, vs_traits, fs_traits);
+    cso_set_vertex_shader_handle(r->cso, shader.vs);
+    cso_set_fragment_shader_handle(r->cso, shader.fs);
+}
+
+static void
+xa_yuv_bind_samplers(struct xa_context *r, struct xa_surface *yuv[])
+{
+    struct pipe_sampler_state *samplers[3];
+    struct pipe_sampler_state sampler;
+    struct pipe_sampler_view *views[3];
+    struct pipe_sampler_view view_templ;
+    unsigned int i;
+
+    memset(&sampler, 0, sizeof(struct pipe_sampler_state));
+
+    sampler.wrap_s = PIPE_TEX_WRAP_CLAMP;
+    sampler.wrap_t = PIPE_TEX_WRAP_CLAMP;
+    sampler.min_img_filter = PIPE_TEX_FILTER_LINEAR;
+    sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR;
+    sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NEAREST;
+    sampler.normalized_coords = 1;
+
+    for (i = 0; i < 3; ++i) {
+	samplers[i] = &sampler;
+	if (!yuv[i]->view) {
+	    u_sampler_view_default_template(&view_templ,
+					    yuv[i]->tex, yuv[i]->tex->format);
+
+	    yuv[i]->view = r->pipe->create_sampler_view(r->pipe,
+							yuv[i]->tex,
+							&view_templ);
+	}
+	views[i] = yuv[i]->view;
+    }
+
+    cso_set_samplers(r->cso, 3, (const struct pipe_sampler_state **)samplers);
+    cso_set_fragment_sampler_views(r->cso, 3, views);
+}
+
+static void
+xa_yuv_fs_constants(struct xa_context *r, const float conversion_matrix[])
+{
+    const int param_bytes = 12 * sizeof(float);
+
+    renderer_set_constants(r, PIPE_SHADER_FRAGMENT,
+			   conversion_matrix, param_bytes);
+}
+
+static void
+xa_yuv_destroy_sampler_views(struct xa_surface *yuv[])
+{
+    unsigned int i;
+
+    for (i = 0; i < 3; ++i) {
+	pipe_sampler_view_reference(&yuv[i]->view, NULL);
+    }
+}
+
+extern int
+xa_yuv_planar_blit(struct xa_context *r,
+		   int src_x,
+		   int src_y,
+		   int src_w,
+		   int src_h,
+		   int dst_x,
+		   int dst_y,
+		   int dst_w,
+		   int dst_h,
+		   struct xa_box *box,
+		   unsigned int num_boxes,
+		   const float conversion_matrix[],
+		   struct xa_surface *dst, struct xa_surface *yuv[])
+{
+    float scale_x;
+    float scale_y;
+    struct pipe_surface srf_templ;
+
+    if (dst_w == 0 || dst_h == 0)
+	return XA_ERR_NONE;
+
+    memset(&srf_templ, 0, sizeof(srf_templ));
+    u_surface_default_template(&srf_templ, dst->tex, PIPE_BIND_RENDER_TARGET);
+    dst->srf = r->pipe->create_surface(r->pipe, dst->tex, &srf_templ);
+    if (!dst->srf)
+	return -XA_ERR_NORES;
+
+    renderer_bind_destination(r, dst->srf, dst->srf->width, dst->srf->height);
+    xa_yuv_bind_blend_state(r);
+    xa_yuv_bind_shaders(r);
+    xa_yuv_bind_samplers(r, yuv);
+    xa_yuv_fs_constants(r, conversion_matrix);
+
+    scale_x = (float)src_w / (float)dst_w;
+    scale_y = (float)src_h / (float)dst_h;
+
+    while (num_boxes--) {
+	int x = box->x1;
+	int y = box->y1;
+	int w = box->x2 - box->x1;
+	int h = box->y2 - box->y1;
+
+	renderer_draw_yuv(r,
+			  (float)src_x + scale_x * (x - dst_x),
+			  (float)src_y + scale_y * (y - dst_y),
+			  scale_x * w, scale_y * h, x, y, w, h, yuv);
+	box++;
+    }
+
+    r->pipe->flush(r->pipe, &r->last_fence);
+
+    xa_yuv_destroy_sampler_views(yuv);
+    pipe_surface_reference(&dst->srf, NULL);
+
+    return XA_ERR_NONE;
+}