From c11a7ec821d41b91a3825c5abfb4687c98b5bf98 Mon Sep 17 00:00:00 2001
From: Younes Manton <younes.m@gmail.com>
Date: Sun, 8 Jun 2008 03:04:14 -0400
Subject: Initial commit for g3dvl.

Initial commit for g3dvl, contains support for basic XvMC features.

- Context, surface, block, macroblock creation and deletion
- Surface rendering
  - Frame pictures
  - Frame based motion compensation
  - Intra-coded macroblocks
  - Predicted macroblocks
  - Bi-directionally predicted macroblocks
- Surface display
  - Color conversion
  - Scaling
---
 src/gallium/state_trackers/g3dvl/Makefile | 18 ++++++++++++++++++
 1 file changed, 18 insertions(+)
 create mode 100644 src/gallium/state_trackers/g3dvl/Makefile

(limited to 'src/gallium/state_trackers/g3dvl/Makefile')

diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
new file mode 100644
index 0000000000..a0d85fbcc8
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -0,0 +1,18 @@
+TARGET		= libg3dvl.a
+OBJECTS		= vl_context.o vl_data.o vl_surface.o
+GALLIUMDIR	= ../..
+
+CFLAGS		+= -g -Wall -fPIC -Werror -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary
+
+#############################################
+
+.PHONY	= all clean
+
+all: ${TARGET}
+
+${TARGET}: ${OBJECTS}
+	ar rcs $@ $^
+
+clean:
+	rm -rf ${OBJECTS} ${TARGET}
+
-- 
cgit v1.2.3


From 1c893fd513f5335a81dd72db70d64763634ea856 Mon Sep 17 00:00:00 2001
From: Younes Manton <younes.m@gmail.com>
Date: Sun, 29 Jun 2008 20:52:58 -0400
Subject: g3dvl: Simplify shader code.

---
 src/gallium/state_trackers/g3dvl/Makefile          |    2 +-
 src/gallium/state_trackers/g3dvl/vl_context.c      | 2968 ++++----------------
 src/gallium/state_trackers/g3dvl/vl_shader_build.c |  205 ++
 src/gallium/state_trackers/g3dvl/vl_shader_build.h |   61 +
 4 files changed, 890 insertions(+), 2346 deletions(-)
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_shader_build.c
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_shader_build.h

(limited to 'src/gallium/state_trackers/g3dvl/Makefile')

diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
index a0d85fbcc8..bd46d004e2 100644
--- a/src/gallium/state_trackers/g3dvl/Makefile
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -1,5 +1,5 @@
 TARGET		= libg3dvl.a
-OBJECTS		= vl_context.o vl_data.o vl_surface.o
+OBJECTS		= vl_context.o vl_data.o vl_surface.o vl_shader_build.o
 GALLIUMDIR	= ../..
 
 CFLAGS		+= -g -Wall -fPIC -Werror -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c
index d2b1ad7948..88da47c06a 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.c
+++ b/src/gallium/state_trackers/g3dvl/vl_context.c
@@ -9,6 +9,7 @@
 #include <pipe/p_shader_tokens.h>
 #include <tgsi/util/tgsi_parse.h>
 #include <tgsi/util/tgsi_build.h>
+#include "vl_shader_build.h"
 #include "vl_data.h"
 
 static int vlInitIDCT(struct VL_CONTEXT *context)
@@ -32,21 +33,11 @@ static int vlDestroyIDCT(struct VL_CONTEXT *context)
 static int vlCreateVertexShaderIMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 50;
-	const unsigned int		num_attribs = 3;
-	const unsigned int		semantic_names[3] =
-					{
-						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,	/* Luma texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Chroma texcoords */
-					};
-	const unsigned int		semantic_indexes[3] = {0, 1, 2};
-	const unsigned int		proc_type = TGSI_PROCESSOR_VERTEX;
 	
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	vs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 	
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -57,158 +48,71 @@ static int vlCreateVertexShaderIMC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-	
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
 	ti = 3;
 
-	/* Declare inputs (pos, texcoords) */
-	for (i = 0; i < num_attribs; i++)
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = semantic_names[i];
-		decl.Semantic.SemanticIndex = semantic_indexes[i];
-
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* Declare constant inputs */
-	/* C[0] scales the normalized MB to cover 16x16 pixels,
-	   C[1] translates the macroblock into position on the surface */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 1;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
-	/* Declare outputs (pos, texcoords) */
-	for (i = 0; i < num_attribs; i++)
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_OUTPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = semantic_names[i];
-		decl.Semantic.SemanticIndex = semantic_indexes[i];
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul t0, i0, c0	; Scale normalized coords to window coords */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* add o0, t0, c1	; Translate vertex into position */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add o0, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	/*
-	mov o1, i1		; Move texcoords to output
-	mov o2, i2
-	*/
-	for (i = 1; i < num_attribs; ++i)
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma texcoords to output
+	 */
+	for (i = 1; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	vs.tokens = tokens;
-	
 	context->states.mc.i_vs = pipe->create_vs_state(pipe, &vs);
-	
 	free(tokens);
 	
 	return 0;
@@ -217,13 +121,11 @@ static int vlCreateVertexShaderIMC(struct VL_CONTEXT *context)
 static int vlCreateFragmentShaderIMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 50;
-	const unsigned int		proc_type = TGSI_PROCESSOR_FRAGMENT;
-
+	
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	fs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -234,120 +136,61 @@ static int vlCreateFragmentShaderIMC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
 	ti = 3;
 
-	/* Declare inputs (texcoords) */
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 */
 	for (i = 0; i < 2; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-		decl.Semantic.SemanticIndex = i + 1;
-		decl.Declaration.Interpolate = 1;
-		decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* Declare output (color) */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_OUTPUT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 0;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
-	/* Declare samplers */
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 */
 	for (i = 0; i < 3; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_SAMPLER;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header,max_tokens - ti);
 	}
 	
 	/*
-	tex2d o0.x, i0, s0		; Read texel from luma texture into .x channel
-	tex2d o0.y, i1, s1		; Read texel from chroma Cb texture into .y channel
-	tex2d o0.z, i1, s2		; Read texel from chroma Cr texture into .z channel
+	 * tex2d o0.x, i0, s0		; Read texel from luma texture into .x channel
+	 * tex2d o0.y, i1, s1		; Read texel from chroma Cb texture into .y channel
+	 * tex2d o0.z, i1, s2		; Read texel from chroma Cr texture into .z channel
 	*/
 	for (i = 0; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = 0;
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_OUTPUT, 0, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i > 0 ? 1 : 0;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
 	fs.tokens = tokens;
-	
 	context->states.mc.i_fs = pipe->create_fs_state(pipe, &fs);
-	
 	free(tokens);
 	
 	return 0;
@@ -356,30 +199,11 @@ static int vlCreateFragmentShaderIMC(struct VL_CONTEXT *context)
 static int vlCreateVertexShaderFramePMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 100;
-	const unsigned int		num_input_attribs = 3;
-	const unsigned int		num_output_attribs = 4;
-	const unsigned int		input_semantic_names[3] =
-					{
-						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,	/* Luma texcoords */
-						TGSI_SEMANTIC_GENERIC	/* Chroma texcoords */
-					};
-	const unsigned int		output_semantic_names[4] =
-					{
-						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,	/* Luma texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Chroma texcoords */
-						TGSI_SEMANTIC_GENERIC	/* Ref surface texcoords */
-					};
-	const unsigned int		input_semantic_indexes[3] = {0, 1, 2};
-	const unsigned int		output_semantic_indexes[4] = {0, 1, 2, 3};
-	const unsigned int		proc_type = TGSI_PROCESSOR_VERTEX;
 	
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	vs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 	
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -390,179 +214,78 @@ static int vlCreateVertexShaderFramePMC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-	
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
 	ti = 3;
 
-	/* Declare inputs (pos, texcoords) */
-	for (i = 0; i < num_input_attribs; i++)
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = input_semantic_names[i];
-		decl.Semantic.SemanticIndex = input_semantic_indexes[i];
-
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* Declare constant inputs */
-	/* C[0] scales the normalized MB to cover 16x16 pixels,
-	   C[1] translates the macroblock into position on the surface
-	   C[2] unused
-	   C[3] translates the ref surface texcoords to the ref macroblock */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 3;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 * decl c2		; Unused
+	 * decl c3		; Translation vector to move ref macroblock texcoords into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 3);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
-	/* Declare outputs (pos, texcoords) */
-	for (i = 0; i < num_output_attribs; i++)
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 * decl o3		; Ref macroblock texcoords
+	 */
+	for (i = 0; i < 4; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_OUTPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = output_semantic_names[i];
-		decl.Semantic.SemanticIndex = output_semantic_indexes[i];
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul t0, i0, c0	; Scale normalized coords to window coords */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* add o0, t0, c1	; Translate vertex into position */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add o0, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	/*
-	mov o1, i1		; Move luma & chroma texcoords to output
-	mov o2, i2
-	*/
-	for (i = 1; i < num_output_attribs - 1; ++i)
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma texcoords to output
+	 */
+	for (i = 1; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* add o3, t0, c3	; Translate texcoords into position */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 3;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 3;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add o3, t0, c3	; Translate rect into position on ref macroblock */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	vs.tokens = tokens;
-	
 	context->states.mc.p_vs[0] = pipe->create_vs_state(pipe, &vs);
-	
 	free(tokens);
 	
 	return 0;
@@ -571,32 +294,11 @@ static int vlCreateVertexShaderFramePMC(struct VL_CONTEXT *context)
 static int vlCreateVertexShaderFieldPMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 100;
-	const unsigned int		num_input_attribs = 3;
-	const unsigned int		num_output_attribs = 6;
-	const unsigned int		input_semantic_names[3] =
-					{
-						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,	/* Luma texcoords */
-						TGSI_SEMANTIC_GENERIC	/* Chroma texcoords */
-					};
-	const unsigned int		output_semantic_names[6] =
-					{
-						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,	/* Luma texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Chroma texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Top field surface texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Bottom field surface texcoords */
-						TGSI_SEMANTIC_POSITION	/* Pos */
-					};
-	const unsigned int		input_semantic_indexes[3] = {0, 1, 2};
-	const unsigned int		output_semantic_indexes[6] = {0, 1, 2, 3, 4, 5};
-	const unsigned int		proc_type = TGSI_PROCESSOR_VERTEX;
 	
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	vs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 	
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -607,34 +309,26 @@ static int vlCreateVertexShaderFieldPMC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-	
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
 	ti = 3;
 
-	/* Declare inputs (pos, texcoords) */
-	for (i = 0; i < num_input_attribs; i++)
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = input_semantic_names[i];
-		decl.Semantic.SemanticIndex = input_semantic_indexes[i];
-
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
 		ti += tgsi_build_full_declaration
 		(
 			&decl,
@@ -644,183 +338,70 @@ static int vlCreateVertexShaderFieldPMC(struct VL_CONTEXT *context)
 		);
 	}
 	
-	/* Declare constant inputs */
-	/* C[0] scales the normalized MB to cover 16x16 pixels,
-	   C[1] translates the macroblock into position on the surface
-	   C[2] denormalizes pos components
-	   C[3] translates the ref surface top field texcoords to the ref macroblock
-	   C[4] translates the ref surface bottom field texcoords to the ref macroblock */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 4;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 * decl c2		; Denorm coefficients
+	 * decl c3		; Translation vector to move top field ref macroblock texcoords into position
+	 * decl c4		; Translation vector to move bottom field ref macroblock texcoords into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
-	/* Declare outputs (pos, texcoords) */
-	for (i = 0; i < num_output_attribs; i++)
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 * decl o3		; Top field ref macroblock texcoords
+	 * decl o4		; Bottom field ref macroblock texcoords
+	 * decl o5		; Denormalized vertex pos
+	 */
+	for (i = 0; i < 6; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_OUTPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = output_semantic_names[i];
-		decl.Semantic.SemanticIndex = output_semantic_indexes[i];
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_output((i == 0 || i == 5) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul t0, i0, c0	; Scale normalized coords to window coords */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* add t1, t0, c1	; Translate vertex into position */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add t1, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	/* mov o0, t1		; Move vertex pos to output */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 1;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	/*
-	mov o1, i1		; Move luma & chroma texcoords to output
-	mov o2, i2
+	mov o1, i1		; Move input luma texcoords to output
+	mov o2, i2		; Move input chroma texcoords to output
 	*/
-	for (i = 1; i < num_output_attribs - 1; ++i)
+	for (i = 1; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* add o3, t0, c3	; Translate top field texcoords into position
-	   add o4, t0, c4	; Translate bottom field texcoords into position */
+	/* add o3, t0, c3	; Translate top field rect into position on ref macroblock
+	   add o4, t0, c4	; Translate bottom field rect into position on ref macroblock */
 	for (i = 0; i < 2; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = i + 3;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i + 3;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul o5, t1, c2	; Denorm pos for fragment shader */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 5;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 2;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul o5, t1, c2	; Denorm vertex pos */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 5, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	vs.tokens = tokens;
-	
 	context->states.mc.p_vs[1] = pipe->create_vs_state(pipe, &vs);
-	
 	free(tokens);
 	
 	return 0;
@@ -829,13 +410,11 @@ static int vlCreateVertexShaderFieldPMC(struct VL_CONTEXT *context)
 static int vlCreateFragmentShaderFramePMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 100;
-	const unsigned int		proc_type = TGSI_PROCESSOR_FRAGMENT;
 
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	fs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -846,258 +425,101 @@ static int vlCreateFragmentShaderFramePMC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
 	ti = 3;
 
-	/* Declare inputs (texcoords) */
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 * decl i2			; Texcoords for s3
+	 */
 	for (i = 0; i < 3; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-		decl.Semantic.SemanticIndex = i + 1;
-		decl.Declaration.Interpolate = 1;
-		decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* Declare constant input */
-	/* C[0] is a multiplier to use when concatenating differential into a single channel
-	   C[0] is a bias to get differential back to -1,1 range*/
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 1;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0			; Multiplier to shift 9th bit of differential into place
+	 * decl c1			; Bias to get differential back to a signed value
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
-	/* Declare output */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_OUTPUT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 0;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
-	/* Declare samplers */
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for ref surface texture
+	 */
 	for (i = 0; i < 4; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_SAMPLER;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
 	/*
-	tex2d t0.xw, i0, s0	; Read texel from luma texture into .x and .w channels
-	mov t1.x, t0.w		; Move high part from .w channel to .x
-	tex2d t0.yw, i1, s1	; Read texel from chroma Cb texture into .y and .w channels
-	mov t1.y, t0.w		; Move high part from .w channel to .y
-	tex2d t0.zw, i1, s2	; Read texel from chroma Cr texture into .z and .w channels
-	mov t1.z, t0.w		; Move high part from .w channel to .z
-	*/
+	 * tex2d t0.xw, i0, s0		; Read texel from luma texture into .x and .w channels
+	 * mov t1.x, t0.w		; Move 9th bit from .w channel to .x
+	 * tex2d t0.yw, i1, s1		; Read texel from chroma Cb texture into .y and .w channels
+	 * mov t1.y, t0.w		; Move 9th bit from .w channel to .y
+	 * tex2d t0.zw, i1, s2		; Read texel from chroma Cr texture into .z and .w channels
+	 * mov t1.z, t0.w		; Move 9th bit from .w channel to .z
+	 */
 	for (i = 0; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = 0;
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = (TGSI_WRITEMASK_X << i) | TGSI_WRITEMASK_W;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i > 0 ? 1 : 0;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 		
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = 1;
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullSrcRegisters[0].SrcRegister.Index = 0;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul t1, t1, c0	; Muliply high part by multiplier to get back its full value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul t1, t1, c0		; Muliply 9th bit by multiplier to shift it into place */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* add t0, t0, t1	; Add luma and chroma low and high parts to get a single value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add t0, t0, t1		; Add luma and chroma low and high parts to get normalized unsigned 9-bit values */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* sub t0, t0, c1	; Subtract bias to get back the signed value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_SUB;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* sub t0, t0, c1		; Subtract bias to get back signed values */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* tex2d t1, i2, s3	; Read texel from ref macroblock */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 2;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 3;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* tex2d t1, i2, s3		; Read texel from ref macroblock */
+	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 2, TGSI_FILE_SAMPLER, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* add o0, t0, t1	; Add ref and differential to form final output */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add o0, t0, t1		; Add ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
 	fs.tokens = tokens;
-	
 	context->states.mc.p_fs[0] = pipe->create_fs_state(pipe, &fs);
-	
 	free(tokens);
 	
 	return 0;
@@ -1106,13 +528,11 @@ static int vlCreateFragmentShaderFramePMC(struct VL_CONTEXT *context)
 static int vlCreateFragmentShaderFieldPMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 200;
-	const unsigned int		proc_type = TGSI_PROCESSOR_FRAGMENT;
 
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	fs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -1123,399 +543,150 @@ static int vlCreateFragmentShaderFieldPMC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
 	ti = 3;
 
-	/* Declare inputs (texcoords) 
-	   I[0] Luma texcoords
-	   I[1] Chroma texcoords
-	   I[2] Ref top field surface texcoords
-	   I[3] Ref bottom field surface texcoords
-	   I[4] Denormalized texel pos */
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 * decl i2			; Texcoords for s3
+	 * decl i3			; Texcoords for s3
+	 * decl i4			; Denormalized vertex pos
+	 */
 	for (i = 0; i < 5; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-		decl.Semantic.SemanticIndex = i + 1;
-		decl.Declaration.Interpolate = 1;
-		decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* Declare constant input */
-	/* C[0] is a multiplier to use when concatenating differential into a single channel
-	   C[1] is a bias to get differential back to -1,1 range
-	   C[2] is constants 2 and 1/2 for Y%2 field selector */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 2;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0			; Multiplier to shift 9th bit of differential into place
+	 * decl c1			; Bias to get differential back to a signed value
+	 * decl c2			; Constants 1/2 & 2 in .x, .y channels for Y-mod-2 top/bottom field selection
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 2);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
-	/* Declare output */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_OUTPUT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 0;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
-	/* Declare samplers */
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for ref surface texture
+	 */
 	for (i = 0; i < 4; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_SAMPLER;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
 	/*
-	tex2d t0.xw, i0, s0	; Read texel from luma texture into .x and .w channels
-	mov t1.x, t0.w		; Move high part from .w channel to .x
-	tex2d t0.yw, i1, s1	; Read texel from chroma Cb texture into .y and .w channels
-	mov t1.y, t0.w		; Move high part from .w channel to .y
-	tex2d t0.zw, i1, s2	; Read texel from chroma Cr texture into .z and .w channels
-	mov t1.z, t0.w		; Move high part from .w channel to .z
-	*/
+	 * tex2d t0.xw, i0, s0		; Read texel from luma texture into .x and .w channels
+	 * mov t1.x, t0.w		; Move 9th bit from .w channel to .x
+	 * tex2d t0.yw, i1, s1		; Read texel from chroma Cb texture into .y and .w channels
+	 * mov t1.y, t0.w		; Move 9th bit from .w channel to .y
+	 * tex2d t0.zw, i1, s2		; Read texel from chroma Cr texture into .z and .w channels
+	 * mov t1.z, t0.w		; Move 9th bit from .w channel to .z
+	 */
 	for (i = 0; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = 0;
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = (TGSI_WRITEMASK_X << i) | TGSI_WRITEMASK_W;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i > 0 ? 1 : 0;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 		
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = 1;
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullSrcRegisters[0].SrcRegister.Index = 0;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul t1, t1, c0	; Muliply high part by multiplier to get back its full value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul t1, t1, c0		; Muliply 9th bit by multiplier to shift it into place */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* add t0, t0, t1	; Add luma and chroma low and high parts to get a single value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add t0, t0, t1		; Add luma and chroma low and high parts to get normalized unsigned 9-bit values */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* sub t0, t0, c1	; Subtract bias to get back the signed value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_SUB;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* sub t0, t0, c1		; Subtract bias to get back signed values */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* tex2d t1, i2, s3	; Read texel from ref macroblock top field
-	   tex2d t2, i3, s3	; Read texel from ref macroblock bottom field */
+	/* tex2d t1, i2, s3		; Read texel from ref macroblock top field
+	   tex2d t2, i3, s3		; Read texel from ref macroblock bottom field */
 	for (i = 0; i < 2; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = i + 1;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i + 2;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-		inst.FullSrcRegisters[1].SrcRegister.Index = 3;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* XXX: Pos values off by 0.5 for rounding? */
-	/* sub t4, i4.y, c2.x	; Sub 0.5 from position */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_SUB;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 4;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 4;
+	/* XXX: Pos values off by 0.5? */
+	/* sub t4, i4.y, c2.x		; Sub 0.5 from denormalized pos */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 4, TGSI_FILE_INPUT, 2);
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 2;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* mul t3, t4, c2.x	; Divide pos y coord by 2 (mul by 0.5) */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 3;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 4;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 2;
+	/* mul t3, t4, c2.x		; Multiply pos Y-coord by 1/2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 2);
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* floor t3, t3		; Get rid of fractional part */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_FLOOR;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 3;
-	inst.Instruction.NumSrcRegs = 1;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 3;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* floor t3, t3			; Get rid of fractional part */
+	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* mul t3, t3, c2.y	; Multiply by 2 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 3;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 3;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 2;
+	/* mul t3, t3, c2.y		; Multiply by 2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 2);
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* sub t3, t4, t3	; Subtract from y to get y % 2 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_SUB;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 3;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 4;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 3;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* lerp t1, t3, t1, t2 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_LERP;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 3;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 3;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[2].SrcRegister.Index = 2;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* add o0, t0, t1	; Add ref and differential to form final output */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add o0, t0, t1		; Add ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
 	fs.tokens = tokens;
-	
 	context->states.mc.p_fs[1] = pipe->create_fs_state(pipe, &fs);
-	
 	free(tokens);
 	
 	return 0;
@@ -1524,31 +695,11 @@ static int vlCreateFragmentShaderFieldPMC(struct VL_CONTEXT *context)
 static int vlCreateVertexShaderFrameBMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 100;
-	const unsigned int		num_input_attribs = 3;
-	const unsigned int		num_output_attribs = 5;
-	const unsigned int		input_semantic_names[3] =
-					{
-						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,
-						TGSI_SEMANTIC_GENERIC
-					};
-	const unsigned int		output_semantic_names[5] =
-					{
-						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,
-						TGSI_SEMANTIC_GENERIC,
-						TGSI_SEMANTIC_GENERIC,
-						TGSI_SEMANTIC_GENERIC
-					};
-	const unsigned int		input_semantic_indexes[3] = {0, 1, 2};
-	const unsigned int		output_semantic_indexes[5] = {0, 1, 2, 3, 4};
-	const unsigned int		proc_type = TGSI_PROCESSOR_VERTEX;
 	
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	vs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 	
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -1559,185 +710,85 @@ static int vlCreateVertexShaderFrameBMC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-	
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
 	ti = 3;
 
-	/* Declare inputs (pos, texcoords) */
-	for (i = 0; i < num_input_attribs; i++)
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = input_semantic_names[i];
-		decl.Semantic.SemanticIndex = input_semantic_indexes[i];
-
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* Declare constant inputs */
-	/* C[0] scales the normalized MB to cover 16x16 pixels,
-	   C[1] translates the macroblock into position on the surface
-	   C[2] unused
-	   C[3] translates the past surface texcoords to the ref macroblock
-	   C[4] unused
-	   C[5] translates the future surface texcoords to the ref macroblock */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 5;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 * decl c2		; Unused
+	 * decl c3		; Translation vector to move past ref macroblock texcoords into position
+	 * decl c4		; Unused
+	 * decl c5		; Translation vector to move future ref macroblock texcoords into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 5);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
-	/* Declare outputs (pos, texcoords) */
-	for (i = 0; i < num_output_attribs; i++)
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 * decl o3		; Past ref macroblock texcoords
+	 * decl o4		; Future ref macroblock texcoords
+	 */
+	for (i = 0; i < 5; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_OUTPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = output_semantic_names[i];
-		decl.Semantic.SemanticIndex = output_semantic_indexes[i];
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul t0, i0, c0	; Scale normalized coords to window coords */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
-
-	/* add o0, t0, c1	; Translate vertex into position */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	/*
-	mov o1, i1		; Move luma & chroma texcoords to output
-	mov o2, i2
-	*/
-	for (i = 1; i < num_output_attribs - 2; ++i)
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma texcoords to output
+	 */
+	for (i = 1; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* add o3, t0, c3	; Translate past surface texcoords into position
-	   add o4, t0, c5	; Repeat for future surface texcoords */
+	/* add o3, t0, c3	; Translate rect into position on past ref macroblock
+	   add o4, t0, c5	; Translate rect into position on future ref macroblock */
 	for (i = 0; i < 2; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = i + 3;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i * 2 + 3;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i * 2 + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	vs.tokens = tokens;
-	
 	context->states.mc.b_vs[0] = pipe->create_vs_state(pipe, &vs);
-	
 	free(tokens);
 	
 	return 0;
@@ -1746,34 +797,11 @@ static int vlCreateVertexShaderFrameBMC(struct VL_CONTEXT *context)
 static int vlCreateVertexShaderFieldBMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 100;
-	const unsigned int		num_input_attribs = 3;
-	const unsigned int		num_output_attribs = 8;
-	const unsigned int		input_semantic_names[3] =
-					{
-						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,	/* Luma texcoords */
-						TGSI_SEMANTIC_GENERIC	/* Chroma texcoords */
-					};
-	const unsigned int		output_semantic_names[8] =
-					{
-						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,	/* Luma texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Chroma texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Top field past surface texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Bottom field past surface texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Top field future surface texcoords */
-						TGSI_SEMANTIC_GENERIC,	/* Bottom field future surface texcoords */
-						TGSI_SEMANTIC_POSITION	/* Pos */
-					};
-	const unsigned int		input_semantic_indexes[3] = {0, 1, 2};
-	const unsigned int		output_semantic_indexes[8] = {0, 1, 2, 3, 4, 5, 6, 7};
-	const unsigned int		proc_type = TGSI_PROCESSOR_VERTEX;
 	
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	vs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 	
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -1783,225 +811,102 @@ static int vlCreateVertexShaderFieldBMC(struct VL_CONTEXT *context)
 	
 	assert(context);
 	
-	pipe = context->pipe;
-	
+	pipe = context->pipe;	
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
 	ti = 3;
 
-	/* Declare inputs (pos, texcoords) */
-	for (i = 0; i < num_input_attribs; i++)
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = input_semantic_names[i];
-		decl.Semantic.SemanticIndex = input_semantic_indexes[i];
-
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* Declare constant inputs */
-	/* C[0] scales the normalized MB to cover 16x16 pixels,
-	   C[1] translates the macroblock into position on the surface
-	   C[2] denormalizes pos components
-	   C[3] translates the past surface top field texcoords to the ref macroblock
-	   C[4] translates the past surface bottom field texcoords to the ref macroblock
-	   C[5] translates the future surface top field texcoords to the ref macroblock
-	   C[6] translates the future surface bottom field texcoords to the ref macroblock */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 6;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 * decl c2		; Denorm coefficients
+	 * decl c3		; Translation vector to move top field past ref macroblock texcoords into position
+	 * decl c4		; Translation vector to move bottom field past ref macroblock texcoords into position
+	 * decl c5		; Translation vector to move top field future ref macroblock texcoords into position
+	 * decl c6		; Translation vector to move bottom field future ref macroblock texcoords into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 6);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
-	/* Declare outputs (pos, texcoords) */
-	for (i = 0; i < num_output_attribs; i++)
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 * decl o3		; Top field past ref macroblock texcoords
+	 * decl o4		; Bottom field past ref macroblock texcoords
+	 * decl o5		; Top field future ref macroblock texcoords
+	 * decl o6		; Bottom field future ref macroblock texcoords
+	 * decl o7		; Denormalized vertex pos
+	 */
+	for (i = 0; i < 8; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_OUTPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = output_semantic_names[i];
-		decl.Semantic.SemanticIndex = output_semantic_indexes[i];
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_output((i == 0 || i == 7) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul t0, i0, c0	; Scale normalized coords to window coords */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* add t1, t0, c1	; Translate vertex into position */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add t1, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	/* mov o0, t1		; Move vertex pos to output */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 1;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	/*
-	mov o1, i1		; Move luma & chroma texcoords to output
-	mov o2, i2
-	*/
-	for (i = 1; i < num_output_attribs - 1; ++i)
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma texcoords to output
+	 */
+	for (i = 1; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* add o3, t0, c3	; Translate top field past texcoords into position
-	   add o4, t0, c4	; Translate bottom field past texcoords into position
-	   add o5, t0, c5	; Translate top field past texcoords into position
-	   add o6, t0, c6	; Translate bottom field past texcoords into position */
+	/*
+	 * add o3, t0, c3	; Translate top field rect into position on past ref macroblock
+	 * add o4, t0, c4	; Translate bottom field rect into position on past ref macroblock
+	 * add o5, t0, c5	; Translate top field rect into position on future ref macroblock
+	 * add o6, t0, c6	; Translate bottom field rect into position on future ref macroblock
+	 */
 	for (i = 0; i < 4; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = i + 3;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i + 3;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul o7, t1, c2	; Denorm pos for fragment shader */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 7;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 2;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul o7, t1, c2	; Denorm vertex pos */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 7, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	vs.tokens = tokens;
-	
 	context->states.mc.b_vs[1] = pipe->create_vs_state(pipe, &vs);
-	
 	free(tokens);
 	
 	return 0;
@@ -2010,13 +915,11 @@ static int vlCreateVertexShaderFieldBMC(struct VL_CONTEXT *context)
 static int vlCreateFragmentShaderFrameBMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 100;
-	const unsigned int		proc_type = TGSI_PROCESSOR_FRAGMENT;
 
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	fs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -2027,288 +930,118 @@ static int vlCreateFragmentShaderFrameBMC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
 	ti = 3;
 
-	/* Declare inputs (texcoords) */
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 * decl i2			; Texcoords for s3
+	 * decl i3			; Texcoords for s4
+	 */
 	for (i = 0; i < 4; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-		decl.Semantic.SemanticIndex = i + 1;
-		decl.Declaration.Interpolate = 1;
-		decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* Declare constant input */
-	/* C[0] is a multiplier to use when concatenating differential into a single channel
-	   C[1] is a bias to get differential back to -1,1 range
-	   C[2] contains 0.5 in channel X for use as a weight to blend past and future samples */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 2;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0			; Multiplier to shift 9th bit of differential into place
+	 * decl c1			; Bias to get differential back to a signed value
+	 * decl c2			; Constant 1/2 in .x channel to use as weight to blend past and future texels
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 2);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
-	/* Declare output */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_OUTPUT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 0;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
-	/* Declare samplers */
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for past ref surface texture
+	 * decl s4			; Sampler for future ref surface texture
+	 */
 	for (i = 0; i < 5; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_SAMPLER;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
 	/*
-	tex2d t0.xw, i0, s0	; Read texel from luma texture into .x and .w channels
-	mov t1.x, t0.w		; Move high part from .w channel to .x
-	tex2d t0.yw, i1, s1	; Read texel from chroma Cb texture into .y and .w channels
-	mov t1.y, t0.w		; Move high part from .w channel to .y
-	tex2d t0.zw, i1, s2	; Read texel from chroma Cr texture into .z and .w channels
-	mov t1.z, t0.w		; Move high part from .w channel to .z
-	*/
+	 * tex2d t0.xw, i0, s0		; Read texel from luma texture into .x and .w channels
+	 * mov t1.x, t0.w		; Move 9th bit from .w channel to .x
+	 * tex2d t0.yw, i1, s1		; Read texel from chroma Cb texture into .y and .w channels
+	 * mov t1.y, t0.w		; Move 9th bit from .w channel to .y
+	 * tex2d t0.zw, i1, s2		; Read texel from chroma Cr texture into .z and .w channels
+	 * mov t1.z, t0.w		; Move 9th bit from .w channel to .z
+	 */
 	for (i = 0; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = 0;
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = (TGSI_WRITEMASK_X << i) | TGSI_WRITEMASK_W;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i > 0 ? 1 : 0;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 		
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = 1;
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullSrcRegisters[0].SrcRegister.Index = 0;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul t1, t1, c0	; Muliply high part by multiplier to get back its full value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul t1, t1, c0		; Muliply 9th bit by multiplier to shift it into place */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* add t0, t0, t1	; Add luma and chroma low and high parts to get a single value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add t0, t0, t1		; Add luma and chroma low and high parts to get normalized unsigned 9-bit values */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* sub t0, t0, c1	; Subtract bias to get back the signed value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_SUB;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* sub t0, t0, c1		; Subtract bias to get back signed values */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* tex2d t1, i2, s3	; Read texel from past macroblock
-	   tex2d t2, i3, s4	; Read texel from future macroblock */
+	/*
+	 * tex2d t1, i2, s3		; Read texel from past ref macroblock
+	 * tex2d t2, i3, s4		; Read texel from future ref macroblock
+	 */
 	for (i = 0; i < 2; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = i + 1;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i + 2;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i + 3;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, i + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
 	/* lerp t1, c2.x, t1, t2	; Blend past and future texels */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_LERP;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 3;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 2;
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[2].SrcRegister.Index = 2;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* add o0, t0, t1	; Add ref and differential to form final output */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
 	fs.tokens = tokens;
-	
 	context->states.mc.b_fs[0] = pipe->create_fs_state(pipe, &fs);
-	
 	free(tokens);
 	
 	return 0;
@@ -2317,13 +1050,11 @@ static int vlCreateFragmentShaderFrameBMC(struct VL_CONTEXT *context)
 static int vlCreateFragmentShaderFieldBMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 200;
-	const unsigned int		proc_type = TGSI_PROCESSOR_FRAGMENT;
 
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	fs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -2334,471 +1065,179 @@ static int vlCreateFragmentShaderFieldBMC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
 	ti = 3;
 
-	/* Declare inputs (texcoords) 
-	   I[0] Luma texcoords
-	   I[1] Chroma texcoords
-	   I[2] Past top field surface texcoords
-	   I[3] Past bottom field surface texcoords
-	   I[4] Future top field surface texcoords
-	   I[5] Future bottom field surface texcoords
-	   I[6] Denormalized texel pos */
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 * decl i2			; Texcoords for s3
+	 * decl i3			; Texcoords for s3
+	 * decl i4			; Texcoords for s4
+	 * decl i5			; Texcoords for s4
+	 * decl i6			; Denormalized vertex pos
+	 */
 	for (i = 0; i < 7; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-		decl.Semantic.SemanticIndex = i + 1;
-		decl.Declaration.Interpolate = 1;
-		decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* Declare constant input */
-	/* C[0] is a multiplier to use when concatenating differential into a single channel
-	   C[1] is a bias to get differential back to -1,1 range
-	   C[2] is constants 2 and 1/2 for Y%2 field selector */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 2;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
-
-	/* Declare output */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_OUTPUT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 0;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0			; Multiplier to shift 9th bit of differential into place
+	 * decl c1			; Bias to get differential back to a signed value
+	 * decl c2			; Constants 1/2 & 2 in .x, .y channels to use as weight to blend past and future texels
+	 *				; and for Y-mod-2 top/bottom field selection
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 2);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
-	/* Declare samplers */
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for past ref surface texture
+	 * decl s4			; Sampler for future ref surface texture
+	 */
 	for (i = 0; i < 5; ++i)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_SAMPLER;
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
 	/*
-	tex2d t0.xw, i0, s0	; Read texel from luma texture into .x and .w channels
-	mov t1.x, t0.w		; Move high part from .w channel to .x
-	tex2d t0.yw, i1, s1	; Read texel from chroma Cb texture into .y and .w channels
-	mov t1.y, t0.w		; Move high part from .w channel to .y
-	tex2d t0.zw, i1, s2	; Read texel from chroma Cr texture into .z and .w channels
-	mov t1.z, t0.w		; Move high part from .w channel to .z
-	*/
+	 * tex2d t0.xw, i0, s0		; Read texel from luma texture into .x and .w channels
+	 * mov t1.x, t0.w		; Move 9th bit from .w channel to .x
+	 * tex2d t0.yw, i1, s1		; Read texel from chroma Cb texture into .y and .w channels
+	 * mov t1.y, t0.w		; Move 9th bit from .w channel to .y
+	 * tex2d t0.zw, i1, s2		; Read texel from chroma Cr texture into .z and .w channels
+	 * mov t1.z, t0.w		; Move 9th bit from .w channel to .z
+	 */
 	for (i = 0; i < 3; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = 0;
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = (TGSI_WRITEMASK_X << i) | TGSI_WRITEMASK_W;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i > 0 ? 1 : 0;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 		
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = 1;
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullSrcRegisters[0].SrcRegister.Index = 0;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_W;
 		inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_W;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* mul t1, t1, c0	; Muliply high part by multiplier to get back its full value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* mul t1, t1, c0		; Muliply 9th bit by multiplier to shift it into place */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* add t0, t0, t1	; Add luma and chroma low and high parts to get a single value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add t0, t0, t1		; Add luma and chroma low and high parts to get normalized unsigned 9-bit values */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* sub t0, t0, c1	; Subtract bias to get back the signed value */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_SUB;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* sub t0, t0, c1		; Subtract bias to get back signed values */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* XXX: Pos values off by 0.5 for rounding? */
-	/* sub t4, i6.y, c2.x	; Sub 0.5 from position */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_SUB;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 4;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 6;
+	/* XXX: Pos values off by 0.5? */
+	/* sub t4, i6.y, c2.x		; Sub 0.5 from denormalized pos */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 6, TGSI_FILE_CONSTANT, 2);
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 2;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* mul t3, t4, c2.x	; Divide pos y coord by 2 (mul by 0.5) */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 3;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 4;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 2;
+	/* mul t3, t4, c2.x		; Multiply pos Y-coord by 1/2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 2);
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* floor t3, t3		; Get rid of fractional part */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_FLOOR;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 3;
-	inst.Instruction.NumSrcRegs = 1;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 3;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* floor t3, t3			; Get rid of fractional part */
+	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* mul t3, t3, c2.y	; Multiply by 2 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_MUL;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 3;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 3;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 2;
+	/* mul t3, t3, c2.y		; Multiply by 2 */
+	inst = vl_inst3( TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 2);
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
 	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* sub t3, t4, t3	; Subtract from y to get y % 2 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_SUB;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 3;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 4;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 3;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* tex2d t1, i2, s3	; Read texel from past macroblock top field
-	   tex2d t2, i3, s3	; Read texel from past macroblock bottom field */
+	/*
+	 * tex2d t1, i2, s3		; Read texel from past ref macroblock top field
+	 * tex2d t2, i3, s3		; Read texel from past ref macroblock bottom field
+	 */
 	for (i = 0; i < 2; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = i + 1;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i + 2;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-		inst.FullSrcRegisters[1].SrcRegister.Index = 3;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* lerp t1, t3, t1, t2 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_LERP;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 3;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 3;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[2].SrcRegister.Index = 2;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* tex2d t4, i4, s4	; Read texel from future macroblock top field
-	   tex2d t5, i5, s4	; Read texel from future macroblock bottom field */
+	/*
+	 * tex2d t4, i4, s4		; Read texel from future ref macroblock top field
+	 * tex2d t5, i5, s4		; Read texel from future ref macroblock bottom field
+	 */
 	for (i = 0; i < 2; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullDstRegisters[0].DstRegister.Index = i + 4;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i + 4;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-		inst.FullSrcRegisters[1].SrcRegister.Index = 4;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 4, TGSI_FILE_INPUT, i + 4, TGSI_FILE_SAMPLER, 4);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 	
-	/* lerp t2, t3, t4, t5 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_LERP;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 2;
-	inst.Instruction.NumSrcRegs = 3;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 3;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 4;
-	inst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[2].SrcRegister.Index = 5;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t2, t3, t4, t5		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 2, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 5);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* lerp t1, c2.x, t1, t2 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_LERP;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 1;
-	inst.Instruction.NumSrcRegs = 3;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 2;
+	/* lerp t1, c2.x, t1, t2	; Blend past and future texels */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
 	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	inst.FullSrcRegisters[2].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[2].SrcRegister.Index = 2;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* add o0, t0, t1	; Add future and differential to form final output */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/* END */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
 	fs.tokens = tokens;
-	
 	context->states.mc.b_fs[1] = pipe->create_fs_state(pipe, &fs);
-	
 	free(tokens);
 	
 	return 0;
@@ -2936,8 +1375,8 @@ static int vlInitMC(struct VL_CONTEXT *context)
 	filters[0] = PIPE_TEX_FILTER_NEAREST;
 	filters[1] = context->video_format == VL_FORMAT_YCBCR_444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
 	filters[2] = context->video_format == VL_FORMAT_YCBCR_444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
-	filters[3] = PIPE_TEX_FILTER_NEAREST;
-	filters[4] = PIPE_TEX_FILTER_NEAREST;
+	filters[3] = PIPE_TEX_FILTER_LINEAR;
+	filters[4] = PIPE_TEX_FILTER_LINEAR;
 	
 	for (i = 0; i < 5; ++i)
 	{
@@ -3037,16 +1476,11 @@ static int vlDestroyMC(struct VL_CONTEXT *context)
 static int vlCreateVertexShaderCSC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 50;
-	const unsigned int		num_attribs = 2;
-	const unsigned int		semantic_names[2] = {TGSI_SEMANTIC_POSITION, TGSI_SEMANTIC_GENERIC};
-	const unsigned int		semantic_indexes[2] = {0, 1};
-	const unsigned int		proc_type = TGSI_PROCESSOR_VERTEX;
 	
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	vs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 	
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -3057,101 +1491,54 @@ static int vlCreateVertexShaderCSC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-	
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
 
 	ti = 3;
 
-	/* Declare inputs (pos, texcoords) */
-	for (i = 0; i < num_attribs; i++)
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Vertex texcoords
+	 */
+	for (i = 0; i < 2; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_INPUT;
-
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = semantic_names[i];
-		decl.Semantic.SemanticIndex = semantic_indexes[i];
-
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* Declare outputs (pos, texcoords) */
-	for (i = 0; i < num_attribs; i++)
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Vertex texcoords
+	 */
+	for (i = 0; i < 2; i++)
 	{
-		decl = tgsi_default_full_declaration();
-		decl.Declaration.File = TGSI_FILE_OUTPUT;
-		decl.Declaration.Semantic = 1;
-		decl.Semantic.SemanticName = semantic_names[i];
-		decl.Semantic.SemanticIndex = semantic_indexes[i];
-		decl.u.DeclarationRange.First = i;
-		decl.u.DeclarationRange.Last = i;
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* MOV instructions */
-	/* mov o0, i0
-	   mov o1, i1 */
-	for (i = 0; i < num_attribs; i++)
+	/*
+	 * mov o0, i0		; Move pos in to pos out
+	 * mov o1, i1		; Move input texcoords to output
+	 */
+	for (i = 0; i < 2; i++)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_MOV;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = i;
-		inst.Instruction.NumSrcRegs = 1;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-		inst.FullSrcRegisters[0].SrcRegister.Index = i;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* END instruction */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
 	vs.tokens = tokens;
-	
 	context->states.csc.vertex_shader = pipe->create_vs_state(pipe, &vs);
-	
 	free(tokens);
 	
 	return 0;
@@ -3160,13 +1547,11 @@ static int vlCreateVertexShaderCSC(struct VL_CONTEXT *context)
 static int vlCreateFragmentShaderCSC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 50;
-	const unsigned int		proc_type = TGSI_PROCESSOR_FRAGMENT;
 
 	struct pipe_context		*pipe;
 	struct pipe_shader_state	fs;
 	struct tgsi_token		*tokens;
 	struct tgsi_header		*header;
-	struct tgsi_processor		*processor;
 
 	struct tgsi_full_declaration	decl;
 	struct tgsi_full_instruction	inst;
@@ -3177,171 +1562,64 @@ static int vlCreateFragmentShaderCSC(struct VL_CONTEXT *context)
 	assert(context);
 	
 	pipe = context->pipe;
-
 	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	
 	/* Header */
 	header = (struct tgsi_header*)&tokens[1];
 	*header = tgsi_build_header();
-
 	/* Processor */
-	processor = (struct tgsi_processor*)&tokens[2];
-	*processor = tgsi_build_processor(proc_type, header);
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
 
 	ti = 3;
 
-	/* Declare TEX[0] input */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_INPUT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 1;
-	decl.Declaration.Interpolate = 1;
-	decl.Interpolation.Interpolate = TGSI_INTERPOLATE_LINEAR;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 0;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* decl i0		; Texcoords for s0 */
+	decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, 1, 0, 0, TGSI_INTERPOLATE_LINEAR);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
-	/* Declare constant input */
-	/* Constants include bias vector, 4x4 csc matrix, total 5 vectors */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_CONSTANT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_GENERIC;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 4;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/*
+	 * decl c0		; Bias vector for CSC
+	 * decl c1-c4		; CSC matrix c1-c4
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
-	/* Declare output */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_OUTPUT;
-	decl.Declaration.Semantic = 1;
-	decl.Semantic.SemanticName = TGSI_SEMANTIC_COLOR;
-	decl.Semantic.SemanticIndex = 0;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 0;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* decl o0		; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
-	/* Declare sampler */
-	decl = tgsi_default_full_declaration();
-	decl.Declaration.File = TGSI_FILE_SAMPLER;
-	decl.u.DeclarationRange.First = 0;
-	decl.u.DeclarationRange.Last = 0;
-	ti += tgsi_build_full_declaration
-	(
-		&decl,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* decl s0		; Sampler for tex containing picture to display */
+	decl = vl_decl_samplers(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
-	/* TEX instruction */
-	/* tex2d t0, i0, s0 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_TEX;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.InstructionExtTexture.Texture = TGSI_TEXTURE_2D;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_INPUT;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_SAMPLER;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* tex2d t0, i0, s0	; Read src pixel */
+	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* SUB instruction */
-	/* sub t0, t0, c0 */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_SUB;
-	inst.Instruction.NumDstRegs = 1;
-	inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullDstRegisters[0].DstRegister.Index = 0;
-	inst.Instruction.NumSrcRegs = 2;
-	inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-	inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-	inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-	inst.FullSrcRegisters[1].SrcRegister.Index = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* sub t0, t0, c0	; Subtract bias vector from pixel */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	
-	/* DP4 instruction */
-	/* dp4 o0.x, t0, c1
-	   dp4 o0.y, t0, c2
-	   dp4 o0.z, t0, c3
-	   dp4 o0.w, t0, c4 */
+	/*
+	 * dp4 o0.x, t0, c1	; Multiply pixel by the color conversion matrix
+	 * dp4 o0.y, t0, c2
+	 * dp4 o0.z, t0, c3
+	 * dp4 o0.w, t0, c4	; XXX: Don't need 4th coefficient
+	 */
 	for (i = 0; i < 4; ++i)
 	{
-		inst = tgsi_default_full_instruction();
-		inst.Instruction.Opcode = TGSI_OPCODE_DP4;
-		inst.Instruction.NumDstRegs = 1;
-		inst.FullDstRegisters[0].DstRegister.File = TGSI_FILE_OUTPUT;
-		inst.FullDstRegisters[0].DstRegister.Index = 0;
+		inst = vl_inst3(TGSI_OPCODE_DP4, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 1);
 		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		inst.Instruction.NumSrcRegs = 2;
-		inst.FullSrcRegisters[0].SrcRegister.File = TGSI_FILE_TEMPORARY;
-		inst.FullSrcRegisters[0].SrcRegister.Index = 0;
-		inst.FullSrcRegisters[1].SrcRegister.File = TGSI_FILE_CONSTANT;
-		inst.FullSrcRegisters[1].SrcRegister.Index = i + 1;
-		ti += tgsi_build_full_instruction
-		(
-			&inst,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 	}
 
-	/* END instruction */
-	inst = tgsi_default_full_instruction();
-	inst.Instruction.Opcode = TGSI_OPCODE_END;
-	inst.Instruction.NumDstRegs = 0;
-	inst.Instruction.NumSrcRegs = 0;
-	ti += tgsi_build_full_instruction
-	(
-		&inst,
-		&tokens[ti],
-		header,
-		max_tokens - ti
-	);
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
 	fs.tokens = tokens;
-	
 	context->states.csc.fragment_shader = pipe->create_fs_state(pipe, &fs);
-	
 	free(tokens);
 	
 	return 0;
diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.c b/src/gallium/state_trackers/g3dvl/vl_shader_build.c
new file mode 100644
index 0000000000..fd9de7dab8
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.c
@@ -0,0 +1,205 @@
+#include "vl_shader_build.h"
+#include <assert.h>
+#include <tgsi/util/tgsi_parse.h>
+#include <tgsi/util/tgsi_build.h>
+
+struct tgsi_full_declaration vl_decl_input(unsigned int name, unsigned int index, unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	decl.Declaration.File = TGSI_FILE_INPUT;
+	decl.Declaration.Semantic = 1;
+	decl.Semantic.SemanticName = name;
+	decl.Semantic.SemanticIndex = index;
+	decl.u.DeclarationRange.First = first;
+	decl.u.DeclarationRange.Last = last;
+	
+	return decl;
+}
+
+struct tgsi_full_declaration vl_decl_interpolated_input
+(
+	unsigned int name,
+	unsigned int index,
+	unsigned int first,
+	unsigned int last,
+	int interpolation
+)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+	
+	assert
+	(
+		interpolation == TGSI_INTERPOLATE_CONSTANT ||
+		interpolation == TGSI_INTERPOLATE_LINEAR ||
+		interpolation == TGSI_INTERPOLATE_PERSPECTIVE
+	);
+
+	decl.Declaration.File = TGSI_FILE_INPUT;
+	decl.Declaration.Semantic = 1;
+	decl.Semantic.SemanticName = name;
+	decl.Semantic.SemanticIndex = index;
+	decl.Declaration.Interpolate = 1;
+	decl.Interpolation.Interpolate = interpolation;
+	decl.u.DeclarationRange.First = first;
+	decl.u.DeclarationRange.Last = last;
+	
+	return decl;
+}
+
+struct tgsi_full_declaration vl_decl_constants(unsigned int name, unsigned int index, unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+	
+	decl.Declaration.File = TGSI_FILE_CONSTANT;
+	decl.Declaration.Semantic = 1;
+	decl.Semantic.SemanticName = name;
+	decl.Semantic.SemanticIndex = index;
+	decl.u.DeclarationRange.First = first;
+	decl.u.DeclarationRange.Last = last;
+	
+	return decl;
+}
+
+struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int index, unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	decl.Declaration.File = TGSI_FILE_OUTPUT;
+	decl.Declaration.Semantic = 1;
+	decl.Semantic.SemanticName = name;
+	decl.Semantic.SemanticIndex = index;
+	decl.u.DeclarationRange.First = first;
+	decl.u.DeclarationRange.Last = last;
+	
+	return decl;
+}
+
+struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	decl = tgsi_default_full_declaration();
+	decl.Declaration.File = TGSI_FILE_SAMPLER;
+	decl.u.DeclarationRange.First = first;
+	decl.u.DeclarationRange.Last = last;
+	
+	return decl;
+}
+
+struct tgsi_full_instruction vl_inst2
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src_file,
+	unsigned int src_index
+)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+	
+	inst.Instruction.Opcode = opcode;
+	inst.Instruction.NumDstRegs = 1;
+	inst.FullDstRegisters[0].DstRegister.File = dst_file;
+	inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+	inst.Instruction.NumSrcRegs = 1;
+	inst.FullSrcRegisters[0].SrcRegister.File = src_file;
+	inst.FullSrcRegisters[0].SrcRegister.Index = src_index;
+	
+	return inst;
+}
+
+struct tgsi_full_instruction vl_inst3
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index
+)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+	
+	inst.Instruction.Opcode = opcode;
+	inst.Instruction.NumDstRegs = 1;
+	inst.FullDstRegisters[0].DstRegister.File = dst_file;
+	inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+	inst.Instruction.NumSrcRegs = 2;
+	inst.FullSrcRegisters[0].SrcRegister.File = src1_file;
+	inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
+	inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
+	inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
+	
+	return inst;
+}
+
+struct tgsi_full_instruction vl_tex
+(
+	int tex,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index
+)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+	
+	inst.Instruction.Opcode = TGSI_OPCODE_TEX;
+	inst.Instruction.NumDstRegs = 1;
+	inst.FullDstRegisters[0].DstRegister.File = dst_file;
+	inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+	inst.Instruction.NumSrcRegs = 2;
+	inst.InstructionExtTexture.Texture = tex;
+	inst.FullSrcRegisters[0].SrcRegister.File = src1_file;
+	inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
+	inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
+	inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
+	
+	return inst;
+}
+
+struct tgsi_full_instruction vl_inst4
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index,
+	enum tgsi_file_type src3_file,
+	unsigned int src3_index
+)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+	
+	inst.Instruction.Opcode = opcode;
+	inst.Instruction.NumDstRegs = 1;
+	inst.FullDstRegisters[0].DstRegister.File = dst_file;
+	inst.FullDstRegisters[0].DstRegister.Index = dst_index;
+	inst.Instruction.NumSrcRegs = 3;
+	inst.FullSrcRegisters[0].SrcRegister.File = src1_file;
+	inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
+	inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
+	inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
+	inst.FullSrcRegisters[2].SrcRegister.File = src3_file;
+	inst.FullSrcRegisters[2].SrcRegister.Index = src3_index;
+	
+	return inst;
+}
+
+struct tgsi_full_instruction vl_end(void)
+{
+	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
+	
+	inst.Instruction.Opcode = TGSI_OPCODE_END;
+	inst.Instruction.NumDstRegs = 0;
+	inst.Instruction.NumSrcRegs = 0;
+	
+	return inst;
+}
+
diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.h b/src/gallium/state_trackers/g3dvl/vl_shader_build.h
new file mode 100644
index 0000000000..9e64bbeeae
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.h
@@ -0,0 +1,61 @@
+#ifndef vl_shader_build_h
+#define vl_shader_build_h
+
+#include <pipe/p_shader_tokens.h>
+
+struct tgsi_full_declaration vl_decl_input(unsigned int name, unsigned int index, unsigned int first, unsigned int last);
+struct tgsi_full_declaration vl_decl_interpolated_input
+(
+	unsigned int name,
+	unsigned int index,
+	unsigned int first,
+	unsigned int last,
+	int interpolation
+);
+struct tgsi_full_declaration vl_decl_constants(unsigned int name, unsigned int index, unsigned int first, unsigned int last);
+struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int index, unsigned int first, unsigned int last);
+struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int last);
+struct tgsi_full_instruction vl_inst2
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src_file,
+	unsigned int src_index
+);
+struct tgsi_full_instruction vl_inst3
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index
+);
+struct tgsi_full_instruction vl_tex
+(
+	int tex,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index
+);
+struct tgsi_full_instruction vl_inst4
+(
+	int opcode,
+	enum tgsi_file_type dst_file,
+	unsigned int dst_index,
+	enum tgsi_file_type src1_file,
+	unsigned int src1_index,
+	enum tgsi_file_type src2_file,
+	unsigned int src2_index,
+	enum tgsi_file_type src3_file,
+	unsigned int src3_index
+);
+struct tgsi_full_instruction vl_end(void);
+
+#endif
+
-- 
cgit v1.2.3


From 49937b99855984dd01a431c026f9308b6c0dac4f Mon Sep 17 00:00:00 2001
From: Younes Manton <younes.m@gmail.com>
Date: Thu, 3 Jul 2008 20:05:32 -0400
Subject: g3dvl: Round surfaces up to POT, use src rect when outputting
 surfaces.

---
 src/gallium/state_trackers/g3dvl/Makefile     |  2 +-
 src/gallium/state_trackers/g3dvl/vl_context.c | 48 +++++++++++++++++++++------
 src/gallium/state_trackers/g3dvl/vl_context.h |  2 +-
 src/gallium/state_trackers/g3dvl/vl_surface.c | 29 +++++++++++++---
 src/gallium/state_trackers/g3dvl/vl_types.h   |  6 ++++
 src/gallium/state_trackers/g3dvl/vl_util.c    | 17 ++++++++++
 src/gallium/state_trackers/g3dvl/vl_util.h    |  7 ++++
 7 files changed, 93 insertions(+), 18 deletions(-)
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_util.c
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_util.h

(limited to 'src/gallium/state_trackers/g3dvl/Makefile')

diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
index bd46d004e2..50e3c843b5 100644
--- a/src/gallium/state_trackers/g3dvl/Makefile
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -1,5 +1,5 @@
 TARGET		= libg3dvl.a
-OBJECTS		= vl_context.o vl_data.o vl_surface.o vl_shader_build.o
+OBJECTS		= vl_context.o vl_data.o vl_surface.o vl_shader_build.o vl_util.o
 GALLIUMDIR	= ../..
 
 CFLAGS		+= -g -Wall -fPIC -Werror -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c
index bd8743dc9a..58971bd7c7 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.c
+++ b/src/gallium/state_trackers/g3dvl/vl_context.c
@@ -11,6 +11,7 @@
 #include <tgsi/util/tgsi_build.h>
 #include "vl_shader_build.h"
 #include "vl_data.h"
+#include "vl_util.h"
 
 static int vlInitIDCT(struct VL_CONTEXT *context)
 {
@@ -1357,8 +1358,9 @@ static int vlInitMC(struct VL_CONTEXT *context)
 	
 	pipe = context->pipe;
 	
-	context->states.mc.viewport.scale[0] = context->video_width;
-	context->states.mc.viewport.scale[1] = context->video_height;
+	/* For MC we render to textures, which are rounded up to nearest POT */
+	context->states.mc.viewport.scale[0] = vlRoundUpPOT(context->video_width);
+	context->states.mc.viewport.scale[1] = vlRoundUpPOT(context->video_height);
 	context->states.mc.viewport.scale[2] = 1;
 	context->states.mc.viewport.scale[3] = 1;
 	context->states.mc.viewport.translate[0] = 0;
@@ -1512,6 +1514,13 @@ static int vlCreateVertexShaderCSC(struct VL_CONTEXT *context)
 		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
 		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
+	
+	/*
+	 * decl c0		; Scaling vector to scale texcoord rect to source size
+	 * decl c1		; Translation vector to move texcoord rect into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 
 	/*
 	 * decl o0		; Vertex pos
@@ -1522,16 +1531,18 @@ static int vlCreateVertexShaderCSC(struct VL_CONTEXT *context)
 		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
 		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
+	
+	/* mov o0, i0		; Move pos in to pos out */
+	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_INPUT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	
+	/* mul t0, i1, c0	; Scale unit texcoord rect to source size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 1, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
-	/*
-	 * mov o0, i0		; Move pos in to pos out
-	 * mov o1, i1		; Move input texcoords to output
-	 */
-	for (i = 0; i < 2; i++)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
+	/* add o1, t0, c1	; Translate texcoord rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
 
 	/* end */
 	inst = vl_end();
@@ -1693,6 +1704,19 @@ static int vlCreateDataBufsCSC(struct VL_CONTEXT *context)
 	context->states.csc.vertex_buf_elems[1].nr_components = 2;
 	context->states.csc.vertex_buf_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT;
 	
+	/*
+	Create our vertex shader's constant buffer
+	Const buffer contains scaling and translation vectors
+	*/
+	context->states.csc.vs_const_buf.size = sizeof(struct VL_CSC_VS_CONSTS);
+	context->states.csc.vs_const_buf.buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_CONSTANT,
+		context->states.csc.vs_const_buf.size
+	);
+	
 	/*
 	Create our fragment shader's constant buffer
 	Const buffer contains the color conversion matrix and bias vectors
@@ -1776,6 +1800,7 @@ static int vlDestroyCSC(struct VL_CONTEXT *context)
 	context->pipe->delete_fs_state(context->pipe, context->states.csc.fragment_shader);
 	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.csc.vertex_bufs[0].buffer);
 	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.csc.vertex_bufs[1].buffer);
+	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.csc.vs_const_buf.buffer);
 	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.csc.fs_const_buf.buffer);
 	
 	return 0;
@@ -1986,6 +2011,7 @@ int vlEndRender(struct VL_CONTEXT *context)
 	pipe->bind_fs_state(pipe, context->states.csc.fragment_shader);
 	pipe->set_vertex_buffers(pipe, 2, context->states.csc.vertex_bufs);
 	pipe->set_vertex_elements(pipe, 2, context->states.csc.vertex_buf_elems);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &context->states.csc.vs_const_buf);
 	pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &context->states.csc.fs_const_buf);
 	
 	return 0;
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.h b/src/gallium/state_trackers/g3dvl/vl_context.h
index 8a12318073..9ebda21a1c 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.h
+++ b/src/gallium/state_trackers/g3dvl/vl_context.h
@@ -49,7 +49,7 @@ struct VL_CONTEXT
 			struct pipe_shader_state		*vertex_shader, *fragment_shader;
 			struct pipe_vertex_buffer 		vertex_bufs[2];
 			struct pipe_vertex_element		vertex_buf_elems[2];
-			struct pipe_constant_buffer		fs_const_buf;
+			struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
 		} csc;
 	} states;
 };
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c
index 68313cc750..13f7301f07 100644
--- a/src/gallium/state_trackers/g3dvl/vl_surface.c
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.c
@@ -7,6 +7,7 @@
 #include <pipe/p_inlines.h>
 #include "vl_context.h"
 #include "vl_defs.h"
+#include "vl_util.h"
 
 static int vlGrabFrameCodedFullBlock(short *src, short *dst, unsigned int dst_pitch)
 {
@@ -194,9 +195,6 @@ static int vlGrabBlocks
 		pipe_surface_unmap(tex_surface);
 	}
 	
-	/* XXX: Texture cache is not invalidated when texture contents change */
-	context->pipe->flush(context->pipe, PIPE_FLUSH_TEXTURE_CACHE, NULL);
-	
 	return 0;
 }
 
@@ -214,8 +212,8 @@ int vlCreateSurface(struct VL_CONTEXT *context, struct VL_SURFACE **surface)
 	sfc = calloc(1, sizeof(struct VL_SURFACE));
 	
 	sfc->context = context;
-	sfc->width = context->video_width;
-	sfc->height = context->video_height;
+	sfc->width = vlRoundUpPOT(context->video_width);
+	sfc->height = vlRoundUpPOT(context->video_height);
 	sfc->format = context->video_format;
 	
 	memset(&template, 0, sizeof(struct pipe_texture));
@@ -227,6 +225,7 @@ int vlCreateSurface(struct VL_CONTEXT *context, struct VL_SURFACE **surface)
 	template.depth[0] = 1;
 	template.compressed = 0;
 	pf_get_block(template.format, &template.block);
+	/* XXX: Needed? */
 	template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_RENDER_TARGET;
 	
 	sfc->texture = pipe->screen->texture_create(pipe->screen, &template);
@@ -517,6 +516,7 @@ int vlPutSurface
 {
 	unsigned int		create_fb = 0;
 	struct pipe_context	*pipe;
+	struct VL_CSC_VS_CONSTS	*vs_consts;
 	
 	assert(surface);
 	
@@ -568,9 +568,28 @@ int vlPutSurface
 	
 	vlEndRender(surface->context);
 	
+	vs_consts = pipe->winsys->buffer_map
+	(
+		pipe->winsys,
+		surface->context->states.csc.vs_const_buf.buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE
+	);
+	
+	vs_consts->src_scale.x = srcw / (float)surface->width;
+	vs_consts->src_scale.y = srch / (float)surface->height;
+	vs_consts->src_scale.z = 1;
+	vs_consts->src_scale.w = 1;
+	vs_consts->src_trans.x = srcx / (float)surface->width;
+	vs_consts->src_trans.y = srcy / (float)surface->height;
+	vs_consts->src_trans.z = 0;
+	vs_consts->src_trans.w = 0;
+	
+	pipe->winsys->buffer_unmap(pipe->winsys, surface->context->states.csc.vs_const_buf.buffer);
+	
 	pipe->set_sampler_textures(pipe, 1, &surface->texture);
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_STRIP, 0, 4);
 	pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+	/* XXX: Need to take destx, desty into consideration */
 	pipe->winsys->flush_frontbuffer
 	(
 		pipe->winsys,
diff --git a/src/gallium/state_trackers/g3dvl/vl_types.h b/src/gallium/state_trackers/g3dvl/vl_types.h
index 97753699db..4d210c9e0a 100644
--- a/src/gallium/state_trackers/g3dvl/vl_types.h
+++ b/src/gallium/state_trackers/g3dvl/vl_types.h
@@ -75,6 +75,12 @@ struct VL_MC_FS_CONSTS
 	struct VL_VERTEX4F	y_divider;
 };
 
+struct VL_CSC_VS_CONSTS
+{
+	struct VL_VERTEX4F	src_scale;
+	struct VL_VERTEX4F	src_trans;
+};
+
 struct VL_CSC_FS_CONSTS
 {
 	struct VL_VERTEX4F	bias;
diff --git a/src/gallium/state_trackers/g3dvl/vl_util.c b/src/gallium/state_trackers/g3dvl/vl_util.c
new file mode 100644
index 0000000000..2421ae2210
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_util.c
@@ -0,0 +1,17 @@
+#include "vl_util.h"
+#include <assert.h>
+
+unsigned int vlRoundUpPOT(unsigned int x)
+{
+	unsigned int i;
+	
+	assert(x > 0);
+	
+	--x;
+	
+	for (i = 1; i < sizeof(unsigned int) * 8; i <<= 1)
+		x |= x >> i;
+	
+	return x + 1;
+}
+
diff --git a/src/gallium/state_trackers/g3dvl/vl_util.h b/src/gallium/state_trackers/g3dvl/vl_util.h
new file mode 100644
index 0000000000..e4b72c4f87
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_util.h
@@ -0,0 +1,7 @@
+#ifndef vl_util_h
+#define vl_util_h
+
+unsigned int vlRoundUpPOT(unsigned int x);
+
+#endif
+
-- 
cgit v1.2.3


From 0c25ac52425e6d6eb037b99ab90f41b47e3f4491 Mon Sep 17 00:00:00 2001
From: Younes Manton <younes.m@gmail.com>
Date: Tue, 22 Jul 2008 22:26:26 -0400
Subject: g3dvl: Add Nouveau winsys, libdriclient.

Nouveau winsys is based on Mesa's Nouveau winsys and soft-links to most of
it. The 'nouveau_context' and 'nouveau_screen' code contains most of the
changes, 'nouveau_winsys_pipe', 'nouveau_swapbuffers' and 'nouveau_lock'
contain some minor changes. The driclient library contains the DRI
userland stuff, most of which was based on Mesa's DRI code.
---
 src/driclient/include/driclient.h                  |  98 ++++
 src/driclient/include/xf86dri.h                    | 119 ++++
 src/driclient/src/Makefile                         |  22 +
 src/driclient/src/XF86dri.c                        | 619 +++++++++++++++++++++
 src/driclient/src/driclient.c                      | 292 ++++++++++
 src/driclient/src/test                             | Bin 0 -> 68389 bytes
 src/driclient/src/test.c                           |  41 ++
 src/driclient/src/xf86dri.h                        | 119 ++++
 src/driclient/src/xf86dristr.h                     | 342 ++++++++++++
 src/gallium/state_trackers/g3dvl/Makefile          |   2 +-
 src/gallium/state_trackers/g3dvl/vl_context.c      |  68 ++-
 src/gallium/state_trackers/g3dvl/vl_shader_build.c |  12 +
 src/gallium/state_trackers/g3dvl/vl_shader_build.h |   1 +
 src/gallium/state_trackers/g3dvl/vl_surface.c      |   4 +-
 src/gallium/winsys/g3dvl/nouveau/Makefile          |  46 ++
 src/gallium/winsys/g3dvl/nouveau/nouveau_bo.c      |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_channel.c |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_context.c | 371 ++++++++++++
 src/gallium/winsys/g3dvl/nouveau/nouveau_context.h | 105 ++++
 src/gallium/winsys/g3dvl/nouveau/nouveau_device.c  |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_dma.c     |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_dma.h     |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_dri.h     |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_drmif.h   |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_fence.c   |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_grobj.c   |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_local.h   |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_lock.c    |  92 +++
 .../winsys/g3dvl/nouveau/nouveau_notifier.c        |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_pushbuf.c |   1 +
 .../winsys/g3dvl/nouveau/nouveau_resource.c        |   1 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_screen.c  |  88 +++
 src/gallium/winsys/g3dvl/nouveau/nouveau_screen.h  |  24 +
 .../winsys/g3dvl/nouveau/nouveau_swapbuffers.c     |  64 +++
 .../winsys/g3dvl/nouveau/nouveau_swapbuffers.h     |  10 +
 src/gallium/winsys/g3dvl/nouveau/nouveau_winsys.c  |   1 +
 .../winsys/g3dvl/nouveau/nouveau_winsys_pipe.c     | 261 +++++++++
 .../winsys/g3dvl/nouveau/nouveau_winsys_pipe.h     |   1 +
 .../winsys/g3dvl/nouveau/nouveau_winsys_softpipe.c |   1 +
 src/gallium/winsys/g3dvl/nouveau/nv04_surface.c    |   1 +
 src/gallium/winsys/g3dvl/nouveau/nv50_surface.c    |   1 +
 src/gallium/winsys/g3dvl/vl_winsys.h               |  14 +
 src/gallium/winsys/g3dvl/xsp_winsys.c              |  72 ++-
 src/gallium/winsys/g3dvl/xsp_winsys.h              |  12 -
 src/libXvMC/Makefile                               |  19 +-
 src/libXvMC/context.c                              |   4 +-
 46 files changed, 2896 insertions(+), 43 deletions(-)
 create mode 100644 src/driclient/include/driclient.h
 create mode 100644 src/driclient/include/xf86dri.h
 create mode 100644 src/driclient/src/Makefile
 create mode 100644 src/driclient/src/XF86dri.c
 create mode 100644 src/driclient/src/driclient.c
 create mode 100755 src/driclient/src/test
 create mode 100644 src/driclient/src/test.c
 create mode 100644 src/driclient/src/xf86dri.h
 create mode 100644 src/driclient/src/xf86dristr.h
 create mode 100644 src/gallium/winsys/g3dvl/nouveau/Makefile
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_bo.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_channel.c
 create mode 100644 src/gallium/winsys/g3dvl/nouveau/nouveau_context.c
 create mode 100644 src/gallium/winsys/g3dvl/nouveau/nouveau_context.h
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_device.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_dma.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_dma.h
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_dri.h
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_drmif.h
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_fence.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_grobj.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_local.h
 create mode 100644 src/gallium/winsys/g3dvl/nouveau/nouveau_lock.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_notifier.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_pushbuf.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_resource.c
 create mode 100644 src/gallium/winsys/g3dvl/nouveau/nouveau_screen.c
 create mode 100644 src/gallium/winsys/g3dvl/nouveau/nouveau_screen.h
 create mode 100644 src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c
 create mode 100644 src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.h
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_winsys.c
 create mode 100644 src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_pipe.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_pipe.h
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_softpipe.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nv04_surface.c
 create mode 120000 src/gallium/winsys/g3dvl/nouveau/nv50_surface.c
 create mode 100644 src/gallium/winsys/g3dvl/vl_winsys.h
 delete mode 100644 src/gallium/winsys/g3dvl/xsp_winsys.h

(limited to 'src/gallium/state_trackers/g3dvl/Makefile')

diff --git a/src/driclient/include/driclient.h b/src/driclient/include/driclient.h
new file mode 100644
index 0000000000..36438a9f79
--- /dev/null
+++ b/src/driclient/include/driclient.h
@@ -0,0 +1,98 @@
+#ifndef driclient_h
+#define driclient_h
+
+#include <stdint.h>
+#include <X11/Xlib.h>
+#include <drm_sarea.h>
+#include "xf86dri.h"
+
+/* TODO: Bring in DRI XML options */
+
+typedef struct dri_version
+{
+	int major;
+	int minor;
+	int patch;
+} dri_version_t;
+
+typedef struct dri_screen
+{
+	Display			*display;
+	unsigned int		num;
+	dri_version_t		ddx, dri, drm;
+	int			draw_lock_id;
+	int			fd;
+	drm_sarea_t		*sarea;
+	void			*drawable_hash;
+	void			*private;
+} dri_screen_t;
+
+struct dri_context;
+
+typedef struct dri_drawable
+{
+	drm_drawable_t		drm_drawable;
+	Drawable		x_drawable;
+	unsigned int		sarea_index;
+	unsigned int		*sarea_stamp;
+	unsigned int		last_sarea_stamp;
+	int			x, y, w, h;
+	int			back_x, back_y;
+	int			num_cliprects, num_back_cliprects;
+	drm_clip_rect_t		*cliprects, *back_cliprects;
+	dri_screen_t		*dri_screen;
+	unsigned int		refcount;
+	void			*private;
+} dri_drawable_t;
+
+typedef struct dri_context
+{
+	XID			id;
+	drm_context_t		drm_context;
+	dri_screen_t		*dri_screen;
+	void			*private;
+} dri_context_t;
+
+typedef struct dri_framebuffer
+{
+	drm_handle_t		drm_handle;
+	int			base, size, stride;
+	int			private_size;
+	void			*private;
+} dri_framebuffer_t;
+
+int driCreateScreen(Display *display, int screen, dri_screen_t **dri_screen, dri_framebuffer_t *dri_framebuf);
+int driDestroyScreen(dri_screen_t *dri_screen);
+int driCreateDrawable(dri_screen_t *dri_screen, Drawable drawable, dri_drawable_t **dri_drawable);
+int driUpdateDrawableInfo(dri_drawable_t *dri_drawable);
+int driDestroyDrawable(dri_drawable_t *dri_drawable);
+int driCreateContext(dri_screen_t *dri_screen, Visual *visual, dri_context_t **dri_context);
+int driDestroyContext(dri_context_t *dri_context);
+int driCompareVersions(const dri_version_t *v1, const dri_version_t *v2);
+
+#define DRI_VALIDATE_DRAWABLE_INFO_ONCE(dri_drawable)					\
+do											\
+{											\
+	if (*(dri_drawable->sarea_stamp) != dri_drawable->last_sarea_stamp)		\
+		driUpdateDrawableInfo(dri_drawable);					\
+} while (0)
+
+#define DRI_VALIDATE_DRAWABLE_INFO(dri_screen, dri_drawable)					\
+do												\
+{												\
+	while (*(dri_drawable->sarea_stamp) != dri_drawable->last_sarea_stamp)			\
+	{											\
+		register unsigned int hwContext = dri_screen->sarea->lock.lock &		\
+		~(DRM_LOCK_HELD | DRM_LOCK_CONT);						\
+		DRM_UNLOCK(dri_screen->fd, &dri_screen->sarea->lock, hwContext);		\
+												\
+		DRM_SPINLOCK(&dri_screen->sarea->drawable_lock, dri_screen->draw_lock_id);	\
+		DRI_VALIDATE_DRAWABLE_INFO_ONCE(dri_drawable);					\
+		DRM_SPINUNLOCK(&dri_screen->sarea->drawable_lock, dri_screen->draw_lock_id);	\
+												\
+		DRM_LIGHT_LOCK(dri_screen->fd, &dri_screen->sarea->lock, hwContext);		\
+	}											\
+} while (0)
+
+#endif
+
diff --git a/src/driclient/include/xf86dri.h b/src/driclient/include/xf86dri.h
new file mode 100644
index 0000000000..baf80a7a9d
--- /dev/null
+++ b/src/driclient/include/xf86dri.h
@@ -0,0 +1,119 @@
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, 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, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file xf86dri.h
+ * Protocol numbers and function prototypes for DRI X protocol.
+ *
+ * \author Kevin E. Martin <martin@valinux.com>
+ * \author Jens Owen <jens@tungstengraphics.com>
+ * \author Rickard E. (Rik) Faith <faith@valinux.com>
+ */
+
+#ifndef _XF86DRI_H_
+#define _XF86DRI_H_
+
+#include <X11/Xfuncproto.h>
+#include <xf86drm.h>
+
+#define X_XF86DRIQueryVersion			0
+#define X_XF86DRIQueryDirectRenderingCapable	1
+#define X_XF86DRIOpenConnection			2
+#define X_XF86DRICloseConnection		3
+#define X_XF86DRIGetClientDriverName		4
+#define X_XF86DRICreateContext			5
+#define X_XF86DRIDestroyContext			6
+#define X_XF86DRICreateDrawable			7
+#define X_XF86DRIDestroyDrawable		8
+#define X_XF86DRIGetDrawableInfo		9
+#define X_XF86DRIGetDeviceInfo			10
+#define X_XF86DRIAuthConnection                 11
+#define X_XF86DRIOpenFullScreen                 12   /* Deprecated */
+#define X_XF86DRICloseFullScreen                13   /* Deprecated */
+
+#define XF86DRINumberEvents		0
+
+#define XF86DRIClientNotLocal		0
+#define XF86DRIOperationNotSupported	1
+#define XF86DRINumberErrors		(XF86DRIOperationNotSupported + 1)
+
+#ifndef _XF86DRI_SERVER_
+
+_XFUNCPROTOBEGIN
+
+Bool XF86DRIQueryExtension( Display *dpy, int *event_base, int *error_base );
+
+Bool XF86DRIQueryVersion( Display *dpy, int *majorVersion, int *minorVersion,
+    int *patchVersion );
+
+Bool XF86DRIQueryDirectRenderingCapable( Display *dpy, int screen,
+    Bool *isCapable );
+
+Bool XF86DRIOpenConnection( Display *dpy, int screen, drm_handle_t *hSAREA,
+    char **busIDString );
+
+Bool XF86DRIAuthConnection( Display *dpy, int screen, drm_magic_t magic );
+
+Bool XF86DRICloseConnection( Display *dpy, int screen );
+
+Bool XF86DRIGetClientDriverName( Display *dpy, int screen,
+    int *ddxDriverMajorVersion, int *ddxDriverMinorVersion,
+    int *ddxDriverPatchVersion, char **clientDriverName );
+
+Bool XF86DRICreateContext( Display *dpy, int screen, Visual *visual,
+    XID *ptr_to_returned_context_id, drm_context_t *hHWContext );
+
+Bool XF86DRICreateContextWithConfig( Display *dpy, int screen, int configID,
+    XID *ptr_to_returned_context_id, drm_context_t *hHWContext );
+
+Bool XF86DRIDestroyContext( Display *dpy, int screen,
+    XID context_id );
+
+Bool XF86DRICreateDrawable( Display *dpy, int screen,
+    Drawable drawable, drm_drawable_t *hHWDrawable );
+
+Bool XF86DRIDestroyDrawable( Display *dpy, int screen, 
+    Drawable drawable);
+
+Bool XF86DRIGetDrawableInfo( Display *dpy, int screen, Drawable drawable,
+    unsigned int *index, unsigned int *stamp, 
+    int *X, int *Y, int *W, int *H,
+    int *numClipRects, drm_clip_rect_t ** pClipRects,
+    int *backX, int *backY,
+    int *numBackClipRects, drm_clip_rect_t **pBackClipRects );
+
+Bool XF86DRIGetDeviceInfo( Display *dpy, int screen,
+    drm_handle_t *hFrameBuffer, int *fbOrigin, int *fbSize,
+    int *fbStride, int *devPrivateSize, void **pDevPrivate );
+
+_XFUNCPROTOEND
+
+#endif /* _XF86DRI_SERVER_ */
+
+#endif /* _XF86DRI_H_ */
+
diff --git a/src/driclient/src/Makefile b/src/driclient/src/Makefile
new file mode 100644
index 0000000000..0fac552de5
--- /dev/null
+++ b/src/driclient/src/Makefile
@@ -0,0 +1,22 @@
+TARGET		= libdriclient.a
+OBJECTS		= driclient.o XF86dri.o
+DRMDIR		?= /usr
+
+CFLAGS		+= -g -Wall -fPIC -Werror -I../include -I${DRMDIR}/include -I${DRMDIR}/include/drm
+
+#############################################
+
+.PHONY	= all clean
+
+all: ${TARGET} test
+
+${TARGET}: ${OBJECTS}
+	ar rcs $@ $^
+	cp ${TARGET} ../lib
+
+test: test.o
+	$(CC) -L../lib -L${DRMDIR}/lib ${LDFLAGS} -o $@ $^ -ldriclient -lX11 -lXext -ldrm
+
+clean:
+	rm -rf ${OBJECTS} ${TARGET} test test.o
+
diff --git a/src/driclient/src/XF86dri.c b/src/driclient/src/XF86dri.c
new file mode 100644
index 0000000000..9e359a9238
--- /dev/null
+++ b/src/driclient/src/XF86dri.c
@@ -0,0 +1,619 @@
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, 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, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <martin@valinux.com>
+ *   Jens Owen <jens@tungstengraphics.com>
+ *   Rickard E. (Rik) Faith <faith@valinux.com>
+ *
+ */
+
+/* THIS IS NOT AN X CONSORTIUM STANDARD */
+
+#define NEED_REPLIES
+#include <X11/Xlibint.h>
+#include <X11/extensions/Xext.h>
+#include <X11/extensions/extutil.h>
+#include "xf86dristr.h"
+
+static XExtensionInfo _xf86dri_info_data;
+static XExtensionInfo *xf86dri_info = &_xf86dri_info_data;
+static char xf86dri_extension_name[] = XF86DRINAME;
+
+#define XF86DRICheckExtension(dpy,i,val) \
+  XextCheckExtension (dpy, i, xf86dri_extension_name, val)
+
+/*****************************************************************************
+ *                                                                           *
+ *			   private utility routines                          *
+ *                                                                           *
+ *****************************************************************************/
+
+static int close_display(Display *dpy, XExtCodes *extCodes);
+static /* const */ XExtensionHooks xf86dri_extension_hooks = {
+    NULL,				/* create_gc */
+    NULL,				/* copy_gc */
+    NULL,				/* flush_gc */
+    NULL,				/* free_gc */
+    NULL,				/* create_font */
+    NULL,				/* free_font */
+    close_display,			/* close_display */
+    NULL,				/* wire_to_event */
+    NULL,				/* event_to_wire */
+    NULL,				/* error */
+    NULL,				/* error_string */
+};
+
+static XEXT_GENERATE_FIND_DISPLAY (find_display, xf86dri_info, 
+				   xf86dri_extension_name, 
+				   &xf86dri_extension_hooks, 
+				   0, NULL)
+
+static XEXT_GENERATE_CLOSE_DISPLAY (close_display, xf86dri_info)
+
+
+/*****************************************************************************
+ *                                                                           *
+ *		    public XFree86-DRI Extension routines                    *
+ *                                                                           *
+ *****************************************************************************/
+
+#if 0
+#include <stdio.h>
+#define TRACE(msg)  fprintf(stderr,"XF86DRI%s\n", msg);
+#else
+#define TRACE(msg)
+#endif
+
+#define PUBLIC
+
+PUBLIC Bool XF86DRIQueryExtension (dpy, event_basep, error_basep)
+    Display *dpy;
+    int *event_basep, *error_basep;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+
+    TRACE("QueryExtension...");
+    if (XextHasExtension(info)) {
+	*event_basep = info->codes->first_event;
+	*error_basep = info->codes->first_error;
+        TRACE("QueryExtension... return True");
+	return True;
+    } else {
+        TRACE("QueryExtension... return False");
+	return False;
+    }
+}
+
+PUBLIC Bool XF86DRIQueryVersion(dpy, majorVersion, minorVersion, patchVersion)
+    Display* dpy;
+    int* majorVersion; 
+    int* minorVersion;
+    int* patchVersion;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIQueryVersionReply rep;
+    xXF86DRIQueryVersionReq *req;
+
+    TRACE("QueryVersion...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIQueryVersion, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIQueryVersion;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("QueryVersion... return False");
+	return False;
+    }
+    *majorVersion = rep.majorVersion;
+    *minorVersion = rep.minorVersion;
+    *patchVersion = rep.patchVersion;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("QueryVersion... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIQueryDirectRenderingCapable(dpy, screen, isCapable)
+    Display* dpy;
+    int screen;
+    Bool* isCapable;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIQueryDirectRenderingCapableReply rep;
+    xXF86DRIQueryDirectRenderingCapableReq *req;
+
+    TRACE("QueryDirectRenderingCapable...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIQueryDirectRenderingCapable, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIQueryDirectRenderingCapable;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("QueryDirectRenderingCapable... return False");
+	return False;
+    }
+    *isCapable = rep.isCapable;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("QueryDirectRenderingCapable... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIOpenConnection(dpy, screen, hSAREA, busIdString)
+    Display* dpy;
+    int screen;
+    drm_handle_t * hSAREA;
+    char **busIdString;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIOpenConnectionReply rep;
+    xXF86DRIOpenConnectionReq *req;
+
+    TRACE("OpenConnection...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIOpenConnection, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIOpenConnection;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("OpenConnection... return False");
+	return False;
+    }
+
+    *hSAREA = rep.hSAREALow;
+    if (sizeof(drm_handle_t) == 8) {
+       int shift = 32; /* var to prevent warning on next line */
+       *hSAREA |= ((drm_handle_t) rep.hSAREAHigh) << shift;
+    }
+
+    if (rep.length) {
+        if (!(*busIdString = (char *)Xcalloc(rep.busIdStringLength + 1, 1))) {
+            _XEatData(dpy, ((rep.busIdStringLength+3) & ~3));
+            UnlockDisplay(dpy);
+            SyncHandle();
+            TRACE("OpenConnection... return False");
+            return False;
+        }
+	_XReadPad(dpy, *busIdString, rep.busIdStringLength);
+    } else {
+        *busIdString = NULL;
+    }
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("OpenConnection... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIAuthConnection(dpy, screen, magic)
+    Display* dpy;
+    int screen;
+    drm_magic_t magic;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIAuthConnectionReq *req;
+    xXF86DRIAuthConnectionReply rep;
+
+    TRACE("AuthConnection...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIAuthConnection, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIAuthConnection;
+    req->screen = screen;
+    req->magic = magic;
+    rep.authenticated = 0;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse) || !rep.authenticated) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("AuthConnection... return False");
+	return False;
+    }
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("AuthConnection... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRICloseConnection(dpy, screen)
+    Display* dpy;
+    int screen;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRICloseConnectionReq *req;
+
+    TRACE("CloseConnection...");
+
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRICloseConnection, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRICloseConnection;
+    req->screen = screen;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("CloseConnection... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIGetClientDriverName(dpy, screen, ddxDriverMajorVersion, 
+	ddxDriverMinorVersion, ddxDriverPatchVersion, clientDriverName)
+    Display* dpy;
+    int screen;
+    int* ddxDriverMajorVersion;
+    int* ddxDriverMinorVersion;
+    int* ddxDriverPatchVersion;
+    char** clientDriverName;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIGetClientDriverNameReply rep;
+    xXF86DRIGetClientDriverNameReq *req;
+
+    TRACE("GetClientDriverName...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIGetClientDriverName, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIGetClientDriverName;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetClientDriverName... return False");
+	return False;
+    }
+
+    *ddxDriverMajorVersion = rep.ddxDriverMajorVersion;
+    *ddxDriverMinorVersion = rep.ddxDriverMinorVersion;
+    *ddxDriverPatchVersion = rep.ddxDriverPatchVersion;
+
+    if (rep.length) {
+        if (!(*clientDriverName = (char *)Xcalloc(rep.clientDriverNameLength + 1, 1))) {
+            _XEatData(dpy, ((rep.clientDriverNameLength+3) & ~3));
+            UnlockDisplay(dpy);
+            SyncHandle();
+            TRACE("GetClientDriverName... return False");
+            return False;
+        }
+	_XReadPad(dpy, *clientDriverName, rep.clientDriverNameLength);
+    } else {
+        *clientDriverName = NULL;
+    }
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("GetClientDriverName... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRICreateContextWithConfig(dpy, screen, configID, context,
+	hHWContext)
+    Display* dpy;
+    int screen;
+    int configID;
+    XID* context;
+    drm_context_t * hHWContext;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRICreateContextReply rep;
+    xXF86DRICreateContextReq *req;
+
+    TRACE("CreateContext...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRICreateContext, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRICreateContext;
+    req->visual = configID;
+    req->screen = screen;
+    *context = XAllocID(dpy);
+    req->context = *context;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("CreateContext... return False");
+	return False;
+    }
+    *hHWContext = rep.hHWContext;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("CreateContext... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRICreateContext(dpy, screen, visual, context, hHWContext)
+    Display* dpy;
+    int screen;
+    Visual* visual;
+    XID* context;
+    drm_context_t * hHWContext;
+{
+    return XF86DRICreateContextWithConfig( dpy, screen, visual->visualid,
+					   context, hHWContext );
+}
+
+PUBLIC Bool XF86DRIDestroyContext( Display * ndpy, int screen, 
+    XID context )
+{
+    Display * const dpy = (Display *) ndpy;
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIDestroyContextReq *req;
+
+    TRACE("DestroyContext...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIDestroyContext, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIDestroyContext;
+    req->screen = screen;
+    req->context = context;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("DestroyContext... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRICreateDrawable( Display * ndpy, int screen, 
+    Drawable drawable, drm_drawable_t * hHWDrawable )
+{
+    Display * const dpy = (Display *) ndpy;
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRICreateDrawableReply rep;
+    xXF86DRICreateDrawableReq *req;
+
+    TRACE("CreateDrawable...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRICreateDrawable, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRICreateDrawable;
+    req->screen = screen;
+    req->drawable = drawable;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("CreateDrawable... return False");
+	return False;
+    }
+    *hHWDrawable = rep.hHWDrawable;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("CreateDrawable... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIDestroyDrawable( Display * ndpy, int screen,
+    Drawable drawable )
+{
+    Display * const dpy = (Display *) ndpy;
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIDestroyDrawableReq *req;
+
+    TRACE("DestroyDrawable...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIDestroyDrawable, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIDestroyDrawable;
+    req->screen = screen;
+    req->drawable = drawable;
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("DestroyDrawable... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIGetDrawableInfo(Display* dpy, int screen, Drawable drawable,
+    unsigned int* index, unsigned int* stamp,
+    int* X, int* Y, int* W, int* H,
+    int* numClipRects, drm_clip_rect_t ** pClipRects,
+    int* backX, int* backY,
+    int* numBackClipRects, drm_clip_rect_t ** pBackClipRects )
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIGetDrawableInfoReply rep;
+    xXF86DRIGetDrawableInfoReq *req;
+    int total_rects;
+
+    TRACE("GetDrawableInfo...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIGetDrawableInfo, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIGetDrawableInfo;
+    req->screen = screen;
+    req->drawable = drawable;
+
+    if (!_XReply(dpy, (xReply *)&rep, 1, xFalse)) 
+    {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetDrawableInfo... return False");
+	return False;
+    }
+    *index = rep.drawableTableIndex;
+    *stamp = rep.drawableTableStamp;
+    *X = (int)rep.drawableX;
+    *Y = (int)rep.drawableY;
+    *W = (int)rep.drawableWidth;
+    *H = (int)rep.drawableHeight;
+    *numClipRects = rep.numClipRects;
+    total_rects = *numClipRects;
+
+    *backX = rep.backX;
+    *backY = rep.backY;
+    *numBackClipRects = rep.numBackClipRects;
+    total_rects += *numBackClipRects;
+
+#if 0
+    /* Because of the fix in Xserver/GL/dri/xf86dri.c, this check breaks
+     * backwards compatibility (Because of the >> 2 shift) but the fix
+     * enables multi-threaded apps to work.
+     */
+    if (rep.length !=  ((((SIZEOF(xXF86DRIGetDrawableInfoReply) - 
+		       SIZEOF(xGenericReply) + 
+		       total_rects * sizeof(drm_clip_rect_t)) + 3) & ~3) >> 2)) {
+        _XEatData(dpy, rep.length);
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetDrawableInfo... return False");
+        return False;
+    }
+#endif
+
+    if (*numClipRects) {
+       int len = sizeof(drm_clip_rect_t) * (*numClipRects);
+
+       *pClipRects = (drm_clip_rect_t *)Xcalloc(len, 1);
+       if (*pClipRects) 
+	  _XRead(dpy, (char*)*pClipRects, len);
+    } else {
+        *pClipRects = NULL;
+    }
+
+    if (*numBackClipRects) {
+       int len = sizeof(drm_clip_rect_t) * (*numBackClipRects);
+
+       *pBackClipRects = (drm_clip_rect_t *)Xcalloc(len, 1);
+       if (*pBackClipRects) 
+	  _XRead(dpy, (char*)*pBackClipRects, len);
+    } else {
+        *pBackClipRects = NULL;
+    }
+
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("GetDrawableInfo... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIGetDeviceInfo(dpy, screen, hFrameBuffer, 
+	fbOrigin, fbSize, fbStride, devPrivateSize, pDevPrivate)
+    Display* dpy;
+    int screen;
+    drm_handle_t * hFrameBuffer;
+    int* fbOrigin;
+    int* fbSize;
+    int* fbStride;
+    int* devPrivateSize;
+    void** pDevPrivate;
+{
+    XExtDisplayInfo *info = find_display (dpy);
+    xXF86DRIGetDeviceInfoReply rep;
+    xXF86DRIGetDeviceInfoReq *req;
+
+    TRACE("GetDeviceInfo...");
+    XF86DRICheckExtension (dpy, info, False);
+
+    LockDisplay(dpy);
+    GetReq(XF86DRIGetDeviceInfo, req);
+    req->reqType = info->codes->major_opcode;
+    req->driReqType = X_XF86DRIGetDeviceInfo;
+    req->screen = screen;
+    if (!_XReply(dpy, (xReply *)&rep, 0, xFalse)) {
+	UnlockDisplay(dpy);
+	SyncHandle();
+        TRACE("GetDeviceInfo... return False");
+	return False;
+    }
+
+    *hFrameBuffer = rep.hFrameBufferLow;
+    if (sizeof(drm_handle_t) == 8) {
+       int shift = 32; /* var to prevent warning on next line */
+       *hFrameBuffer |= ((drm_handle_t) rep.hFrameBufferHigh) << shift;
+    }
+
+    *fbOrigin = rep.framebufferOrigin;
+    *fbSize = rep.framebufferSize;
+    *fbStride = rep.framebufferStride;
+    *devPrivateSize = rep.devPrivateSize;
+
+    if (rep.length) {
+        if (!(*pDevPrivate = (void *)Xcalloc(rep.devPrivateSize, 1))) {
+            _XEatData(dpy, ((rep.devPrivateSize+3) & ~3));
+            UnlockDisplay(dpy);
+            SyncHandle();
+            TRACE("GetDeviceInfo... return False");
+            return False;
+        }
+	_XRead(dpy, (char*)*pDevPrivate, rep.devPrivateSize);
+    } else {
+        *pDevPrivate = NULL;
+    }
+
+    UnlockDisplay(dpy);
+    SyncHandle();
+    TRACE("GetDeviceInfo... return True");
+    return True;
+}
+
+PUBLIC Bool XF86DRIOpenFullScreen(dpy, screen, drawable)
+    Display* dpy;
+    int screen;
+    Drawable drawable;
+{
+    /* This function and the underlying X protocol are deprecated.
+     */
+    (void) dpy;
+    (void) screen;
+    (void) drawable;
+    return False;
+}
+
+PUBLIC Bool XF86DRICloseFullScreen(dpy, screen, drawable)
+    Display* dpy;
+    int screen;
+    Drawable drawable;
+{
+    /* This function and the underlying X protocol are deprecated.
+     */
+    (void) dpy;
+    (void) screen;
+    (void) drawable;
+    return True;
+}
+
diff --git a/src/driclient/src/driclient.c b/src/driclient/src/driclient.c
new file mode 100644
index 0000000000..7a7ca95702
--- /dev/null
+++ b/src/driclient/src/driclient.c
@@ -0,0 +1,292 @@
+#include "driclient.h"
+#include <assert.h>
+#include <stdlib.h>
+
+int driCreateScreen(Display *display, int screen, dri_screen_t **dri_screen, dri_framebuffer_t *dri_framebuf)
+{
+	int		evbase, errbase;
+	char		*driver_name;
+	int		newly_opened;
+	drm_magic_t	magic;
+	drmVersionPtr	drm_version;
+	drm_handle_t	sarea_handle;
+	char		*bus_id;
+	dri_screen_t	*dri_scrn;
+	
+	assert(display);
+	assert(dri_screen);
+	
+	if (!XF86DRIQueryExtension(display, &evbase, &errbase))
+		return 1;
+	
+	dri_scrn = calloc(1, sizeof(dri_screen_t));
+	
+	if (!dri_scrn)
+		return 1;
+
+	if (!XF86DRIQueryVersion(display, &dri_scrn->dri.major, &dri_scrn->dri.minor, &dri_scrn->dri.patch))
+		goto free_screen;
+	
+	dri_scrn->display = display;
+	dri_scrn->num = screen;
+	dri_scrn->draw_lock_id = 1;
+	
+	if (!XF86DRIOpenConnection(display, screen, &sarea_handle, &bus_id))
+		goto free_screen;
+	
+	dri_scrn->fd = -1;
+	dri_scrn->fd = drmOpenOnce(NULL, bus_id, &newly_opened);
+	XFree(bus_id);
+
+	if (dri_scrn->fd < 0)
+		goto close_connection;
+	
+	if (drmGetMagic(dri_scrn->fd, &magic))
+		goto close_drm;
+	
+	drm_version = drmGetVersion(dri_scrn->fd);
+	
+	if (!drm_version)
+		goto close_drm;
+
+	dri_scrn->drm.major = drm_version->version_major;
+	dri_scrn->drm.minor = drm_version->version_minor;
+	dri_scrn->drm.patch = drm_version->version_patchlevel;
+	drmFreeVersion(drm_version);
+				
+	if (!XF86DRIAuthConnection(display, screen, magic))
+		goto close_drm;
+	
+	if (!XF86DRIGetClientDriverName
+	(
+		display,
+		screen,
+		&dri_scrn->ddx.major,
+		&dri_scrn->ddx.minor,
+		&dri_scrn->ddx.patch,
+		&driver_name
+	))
+		goto close_drm;
+	
+	if (drmMap(dri_scrn->fd, sarea_handle, SAREA_MAX, (drmAddress)&dri_scrn->sarea))
+		goto close_drm;
+	
+	dri_scrn->drawable_hash = drmHashCreate();
+	
+	if (!dri_scrn->drawable_hash)
+		goto unmap_sarea;
+	
+	if (dri_framebuf)
+	{
+		if (!XF86DRIGetDeviceInfo
+		(
+			display,
+			screen, &dri_framebuf->drm_handle,
+			&dri_framebuf->base,
+			&dri_framebuf->size,
+			&dri_framebuf->stride,
+			&dri_framebuf->private_size,
+			&dri_framebuf->private
+		))
+			goto destroy_hash;
+	}
+	
+	*dri_screen = dri_scrn;
+	
+	return 0;
+
+destroy_hash:
+	drmHashDestroy(dri_scrn->drawable_hash);
+unmap_sarea:
+	drmUnmap(dri_scrn->sarea, SAREA_MAX);
+close_drm:
+	drmCloseOnce(dri_scrn->fd);
+close_connection:
+	XF86DRICloseConnection(display, screen);
+free_screen:
+	free(dri_scrn);
+	
+	return 1;
+}
+
+int driDestroyScreen(dri_screen_t *dri_screen)
+{
+	assert(dri_screen);
+	
+	drmHashDestroy(dri_screen->drawable_hash);
+	drmUnmap(dri_screen->sarea, SAREA_MAX);
+	drmCloseOnce(dri_screen->fd);
+	XF86DRICloseConnection(dri_screen->display, dri_screen->num);
+	free(dri_screen);
+	
+	return 0;
+}
+
+int driCreateDrawable(dri_screen_t *dri_screen, Drawable drawable, dri_drawable_t **dri_drawable)
+{
+	int		evbase, errbase;
+	dri_drawable_t	*dri_draw;
+	
+	assert(dri_screen);
+	assert(dri_drawable);
+	
+	if (!XF86DRIQueryExtension(dri_screen->display, &evbase, &errbase))
+		return 1;
+	
+	if (!drmHashLookup(dri_screen->drawable_hash, drawable, (void**)dri_drawable))
+	{
+		/* Found */
+		(*dri_drawable)->refcount++;
+		return 0;
+	}
+	
+	dri_draw = calloc(1, sizeof(dri_drawable_t));
+	
+	if (!dri_draw)
+		return 1;
+	
+	if (!XF86DRICreateDrawable(dri_screen->display, 0, drawable, &dri_draw->drm_drawable))
+	{
+		free(dri_draw);
+		return 1;
+	}
+	
+	dri_draw->x_drawable = drawable;
+	dri_draw->sarea_index = 0;
+	dri_draw->sarea_stamp = NULL;
+	dri_draw->last_sarea_stamp = 0;
+	dri_draw->dri_screen = dri_screen;
+	dri_draw->refcount = 1;
+	
+	if (drmHashInsert(dri_screen->drawable_hash, drawable, dri_draw))
+	{
+		XF86DRIDestroyDrawable(dri_screen->display, dri_screen->num, drawable);
+		free(dri_draw);
+		return 1;
+	}
+	
+	/*
+	 * XXX: Need this to initialize sarea pointer and other stuff in dri_drawable_t
+	 * to be able to use the DRI_VALIDATE_DRAWABLE_INFO macro, but is it safe to
+	 * call without any sync?
+	 */
+	if (driUpdateDrawableInfo(dri_draw))
+	{
+		XF86DRIDestroyDrawable(dri_screen->display, dri_screen->num, drawable);
+		free(dri_draw);
+		return 1;
+	}
+	
+	*dri_drawable = dri_draw;
+	
+	return 0;
+}
+
+int driUpdateDrawableInfo(dri_drawable_t *dri_drawable)
+{
+	assert(dri_drawable);
+	
+	if (dri_drawable->cliprects)
+		XFree(dri_drawable->cliprects);
+	if (dri_drawable->back_cliprects)
+		XFree(dri_drawable->back_cliprects);
+	
+	DRM_SPINUNLOCK(&dri_drawable->dri_screen->sarea->drawable_lock, dri_drawable->dri_screen->draw_lock_id);
+	
+	if (!XF86DRIGetDrawableInfo
+	(
+		dri_drawable->dri_screen->display,
+		dri_drawable->dri_screen->num,
+		dri_drawable->x_drawable,
+    		&dri_drawable->sarea_index,
+    		&dri_drawable->last_sarea_stamp,
+    		&dri_drawable->x,
+    		&dri_drawable->y,
+    		&dri_drawable->w,
+    		&dri_drawable->h,
+		&dri_drawable->num_cliprects,
+		&dri_drawable->cliprects,
+		&dri_drawable->back_x,
+		&dri_drawable->back_y,
+		&dri_drawable->num_back_cliprects,
+		&dri_drawable->back_cliprects
+	))
+	{
+		dri_drawable->sarea_stamp = &dri_drawable->last_sarea_stamp;
+		dri_drawable->num_cliprects = 0;
+		dri_drawable->cliprects = NULL;
+		dri_drawable->num_back_cliprects = 0;
+		dri_drawable->back_cliprects = 0;
+		
+		return 1;
+	}
+	else
+		dri_drawable->sarea_stamp = &dri_drawable->dri_screen->sarea->drawableTable[dri_drawable->sarea_index].stamp;
+	
+	DRM_SPINLOCK(&dri_drawable->dri_screen->sarea->drawable_lock, dri_drawable->dri_screen->draw_lock_id);
+	
+	return 0;
+}
+
+int driDestroyDrawable(dri_drawable_t *dri_drawable)
+{
+	assert(dri_drawable);
+	
+	if (--dri_drawable->refcount == 0)
+	{
+		if (dri_drawable->cliprects)
+			XFree(dri_drawable->cliprects);
+		if (dri_drawable->back_cliprects)
+			XFree(dri_drawable->back_cliprects);
+		drmHashDelete(dri_drawable->dri_screen->drawable_hash, dri_drawable->x_drawable);
+		XF86DRIDestroyDrawable(dri_drawable->dri_screen->display, dri_drawable->dri_screen->num, dri_drawable->x_drawable);
+		free(dri_drawable);
+	}
+	
+	return 0;
+}
+
+int driCreateContext(dri_screen_t *dri_screen, Visual *visual, dri_context_t **dri_context)
+{
+	int		evbase, errbase;
+	dri_context_t	*dri_ctx;
+	
+	assert(dri_screen);
+	assert(visual);
+	assert(dri_context);
+	
+	if (!XF86DRIQueryExtension(dri_screen->display, &evbase, &errbase))
+		return 1;
+	
+	dri_ctx = calloc(1, sizeof(dri_context_t));
+	
+	if (!dri_ctx)
+		return 1;
+	
+	if (!XF86DRICreateContext(dri_screen->display, dri_screen->num, visual, &dri_ctx->id, &dri_ctx->drm_context))
+	{
+		free(dri_ctx);
+		return 1;
+	}
+	
+	dri_ctx->dri_screen = dri_screen;	
+	*dri_context = dri_ctx;
+	
+	return 0;
+}
+
+int driDestroyContext(dri_context_t *dri_context)
+{
+	assert(dri_context);
+	
+	XF86DRIDestroyContext(dri_context->dri_screen->display, dri_context->dri_screen->num, dri_context->id);
+	free(dri_context);
+	
+	return 0;
+}
+
+int driCompareVersions(const dri_version_t *v1, const dri_version_t *v2)
+{
+	return (v1->major == v2->major) && (v1->minor == v2->minor) && (v1->patch == v2->patch);
+}
+
diff --git a/src/driclient/src/test b/src/driclient/src/test
new file mode 100755
index 0000000000..57cddf8e00
Binary files /dev/null and b/src/driclient/src/test differ
diff --git a/src/driclient/src/test.c b/src/driclient/src/test.c
new file mode 100644
index 0000000000..15f75d928b
--- /dev/null
+++ b/src/driclient/src/test.c
@@ -0,0 +1,41 @@
+#include <assert.h>
+#include <stdio.h>
+#include "driclient.h"
+
+int main(int argc, char **argv)
+{
+	Display		*dpy;
+	Window		root, window;
+	
+	dri_screen_t	*screen;
+	dri_drawable_t	*dri_drawable;
+	dri_context_t	*context;
+	
+	dpy = XOpenDisplay(NULL);
+	root = XDefaultRootWindow(dpy);
+	window = XCreateSimpleWindow(dpy, root, 0, 0, 100, 100, 0, 0, 0);
+
+	XSelectInput(dpy, window, 0);
+	XMapWindow(dpy, window);
+	XSync(dpy, 0);
+	
+	assert(driCreateScreen(dpy, 0, &screen, NULL) == 0);
+	assert(driCreateDrawable(screen, window, &dri_drawable) == 0);
+	assert(driCreateContext(screen, XDefaultVisual(dpy, 0), &context) == 0);
+	assert(driUpdateDrawableInfo(dri_drawable) == 0);
+	
+	DRI_VALIDATE_DRAWABLE_INFO(screen, dri_drawable);
+	
+	assert(drmGetLock(screen->fd, context->drm_context, 0) == 0);
+	assert(drmUnlock(screen->fd, context->drm_context) == 0);
+	
+	assert(driDestroyContext(context) == 0);
+	assert(driDestroyDrawable(dri_drawable) == 0);
+	assert(driDestroyScreen(screen) == 0);
+	
+	XDestroyWindow(dpy, window);
+	XCloseDisplay(dpy);
+	
+	return 0;
+}
+
diff --git a/src/driclient/src/xf86dri.h b/src/driclient/src/xf86dri.h
new file mode 100644
index 0000000000..baf80a7a9d
--- /dev/null
+++ b/src/driclient/src/xf86dri.h
@@ -0,0 +1,119 @@
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, 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, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/**
+ * \file xf86dri.h
+ * Protocol numbers and function prototypes for DRI X protocol.
+ *
+ * \author Kevin E. Martin <martin@valinux.com>
+ * \author Jens Owen <jens@tungstengraphics.com>
+ * \author Rickard E. (Rik) Faith <faith@valinux.com>
+ */
+
+#ifndef _XF86DRI_H_
+#define _XF86DRI_H_
+
+#include <X11/Xfuncproto.h>
+#include <xf86drm.h>
+
+#define X_XF86DRIQueryVersion			0
+#define X_XF86DRIQueryDirectRenderingCapable	1
+#define X_XF86DRIOpenConnection			2
+#define X_XF86DRICloseConnection		3
+#define X_XF86DRIGetClientDriverName		4
+#define X_XF86DRICreateContext			5
+#define X_XF86DRIDestroyContext			6
+#define X_XF86DRICreateDrawable			7
+#define X_XF86DRIDestroyDrawable		8
+#define X_XF86DRIGetDrawableInfo		9
+#define X_XF86DRIGetDeviceInfo			10
+#define X_XF86DRIAuthConnection                 11
+#define X_XF86DRIOpenFullScreen                 12   /* Deprecated */
+#define X_XF86DRICloseFullScreen                13   /* Deprecated */
+
+#define XF86DRINumberEvents		0
+
+#define XF86DRIClientNotLocal		0
+#define XF86DRIOperationNotSupported	1
+#define XF86DRINumberErrors		(XF86DRIOperationNotSupported + 1)
+
+#ifndef _XF86DRI_SERVER_
+
+_XFUNCPROTOBEGIN
+
+Bool XF86DRIQueryExtension( Display *dpy, int *event_base, int *error_base );
+
+Bool XF86DRIQueryVersion( Display *dpy, int *majorVersion, int *minorVersion,
+    int *patchVersion );
+
+Bool XF86DRIQueryDirectRenderingCapable( Display *dpy, int screen,
+    Bool *isCapable );
+
+Bool XF86DRIOpenConnection( Display *dpy, int screen, drm_handle_t *hSAREA,
+    char **busIDString );
+
+Bool XF86DRIAuthConnection( Display *dpy, int screen, drm_magic_t magic );
+
+Bool XF86DRICloseConnection( Display *dpy, int screen );
+
+Bool XF86DRIGetClientDriverName( Display *dpy, int screen,
+    int *ddxDriverMajorVersion, int *ddxDriverMinorVersion,
+    int *ddxDriverPatchVersion, char **clientDriverName );
+
+Bool XF86DRICreateContext( Display *dpy, int screen, Visual *visual,
+    XID *ptr_to_returned_context_id, drm_context_t *hHWContext );
+
+Bool XF86DRICreateContextWithConfig( Display *dpy, int screen, int configID,
+    XID *ptr_to_returned_context_id, drm_context_t *hHWContext );
+
+Bool XF86DRIDestroyContext( Display *dpy, int screen,
+    XID context_id );
+
+Bool XF86DRICreateDrawable( Display *dpy, int screen,
+    Drawable drawable, drm_drawable_t *hHWDrawable );
+
+Bool XF86DRIDestroyDrawable( Display *dpy, int screen, 
+    Drawable drawable);
+
+Bool XF86DRIGetDrawableInfo( Display *dpy, int screen, Drawable drawable,
+    unsigned int *index, unsigned int *stamp, 
+    int *X, int *Y, int *W, int *H,
+    int *numClipRects, drm_clip_rect_t ** pClipRects,
+    int *backX, int *backY,
+    int *numBackClipRects, drm_clip_rect_t **pBackClipRects );
+
+Bool XF86DRIGetDeviceInfo( Display *dpy, int screen,
+    drm_handle_t *hFrameBuffer, int *fbOrigin, int *fbSize,
+    int *fbStride, int *devPrivateSize, void **pDevPrivate );
+
+_XFUNCPROTOEND
+
+#endif /* _XF86DRI_SERVER_ */
+
+#endif /* _XF86DRI_H_ */
+
diff --git a/src/driclient/src/xf86dristr.h b/src/driclient/src/xf86dristr.h
new file mode 100644
index 0000000000..b834bd1a1a
--- /dev/null
+++ b/src/driclient/src/xf86dristr.h
@@ -0,0 +1,342 @@
+/**************************************************************************
+
+Copyright 1998-1999 Precision Insight, Inc., Cedar Park, Texas.
+Copyright 2000 VA Linux Systems, 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, sub license, and/or sell copies of the Software, and to
+permit persons to whom the Software is furnished to do so, subject to
+the following conditions:
+
+The above copyright notice and this permission notice (including the
+next paragraph) shall be included in all copies or substantial portions
+of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+
+**************************************************************************/
+
+/*
+ * Authors:
+ *   Kevin E. Martin <martin@valinux.com>
+ *   Jens Owen <jens@tungstengraphics.com>
+ *   Rickard E. (Rik) Fiath <faith@valinux.com>
+ *
+ */
+
+#ifndef _XF86DRISTR_H_
+#define _XF86DRISTR_H_
+
+#include "xf86dri.h"
+
+#define XF86DRINAME "XFree86-DRI"
+
+/* The DRI version number.  This was originally set to be the same of the
+ * XFree86 version number.  However, this version is really indepedent of
+ * the XFree86 version.
+ *
+ * Version History:
+ *    4.0.0: Original
+ *    4.0.1: Patch to bump clipstamp when windows are destroyed, 28 May 02
+ *    4.1.0: Add transition from single to multi in DRMInfo rec, 24 Jun 02
+ */
+#define XF86DRI_MAJOR_VERSION	4
+#define XF86DRI_MINOR_VERSION	1
+#define XF86DRI_PATCH_VERSION	0
+
+typedef struct _XF86DRIQueryVersion {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIQueryVersion */
+    CARD16	length B16;
+} xXF86DRIQueryVersionReq;
+#define sz_xXF86DRIQueryVersionReq	4
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD16	majorVersion B16;	/* major version of DRI protocol */
+    CARD16	minorVersion B16;	/* minor version of DRI protocol */
+    CARD32	patchVersion B32;       /* patch version of DRI protocol */
+    CARD32	pad3 B32;
+    CARD32	pad4 B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRIQueryVersionReply;
+#define sz_xXF86DRIQueryVersionReply	32
+
+typedef struct _XF86DRIQueryDirectRenderingCapable {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* X_DRIQueryDirectRenderingCapable */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIQueryDirectRenderingCapableReq;
+#define sz_xXF86DRIQueryDirectRenderingCapableReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    BOOL	isCapable;
+    BOOL	pad2;
+    BOOL	pad3;
+    BOOL	pad4;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+    CARD32	pad7 B32;
+    CARD32	pad8 B32;
+    CARD32	pad9 B32;
+} xXF86DRIQueryDirectRenderingCapableReply;
+#define sz_xXF86DRIQueryDirectRenderingCapableReply	32
+
+typedef struct _XF86DRIOpenConnection {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIOpenConnection */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIOpenConnectionReq;
+#define sz_xXF86DRIOpenConnectionReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hSAREALow B32;
+    CARD32	hSAREAHigh B32;
+    CARD32	busIdStringLength B32;
+    CARD32	pad6 B32;
+    CARD32	pad7 B32;
+    CARD32	pad8 B32;
+} xXF86DRIOpenConnectionReply;
+#define sz_xXF86DRIOpenConnectionReply	32
+
+typedef struct _XF86DRIAuthConnection {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICloseConnection */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32      magic B32;
+} xXF86DRIAuthConnectionReq;
+#define sz_xXF86DRIAuthConnectionReq	12
+
+typedef struct {
+    BYTE        type;
+    BOOL        pad1;
+    CARD16      sequenceNumber B16;
+    CARD32      length B32;
+    CARD32      authenticated B32;
+    CARD32      pad2 B32;
+    CARD32      pad3 B32;
+    CARD32      pad4 B32;
+    CARD32      pad5 B32;
+    CARD32      pad6 B32;
+} xXF86DRIAuthConnectionReply;
+#define zx_xXF86DRIAuthConnectionReply  32
+
+typedef struct _XF86DRICloseConnection {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICloseConnection */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRICloseConnectionReq;
+#define sz_xXF86DRICloseConnectionReq	8
+
+typedef struct _XF86DRIGetClientDriverName {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIGetClientDriverName */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIGetClientDriverNameReq;
+#define sz_xXF86DRIGetClientDriverNameReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	ddxDriverMajorVersion B32;
+    CARD32	ddxDriverMinorVersion B32;
+    CARD32	ddxDriverPatchVersion B32;
+    CARD32	clientDriverNameLength B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRIGetClientDriverNameReply;
+#define sz_xXF86DRIGetClientDriverNameReply	32
+
+typedef struct _XF86DRICreateContext {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICreateContext */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	visual B32;
+    CARD32	context B32;
+} xXF86DRICreateContextReq;
+#define sz_xXF86DRICreateContextReq	16
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hHWContext B32;
+    CARD32	pad2 B32;
+    CARD32	pad3 B32;
+    CARD32	pad4 B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRICreateContextReply;
+#define sz_xXF86DRICreateContextReply	32
+
+typedef struct _XF86DRIDestroyContext {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIDestroyContext */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	context B32;
+} xXF86DRIDestroyContextReq;
+#define sz_xXF86DRIDestroyContextReq	12
+
+typedef struct _XF86DRICreateDrawable {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRICreateDrawable */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	drawable B32;
+} xXF86DRICreateDrawableReq;
+#define sz_xXF86DRICreateDrawableReq	12
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hHWDrawable B32;
+    CARD32	pad2 B32;
+    CARD32	pad3 B32;
+    CARD32	pad4 B32;
+    CARD32	pad5 B32;
+    CARD32	pad6 B32;
+} xXF86DRICreateDrawableReply;
+#define sz_xXF86DRICreateDrawableReply	32
+
+typedef struct _XF86DRIDestroyDrawable {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIDestroyDrawable */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	drawable B32;
+} xXF86DRIDestroyDrawableReq;
+#define sz_xXF86DRIDestroyDrawableReq	12
+
+typedef struct _XF86DRIGetDrawableInfo {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIGetDrawableInfo */
+    CARD16	length B16;
+    CARD32	screen B32;
+    CARD32	drawable B32;
+} xXF86DRIGetDrawableInfoReq;
+#define sz_xXF86DRIGetDrawableInfoReq	12
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	drawableTableIndex B32;
+    CARD32	drawableTableStamp B32;
+    INT16	drawableX B16;
+    INT16	drawableY B16;
+    INT16	drawableWidth B16;
+    INT16	drawableHeight B16;
+    CARD32	numClipRects B32;
+    INT16       backX B16;
+    INT16       backY B16;
+    CARD32      numBackClipRects B32;
+} xXF86DRIGetDrawableInfoReply;
+
+#define sz_xXF86DRIGetDrawableInfoReply	36
+
+
+typedef struct _XF86DRIGetDeviceInfo {
+    CARD8	reqType;		/* always DRIReqCode */
+    CARD8	driReqType;		/* always X_DRIGetDeviceInfo */
+    CARD16	length B16;
+    CARD32	screen B32;
+} xXF86DRIGetDeviceInfoReq;
+#define sz_xXF86DRIGetDeviceInfoReq	8
+
+typedef struct {
+    BYTE	type;			/* X_Reply */
+    BOOL	pad1;
+    CARD16	sequenceNumber B16;
+    CARD32	length B32;
+    CARD32	hFrameBufferLow B32;
+    CARD32	hFrameBufferHigh B32;
+    CARD32	framebufferOrigin B32;
+    CARD32	framebufferSize B32;
+    CARD32	framebufferStride B32;
+    CARD32	devPrivateSize B32;
+} xXF86DRIGetDeviceInfoReply;
+#define sz_xXF86DRIGetDeviceInfoReply	32
+
+typedef struct _XF86DRIOpenFullScreen {
+    CARD8       reqType;	/* always DRIReqCode */
+    CARD8       driReqType;	/* always X_DRIOpenFullScreen */
+    CARD16      length B16;
+    CARD32      screen B32;
+    CARD32      drawable B32;
+} xXF86DRIOpenFullScreenReq;
+#define sz_xXF86DRIOpenFullScreenReq    12
+
+typedef struct {
+    BYTE        type;
+    BOOL        pad1;
+    CARD16      sequenceNumber B16;
+    CARD32      length B32;
+    CARD32      isFullScreen B32;
+    CARD32      pad2 B32;
+    CARD32      pad3 B32;
+    CARD32      pad4 B32;
+    CARD32      pad5 B32;
+    CARD32      pad6 B32;
+} xXF86DRIOpenFullScreenReply;
+#define sz_xXF86DRIOpenFullScreenReply  32
+
+typedef struct _XF86DRICloseFullScreen {
+    CARD8       reqType;	/* always DRIReqCode */
+    CARD8       driReqType;	/* always X_DRICloseFullScreen */
+    CARD16      length B16;
+    CARD32      screen B32;
+    CARD32      drawable B32;
+} xXF86DRICloseFullScreenReq;
+#define sz_xXF86DRICloseFullScreenReq   12
+
+typedef struct {
+    BYTE        type;
+    BOOL        pad1;
+    CARD16      sequenceNumber B16;
+    CARD32      length B32;
+    CARD32      pad2 B32;
+    CARD32      pad3 B32;
+    CARD32      pad4 B32;
+    CARD32      pad5 B32;
+    CARD32      pad6 B32;
+    CARD32      pad7 B32;
+} xXF86DRICloseFullScreenReply;
+#define sz_xXF86DRICloseFullScreenReply  32
+
+
+#endif /* _XF86DRISTR_H_ */
diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
index 50e3c843b5..c6a22cad4e 100644
--- a/src/gallium/state_trackers/g3dvl/Makefile
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -2,7 +2,7 @@ TARGET		= libg3dvl.a
 OBJECTS		= vl_context.o vl_data.o vl_surface.o vl_shader_build.o vl_util.o
 GALLIUMDIR	= ../..
 
-CFLAGS		+= -g -Wall -fPIC -Werror -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary
+CFLAGS		+= -g -Wall -fPIC -Werror -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary -I${GALLIUMDIR}/winsys/g3dvl
 
 #############################################
 
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c
index 3d4ca7cf4e..850a769376 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.c
+++ b/src/gallium/state_trackers/g3dvl/vl_context.c
@@ -330,6 +330,10 @@ static int vlCreateVertexShaderIMC(struct VL_CONTEXT *context)
 		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
 	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
 	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
@@ -354,7 +358,7 @@ static int vlCreateVertexShaderIMC(struct VL_CONTEXT *context)
 	
 	vs.tokens = tokens;
 	context->states.mc.i_vs = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -432,7 +436,7 @@ static int vlCreateFragmentShaderIMC(struct VL_CONTEXT *context)
 
 	fs.tokens = tokens;
 	context->states.mc.i_fs = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -499,6 +503,10 @@ static int vlCreateVertexShaderFramePMC(struct VL_CONTEXT *context)
 		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
 	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
 	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
@@ -527,7 +535,7 @@ static int vlCreateVertexShaderFramePMC(struct VL_CONTEXT *context)
 	
 	vs.tokens = tokens;
 	context->states.mc.p_vs[0] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -603,6 +611,10 @@ static int vlCreateVertexShaderFieldPMC(struct VL_CONTEXT *context)
 		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
 	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
 	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
@@ -643,7 +655,7 @@ static int vlCreateVertexShaderFieldPMC(struct VL_CONTEXT *context)
 	
 	vs.tokens = tokens;
 	context->states.mc.p_vs[1] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -700,6 +712,10 @@ static int vlCreateFragmentShaderFramePMC(struct VL_CONTEXT *context)
 	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
 	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/*
 	 * decl s0			; Sampler for luma texture
 	 * decl s1			; Sampler for chroma Cb texture
@@ -761,7 +777,7 @@ static int vlCreateFragmentShaderFramePMC(struct VL_CONTEXT *context)
 
 	fs.tokens = tokens;
 	context->states.mc.p_fs[0] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -821,6 +837,10 @@ static int vlCreateFragmentShaderFieldPMC(struct VL_CONTEXT *context)
 	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
 	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
+	/* decl t0-t4 */
+	decl = vl_decl_temps(0, 4);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/*
 	 * decl s0			; Sampler for luma texture
 	 * decl s1			; Sampler for chroma Cb texture
@@ -928,7 +948,7 @@ static int vlCreateFragmentShaderFieldPMC(struct VL_CONTEXT *context)
 
 	fs.tokens = tokens;
 	context->states.mc.p_fs[1] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -998,6 +1018,10 @@ static int vlCreateVertexShaderFrameBMC(struct VL_CONTEXT *context)
 		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
 	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
 	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
@@ -1030,7 +1054,7 @@ static int vlCreateVertexShaderFrameBMC(struct VL_CONTEXT *context)
 	
 	vs.tokens = tokens;
 	context->states.mc.b_vs[0] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -1104,6 +1128,10 @@ static int vlCreateVertexShaderFieldBMC(struct VL_CONTEXT *context)
 		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
 	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
 	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
@@ -1148,7 +1176,7 @@ static int vlCreateVertexShaderFieldBMC(struct VL_CONTEXT *context)
 	
 	vs.tokens = tokens;
 	context->states.mc.b_vs[1] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -1207,6 +1235,10 @@ static int vlCreateFragmentShaderFrameBMC(struct VL_CONTEXT *context)
 	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
 	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
+	/* decl t0-t2 */
+	decl = vl_decl_temps(0, 2);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/*
 	 * decl s0			; Sampler for luma texture
 	 * decl s1			; Sampler for chroma Cb texture
@@ -1283,7 +1315,7 @@ static int vlCreateFragmentShaderFrameBMC(struct VL_CONTEXT *context)
 
 	fs.tokens = tokens;
 	context->states.mc.b_fs[0] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -1346,6 +1378,10 @@ static int vlCreateFragmentShaderFieldBMC(struct VL_CONTEXT *context)
 	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
 	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
+	/* decl t0-t5 */
+	decl = vl_decl_temps(0, 5);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/*
 	 * decl s0			; Sampler for luma texture
 	 * decl s1			; Sampler for chroma Cb texture
@@ -1479,7 +1515,7 @@ static int vlCreateFragmentShaderFieldBMC(struct VL_CONTEXT *context)
 
 	fs.tokens = tokens;
 	context->states.mc.b_fs[1] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -1771,6 +1807,10 @@ static int vlCreateVertexShaderCSC(struct VL_CONTEXT *context)
 		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	}
 	
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/* mov o0, i0		; Move pos in to pos out */
 	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_INPUT, 0);
 	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
@@ -1789,7 +1829,7 @@ static int vlCreateVertexShaderCSC(struct VL_CONTEXT *context)
 	
 	vs.tokens = tokens;
 	context->states.csc.vertex_shader = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
@@ -1839,6 +1879,10 @@ static int vlCreateFragmentShaderCSC(struct VL_CONTEXT *context)
 	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
 	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
 	
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	
 	/* decl s0		; Sampler for tex containing picture to display */
 	decl = vl_decl_samplers(0, 0);
 	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
@@ -1870,7 +1914,7 @@ static int vlCreateFragmentShaderCSC(struct VL_CONTEXT *context)
 
 	fs.tokens = tokens;
 	context->states.csc.fragment_shader = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
+	//free(tokens);
 	
 	return 0;
 }
diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.c b/src/gallium/state_trackers/g3dvl/vl_shader_build.c
index 365ad69725..1dc5be6fdb 100644
--- a/src/gallium/state_trackers/g3dvl/vl_shader_build.c
+++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.c
@@ -74,6 +74,18 @@ struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int inde
 	return decl;
 }
 
+struct tgsi_full_declaration vl_decl_temps(unsigned int first, unsigned int last)
+{
+	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
+
+	decl = tgsi_default_full_declaration();
+	decl.Declaration.File = TGSI_FILE_TEMPORARY;
+	decl.DeclarationRange.First = first;
+	decl.DeclarationRange.Last = last;
+	
+	return decl;
+}
+
 struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int last)
 {
 	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.h b/src/gallium/state_trackers/g3dvl/vl_shader_build.h
index 9e64bbeeae..878d7e2c45 100644
--- a/src/gallium/state_trackers/g3dvl/vl_shader_build.h
+++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.h
@@ -14,6 +14,7 @@ struct tgsi_full_declaration vl_decl_interpolated_input
 );
 struct tgsi_full_declaration vl_decl_constants(unsigned int name, unsigned int index, unsigned int first, unsigned int last);
 struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int index, unsigned int first, unsigned int last);
+struct tgsi_full_declaration vl_decl_temps(unsigned int first, unsigned int last);
 struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int last);
 struct tgsi_full_instruction vl_inst2
 (
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c
index 145ea32892..9b91ab4e22 100644
--- a/src/gallium/state_trackers/g3dvl/vl_surface.c
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.c
@@ -5,6 +5,7 @@
 #include <pipe/p_state.h>
 #include <pipe/p_format.h>
 #include <pipe/p_inlines.h>
+#include <vl_winsys.h>
 #include "vl_context.h"
 #include "vl_defs.h"
 #include "vl_util.h"
@@ -711,12 +712,13 @@ int vlPutSurface
 	pipe->set_sampler_textures(pipe, 1, &surface->texture);
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_STRIP, 0, 4);
 	pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+	bind_pipe_drawable(pipe, drawable);
 	/* TODO: Need to take destx, desty into consideration */
 	pipe->winsys->flush_frontbuffer
 	(
 		pipe->winsys,
 		surface->context->states.csc.framebuffer.cbufs[0],
-		&drawable
+		pipe->priv
 	);
 	
 	vlBeginRender(surface->context);
diff --git a/src/gallium/winsys/g3dvl/nouveau/Makefile b/src/gallium/winsys/g3dvl/nouveau/Makefile
new file mode 100644
index 0000000000..644ac083e6
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/Makefile
@@ -0,0 +1,46 @@
+TARGET		= libnouveau_dri.so
+GALLIUMDIR	= ../../..
+DRMDIR		?= /usr
+DRIDIR		= ../../../../driclient
+
+OBJECTS		= nouveau_bo.o nouveau_fence.o nouveau_swapbuffers.o nouveau_channel.o		\
+		  nouveau_grobj.o nouveau_context.o nouveau_winsys.o nouveau_lock.o		\
+		  nouveau_winsys_pipe.o nouveau_device.o nouveau_notifier.o nouveau_dma.o	\
+		  nouveau_pushbuf.o nouveau_resource.o nouveau_screen.o nv04_surface.o		\
+		  nv50_surface.o #nouveau_winsys_softpipe.o
+
+CFLAGS		+= -g -Wall -Werror -fPIC		\
+		   -I${GALLIUMDIR}/include		\
+		   -I${GALLIUMDIR}/winsys/g3dvl		\
+		   -I${DRMDIR}/include			\
+		   -I${DRMDIR}/include/drm		\
+		   -I${GALLIUMDIR}/drivers		\
+		   -I${GALLIUMDIR}/auxiliary		\
+		   -I${DRIDIR}/include
+
+LDFLAGS		+= -L${DRMDIR}/lib			\
+		   -L${DRIDIR}/lib			\
+		   -L${GALLIUMDIR}/auxiliary/draw	\
+		   -L${GALLIUMDIR}/auxiliary/tgsi	\
+		   -L${GALLIUMDIR}/auxiliary/translate	\
+		   -L${GALLIUMDIR}/auxiliary/rtasm	\
+		   -L${GALLIUMDIR}/auxiliary/cso_cache	\
+		   -L${GALLIUMDIR}/drivers/nv10		\
+		   -L${GALLIUMDIR}/drivers/nv30		\
+		   -L${GALLIUMDIR}/drivers/nv40		\
+		   -L${GALLIUMDIR}/drivers/nv50
+
+LIBS		+= -ldriclient -ldrm -lnv10 -lnv30 -lnv40 -lnv50 -ldraw -ltgsi -ltranslate -lrtasm -lcso_cache -lm
+
+#############################################
+
+.PHONY	= all clean
+
+all: ${TARGET}
+
+${TARGET}: ${OBJECTS}
+	$(CC) ${LDFLAGS} -shared -o $@ $^ ${LIBS}
+
+clean:
+	rm -rf ${OBJECTS} ${TARGET}
+
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_bo.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_bo.c
new file mode 120000
index 0000000000..73ac4a4171
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_bo.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_bo.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_channel.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_channel.c
new file mode 120000
index 0000000000..6c9b2c48d8
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_channel.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_channel.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_context.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_context.c
new file mode 100644
index 0000000000..5e173c7672
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_context.c
@@ -0,0 +1,371 @@
+#include "pipe/p_defines.h"
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "pipe/p_util.h"
+
+#include "nouveau_context.h"
+#include "nouveau_dri.h"
+#include "nouveau_local.h"
+#include "nouveau_screen.h"
+#include "nouveau_winsys_pipe.h"
+
+/*
+#ifdef DEBUG
+static const struct dri_debug_control debug_control[] = {
+	{ "bo", DEBUG_BO },
+	{ NULL, 0 }
+};
+int __nouveau_debug = 0;
+#endif
+*/
+
+/*
+ * TODO: Re-examine dri_screen, dri_context, nouveau_screen, nouveau_context
+ * relationships, seems like there is a lot of room for simplification there.
+ */
+
+static void
+nouveau_channel_context_destroy(struct nouveau_channel_context *nvc)
+{
+	nouveau_grobj_free(&nvc->NvCtxSurf2D);
+	nouveau_grobj_free(&nvc->NvImageBlit);
+	nouveau_grobj_free(&nvc->NvGdiRect);
+	nouveau_grobj_free(&nvc->NvM2MF);
+	nouveau_grobj_free(&nvc->Nv2D);
+	nouveau_grobj_free(&nvc->NvSwzSurf);
+	nouveau_grobj_free(&nvc->NvSIFM);
+
+	nouveau_notifier_free(&nvc->sync_notifier);
+
+	nouveau_channel_free(&nvc->channel);
+
+	FREE(nvc);
+}
+
+static struct nouveau_channel_context *
+nouveau_channel_context_create(struct nouveau_device *dev)
+{
+	struct nouveau_channel_context *nvc;
+	int ret;
+
+	nvc = CALLOC_STRUCT(nouveau_channel_context);
+	if (!nvc)
+		return NULL;
+
+	if ((ret = nouveau_channel_alloc(dev, 0x8003d001, 0x8003d002,
+					 &nvc->channel))) {
+		NOUVEAU_ERR("Error creating GPU channel: %d\n", ret);
+		nouveau_channel_context_destroy(nvc);
+		return NULL;
+	}
+
+	nvc->next_handle = 0x80000000;
+
+	if ((ret = nouveau_notifier_alloc(nvc->channel, nvc->next_handle++, 1,
+					  &nvc->sync_notifier))) {
+		NOUVEAU_ERR("Error creating channel sync notifier: %d\n", ret);
+		nouveau_channel_context_destroy(nvc);
+		return NULL;
+	}
+
+	switch (dev->chipset & 0xf0) {
+	case 0x50:
+	case 0x80:
+	case 0x90:
+		ret = nouveau_surface_channel_create_nv50(nvc);
+		break;
+	default:
+		ret = nouveau_surface_channel_create_nv04(nvc);
+		break;
+	}
+
+	if (ret) {
+		NOUVEAU_ERR("Error initialising surface objects: %d\n", ret);
+		nouveau_channel_context_destroy(nvc);
+		return NULL;
+	}
+
+	return nvc;
+}
+
+int
+nouveau_context_create(dri_context_t *dri_context)
+{
+	dri_screen_t			*dri_screen = dri_context->dri_screen;
+	struct nouveau_screen 		*nv_screen = dri_screen->private;
+	struct nouveau_context		*nv = CALLOC_STRUCT(nouveau_context);
+	struct pipe_context		*pipe = NULL;
+	struct nouveau_channel_context	*nvc = NULL;
+	struct nouveau_device		*dev = nv_screen->device;
+	int				i;
+
+	switch (dev->chipset & 0xf0) {
+	case 0x10:
+	case 0x20:
+		/* NV10 */
+	case 0x30:
+		/* NV30 */
+	case 0x40:
+	case 0x60:
+		/* NV40 */
+	case 0x50:
+	case 0x80:
+	case 0x90:
+		/* G80 */
+		break;
+	default:
+		NOUVEAU_ERR("Unsupported chipset: NV%02x\n", dev->chipset);
+		return 1;
+	}
+
+	dri_context->private = (void*)nv;
+	nv->dri_context = dri_context;
+	nv->nv_screen  = nv_screen;
+
+	{
+		struct nouveau_device_priv *nvdev = nouveau_device(dev);
+
+		nvdev->ctx  = dri_context->drm_context;
+		nvdev->lock = (drmLock*)&dri_screen->sarea->lock;
+	}
+
+	/*
+	driParseConfigFiles(&nv->dri_option_cache, &nv_screen->option_cache,
+			    nv->dri_screen->myNum, "nouveau");
+#ifdef DEBUG
+	__nouveau_debug = driParseDebugString(getenv("NOUVEAU_DEBUG"),
+					      debug_control);
+#endif
+	*/
+
+	/*XXX: Hack up a fake region and buffer object for front buffer.
+	 *     This will go away with TTM, replaced with a simple reference
+	 *     of the front buffer handle passed to us by the DDX.
+	 */
+	{
+		struct pipe_surface *fb_surf;
+		struct nouveau_pipe_buffer *fb_buf;
+		struct nouveau_bo_priv *fb_bo;
+
+		fb_bo = calloc(1, sizeof(struct nouveau_bo_priv));
+		fb_bo->drm.offset = nv_screen->front_offset;
+		fb_bo->drm.flags = NOUVEAU_MEM_FB;
+		fb_bo->drm.size = nv_screen->front_pitch * 
+				  nv_screen->front_height;
+		fb_bo->refcount = 1;
+		fb_bo->base.flags = NOUVEAU_BO_PIN | NOUVEAU_BO_VRAM;
+		fb_bo->base.offset = fb_bo->drm.offset;
+		fb_bo->base.handle = (unsigned long)fb_bo;
+		fb_bo->base.size = fb_bo->drm.size;
+		fb_bo->base.device = nv_screen->device;
+
+		fb_buf = calloc(1, sizeof(struct nouveau_pipe_buffer));
+		fb_buf->bo = &fb_bo->base;
+
+		fb_surf = calloc(1, sizeof(struct pipe_surface));
+		if (nv_screen->front_cpp == 2)
+			fb_surf->format = PIPE_FORMAT_R5G6B5_UNORM;
+		else
+			fb_surf->format = PIPE_FORMAT_A8R8G8B8_UNORM;
+		pf_get_block(fb_surf->format, &fb_surf->block);
+		fb_surf->width = nv_screen->front_pitch / nv_screen->front_cpp;
+		fb_surf->height = nv_screen->front_height;
+		fb_surf->stride = fb_surf->width * fb_surf->block.size;
+		fb_surf->refcount = 1;
+		fb_surf->buffer = &fb_buf->base;
+
+		nv->frontbuffer = fb_surf;
+	}
+
+	nvc = nv_screen->nvc;
+
+	if (!nvc) {
+		nvc = nouveau_channel_context_create(dev);
+		if (!nvc) {
+			NOUVEAU_ERR("Failed initialising GPU context\n");
+			return 1;
+		}
+		nv_screen->nvc = nvc;
+	}
+
+	nvc->refcount++;
+	nv->nvc = nvc;
+
+	/* Find a free slot for a pipe context, allocate a new one if needed */
+	nv->pctx_id = -1;
+	for (i = 0; i < nvc->nr_pctx; i++) {
+		if (nvc->pctx[i] == NULL) {
+			nv->pctx_id = i;
+			break;
+		}
+	}
+
+	if (nv->pctx_id < 0) {
+		nv->pctx_id = nvc->nr_pctx++;
+		nvc->pctx =
+			realloc(nvc->pctx,
+				sizeof(struct pipe_context *) * nvc->nr_pctx);
+	}
+
+	/* Create pipe */
+	switch (dev->chipset & 0xf0) {
+	case 0x50:
+	case 0x80:
+	case 0x90:
+		if (nouveau_surface_init_nv50(nv))
+			return 1;
+		break;
+	default:
+		if (nouveau_surface_init_nv04(nv))
+			return 1;
+		break;
+	}
+
+	if (!getenv("NOUVEAU_FORCE_SOFTPIPE")) {
+		struct pipe_screen *pscreen;
+
+		pipe = nouveau_pipe_create(nv);
+		if (!pipe)
+			NOUVEAU_ERR("Couldn't create hw pipe\n");
+		pscreen = nvc->pscreen;
+
+		nv->cap.hw_vertex_buffer =
+			pscreen->get_param(pscreen, NOUVEAU_CAP_HW_VTXBUF);
+		nv->cap.hw_index_buffer =
+			pscreen->get_param(pscreen, NOUVEAU_CAP_HW_IDXBUF);
+	}
+
+	/* XXX: nouveau_winsys_softpipe needs a mesa header removed before we can compile it. */
+	/*
+	if (!pipe) {
+		NOUVEAU_MSG("Using softpipe\n");
+		pipe = nouveau_create_softpipe(nv);
+		if (!pipe) {
+			NOUVEAU_ERR("Error creating pipe, bailing\n");
+			return 1;
+		}
+	}
+	*/
+	if (!pipe) {
+		NOUVEAU_ERR("Error creating pipe, bailing\n");
+		return 1;
+	}
+
+	pipe->priv = nv;
+
+	return 0;
+}
+
+void
+nouveau_context_destroy(dri_context_t *dri_context)
+{
+	struct nouveau_context *nv = dri_context->private;
+	struct nouveau_channel_context *nvc = nv->nvc;
+
+	assert(nv);
+
+	if (nv->pctx_id >= 0) {
+		nvc->pctx[nv->pctx_id] = NULL;
+		if (--nvc->refcount <= 0) {
+			nouveau_channel_context_destroy(nvc);
+			nv->nv_screen->nvc = NULL;
+		}
+	}
+
+	free(nv);
+}
+
+int
+nouveau_context_bind(struct nouveau_context *nv, dri_drawable_t *dri_drawable)
+{
+	assert(nv);
+	assert(dri_drawable);
+	
+	if (nv->dri_drawable != dri_drawable)
+	{
+		nv->dri_drawable = dri_drawable;
+		dri_drawable->private = nv;
+	}
+
+	return 0;
+}
+
+int
+nouveau_context_unbind(struct nouveau_context *nv)
+{
+	assert(nv);
+	
+	nv->dri_drawable = NULL;
+	
+	return 0;
+}
+
+/* Show starts here */
+
+int bind_pipe_drawable(struct pipe_context *pipe, Drawable drawable)
+{
+	struct nouveau_context	*nv;
+	dri_drawable_t		*dri_drawable;
+	
+	nv = pipe->priv;
+	
+	driCreateDrawable(nv->nv_screen->dri_screen, drawable, &dri_drawable);
+	
+	nouveau_context_bind(nv, dri_drawable);
+	
+	return 0;
+}
+
+int unbind_pipe_drawable(struct pipe_context *pipe)
+{
+	nouveau_context_unbind(pipe->priv);
+	
+	return 0;
+}
+
+struct pipe_context* create_pipe_context(Display *display, int screen)
+{
+	dri_screen_t		*dri_screen;
+	dri_framebuffer_t	dri_framebuf;
+	dri_context_t		*dri_context;
+	struct nouveau_context	*nv;
+	
+	driCreateScreen(display, screen, &dri_screen, &dri_framebuf);
+	driCreateContext(dri_screen, XDefaultVisual(display, screen), &dri_context);
+	
+	nouveau_screen_create(dri_screen, &dri_framebuf);
+	nouveau_context_create(dri_context);
+	
+	nv = dri_context->private;
+	
+	return nv->nvc->pctx[nv->pctx_id];
+}
+
+int destroy_pipe_context(struct pipe_context *pipe)
+{
+	struct pipe_screen	*screen;
+	struct pipe_winsys	*winsys;
+	struct nouveau_context	*nv;
+	dri_screen_t		*dri_screen;
+	dri_context_t		*dri_context;
+	
+	assert(pipe);
+	
+	screen = pipe->screen;
+	winsys = pipe->winsys;
+	nv = pipe->priv;
+	dri_context = nv->dri_context;
+	dri_screen = dri_context->dri_screen;
+	
+	pipe->destroy(pipe);
+	screen->destroy(screen);
+	free(winsys);
+
+	nouveau_context_destroy(dri_context);
+	nouveau_screen_destroy(dri_screen);
+	driDestroyContext(dri_context);
+	driDestroyScreen(dri_screen);
+	
+	return 0;
+}
+
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_context.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_context.h
new file mode 100644
index 0000000000..395a3ab790
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_context.h
@@ -0,0 +1,105 @@
+#ifndef __NOUVEAU_CONTEXT_H__
+#define __NOUVEAU_CONTEXT_H__
+
+/*#include "xmlconfig.h"*/
+
+#include <driclient.h>
+#include "nouveau/nouveau_winsys.h"
+#include "nouveau_drmif.h"
+#include "nouveau_dma.h"
+
+struct nouveau_channel_context {
+	struct pipe_screen *pscreen;
+	int refcount;
+
+	unsigned cur_pctx;
+	unsigned nr_pctx;
+	struct pipe_context **pctx;
+
+	struct nouveau_channel  *channel;
+
+	struct nouveau_notifier *sync_notifier;
+
+	/* Common */
+	struct nouveau_grobj    *NvM2MF;
+	/* NV04-NV40 */
+	struct nouveau_grobj    *NvCtxSurf2D;
+	struct nouveau_grobj	*NvSwzSurf;
+	struct nouveau_grobj    *NvImageBlit;
+	struct nouveau_grobj    *NvGdiRect;
+	struct nouveau_grobj	*NvSIFM;
+	/* G80 */
+	struct nouveau_grobj    *Nv2D;
+
+	uint32_t                 next_handle;
+	uint32_t                 next_subchannel;
+	uint32_t                 next_sequence;
+};
+
+struct nouveau_context {
+	/* DRI stuff */
+	dri_context_t		*dri_context;
+	dri_drawable_t		*dri_drawable;
+	unsigned int		last_stamp;
+	/*driOptionCache	dri_option_cache;*/
+	drm_context_t		drm_context;
+	drmLock			drm_lock;
+	int			locked;
+	struct nouveau_screen	*nv_screen;
+	struct pipe_surface	*frontbuffer;
+
+	struct {
+		int hw_vertex_buffer;
+		int hw_index_buffer;
+	} cap;
+
+	/* Hardware context */
+	struct nouveau_channel_context	*nvc;
+	int				pctx_id;
+
+	/* pipe_surface accel */
+	struct pipe_surface		*surf_src, *surf_dst;
+	unsigned			surf_src_offset, surf_dst_offset;
+	
+	int  (*surface_copy_prep)(struct nouveau_context *,
+				  struct pipe_surface *dst,
+				  struct pipe_surface *src);
+	void (*surface_copy)(struct nouveau_context *, unsigned dx, unsigned dy,
+			     unsigned sx, unsigned sy, unsigned w, unsigned h);
+	void (*surface_copy_done)(struct nouveau_context *);
+	int (*surface_fill)(struct nouveau_context *, struct pipe_surface *,
+			    unsigned, unsigned, unsigned, unsigned, unsigned);
+};
+
+extern int nouveau_context_create(dri_context_t *);
+extern void nouveau_context_destroy(dri_context_t *);
+extern int nouveau_context_bind(struct nouveau_context *, dri_drawable_t *);
+extern int nouveau_context_unbind(struct nouveau_context *);
+
+#ifdef DEBUG
+extern int __nouveau_debug;
+
+#define DEBUG_BO (1 << 0)
+
+#define DBG(flag, ...) do {                   \
+	if (__nouveau_debug & (DEBUG_##flag)) \
+		NOUVEAU_ERR(__VA_ARGS__);     \
+} while(0)
+#else
+#define DBG(flag, ...)
+#endif
+
+extern void LOCK_HARDWARE(struct nouveau_context *);
+extern void UNLOCK_HARDWARE(struct nouveau_context *);
+
+extern int
+nouveau_surface_channel_create_nv04(struct nouveau_channel_context *);
+extern int
+nouveau_surface_channel_create_nv50(struct nouveau_channel_context *);
+extern int nouveau_surface_init_nv04(struct nouveau_context *);
+extern int nouveau_surface_init_nv50(struct nouveau_context *);
+
+extern uint32_t *nouveau_pipe_dma_beginp(struct nouveau_grobj *, int, int);
+extern void nouveau_pipe_dma_kickoff(struct nouveau_channel *);
+
+#endif
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_device.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_device.c
new file mode 120000
index 0000000000..47d52dafa8
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_device.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_device.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_dma.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_dma.c
new file mode 120000
index 0000000000..45078c964f
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_dma.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_dma.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_dma.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_dma.h
new file mode 120000
index 0000000000..6b9ec77741
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_dma.h
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_dma.h
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_dri.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_dri.h
new file mode 120000
index 0000000000..0e6c9fce35
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_dri.h
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_dri.h
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_drmif.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_drmif.h
new file mode 120000
index 0000000000..473b7d4812
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_drmif.h
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_drmif.h
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_fence.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_fence.c
new file mode 120000
index 0000000000..ef1f0c6afe
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_fence.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_fence.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_grobj.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_grobj.c
new file mode 120000
index 0000000000..428186544c
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_grobj.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_grobj.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_local.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_local.h
new file mode 120000
index 0000000000..6b878aad22
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_local.h
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_local.h
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_lock.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_lock.c
new file mode 100644
index 0000000000..375634bd05
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_lock.c
@@ -0,0 +1,92 @@
+/**************************************************************************
+ * 
+ * Copyright 2003 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * 
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ * 
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ * 
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ * 
+ **************************************************************************/
+
+#include <pthread.h>
+#include <driclient.h>
+#include "nouveau_context.h"
+#include "nouveau_screen.h"
+
+static pthread_mutex_t lockMutex = PTHREAD_MUTEX_INITIALIZER;
+
+static void
+nouveau_contended_lock(struct nouveau_context *nv, unsigned int flags)
+{
+	dri_drawable_t			*dri_drawable = nv->dri_drawable;
+	dri_screen_t			*dri_screen = nv->dri_context->dri_screen;
+	struct nouveau_screen		*nv_screen = nv->nv_screen;
+	struct nouveau_device		*dev = nv_screen->device;
+	struct nouveau_device_priv	*nvdev = nouveau_device(dev);
+
+	drmGetLock(nvdev->fd, nvdev->ctx, flags);
+
+	/* If the window moved, may need to set a new cliprect now.
+	 *
+	 * NOTE: This releases and regains the hw lock, so all state
+	 * checking must be done *after* this call:
+	 */
+	if (dri_drawable)
+		DRI_VALIDATE_DRAWABLE_INFO(dri_screen, dri_drawable);
+}
+
+/* Lock the hardware and validate our state.
+ */
+void
+LOCK_HARDWARE(struct nouveau_context *nv)
+{
+	struct nouveau_screen		*nv_screen = nv->nv_screen;
+	struct nouveau_device		*dev = nv_screen->device;
+	struct nouveau_device_priv	*nvdev = nouveau_device(dev);
+	char				__ret=0;
+
+	pthread_mutex_lock(&lockMutex);
+	assert(!nv->locked);
+	
+	DRM_CAS(nvdev->lock, nvdev->ctx,
+		(DRM_LOCK_HELD | nvdev->ctx), __ret);
+	
+	if (__ret)
+		nouveau_contended_lock(nv, 0);
+	nv->locked = 1;
+}
+
+
+/* Unlock the hardware using the global current context 
+ */
+void
+UNLOCK_HARDWARE(struct nouveau_context *nv)
+{
+	struct nouveau_screen		*nv_screen = nv->nv_screen;
+	struct nouveau_device		*dev = nv_screen->device;
+	struct nouveau_device_priv	*nvdev = nouveau_device(dev);
+
+	assert(nv->locked);
+	nv->locked = 0;
+
+	DRM_UNLOCK(nvdev->fd, nvdev->lock, nvdev->ctx);
+
+	pthread_mutex_unlock(&lockMutex);
+} 
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_notifier.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_notifier.c
new file mode 120000
index 0000000000..3024a612a7
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_notifier.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_notifier.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_pushbuf.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_pushbuf.c
new file mode 120000
index 0000000000..dae31d9799
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_pushbuf.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_pushbuf.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_resource.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_resource.c
new file mode 120000
index 0000000000..e0d71e9d2b
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_resource.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_resource.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_screen.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_screen.c
new file mode 100644
index 0000000000..0087ce0056
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_screen.c
@@ -0,0 +1,88 @@
+#include "pipe/p_context.h"
+#include "pipe/p_util.h"
+#include "nouveau_context.h"
+#include "nouveau_drm.h"
+#include "nouveau_dri.h"
+#include "nouveau_local.h"
+#include "nouveau_screen.h"
+#include "nouveau_swapbuffers.h"
+
+#if NOUVEAU_DRM_HEADER_PATCHLEVEL != 11
+#error nouveau_drm.h version does not match expected version
+#endif
+
+/*
+PUBLIC const char __driConfigOptions[] =
+DRI_CONF_BEGIN
+DRI_CONF_END;
+static const GLuint __driNConfigOptions = 0;
+*/
+
+int
+nouveau_screen_create(dri_screen_t *dri_screen, dri_framebuffer_t *dri_framebuf)
+{
+	/* XXX: Someone forgot to bump this? */
+	static const dri_version_t ddx_expected = {0, 0, 10 /*NOUVEAU_DRM_HEADER_PATCHLEVEL*/};
+	static const dri_version_t dri_expected = {4, 1, 0};
+	static const dri_version_t drm_expected = {0, 0, NOUVEAU_DRM_HEADER_PATCHLEVEL};
+	
+	struct nouveau_dri	*nv_dri = dri_framebuf->private;
+	struct nouveau_screen	*nv_screen;
+	int			ret;
+	
+	if (!driCompareVersions(&ddx_expected, &dri_screen->ddx))
+	{
+		NOUVEAU_ERR("Unexpected DDX version.\n");
+		return 1;
+	}
+	
+	if (!driCompareVersions(&drm_expected, &dri_screen->drm))
+	{
+		NOUVEAU_ERR("Unexpected DRM version.\n");
+		return 1;
+	}
+	
+	if (!driCompareVersions(&dri_expected, &dri_screen->dri))
+	{
+		NOUVEAU_ERR("Unexpected DRI version.\n");
+		return 1;
+	}
+
+	if (dri_framebuf->private_size != sizeof(struct nouveau_dri)) {
+		NOUVEAU_ERR("DRI struct mismatch between DDX/DRI.\n");
+		return 1;
+	}
+
+	nv_screen = CALLOC_STRUCT(nouveau_screen);
+	if (!nv_screen)
+		return 1;
+	nv_screen->dri_screen = dri_screen;
+	dri_screen->private = (void*)nv_screen;
+
+	/*
+	driParseOptionInfo(&nv_screen->option_cache,
+			   __driConfigOptions, __driNConfigOptions);
+	*/
+
+	if ((ret = nouveau_device_open_existing(&nv_screen->device, 0,
+						dri_screen->fd, 0))) {
+		NOUVEAU_ERR("Failed opening nouveau device: %d.\n", ret);
+		return 1;
+	}
+
+	nv_screen->front_offset = nv_dri->front_offset;
+	nv_screen->front_pitch  = nv_dri->front_pitch * (nv_dri->bpp / 8);
+	nv_screen->front_cpp = nv_dri->bpp / 8;
+	nv_screen->front_height = nv_dri->height;
+
+	return 0;
+}
+
+void
+nouveau_screen_destroy(dri_screen_t *dri_screen)
+{
+	struct nouveau_screen *nv_screen = dri_screen->private;
+
+	FREE(nv_screen);
+}
+
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_screen.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_screen.h
new file mode 100644
index 0000000000..8a58bb7556
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_screen.h
@@ -0,0 +1,24 @@
+#ifndef __NOUVEAU_SCREEN_H__
+#define __NOUVEAU_SCREEN_H__
+
+/* TODO: Investigate using DRI options for interesting things */
+/*#include "xmlconfig.h"*/
+
+struct nouveau_screen {
+	dri_screen_t			*dri_screen;
+	struct nouveau_device		*device;
+	struct nouveau_channel_context	*nvc;
+
+	uint32_t			front_offset;
+	uint32_t			front_pitch;
+	uint32_t			front_cpp;
+	uint32_t			front_height;
+	
+	/*driOptionCache		option_cache;*/
+};
+
+int nouveau_screen_create(dri_screen_t *dri_screen, dri_framebuffer_t *dri_framebuf);
+void nouveau_screen_destroy(dri_screen_t *dri_screen);
+
+#endif
+
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c
new file mode 100644
index 0000000000..7916c80615
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.c
@@ -0,0 +1,64 @@
+#include "pipe/p_context.h"
+#include "nouveau_context.h"
+#include "nouveau_local.h"
+#include "nouveau_screen.h"
+#include "nouveau_swapbuffers.h"
+
+void
+nouveau_copy_buffer(dri_drawable_t *dri_drawable, struct pipe_surface *surf,
+		    const drm_clip_rect_t *rect)
+{
+	struct nouveau_context	*nv = dri_drawable->private;
+	drm_clip_rect_t		*pbox;
+	int			nbox, i;
+
+	LOCK_HARDWARE(nv);
+	if (!dri_drawable->num_cliprects) {
+		UNLOCK_HARDWARE(nv);
+		return;
+	}
+	pbox = dri_drawable->cliprects;
+	nbox = dri_drawable->num_cliprects;
+
+	nv->surface_copy_prep(nv, nv->frontbuffer, surf);
+	for (i = 0; i < nbox; i++, pbox++) {
+		int sx, sy, dx, dy, w, h;
+
+		sx = pbox->x1 - dri_drawable->x;
+		sy = pbox->y1 - dri_drawable->y;
+		dx = pbox->x1;
+		dy = pbox->y1;
+		w  = pbox->x2 - pbox->x1;
+		h  = pbox->y2 - pbox->y1;
+
+		nv->surface_copy(nv, dx, dy, sx, sy, w, h);
+	}
+
+	FIRE_RING(nv->nvc->channel);
+	UNLOCK_HARDWARE(nv);
+
+	//if (nv->last_stamp != dri_drawable->last_sarea_stamp)
+		//nv->last_stamp = dri_drawable->last_sarea_stamp;
+}
+
+void
+nouveau_copy_sub_buffer(dri_drawable_t *dri_drawable, struct pipe_surface *surf, int x, int y, int w, int h)
+{
+	if (surf) {
+		drm_clip_rect_t rect;
+		rect.x1 = x;
+		rect.y1 = y;
+		rect.x2 = x + w;
+		rect.y2 = y + h;
+
+		nouveau_copy_buffer(dri_drawable, surf, &rect);
+	}
+}
+
+void
+nouveau_swap_buffers(dri_drawable_t *dri_drawable, struct pipe_surface *surf)
+{
+	if (surf)
+		nouveau_copy_buffer(dri_drawable, surf, NULL);
+}
+
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.h
new file mode 100644
index 0000000000..35e934adba
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_swapbuffers.h
@@ -0,0 +1,10 @@
+#ifndef __NOUVEAU_SWAPBUFFERS_H__
+#define __NOUVEAU_SWAPBUFFERS_H__
+
+extern void nouveau_copy_buffer(dri_drawable_t *, struct pipe_surface *,
+				const drm_clip_rect_t *);
+extern void nouveau_copy_sub_buffer(dri_drawable_t *, struct pipe_surface *,
+				    int x, int y, int w, int h);
+extern void nouveau_swap_buffers(dri_drawable_t *, struct pipe_surface *);
+
+#endif
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys.c
new file mode 120000
index 0000000000..43de49b98b
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_winsys.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_pipe.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_pipe.c
new file mode 100644
index 0000000000..b835bd5760
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_pipe.c
@@ -0,0 +1,261 @@
+#include "pipe/p_winsys.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_util.h"
+#include "pipe/p_inlines.h"
+
+#include "nouveau_context.h"
+#include "nouveau_local.h"
+#include "nouveau_screen.h"
+#include "nouveau_swapbuffers.h"
+#include "nouveau_winsys_pipe.h"
+
+static void
+nouveau_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surf,
+			  void *context_private)
+{
+	struct nouveau_context *nv = context_private;
+	dri_drawable_t *dri_drawable = nv->dri_drawable;
+
+	nouveau_copy_buffer(dri_drawable, surf, NULL);
+}
+
+static const char *
+nouveau_get_name(struct pipe_winsys *pws)
+{
+	return "Nouveau/DRI";
+}
+
+static struct pipe_surface *
+nouveau_surface_alloc(struct pipe_winsys *ws)
+{
+	struct pipe_surface *surf;
+	
+	surf = CALLOC_STRUCT(pipe_surface);
+	if (!surf)
+		return NULL;
+
+	surf->refcount = 1;
+	surf->winsys = ws;
+	return surf;
+}
+
+/* Borrowed from Mesa's xm_winsys */
+static unsigned int
+round_up(unsigned n, unsigned multiple)
+{
+   return (n + multiple - 1) & ~(multiple - 1);
+}
+
+static int
+nouveau_surface_alloc_storage
+(
+	struct pipe_winsys *pws,
+	struct pipe_surface *surface,
+	unsigned width,
+	unsigned height,
+	enum pipe_format format,
+	unsigned flags,
+	unsigned tex_usage
+)
+{
+	const unsigned int ALIGNMENT = 256;
+	
+	assert(pws);
+	assert(surface);
+	
+	surface->width = width;
+	surface->height = height;
+	surface->format = format;
+	pf_get_block(format, &surface->block);
+	surface->nblocksx = pf_get_nblocksx(&surface->block, width);
+	surface->nblocksy = pf_get_nblocksy(&surface->block, height);
+	surface->stride = round_up(surface->nblocksx * surface->block.size, ALIGNMENT);
+	surface->usage = flags;
+	surface->buffer = pws->buffer_create(pws, ALIGNMENT, PIPE_BUFFER_USAGE_PIXEL, surface->stride * surface->nblocksy);
+	
+	return 0;
+}
+
+static void
+nouveau_surface_release(struct pipe_winsys *ws, struct pipe_surface **s)
+{
+	struct pipe_surface *surf = *s;
+
+	*s = NULL;
+	if (--surf->refcount <= 0) {
+		if (surf->buffer)
+			pipe_buffer_reference(ws, &surf->buffer, NULL);
+		free(surf);
+	}
+}
+
+static struct pipe_buffer *
+nouveau_pipe_bo_create(struct pipe_winsys *pws, unsigned alignment,
+		       unsigned usage, unsigned size)
+{
+	struct nouveau_pipe_winsys *nvpws = (struct nouveau_pipe_winsys *)pws;
+	struct nouveau_context *nv = nvpws->nv;
+	struct nouveau_device *dev = nv->nv_screen->device;
+	struct nouveau_pipe_buffer *nvbuf;
+	uint32_t flags;
+
+	nvbuf = calloc(1, sizeof(*nvbuf));
+	if (!nvbuf)
+		return NULL;
+	nvbuf->base.refcount = 1;
+	nvbuf->base.alignment = alignment;
+	nvbuf->base.usage = usage;
+	nvbuf->base.size = size;
+
+	flags = NOUVEAU_BO_LOCAL;
+
+	if (usage & PIPE_BUFFER_USAGE_PIXEL) {
+		if (usage & NOUVEAU_BUFFER_USAGE_TEXTURE)
+			flags |= NOUVEAU_BO_GART;
+		flags |= NOUVEAU_BO_VRAM;
+	}
+
+	if (usage & PIPE_BUFFER_USAGE_VERTEX) {
+		if (nv->cap.hw_vertex_buffer)
+			flags |= NOUVEAU_BO_GART;
+	}
+
+	if (usage & PIPE_BUFFER_USAGE_INDEX) {
+		if (nv->cap.hw_index_buffer)
+			flags |= NOUVEAU_BO_GART;
+	}
+
+	if (nouveau_bo_new(dev, flags, alignment, size, &nvbuf->bo)) {
+		free(nvbuf);
+		return NULL;
+	}
+
+	return &nvbuf->base;
+}
+
+static struct pipe_buffer *
+nouveau_pipe_bo_user_create(struct pipe_winsys *pws, void *ptr, unsigned bytes)
+{
+	struct nouveau_pipe_winsys *nvpws = (struct nouveau_pipe_winsys *)pws;
+	struct nouveau_device *dev = nvpws->nv->nv_screen->device;
+	struct nouveau_pipe_buffer *nvbuf;
+
+	nvbuf = calloc(1, sizeof(*nvbuf));
+	if (!nvbuf)
+		return NULL;
+	nvbuf->base.refcount = 1;
+	nvbuf->base.size = bytes;
+
+	if (nouveau_bo_user(dev, ptr, bytes, &nvbuf->bo)) {
+		free(nvbuf);
+		return NULL;
+	}
+
+	return &nvbuf->base;
+}
+
+static void
+nouveau_pipe_bo_del(struct pipe_winsys *ws, struct pipe_buffer *buf)
+{
+	struct nouveau_pipe_buffer *nvbuf = nouveau_buffer(buf);
+
+	nouveau_bo_del(&nvbuf->bo);
+	free(nvbuf);
+}
+
+static void *
+nouveau_pipe_bo_map(struct pipe_winsys *pws, struct pipe_buffer *buf,
+		    unsigned flags)
+{
+	struct nouveau_pipe_buffer *nvbuf = nouveau_buffer(buf);
+	uint32_t map_flags = 0;
+
+	if (flags & PIPE_BUFFER_USAGE_CPU_READ)
+		map_flags |= NOUVEAU_BO_RD;
+	if (flags & PIPE_BUFFER_USAGE_CPU_WRITE)
+		map_flags |= NOUVEAU_BO_WR;
+
+	if (nouveau_bo_map(nvbuf->bo, map_flags))
+		return NULL;
+	return nvbuf->bo->map;
+}
+
+static void
+nouveau_pipe_bo_unmap(struct pipe_winsys *pws, struct pipe_buffer *buf)
+{
+	struct nouveau_pipe_buffer *nvbuf = nouveau_buffer(buf);
+
+	nouveau_bo_unmap(nvbuf->bo);
+}
+
+static INLINE struct nouveau_fence *
+nouveau_pipe_fence(struct pipe_fence_handle *pfence)
+{
+	return (struct nouveau_fence *)pfence;
+}
+
+static void
+nouveau_pipe_fence_reference(struct pipe_winsys *ws,
+			     struct pipe_fence_handle **ptr,
+			     struct pipe_fence_handle *pfence)
+{
+	nouveau_fence_ref((void *)pfence, (void *)ptr);
+}
+
+static int
+nouveau_pipe_fence_signalled(struct pipe_winsys *ws,
+			     struct pipe_fence_handle *pfence, unsigned flag)
+{
+	struct nouveau_pipe_winsys *nvpws = (struct nouveau_pipe_winsys *)ws;
+	struct nouveau_fence *fence = nouveau_pipe_fence(pfence);
+
+	if (nouveau_fence(fence)->signalled == 0)
+		nouveau_fence_flush(nvpws->nv->nvc->channel);
+
+	return !nouveau_fence(fence)->signalled;
+}
+
+static int
+nouveau_pipe_fence_finish(struct pipe_winsys *ws,
+			  struct pipe_fence_handle *pfence, unsigned flag)
+{
+	struct nouveau_fence *fence = nouveau_pipe_fence(pfence);
+	struct nouveau_fence *ref = NULL;
+
+	nouveau_fence_ref(fence, &ref);
+	return nouveau_fence_wait(&ref);
+}
+
+struct pipe_winsys *
+nouveau_create_pipe_winsys(struct nouveau_context *nv)
+{
+	struct nouveau_pipe_winsys *nvpws;
+	struct pipe_winsys *pws;
+
+	nvpws = CALLOC_STRUCT(nouveau_pipe_winsys);
+	if (!nvpws)
+		return NULL;
+	nvpws->nv = nv;
+	pws = &nvpws->pws;
+
+	pws->flush_frontbuffer = nouveau_flush_frontbuffer;
+
+	pws->surface_alloc = nouveau_surface_alloc;
+	pws->surface_alloc_storage = nouveau_surface_alloc_storage;
+	pws->surface_release = nouveau_surface_release;
+
+	pws->buffer_create = nouveau_pipe_bo_create;
+	pws->buffer_destroy = nouveau_pipe_bo_del;
+	pws->user_buffer_create = nouveau_pipe_bo_user_create;
+	pws->buffer_map = nouveau_pipe_bo_map;
+	pws->buffer_unmap = nouveau_pipe_bo_unmap;
+
+	pws->fence_reference = nouveau_pipe_fence_reference;
+	pws->fence_signalled = nouveau_pipe_fence_signalled;
+	pws->fence_finish = nouveau_pipe_fence_finish;
+
+	pws->get_name = nouveau_get_name;
+
+	return &nvpws->pws;
+}
+
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_pipe.h b/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_pipe.h
new file mode 120000
index 0000000000..264716fef0
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_pipe.h
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_winsys_pipe.h
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_softpipe.c b/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_softpipe.c
new file mode 120000
index 0000000000..83faccde96
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nouveau_winsys_softpipe.c
@@ -0,0 +1 @@
+../../dri/nouveau/nouveau_winsys_softpipe.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nv04_surface.c b/src/gallium/winsys/g3dvl/nouveau/nv04_surface.c
new file mode 120000
index 0000000000..e05f0671d6
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nv04_surface.c
@@ -0,0 +1 @@
+../../dri/nouveau/nv04_surface.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/nouveau/nv50_surface.c b/src/gallium/winsys/g3dvl/nouveau/nv50_surface.c
new file mode 120000
index 0000000000..3850748229
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/nouveau/nv50_surface.c
@@ -0,0 +1 @@
+../../dri/nouveau/nv50_surface.c
\ No newline at end of file
diff --git a/src/gallium/winsys/g3dvl/vl_winsys.h b/src/gallium/winsys/g3dvl/vl_winsys.h
new file mode 100644
index 0000000000..c83db28dd9
--- /dev/null
+++ b/src/gallium/winsys/g3dvl/vl_winsys.h
@@ -0,0 +1,14 @@
+#ifndef vl_winsys_h
+#define vl_winsys_h
+
+#include <X11/Xlib.h>
+
+struct pipe_context;
+
+struct pipe_context* create_pipe_context(Display *display, int screen);
+int destroy_pipe_context(struct pipe_context *pipe);
+int bind_pipe_drawable(struct pipe_context *pipe, Drawable drawable);
+int unbind_pipe_drawable(struct pipe_context *pipe);
+
+#endif
+
diff --git a/src/gallium/winsys/g3dvl/xsp_winsys.c b/src/gallium/winsys/g3dvl/xsp_winsys.c
index c660c6e018..0100fe37bd 100644
--- a/src/gallium/winsys/g3dvl/xsp_winsys.c
+++ b/src/gallium/winsys/g3dvl/xsp_winsys.c
@@ -1,4 +1,4 @@
-#include "xsp_winsys.h"
+#include "vl_winsys.h"
 #include <X11/Xutil.h>
 #include <pipe/p_winsys.h>
 #include <pipe/p_state.h>
@@ -11,10 +11,17 @@
 struct xsp_pipe_winsys
 {
 	struct pipe_winsys	base;
-	Display			*display;
 	XImage			fbimage;
 };
 
+struct xsp_context
+{
+	Display			*display;
+	int			screen;
+	Drawable		drawable;
+	int			drawable_bound;
+};
+
 struct xsp_buffer
 {
 	struct pipe_buffer	base;
@@ -183,13 +190,18 @@ static int xsp_fence_finish(struct pipe_winsys *pws, struct pipe_fence_handle *f
 
 static void xsp_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *surface, void *context_private)
 {	
-	struct xsp_pipe_winsys *xsp_winsys;
+	struct xsp_pipe_winsys	*xsp_winsys;
+	struct xsp_context	*xsp_context;
 	
 	assert(pws);
 	assert(surface);
 	assert(context_private);
 	
 	xsp_winsys = (struct xsp_pipe_winsys*)pws;
+	xsp_context = (struct xsp_context*)context_private;
+	
+	if (!xsp_context->drawable_bound)
+		return;
 	
 	xsp_winsys->fbimage.width = surface->width;
 	xsp_winsys->fbimage.height = surface->height;
@@ -198,9 +210,9 @@ static void xsp_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *
 	
 	XPutImage
 	(
-		xsp_winsys->display,
-		*(Drawable*)context_private,
-		XDefaultGC(xsp_winsys->display, XDefaultScreen(xsp_winsys->display)),
+		xsp_context->display,
+		xsp_context->drawable,
+		XDefaultGC(xsp_context->display, xsp_context->screen),
 		&xsp_winsys->fbimage,
 		0,
 		0,
@@ -209,7 +221,7 @@ static void xsp_flush_frontbuffer(struct pipe_winsys *pws, struct pipe_surface *
 		surface->width,
 		surface->height
 	);
-	XFlush(xsp_winsys->display);
+	XFlush(xsp_context->display);
 	pipe_surface_unmap(surface);
 }
 
@@ -221,11 +233,37 @@ static const char* xsp_get_name(struct pipe_winsys *pws)
 
 /* Show starts here */
 
-struct pipe_context* create_pipe_context(Display *display)
+int bind_pipe_drawable(struct pipe_context *pipe, Drawable drawable)
+{
+	struct xsp_context *xsp_context;
+	
+	assert(pipe);
+	
+	xsp_context = pipe->priv;
+	xsp_context->drawable = drawable;
+	xsp_context->drawable_bound = 1;
+	
+	return 0;
+}
+
+int unbind_pipe_drawable(struct pipe_context *pipe)
+{
+	struct xsp_context *xsp_context;
+	
+	assert(pipe);
+	
+	xsp_context = pipe->priv;
+	xsp_context->drawable_bound = 0;
+	
+	return 0;
+}
+
+struct pipe_context* create_pipe_context(Display *display, int screen)
 {
 	struct xsp_pipe_winsys	*xsp_winsys;
-	struct pipe_screen	*screen;
-	struct pipe_context	*pipe;
+	struct xsp_context	*xsp_context;
+	struct pipe_screen	*sp_screen;
+	struct pipe_context	*sp_pipe;
 	
 	assert(display);
 	
@@ -243,7 +281,6 @@ struct pipe_context* create_pipe_context(Display *display)
 	xsp_winsys->base.fence_finish = xsp_fence_finish;
 	xsp_winsys->base.flush_frontbuffer = xsp_flush_frontbuffer;
 	xsp_winsys->base.get_name = xsp_get_name;
-	xsp_winsys->display = display;
 	
 	{
 		/* XXX: Can't use the returned XImage* directly,
@@ -269,10 +306,16 @@ struct pipe_context* create_pipe_context(Display *display)
 		XDestroyImage(template);
 	}
 	
-	screen = softpipe_create_screen((struct pipe_winsys*)xsp_winsys);
-	pipe = softpipe_create(screen, (struct pipe_winsys*)xsp_winsys, NULL);
+	sp_screen = softpipe_create_screen((struct pipe_winsys*)xsp_winsys);
+	sp_pipe = softpipe_create(sp_screen, (struct pipe_winsys*)xsp_winsys, NULL);
+	
+	xsp_context = calloc(1, sizeof(struct xsp_context));
+	xsp_context->display = display;
+	xsp_context->screen = screen;
+	
+	sp_pipe->priv = xsp_context;
 	
-	return pipe;
+	return sp_pipe;
 }
 
 int destroy_pipe_context(struct pipe_context *pipe)
@@ -284,6 +327,7 @@ int destroy_pipe_context(struct pipe_context *pipe)
 	
 	screen = pipe->screen;
 	winsys = pipe->winsys;
+	free(pipe->priv);
 	pipe->destroy(pipe);
 	screen->destroy(screen);
 	free(winsys);
diff --git a/src/gallium/winsys/g3dvl/xsp_winsys.h b/src/gallium/winsys/g3dvl/xsp_winsys.h
deleted file mode 100644
index cb163dc24d..0000000000
--- a/src/gallium/winsys/g3dvl/xsp_winsys.h
+++ /dev/null
@@ -1,12 +0,0 @@
-#ifndef xsp_winsys_h
-#define xsp_winsys_h
-
-#include <X11/Xlib.h>
-
-struct pipe_context;
-
-struct pipe_context* create_pipe_context(Display *display);
-int destroy_pipe_context(struct pipe_context *pipe);
-
-#endif
-
diff --git a/src/libXvMC/Makefile b/src/libXvMC/Makefile
index 4985ecd3e9..83fcfc523c 100644
--- a/src/libXvMC/Makefile
+++ b/src/libXvMC/Makefile
@@ -1,7 +1,12 @@
 TARGET		= libXvMCg3dvl.so
 SONAME		= libXvMCg3dvl.so.1
 GALLIUMDIR	= ../gallium
-OBJECTS		= block.o surface.o context.o subpicture.o attributes.o ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
+
+OBJECTS         = block.o surface.o context.o subpicture.o attributes.o
+
+ifeq (${DRIVER}, softpipe)
+OBJECTS         += ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
+endif
 
 CFLAGS	+= -g -fPIC -Wall -Werror			\
 	   -I${GALLIUMDIR}/state_trackers/g3dvl		\
@@ -9,6 +14,8 @@ CFLAGS	+= -g -fPIC -Wall -Werror			\
 	   -I${GALLIUMDIR}/include			\
 	   -I${GALLIUMDIR}/auxiliary			\
 	   -I${GALLIUMDIR}/drivers
+
+ifeq (${DRIVER}, softpipe)
 LDFLAGS	+= -L${GALLIUMDIR}/state_trackers/g3dvl		\
 	   -L${GALLIUMDIR}/drivers/softpipe		\
 	   -L${GALLIUMDIR}/auxiliary/tgsi		\
@@ -17,7 +24,17 @@ LDFLAGS	+= -L${GALLIUMDIR}/state_trackers/g3dvl		\
 	   -L${GALLIUMDIR}/auxiliary/cso_cache		\
 	   -L${GALLIUMDIR}/auxiliary/util		\
 	   -L${GALLIUMDIR}/auxiliary/rtasm
+else
+LDFLAGS	+= -L${GALLIUMDIR}/state_trackers/g3dvl		\
+   	   -L${GALLIUMDIR}/winsys/g3dvl/nouveau		\
+   	   -L${GALLIUMDIR}/auxiliary/util
+endif
+
+ifeq (${DRIVER}, softpipe)
 LIBS	+= -lg3dvl -lsoftpipe -ldraw -ltgsi -ltranslate -lrtasm -lcso_cache -lutil -lm
+else
+LIBS	+= -lg3dvl -lnouveau_dri -lutil
+endif
 
 #############################################
 
diff --git a/src/libXvMC/context.c b/src/libXvMC/context.c
index 22eadc1889..59e1cb2b25 100644
--- a/src/libXvMC/context.c
+++ b/src/libXvMC/context.c
@@ -2,7 +2,7 @@
 #include <X11/Xlib.h>
 #include <X11/extensions/XvMClib.h>
 #include <vl_context.h>
-#include <xsp_winsys.h>
+#include <vl_winsys.h>
 
 static Status Validate(Display *display, XvPortID port, int surface_type_id, unsigned int width, unsigned int height, int flags, int *chroma_format)
 {
@@ -117,7 +117,7 @@ Status XvMCCreateContext(Display *display, XvPortID port, int surface_type_id, i
 	if (ret != Success)
 		return ret;
 	
-	pipe = create_pipe_context(display);
+	pipe = create_pipe_context(display, XDefaultScreen(display));
 	
 	assert(pipe);
 	
-- 
cgit v1.2.3


From 6858dd50c9b696c1c6044f5a403000f9d20b286b Mon Sep 17 00:00:00 2001
From: Younes Manton <younes.m@gmail.com>
Date: Sat, 16 Aug 2008 13:04:23 -0400
Subject: g3dvl: Modularized rendering, refactored to accommodate VAAPI, other
 APIs.

---
 src/gallium/state_trackers/g3dvl/Makefile          |    4 +-
 src/gallium/state_trackers/g3dvl/vl_basic_csc.c    |  694 ++++++
 src/gallium/state_trackers/g3dvl/vl_basic_csc.h    |   13 +
 src/gallium/state_trackers/g3dvl/vl_context.c      | 2272 +------------------
 src/gallium/state_trackers/g3dvl/vl_context.h      |  118 +-
 src/gallium/state_trackers/g3dvl/vl_csc.h          |   53 +
 src/gallium/state_trackers/g3dvl/vl_data.c         |  130 +-
 src/gallium/state_trackers/g3dvl/vl_data.h         |   19 +-
 src/gallium/state_trackers/g3dvl/vl_defs.h         |    1 -
 src/gallium/state_trackers/g3dvl/vl_display.c      |   48 +
 src/gallium/state_trackers/g3dvl/vl_display.h      |   29 +
 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c  | 2315 ++++++++++++++++++++
 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h  |   18 +
 src/gallium/state_trackers/g3dvl/vl_render.h       |   33 +
 src/gallium/state_trackers/g3dvl/vl_screen.c       |  115 +
 src/gallium/state_trackers/g3dvl/vl_screen.h       |   63 +
 src/gallium/state_trackers/g3dvl/vl_shader_build.c |   37 +-
 src/gallium/state_trackers/g3dvl/vl_shader_build.h |    1 -
 src/gallium/state_trackers/g3dvl/vl_surface.c      |  671 +-----
 src/gallium/state_trackers/g3dvl/vl_surface.h      |   91 +-
 src/gallium/state_trackers/g3dvl/vl_types.h        |  124 +-
 src/gallium/state_trackers/g3dvl/vl_util.c         |    9 +-
 src/gallium/state_trackers/g3dvl/vl_util.h         |    1 -
 src/libXvMC/block.c                                |   47 +-
 src/libXvMC/context.c                              |  194 +-
 src/libXvMC/surface.c                              |  403 ++--
 26 files changed, 4061 insertions(+), 3442 deletions(-)
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_basic_csc.c
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_basic_csc.h
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_csc.h
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_display.c
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_display.h
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_render.h
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_screen.c
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_screen.h

(limited to 'src/gallium/state_trackers/g3dvl/Makefile')

diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
index c6a22cad4e..9995c554ab 100644
--- a/src/gallium/state_trackers/g3dvl/Makefile
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -1,5 +1,6 @@
 TARGET		= libg3dvl.a
-OBJECTS		= vl_context.o vl_data.o vl_surface.o vl_shader_build.o vl_util.o
+OBJECTS		= vl_display.o vl_screen.o vl_context.o vl_surface.o vl_data.o vl_shader_build.o vl_util.o vl_basic_csc.o \
+		  vl_r16snorm_mc.o
 GALLIUMDIR	= ../..
 
 CFLAGS		+= -g -Wall -fPIC -Werror -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary -I${GALLIUMDIR}/winsys/g3dvl
@@ -15,4 +16,3 @@ ${TARGET}: ${OBJECTS}
 
 clean:
 	rm -rf ${OBJECTS} ${TARGET}
-
diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
new file mode 100644
index 0000000000..ea003a31d1
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
@@ -0,0 +1,694 @@
+#define VL_INTERNAL
+#include "vl_basic_csc.h"
+#include <assert.h>
+#include <stdlib.h>
+#include <pipe/p_context.h>
+#include <pipe/p_winsys.h>
+#include <pipe/p_state.h>
+#include <tgsi/tgsi_parse.h>
+#include <tgsi/tgsi_build.h>
+#include "vl_csc.h"
+#include "vl_surface.h"
+#include "vl_shader_build.h"
+#include "vl_types.h"
+
+struct vlVertexShaderConsts
+{
+	struct vlVertex4f	src_scale;
+	struct vlVertex4f	src_trans;
+};
+
+struct vlFragmentShaderConsts
+{
+	struct vlVertex4f	bias;
+	float			matrix[16];
+};
+
+struct vlBasicCSC
+{
+	struct vlCSC				base;
+
+	struct pipe_context			*pipe;
+	struct pipe_viewport_state		viewport;
+	struct pipe_framebuffer_state		framebuffer;
+	void					*sampler;
+	void					*vertex_shader, *fragment_shader;
+	struct pipe_vertex_buffer 		vertex_bufs[2];
+	struct pipe_vertex_element		vertex_elems[2];
+	struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
+};
+
+static int vlResizeFrameBuffer
+(
+	struct vlCSC *csc,
+	unsigned int width,
+	unsigned int height
+)
+{
+	struct vlBasicCSC	*basic_csc;
+	struct pipe_context	*pipe;
+
+	assert(csc);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+	pipe = basic_csc->pipe;
+
+	if (basic_csc->framebuffer.width == width && basic_csc->framebuffer.height == height)
+		return 0;
+
+	if (basic_csc->framebuffer.cbufs[0])
+		pipe->winsys->surface_release
+		(
+			pipe->winsys,
+			&basic_csc->framebuffer.cbufs[0]
+		);
+
+	basic_csc->viewport.scale[0] = width;
+	basic_csc->viewport.scale[1] = height;
+	basic_csc->viewport.scale[2] = 1;
+	basic_csc->viewport.scale[3] = 1;
+	basic_csc->viewport.translate[0] = 0;
+	basic_csc->viewport.translate[1] = 0;
+	basic_csc->viewport.translate[2] = 0;
+	basic_csc->viewport.translate[3] = 0;
+
+	basic_csc->framebuffer.width = width;
+	basic_csc->framebuffer.height = height;
+	basic_csc->framebuffer.cbufs[0] = pipe->winsys->surface_alloc(pipe->winsys);
+	pipe->winsys->surface_alloc_storage
+	(
+		pipe->winsys,
+		basic_csc->framebuffer.cbufs[0],
+		width,
+		height,
+		PIPE_FORMAT_A8R8G8B8_UNORM,
+		/* XXX: SoftPipe doesn't change GPU usage to CPU like it does for textures */
+		PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE,
+		0
+	);
+
+	return 0;
+}
+
+static int vlBegin
+(
+	struct vlCSC *csc
+)
+{
+	struct vlBasicCSC	*basic_csc;
+	struct pipe_context	*pipe;
+
+	assert(csc);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+	pipe = basic_csc->pipe;
+
+	pipe->set_framebuffer_state(pipe, &basic_csc->framebuffer);
+	pipe->set_viewport_state(pipe, &basic_csc->viewport);
+	pipe->bind_sampler_states(pipe, 1, (void**)&basic_csc->sampler);
+	/* Source texture set in vlPutSurface() */
+	pipe->bind_vs_state(pipe, basic_csc->vertex_shader);
+	pipe->bind_fs_state(pipe, basic_csc->fragment_shader);
+	pipe->set_vertex_buffers(pipe, 2, basic_csc->vertex_bufs);
+	pipe->set_vertex_elements(pipe, 2, basic_csc->vertex_elems);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &basic_csc->vs_const_buf);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &basic_csc->fs_const_buf);
+
+	return 0;
+}
+
+static int vlPutPictureCSC
+(
+	struct vlCSC *csc,
+	struct vlSurface *surface,
+	int srcx,
+	int srcy,
+	int srcw,
+	int srch,
+	int destx,
+	int desty,
+	int destw,
+	int desth,
+	enum vlPictureType picture_type
+)
+{
+	struct vlBasicCSC		*basic_csc;
+	struct pipe_context		*pipe;
+	struct vlVertexShaderConsts	*vs_consts;
+
+	assert(csc);
+	assert(surface);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+	pipe = basic_csc->pipe;
+
+	vs_consts = pipe->winsys->buffer_map
+	(
+		pipe->winsys,
+		basic_csc->vs_const_buf.buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE
+	);
+
+	vs_consts->src_scale.x = srcw / (float)surface->texture->width[0];
+	vs_consts->src_scale.y = srch / (float)surface->texture->height[0];
+	vs_consts->src_scale.z = 1;
+	vs_consts->src_scale.w = 1;
+	vs_consts->src_trans.x = srcx / (float)surface->texture->width[0];
+	vs_consts->src_trans.y = srcy / (float)surface->texture->height[0];
+	vs_consts->src_trans.z = 0;
+	vs_consts->src_trans.w = 0;
+
+	pipe->winsys->buffer_unmap(pipe->winsys, basic_csc->vs_const_buf.buffer);
+
+	pipe->set_sampler_textures(pipe, 1, &surface->texture);
+	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_STRIP, 0, 4);
+
+	return 0;
+}
+
+static int vlEnd
+(
+	struct vlCSC *csc
+)
+{
+	assert(csc);
+
+	return 0;
+}
+
+static struct pipe_surface* vlGetFrameBuffer
+(
+	struct vlCSC *csc
+)
+{
+	struct vlBasicCSC	*basic_csc;
+
+	assert(csc);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+
+	return basic_csc->framebuffer.cbufs[0];
+}
+
+static int vlDestroy
+(
+	struct vlCSC *csc
+)
+{
+	struct vlBasicCSC	*basic_csc;
+	struct pipe_context	*pipe;
+	unsigned int		i;
+
+	assert(csc);
+
+	basic_csc = (struct vlBasicCSC*)csc;
+	pipe = basic_csc->pipe;
+
+	if (basic_csc->framebuffer.cbufs[0])
+		pipe->winsys->surface_release
+		(
+			pipe->winsys,
+			&basic_csc->framebuffer.cbufs[0]
+		);
+
+	pipe->delete_sampler_state(pipe, basic_csc->sampler);
+	pipe->delete_vs_state(pipe, basic_csc->vertex_shader);
+	pipe->delete_fs_state(pipe, basic_csc->fragment_shader);
+
+	for (i = 0; i < 2; ++i)
+		pipe->winsys->buffer_destroy(pipe->winsys, basic_csc->vertex_bufs[i].buffer);
+
+	pipe->winsys->buffer_destroy(pipe->winsys, basic_csc->vs_const_buf.buffer);
+	pipe->winsys->buffer_destroy(pipe->winsys, basic_csc->fs_const_buf.buffer);
+
+	free(basic_csc);
+
+	return 0;
+}
+
+/*
+ * Represents 2 triangles in a strip in normalized coords.
+ * Used to render the surface onto the frame buffer.
+ */
+static const struct vlVertex2f surface_verts[4] =
+{
+	{0.0f, 0.0f},
+	{0.0f, 1.0f},
+	{1.0f, 0.0f},
+	{1.0f, 1.0f}
+};
+
+/*
+ * Represents texcoords for the above. We can use the position values directly.
+ * TODO: Duplicate these in the shader, no need to create a buffer.
+ */
+static const struct vlVertex2f *surface_texcoords = surface_verts;
+
+/*
+ * Identity color conversion constants, for debugging
+ */
+static const struct vlFragmentShaderConsts identity =
+{
+	{
+		0.0f, 0.0f, 0.0f, 0.0f
+	},
+	{
+		1.0f, 0.0f, 0.0f, 0.0f,
+		0.0f, 1.0f, 0.0f, 0.0f,
+		0.0f, 0.0f, 1.0f, 0.0f,
+		0.0f, 0.0f, 0.0f, 1.0f
+	}
+};
+
+/*
+ * Converts ITU-R BT.601 YCbCr pixels to RGB pixels where:
+ * Y is in [16,235], Cb and Cr are in [16,240]
+ * R, G, and B are in [16,235]
+ */
+static const struct vlFragmentShaderConsts bt_601 =
+{
+	{
+		0.0f,		0.501960784f,	0.501960784f,	0.0f
+	},
+	{
+		1.0f,		0.0f,		1.371f,		0.0f,
+		1.0f,		-0.336f,	-0.698f,	0.0f,
+		1.0f,		1.732f,		0.0f,		0.0f,
+		0.0f,		0.0f,		0.0f,		1.0f
+	}
+};
+
+/*
+ * Converts ITU-R BT.601 YCbCr pixels to RGB pixels where:
+ * Y is in [16,235], Cb and Cr are in [16,240]
+ * R, G, and B are in [0,255]
+ */
+static const struct vlFragmentShaderConsts bt_601_full =
+{
+	{
+		0.062745098f,	0.501960784f,	0.501960784f,	0.0f
+	},
+	{
+		1.164f,		0.0f,		1.596f,		0.0f,
+		1.164f,		-0.391f,	-0.813f,	0.0f,
+		1.164f,		2.018f,		0.0f,		0.0f,
+		0.0f,		0.0f,		0.0f,		1.0f
+	}
+};
+
+/*
+ * Converts ITU-R BT.709 YCbCr pixels to RGB pixels where:
+ * Y is in [16,235], Cb and Cr are in [16,240]
+ * R, G, and B are in [16,235]
+ */
+static const struct vlFragmentShaderConsts bt_709 =
+{
+	{
+		0.0f,		0.501960784f,	0.501960784f,	0.0f
+	},
+	{
+		1.0f,		0.0f,		1.540f,		0.0f,
+		1.0f,		-0.183f,	-0.459f,	0.0f,
+		1.0f,		1.816f,		0.0f,		0.0f,
+		0.0f,		0.0f,		0.0f,		1.0f
+	}
+};
+
+/*
+ * Converts ITU-R BT.709 YCbCr pixels to RGB pixels where:
+ * Y is in [16,235], Cb and Cr are in [16,240]
+ * R, G, and B are in [0,255]
+ */
+const struct vlFragmentShaderConsts bt_709_full =
+{
+	{
+		0.062745098f,	0.501960784f,	0.501960784f,	0.0f
+	},
+	{
+		1.164f,		0.0f,		1.793f,		0.0f,
+		1.164f,		-0.213f,	-0.534f,	0.0f,
+		1.164f,		2.115f,		0.0f,		0.0f,
+		0.0f,		0.0f,		0.0f,		1.0f
+	}
+};
+
+static int vlCreateVertexShader
+(
+	struct vlBasicCSC *csc
+)
+{
+	const unsigned int		max_tokens = 50;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(context);
+
+	pipe = csc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Vertex texcoords
+	 */
+	for (i = 0; i < 2; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0		; Scaling vector to scale texcoord rect to source size
+	 * decl c1		; Translation vector to move texcoord rect into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Vertex texcoords
+	 */
+	for (i = 0; i < 2; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* mov o0, i0		; Move pos in to pos out */
+	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_INPUT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t0, i1, c0	; Scale unit texcoord rect to source size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 1, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o1, t0, c1	; Translate texcoord rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	csc->vertex_shader = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShader
+(
+	struct vlBasicCSC *csc
+)
+{
+	const unsigned int		max_tokens = 50;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(context);
+
+	pipe = csc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/* decl i0		; Texcoords for s0 */
+	decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, 1, 0, 0, TGSI_INTERPOLATE_LINEAR);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl c0		; Bias vector for CSC
+	 * decl c1-c4		; CSC matrix c1-c4
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0		; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl s0		; Sampler for tex containing picture to display */
+	decl = vl_decl_samplers(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* tex2d t0, i0, s0	; Read src pixel */
+	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* sub t0, t0, c0	; Subtract bias vector from pixel */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * dp4 o0.x, t0, c1	; Multiply pixel by the color conversion matrix
+	 * dp4 o0.y, t0, c2
+	 * dp4 o0.z, t0, c3
+	 * dp4 o0.w, t0, c4	; XXX: Don't need 4th coefficient
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_DP4, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 1);
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	csc->fragment_shader = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateDataBufs
+(
+	struct vlBasicCSC *csc
+)
+{
+	struct pipe_context *pipe;
+
+	assert(csc);
+
+	pipe = csc->pipe;
+
+	/*
+	Create our vertex buffer and vertex buffer element
+	VB contains 4 vertices that render a quad covering the entire window
+	to display a rendered surface
+	Quad is rendered as a tri strip
+	*/
+	csc->vertex_bufs[0].pitch = sizeof(struct vlVertex2f);
+	csc->vertex_bufs[0].max_index = 3;
+	csc->vertex_bufs[0].buffer_offset = 0;
+	csc->vertex_bufs[0].buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_VERTEX,
+		sizeof(struct vlVertex2f) * 4
+	);
+
+	memcpy
+	(
+		pipe->winsys->buffer_map(pipe->winsys, csc->vertex_bufs[0].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		surface_verts,
+		sizeof(struct vlVertex2f) * 4
+	);
+
+	pipe->winsys->buffer_unmap(pipe->winsys, csc->vertex_bufs[0].buffer);
+
+	csc->vertex_elems[0].src_offset = 0;
+	csc->vertex_elems[0].vertex_buffer_index = 0;
+	csc->vertex_elems[0].nr_components = 2;
+	csc->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/*
+	Create our texcoord buffer and texcoord buffer element
+	Texcoord buffer contains the TCs for mapping the rendered surface to the 4 vertices
+	*/
+	csc->vertex_bufs[1].pitch = sizeof(struct vlVertex2f);
+	csc->vertex_bufs[1].max_index = 3;
+	csc->vertex_bufs[1].buffer_offset = 0;
+	csc->vertex_bufs[1].buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_VERTEX,
+		sizeof(struct vlVertex2f) * 4
+	);
+
+	memcpy
+	(
+		pipe->winsys->buffer_map(pipe->winsys, csc->vertex_bufs[1].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		surface_texcoords,
+		sizeof(struct vlVertex2f) * 4
+	);
+
+	pipe->winsys->buffer_unmap(pipe->winsys, csc->vertex_bufs[1].buffer);
+
+	csc->vertex_elems[1].src_offset = 0;
+	csc->vertex_elems[1].vertex_buffer_index = 1;
+	csc->vertex_elems[1].nr_components = 2;
+	csc->vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/*
+	Create our vertex shader's constant buffer
+	Const buffer contains scaling and translation vectors
+	*/
+	csc->vs_const_buf.size = sizeof(struct vlVertexShaderConsts);
+	csc->vs_const_buf.buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_CONSTANT,
+		csc->vs_const_buf.size
+	);
+
+	/*
+	Create our fragment shader's constant buffer
+	Const buffer contains the color conversion matrix and bias vectors
+	*/
+	csc->fs_const_buf.size = sizeof(struct vlFragmentShaderConsts);
+	csc->fs_const_buf.buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_CONSTANT,
+		csc->fs_const_buf.size
+	);
+
+	/*
+	TODO: Refactor this into a seperate function,
+	allow changing the CSC matrix at runtime to switch between regular & full versions
+	*/
+	memcpy
+	(
+		pipe->winsys->buffer_map(pipe->winsys, csc->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		&bt_601,
+		sizeof(struct vlFragmentShaderConsts)
+	);
+
+	pipe->winsys->buffer_unmap(pipe->winsys, csc->fs_const_buf.buffer);
+
+	return 0;
+}
+
+static int vlInit
+(
+	struct vlBasicCSC *csc
+)
+{
+	struct pipe_context		*pipe;
+	struct pipe_sampler_state	sampler;
+
+	assert(csc);
+
+	pipe = csc->pipe;
+
+	/* Delay creating the FB until vlPutSurface() so we know window size */
+	csc->framebuffer.num_cbufs = 1;
+	csc->framebuffer.cbufs[0] = NULL;
+	csc->framebuffer.zsbuf = NULL;
+
+	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_img_filter = PIPE_TEX_FILTER_LINEAR;
+	sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+	sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR;
+	sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
+	sampler.compare_func = PIPE_FUNC_ALWAYS;
+	sampler.normalized_coords = 1;
+	/*sampler.prefilter = ;*/
+	/*sampler.shadow_ambient = ;*/
+	/*sampler.lod_bias = ;*/
+	/*sampler.min_lod = ;*/
+	/*sampler.max_lod = ;*/
+	/*sampler.border_color[i] = ;*/
+	/*sampler.max_anisotropy = ;*/
+	csc->sampler = pipe->create_sampler_state(pipe, &sampler);
+
+	vlCreateVertexShader(csc);
+	vlCreateFragmentShader(csc);
+	vlCreateDataBufs(csc);
+
+	return 0;
+}
+
+int vlCreateBasicCSC
+(
+	struct pipe_context *pipe,
+	struct vlCSC **csc
+)
+{
+	struct vlBasicCSC *basic_csc;
+
+	assert(pipe);
+	assert(csc);
+
+	basic_csc = calloc(1, sizeof(struct vlBasicCSC));
+
+	if (!basic_csc)
+		return 1;
+
+	basic_csc->base.vlResizeFrameBuffer = &vlResizeFrameBuffer;
+	basic_csc->base.vlBegin = &vlBegin;
+	basic_csc->base.vlPutPicture = &vlPutPictureCSC;
+	basic_csc->base.vlEnd = &vlEnd;
+	basic_csc->base.vlGetFrameBuffer = &vlGetFrameBuffer;
+	basic_csc->base.vlDestroy = &vlDestroy;
+	basic_csc->pipe = pipe;
+
+	vlInit(basic_csc);
+
+	*csc = &basic_csc->base;
+
+	return 0;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.h b/src/gallium/state_trackers/g3dvl/vl_basic_csc.h
new file mode 100644
index 0000000000..2e17f1d814
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.h
@@ -0,0 +1,13 @@
+#ifndef vl_basic_csc_h
+#define vl_basic_csc_h
+
+struct pipe_context;
+struct vlCSC;
+
+int vlCreateBasicCSC
+(
+	struct pipe_context *pipe,
+	struct vlCSC **csc
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c
index 5616de0ba4..56d360c05b 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.c
+++ b/src/gallium/state_trackers/g3dvl/vl_context.c
@@ -1,2088 +1,26 @@
+#define VL_INTERNAL
 #include "vl_context.h"
 #include <assert.h>
 #include <stdlib.h>
 #include <pipe/p_context.h>
-#include <pipe/p_winsys.h>
-#include <pipe/p_screen.h>
 #include <pipe/p_state.h>
-#include <pipe/p_inlines.h>
-#include <pipe/p_shader_tokens.h>
-#include <tgsi/tgsi_parse.h>
-#include <tgsi/tgsi_build.h>
-#include "vl_shader_build.h"
-#include "vl_data.h"
-#include "vl_defs.h"
-#include "vl_util.h"
+#include "vl_render.h"
+#include "vl_r16snorm_mc.h"
+#include "vl_csc.h"
+#include "vl_basic_csc.h"
 
-static int vlCreateVertexShaderFrameIDCT(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 50;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-	
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Vertex texcoords
-	 */
-	for (i = 0; i < 2; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Vertex texcoords
-	 */
-	for (i = 0; i < 2; i++)
-	{
-		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * mov o0, i0		; Move pos in to pos out
-	 * mov o1, i1		; Move texcoord in to texcoord out */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	//context->states.idct.frame_vs = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateFragmentShaderFrameIDCT(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 50;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/* decl i0		; Texcoords for s0 */
-	decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, 1, 0, 0, TGSI_INTERPOLATE_LINEAR);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0		; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* decl s0		; Sampler for tex containing picture to display */
-	decl = vl_decl_samplers(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* tex2d t0, i0, s0	; Read src pixel */
-	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* sub t0, t0, c0	; Subtract bias vector from pixel */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * dp4 o0.x, t0, c1	; Multiply pixel by the color conversion matrix
-	 * dp4 o0.y, t0, c2
-	 * dp4 o0.z, t0, c3
-	 * dp4 o0.w, t0, c4	; XXX: Don't need 4th coefficient
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		inst = vl_inst3(TGSI_OPCODE_DP4, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 1);
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	//context->states.idct.frame_fs = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlInitIDCT(struct VL_CONTEXT *context)
-{
-	struct pipe_context		*pipe;
-	struct pipe_sampler_state	sampler;
-	struct pipe_texture		template;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	
-	context->states.idct.viewport.scale[0] = VL_BLOCK_WIDTH;
-	context->states.idct.viewport.scale[1] = VL_BLOCK_HEIGHT;
-	context->states.idct.viewport.scale[2] = 1;
-	context->states.idct.viewport.scale[3] = 1;
-	context->states.idct.viewport.translate[0] = 0;
-	context->states.idct.viewport.translate[1] = 0;
-	context->states.idct.viewport.translate[2] = 0;
-	context->states.idct.viewport.translate[3] = 0;
-	
-	context->states.idct.render_target.width = VL_BLOCK_WIDTH;
-	context->states.idct.render_target.height = VL_BLOCK_HEIGHT;
-	context->states.idct.render_target.num_cbufs = 1;
-	context->states.idct.render_target.zsbuf = NULL;
-	
-	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_img_filter = PIPE_TEX_FILTER_NEAREST;
-	sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
-	sampler.mag_img_filter = PIPE_TEX_FILTER_NEAREST;
-	sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
-	sampler.compare_func = PIPE_FUNC_ALWAYS;
-	sampler.normalized_coords = 1;
-	/*sampler.prefilter = ;*/
-	/*sampler.shadow_ambient = ;*/
-	/*sampler.lod_bias = ;*/
-	sampler.min_lod = 0;
-	/*sampler.max_lod = ;*/
-	/*sampler.border_color[i] = ;*/
-	/*sampler.max_anisotropy = ;*/
-	context->states.idct.sampler = pipe->create_sampler_state(pipe, &sampler);
-	
-	memset(&template, 0, sizeof(struct pipe_texture));
-	template.target = PIPE_TEXTURE_2D;
-	template.format = PIPE_FORMAT_A8L8_UNORM;
-	template.last_level = 0;
-	template.width[0] = 8;
-	template.height[0] = 8;
-	template.depth[0] = 1;
-	template.compressed = 0;
-	pf_get_block(template.format, &template.block);
-	
-	context->states.idct.texture = pipe->screen->texture_create(pipe->screen, &template);
-	
-	template.format = PIPE_FORMAT_A8R8G8B8_UNORM;
-	template.width[0] = 16;
-	template.height[0] = 1;
-	
-	context->states.idct.basis = pipe->screen->texture_create(pipe->screen, &template);
-	
-	for (i = 0; i < 2; ++i)
-	{
-		context->states.idct.vertex_bufs[i] = &context->states.csc.vertex_bufs[i];
-		context->states.idct.vertex_buf_elems[i] = &context->states.csc.vertex_buf_elems[i];
-		/*
-		context->states.idct.vertex_bufs[i].pitch = sizeof(struct VL_VERTEX2F);
-		context->states.idct.vertex_bufs[i].max_index = 3;
-		context->states.idct.vertex_bufs[i].buffer_offset = 0;
-		context->states.idct.vertex_bufs[i].buffer = pipe->winsys->buffer_create
-		(
-			pipe->winsys,
-			1,
-			PIPE_BUFFER_USAGE_VERTEX,
-			sizeof(struct VL_VERTEX2F) * 4
-		);
-	
-		context->states.idct.vertex_buf_elems[i].src_offset = 0;
-		context->states.idct.vertex_buf_elems[i].vertex_buffer_index = i;
-		context->states.idct.vertex_buf_elems[i].nr_components = 2;
-		context->states.idct.vertex_buf_elems[i].src_format = PIPE_FORMAT_R32G32_FLOAT;
-		*/
-	}
-	
-	vlCreateVertexShaderFrameIDCT(context);
-	vlCreateFragmentShaderFrameIDCT(context);
-	
-	return 0;
-}
-
-static int vlDestroyIDCT(struct VL_CONTEXT *context)
-{
-	//unsigned int i;
-	
-	assert(context);
-	
-	context->pipe->delete_sampler_state(context->pipe, context->states.idct.sampler);
-	
-	//for (i = 0; i < 2; ++i)
-		//context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.idct.vertex_bufs[i].buffer);
-	
-	pipe_texture_release(&context->states.idct.texture);
-	pipe_texture_release(&context->states.idct.basis);
-	
-	//context->pipe->delete_vs_state(context->pipe, context->states.idct.frame_vs);
-	//context->pipe->delete_fs_state(context->pipe, context->states.idct.frame_fs);
-	
-	//context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.idct.vs_const_buf.buffer);
-	//context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.idct.fs_const_buf.buffer);
-	
-	return 0;
-}
-
-static int vlCreateVertexShaderIMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 50;
-	
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-	
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* decl t0 */
-	decl = vl_decl_temps(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * mov o1, i1		; Move input luma texcoords to output
-	 * mov o2, i2		; Move input chroma texcoords to output
-	 */
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	vs.tokens = tokens;
-	context->states.mc.i_vs = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateFragmentShaderIMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 100;
-	
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* decl t0, t1 */
-	decl = vl_decl_temps(0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header,max_tokens - ti);
-	}
-	
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-	}
-	
-	/* mul o0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	context->states.mc.i_fs = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateVertexShaderFramePMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 100;
-	
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-	
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 * decl c2		; Unused
-	 * decl c3		; Translation vector to move ref macroblock texcoords into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 3);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 * decl o3		; Ref macroblock texcoords
-	 */
-	for (i = 0; i < 4; i++)
-	{
-		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* decl t0 */
-	decl = vl_decl_temps(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * mov o1, i1		; Move input luma texcoords to output
-	 * mov o2, i2		; Move input chroma texcoords to output
-	 */
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* add o3, t0, c3	; Translate rect into position on ref macroblock */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	vs.tokens = tokens;
-	context->states.mc.p_vs[0] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateVertexShaderFieldPMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 100;
-	
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-	
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
-	}
-	
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 * decl c2		; Denorm coefficients
-	 * decl c3		; Translation vector to move top field ref macroblock texcoords into position
-	 * decl c4		; Translation vector to move bottom field ref macroblock texcoords into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 * decl o3		; Top field ref macroblock texcoords
-	 * decl o4		; Bottom field ref macroblock texcoords
-	 * decl o5		; Denormalized vertex pos
-	 */
-	for (i = 0; i < 6; i++)
-	{
-		decl = vl_decl_output((i == 0 || i == 5) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* decl t0, t1 */
-	decl = vl_decl_temps(0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add t1, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* mov o0, t1		; Move vertex pos to output */
-	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	mov o1, i1		; Move input luma texcoords to output
-	mov o2, i2		; Move input chroma texcoords to output
-	*/
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* add o3, t0, c3	; Translate top field rect into position on ref macroblock
-	   add o4, t0, c4	; Translate bottom field rect into position on ref macroblock */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* mul o5, t1, c2	; Denorm vertex pos */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 5, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	vs.tokens = tokens;
-	context->states.mc.p_vs[1] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateFragmentShaderFramePMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 100;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 * decl i2			; Texcoords for s3
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* decl t0, t1 */
-	decl = vl_decl_temps(0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 * decl s3			; Sampler for ref surface texture
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-	}
-	
-	/* mul t0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* tex2d t1, i2, s3		; Read texel from ref macroblock */
-	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 2, TGSI_FILE_SAMPLER, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* add o0, t0, t1		; Add ref and differential to form final output */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	context->states.mc.p_fs[0] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateFragmentShaderFieldPMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 200;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 * decl i2			; Texcoords for s3
-	 * decl i3			; Texcoords for s3
-	 * decl i4			; Denormalized vertex pos
-	 */
-	for (i = 0; i < 5; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
-	 * decl c1			; Constants 1/2 & 2 in .x, .y channels for Y-mod-2 top/bottom field selection
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* decl t0-t4 */
-	decl = vl_decl_temps(0, 4);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 * decl s3			; Sampler for ref surface texture
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-	}
-	
-	/* mul t0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * tex2d t1, i2, s3		; Read texel from ref macroblock top field
-	 * tex2d t2, i3, s3		; Read texel from ref macroblock bottom field
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* XXX: Pos values off by 0.5? */
-	/* sub t4, i4.y, c1.x		; Sub 0.5 from denormalized pos */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 4, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* floor t3, t3			; Get rid of fractional part */
-	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t3, t3, c1.y		; Multiply by 2 */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
-	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* add o0, t0, t1		; Add ref and differential to form final output */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	context->states.mc.p_fs[1] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateVertexShaderFrameBMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 100;
-	
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-	
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 * decl c2		; Unused
-	 * decl c3		; Translation vector to move past ref macroblock texcoords into position
-	 * decl c4		; Unused
-	 * decl c5		; Translation vector to move future ref macroblock texcoords into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 5);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 * decl o3		; Past ref macroblock texcoords
-	 * decl o4		; Future ref macroblock texcoords
-	 */
-	for (i = 0; i < 5; i++)
-	{
-		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* decl t0 */
-	decl = vl_decl_temps(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * mov o1, i1		; Move input luma texcoords to output
-	 * mov o2, i2		; Move input chroma texcoords to output
-	 */
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* add o3, t0, c3	; Translate rect into position on past ref macroblock
-	   add o4, t0, c5	; Translate rect into position on future ref macroblock */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i * 2 + 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	vs.tokens = tokens;
-	context->states.mc.b_vs[0] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateVertexShaderFieldBMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 100;
-	
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-	
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;	
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 * decl c2		; Denorm coefficients
-	 * decl c3		; Translation vector to move top field past ref macroblock texcoords into position
-	 * decl c4		; Translation vector to move bottom field past ref macroblock texcoords into position
-	 * decl c5		; Translation vector to move top field future ref macroblock texcoords into position
-	 * decl c6		; Translation vector to move bottom field future ref macroblock texcoords into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 6);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 * decl o3		; Top field past ref macroblock texcoords
-	 * decl o4		; Bottom field past ref macroblock texcoords
-	 * decl o5		; Top field future ref macroblock texcoords
-	 * decl o6		; Bottom field future ref macroblock texcoords
-	 * decl o7		; Denormalized vertex pos
-	 */
-	for (i = 0; i < 8; i++)
-	{
-		decl = vl_decl_output((i == 0 || i == 7) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* decl t0, t1 */
-	decl = vl_decl_temps(0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add t1, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* mov o0, t1		; Move vertex pos to output */
-	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * mov o1, i1		; Move input luma texcoords to output
-	 * mov o2, i2		; Move input chroma texcoords to output
-	 */
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * add o3, t0, c3	; Translate top field rect into position on past ref macroblock
-	 * add o4, t0, c4	; Translate bottom field rect into position on past ref macroblock
-	 * add o5, t0, c5	; Translate top field rect into position on future ref macroblock
-	 * add o6, t0, c6	; Translate bottom field rect into position on future ref macroblock
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* mul o7, t1, c2	; Denorm vertex pos */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 7, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	vs.tokens = tokens;
-	context->states.mc.b_vs[1] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateFragmentShaderFrameBMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 100;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 * decl i2			; Texcoords for s3
-	 * decl i3			; Texcoords for s4
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
-	 * decl c1			; Constant 1/2 in .x channel to use as weight to blend past and future texels
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* decl t0-t2 */
-	decl = vl_decl_temps(0, 2);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 * decl s3			; Sampler for past ref surface texture
-	 * decl s4			; Sampler for future ref surface texture
-	 */
-	for (i = 0; i < 5; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-	}
-	
-	/* mul t0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * tex2d t1, i2, s3		; Read texel from past ref macroblock
-	 * tex2d t2, i3, s4		; Read texel from future ref macroblock
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, i + 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	context->states.mc.b_fs[0] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateFragmentShaderFieldBMC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 200;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 * decl i2			; Texcoords for s3
-	 * decl i3			; Texcoords for s3
-	 * decl i4			; Texcoords for s4
-	 * decl i5			; Texcoords for s4
-	 * decl i6			; Denormalized vertex pos
-	 */
-	for (i = 0; i < 7; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
-	 * decl c1			; Constants 1/2 & 2 in .x, .y channels to use as weight to blend past and future texels
-	 *				; and for Y-mod-2 top/bottom field selection
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* decl t0-t5 */
-	decl = vl_decl_temps(0, 5);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 * decl s3			; Sampler for past ref surface texture
-	 * decl s4			; Sampler for future ref surface texture
-	 */
-	for (i = 0; i < 5; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-		
-	}
-	
-	/* mul t0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* XXX: Pos values off by 0.5? */
-	/* sub t4, i6.y, c1.x		; Sub 0.5 from denormalized pos */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 6, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* floor t3, t3			; Get rid of fractional part */
-	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t3, t3, c1.y		; Multiply by 2 */
-	inst = vl_inst3( TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * tex2d t1, i2, s3		; Read texel from past ref macroblock top field
-	 * tex2d t2, i3, s3		; Read texel from past ref macroblock bottom field
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
-	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * tex2d t4, i4, s4		; Read texel from future ref macroblock top field
-	 * tex2d t5, i5, s4		; Read texel from future ref macroblock bottom field
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 4, TGSI_FILE_INPUT, i + 4, TGSI_FILE_SAMPLER, 4);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
-	/* lerp t2, t3, t4, t5		; Choose between top and bottom fields based on Y % 2 */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 2, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 5);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	context->states.mc.b_fs[1] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-	
-	return 0;
-}
-
-int vlCreateDataBufsMC(struct VL_CONTEXT *context)
-{
-	struct pipe_context	*pipe;
-	unsigned int		i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	
-	/* Create our vertex buffer and vertex buffer element */
-	context->states.mc.vertex_bufs[0].pitch = sizeof(struct VL_VERTEX2F);
-	context->states.mc.vertex_bufs[0].max_index = 23;
-	context->states.mc.vertex_bufs[0].buffer_offset = 0;
-	context->states.mc.vertex_bufs[0].buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_VERTEX,
-		sizeof(struct VL_VERTEX2F) * 24
-	);
-	
-	context->states.mc.vertex_buf_elems[0].src_offset = 0;
-	context->states.mc.vertex_buf_elems[0].vertex_buffer_index = 0;
-	context->states.mc.vertex_buf_elems[0].nr_components = 2;
-	context->states.mc.vertex_buf_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
-	
-	/* Create our texcoord buffers and texcoord buffer elements */
-	for (i = 1; i < 3; ++i)
-	{
-		context->states.mc.vertex_bufs[i].pitch = sizeof(struct VL_TEXCOORD2F);
-		context->states.mc.vertex_bufs[i].max_index = 23;
-		context->states.mc.vertex_bufs[i].buffer_offset = 0;
-		context->states.mc.vertex_bufs[i].buffer = pipe->winsys->buffer_create
-		(
-			pipe->winsys,
-			1,
-			PIPE_BUFFER_USAGE_VERTEX,
-			sizeof(struct VL_TEXCOORD2F) * 24
-		);
-	
-		context->states.mc.vertex_buf_elems[i].src_offset = 0;
-		context->states.mc.vertex_buf_elems[i].vertex_buffer_index = i;
-		context->states.mc.vertex_buf_elems[i].nr_components = 2;
-		context->states.mc.vertex_buf_elems[i].src_format = PIPE_FORMAT_R32G32_FLOAT;
-	}
-	
-	/* Fill buffers */
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, context->states.mc.vertex_bufs[0].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		vl_chroma_420_texcoords,
-		sizeof(struct VL_VERTEX2F) * 24
-	);
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, context->states.mc.vertex_bufs[1].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		vl_luma_texcoords,
-		sizeof(struct VL_TEXCOORD2F) * 24
-	);
-	/* TODO: Accomodate 422, 444 */
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, context->states.mc.vertex_bufs[2].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		vl_chroma_420_texcoords,
-		sizeof(struct VL_TEXCOORD2F) * 24
-	);
-	
-	for (i = 0; i < 3; ++i)
-		pipe->winsys->buffer_unmap(pipe->winsys, context->states.mc.vertex_bufs[i].buffer);
-	
-	/* Create our constant buffer */
-	context->states.mc.vs_const_buf.size = sizeof(struct VL_MC_VS_CONSTS);
-	context->states.mc.vs_const_buf.buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_CONSTANT,
-		context->states.mc.vs_const_buf.size
-	);
-	
-	context->states.mc.fs_const_buf.size = sizeof(struct VL_MC_FS_CONSTS);
-	context->states.mc.fs_const_buf.buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_CONSTANT,
-		context->states.mc.fs_const_buf.size
-	);
-	
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, context->states.mc.fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		&vl_mc_fs_consts,
-		sizeof(struct VL_MC_FS_CONSTS)
-	);
-	
-	pipe->winsys->buffer_unmap(pipe->winsys, context->states.mc.fs_const_buf.buffer);
-	
-	return 0;
-}
-
-static int vlInitMC(struct VL_CONTEXT *context)
-{	
-	struct pipe_context		*pipe;
-	struct pipe_sampler_state	sampler;
-	struct pipe_texture		template;
-	unsigned int			filters[5];
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	
-	/* For MC we render to textures, which are rounded up to nearest POT */
-	context->states.mc.viewport.scale[0] = vlRoundUpPOT(context->video_width);
-	context->states.mc.viewport.scale[1] = vlRoundUpPOT(context->video_height);
-	context->states.mc.viewport.scale[2] = 1;
-	context->states.mc.viewport.scale[3] = 1;
-	context->states.mc.viewport.translate[0] = 0;
-	context->states.mc.viewport.translate[1] = 0;
-	context->states.mc.viewport.translate[2] = 0;
-	context->states.mc.viewport.translate[3] = 0;
-	
-	context->states.mc.render_target.width = vlRoundUpPOT(context->video_width);
-	context->states.mc.render_target.height = vlRoundUpPOT(context->video_height);
-	context->states.mc.render_target.num_cbufs = 1;
-	/* FB for MC stage is a VL_SURFACE, set in vlSetRenderSurface() */
-	context->states.mc.render_target.zsbuf = NULL;
-	
-	filters[0] = PIPE_TEX_FILTER_NEAREST;
-	filters[1] = context->video_format == VL_FORMAT_YCBCR_444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
-	filters[2] = context->video_format == VL_FORMAT_YCBCR_444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
-	filters[3] = PIPE_TEX_FILTER_LINEAR;
-	filters[4] = PIPE_TEX_FILTER_LINEAR;
-	
-	for (i = 0; i < 5; ++i)
-	{
-		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_img_filter = filters[i];
-		sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
-		sampler.mag_img_filter = filters[i];
-		sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
-		sampler.compare_func = PIPE_FUNC_ALWAYS;
-		sampler.normalized_coords = 1;
-		/*sampler.prefilter = ;*/
-		/*sampler.shadow_ambient = ;*/
-		/*sampler.lod_bias = ;*/
-		sampler.min_lod = 0;
-		/*sampler.max_lod = ;*/
-		/*sampler.border_color[i] = ;*/
-		/*sampler.max_anisotropy = ;*/
-		context->states.mc.samplers[i] = pipe->create_sampler_state(pipe, &sampler);
-	}
-	
-	memset(&template, 0, sizeof(struct pipe_texture));
-	template.target = PIPE_TEXTURE_2D;
-	template.format = PIPE_FORMAT_R16_SNORM;
-	template.last_level = 0;
-	template.width[0] = 8;
-	template.height[0] = 8 * 4;
-	template.depth[0] = 1;
-	template.compressed = 0;
-	pf_get_block(template.format, &template.block);
-	
-	context->states.mc.textures[0] = pipe->screen->texture_create(pipe->screen, &template);
-	
-	if (context->video_format == VL_FORMAT_YCBCR_420)
-		template.height[0] = 8;
-	else if (context->video_format == VL_FORMAT_YCBCR_422)
-		template.height[0] = 8 * 2;
-	else if (context->video_format == VL_FORMAT_YCBCR_444)
-		template.height[0] = 8 * 4;
-	else
-		assert(0);
-		
-	context->states.mc.textures[1] = pipe->screen->texture_create(pipe->screen, &template);
-	context->states.mc.textures[2] = pipe->screen->texture_create(pipe->screen, &template);
-	
-	/* textures[3] & textures[4] are assigned from VL_SURFACEs for P and B macroblocks at render time */
-	
-	vlCreateVertexShaderIMC(context);
-	vlCreateFragmentShaderIMC(context);
-	vlCreateVertexShaderFramePMC(context);
-	vlCreateVertexShaderFieldPMC(context);
-	vlCreateFragmentShaderFramePMC(context);
-	vlCreateFragmentShaderFieldPMC(context);
-	vlCreateVertexShaderFrameBMC(context);
-	vlCreateVertexShaderFieldBMC(context);
-	vlCreateFragmentShaderFrameBMC(context);
-	vlCreateFragmentShaderFieldBMC(context);
-	vlCreateDataBufsMC(context);
-	
-	return 0;
-}
-
-static int vlDestroyMC(struct VL_CONTEXT *context)
-{
-	unsigned int i;
-	
-	assert(context);
-	
-	for (i = 0; i < 5; ++i)
-		context->pipe->delete_sampler_state(context->pipe, context->states.mc.samplers[i]);
-	
-	for (i = 0; i < 3; ++i)
-		context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.mc.vertex_bufs[i].buffer);
-	
-	/* Textures 3 & 4 are not created directly, no need to release them here */
-	for (i = 0; i < 3; ++i)
-		pipe_texture_release(&context->states.mc.textures[i]);
-	
-	context->pipe->delete_vs_state(context->pipe, context->states.mc.i_vs);
-	context->pipe->delete_fs_state(context->pipe, context->states.mc.i_fs);
-	
-	for (i = 0; i < 2; ++i)
-	{
-		context->pipe->delete_vs_state(context->pipe, context->states.mc.p_vs[i]);
-		context->pipe->delete_fs_state(context->pipe, context->states.mc.p_fs[i]);
-		context->pipe->delete_vs_state(context->pipe, context->states.mc.b_vs[i]);
-		context->pipe->delete_fs_state(context->pipe, context->states.mc.b_fs[i]);
-	}
-	
-	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.mc.vs_const_buf.buffer);
-	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.mc.fs_const_buf.buffer);
-	
-	return 0;
-}
-
-static int vlCreateVertexShaderCSC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 50;
-	
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-	
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Vertex texcoords
-	 */
-	for (i = 0; i < 2; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/*
-	 * decl c0		; Scaling vector to scale texcoord rect to source size
-	 * decl c1		; Translation vector to move texcoord rect into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Vertex texcoords
-	 */
-	for (i = 0; i < 2; i++)
-	{
-		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-	
-	/* decl t0 */
-	decl = vl_decl_temps(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* mov o0, i0		; Move pos in to pos out */
-	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_INPUT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* mul t0, i1, c0	; Scale unit texcoord rect to source size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 1, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o1, t0, c1	; Translate texcoord rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	vs.tokens = tokens;
-	context->states.csc.vertex_shader = pipe->create_vs_state(pipe, &vs);
-	//free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateFragmentShaderCSC(struct VL_CONTEXT *context)
-{
-	const unsigned int		max_tokens = 50;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-	
-	unsigned int			ti;
-	unsigned int			i;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/* decl i0		; Texcoords for s0 */
-	decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, 1, 0, 0, TGSI_INTERPOLATE_LINEAR);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * decl c0		; Bias vector for CSC
-	 * decl c1-c4		; CSC matrix c1-c4
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0		; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* decl t0 */
-	decl = vl_decl_temps(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* decl s0		; Sampler for tex containing picture to display */
-	decl = vl_decl_samplers(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	
-	/* tex2d t0, i0, s0	; Read src pixel */
-	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/* sub t0, t0, c0	; Subtract bias vector from pixel */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	
-	/*
-	 * dp4 o0.x, t0, c1	; Multiply pixel by the color conversion matrix
-	 * dp4 o0.y, t0, c2
-	 * dp4 o0.z, t0, c3
-	 * dp4 o0.w, t0, c4	; XXX: Don't need 4th coefficient
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		inst = vl_inst3(TGSI_OPCODE_DP4, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 1);
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	context->states.csc.fragment_shader = pipe->create_fs_state(pipe, &fs);
-	//free(tokens);
-	
-	return 0;
-}
-
-static int vlCreateDataBufsCSC(struct VL_CONTEXT *context)
-{
-	struct pipe_context *pipe;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	
-	/*
-	Create our vertex buffer and vertex buffer element
-	VB contains 4 vertices that render a quad covering the entire window
-	to display a rendered surface
-	Quad is rendered as a tri strip
-	*/
-	context->states.csc.vertex_bufs[0].pitch = sizeof(struct VL_VERTEX2F);
-	context->states.csc.vertex_bufs[0].max_index = 3;
-	context->states.csc.vertex_bufs[0].buffer_offset = 0;
-	context->states.csc.vertex_bufs[0].buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_VERTEX,
-		sizeof(struct VL_VERTEX2F) * 4
-	);
-	
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, context->states.csc.vertex_bufs[0].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		vl_surface_vertex_positions,
-		sizeof(struct VL_VERTEX2F) * 4
-	);
-	
-	pipe->winsys->buffer_unmap(pipe->winsys, context->states.csc.vertex_bufs[0].buffer);
-	
-	context->states.csc.vertex_buf_elems[0].src_offset = 0;
-	context->states.csc.vertex_buf_elems[0].vertex_buffer_index = 0;
-	context->states.csc.vertex_buf_elems[0].nr_components = 2;
-	context->states.csc.vertex_buf_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
-	
-	/*
-	Create our texcoord buffer and texcoord buffer element
-	Texcoord buffer contains the TCs for mapping the rendered surface to the 4 vertices
-	*/
-	context->states.csc.vertex_bufs[1].pitch = sizeof(struct VL_TEXCOORD2F);
-	context->states.csc.vertex_bufs[1].max_index = 3;
-	context->states.csc.vertex_bufs[1].buffer_offset = 0;
-	context->states.csc.vertex_bufs[1].buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_VERTEX,
-		sizeof(struct VL_TEXCOORD2F) * 4
-	);
-	
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, context->states.csc.vertex_bufs[1].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		vl_surface_texcoords,
-		sizeof(struct VL_TEXCOORD2F) * 4
-	);
-	
-	pipe->winsys->buffer_unmap(pipe->winsys, context->states.csc.vertex_bufs[1].buffer);
-	
-	context->states.csc.vertex_buf_elems[1].src_offset = 0;
-	context->states.csc.vertex_buf_elems[1].vertex_buffer_index = 1;
-	context->states.csc.vertex_buf_elems[1].nr_components = 2;
-	context->states.csc.vertex_buf_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT;
-	
-	/*
-	Create our vertex shader's constant buffer
-	Const buffer contains scaling and translation vectors
-	*/
-	context->states.csc.vs_const_buf.size = sizeof(struct VL_CSC_VS_CONSTS);
-	context->states.csc.vs_const_buf.buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_CONSTANT,
-		context->states.csc.vs_const_buf.size
-	);
-	
-	/*
-	Create our fragment shader's constant buffer
-	Const buffer contains the color conversion matrix and bias vectors
-	*/
-	context->states.csc.fs_const_buf.size = sizeof(struct VL_CSC_FS_CONSTS);
-	context->states.csc.fs_const_buf.buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_CONSTANT,
-		context->states.csc.fs_const_buf.size
-	);
-	
-	/*
-	TODO: Refactor this into a seperate function,
-	allow changing the CSC matrix at runtime to switch between regular & full versions
-	*/
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, context->states.csc.fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		&vl_csc_fs_consts_601,
-		sizeof(struct VL_CSC_FS_CONSTS)
-	);
-	
-	pipe->winsys->buffer_unmap(pipe->winsys, context->states.csc.fs_const_buf.buffer);
-	
-	return 0;
-}
-
-static int vlInitCSC(struct VL_CONTEXT *context)
-{	
-	struct pipe_context		*pipe;
-	struct pipe_sampler_state	sampler;
-	
-	assert(context);
-	
-	pipe = context->pipe;
-	
-	/* Delay creating the FB until vlPutSurface() so we know window size */
-	context->states.csc.framebuffer.num_cbufs = 1;
-	context->states.csc.framebuffer.cbufs[0] = NULL;
-	context->states.csc.framebuffer.zsbuf = NULL;
-
-	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_img_filter = PIPE_TEX_FILTER_LINEAR;
-	sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
-	sampler.mag_img_filter = PIPE_TEX_FILTER_LINEAR;
-	sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
-	sampler.compare_func = PIPE_FUNC_ALWAYS;
-	sampler.normalized_coords = 1;
-	/*sampler.prefilter = ;*/
-	/*sampler.shadow_ambient = ;*/
-	/*sampler.lod_bias = ;*/
-	/*sampler.min_lod = ;*/
-	/*sampler.max_lod = ;*/
-	/*sampler.border_color[i] = ;*/
-	/*sampler.max_anisotropy = ;*/
-	context->states.csc.sampler = pipe->create_sampler_state(pipe, &sampler);
-	
-	vlCreateVertexShaderCSC(context);
-	vlCreateFragmentShaderCSC(context);
-	vlCreateDataBufsCSC(context);
-	
-	return 0;
-}
-
-static int vlDestroyCSC(struct VL_CONTEXT *context)
-{
-	assert(context);
-	
-	/*
-	Since we create the final FB when we display our first surface,
-	it may not be created if vlPutSurface() is never called
-	*/
-	if (context->states.csc.framebuffer.cbufs[0])
-		context->pipe->winsys->surface_release(context->pipe->winsys, &context->states.csc.framebuffer.cbufs[0]);
-	context->pipe->delete_sampler_state(context->pipe, context->states.csc.sampler);
-	context->pipe->delete_vs_state(context->pipe, context->states.csc.vertex_shader);
-	context->pipe->delete_fs_state(context->pipe, context->states.csc.fragment_shader);
-	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.csc.vertex_bufs[0].buffer);
-	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.csc.vertex_bufs[1].buffer);
-	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.csc.vs_const_buf.buffer);
-	context->pipe->winsys->buffer_destroy(context->pipe->winsys, context->states.csc.fs_const_buf.buffer);
-	
-	return 0;
-}
-
-static int vlInitCommon(struct VL_CONTEXT *context)
+static int vlInitCommon(struct vlContext *context)
 {
 	struct pipe_context			*pipe;
 	struct pipe_rasterizer_state		rast;
 	struct pipe_blend_state			blend;
 	struct pipe_depth_stencil_alpha_state	dsa;
 	unsigned int				i;
-	
+
 	assert(context);
-	
+
 	pipe = context->pipe;
-	
+
 	rast.flatshade = 1;
 	rast.flatshade_first = 0;
 	rast.light_twoside = 0;
@@ -2113,9 +51,9 @@ static int vlInitCommon(struct VL_CONTEXT *context)
 	rast.offset_units = 1;
 	rast.offset_scale = 1;
 	/*rast.sprite_coord_mode[i] = ;*/
-	context->states.common.raster = pipe->create_rasterizer_state(pipe, &rast);
-	pipe->bind_rasterizer_state(pipe, context->states.common.raster);
-	
+	context->raster = pipe->create_rasterizer_state(pipe, &rast);
+	pipe->bind_rasterizer_state(pipe, context->raster);
+
 	blend.blend_enable = 0;
 	blend.rgb_func = PIPE_BLEND_ADD;
 	blend.rgb_src_factor = PIPE_BLENDFACTOR_ONE;
@@ -2128,9 +66,9 @@ static int vlInitCommon(struct VL_CONTEXT *context)
 	/* Needed to allow color writes to FB, even if blending disabled */
 	blend.colormask = PIPE_MASK_RGBA;
 	blend.dither = 0;
-	context->states.common.blend = pipe->create_blend_state(pipe, &blend);
-	pipe->bind_blend_state(pipe, context->states.common.blend);
-	
+	context->blend = pipe->create_blend_state(pipe, &blend);
+	pipe->bind_blend_state(pipe, context->blend);
+
 	dsa.depth.enabled = 0;
 	dsa.depth.writemask = 0;
 	dsa.depth.func = PIPE_FUNC_ALWAYS;
@@ -2149,134 +87,122 @@ static int vlInitCommon(struct VL_CONTEXT *context)
 	dsa.alpha.enabled = 0;
 	dsa.alpha.func = PIPE_FUNC_ALWAYS;
 	dsa.alpha.ref = 0;
-	context->states.common.dsa = pipe->create_depth_stencil_alpha_state(pipe, &dsa);
-	pipe->bind_depth_stencil_alpha_state(pipe, context->states.common.dsa);
-	
-	return 0;
-}
+	context->dsa = pipe->create_depth_stencil_alpha_state(pipe, &dsa);
+	pipe->bind_depth_stencil_alpha_state(pipe, context->dsa);
 
-static int vlDestroyCommon(struct VL_CONTEXT *context)
-{
-	assert(context);
-	
-	context->pipe->delete_blend_state(context->pipe, context->states.common.blend);
-	context->pipe->delete_rasterizer_state(context->pipe, context->states.common.raster);
-	context->pipe->delete_depth_stencil_alpha_state(context->pipe, context->states.common.dsa);
-	
 	return 0;
 }
 
-static int vlInit(struct VL_CONTEXT *context)
+int vlCreateContext
+(
+	struct vlScreen *screen,
+	struct pipe_context *pipe,
+	unsigned int picture_width,
+	unsigned int picture_height,
+	enum vlFormat picture_format,
+	enum vlProfile profile,
+	enum vlEntryPoint entry_point,
+	struct vlContext **context
+)
 {
+	struct vlContext *ctx;
+
+	assert(screen);
 	assert(context);
-	
-	vlInitCommon(context);
-	vlInitCSC(context);
-	vlInitMC(context);
-	vlInitIDCT(context);
-	
+	assert(pipe);
+
+	ctx = calloc(1, sizeof(struct vlContext));
+
+	if (!ctx)
+		return 1;
+
+	ctx->screen = screen;
+	ctx->pipe = pipe;
+	ctx->picture_width = picture_width;
+	ctx->picture_height = picture_height;
+	ctx->picture_format = picture_format;
+	ctx->profile = profile;
+	ctx->entry_point = entry_point;
+
+	vlInitCommon(ctx);
+
+	vlCreateR16SNormMC(pipe, picture_width, picture_height, picture_format, &ctx->render);
+	vlCreateBasicCSC(pipe, &ctx->csc);
+
+	*context = ctx;
+
 	return 0;
 }
 
-static int vlDestroy(struct VL_CONTEXT *context)
+int vlDestroyContext
+(
+	struct vlContext *context
+)
 {
 	assert(context);
-	
+
 	/* XXX: Must unbind shaders before we can delete them for some reason */
 	context->pipe->bind_vs_state(context->pipe, NULL);
 	context->pipe->bind_fs_state(context->pipe, NULL);
-	
-	vlDestroyCommon(context);
-	vlDestroyCSC(context);
-	vlDestroyMC(context);
-	vlDestroyIDCT(context);
-	
+
+	context->render->vlDestroy(context->render);
+	context->csc->vlDestroy(context->csc);
+
+	context->pipe->delete_blend_state(context->pipe, context->blend);
+	context->pipe->delete_rasterizer_state(context->pipe, context->raster);
+	context->pipe->delete_depth_stencil_alpha_state(context->pipe, context->dsa);
+
+	free(context);
+
 	return 0;
 }
 
-int vlCreateContext
+struct vlScreen* vlContextGetScreen
 (
-	Display *display,
-	struct pipe_context *pipe,
-	unsigned int video_width,
-	unsigned int video_height,
-	enum VL_FORMAT video_format,
-	struct VL_CONTEXT **context
+	struct vlContext *context
 )
 {
-	struct VL_CONTEXT *ctx;
-	
-	assert(display);
-	assert(pipe);
 	assert(context);
-	
-	ctx = calloc(1, sizeof(struct VL_CONTEXT));
-	
-	ctx->display = display;
-	ctx->pipe = pipe;
-	ctx->video_width = video_width;
-	ctx->video_height = video_height;
-	ctx->video_format = video_format;
-	
-	vlInit(ctx);
-	
-	/* Since we only change states in vlPutSurface() we need to start in render mode */
-	vlBeginRender(ctx);
-	
-	*context = ctx;
-	
-	return 0;
+
+	return context->screen;
 }
 
-int vlDestroyContext(struct VL_CONTEXT *context)
+struct pipe_context* vlGetPipeContext
+(
+	struct vlContext *context
+)
 {
 	assert(context);
-	
-	vlDestroy(context);
-	
-	free(context);
-	
-	return 0;
+
+	return context->pipe;
 }
 
-int vlBeginRender(struct VL_CONTEXT *context)
+unsigned int vlGetPictureWidth
+(
+	struct vlContext *context
+)
 {
-	struct pipe_context	*pipe;
-	
 	assert(context);
-	
-	pipe = context->pipe;
-	
-	/* Frame buffer set in vlRender*Macroblock() */
-	/* Shaders, samplers, textures set in vlRender*Macroblock() */
-	pipe->set_vertex_buffers(pipe, 3, context->states.mc.vertex_bufs);
-	pipe->set_vertex_elements(pipe, 3, context->states.mc.vertex_buf_elems);
-	pipe->set_viewport_state(pipe, &context->states.mc.viewport);
-	pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &context->states.mc.vs_const_buf);
-	pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &context->states.mc.fs_const_buf);
-	
-	return 0;
+
+	return context->picture_width;
 }
 
-int vlEndRender(struct VL_CONTEXT *context)
+unsigned int vlGetPictureHeight
+(
+	struct vlContext *context
+)
 {
-	struct pipe_context *pipe;
-	
 	assert(context);
-	
-	pipe = context->pipe;
-	
-	pipe->set_framebuffer_state(pipe, &context->states.csc.framebuffer);
-	pipe->set_viewport_state(pipe, &context->states.csc.viewport);
-	pipe->bind_sampler_states(pipe, 1, (void**)&context->states.csc.sampler);
-	/* Source texture set in vlPutSurface() */
-	pipe->bind_vs_state(pipe, context->states.csc.vertex_shader);
-	pipe->bind_fs_state(pipe, context->states.csc.fragment_shader);
-	pipe->set_vertex_buffers(pipe, 2, context->states.csc.vertex_bufs);
-	pipe->set_vertex_elements(pipe, 2, context->states.csc.vertex_buf_elems);
-	pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &context->states.csc.vs_const_buf);
-	pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &context->states.csc.fs_const_buf);
-	
-	return 0;
+
+	return context->picture_height;
 }
 
+enum vlFormat vlGetPictureFormat
+(
+	struct vlContext *context
+)
+{
+	assert(context);
+
+	return context->picture_format;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.h b/src/gallium/state_trackers/g3dvl/vl_context.h
index bff318854a..3d14634c44 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.h
+++ b/src/gallium/state_trackers/g3dvl/vl_context.h
@@ -1,83 +1,73 @@
 #ifndef vl_context_h
 #define vl_context_h
 
-#include <X11/Xlib.h>
-#include <pipe/p_state.h>
 #include "vl_types.h"
 
 struct pipe_context;
 
-struct VL_CONTEXT
+#ifdef VL_INTERNAL
+struct vlRender;
+struct vlCSC;
+
+struct vlContext
 {
-	Display			*display;
+	struct vlScreen		*screen;
 	struct pipe_context	*pipe;
-	unsigned int		video_width;
-	unsigned int		video_height;
-	enum VL_FORMAT		video_format;
-	
-	struct
-	{
-		struct
-		{
-			struct pipe_rasterizer_state		*raster;
-			struct pipe_depth_stencil_alpha_state	*dsa;
-			struct pipe_blend_state			*blend;
-		} common;
-		
-		struct
-		{
-			struct pipe_viewport_state		viewport;
-			struct pipe_framebuffer_state		render_target;
-			struct pipe_sampler_state		*sampler;
-			struct pipe_texture			*texture;
-			struct pipe_texture			*basis;
-			struct pipe_shader_state		*frame_vs;
-			struct pipe_shader_state		*frame_fs;
-			struct pipe_vertex_buffer 		*vertex_bufs[2];
-			struct pipe_vertex_element		*vertex_buf_elems[2];
-			//struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
-		} idct;
-		
-		struct
-		{
-			struct pipe_viewport_state		viewport;
-			struct pipe_framebuffer_state		render_target;
-			struct pipe_sampler_state		*samplers[5];
-			struct pipe_texture			*textures[5];
-			struct pipe_shader_state		*i_vs, *p_vs[2], *b_vs[2];
-			struct pipe_shader_state		*i_fs, *p_fs[2], *b_fs[2];
-			struct pipe_vertex_buffer 		vertex_bufs[3];
-			struct pipe_vertex_element		vertex_buf_elems[3];
-			struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
-		} mc;
-		
-		struct
-		{
-			struct pipe_viewport_state		viewport;
-			struct pipe_framebuffer_state		framebuffer;
-			struct pipe_sampler_state		*sampler;
-			struct pipe_shader_state		*vertex_shader, *fragment_shader;
-			struct pipe_vertex_buffer 		vertex_bufs[2];
-			struct pipe_vertex_element		vertex_buf_elems[2];
-			struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
-		} csc;
-	} states;
+	unsigned int		picture_width;
+	unsigned int		picture_height;
+	enum vlFormat		picture_format;
+	enum vlProfile		profile;
+	enum vlEntryPoint	entry_point;
+
+	void			*raster;
+	void			*dsa;
+	void			*blend;
+
+	struct vlRender		*render;
+	struct vlCSC		*csc;
 };
+#endif
 
 int vlCreateContext
 (
-	Display *display,
+	struct vlScreen *screen,
 	struct pipe_context *pipe,
-	unsigned int video_width,
-	unsigned int video_height,
-	enum VL_FORMAT video_format,
-	struct VL_CONTEXT **context
+	unsigned int picture_width,
+	unsigned int picture_height,
+	enum vlFormat picture_format,
+	enum vlProfile profile,
+	enum vlEntryPoint entry_point,
+	struct vlContext **context
 );
 
-int vlDestroyContext(struct VL_CONTEXT *context);
+int vlDestroyContext
+(
+	struct vlContext *context
+);
 
-int vlBeginRender(struct VL_CONTEXT *context);
-int vlEndRender(struct VL_CONTEXT *context);
+struct vlScreen* vlContextGetScreen
+(
+	struct vlContext *context
+);
 
-#endif
+struct pipe_context* vlGetPipeContext
+(
+	struct vlContext *context
+);
 
+unsigned int vlGetPictureWidth
+(
+	struct vlContext *context
+);
+
+unsigned int vlGetPictureHeight
+(
+	struct vlContext *context
+);
+
+enum vlFormat vlGetPictureFormat
+(
+	struct vlContext *context
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_csc.h b/src/gallium/state_trackers/g3dvl/vl_csc.h
new file mode 100644
index 0000000000..36417a2792
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_csc.h
@@ -0,0 +1,53 @@
+#ifndef vl_csc_h
+#define vl_csc_h
+
+#include "vl_types.h"
+
+struct pipe_surface;
+
+struct vlCSC
+{
+	int (*vlResizeFrameBuffer)
+	(
+		struct vlCSC *csc,
+		unsigned int width,
+		unsigned int height
+	);
+
+	int (*vlBegin)
+	(
+		struct vlCSC *csc
+	);
+
+	int (*vlPutPicture)
+	(
+		struct vlCSC *csc,
+		struct vlSurface *surface,
+		int srcx,
+		int srcy,
+		int srcw,
+		int srch,
+		int destx,
+		int desty,
+		int destw,
+		int desth,
+		enum vlPictureType picture_type
+	);
+
+	int (*vlEnd)
+	(
+		struct vlCSC *csc
+	);
+
+	struct pipe_surface* (*vlGetFrameBuffer)
+	(
+		struct vlCSC *csc
+	);
+
+	int (*vlDestroy)
+	(
+		struct vlCSC *csc
+	);
+};
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_data.c b/src/gallium/state_trackers/g3dvl/vl_data.c
index 0e5c8c77f9..f2476dbf1e 100644
--- a/src/gallium/state_trackers/g3dvl/vl_data.c
+++ b/src/gallium/state_trackers/g3dvl/vl_data.c
@@ -6,17 +6,17 @@
  * Need to be scaled to cover mbW*mbH macroblock pixels and translated into
  * position on target surface.
  */
-const struct VL_VERTEX2F vl_mb_vertex_positions[24] =
+const struct vlVertex2f macroblock_verts[24] =
 {
 	{0.0f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.0f},
 	{0.5f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.5f},
-	
+
 	{0.5f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.0f},
 	{1.0f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.5f},
-	
+
 	{0.0f, 0.5f}, {0.0f, 1.0f}, {0.5f, 0.5f},
 	{0.5f, 0.5f}, {0.0f, 1.0f}, {0.5f, 1.0f},
-	
+
 	{0.5f, 0.5f}, {0.5f, 1.0f}, {1.0f, 0.5f},
 	{1.0f, 0.5f}, {0.5f, 1.0f}, {1.0f, 1.0f}
 };
@@ -26,17 +26,17 @@ const struct VL_VERTEX2F vl_mb_vertex_positions[24] =
  * in a bW*(bH*4) texture. First luma block located at 0,0->bW,bH; second at
  * 0,bH->bW,2bH; third at 0,2bH->bW,3bH; fourth at 0,3bH->bW,4bH.
  */
-const struct VL_TEXCOORD2F vl_luma_texcoords[24] =
+const struct vlVertex2f macroblock_luma_texcoords[24] =
 {
 	{0.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.0f},
 	{1.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.25f},
-	
+
 	{0.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.25f},
 	{1.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.5f},
-	
+
 	{0.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.5f},
 	{1.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.75f},
-	
+
 	{0.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 0.75f},
 	{1.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 1.0f}
 };
@@ -45,7 +45,7 @@ const struct VL_TEXCOORD2F vl_luma_texcoords[24] =
  * Represents texcoords for the above for rendering 1 chroma block.
  * Straight forward 0,0->1,1 mapping so we can reuse the MB pos vectors.
  */
-const struct VL_TEXCOORD2F *vl_chroma_420_texcoords = (const struct VL_TEXCOORD2F*)vl_mb_vertex_positions;
+const struct vlVertex2f *macroblock_chroma_420_texcoords = macroblock_verts;
 
 /*
  * Represents texcoords for the above for rendering 2 chroma blocks arranged
@@ -53,30 +53,13 @@ const struct VL_TEXCOORD2F *vl_chroma_420_texcoords = (const struct VL_TEXCOORD2
  * 0,bH->bW,2bH. We can render this with 0,0->1,1 mapping.
  * Straight forward 0,0->1,1 mapping so we can reuse MB pos vectors.
  */
-const struct VL_TEXCOORD2F *vl_chroma_422_texcoords = (const struct VL_TEXCOORD2F*)vl_mb_vertex_positions;
+const struct vlVertex2f *macroblock_chroma_422_texcoords = macroblock_verts;
 
 /*
  * Represents texcoords for the above for rendering 4 chroma blocks.
  * Same case as 4 luma blocks.
  */
-const struct VL_TEXCOORD2F *vl_chroma_444_texcoords = vl_luma_texcoords;
-
-/*
- * Represents 2 triangles in a strip in normalized coords.
- * Used to render the surface onto the frame buffer.
- */
-const struct VL_VERTEX2F vl_surface_vertex_positions[4] =
-{
-	{0.0f, 0.0f},
-	{0.0f, 1.0f},
-	{1.0f, 0.0f},
-	{1.0f, 1.0f}
-};
-
-/*
- * Represents texcoords for the above. We can use the position values directly.
- */
-const struct VL_TEXCOORD2F *vl_surface_texcoords = (const struct VL_TEXCOORD2F*)vl_surface_vertex_positions;
+const struct vlVertex2f *macroblock_chroma_444_texcoords = macroblock_luma_texcoords;
 
 /*
  * Used when rendering P and B macroblocks, multiplier is applied to the A channel,
@@ -84,97 +67,10 @@ const struct VL_TEXCOORD2F *vl_surface_texcoords = (const struct VL_TEXCOORD2F*)
  * get back the differential. The differential is then added to the samples from the
  * reference surface(s).
  */
+#if 0
 const struct VL_MC_FS_CONSTS vl_mc_fs_consts =
 {
 	{32767.0f / 255.0f, 32767.0f / 255.0f, 32767.0f / 255.0f, 0.0f},
 	{0.5f, 2.0f, 0.0f, 0.0f}
 };
-
-/*
- * Identity color conversion constants, for debugging
- */
-const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_identity =
-{
-	{
-		0.0f, 0.0f, 0.0f, 0.0f
-	},
-	{
-		1.0f, 0.0f, 0.0f, 0.0f,
-		0.0f, 1.0f, 0.0f, 0.0f,
-		0.0f, 0.0f, 1.0f, 0.0f,
-		0.0f, 0.0f, 0.0f, 1.0f
-	}
-};
-
-/*
- * Converts ITU-R BT.601 YCbCr pixels to RGB pixels where:
- * Y is in [16,235], Cb and Cr are in [16,240]
- * R, G, and B are in [16,235]
- */
-const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_601 =
-{
-	{
-		0.0f,		0.501960784f,	0.501960784f,	0.0f
-	},
-	{
-		1.0f,		0.0f,		1.371f,		0.0f,
-		1.0f,		-0.336f,	-0.698f,	0.0f,
-		1.0f,		1.732f,		0.0f,		0.0f,
-		0.0f,		0.0f,		0.0f,		1.0f
-	}
-};
-
-/*
- * Converts ITU-R BT.601 YCbCr pixels to RGB pixels where:
- * Y is in [16,235], Cb and Cr are in [16,240]
- * R, G, and B are in [0,255]
- */
-const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_601_full =
-{
-	{
-		0.062745098f,	0.501960784f,	0.501960784f,	0.0f
-	},
-	{
-		1.164f,		0.0f,		1.596f,		0.0f,
-		1.164f,		-0.391f,	-0.813f,	0.0f,
-		1.164f,		2.018f,		0.0f,		0.0f,
-		0.0f,		0.0f,		0.0f,		1.0f
-	}
-};
-
-/*
- * Converts ITU-R BT.709 YCbCr pixels to RGB pixels where:
- * Y is in [16,235], Cb and Cr are in [16,240]
- * R, G, and B are in [16,235]
- */
-const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_709 =
-{
-	{
-		0.0f,		0.501960784f,	0.501960784f,	0.0f
-	},
-	{
-		1.0f,		0.0f,		1.540f,		0.0f,
-		1.0f,		-0.183f,	-0.459f,	0.0f,
-		1.0f,		1.816f,		0.0f,		0.0f,
-		0.0f,		0.0f,		0.0f,		1.0f
-	}
-};
-
-/*
- * Converts ITU-R BT.709 YCbCr pixels to RGB pixels where:
- * Y is in [16,235], Cb and Cr are in [16,240]
- * R, G, and B are in [0,255]
- */
-const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_709_full =
-{
-	{
-		0.062745098f,	0.501960784f,	0.501960784f,	0.0f
-	},
-	{
-		1.164f,		0.0f,		1.793f,		0.0f,
-		1.164f,		-0.213f,	-0.534f,	0.0f,
-		1.164f,		2.115f,		0.0f,		0.0f,
-		0.0f,		0.0f,		0.0f,		1.0f
-	}
-};
-
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_data.h b/src/gallium/state_trackers/g3dvl/vl_data.h
index 8f347273ad..f0de2e976c 100644
--- a/src/gallium/state_trackers/g3dvl/vl_data.h
+++ b/src/gallium/state_trackers/g3dvl/vl_data.h
@@ -3,15 +3,18 @@
 
 #include "vl_types.h"
 
-extern const struct VL_VERTEX2F vl_mb_vertex_positions[24];
-extern const struct VL_TEXCOORD2F vl_luma_texcoords[24];
-extern const struct VL_TEXCOORD2F *vl_chroma_420_texcoords;
-extern const struct VL_TEXCOORD2F *vl_chroma_422_texcoords;
-extern const struct VL_TEXCOORD2F *vl_chroma_444_texcoords;
+/* TODO: Needs to be rolled into the appropriate stage */
 
-extern const struct VL_VERTEX2F vl_surface_vertex_positions[4];
-extern const struct VL_TEXCOORD2F *vl_surface_texcoords;
+extern const struct vlVertex2f macroblock_verts[24];
+extern const struct vlVertex2f macroblock_luma_texcoords[24];
+extern const struct vlVertex2f *macroblock_chroma_420_texcoords;
+extern const struct vlVertex2f *macroblock_chroma_422_texcoords;
+extern const struct vlVertex2f *macroblock_chroma_444_texcoords;
 
+extern const struct vlVertex2f surface_verts[4];
+extern const struct vlVertex2f *surface_texcoords;
+
+/*
 extern const struct VL_MC_FS_CONSTS vl_mc_fs_consts;
 
 extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_identity;
@@ -19,6 +22,6 @@ extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_601;
 extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_601_full;
 extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_709;
 extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_709_full;
+*/
 
 #endif
-
diff --git a/src/gallium/state_trackers/g3dvl/vl_defs.h b/src/gallium/state_trackers/g3dvl/vl_defs.h
index e668a7a10e..d612d02502 100644
--- a/src/gallium/state_trackers/g3dvl/vl_defs.h
+++ b/src/gallium/state_trackers/g3dvl/vl_defs.h
@@ -9,4 +9,3 @@
 #define VL_MACROBLOCK_SIZE	(VL_MACROBLOCK_WIDTH * VL_MACROBLOCK_HEIGHT)
 
 #endif
-
diff --git a/src/gallium/state_trackers/g3dvl/vl_display.c b/src/gallium/state_trackers/g3dvl/vl_display.c
new file mode 100644
index 0000000000..af80faa7f5
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_display.c
@@ -0,0 +1,48 @@
+#define VL_INTERNAL
+#include "vl_display.h"
+#include <assert.h>
+#include <stdlib.h>
+
+int vlCreateDisplay
+(
+	vlNativeDisplay native_display,
+	struct vlDisplay **display
+)
+{
+	struct vlDisplay *dpy;
+
+	assert(native_display);
+	assert(display);
+
+	dpy = calloc(1, sizeof(struct vlDisplay));
+
+	if (!dpy)
+		return 1;
+
+	dpy->native = native_display;
+	*display = dpy;
+
+	return 0;
+}
+
+int vlDestroyDisplay
+(
+	struct vlDisplay *display
+)
+{
+	assert(display);
+
+	free(display);
+
+	return 0;
+}
+
+vlNativeDisplay vlGetNativeDisplay
+(
+	struct vlDisplay *display
+)
+{
+	assert(display);
+
+	return display->native;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_display.h b/src/gallium/state_trackers/g3dvl/vl_display.h
new file mode 100644
index 0000000000..e11fd40799
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_display.h
@@ -0,0 +1,29 @@
+#ifndef vl_display_h
+#define vl_display_h
+
+#include "vl_types.h"
+
+#ifdef VL_INTERNAL
+struct vlDisplay
+{
+	vlNativeDisplay native;
+};
+#endif
+
+int vlCreateDisplay
+(
+	vlNativeDisplay native_display,
+	struct vlDisplay **display
+);
+
+int vlDestroyDisplay
+(
+	struct vlDisplay *display
+);
+
+vlNativeDisplay vlGetNativeDisplay
+(
+	struct vlDisplay *display
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
new file mode 100644
index 0000000000..4fae224431
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
@@ -0,0 +1,2315 @@
+#define VL_INTERNAL
+#include "vl_r16snorm_mc.h"
+#include <assert.h>
+#include <stdlib.h>
+#include <pipe/p_context.h>
+#include <pipe/p_winsys.h>
+#include <pipe/p_state.h>
+#include <pipe/p_inlines.h>
+#include <tgsi/tgsi_parse.h>
+#include <tgsi/tgsi_build.h>
+#include "vl_render.h"
+#include "vl_shader_build.h"
+#include "vl_surface.h"
+#include "vl_util.h"
+#include "vl_types.h"
+#include "vl_defs.h"
+
+struct vlVertexShaderConsts
+{
+	/*struct vlVertex4f scale;
+	struct vlVertex4f denorm;*/
+	struct vlVertex4f	scale;
+	struct vlVertex4f	mb_pos_trans;
+	struct vlVertex4f	denorm;
+	struct
+	{
+		struct vlVertex4f	top_field;
+		struct vlVertex4f	bottom_field;
+	} mb_tc_trans[2];
+};
+
+struct vlFragmentShaderConsts
+{
+	struct vlVertex4f multiplier;
+	struct vlVertex4f div;
+};
+
+struct vlR16SnormMC
+{
+	struct vlRender				base;
+
+	unsigned int				video_width, video_height;
+	enum vlFormat				video_format;
+
+	struct pipe_context			*pipe;
+	struct pipe_viewport_state		viewport;
+	struct pipe_framebuffer_state		render_target;
+	struct pipe_sampler_state		*samplers[5];
+	struct pipe_texture			*textures[5];
+	void					*i_vs, *p_vs[2], *b_vs[2];
+	void					*i_fs, *p_fs[2], *b_fs[2];
+	struct pipe_vertex_buffer 		vertex_bufs[3];
+	struct pipe_vertex_element		vertex_elems[3];
+	struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
+};
+
+int vlBegin
+(
+	struct vlRender *render
+)
+{
+	struct vlR16SnormMC	*mc;
+	struct pipe_context	*pipe;
+
+	assert(render);
+
+	mc = (struct vlR16SnormMC*)render;
+	pipe = mc->pipe;
+
+	/* Frame buffer set in vlRender*Macroblock() */
+	/* Shaders, samplers, textures set in vlRender*Macroblock() */
+	pipe->set_vertex_buffers(pipe, 3, mc->vertex_bufs);
+	pipe->set_vertex_elements(pipe, 3, mc->vertex_elems);
+	pipe->set_viewport_state(pipe, &mc->viewport);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &mc->vs_const_buf);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &mc->fs_const_buf);
+
+	return 0;
+}
+
+/*static int vlGrabMacroBlock
+(
+	struct vlR16SnormMC *mc,
+	struct vlMpeg2MacroBlock *macroblock
+)
+{
+	assert(mc);
+	assert(macroblock);
+
+
+
+	return 0;
+}*/
+
+/*#define DO_IDCT*/
+
+#ifdef DO_IDCT
+static int vlTransformBlock(short *src, short *dst, short bias)
+{
+	static const float basis[8][8] =
+	{
+		{0.3536,   0.4904,   0.4619,   0.4157,   0.3536,   0.2778,   0.1913,   0.0975},
+		{0.3536,   0.4157,   0.1913,  -0.0975,  -0.3536,  -0.4904,  -0.4619,  -0.2778},
+		{0.3536,   0.2778,  -0.1913,  -0.4904,  -0.3536,   0.0975,   0.4619,   0.4157},
+		{0.3536,   0.0975,  -0.4619,  -0.2778,   0.3536,   0.4157,  -0.1913,  -0.4904},
+		{0.3536,  -0.0975,  -0.4619,   0.2778,   0.3536,  -0.4157,  -0.1913,   0.4904},
+		{0.3536,  -0.2778,  -0.1913,   0.4904,  -0.3536,  -0.0975,   0.4619,  -0.4157},
+		{0.3536,  -0.4157,   0.1913,   0.0975,  -0.3536,   0.4904,  -0.4619,   0.2778},
+		{0.3536,  -0.4904,   0.4619,  -0.4157,   0.3536,  -0.2778,   0.1913,  -0.0975}
+	};
+
+	unsigned int	x, y;
+	short		tmp[64];
+
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		for (x = 0; x < VL_BLOCK_WIDTH; ++x)
+			tmp[y * VL_BLOCK_WIDTH + x] = (short)
+			(
+				src[y * VL_BLOCK_WIDTH + 0] * basis[x][0] +
+				src[y * VL_BLOCK_WIDTH + 1] * basis[x][1] +
+				src[y * VL_BLOCK_WIDTH + 2] * basis[x][2] +
+				src[y * VL_BLOCK_WIDTH + 3] * basis[x][3] +
+				src[y * VL_BLOCK_WIDTH + 4] * basis[x][4] +
+				src[y * VL_BLOCK_WIDTH + 5] * basis[x][5] +
+				src[y * VL_BLOCK_WIDTH + 6] * basis[x][6] +
+				src[y * VL_BLOCK_WIDTH + 7] * basis[x][7]
+			);
+
+	for (x = 0; x < VL_BLOCK_WIDTH; ++x)
+		for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		{
+			dst[y * VL_BLOCK_WIDTH + x] = bias + (short)
+			(
+				tmp[0 * VL_BLOCK_WIDTH + x] * basis[y][0] +
+				tmp[1 * VL_BLOCK_WIDTH + x] * basis[y][1] +
+				tmp[2 * VL_BLOCK_WIDTH + x] * basis[y][2] +
+				tmp[3 * VL_BLOCK_WIDTH + x] * basis[y][3] +
+				tmp[4 * VL_BLOCK_WIDTH + x] * basis[y][4] +
+				tmp[5 * VL_BLOCK_WIDTH + x] * basis[y][5] +
+				tmp[6 * VL_BLOCK_WIDTH + x] * basis[y][6] +
+				tmp[7 * VL_BLOCK_WIDTH + x] * basis[y][7]
+			);
+			if (dst[y * VL_BLOCK_WIDTH + x] > 255)
+				dst[y * VL_BLOCK_WIDTH + x] = 255;
+			else if (bias > 0 && dst[y * VL_BLOCK_WIDTH + x] < 0)
+				dst[y * VL_BLOCK_WIDTH + x] = 0;
+		}
+	return 0;
+}
+#endif
+
+static int vlGrabFrameCodedBlock(short *src, short *dst, unsigned int dst_pitch)
+{
+	unsigned int y;
+
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		memcpy
+		(
+			dst + y * dst_pitch,
+			src + y * VL_BLOCK_WIDTH,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	return 0;
+}
+
+static int vlGrabFieldCodedBlock(short *src, short *dst, unsigned int dst_pitch)
+{
+	unsigned int y;
+
+	for (y = 0; y < VL_BLOCK_HEIGHT / 2; ++y)
+		memcpy
+		(
+			dst + y * dst_pitch * 2,
+			src + y * VL_BLOCK_WIDTH,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	dst += VL_BLOCK_HEIGHT * dst_pitch;
+
+	for (; y < VL_BLOCK_HEIGHT; ++y)
+		memcpy
+		(
+			dst + y * dst_pitch * 2,
+			src + y * VL_BLOCK_WIDTH,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	return 0;
+}
+
+static int vlGrabNoBlock(short *dst, unsigned int dst_pitch)
+{
+	unsigned int y;
+
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		memset
+		(
+			dst + y * dst_pitch,
+			0,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	return 0;
+}
+
+enum vlSampleType
+{
+	vlSampleTypeFull,
+	vlSampleTypeDiff
+};
+
+static int vlGrabBlocks
+(
+	struct vlR16SnormMC *mc,
+	unsigned int coded_block_pattern,
+	enum vlDCTType dct_type,
+	enum vlSampleType sample_type,
+	short *blocks
+)
+{
+	struct pipe_surface	*tex_surface;
+	short			*texels;
+	unsigned int		tex_pitch;
+	unsigned int		tb, sb = 0;
+
+	assert(mc);
+	assert(blocks);
+
+	tex_surface = mc->pipe->screen->get_tex_surface
+	(
+		mc->pipe->screen,
+		mc->textures[0],
+		0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
+	);
+
+	texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE);
+	tex_pitch = tex_surface->stride / tex_surface->block.size;
+
+	for (tb = 0; tb < 4; ++tb)
+	{
+		if ((coded_block_pattern >> (5 - tb)) & 1)
+		{
+			short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
+
+#ifdef DO_IDCT
+			vlTransformBlock(cur_block, cur_block, sample_type == vlSampleTypeFull ? 128 : 0);
+#endif
+
+			if (dct_type == vlDCTTypeFrameCoded)
+				vlGrabFrameCodedBlock
+				(
+					cur_block,
+					texels + tb * tex_pitch * VL_BLOCK_HEIGHT,
+					tex_pitch
+				);
+			else
+				vlGrabFieldCodedBlock
+				(
+					cur_block,
+					texels + (tb % 2) * tex_pitch * VL_BLOCK_HEIGHT + (tb / 2) * tex_pitch,
+					tex_pitch
+				);
+
+			++sb;
+		}
+		else
+			vlGrabNoBlock(texels + tb * tex_pitch * VL_BLOCK_HEIGHT, tex_pitch);
+	}
+
+	pipe_surface_unmap(tex_surface);
+
+	/* TODO: Implement 422, 444 */
+	for (tb = 0; tb < 2; ++tb)
+	{
+		tex_surface = mc->pipe->screen->get_tex_surface
+		(
+			mc->pipe->screen,
+			mc->textures[tb + 1],
+			0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
+		);
+
+		texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE);
+		tex_pitch = tex_surface->stride / tex_surface->block.size;
+
+		if ((coded_block_pattern >> (1 - tb)) & 1)
+		{
+			short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
+
+#ifdef DO_IDCT
+			vlTransformBlock(cur_block, cur_block, sample_type == vlSampleTypeFull ? 128 : 0);
+#endif
+
+			vlGrabFrameCodedBlock
+			(
+				cur_block,
+				texels,
+				tex_pitch
+			);
+
+			++sb;
+		}
+		else
+			vlGrabNoBlock(texels, tex_pitch);
+
+		pipe_surface_unmap(tex_surface);
+	}
+
+	return 0;
+}
+
+int vlRenderIMacroBlock
+(
+	struct vlR16SnormMC *mc,
+	enum vlPictureType picture_type,
+	enum vlFieldOrder field_order,
+	unsigned int mbx,
+	unsigned int mby,
+	unsigned int coded_block_pattern,
+	enum vlDCTType dct_type,
+	short *blocks,
+	struct vlSurface *surface
+)
+{
+	struct pipe_context		*pipe;
+	struct vlVertexShaderConsts	*vs_consts;
+
+	assert(blocks);
+	assert(surface);
+
+	/* TODO: Implement interlaced rendering */
+	if (picture_type != vlPictureTypeFrame)
+		return 0;
+
+	vlGrabBlocks(mc, coded_block_pattern, dct_type, vlSampleTypeFull, blocks);
+
+	pipe = mc->pipe;
+
+	vs_consts = pipe->winsys->buffer_map
+	(
+		pipe->winsys,
+		mc->vs_const_buf.buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE
+	);
+
+	vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->texture->width[0];
+	vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->texture->height[0];
+	vs_consts->scale.z = 1.0f;
+	vs_consts->scale.w = 1.0f;
+	vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->texture->width[0];
+	vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->texture->height[0];
+	vs_consts->mb_pos_trans.z = 0.0f;
+	vs_consts->mb_pos_trans.w = 0.0f;
+
+	pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer);
+
+	mc->render_target.cbufs[0] = pipe->screen->get_tex_surface
+	(
+		pipe->screen,
+		surface->texture,
+		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
+	);
+	pipe->set_framebuffer_state(pipe, &mc->render_target);
+	pipe->set_sampler_textures(pipe, 3, mc->textures);
+	pipe->bind_sampler_states(pipe, 3, (void**)mc->samplers);
+	pipe->bind_vs_state(pipe, mc->i_vs);
+	pipe->bind_fs_state(pipe, mc->i_fs);
+
+	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
+
+	return 0;
+}
+
+int vlRenderPMacroBlock
+(
+	struct vlR16SnormMC *mc,
+	enum vlPictureType picture_type,
+	enum vlFieldOrder field_order,
+	unsigned int mbx,
+	unsigned int mby,
+	enum vlMotionType mc_type,
+	short top_x,
+	short top_y,
+	short bottom_x,
+	short bottom_y,
+	unsigned int coded_block_pattern,
+	enum vlDCTType dct_type,
+	short *blocks,
+	struct vlSurface *ref_surface,
+	struct vlSurface *surface
+)
+{
+	struct pipe_context		*pipe;
+	struct vlVertexShaderConsts	*vs_consts;
+
+	assert(motion_vectors);
+	assert(blocks);
+	assert(ref_surface);
+	assert(surface);
+
+	/* TODO: Implement interlaced rendering */
+	if (picture_type != vlPictureTypeFrame)
+		return 0;
+	/* TODO: Implement other MC types */
+	if (mc_type != vlMotionTypeFrame && mc_type != vlMotionTypeField)
+		return 0;
+
+	vlGrabBlocks(mc, coded_block_pattern, dct_type, vlSampleTypeDiff, blocks);
+
+	pipe = mc->pipe;
+
+	vs_consts = pipe->winsys->buffer_map
+	(
+		pipe->winsys,
+		mc->vs_const_buf.buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE
+	);
+
+	vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->texture->width[0];
+	vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->texture->height[0];
+	vs_consts->scale.z = 1.0f;
+	vs_consts->scale.w = 1.0f;
+	vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->texture->width[0];
+	vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->texture->height[0];
+	vs_consts->mb_pos_trans.z = 0.0f;
+	vs_consts->mb_pos_trans.w = 0.0f;
+	vs_consts->mb_tc_trans[0].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + top_x * 0.5f) / (float)surface->texture->width[0];
+	vs_consts->mb_tc_trans[0].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + top_y * 0.5f) / (float)surface->texture->height[0];
+	vs_consts->mb_tc_trans[0].top_field.z = 0.0f;
+	vs_consts->mb_tc_trans[0].top_field.w = 0.0f;
+
+	if (mc_type == vlMotionTypeField)
+	{
+		vs_consts->denorm.x = (float)surface->texture->width[0];
+		vs_consts->denorm.y = (float)surface->texture->height[0];
+
+		vs_consts->mb_tc_trans[0].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + bottom_x * 0.5f) / (float)surface->texture->width[0];
+		vs_consts->mb_tc_trans[0].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + bottom_y * 0.5f) / (float)surface->texture->height[0];
+		vs_consts->mb_tc_trans[0].bottom_field.z = 0.0f;
+		vs_consts->mb_tc_trans[0].bottom_field.w = 0.0f;
+
+		pipe->bind_vs_state(pipe, mc->p_vs[1]);
+		pipe->bind_fs_state(pipe, mc->p_fs[1]);
+	}
+	else
+	{
+		pipe->bind_vs_state(pipe, mc->p_vs[0]);
+		pipe->bind_fs_state(pipe, mc->p_fs[0]);
+	}
+
+	pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer);
+
+	mc->render_target.cbufs[0] = pipe->screen->get_tex_surface
+	(
+		pipe->screen,
+		surface->texture,
+		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
+	);
+	pipe->set_framebuffer_state(pipe, &mc->render_target);
+
+	mc->textures[3] = ref_surface->texture;
+	pipe->set_sampler_textures(pipe, 4, mc->textures);
+	pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers);
+
+	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
+
+	return 0;
+}
+
+int vlRenderBMacroBlock
+(
+	struct vlR16SnormMC *mc,
+	enum vlPictureType picture_type,
+	enum vlFieldOrder field_order,
+	unsigned int mbx,
+	unsigned int mby,
+	enum vlMotionType mc_type,
+	short top_past_x,
+	short top_past_y,
+	short bottom_past_x,
+	short bottom_past_y,
+	short top_future_x,
+	short top_future_y,
+	short bottom_future_x,
+	short bottom_future_y,
+	unsigned int coded_block_pattern,
+	enum vlDCTType dct_type,
+	short *blocks,
+	struct vlSurface *past_surface,
+	struct vlSurface *future_surface,
+	struct vlSurface *surface
+)
+{
+	struct pipe_context		*pipe;
+	struct vlVertexShaderConsts	*vs_consts;
+
+	assert(motion_vectors);
+	assert(blocks);
+	assert(ref_surface);
+	assert(surface);
+
+	/* TODO: Implement interlaced rendering */
+	if (picture_type != vlPictureTypeFrame)
+		return 0;
+	/* TODO: Implement other MC types */
+	if (mc_type != vlMotionTypeFrame && mc_type != vlMotionTypeField)
+		return 0;
+
+	vlGrabBlocks(mc, coded_block_pattern, dct_type, vlSampleTypeDiff, blocks);
+
+	pipe = mc->pipe;
+
+	vs_consts = pipe->winsys->buffer_map
+	(
+		pipe->winsys,
+		mc->vs_const_buf.buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE
+	);
+
+	vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->texture->width[0];
+	vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->texture->height[0];
+	vs_consts->scale.z = 1.0f;
+	vs_consts->scale.w = 1.0f;
+	vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->texture->width[0];
+	vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->texture->height[0];
+	vs_consts->mb_pos_trans.z = 0.0f;
+	vs_consts->mb_pos_trans.w = 0.0f;
+	vs_consts->mb_tc_trans[0].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + top_past_x * 0.5f) / (float)surface->texture->width[0];
+	vs_consts->mb_tc_trans[0].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + top_past_y * 0.5f) / (float)surface->texture->height[0];
+	vs_consts->mb_tc_trans[0].top_field.z = 0.0f;
+	vs_consts->mb_tc_trans[0].top_field.w = 0.0f;
+	vs_consts->mb_tc_trans[1].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + top_future_x * 0.5f) / (float)surface->texture->width[0];
+	vs_consts->mb_tc_trans[1].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + top_future_y * 0.5f) / (float)surface->texture->height[0];
+	vs_consts->mb_tc_trans[1].top_field.z = 0.0f;
+	vs_consts->mb_tc_trans[1].top_field.w = 0.0f;
+
+	if (mc_type == vlMotionTypeField)
+	{
+		vs_consts->denorm.x = (float)surface->texture->width[0];
+		vs_consts->denorm.y = (float)surface->texture->height[0];
+
+		vs_consts->mb_tc_trans[0].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + bottom_past_x * 0.5f) / (float)surface->texture->width[0];
+		vs_consts->mb_tc_trans[0].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + bottom_past_y * 0.5f) / (float)surface->texture->height[0];
+		vs_consts->mb_tc_trans[0].bottom_field.z = 0.0f;
+		vs_consts->mb_tc_trans[0].bottom_field.w = 0.0f;
+		vs_consts->mb_tc_trans[1].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + bottom_future_x * 0.5f) / (float)surface->texture->width[0];
+		vs_consts->mb_tc_trans[1].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + bottom_future_y * 0.5f) / (float)surface->texture->height[0];
+		vs_consts->mb_tc_trans[1].bottom_field.z = 0.0f;
+		vs_consts->mb_tc_trans[1].bottom_field.w = 0.0f;
+
+		pipe->bind_vs_state(pipe, mc->b_vs[1]);
+		pipe->bind_fs_state(pipe, mc->b_fs[1]);
+	}
+	else
+	{
+		pipe->bind_vs_state(pipe, mc->b_vs[0]);
+		pipe->bind_fs_state(pipe, mc->b_fs[0]);
+	}
+
+	pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer);
+
+	mc->render_target.cbufs[0] = pipe->screen->get_tex_surface
+	(
+		pipe->screen,
+		surface->texture,
+		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
+	);
+	pipe->set_framebuffer_state(pipe, &mc->render_target);
+
+	mc->textures[3] = past_surface->texture;
+	mc->textures[4] = future_surface->texture;
+	pipe->set_sampler_textures(pipe, 5, mc->textures);
+	pipe->bind_sampler_states(pipe, 5, (void**)mc->samplers);
+
+	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
+
+	return 0;
+}
+
+int vlRenderMacroBlocksMpeg2R16Snorm
+(
+	struct vlRender *render,
+	struct vlMpeg2MacroBlockBatch *batch,
+	struct vlSurface *surface
+)
+{
+	struct vlR16SnormMC	*mc;
+	unsigned int		i;
+
+	assert(render);
+
+	mc = (struct vlR16SnormMC*)render;
+
+	/*for (i = 0; i < batch->num_macroblocks; ++i)
+		vlGrabMacroBlock(batch->macroblocks[i]);*/
+
+	for (i = 0; i < batch->num_macroblocks; ++i)
+	{
+		switch (batch->macroblocks[i].mb_type)
+		{
+			case vlMacroBlockTypeIntra:
+			{
+				vlRenderIMacroBlock
+				(
+					mc,
+					batch->picture_type,
+					batch->field_order,
+					batch->macroblocks[i].mbx,
+					batch->macroblocks[i].mby,
+					batch->macroblocks[i].cbp,
+					batch->macroblocks[i].dct_type,
+					batch->macroblocks[i].blocks,
+					surface
+				);
+				break;
+			}
+			case vlMacroBlockTypeFwdPredicted:
+			{
+				vlRenderPMacroBlock
+				(
+					mc,
+					batch->picture_type,
+					batch->field_order,
+					batch->macroblocks[i].mbx,
+					batch->macroblocks[i].mby,
+					batch->macroblocks[i].mo_type,
+					batch->macroblocks[i].PMV[0][0][0],
+					batch->macroblocks[i].PMV[0][0][1],
+					batch->macroblocks[i].PMV[1][0][0],
+					batch->macroblocks[i].PMV[1][0][1],
+					batch->macroblocks[i].cbp,
+					batch->macroblocks[i].dct_type,
+					batch->macroblocks[i].blocks,
+					batch->past_surface,
+					surface
+				);
+				break;
+			}
+			case vlMacroBlockTypeBkwdPredicted:
+			{
+				vlRenderPMacroBlock
+				(
+					mc,
+					batch->picture_type,
+					batch->field_order,
+					batch->macroblocks[i].mbx,
+					batch->macroblocks[i].mby,
+					batch->macroblocks[i].mo_type,
+					batch->macroblocks[i].PMV[0][1][0],
+					batch->macroblocks[i].PMV[0][1][1],
+					batch->macroblocks[i].PMV[1][1][0],
+					batch->macroblocks[i].PMV[1][1][1],
+					batch->macroblocks[i].cbp,
+					batch->macroblocks[i].dct_type,
+					batch->macroblocks[i].blocks,
+					batch->future_surface,
+					surface
+				);
+				break;
+			}
+			case vlMacroBlockTypeBiPredicted:
+			{
+				vlRenderBMacroBlock
+				(
+					mc,
+					batch->picture_type,
+					batch->field_order,
+					batch->macroblocks[i].mbx,
+					batch->macroblocks[i].mby,
+					batch->macroblocks[i].mo_type,
+					batch->macroblocks[i].PMV[0][0][0],
+					batch->macroblocks[i].PMV[0][0][1],
+					batch->macroblocks[i].PMV[1][0][0],
+					batch->macroblocks[i].PMV[1][0][1],
+					batch->macroblocks[i].PMV[0][1][0],
+					batch->macroblocks[i].PMV[0][1][1],
+					batch->macroblocks[i].PMV[1][1][0],
+					batch->macroblocks[i].PMV[1][1][1],
+					batch->macroblocks[i].cbp,
+					batch->macroblocks[i].dct_type,
+					batch->macroblocks[i].blocks,
+					batch->past_surface,
+					batch->future_surface,
+					surface
+				);
+				break;
+			}
+			default:
+				assert(0);
+		}
+	}
+
+	return 0;
+}
+
+int vlEnd
+(
+	struct vlRender *render
+)
+{
+	assert(render);
+
+	return 0;
+}
+
+int vlDestroy
+(
+	struct vlRender *render
+)
+{
+	struct vlR16SnormMC	*mc;
+	struct pipe_context	*pipe;
+	unsigned int		i;
+
+	assert(render);
+
+	mc = (struct vlR16SnormMC*)render;
+	pipe = mc->pipe;
+
+	for (i = 0; i < 5; ++i)
+		pipe->delete_sampler_state(pipe, mc->samplers[i]);
+
+	for (i = 0; i < 3; ++i)
+		pipe->winsys->buffer_destroy(pipe->winsys, mc->vertex_bufs[i].buffer);
+
+	/* Textures 3 & 4 are not created directly, no need to release them here */
+	for (i = 0; i < 3; ++i)
+		pipe_texture_release(&mc->textures[i]);
+
+	pipe->delete_vs_state(pipe, mc->i_vs);
+	pipe->delete_fs_state(pipe, mc->i_fs);
+
+	for (i = 0; i < 2; ++i)
+	{
+		pipe->delete_vs_state(pipe, mc->p_vs[i]);
+		pipe->delete_fs_state(pipe, mc->p_fs[i]);
+		pipe->delete_vs_state(pipe, mc->b_vs[i]);
+		pipe->delete_fs_state(pipe, mc->b_fs[i]);
+	}
+
+	pipe->winsys->buffer_destroy(pipe->winsys, mc->vs_const_buf.buffer);
+	pipe->winsys->buffer_destroy(pipe->winsys, mc->fs_const_buf.buffer);
+
+	free(mc);
+
+	return 0;
+}
+
+/*
+ * Represents 8 triangles (4 quads, 1 per block) in noormalized coords
+ * that render a macroblock.
+ * Need to be scaled to cover mbW*mbH macroblock pixels and translated into
+ * position on target surface.
+ */
+const struct vlVertex2f macroblock_verts[24] =
+{
+	{0.0f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.0f},
+	{0.5f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.5f},
+
+	{0.5f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.0f},
+	{1.0f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.5f},
+
+	{0.0f, 0.5f}, {0.0f, 1.0f}, {0.5f, 0.5f},
+	{0.5f, 0.5f}, {0.0f, 1.0f}, {0.5f, 1.0f},
+
+	{0.5f, 0.5f}, {0.5f, 1.0f}, {1.0f, 0.5f},
+	{1.0f, 0.5f}, {0.5f, 1.0f}, {1.0f, 1.0f}
+};
+
+/*
+ * Represents texcoords for the above for rendering 4 luma blocks arranged
+ * in a bW*(bH*4) texture. First luma block located at 0,0->bW,bH; second at
+ * 0,bH->bW,2bH; third at 0,2bH->bW,3bH; fourth at 0,3bH->bW,4bH.
+ */
+const struct vlVertex2f macroblock_luma_texcoords[24] =
+{
+	{0.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.0f},
+	{1.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.25f},
+
+	{0.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.25f},
+	{1.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.5f},
+
+	{0.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.5f},
+	{1.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.75f},
+
+	{0.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 0.75f},
+	{1.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 1.0f}
+};
+
+/*
+ * Represents texcoords for the above for rendering 1 chroma block.
+ * Straight forward 0,0->1,1 mapping so we can reuse the MB pos vectors.
+ */
+const struct vlVertex2f *macroblock_chroma_420_texcoords = macroblock_verts;
+
+/*
+ * Represents texcoords for the above for rendering 2 chroma blocks arranged
+ * in a bW*(bH*2) texture. First chroma block located at 0,0->bW,bH; second at
+ * 0,bH->bW,2bH. We can render this with 0,0->1,1 mapping.
+ * Straight forward 0,0->1,1 mapping so we can reuse MB pos vectors.
+ */
+const struct vlVertex2f *macroblock_chroma_422_texcoords = macroblock_verts;
+
+/*
+ * Represents texcoords for the above for rendering 4 chroma blocks.
+ * Same case as 4 luma blocks.
+ */
+const struct vlVertex2f *macroblock_chroma_444_texcoords = macroblock_luma_texcoords;
+
+/*
+ * Used when rendering P and B macroblocks, multiplier is applied to the A channel,
+ * which is then added to the L channel, then the bias is subtracted from that to
+ * get back the differential. The differential is then added to the samples from the
+ * reference surface(s).
+ */
+const struct vlFragmentShaderConsts fs_consts =
+{
+	{32767.0f / 255.0f, 32767.0f / 255.0f, 32767.0f / 255.0f, 0.0f},
+	{0.5f, 2.0f, 0.0f, 0.0f}
+};
+
+static int vlCreateVertexShaderIMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 50;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma texcoords to output
+	 */
+	for (i = 1; i < 3; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->i_vs = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderIMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header,max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul o0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->i_fs = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFramePMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 * decl c2		; Unused
+	 * decl c3		; Translation vector to move ref macroblock texcoords into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 3);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 * decl o3		; Ref macroblock texcoords
+	 */
+	for (i = 0; i < 4; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma texcoords to output
+	 */
+	for (i = 1; i < 3; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* add o3, t0, c3	; Translate rect into position on ref macroblock */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->p_vs[0] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFieldPMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration
+		(
+			&decl,
+			&tokens[ti],
+			header,
+			max_tokens - ti
+		);
+	}
+
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 * decl c2		; Denorm coefficients
+	 * decl c3		; Translation vector to move top field ref macroblock texcoords into position
+	 * decl c4		; Translation vector to move bottom field ref macroblock texcoords into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 * decl o3		; Top field ref macroblock texcoords
+	 * decl o4		; Bottom field ref macroblock texcoords
+	 * decl o5		; Denormalized vertex pos
+	 */
+	for (i = 0; i < 6; i++)
+	{
+		decl = vl_decl_output((i == 0 || i == 5) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add t1, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mov o0, t1		; Move vertex pos to output */
+	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	mov o1, i1		; Move input luma texcoords to output
+	mov o2, i2		; Move input chroma texcoords to output
+	*/
+	for (i = 1; i < 3; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* add o3, t0, c3	; Translate top field rect into position on ref macroblock
+	   add o4, t0, c4	; Translate bottom field rect into position on ref macroblock */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul o5, t1, c2	; Denorm vertex pos */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 5, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->p_vs[1] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFramePMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 * decl i2			; Texcoords for s3
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for ref surface texture
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* tex2d t1, i2, s3		; Read texel from ref macroblock */
+	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 2, TGSI_FILE_SAMPLER, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->p_fs[0] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFieldPMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 200;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 * decl i2			; Texcoords for s3
+	 * decl i3			; Texcoords for s3
+	 * decl i4			; Denormalized vertex pos
+	 */
+	for (i = 0; i < 5; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
+	 * decl c1			; Constants 1/2 & 2 in .x, .y channels for Y-mod-2 top/bottom field selection
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0-t4 */
+	decl = vl_decl_temps(0, 4);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for ref surface texture
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t1, i2, s3		; Read texel from ref macroblock top field
+	 * tex2d t2, i3, s3		; Read texel from ref macroblock bottom field
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* XXX: Pos values off by 0.5? */
+	/* sub t4, i4.y, c1.x		; Sub 0.5 from denormalized pos */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 4, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* floor t3, t3			; Get rid of fractional part */
+	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t3, c1.y		; Multiply by 2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->p_fs[1] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFrameBMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 * decl c2		; Unused
+	 * decl c3		; Translation vector to move past ref macroblock texcoords into position
+	 * decl c4		; Unused
+	 * decl c5		; Translation vector to move future ref macroblock texcoords into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 5);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 * decl o3		; Past ref macroblock texcoords
+	 * decl o4		; Future ref macroblock texcoords
+	 */
+	for (i = 0; i < 5; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl t0 */
+	decl = vl_decl_temps(0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma texcoords to output
+	 */
+	for (i = 1; i < 3; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* add o3, t0, c3	; Translate rect into position on past ref macroblock
+	   add o4, t0, c5	; Translate rect into position on future ref macroblock */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i * 2 + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->b_vs[0] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFieldBMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos
+	 * decl i1		; Luma texcoords
+	 * decl i2		; Chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0		; Scaling vector to scale unit rect to macroblock size
+	 * decl c1		; Translation vector to move macroblock into position
+	 * decl c2		; Denorm coefficients
+	 * decl c3		; Translation vector to move top field past ref macroblock texcoords into position
+	 * decl c4		; Translation vector to move bottom field past ref macroblock texcoords into position
+	 * decl c5		; Translation vector to move top field future ref macroblock texcoords into position
+	 * decl c6		; Translation vector to move bottom field future ref macroblock texcoords into position
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 6);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma texcoords
+	 * decl o2		; Chroma texcoords
+	 * decl o3		; Top field past ref macroblock texcoords
+	 * decl o4		; Bottom field past ref macroblock texcoords
+	 * decl o5		; Top field future ref macroblock texcoords
+	 * decl o6		; Bottom field future ref macroblock texcoords
+	 * decl o7		; Denormalized vertex pos
+	 */
+	for (i = 0; i < 8; i++)
+	{
+		decl = vl_decl_output((i == 0 || i == 7) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add t1, t0, c1	; Translate rect into position */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mov o0, t1		; Move vertex pos to output */
+	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * mov o1, i1		; Move input luma texcoords to output
+	 * mov o2, i2		; Move input chroma texcoords to output
+	 */
+	for (i = 1; i < 3; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * add o3, t0, c3	; Translate top field rect into position on past ref macroblock
+	 * add o4, t0, c4	; Translate bottom field rect into position on past ref macroblock
+	 * add o5, t0, c5	; Translate top field rect into position on future ref macroblock
+	 * add o6, t0, c6	; Translate bottom field rect into position on future ref macroblock
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul o7, t1, c2	; Denorm vertex pos */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 7, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->b_vs[1] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFrameBMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 * decl i2			; Texcoords for s3
+	 * decl i3			; Texcoords for s4
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
+	 * decl c1			; Constant 1/2 in .x channel to use as weight to blend past and future texels
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0-t2 */
+	decl = vl_decl_temps(0, 2);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for past ref surface texture
+	 * decl s4			; Sampler for future ref surface texture
+	 */
+	for (i = 0; i < 5; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t1, i2, s3		; Read texel from past ref macroblock
+	 * tex2d t2, i3, s4		; Read texel from future ref macroblock
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, i + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->b_fs[0] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFieldBMB
+(
+	struct vlR16SnormMC *mc
+)
+{
+	const unsigned int		max_tokens = 200;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Texcoords for s0
+	 * decl i1			; Texcoords for s1, s2
+	 * decl i2			; Texcoords for s3
+	 * decl i3			; Texcoords for s3
+	 * decl i4			; Texcoords for s4
+	 * decl i5			; Texcoords for s4
+	 * decl i6			; Denormalized vertex pos
+	 */
+	for (i = 0; i < 7; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
+	 * decl c1			; Constants 1/2 & 2 in .x, .y channels to use as weight to blend past and future texels
+	 *				; and for Y-mod-2 top/bottom field selection
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0-t5 */
+	decl = vl_decl_temps(0, 5);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for past ref surface texture
+	 * decl s4			; Sampler for future ref surface texture
+	 */
+	for (i = 0; i < 5; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* XXX: Pos values off by 0.5? */
+	/* sub t4, i6.y, c1.x		; Sub 0.5 from denormalized pos */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 6, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* floor t3, t3			; Get rid of fractional part */
+	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t3, c1.y		; Multiply by 2 */
+	inst = vl_inst3( TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t1, i2, s3		; Read texel from past ref macroblock top field
+	 * tex2d t2, i3, s3		; Read texel from past ref macroblock bottom field
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t4, i4, s4		; Read texel from future ref macroblock top field
+	 * tex2d t5, i5, s4		; Read texel from future ref macroblock bottom field
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 4, TGSI_FILE_INPUT, i + 4, TGSI_FILE_SAMPLER, 4);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t2, t3, t4, t5		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 2, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 5);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->b_fs[1] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+int vlCreateDataBufs
+(
+	struct vlR16SnormMC *mc
+)
+{
+	struct pipe_context	*pipe;
+	unsigned int		i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+
+	/* Create our vertex buffer and vertex buffer element */
+	mc->vertex_bufs[0].pitch = sizeof(struct vlVertex2f);
+	mc->vertex_bufs[0].max_index = 23;
+	mc->vertex_bufs[0].buffer_offset = 0;
+	mc->vertex_bufs[0].buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_VERTEX,
+		sizeof(struct vlVertex2f) * 24
+	);
+
+	mc->vertex_elems[0].src_offset = 0;
+	mc->vertex_elems[0].vertex_buffer_index = 0;
+	mc->vertex_elems[0].nr_components = 2;
+	mc->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Create our texcoord buffers and texcoord buffer elements */
+	for (i = 1; i < 3; ++i)
+	{
+		mc->vertex_bufs[i].pitch = sizeof(struct vlVertex2f);
+		mc->vertex_bufs[i].max_index = 23;
+		mc->vertex_bufs[i].buffer_offset = 0;
+		mc->vertex_bufs[i].buffer = pipe->winsys->buffer_create
+		(
+			pipe->winsys,
+			1,
+			PIPE_BUFFER_USAGE_VERTEX,
+			sizeof(struct vlVertex2f) * 24
+		);
+
+		mc->vertex_elems[i].src_offset = 0;
+		mc->vertex_elems[i].vertex_buffer_index = i;
+		mc->vertex_elems[i].nr_components = 2;
+		mc->vertex_elems[i].src_format = PIPE_FORMAT_R32G32_FLOAT;
+	}
+
+	/* Fill buffers */
+	memcpy
+	(
+		pipe->winsys->buffer_map(pipe->winsys, mc->vertex_bufs[0].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		macroblock_verts,
+		sizeof(struct vlVertex2f) * 24
+	);
+	memcpy
+	(
+		pipe->winsys->buffer_map(pipe->winsys, mc->vertex_bufs[1].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		macroblock_luma_texcoords,
+		sizeof(struct vlVertex2f) * 24
+	);
+	/* TODO: Accomodate 422, 444 */
+	memcpy
+	(
+		pipe->winsys->buffer_map(pipe->winsys, mc->vertex_bufs[2].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		macroblock_chroma_420_texcoords,
+		sizeof(struct vlVertex2f) * 24
+	);
+
+	for (i = 0; i < 3; ++i)
+		pipe->winsys->buffer_unmap(pipe->winsys, mc->vertex_bufs[i].buffer);
+
+	/* Create our constant buffer */
+	mc->vs_const_buf.size = sizeof(struct vlVertexShaderConsts);
+	mc->vs_const_buf.buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_CONSTANT,
+		mc->vs_const_buf.size
+	);
+
+	mc->fs_const_buf.size = sizeof(struct vlFragmentShaderConsts);
+	mc->fs_const_buf.buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_CONSTANT,
+		mc->fs_const_buf.size
+	);
+
+	memcpy
+	(
+		pipe->winsys->buffer_map(pipe->winsys, mc->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		&fs_consts,
+		sizeof(struct vlFragmentShaderConsts)
+	);
+
+	pipe->winsys->buffer_unmap(pipe->winsys, mc->fs_const_buf.buffer);
+
+	return 0;
+}
+
+static int vlInit
+(
+	struct vlR16SnormMC *mc
+)
+{
+	struct pipe_context		*pipe;
+	struct pipe_sampler_state	sampler;
+	struct pipe_texture		template;
+	unsigned int			filters[5];
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+
+	/* For MC we render to textures, which are rounded up to nearest POT */
+	mc->viewport.scale[0] = vlRoundUpPOT(mc->video_width);
+	mc->viewport.scale[1] = vlRoundUpPOT(mc->video_height);
+	mc->viewport.scale[2] = 1;
+	mc->viewport.scale[3] = 1;
+	mc->viewport.translate[0] = 0;
+	mc->viewport.translate[1] = 0;
+	mc->viewport.translate[2] = 0;
+	mc->viewport.translate[3] = 0;
+
+	mc->render_target.width = vlRoundUpPOT(mc->video_width);
+	mc->render_target.height = vlRoundUpPOT(mc->video_height);
+	mc->render_target.num_cbufs = 1;
+	/* FB for MC stage is a vlSurface, set in vlSetRenderSurface() */
+	mc->render_target.zsbuf = NULL;
+
+	filters[0] = PIPE_TEX_FILTER_NEAREST;
+	filters[1] = mc->video_format == vlFormatYCbCr444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
+	filters[2] = mc->video_format == vlFormatYCbCr444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
+	filters[3] = PIPE_TEX_FILTER_LINEAR;
+	filters[4] = PIPE_TEX_FILTER_LINEAR;
+
+	for (i = 0; i < 5; ++i)
+	{
+		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_img_filter = filters[i];
+		sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+		sampler.mag_img_filter = filters[i];
+		sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
+		sampler.compare_func = PIPE_FUNC_ALWAYS;
+		sampler.normalized_coords = 1;
+		/*sampler.prefilter = ;*/
+		/*sampler.shadow_ambient = ;*/
+		/*sampler.lod_bias = ;*/
+		sampler.min_lod = 0;
+		/*sampler.max_lod = ;*/
+		/*sampler.border_color[i] = ;*/
+		/*sampler.max_anisotropy = ;*/
+		mc->samplers[i] = pipe->create_sampler_state(pipe, &sampler);
+	}
+
+	memset(&template, 0, sizeof(struct pipe_texture));
+	template.target = PIPE_TEXTURE_2D;
+	template.format = PIPE_FORMAT_R16_SNORM;
+	template.last_level = 0;
+	template.width[0] = 8;
+	template.height[0] = 8 * 4;
+	template.depth[0] = 1;
+	template.compressed = 0;
+	pf_get_block(template.format, &template.block);
+
+	mc->textures[0] = pipe->screen->texture_create(pipe->screen, &template);
+
+	if (mc->video_format == vlFormatYCbCr420)
+		template.height[0] = 8;
+	else if (mc->video_format == vlFormatYCbCr422)
+		template.height[0] = 8 * 2;
+	else if (mc->video_format == vlFormatYCbCr444)
+		template.height[0] = 8 * 4;
+	else
+		assert(0);
+
+	mc->textures[1] = pipe->screen->texture_create(pipe->screen, &template);
+	mc->textures[2] = pipe->screen->texture_create(pipe->screen, &template);
+
+	/* textures[3] & textures[4] are assigned from vlSurfaces for P and B macroblocks at render time */
+
+	vlCreateVertexShaderIMB(mc);
+	vlCreateFragmentShaderIMB(mc);
+	vlCreateVertexShaderFramePMB(mc);
+	vlCreateVertexShaderFieldPMB(mc);
+	vlCreateFragmentShaderFramePMB(mc);
+	vlCreateFragmentShaderFieldPMB(mc);
+	vlCreateVertexShaderFrameBMB(mc);
+	vlCreateVertexShaderFieldBMB(mc);
+	vlCreateFragmentShaderFrameBMB(mc);
+	vlCreateFragmentShaderFieldBMB(mc);
+	vlCreateDataBufs(mc);
+
+	return 0;
+}
+
+int vlCreateR16SNormMC
+(
+	struct pipe_context *pipe,
+	unsigned int video_width,
+	unsigned int video_height,
+	enum vlFormat video_format,
+	struct vlRender **render
+)
+{
+	struct vlR16SnormMC *mc;
+
+	assert(pipe);
+	assert(render);
+
+	mc = calloc(1, sizeof(struct vlR16SnormMC));
+
+	mc->base.vlBegin = &vlBegin;
+	mc->base.vlRenderMacroBlocksMpeg2 = &vlRenderMacroBlocksMpeg2R16Snorm;
+	mc->base.vlEnd = &vlEnd;
+	mc->base.vlDestroy = &vlDestroy;
+	mc->pipe = pipe;
+	mc->video_width = video_width;
+	mc->video_height = video_height;
+
+	vlInit(mc);
+
+	*render = &mc->base;
+
+	return 0;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h
new file mode 100644
index 0000000000..a6eecf05b6
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h
@@ -0,0 +1,18 @@
+#ifndef vl_mc_h
+#define vl_mc_h
+
+#include "vl_types.h"
+
+struct pipe_context;
+struct vlRender;
+
+int vlCreateR16SNormMC
+(
+	struct pipe_context *pipe,
+	unsigned int video_width,
+	unsigned int video_height,
+	enum vlFormat video_format,
+	struct vlRender **render
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_render.h b/src/gallium/state_trackers/g3dvl/vl_render.h
new file mode 100644
index 0000000000..63016b5cbe
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_render.h
@@ -0,0 +1,33 @@
+#ifndef vl_render_h
+#define vl_render_h
+
+#include "vl_types.h"
+
+struct pipe_surface;
+
+struct vlRender
+{
+	int (*vlBegin)
+	(
+		struct vlRender *render
+	);
+
+	int (*vlRenderMacroBlocksMpeg2)
+	(
+		struct vlRender *render,
+		struct vlMpeg2MacroBlockBatch *batch,
+		struct vlSurface *surface
+	);
+
+	int (*vlEnd)
+	(
+		struct vlRender *render
+	);
+
+	int (*vlDestroy)
+	(
+		struct vlRender *render
+	);
+};
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_screen.c b/src/gallium/state_trackers/g3dvl/vl_screen.c
new file mode 100644
index 0000000000..484f63b0d4
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_screen.c
@@ -0,0 +1,115 @@
+#define VL_INTERNAL
+#include "vl_screen.h"
+#include <assert.h>
+#include <stdlib.h>
+
+int vlCreateScreen
+(
+	struct vlDisplay *display,
+	int screen,
+	struct pipe_screen *pscreen,
+	struct vlScreen **vl_screen
+)
+{
+	struct vlScreen *scrn;
+
+	assert(display);
+	assert(pscreen);
+	assert(vl_screen);
+
+	scrn = calloc(1, sizeof(struct vlScreen));
+
+	if (!scrn)
+		return 1;
+
+	scrn->display = display;
+	scrn->ordinal = screen;
+	scrn->pscreen = pscreen;
+	*vl_screen = scrn;
+
+	return 0;
+}
+
+int vlDestroyScreen
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	free(screen);
+
+	return 0;
+}
+
+struct vlDisplay* vlGetDisplay
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	return screen->display;
+}
+
+struct pipe_screen* vlGetPipeScreen
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	return screen->pscreen;
+}
+
+unsigned int vlGetMaxProfiles
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	return vlProfileCount;
+}
+
+int vlQueryProfiles
+(
+	struct vlScreen *screen,
+	enum vlProfile *profiles
+)
+{
+	assert(screen);
+	assert(profiles);
+
+	profiles[0] = vlProfileMpeg2Simple;
+	profiles[1] = vlProfileMpeg2Main;
+
+	return 0;
+}
+
+unsigned int vlGetMaxEntryPoints
+(
+	struct vlScreen *screen
+)
+{
+	assert(screen);
+
+	return vlEntryPointCount;
+}
+
+int vlQueryEntryPoints
+(
+	struct vlScreen *screen,
+	enum vlProfile profile,
+	enum vlEntryPoint *entry_points
+)
+{
+	assert(screen);
+	assert(entry_points);
+
+	entry_points[0] = vlEntryPointIDCT;
+	entry_points[1] = vlEntryPointMC;
+	entry_points[2] = vlEntryPointCSC;
+
+	return 0;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_screen.h b/src/gallium/state_trackers/g3dvl/vl_screen.h
new file mode 100644
index 0000000000..98f3d429b6
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_screen.h
@@ -0,0 +1,63 @@
+#ifndef vl_screen_h
+#define vl_screen_h
+
+#include "vl_types.h"
+
+struct pipe_screen;
+
+#ifdef VL_INTERNAL
+struct vlScreen
+{
+	struct vlDisplay	*display;
+	unsigned int		ordinal;
+	struct pipe_screen	*pscreen;
+};
+#endif
+
+int vlCreateScreen
+(
+	struct vlDisplay *display,
+	int screen,
+	struct pipe_screen *pscreen,
+	struct vlScreen **vl_screen
+);
+
+int vlDestroyScreen
+(
+	struct vlScreen *screen
+);
+
+struct vlDisplay* vlGetDisplay
+(
+	struct vlScreen *screen
+);
+
+struct pipe_screen* vlGetPipeScreen
+(
+	struct vlScreen *screen
+);
+
+unsigned int vlGetMaxProfiles
+(
+	struct vlScreen *screen
+);
+
+int vlQueryProfiles
+(
+	struct vlScreen *screen,
+	enum vlProfile *profiles
+);
+
+unsigned int vlGetMaxEntryPoints
+(
+	struct vlScreen *screen
+);
+
+int vlQueryEntryPoints
+(
+	struct vlScreen *screen,
+	enum vlProfile profile,
+	enum vlEntryPoint *entry_points
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.c b/src/gallium/state_trackers/g3dvl/vl_shader_build.c
index 5f30e23ff8..51f1721a33 100644
--- a/src/gallium/state_trackers/g3dvl/vl_shader_build.c
+++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.c
@@ -13,7 +13,7 @@ struct tgsi_full_declaration vl_decl_input(unsigned int name, unsigned int index
 	decl.Semantic.SemanticIndex = index;
 	decl.DeclarationRange.First = first;
 	decl.DeclarationRange.Last = last;
-	
+
 	return decl;
 }
 
@@ -27,7 +27,7 @@ struct tgsi_full_declaration vl_decl_interpolated_input
 )
 {
 	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
-	
+
 	assert
 	(
 		interpolation == TGSI_INTERPOLATE_CONSTANT ||
@@ -42,21 +42,21 @@ struct tgsi_full_declaration vl_decl_interpolated_input
 	decl.Declaration.Interpolate = interpolation;;
 	decl.DeclarationRange.First = first;
 	decl.DeclarationRange.Last = last;
-	
+
 	return decl;
 }
 
 struct tgsi_full_declaration vl_decl_constants(unsigned int name, unsigned int index, unsigned int first, unsigned int last)
 {
 	struct tgsi_full_declaration decl = tgsi_default_full_declaration();
-	
+
 	decl.Declaration.File = TGSI_FILE_CONSTANT;
 	decl.Declaration.Semantic = 1;
 	decl.Semantic.SemanticName = name;
 	decl.Semantic.SemanticIndex = index;
 	decl.DeclarationRange.First = first;
 	decl.DeclarationRange.Last = last;
-	
+
 	return decl;
 }
 
@@ -70,7 +70,7 @@ struct tgsi_full_declaration vl_decl_output(unsigned int name, unsigned int inde
 	decl.Semantic.SemanticIndex = index;
 	decl.DeclarationRange.First = first;
 	decl.DeclarationRange.Last = last;
-	
+
 	return decl;
 }
 
@@ -82,7 +82,7 @@ struct tgsi_full_declaration vl_decl_temps(unsigned int first, unsigned int last
 	decl.Declaration.File = TGSI_FILE_TEMPORARY;
 	decl.DeclarationRange.First = first;
 	decl.DeclarationRange.Last = last;
-	
+
 	return decl;
 }
 
@@ -94,7 +94,7 @@ struct tgsi_full_declaration vl_decl_samplers(unsigned int first, unsigned int l
 	decl.Declaration.File = TGSI_FILE_SAMPLER;
 	decl.DeclarationRange.First = first;
 	decl.DeclarationRange.Last = last;
-	
+
 	return decl;
 }
 
@@ -108,7 +108,7 @@ struct tgsi_full_instruction vl_inst2
 )
 {
 	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
-	
+
 	inst.Instruction.Opcode = opcode;
 	inst.Instruction.NumDstRegs = 1;
 	inst.FullDstRegisters[0].DstRegister.File = dst_file;
@@ -116,7 +116,7 @@ struct tgsi_full_instruction vl_inst2
 	inst.Instruction.NumSrcRegs = 1;
 	inst.FullSrcRegisters[0].SrcRegister.File = src_file;
 	inst.FullSrcRegisters[0].SrcRegister.Index = src_index;
-	
+
 	return inst;
 }
 
@@ -132,7 +132,7 @@ struct tgsi_full_instruction vl_inst3
 )
 {
 	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
-	
+
 	inst.Instruction.Opcode = opcode;
 	inst.Instruction.NumDstRegs = 1;
 	inst.FullDstRegisters[0].DstRegister.File = dst_file;
@@ -142,7 +142,7 @@ struct tgsi_full_instruction vl_inst3
 	inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
 	inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
 	inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
-	
+
 	return inst;
 }
 
@@ -158,7 +158,7 @@ struct tgsi_full_instruction vl_tex
 )
 {
 	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
-	
+
 	inst.Instruction.Opcode = TGSI_OPCODE_TEX;
 	inst.Instruction.NumDstRegs = 1;
 	inst.FullDstRegisters[0].DstRegister.File = dst_file;
@@ -169,7 +169,7 @@ struct tgsi_full_instruction vl_tex
 	inst.FullSrcRegisters[0].SrcRegister.Index = src1_index;
 	inst.FullSrcRegisters[1].SrcRegister.File = src2_file;
 	inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
-	
+
 	return inst;
 }
 
@@ -187,7 +187,7 @@ struct tgsi_full_instruction vl_inst4
 )
 {
 	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
-	
+
 	inst.Instruction.Opcode = opcode;
 	inst.Instruction.NumDstRegs = 1;
 	inst.FullDstRegisters[0].DstRegister.File = dst_file;
@@ -199,18 +199,17 @@ struct tgsi_full_instruction vl_inst4
 	inst.FullSrcRegisters[1].SrcRegister.Index = src2_index;
 	inst.FullSrcRegisters[2].SrcRegister.File = src3_file;
 	inst.FullSrcRegisters[2].SrcRegister.Index = src3_index;
-	
+
 	return inst;
 }
 
 struct tgsi_full_instruction vl_end(void)
 {
 	struct tgsi_full_instruction inst = tgsi_default_full_instruction();
-	
+
 	inst.Instruction.Opcode = TGSI_OPCODE_END;
 	inst.Instruction.NumDstRegs = 0;
 	inst.Instruction.NumSrcRegs = 0;
-	
+
 	return inst;
 }
-
diff --git a/src/gallium/state_trackers/g3dvl/vl_shader_build.h b/src/gallium/state_trackers/g3dvl/vl_shader_build.h
index 878d7e2c45..dc615cb156 100644
--- a/src/gallium/state_trackers/g3dvl/vl_shader_build.h
+++ b/src/gallium/state_trackers/g3dvl/vl_shader_build.h
@@ -59,4 +59,3 @@ struct tgsi_full_instruction vl_inst4
 struct tgsi_full_instruction vl_end(void);
 
 #endif
-
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c
index 1386b1107c..ffc8122172 100644
--- a/src/gallium/state_trackers/g3dvl/vl_surface.c
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.c
@@ -1,628 +1,177 @@
+#define VL_INTERNAL
 #include "vl_surface.h"
 #include <assert.h>
 #include <stdlib.h>
-#include <pipe/p_context.h>
+#include <string.h>
+#include <pipe/p_screen.h>
 #include <pipe/p_state.h>
-#include <pipe/p_format.h>
 #include <pipe/p_inlines.h>
 #include <vl_winsys.h>
+#include "vl_screen.h"
 #include "vl_context.h"
-#include "vl_defs.h"
+#include "vl_render.h"
+#include "vl_csc.h"
 #include "vl_util.h"
 
-/*#define DO_IDCT*/
-
-#ifdef DO_IDCT
-static int vlTransformBlock(short *src, short *dst, short bias)
+int vlCreateSurface
+(
+	struct vlScreen *screen,
+	unsigned int width,
+	unsigned int height,
+	enum vlFormat format,
+	struct vlSurface **surface
+)
 {
-	static const float basis[8][8] =
-	{
-		{0.3536,   0.4904,   0.4619,   0.4157,   0.3536,   0.2778,   0.1913,   0.0975},
-		{0.3536,   0.4157,   0.1913,  -0.0975,  -0.3536,  -0.4904,  -0.4619,  -0.2778},
-		{0.3536,   0.2778,  -0.1913,  -0.4904,  -0.3536,   0.0975,   0.4619,   0.4157},
-		{0.3536,   0.0975,  -0.4619,  -0.2778,   0.3536,   0.4157,  -0.1913,  -0.4904},
-		{0.3536,  -0.0975,  -0.4619,   0.2778,   0.3536,  -0.4157,  -0.1913,   0.4904},
-		{0.3536,  -0.2778,  -0.1913,   0.4904,  -0.3536,  -0.0975,   0.4619,  -0.4157},
-		{0.3536,  -0.4157,   0.1913,   0.0975,  -0.3536,   0.4904,  -0.4619,   0.2778},
-		{0.3536,  -0.4904,   0.4619,  -0.4157,   0.3536,  -0.2778,   0.1913,  -0.0975}
-	};
-	
-	unsigned int	x, y;
-	short		tmp[64];
-	
-	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
-		for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-			tmp[y * VL_BLOCK_WIDTH + x] = (short)
-			(
-				src[y * VL_BLOCK_WIDTH + 0] * basis[x][0] +
-				src[y * VL_BLOCK_WIDTH + 1] * basis[x][1] +
-				src[y * VL_BLOCK_WIDTH + 2] * basis[x][2] +
-				src[y * VL_BLOCK_WIDTH + 3] * basis[x][3] +
-				src[y * VL_BLOCK_WIDTH + 4] * basis[x][4] +
-				src[y * VL_BLOCK_WIDTH + 5] * basis[x][5] +
-				src[y * VL_BLOCK_WIDTH + 6] * basis[x][6] +
-				src[y * VL_BLOCK_WIDTH + 7] * basis[x][7]
-			);
-
-	for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-		for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
-		{
-			dst[y * VL_BLOCK_WIDTH + x] = bias + (short)
-			(
-				tmp[0 * VL_BLOCK_WIDTH + x] * basis[y][0] +
-				tmp[1 * VL_BLOCK_WIDTH + x] * basis[y][1] +
-				tmp[2 * VL_BLOCK_WIDTH + x] * basis[y][2] +
-				tmp[3 * VL_BLOCK_WIDTH + x] * basis[y][3] +
-				tmp[4 * VL_BLOCK_WIDTH + x] * basis[y][4] +
-				tmp[5 * VL_BLOCK_WIDTH + x] * basis[y][5] +
-				tmp[6 * VL_BLOCK_WIDTH + x] * basis[y][6] +
-				tmp[7 * VL_BLOCK_WIDTH + x] * basis[y][7]
-			);
-			if (dst[y * VL_BLOCK_WIDTH + x] > 255)
-				dst[y * VL_BLOCK_WIDTH + x] = 255;
-			else if (bias > 0 && dst[y * VL_BLOCK_WIDTH + x] < 0)
-				dst[y * VL_BLOCK_WIDTH + x] = 0;
-		}
-	return 0;
-}
-#endif
+	struct vlSurface	*sfc;
+	struct pipe_texture	template;
 
-static int vlGrabFrameCodedBlock(short *src, short *dst, unsigned int dst_pitch)
-{
-	unsigned int y;
-	
-	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
-		memcpy
-		(
-			dst + y * dst_pitch,
-			src + y * VL_BLOCK_WIDTH,
-			VL_BLOCK_WIDTH * 2
-		);
-	
-	return 0;
-}
+	assert(screen);
+	assert(surface);
 
-static int vlGrabFieldCodedBlock(short *src, short *dst, unsigned int dst_pitch)
-{
-	unsigned int y;
-	
-	for (y = 0; y < VL_BLOCK_HEIGHT / 2; ++y)
-		memcpy
-		(
-			dst + y * dst_pitch * 2,
-			src + y * VL_BLOCK_WIDTH,
-			VL_BLOCK_WIDTH * 2
-		);
-	
-	dst += VL_BLOCK_HEIGHT * dst_pitch;
-	
-	for (; y < VL_BLOCK_HEIGHT; ++y)
-		memcpy
-		(
-			dst + y * dst_pitch * 2,
-			src + y * VL_BLOCK_WIDTH,
-			VL_BLOCK_WIDTH * 2
-		);
-	
-	return 0;
-}
+	sfc = calloc(1, sizeof(struct vlSurface));
 
-static int vlGrabNoBlock(short *dst, unsigned int dst_pitch)
-{
-	unsigned int y;
-	
-	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
-		memset
-		(
-			dst + y * dst_pitch,
-			0,
-			VL_BLOCK_WIDTH * 2
-		);
-	
-	return 0;
-}
+	if (!sfc)
+		return 1;
 
-static int vlGrabBlocks
-(
-	struct VL_CONTEXT *context,
-	unsigned int coded_block_pattern,
-	enum VL_DCT_TYPE dct_type,
-	enum VL_SAMPLE_TYPE sample_type,
-	short *blocks
-)
-{
-	struct pipe_surface	*tex_surface;
-	short			*texels;
-	unsigned int		tex_pitch;
-	unsigned int		tb, sb = 0;
-	
-	assert(context);
-	assert(blocks);
-	
-	tex_surface = context->pipe->screen->get_tex_surface
-	(
-		context->pipe->screen,
-		context->states.mc.textures[0],
-		0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
-	);
-	
-	texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE);
-	tex_pitch = tex_surface->stride / tex_surface->block.size;
-	
-	for (tb = 0; tb < 4; ++tb)
-	{
-		if ((coded_block_pattern >> (5 - tb)) & 1)
-		{
-			short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
-			
-#ifdef DO_IDCT
-			vlTransformBlock(cur_block, cur_block, sample_type == VL_FULL_SAMPLE ? 128 : 0);
-#endif
-			
-			if (dct_type == VL_DCT_FRAME_CODED)
-				vlGrabFrameCodedBlock
-				(
-					cur_block,
-					texels + tb * tex_pitch * VL_BLOCK_HEIGHT,
-					tex_pitch
-				);
-			else
-				vlGrabFieldCodedBlock
-				(
-					cur_block,
-					texels + (tb % 2) * tex_pitch * VL_BLOCK_HEIGHT + (tb / 2) * tex_pitch,
-					tex_pitch
-				);
-			
-			++sb;
-		}
-		else
-			vlGrabNoBlock(texels + tb * tex_pitch * VL_BLOCK_HEIGHT, tex_pitch);
-	}
-	
-	pipe_surface_unmap(tex_surface);
-	
-	/* TODO: Implement 422, 444 */
-	for (tb = 0; tb < 2; ++tb)
-	{
-		tex_surface = context->pipe->screen->get_tex_surface
-			(
-				context->pipe->screen,
-				context->states.mc.textures[tb + 1],
-				0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
-			);
-	
-		texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE);
-		tex_pitch = tex_surface->stride / tex_surface->block.size;
-		
-		if ((coded_block_pattern >> (1 - tb)) & 1)
-		{
-			short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
-			
-#ifdef DO_IDCT
-			vlTransformBlock(cur_block, cur_block, sample_type == VL_FULL_SAMPLE ? 128 : 0);
-#endif
-			
-			vlGrabFrameCodedBlock
-			(
-				cur_block,
-				texels,
-				tex_pitch
-			);
-			
-			++sb;
-		}
-		else
-			vlGrabNoBlock(texels, tex_pitch);
-		
-		pipe_surface_unmap(tex_surface);
-	}
-	
-	return 0;
-}
+	sfc->screen = screen;
+	sfc->width = width;
+	sfc->height = height;
+	sfc->format = format;
 
-int vlCreateSurface(struct VL_CONTEXT *context, struct VL_SURFACE **surface)
-{
-	struct pipe_context	*pipe;
-	struct pipe_texture	template;
-	struct VL_SURFACE	*sfc;
-	
-	assert(context);
-	assert(surface);
-	
-	pipe = context->pipe;
-	
-	sfc = calloc(1, sizeof(struct VL_SURFACE));
-	
-	sfc->context = context;
-	sfc->width = vlRoundUpPOT(context->video_width);
-	sfc->height = vlRoundUpPOT(context->video_height);
-	sfc->format = context->video_format;
-	
 	memset(&template, 0, sizeof(struct pipe_texture));
 	template.target = PIPE_TEXTURE_2D;
 	template.format = PIPE_FORMAT_A8R8G8B8_UNORM;
 	template.last_level = 0;
-	template.width[0] = sfc->width;
-	template.height[0] = sfc->height;
+	template.width[0] = vlRoundUpPOT(sfc->width);
+	template.height[0] = vlRoundUpPOT(sfc->height);
 	template.depth[0] = 1;
 	template.compressed = 0;
 	pf_get_block(template.format, &template.block);
-	/* XXX: Needed? */
 	template.tex_usage = PIPE_TEXTURE_USAGE_SAMPLER | PIPE_TEXTURE_USAGE_RENDER_TARGET;
-	
-	sfc->texture = pipe->screen->texture_create(pipe->screen, &template);
-	
+
+	sfc->texture = vlGetPipeScreen(screen)->texture_create(vlGetPipeScreen(screen), &template);
+
 	*surface = sfc;
-	
+
 	return 0;
 }
 
-int vlDestroySurface(struct VL_SURFACE *surface)
+int vlDestroySurface
+(
+	struct vlSurface *surface
+)
 {
 	assert(surface);
+
 	pipe_texture_release(&surface->texture);
 	free(surface);
-	
+
 	return 0;
 }
 
-int vlRenderIMacroBlock
+int vlRenderMacroBlocksMpeg2
 (
-	enum VL_PICTURE picture_type,
-	enum VL_FIELD_ORDER field_order,
-	unsigned int mbx,
-	unsigned int mby,
-	unsigned int coded_block_pattern,
-	enum VL_DCT_TYPE dct_type,
-	short *blocks,
-	struct VL_SURFACE *surface
+	struct vlMpeg2MacroBlockBatch *batch,
+	struct vlSurface *surface
 )
 {
-	struct pipe_context	*pipe;
-	struct VL_MC_VS_CONSTS	*vs_consts;
-	
-	assert(blocks);
+	assert(batch);
 	assert(surface);
-	
-	/* TODO: Implement interlaced rendering */
-	if (picture_type != VL_FRAME_PICTURE)
-		return 0;
-	
-	vlGrabBlocks(surface->context, coded_block_pattern, dct_type, VL_FULL_SAMPLE, blocks);
-	
-	pipe = surface->context->pipe;
-	
-	vs_consts = pipe->winsys->buffer_map
-	(
-		pipe->winsys,
-		surface->context->states.mc.vs_const_buf.buffer,
-		PIPE_BUFFER_USAGE_CPU_WRITE
-	);
-	
-	vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->width;
-	vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->height;
-	vs_consts->scale.z = 1.0f;
-	vs_consts->scale.w = 1.0f;
-	vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->width;
-	vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->height;
-	vs_consts->mb_pos_trans.z = 0.0f;
-	vs_consts->mb_pos_trans.w = 0.0f;
-	
-	pipe->winsys->buffer_unmap(pipe->winsys, surface->context->states.mc.vs_const_buf.buffer);
-	
-	surface->context->states.mc.render_target.cbufs[0] = pipe->screen->get_tex_surface
+
+	surface->context->render->vlBegin(surface->context->render);
+
+	surface->context->render->vlRenderMacroBlocksMpeg2
 	(
-		pipe->screen,
-		surface->texture,
-		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
+		surface->context->render,
+		batch,
+		surface
 	);
-	pipe->set_framebuffer_state(pipe, &surface->context->states.mc.render_target);
-	pipe->set_sampler_textures(pipe, 3, surface->context->states.mc.textures);
-	pipe->bind_sampler_states(pipe, 3, (void**)surface->context->states.mc.samplers);
-	pipe->bind_vs_state(pipe, surface->context->states.mc.i_vs);
-	pipe->bind_fs_state(pipe, surface->context->states.mc.i_fs);
-	
-	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
-	
+
+	surface->context->render->vlEnd(surface->context->render);
+
 	return 0;
 }
 
-int vlRenderPMacroBlock
+int vlPutPicture
 (
-	enum VL_PICTURE picture_type,
-	enum VL_FIELD_ORDER field_order,
-	unsigned int mbx,
-	unsigned int mby,
-	enum VL_MC_TYPE mc_type,
-	struct VL_MOTION_VECTOR *motion_vector,
-	unsigned int coded_block_pattern,
-	enum VL_DCT_TYPE dct_type,
-	short *blocks,
-	struct VL_SURFACE *ref_surface,
-	struct VL_SURFACE *surface
+	struct vlSurface *surface,
+	vlNativeDrawable drawable,
+	int srcx,
+	int srcy,
+	int srcw,
+	int srch,
+	int destx,
+	int desty,
+	int destw,
+	int desth,
+	enum vlPictureType picture_type
 )
 {
+	struct vlCSC		*csc;
 	struct pipe_context	*pipe;
-	struct VL_MC_VS_CONSTS	*vs_consts;
-	
-	assert(motion_vectors);
-	assert(blocks);
-	assert(ref_surface);
+
 	assert(surface);
-	
-	/* TODO: Implement interlaced rendering */
-	if (picture_type != VL_FRAME_PICTURE)
-		return 0;
-	/* TODO: Implement other MC types */
-	if (mc_type != VL_FRAME_MC && mc_type != VL_FIELD_MC)
-		return 0;
-	
-	vlGrabBlocks(surface->context, coded_block_pattern, dct_type, VL_DIFFERENCE_SAMPLE, blocks);
-	
+	assert(surface->context);
+
+	csc = surface->context->csc;
 	pipe = surface->context->pipe;
-	
-	vs_consts = pipe->winsys->buffer_map
+
+	csc->vlResizeFrameBuffer(csc, destw, desth);
+
+	csc->vlBegin(csc);
+
+	csc->vlPutPicture
 	(
-		pipe->winsys,
-		surface->context->states.mc.vs_const_buf.buffer,
-		PIPE_BUFFER_USAGE_CPU_WRITE
+		csc,
+		surface,
+		srcx,
+		srcy,
+		srcw,
+		srch,
+		destx,
+		desty,
+		destw,
+		desth,
+		picture_type
 	);
-	
-	vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->width;
-	vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->height;
-	vs_consts->scale.z = 1.0f;
-	vs_consts->scale.w = 1.0f;
-	vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->width;
-	vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->height;
-	vs_consts->mb_pos_trans.z = 0.0f;
-	vs_consts->mb_pos_trans.w = 0.0f;
-	vs_consts->mb_tc_trans[0].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + motion_vector->top_field.x * 0.5f) / (float)surface->width;
-	vs_consts->mb_tc_trans[0].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + motion_vector->top_field.y * 0.5f) / (float)surface->height;
-	vs_consts->mb_tc_trans[0].top_field.z = 0.0f;
-	vs_consts->mb_tc_trans[0].top_field.w = 0.0f;
-	
-	if (mc_type == VL_FIELD_MC)
-	{
-		vs_consts->denorm.x = (float)surface->width;
-		vs_consts->denorm.y = (float)surface->height;
-		
-		vs_consts->mb_tc_trans[0].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + motion_vector->bottom_field.x * 0.5f) / (float)surface->width;
-		vs_consts->mb_tc_trans[0].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + motion_vector->bottom_field.y * 0.5f) / (float)surface->height;
-		vs_consts->mb_tc_trans[0].bottom_field.z = 0.0f;
-		vs_consts->mb_tc_trans[0].bottom_field.w = 0.0f;
-		
-		pipe->bind_vs_state(pipe, surface->context->states.mc.p_vs[1]);
-		pipe->bind_fs_state(pipe, surface->context->states.mc.p_fs[1]);
-	}
-	else
-	{
-		pipe->bind_vs_state(pipe, surface->context->states.mc.p_vs[0]);
-		pipe->bind_fs_state(pipe, surface->context->states.mc.p_fs[0]);
-	}
-	
-	pipe->winsys->buffer_unmap(pipe->winsys, surface->context->states.mc.vs_const_buf.buffer);
-	
-	surface->context->states.mc.render_target.cbufs[0] = pipe->screen->get_tex_surface
+
+	csc->vlEnd(csc);
+
+	pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
+	bind_pipe_drawable(pipe, drawable);
+	/* TODO: Need to take destx, desty into consideration */
+	pipe->winsys->flush_frontbuffer
 	(
-		pipe->screen,
-		surface->texture,
-		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
+		pipe->winsys,
+		csc->vlGetFrameBuffer(csc),
+		pipe->priv
 	);
-	pipe->set_framebuffer_state(pipe, &surface->context->states.mc.render_target);
-	
-	surface->context->states.mc.textures[3] = ref_surface->texture;
-	pipe->set_sampler_textures(pipe, 4, surface->context->states.mc.textures);
-	pipe->bind_sampler_states(pipe, 4, (void**)surface->context->states.mc.samplers);
-	
-	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
-	
+
 	return 0;
 }
 
-int vlRenderBMacroBlock
+struct vlScreen* vlSurfaceGetScreen
 (
-	enum VL_PICTURE picture_type,
-	enum VL_FIELD_ORDER field_order,
-	unsigned int mbx,
-	unsigned int mby,
-	enum VL_MC_TYPE mc_type,
-	struct VL_MOTION_VECTOR *motion_vector,
-	unsigned int coded_block_pattern,
-	enum VL_DCT_TYPE dct_type,
-	short *blocks,
-	struct VL_SURFACE *past_surface,
-	struct VL_SURFACE *future_surface,
-	struct VL_SURFACE *surface
+	struct vlSurface *surface
 )
 {
-	struct pipe_context	*pipe;
-	struct VL_MC_VS_CONSTS	*vs_consts;
-	
-	assert(motion_vectors);
-	assert(blocks);
-	assert(ref_surface);
 	assert(surface);
-	
-	/* TODO: Implement interlaced rendering */
-	if (picture_type != VL_FRAME_PICTURE)
-		return 0;
-	/* TODO: Implement other MC types */
-	if (mc_type != VL_FRAME_MC && mc_type != VL_FIELD_MC)
-		return 0;
-	
-	vlGrabBlocks(surface->context, coded_block_pattern, dct_type, VL_DIFFERENCE_SAMPLE, blocks);
-	
-	pipe = surface->context->pipe;
-	
-	vs_consts = pipe->winsys->buffer_map
-	(
-		pipe->winsys,
-		surface->context->states.mc.vs_const_buf.buffer,
-		PIPE_BUFFER_USAGE_CPU_WRITE
-	);
-	
-	vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->width;
-	vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->height;
-	vs_consts->scale.z = 1.0f;
-	vs_consts->scale.w = 1.0f;
-	vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->width;
-	vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->height;
-	vs_consts->mb_pos_trans.z = 0.0f;
-	vs_consts->mb_pos_trans.w = 0.0f;
-	vs_consts->mb_tc_trans[0].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + motion_vector[0].top_field.x * 0.5f) / (float)surface->width;
-	vs_consts->mb_tc_trans[0].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + motion_vector[0].top_field.y * 0.5f) / (float)surface->height;
-	vs_consts->mb_tc_trans[0].top_field.z = 0.0f;
-	vs_consts->mb_tc_trans[0].top_field.w = 0.0f;
-	vs_consts->mb_tc_trans[1].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + motion_vector[1].top_field.x * 0.5f) / (float)surface->width;
-	vs_consts->mb_tc_trans[1].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + motion_vector[1].top_field.y * 0.5f) / (float)surface->height;
-	vs_consts->mb_tc_trans[1].top_field.z = 0.0f;
-	vs_consts->mb_tc_trans[1].top_field.w = 0.0f;
-	
-	if (mc_type == VL_FIELD_MC)
-	{
-		vs_consts->denorm.x = (float)surface->width;
-		vs_consts->denorm.y = (float)surface->height;
-		
-		vs_consts->mb_tc_trans[0].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + motion_vector[0].bottom_field.x * 0.5f) / (float)surface->width;
-		vs_consts->mb_tc_trans[0].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + motion_vector[0].bottom_field.y * 0.5f) / (float)surface->height;
-		vs_consts->mb_tc_trans[0].bottom_field.z = 0.0f;
-		vs_consts->mb_tc_trans[0].bottom_field.w = 0.0f;
-		vs_consts->mb_tc_trans[1].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + motion_vector[1].bottom_field.x * 0.5f) / (float)surface->width;
-		vs_consts->mb_tc_trans[1].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + motion_vector[1].bottom_field.y * 0.5f) / (float)surface->height;
-		vs_consts->mb_tc_trans[1].bottom_field.z = 0.0f;
-		vs_consts->mb_tc_trans[1].bottom_field.w = 0.0f;
-		
-		pipe->bind_vs_state(pipe, surface->context->states.mc.b_vs[1]);
-		pipe->bind_fs_state(pipe, surface->context->states.mc.b_fs[1]);
-	}
-	else
-	{
-		pipe->bind_vs_state(pipe, surface->context->states.mc.b_vs[0]);
-		pipe->bind_fs_state(pipe, surface->context->states.mc.b_fs[0]);
-	}
-	
-	pipe->winsys->buffer_unmap(pipe->winsys, surface->context->states.mc.vs_const_buf.buffer);
-	
-	surface->context->states.mc.render_target.cbufs[0] = pipe->screen->get_tex_surface
-	(
-		pipe->screen,
-		surface->texture,
-		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
-	);
-	pipe->set_framebuffer_state(pipe, &surface->context->states.mc.render_target);
-	
-	surface->context->states.mc.textures[3] = past_surface->texture;
-	surface->context->states.mc.textures[4] = future_surface->texture;
-	pipe->set_sampler_textures(pipe, 5, surface->context->states.mc.textures);
-	pipe->bind_sampler_states(pipe, 5, (void**)surface->context->states.mc.samplers);
-	
-	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
-	
-	return 0;
+
+	return surface->screen;
 }
 
-int vlPutSurface
+struct vlContext* vlBindToContext
 (
-	struct VL_SURFACE *surface,
-	Drawable drawable,
-	unsigned int srcx,
-	unsigned int srcy,
-	unsigned int srcw,
-	unsigned int srch,
-	unsigned int destx,
-	unsigned int desty,
-	unsigned int destw,
-	unsigned int desth,
-	enum VL_PICTURE picture_type
+	struct vlSurface *surface,
+	struct vlContext *context
 )
 {
-	unsigned int		create_fb = 0;
-	struct pipe_context	*pipe;
-	struct VL_CSC_VS_CONSTS	*vs_consts;
-	
+	struct vlContext *old;
+
 	assert(surface);
-	
-	pipe = surface->context->pipe;
-	
-	if (!surface->context->states.csc.framebuffer.cbufs[0])
-		create_fb = 1;
-	else if
-	(
-		surface->context->states.csc.framebuffer.width != destw ||
-		surface->context->states.csc.framebuffer.height != desth
-	)
-	{
-		pipe->winsys->surface_release
-		(
-			pipe->winsys,
-			&surface->context->states.csc.framebuffer.cbufs[0]
-		);
-		
-		create_fb = 1;
-	}
-	
-	if (create_fb)
-	{
-		surface->context->states.csc.viewport.scale[0] = destw;
-		surface->context->states.csc.viewport.scale[1] = desth;
-		surface->context->states.csc.viewport.scale[2] = 1;
-		surface->context->states.csc.viewport.scale[3] = 1;
-		surface->context->states.csc.viewport.translate[0] = 0;
-		surface->context->states.csc.viewport.translate[1] = 0;
-		surface->context->states.csc.viewport.translate[2] = 0;
-		surface->context->states.csc.viewport.translate[3] = 0;
-		
-		surface->context->states.csc.framebuffer.width = destw;
-		surface->context->states.csc.framebuffer.height = desth;
-		surface->context->states.csc.framebuffer.cbufs[0] = pipe->winsys->surface_alloc(pipe->winsys);
-		pipe->winsys->surface_alloc_storage
-		(
-			pipe->winsys,
-			surface->context->states.csc.framebuffer.cbufs[0],
-			destw,
-			desth,
-			PIPE_FORMAT_A8R8G8B8_UNORM,
-			/* XXX: SoftPipe doesn't change GPU usage to CPU like it does for textures */
-			PIPE_BUFFER_USAGE_CPU_READ | PIPE_BUFFER_USAGE_CPU_WRITE,
-			0
-		);
-	}
-	
-	vlEndRender(surface->context);
-	
-	vs_consts = pipe->winsys->buffer_map
-	(
-		pipe->winsys,
-		surface->context->states.csc.vs_const_buf.buffer,
-		PIPE_BUFFER_USAGE_CPU_WRITE
-	);
-	
-	vs_consts->src_scale.x = srcw / (float)surface->width;
-	vs_consts->src_scale.y = srch / (float)surface->height;
-	vs_consts->src_scale.z = 1;
-	vs_consts->src_scale.w = 1;
-	vs_consts->src_trans.x = srcx / (float)surface->width;
-	vs_consts->src_trans.y = srcy / (float)surface->height;
-	vs_consts->src_trans.z = 0;
-	vs_consts->src_trans.w = 0;
-	
-	pipe->winsys->buffer_unmap(pipe->winsys, surface->context->states.csc.vs_const_buf.buffer);
-	
-	pipe->set_sampler_textures(pipe, 1, &surface->texture);
-	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLE_STRIP, 0, 4);
-	pipe->flush(pipe, PIPE_FLUSH_RENDER_CACHE, NULL);
-	bind_pipe_drawable(pipe, drawable);
-	/* TODO: Need to take destx, desty into consideration */
-	pipe->winsys->flush_frontbuffer
-	(
-		pipe->winsys,
-		surface->context->states.csc.framebuffer.cbufs[0],
-		pipe->priv
-	);
-	
-	vlBeginRender(surface->context);
-	
-	return 0;
-}
 
+	old = surface->context;
+	surface->context = context;
+
+	return old;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.h b/src/gallium/state_trackers/g3dvl/vl_surface.h
index 9f56b77e1e..b975e131fa 100644
--- a/src/gallium/state_trackers/g3dvl/vl_surface.h
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.h
@@ -1,81 +1,66 @@
 #ifndef vl_surface_h
 #define vl_surface_h
 
-#include <X11/Xlib.h>
 #include "vl_types.h"
 
+#ifdef VL_INTERNAL
 struct pipe_texture;
 
-struct VL_SURFACE
+struct vlSurface
 {
-	struct VL_CONTEXT	*context;
+	struct vlScreen		*screen;
+	struct vlContext	*context;
 	unsigned int		width;
 	unsigned int		height;
-	enum VL_FORMAT		format;
+	enum vlFormat		format;
 	struct pipe_texture	*texture;
 };
+#endif
 
-int vlCreateSurface(struct VL_CONTEXT *context, struct VL_SURFACE **surface);
+int vlCreateSurface
+(
+	struct vlScreen *screen,
+	unsigned int width,
+	unsigned int height,
+	enum vlFormat format,
+	struct vlSurface **surface
+);
 
-int vlDestroySurface(struct VL_SURFACE *surface);
+int vlDestroySurface
+(
+	struct vlSurface *surface
+);
 
-int vlRenderIMacroBlock
+int vlRenderMacroBlocksMpeg2
 (
-	enum VL_PICTURE picture_type,
-	enum VL_FIELD_ORDER field_order,
-	unsigned int mbx,
-	unsigned int mby,
-	unsigned int coded_block_pattern,
-	enum VL_DCT_TYPE dct_type,
-	short *blocks,
-	struct VL_SURFACE *surface
+	struct vlMpeg2MacroBlockBatch *batch,
+	struct vlSurface *surface
 );
 
-int vlRenderPMacroBlock
+int vlPutPicture
 (
-	enum VL_PICTURE picture_type,
-	enum VL_FIELD_ORDER field_order,
-	unsigned int mbx,
-	unsigned int mby,
-	enum VL_MC_TYPE mc_type,
-	struct VL_MOTION_VECTOR *motion_vector,
-	unsigned int coded_block_pattern,
-	enum VL_DCT_TYPE dct_type,
-	short *blocks,
-	struct VL_SURFACE *ref_surface,
-	struct VL_SURFACE *surface
+	struct vlSurface *surface,
+	vlNativeDrawable drawable,
+	int srcx,
+	int srcy,
+	int srcw,
+	int srch,
+	int destx,
+	int desty,
+	int destw,
+	int desth,
+	enum vlPictureType picture_type
 );
 
-int vlRenderBMacroBlock
+struct vlScreen* vlSurfaceGetScreen
 (
-	enum VL_PICTURE picture_type,
-	enum VL_FIELD_ORDER field_order,
-	unsigned int mbx,
-	unsigned int mby,
-	enum VL_MC_TYPE mc_type,
-	struct VL_MOTION_VECTOR *motion_vector,
-	unsigned int coded_block_pattern,
-	enum VL_DCT_TYPE dct_type,
-	short *blocks,
-	struct VL_SURFACE *past_surface,
-	struct VL_SURFACE *future_surface,
-	struct VL_SURFACE *surface
+	struct vlSurface *surface
 );
 
-int vlPutSurface
+struct vlContext* vlBindToContext
 (
-	struct VL_SURFACE *surface,
-	Drawable drawable,
-	unsigned int srcx,
-	unsigned int srcy,
-	unsigned int srcw,
-	unsigned int srch,
-	unsigned int destx,
-	unsigned int desty,
-	unsigned int destw,
-	unsigned int desth,
-	enum VL_PICTURE picture_type
+	struct vlSurface *surface,
+	struct vlContext *context
 );
 
 #endif
-
diff --git a/src/gallium/state_trackers/g3dvl/vl_types.h b/src/gallium/state_trackers/g3dvl/vl_types.h
index 4d210c9e0a..504ba8ac81 100644
--- a/src/gallium/state_trackers/g3dvl/vl_types.h
+++ b/src/gallium/state_trackers/g3dvl/vl_types.h
@@ -1,102 +1,106 @@
 #ifndef vl_types_h
 #define vl_types_h
 
-enum VL_FORMAT
-{
-	VL_FORMAT_YCBCR_420,
-	VL_FORMAT_YCBCR_422,
-	VL_FORMAT_YCBCR_444
-};
+#if 1 /*#ifdef X11*/
+#include <X11/Xlib.h>
 
-enum VL_PICTURE
-{
-	VL_TOP_FIELD,
-	VL_BOTTOM_FIELD,
-	VL_FRAME_PICTURE
-};
+typedef Display* vlNativeDisplay;
+typedef Drawable vlNativeDrawable;
+#endif
+
+struct vlDisplay;
+struct vlScreen;
+struct vlContext;
+struct vlSurface;
 
-enum VL_FIELD_ORDER
+enum vlProfile
 {
-	VL_FIELD_FIRST,
-	VL_FIELD_SECOND
+	vlProfileMpeg2Simple,
+	vlProfileMpeg2Main,
+
+	vlProfileCount
 };
 
-enum VL_DCT_TYPE
+enum vlEntryPoint
 {
-	VL_DCT_FIELD_CODED,
-	VL_DCT_FRAME_CODED
+	vlEntryPointIDCT,
+	vlEntryPointMC,
+	vlEntryPointCSC,
+
+	vlEntryPointCount
 };
 
-enum VL_SAMPLE_TYPE
+enum vlFormat
 {
-	VL_FULL_SAMPLE,
-	VL_DIFFERENCE_SAMPLE
+	vlFormatYCbCr420,
+	vlFormatYCbCr422,
+	vlFormatYCbCr444
 };
 
-enum VL_MC_TYPE
+enum vlPictureType
 {
-	VL_FIELD_MC,
-	VL_FRAME_MC,
-	VL_DUAL_PRIME_MC,
-	VL_16x8_MC = VL_FRAME_MC
+	vlPictureTypeTopField,
+	vlPictureTypeBottomField,
+	vlPictureTypeFrame
 };
 
-struct VL_VERTEX4F
+enum vlMotionType
 {
-	float x, y, z, w;
+	vlMotionTypeField,
+	vlMotionTypeFrame,
+	vlMotionTypeDualPrime,
+	vlMotionType16x8
 };
 
-struct VL_VERTEX2F
+enum vlFieldOrder
 {
-	float x, y;
+	vlFieldOrderFirst,
+	vlFieldOrderSecond
 };
 
-struct VL_TEXCOORD2F
+enum vlDCTType
 {
-	float s, t;
+	vlDCTTypeFrameCoded,
+	vlDCTTypeFieldCoded
 };
 
-struct VL_MC_VS_CONSTS
+struct vlVertex2f
 {
-	struct VL_VERTEX4F	scale;
-	struct VL_VERTEX4F	mb_pos_trans;
-	struct VL_VERTEX4F	denorm;
-	struct
-	{
-		struct VL_VERTEX4F	top_field;
-		struct VL_VERTEX4F	bottom_field;
-	} mb_tc_trans[2];
+	float x, y;
 };
 
-struct VL_MC_FS_CONSTS
+struct vlVertex4f
 {
-	struct VL_VERTEX4F	multiplier;
-	struct VL_VERTEX4F	bias;
-	struct VL_VERTEX4F	y_divider;
+	float x, y, z, w;
 };
 
-struct VL_CSC_VS_CONSTS
+enum vlMacroBlockType
 {
-	struct VL_VERTEX4F	src_scale;
-	struct VL_VERTEX4F	src_trans;
+	vlMacroBlockTypeIntra,
+	vlMacroBlockTypeFwdPredicted,
+	vlMacroBlockTypeBkwdPredicted,
+	vlMacroBlockTypeBiPredicted
 };
 
-struct VL_CSC_FS_CONSTS
+struct vlMpeg2MacroBlock
 {
-	struct VL_VERTEX4F	bias;
-	float			matrix[16];
+	unsigned int		mbx, mby;
+	enum vlMacroBlockType	mb_type;
+	enum vlMotionType	mo_type;
+	enum vlDCTType		dct_type;
+	int			PMV[2][2][2];
+	unsigned int		cbp;
+	short			*blocks;
 };
 
-struct VL_MOTION_VECTOR
+struct vlMpeg2MacroBlockBatch
 {
-	struct
-	{
-		int x, y;
-	} top_field, bottom_field;
+	struct vlSurface		*past_surface;
+	struct vlSurface		*future_surface;
+	enum vlPictureType		picture_type;
+	enum vlFieldOrder		field_order;
+	unsigned int			num_macroblocks;
+	struct vlMpeg2MacroBlock	*macroblocks;
 };
 
-struct VL_CONTEXT;
-struct VL_SURFACE;
-
 #endif
-
diff --git a/src/gallium/state_trackers/g3dvl/vl_util.c b/src/gallium/state_trackers/g3dvl/vl_util.c
index 2421ae2210..50aa9af66f 100644
--- a/src/gallium/state_trackers/g3dvl/vl_util.c
+++ b/src/gallium/state_trackers/g3dvl/vl_util.c
@@ -4,14 +4,13 @@
 unsigned int vlRoundUpPOT(unsigned int x)
 {
 	unsigned int i;
-	
+
 	assert(x > 0);
-	
+
 	--x;
-	
+
 	for (i = 1; i < sizeof(unsigned int) * 8; i <<= 1)
 		x |= x >> i;
-	
+
 	return x + 1;
 }
-
diff --git a/src/gallium/state_trackers/g3dvl/vl_util.h b/src/gallium/state_trackers/g3dvl/vl_util.h
index e4b72c4f87..bc98e79df4 100644
--- a/src/gallium/state_trackers/g3dvl/vl_util.h
+++ b/src/gallium/state_trackers/g3dvl/vl_util.h
@@ -4,4 +4,3 @@
 unsigned int vlRoundUpPOT(unsigned int x);
 
 #endif
-
diff --git a/src/libXvMC/block.c b/src/libXvMC/block.c
index deca305bdc..328b035576 100644
--- a/src/libXvMC/block.c
+++ b/src/libXvMC/block.c
@@ -2,83 +2,78 @@
 #include <stdlib.h>
 #include <X11/Xlib.h>
 #include <X11/extensions/XvMC.h>
+#include <vl_display.h>
+#include <vl_screen.h>
 #include <vl_context.h>
 
-/*
- * XvMC defines 64 element blocks (8x8 elements).
- * Elements are 8 bits when they represent color values,
- * 9 bits when they reprecent DCT coefficients, we
- * store them in 2 bytes in either case.
- */
 #define BLOCK_SIZE (64 * 2)
 
 Status XvMCCreateBlocks(Display *display, XvMCContext *context, unsigned int num_blocks, XvMCBlockArray *blocks)
 {
-	struct vl_context *vl_ctx;
-	
+	struct vlContext *vl_ctx;
+
 	assert(display);
-	
+
 	if (!context)
 		return XvMCBadContext;
 	if (num_blocks == 0)
 		return BadValue;
-	
+
 	assert(blocks);
-	
+
 	vl_ctx = context->privData;
-	assert(display == vl_ctx->display);
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
 
 	blocks->context_id = context->context_id;
 	blocks->num_blocks = num_blocks;
 	blocks->blocks = malloc(BLOCK_SIZE * num_blocks);
 	/* Since we don't have a VL type for blocks, set privData to the display so we can catch mismatches */
 	blocks->privData = display;
-	
+
 	return Success;
 }
 
 Status XvMCDestroyBlocks(Display *display, XvMCBlockArray *blocks)
-{	
+{
 	assert(display);
 	assert(blocks);
 	assert(display == blocks->privData);
 	free(blocks->blocks);
-	
+
 	return Success;
 }
 
 Status XvMCCreateMacroBlocks(Display *display, XvMCContext *context, unsigned int num_blocks, XvMCMacroBlockArray *blocks)
 {
-	struct vl_context *vl_ctx;
-	
+	struct vlContext *vl_ctx;
+
 	assert(display);
-	
+
 	if (!context)
 		return XvMCBadContext;
 	if (num_blocks == 0)
 		return BadValue;
-	
+
 	assert(blocks);
-	
+
 	vl_ctx = context->privData;
-	assert(display == vl_ctx->display);
-	
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
+
 	blocks->context_id = context->context_id;
 	blocks->num_blocks = num_blocks;
 	blocks->macro_blocks = malloc(sizeof(XvMCMacroBlock) * num_blocks);
 	/* Since we don't have a VL type for blocks, set privData to the display so we can catch mismatches */
 	blocks->privData = display;
-	
+
 	return Success;
 }
 
 Status XvMCDestroyMacroBlocks(Display *display, XvMCMacroBlockArray *blocks)
-{	
+{
 	assert(display);
 	assert(blocks);
 	assert(display == blocks->privData);
 	free(blocks->macro_blocks);
-	
+
 	return Success;
 }
-
diff --git a/src/libXvMC/context.c b/src/libXvMC/context.c
index 9cf654d6bb..760e012d1a 100644
--- a/src/libXvMC/context.c
+++ b/src/libXvMC/context.c
@@ -1,10 +1,23 @@
 #include <assert.h>
 #include <X11/Xlib.h>
 #include <X11/extensions/XvMClib.h>
+#include <pipe/p_context.h>
+#include <vl_display.h>
+#include <vl_screen.h>
 #include <vl_context.h>
 #include <vl_winsys.h>
 
-static Status Validate(Display *display, XvPortID port, int surface_type_id, unsigned int width, unsigned int height, int flags, int *chroma_format)
+static Status Validate
+(
+	Display *display,
+	XvPortID port,
+	int surface_type_id,
+	unsigned int width,
+	unsigned int height,
+	int flags,
+	int *chroma_format,
+	int *mc_type
+)
 {
 	unsigned int	found_port = 0;
 	unsigned int	found_surface = 0;
@@ -14,13 +27,13 @@ static Status Validate(Display *display, XvPortID port, int surface_type_id, uns
 	unsigned int	max_width, max_height;
 	Status		ret;
 	unsigned int	i, j, k;
-	
+
 	assert(display && chroma_format);
-	
+
 	ret = XvQueryAdaptors(display, XDefaultRootWindow(display), &num_adaptors, &adaptor_info);
 	if (ret != Success)
 		return ret;
-	
+
 	/* Scan through all adaptors looking for this port and surface */
 	for (i = 0; i < num_adaptors && !found_port; ++i)
 	{
@@ -31,10 +44,10 @@ static Status Validate(Display *display, XvPortID port, int surface_type_id, uns
 			if (adaptor_info[i].base_id + j == port)
 			{
 				XvMCSurfaceInfo *surface_info;
-				
+
 				found_port = 1;
 				surface_info = XvMCListSurfaceTypes(display, adaptor_info[i].base_id, &num_types);
-				
+
 				if (surface_info)
 				{
 					for (k = 0; k < num_types && !found_surface; ++k)
@@ -45,9 +58,10 @@ static Status Validate(Display *display, XvPortID port, int surface_type_id, uns
 							max_width = surface_info[k].max_width;
 							max_height = surface_info[k].max_height;
 							*chroma_format = surface_info[k].chroma_format;
+							*mc_type = surface_info[k].mc_type;
 						}
 					}
-					
+
 					XFree(surface_info);
 				}
 				else
@@ -58,9 +72,9 @@ static Status Validate(Display *display, XvPortID port, int surface_type_id, uns
 			}
 		}
 	}
-	
+
 	XvFreeAdaptorInfo(adaptor_info);
-	
+
 	if (!found_port)
 		return XvBadPort;
 	if (!found_surface)
@@ -69,60 +83,86 @@ static Status Validate(Display *display, XvPortID port, int surface_type_id, uns
 		return BadValue;
 	if (flags != XVMC_DIRECT && flags != 0)
 		return BadValue;
-	
+
 	return Success;
 }
 
-static enum VL_FORMAT FormatToVL(int xvmc_format)
+static enum vlProfile ProfileToVL(int xvmc_profile)
+{
+	if (xvmc_profile & XVMC_MPEG_1)
+		assert(0);
+	else if (xvmc_profile & XVMC_MPEG_2)
+		return vlProfileMpeg2Main;
+	else if (xvmc_profile & XVMC_H263)
+		assert(0);
+	else if (xvmc_profile & XVMC_MPEG_4)
+		assert(0);
+	else
+		assert(0);
+
+	return -1;
+}
+
+static enum vlEntryPoint EntryToVL(int xvmc_entry)
+{
+	return xvmc_entry & XVMC_IDCT ? vlEntryPointIDCT : vlEntryPointMC;
+}
+
+static enum vlFormat FormatToVL(int xvmc_format)
 {
-	enum VL_FORMAT vl_format;
-	
 	switch (xvmc_format)
 	{
 		case XVMC_CHROMA_FORMAT_420:
-		{
-			vl_format = VL_FORMAT_YCBCR_420;
-			break;
-		}
+			return vlFormatYCbCr420;
 		case XVMC_CHROMA_FORMAT_422:
-		{
-			vl_format = VL_FORMAT_YCBCR_422;
-			break;
-		}
+			return vlFormatYCbCr422;
 		case XVMC_CHROMA_FORMAT_444:
-		{
-			vl_format = VL_FORMAT_YCBCR_444;
-			break;
-		}
+			return vlFormatYCbCr444;
 		default:
 			assert(0);
 	}
-	
-	return vl_format;
+
+	return -1;
 }
 
 Status XvMCCreateContext(Display *display, XvPortID port, int surface_type_id, int width, int height, int flags, XvMCContext *context)
 {
 	int			chroma_format;
+	int			mc_type;
 	Status			ret;
-	struct VL_CONTEXT	*vl_ctx;
+	struct vlDisplay	*vl_dpy;
+	struct vlScreen		*vl_scrn;
+	struct vlContext	*vl_ctx;
 	struct pipe_context	*pipe;
-	
+
 	assert(display);
-	
+
 	if (!context)
 		return XvMCBadContext;
-	
-	ret = Validate(display, port, surface_type_id, width, height, flags, &chroma_format);
+
+	ret = Validate(display, port, surface_type_id, width, height, flags, &chroma_format, &mc_type);
 	if (ret != Success)
 		return ret;
-	
+
+	/* XXX: Assumes default screen, should check which screen port is on */
 	pipe = create_pipe_context(display, XDefaultScreen(display));
-	
+
 	assert(pipe);
-	
-	vlCreateContext(display, pipe, width, height, FormatToVL(chroma_format), &vl_ctx);
-	
+
+	vlCreateDisplay(display, &vl_dpy);
+	vlCreateScreen(vl_dpy, XDefaultScreen(display), pipe->screen, &vl_scrn);
+	vlCreateContext
+	(
+		vl_scrn,
+		pipe,
+		width,
+		height,
+		FormatToVL(chroma_format),
+		ProfileToVL(mc_type),
+		EntryToVL(mc_type),
+		&vl_ctx
+	);
+
 	context->context_id = XAllocID(display);
 	context->surface_type_id = surface_type_id;
 	context->width = width;
@@ -130,89 +170,27 @@ Status XvMCCreateContext(Display *display, XvPortID port, int surface_type_id, i
 	context->flags = flags;
 	context->port = port;
 	context->privData = vl_ctx;
-	
+
 	return Success;
 }
 
 Status XvMCDestroyContext(Display *display, XvMCContext *context)
 {
-	struct VL_CONTEXT	*vl_ctx;
+	struct vlContext	*vl_ctx;
 	struct pipe_context	*pipe;
-	
+
 	assert(display);
-	
+
 	if (!context)
 		return XvMCBadContext;
-	
+
 	vl_ctx = context->privData;
-	
-	assert(display == vl_ctx->display);
-	
-	pipe = vl_ctx->pipe;
-	vlDestroyContext(vl_ctx);
-	destroy_pipe_context(pipe);
-	
-	return Success;
-}
 
-/* XXX: The following are here temporarily, need to be implemented in the DDX driver */
-/* TODO: Figure out which of these need to be in DDX, which are better off in DDX, which can stay */
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
 
-Bool XvMCQueryExtension(Display *display, int *event_base, int *err_base)
-{
-	*event_base = 0;
-	*err_base = 0;
-	
-	return True;
-}
+	pipe = vlGetPipeContext(vl_ctx);
+	vlDestroyContext(vl_ctx);
+	destroy_pipe_context(pipe);
 
-Status XvMCQueryVersion(Display *display, int *major, int *minor)
-{
-	*major = 1;
-	*minor = 0;
-	
 	return Success;
 }
-
-XvMCSurfaceInfo* XvMCListSurfaceTypes(Display *display, XvPortID port, int *num)
-{
-	XvMCSurfaceInfo	*surface_info = calloc(1, sizeof(XvMCSurfaceInfo));
-	
-	*num = 1;
-	
-	surface_info->chroma_format = XVMC_CHROMA_FORMAT_420;
-	surface_info->max_width = 2048;
-	surface_info->max_height = 2048;
-	surface_info->subpicture_max_width = 2048;
-	surface_info->subpicture_max_height = 2048;
-	surface_info->mc_type = XVMC_IDCT | XVMC_MPEG_2;
-	surface_info->surface_type_id = 123; /* FIXME: XAllocID(display)*/;
-	surface_info->flags = XVMC_INTRA_UNSIGNED | XVMC_SUBPICTURE_INDEPENDENT_SCALING | XVMC_BACKEND_SUBPICTURE;
-	
-	return surface_info;
-}
-
-XvImageFormatValues* XvMCListSubpictureTypes(Display* display, XvPortID port, int surface_type_id, int *count_return)
-{
-	XvImageFormatValues *image_formats = calloc(1, sizeof(XvImageFormatValues));
-	
-	*count_return = 1;
-	
-	image_formats[0].id = 123;
-	image_formats[0].type = XvRGB;
-	image_formats[0].byte_order = LSBFirst;
-	image_formats[0].bits_per_pixel = 8;
-	image_formats[0].format = XvPacked;
-	image_formats[0].num_planes = 1;
-	image_formats[0].depth = 8;
-	image_formats[0].red_mask = 0x0000FF;
-	image_formats[0].green_mask = 0x00FF00;
-	image_formats[0].blue_mask = 0xFF0000;
-	image_formats[0].component_order[0] = 'R';
-	image_formats[0].component_order[0] = 'G';
-	image_formats[0].component_order[0] = 'B';
-	image_formats[0].scanline_order = XvTopToBottom;
-	
-	return image_formats;
-}
-
diff --git a/src/libXvMC/surface.c b/src/libXvMC/surface.c
index 1c07220e84..038befc297 100644
--- a/src/libXvMC/surface.c
+++ b/src/libXvMC/surface.c
@@ -1,90 +1,96 @@
 #include <assert.h>
 #include <X11/Xlib.h>
 #include <X11/extensions/XvMC.h>
+#include <vl_display.h>
+#include <vl_screen.h>
 #include <vl_context.h>
 #include <vl_surface.h>
+#include <vl_types.h>
 
-static enum VL_PICTURE PictureToVL(int xvmc_pic)
+static enum vlMacroBlockType TypeToVL(int xvmc_mb_type)
+{
+	if (xvmc_mb_type & XVMC_MB_TYPE_INTRA)
+		return vlMacroBlockTypeIntra;
+	if ((xvmc_mb_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) == XVMC_MB_TYPE_MOTION_FORWARD)
+		return vlMacroBlockTypeFwdPredicted;
+	if ((xvmc_mb_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) == XVMC_MB_TYPE_MOTION_BACKWARD)
+		return vlMacroBlockTypeBkwdPredicted;
+	if ((xvmc_mb_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)) == (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD))
+		return vlMacroBlockTypeBiPredicted;
+
+	assert(0);
+
+	return -1;
+}
+
+static enum vlPictureType PictureToVL(int xvmc_pic)
 {
-	enum VL_PICTURE vl_pic;
-	
 	switch (xvmc_pic)
 	{
 		case XVMC_TOP_FIELD:
-		{
-			vl_pic = VL_TOP_FIELD;
-			break;
-		}
+			return vlPictureTypeTopField;
 		case XVMC_BOTTOM_FIELD:
-		{
-			vl_pic = VL_BOTTOM_FIELD;
-			break;
-		}
+			return vlPictureTypeBottomField;
 		case XVMC_FRAME_PICTURE:
-		{
-			vl_pic = VL_FRAME_PICTURE;
-			break;
-		}
+			return vlPictureTypeFrame;
 		default:
 			assert(0);
 	}
-	
-	return vl_pic;
+
+	return -1;
 }
 
-static enum VL_MC_TYPE MotionToVL(int xvmc_motion_type)
+static enum vlMotionType MotionToVL(int xvmc_motion_type)
 {
-	enum VL_MC_TYPE vl_mc_type;
-	
 	switch (xvmc_motion_type)
 	{
 		case XVMC_PREDICTION_FRAME:
-		{
-			vl_mc_type = VL_FRAME_MC;
-			break;
-		}
+			return vlMotionTypeFrame;
 		case XVMC_PREDICTION_FIELD:
-		{
-			vl_mc_type = VL_FIELD_MC;
-			break;
-		}
+			return vlMotionTypeField;
 		case XVMC_PREDICTION_DUAL_PRIME:
-		{
-			vl_mc_type = VL_DUAL_PRIME_MC;
-			break;
-		}
+			return vlMotionTypeDualPrime;
 		default:
 			assert(0);
 	}
-	
-	return vl_mc_type;
+
+	return -1;
 }
 
 Status XvMCCreateSurface(Display *display, XvMCContext *context, XvMCSurface *surface)
 {
-	struct VL_CONTEXT *vl_ctx;
-	struct VL_SURFACE *vl_sfc;
-	
+	struct vlContext *vl_ctx;
+	struct vlSurface *vl_sfc;
+
 	assert(display);
-	
+
 	if (!context)
 		return XvMCBadContext;
 	if (!surface)
 		return XvMCBadSurface;
-	
+
 	vl_ctx = context->privData;
-	
-	assert(display == vl_ctx->display);
-	
-	vlCreateSurface(vl_ctx, &vl_sfc);
-	
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
+
+	vlCreateSurface
+	(
+		vlContextGetScreen(vl_ctx),
+		context->width,
+		context->height,
+		vlGetPictureFormat(vl_ctx),
+		&vl_sfc
+	);
+
+	vlBindToContext(vl_sfc, vl_ctx);
+
 	surface->surface_id = XAllocID(display);
 	surface->context_id = context->context_id;
 	surface->surface_type_id = context->surface_type_id;
 	surface->width = context->width;
 	surface->height = context->height;
 	surface->privData = vl_sfc;
-	
+
 	return Success;
 }
 
@@ -103,19 +109,21 @@ Status XvMCRenderSurface
 	XvMCBlockArray *blocks
 )
 {
-	struct VL_CONTEXT	*vl_ctx;
-	struct VL_SURFACE	*target_vl_surface;
-	struct VL_SURFACE	*past_vl_surface;
-	struct VL_SURFACE	*future_vl_surface;
-	unsigned int		i;
-	
+	struct vlContext		*vl_ctx;
+	struct vlSurface		*target_vl_surface;
+	struct vlSurface		*past_vl_surface;
+	struct vlSurface		*future_vl_surface;
+	struct vlMpeg2MacroBlockBatch	batch;
+	struct vlMpeg2MacroBlock	vl_macroblocks[num_macroblocks];
+	unsigned int			i;
+
 	assert(display);
-	
+
 	if (!context)
 		return XvMCBadContext;
 	if (!target_surface)
 		return XvMCBadSurface;
-	
+
 	if
 	(
 		picture_structure != XVMC_TOP_FIELD &&
@@ -125,178 +133,94 @@ Status XvMCRenderSurface
 		return BadValue;
 	if (future_surface && !past_surface)
 		return BadMatch;
-	
+
 	vl_ctx = context->privData;
-	
-	assert(display == vl_ctx->display);
-	
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlContextGetScreen(vl_ctx))));
+
 	target_vl_surface = target_surface->privData;
 	past_vl_surface = past_surface ? past_surface->privData : NULL;
 	future_vl_surface = future_surface ? future_surface->privData : NULL;
-	
-	assert(vl_ctx == target_vl_surface->context);
-	assert(!past_vl_surface || vl_ctx == past_vl_surface->context);
-	assert(!future_vl_surface || vl_ctx == future_vl_surface->context);
-	
+
+	assert(context->context_id == target_surface->context_id);
+	assert(!past_surface || context->context_id == past_surface->context_id);
+	assert(!future_surface || context->context_id == future_surface->context_id);
+
 	assert(macroblocks);
 	assert(blocks);
-	
+
 	assert(macroblocks->context_id == context->context_id);
 	assert(blocks->context_id == context->context_id);
-	
+
 	assert(flags == 0 || flags == XVMC_SECOND_FIELD);
-	
-	/* TODO: Batch macroblocks by type (I,P,B) */
-	
-	for (i = first_macroblock; i < first_macroblock + num_macroblocks; ++i)
-		if (macroblocks->macro_blocks[i].macroblock_type & XVMC_MB_TYPE_INTRA)
-			vlRenderIMacroBlock
-			(
-				PictureToVL(picture_structure),
-				flags == XVMC_SECOND_FIELD ? VL_FIELD_SECOND : VL_FIELD_FIRST,
-				macroblocks->macro_blocks[i].x,
-				macroblocks->macro_blocks[i].y,
-				macroblocks->macro_blocks[i].coded_block_pattern,
-				macroblocks->macro_blocks[i].dct_type == XVMC_DCT_TYPE_FIELD ? VL_DCT_FIELD_CODED : VL_DCT_FRAME_CODED,
-				blocks->blocks + (macroblocks->macro_blocks[i].index * 64),
-				target_vl_surface
-			);
-		else if
-		(
-			(macroblocks->macro_blocks[i].macroblock_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD))
-			== XVMC_MB_TYPE_MOTION_FORWARD
-		)
-		{
-			struct VL_MOTION_VECTOR motion_vector =
-			{
-				{
-					macroblocks->macro_blocks[i].PMV[0][0][0],
-					macroblocks->macro_blocks[i].PMV[0][0][1],
-				},
-				{
-					macroblocks->macro_blocks[i].PMV[1][0][0],
-					macroblocks->macro_blocks[i].PMV[1][0][1],
-				}
-			};
-						
-			vlRenderPMacroBlock
-			(
-				PictureToVL(picture_structure),
-				flags == XVMC_SECOND_FIELD ? VL_FIELD_SECOND : VL_FIELD_FIRST,
-				macroblocks->macro_blocks[i].x,
-				macroblocks->macro_blocks[i].y,
-				MotionToVL(macroblocks->macro_blocks[i].motion_type),
-				&motion_vector,
-				macroblocks->macro_blocks[i].coded_block_pattern,
-				macroblocks->macro_blocks[i].dct_type == XVMC_DCT_TYPE_FIELD ? VL_DCT_FIELD_CODED : VL_DCT_FRAME_CODED,
-				blocks->blocks + (macroblocks->macro_blocks[i].index * 64),
-				past_vl_surface,
-				target_vl_surface
-			);
-		}
-		else if
-		(
-			(macroblocks->macro_blocks[i].macroblock_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD))
-			== XVMC_MB_TYPE_MOTION_BACKWARD
-		)
-		{
-			struct VL_MOTION_VECTOR motion_vector =
-			{
-				{
-					macroblocks->macro_blocks[i].PMV[0][1][0],
-					macroblocks->macro_blocks[i].PMV[0][1][1],
-				},
-				{
-					macroblocks->macro_blocks[i].PMV[1][1][0],
-					macroblocks->macro_blocks[i].PMV[1][1][1],
-				}
-			};
-			
-			vlRenderPMacroBlock
-			(
-				PictureToVL(picture_structure),
-				flags == XVMC_SECOND_FIELD ? VL_FIELD_SECOND : VL_FIELD_FIRST,
-				macroblocks->macro_blocks[i].x,
-				macroblocks->macro_blocks[i].y,
-				MotionToVL(macroblocks->macro_blocks[i].motion_type),
-				&motion_vector,
-				macroblocks->macro_blocks[i].coded_block_pattern,
-				macroblocks->macro_blocks[i].dct_type == XVMC_DCT_TYPE_FIELD ? VL_DCT_FIELD_CODED : VL_DCT_FRAME_CODED,
-				blocks->blocks + (macroblocks->macro_blocks[i].index * 64),
-				future_vl_surface,
-				target_vl_surface
-			);
-		}
-		else if
-		(
-			(macroblocks->macro_blocks[i].macroblock_type & (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD))
-			== (XVMC_MB_TYPE_MOTION_FORWARD | XVMC_MB_TYPE_MOTION_BACKWARD)
-		)
-		{
-			struct VL_MOTION_VECTOR motion_vector[2] =
-			{
-				{
-					{
-						macroblocks->macro_blocks[i].PMV[0][0][0],
-						macroblocks->macro_blocks[i].PMV[0][0][1],
-					},
-					{
-						macroblocks->macro_blocks[i].PMV[1][0][0],
-						macroblocks->macro_blocks[i].PMV[1][0][1],
-					}
-				},
-				{
-					{
-						macroblocks->macro_blocks[i].PMV[0][1][0],
-						macroblocks->macro_blocks[i].PMV[0][1][1],
-					},
-					{
-						macroblocks->macro_blocks[i].PMV[1][1][0],
-						macroblocks->macro_blocks[i].PMV[1][1][1],
-					}
-				}
-			};
-			
-			vlRenderBMacroBlock
-			(
-				PictureToVL(picture_structure),
-				flags == XVMC_SECOND_FIELD ? VL_FIELD_SECOND : VL_FIELD_FIRST,
-				macroblocks->macro_blocks[i].x,
-				macroblocks->macro_blocks[i].y,
-				MotionToVL(macroblocks->macro_blocks[i].motion_type),
-				motion_vector,
-				macroblocks->macro_blocks[i].coded_block_pattern,
-				macroblocks->macro_blocks[i].dct_type == XVMC_DCT_TYPE_FIELD ? VL_DCT_FIELD_CODED : VL_DCT_FRAME_CODED,
-				blocks->blocks + (macroblocks->macro_blocks[i].index * 64),
-				past_vl_surface,
-				future_vl_surface,
-				target_vl_surface
-			);
-		}
-		else
-			fprintf(stderr, "Unrecognized macroblock\n");
-	
+
+	batch.past_surface = past_vl_surface;
+	batch.future_surface = future_vl_surface;
+	batch.picture_type = PictureToVL(picture_structure);
+	batch.field_order = flags & XVMC_SECOND_FIELD ? vlFieldOrderSecond : vlFieldOrderFirst;
+	batch.num_macroblocks = num_macroblocks;
+	batch.macroblocks = vl_macroblocks;
+
+	for (i = 0; i < num_macroblocks; ++i)
+	{
+		unsigned int j = first_macroblock + i;
+
+		unsigned int k, l, m;
+
+		batch.macroblocks[i].mbx = macroblocks->macro_blocks[j].x;
+		batch.macroblocks[i].mby = macroblocks->macro_blocks[j].y;
+		batch.macroblocks[i].mb_type = TypeToVL(macroblocks->macro_blocks[j].macroblock_type);
+		if (batch.macroblocks[i].mb_type != vlMacroBlockTypeIntra)
+			batch.macroblocks[i].mo_type = MotionToVL(macroblocks->macro_blocks[j].motion_type);
+		batch.macroblocks[i].dct_type = macroblocks->macro_blocks[j].dct_type & XVMC_DCT_TYPE_FIELD ? vlDCTTypeFieldCoded : vlDCTTypeFrameCoded;
+
+		for (k = 0; k < 2; ++k)
+			for (l = 0; l < 2; ++l)
+				for (m = 0; m < 2; ++m)
+					batch.macroblocks[i].PMV[k][l][m] = macroblocks->macro_blocks[j].PMV[k][l][m];
+
+		batch.macroblocks[i].cbp = macroblocks->macro_blocks[j].coded_block_pattern;
+		batch.macroblocks[i].blocks = blocks->blocks + (macroblocks->macro_blocks[j].index * 64);
+	}
+
+	vlRenderMacroBlocksMpeg2(&batch, target_vl_surface);
+
 	return Success;
 }
 
 Status XvMCFlushSurface(Display *display, XvMCSurface *surface)
 {
+	struct vlSurface *vl_sfc;
+
 	assert(display);
-	
+
 	if (!surface)
 		return XvMCBadSurface;
-	
-	/* TODO: Check display & surface match */
+
+	vl_sfc = surface->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
+	/* TODO */
+
 	return Success;
 }
 
 Status XvMCSyncSurface(Display *display, XvMCSurface *surface)
 {
+	struct vlSurface *vl_sfc;
+
 	assert(display);
-	
+
 	if (!surface)
 		return XvMCBadSurface;
-	
+
+	vl_sfc = surface->privData;
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
+	/* TODO */
+
 	return Success;
 }
 
@@ -321,92 +245,85 @@ Status XvMCPutSurface
 	unsigned int		width, height;
 	unsigned int		border_width;
 	unsigned int		depth;
-	struct VL_SURFACE	*vl_sfc;
-	
+	struct vlSurface	*vl_sfc;
+
 	assert(display);
-	
+
 	if (!surface)
 		return XvMCBadSurface;
-		
+
 	if (XGetGeometry(display, drawable, &root, &x, &y, &width, &height, &border_width, &depth) == BadDrawable)
 		return BadDrawable;
-	
+
 	assert(flags == XVMC_TOP_FIELD || flags == XVMC_BOTTOM_FIELD || flags == XVMC_FRAME_PICTURE);
-	
+
 	/* TODO: Correct for negative srcx,srcy & destx,desty by clipping */
-	
+
 	assert(srcx + srcw - 1 < surface->width);
 	assert(srcy + srch - 1 < surface->height);
 	assert(destx + destw - 1 < width);
 	assert(desty + desth - 1 < height);
-	
+
 	vl_sfc = surface->privData;
-	
-	vlPutSurface(vl_sfc, drawable, srcx, srcy, srcw, srch, destx, desty, destw, desth, PictureToVL(flags));
-	
+
+	vlPutPicture(vl_sfc, drawable, srcx, srcy, srcw, srch, destx, desty, destw, desth, PictureToVL(flags));
+
 	return Success;
 }
 
 Status XvMCGetSurfaceStatus(Display *display, XvMCSurface *surface, int *status)
 {
-	struct VL_CONTEXT *vl_ctx;
-	struct VL_SURFACE *vl_sfc;
-	
+	struct vlSurface *vl_sfc;
+
 	assert(display);
-	
+
 	if (!surface)
 		return XvMCBadSurface;
-		
+
 	assert(status);
-	
+
 	vl_sfc = surface->privData;
-	vl_ctx = vl_sfc->context;
-	
-	assert(display == vl_ctx->display);
-	
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
 	/* TODO */
 	*status = 0;
-	
+
 	return Success;
 }
 
 Status XvMCDestroySurface(Display *display, XvMCSurface *surface)
 {
-	struct VL_CONTEXT *vl_ctx;
-	struct VL_SURFACE *vl_sfc;
-	
+	struct vlSurface *vl_sfc;
+
 	assert(display);
-	
+
 	if (!surface)
 		return XvMCBadSurface;
-	
+
 	vl_sfc = surface->privData;
-	vl_ctx = vl_sfc->context;
-	
-	assert(display == vl_ctx->display);
-	
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
 	vlDestroySurface(vl_sfc);
-	
+
 	return Success;
 }
 
 Status XvMCHideSurface(Display *display, XvMCSurface *surface)
 {
-	struct VL_CONTEXT *vl_ctx;
-	struct VL_SURFACE *vl_sfc;
-	
+	struct vlSurface *vl_sfc;
+
 	assert(display);
-	
+
 	if (!surface)
 		return XvMCBadSurface;
-	
+
 	vl_sfc = surface->privData;
-	vl_ctx = vl_sfc->context;
-	
-	assert(display == vl_ctx->display);
-	
+
+	assert(display == vlGetNativeDisplay(vlGetDisplay(vlSurfaceGetScreen(vl_sfc))));
+
 	/* No op, only for overlaid rendering */
-	
+
 	return Success;
 }
-
-- 
cgit v1.2.3


From 7f100d04ddacf9f6517c9aff1e2de5257eb77fb0 Mon Sep 17 00:00:00 2001
From: Younes Manton <younes.m@gmail.com>
Date: Mon, 18 Aug 2008 00:04:29 -0400
Subject: g3dvl: Use rotating buffers to avoid waiting for map().

---
 src/driclient/src/Makefile                        |  2 +-
 src/gallium/state_trackers/g3dvl/Makefile         |  4 +-
 src/gallium/state_trackers/g3dvl/vl_data.c        | 76 -----------------------
 src/gallium/state_trackers/g3dvl/vl_data.h        | 27 --------
 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c | 46 +++++++++-----
 src/gallium/winsys/g3dvl/nouveau/Makefile         |  2 +-
 src/libXvMC/Makefile                              |  2 +-
 7 files changed, 37 insertions(+), 122 deletions(-)
 delete mode 100644 src/gallium/state_trackers/g3dvl/vl_data.c
 delete mode 100644 src/gallium/state_trackers/g3dvl/vl_data.h

(limited to 'src/gallium/state_trackers/g3dvl/Makefile')

diff --git a/src/driclient/src/Makefile b/src/driclient/src/Makefile
index 89ffb84477..5d913ad5fa 100644
--- a/src/driclient/src/Makefile
+++ b/src/driclient/src/Makefile
@@ -2,7 +2,7 @@ TARGET		= libdriclient.a
 OBJECTS		= driclient.o XF86dri.o
 DRMDIR		?= /usr
 
-CFLAGS		+= -g -Wall -fPIC -Werror -I../include -I${DRMDIR}/include -I${DRMDIR}/include/drm
+CFLAGS		+= -g -Wall -fPIC -I../include -I${DRMDIR}/include -I${DRMDIR}/include/drm
 
 #############################################
 
diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
index 9995c554ab..bd77c62bc5 100644
--- a/src/gallium/state_trackers/g3dvl/Makefile
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -1,9 +1,9 @@
 TARGET		= libg3dvl.a
-OBJECTS		= vl_display.o vl_screen.o vl_context.o vl_surface.o vl_data.o vl_shader_build.o vl_util.o vl_basic_csc.o \
+OBJECTS		= vl_display.o vl_screen.o vl_context.o vl_surface.o vl_shader_build.o vl_util.o vl_basic_csc.o	\
 		  vl_r16snorm_mc.o
 GALLIUMDIR	= ../..
 
-CFLAGS		+= -g -Wall -fPIC -Werror -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary -I${GALLIUMDIR}/winsys/g3dvl
+CFLAGS		+= -g -Wall -fPIC -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary -I${GALLIUMDIR}/winsys/g3dvl
 
 #############################################
 
diff --git a/src/gallium/state_trackers/g3dvl/vl_data.c b/src/gallium/state_trackers/g3dvl/vl_data.c
deleted file mode 100644
index f2476dbf1e..0000000000
--- a/src/gallium/state_trackers/g3dvl/vl_data.c
+++ /dev/null
@@ -1,76 +0,0 @@
-#include "vl_data.h"
-
-/*
- * Represents 8 triangles (4 quads, 1 per block) in noormalized coords
- * that render a macroblock.
- * Need to be scaled to cover mbW*mbH macroblock pixels and translated into
- * position on target surface.
- */
-const struct vlVertex2f macroblock_verts[24] =
-{
-	{0.0f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.0f},
-	{0.5f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.5f},
-
-	{0.5f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.0f},
-	{1.0f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.5f},
-
-	{0.0f, 0.5f}, {0.0f, 1.0f}, {0.5f, 0.5f},
-	{0.5f, 0.5f}, {0.0f, 1.0f}, {0.5f, 1.0f},
-
-	{0.5f, 0.5f}, {0.5f, 1.0f}, {1.0f, 0.5f},
-	{1.0f, 0.5f}, {0.5f, 1.0f}, {1.0f, 1.0f}
-};
-
-/*
- * Represents texcoords for the above for rendering 4 luma blocks arranged
- * in a bW*(bH*4) texture. First luma block located at 0,0->bW,bH; second at
- * 0,bH->bW,2bH; third at 0,2bH->bW,3bH; fourth at 0,3bH->bW,4bH.
- */
-const struct vlVertex2f macroblock_luma_texcoords[24] =
-{
-	{0.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.0f},
-	{1.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.25f},
-
-	{0.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.25f},
-	{1.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.5f},
-
-	{0.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.5f},
-	{1.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.75f},
-
-	{0.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 0.75f},
-	{1.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 1.0f}
-};
-
-/*
- * Represents texcoords for the above for rendering 1 chroma block.
- * Straight forward 0,0->1,1 mapping so we can reuse the MB pos vectors.
- */
-const struct vlVertex2f *macroblock_chroma_420_texcoords = macroblock_verts;
-
-/*
- * Represents texcoords for the above for rendering 2 chroma blocks arranged
- * in a bW*(bH*2) texture. First chroma block located at 0,0->bW,bH; second at
- * 0,bH->bW,2bH. We can render this with 0,0->1,1 mapping.
- * Straight forward 0,0->1,1 mapping so we can reuse MB pos vectors.
- */
-const struct vlVertex2f *macroblock_chroma_422_texcoords = macroblock_verts;
-
-/*
- * Represents texcoords for the above for rendering 4 chroma blocks.
- * Same case as 4 luma blocks.
- */
-const struct vlVertex2f *macroblock_chroma_444_texcoords = macroblock_luma_texcoords;
-
-/*
- * Used when rendering P and B macroblocks, multiplier is applied to the A channel,
- * which is then added to the L channel, then the bias is subtracted from that to
- * get back the differential. The differential is then added to the samples from the
- * reference surface(s).
- */
-#if 0
-const struct VL_MC_FS_CONSTS vl_mc_fs_consts =
-{
-	{32767.0f / 255.0f, 32767.0f / 255.0f, 32767.0f / 255.0f, 0.0f},
-	{0.5f, 2.0f, 0.0f, 0.0f}
-};
-#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_data.h b/src/gallium/state_trackers/g3dvl/vl_data.h
deleted file mode 100644
index f0de2e976c..0000000000
--- a/src/gallium/state_trackers/g3dvl/vl_data.h
+++ /dev/null
@@ -1,27 +0,0 @@
-#ifndef vl_data_h
-#define vl_data_h
-
-#include "vl_types.h"
-
-/* TODO: Needs to be rolled into the appropriate stage */
-
-extern const struct vlVertex2f macroblock_verts[24];
-extern const struct vlVertex2f macroblock_luma_texcoords[24];
-extern const struct vlVertex2f *macroblock_chroma_420_texcoords;
-extern const struct vlVertex2f *macroblock_chroma_422_texcoords;
-extern const struct vlVertex2f *macroblock_chroma_444_texcoords;
-
-extern const struct vlVertex2f surface_verts[4];
-extern const struct vlVertex2f *surface_texcoords;
-
-/*
-extern const struct VL_MC_FS_CONSTS vl_mc_fs_consts;
-
-extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_identity;
-extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_601;
-extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_601_full;
-extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_709;
-extern const struct VL_CSC_FS_CONSTS vl_csc_fs_consts_709_full;
-*/
-
-#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
index 4fae224431..80b09a6d1d 100644
--- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
@@ -15,6 +15,8 @@
 #include "vl_types.h"
 #include "vl_defs.h"
 
+#define NUM_BUFS 4	/* Number of rotating buffers to use */
+
 struct vlVertexShaderConsts
 {
 	/*struct vlVertex4f scale;
@@ -41,12 +43,13 @@ struct vlR16SnormMC
 
 	unsigned int				video_width, video_height;
 	enum vlFormat				video_format;
+	unsigned int				cur_buf;
 
 	struct pipe_context			*pipe;
 	struct pipe_viewport_state		viewport;
 	struct pipe_framebuffer_state		render_target;
 	struct pipe_sampler_state		*samplers[5];
-	struct pipe_texture			*textures[5];
+	struct pipe_texture			*textures[NUM_BUFS][5];
 	void					*i_vs, *p_vs[2], *b_vs[2];
 	void					*i_fs, *p_fs[2], *b_fs[2];
 	struct pipe_vertex_buffer 		vertex_bufs[3];
@@ -230,7 +233,7 @@ static int vlGrabBlocks
 	tex_surface = mc->pipe->screen->get_tex_surface
 	(
 		mc->pipe->screen,
-		mc->textures[0],
+		mc->textures[mc->cur_buf % NUM_BUFS][0],
 		0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
 	);
 
@@ -276,7 +279,7 @@ static int vlGrabBlocks
 		tex_surface = mc->pipe->screen->get_tex_surface
 		(
 			mc->pipe->screen,
-			mc->textures[tb + 1],
+			mc->textures[mc->cur_buf % NUM_BUFS][tb + 1],
 			0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
 		);
 
@@ -361,12 +364,14 @@ int vlRenderIMacroBlock
 		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
 	);
 	pipe->set_framebuffer_state(pipe, &mc->render_target);
-	pipe->set_sampler_textures(pipe, 3, mc->textures);
+	pipe->set_sampler_textures(pipe, 3, mc->textures[mc->cur_buf % NUM_BUFS]);
 	pipe->bind_sampler_states(pipe, 3, (void**)mc->samplers);
 	pipe->bind_vs_state(pipe, mc->i_vs);
 	pipe->bind_fs_state(pipe, mc->i_fs);
 
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
+	
+	mc->cur_buf++;
 
 	return 0;
 }
@@ -458,11 +463,13 @@ int vlRenderPMacroBlock
 	);
 	pipe->set_framebuffer_state(pipe, &mc->render_target);
 
-	mc->textures[3] = ref_surface->texture;
-	pipe->set_sampler_textures(pipe, 4, mc->textures);
+	mc->textures[mc->cur_buf % NUM_BUFS][3] = ref_surface->texture;
+	pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUFS]);
 	pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers);
 
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
+	
+	mc->cur_buf++;
 
 	return 0;
 }
@@ -567,12 +574,14 @@ int vlRenderBMacroBlock
 	);
 	pipe->set_framebuffer_state(pipe, &mc->render_target);
 
-	mc->textures[3] = past_surface->texture;
-	mc->textures[4] = future_surface->texture;
-	pipe->set_sampler_textures(pipe, 5, mc->textures);
+	mc->textures[mc->cur_buf % NUM_BUFS][3] = past_surface->texture;
+	mc->textures[mc->cur_buf % NUM_BUFS][4] = future_surface->texture;
+	pipe->set_sampler_textures(pipe, 5, mc->textures[mc->cur_buf % NUM_BUFS]);
 	pipe->bind_sampler_states(pipe, 5, (void**)mc->samplers);
 
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
+	
+	mc->cur_buf++;
 
 	return 0;
 }
@@ -724,8 +733,12 @@ int vlDestroy
 		pipe->winsys->buffer_destroy(pipe->winsys, mc->vertex_bufs[i].buffer);
 
 	/* Textures 3 & 4 are not created directly, no need to release them here */
-	for (i = 0; i < 3; ++i)
-		pipe_texture_release(&mc->textures[i]);
+	for (i = 0; i < NUM_BUFS; ++i)
+	{
+		pipe_texture_release(&mc->textures[i][0]);
+		pipe_texture_release(&mc->textures[i][1]);
+		pipe_texture_release(&mc->textures[i][2]);
+	}
 
 	pipe->delete_vs_state(pipe, mc->i_vs);
 	pipe->delete_fs_state(pipe, mc->i_fs);
@@ -2252,7 +2265,8 @@ static int vlInit
 	template.compressed = 0;
 	pf_get_block(template.format, &template.block);
 
-	mc->textures[0] = pipe->screen->texture_create(pipe->screen, &template);
+	for (i = 0; i < NUM_BUFS; ++i)
+		mc->textures[i][0] = pipe->screen->texture_create(pipe->screen, &template);
 
 	if (mc->video_format == vlFormatYCbCr420)
 		template.height[0] = 8;
@@ -2263,8 +2277,11 @@ static int vlInit
 	else
 		assert(0);
 
-	mc->textures[1] = pipe->screen->texture_create(pipe->screen, &template);
-	mc->textures[2] = pipe->screen->texture_create(pipe->screen, &template);
+	for (i = 0; i < NUM_BUFS; ++i)
+	{
+		mc->textures[i][1] = pipe->screen->texture_create(pipe->screen, &template);
+		mc->textures[i][2] = pipe->screen->texture_create(pipe->screen, &template);
+	}
 
 	/* textures[3] & textures[4] are assigned from vlSurfaces for P and B macroblocks at render time */
 
@@ -2306,6 +2323,7 @@ int vlCreateR16SNormMC
 	mc->pipe = pipe;
 	mc->video_width = video_width;
 	mc->video_height = video_height;
+	mc->cur_buf = 0;
 
 	vlInit(mc);
 
diff --git a/src/gallium/winsys/g3dvl/nouveau/Makefile b/src/gallium/winsys/g3dvl/nouveau/Makefile
index 2861bd7db4..7fa29d2f5f 100644
--- a/src/gallium/winsys/g3dvl/nouveau/Makefile
+++ b/src/gallium/winsys/g3dvl/nouveau/Makefile
@@ -9,7 +9,7 @@ OBJECTS		= nouveau_bo.o nouveau_fence.o nouveau_swapbuffers.o nouveau_channel.o
 		  nouveau_pushbuf.o nouveau_resource.o nouveau_screen.o nv04_surface.o		\
 		  nv50_surface.o #nouveau_winsys_softpipe.o
 
-CFLAGS		+= -g -Wall -Werror -fPIC		\
+CFLAGS		+= -g -Wall -fPIC			\
 		   -I${GALLIUMDIR}/include		\
 		   -I${GALLIUMDIR}/winsys/g3dvl		\
 		   -I${DRMDIR}/include			\
diff --git a/src/libXvMC/Makefile b/src/libXvMC/Makefile
index c154e17dc3..b72bb16efb 100644
--- a/src/libXvMC/Makefile
+++ b/src/libXvMC/Makefile
@@ -8,7 +8,7 @@ ifeq (${DRIVER}, softpipe)
 OBJECTS         += ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
 endif
 
-CFLAGS	+= -g -fPIC -Wall -Werror			\
+CFLAGS	+= -g -fPIC -Wall				\
 	   -I${GALLIUMDIR}/state_trackers/g3dvl		\
 	   -I${GALLIUMDIR}/winsys/g3dvl			\
 	   -I${GALLIUMDIR}/include			\
-- 
cgit v1.2.3


From 4d9d192672508eaa9b2a70f84e933f11108bf09f Mon Sep 17 00:00:00 2001
From: Younes Manton <younes.m@gmail.com>
Date: Thu, 28 Aug 2008 23:25:13 -0400
Subject: g3dvl: Buffer the entire frame before rendering.

---
 src/gallium/state_trackers/g3dvl/Makefile          |    2 +-
 src/gallium/state_trackers/g3dvl/vl_context.c      |    4 +-
 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c  |   45 +-
 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h  |    4 +-
 .../state_trackers/g3dvl/vl_r16snorm_mc_buf.c      | 2341 ++++++++++++++++++++
 .../state_trackers/g3dvl/vl_r16snorm_mc_buf.h      |   18 +
 src/gallium/state_trackers/g3dvl/vl_render.h       |    5 +
 src/gallium/state_trackers/g3dvl/vl_surface.c      |    2 +
 8 files changed, 2400 insertions(+), 21 deletions(-)
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
 create mode 100644 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h

(limited to 'src/gallium/state_trackers/g3dvl/Makefile')

diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
index bd77c62bc5..4f7a953484 100644
--- a/src/gallium/state_trackers/g3dvl/Makefile
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -1,6 +1,6 @@
 TARGET		= libg3dvl.a
 OBJECTS		= vl_display.o vl_screen.o vl_context.o vl_surface.o vl_shader_build.o vl_util.o vl_basic_csc.o	\
-		  vl_r16snorm_mc.o
+		  vl_r16snorm_mc.o vl_r16snorm_mc_buf.o
 GALLIUMDIR	= ../..
 
 CFLAGS		+= -g -Wall -fPIC -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary -I${GALLIUMDIR}/winsys/g3dvl
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c
index 56d360c05b..fe107e406d 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.c
+++ b/src/gallium/state_trackers/g3dvl/vl_context.c
@@ -6,6 +6,7 @@
 #include <pipe/p_state.h>
 #include "vl_render.h"
 #include "vl_r16snorm_mc.h"
+#include "vl_r16snorm_mc_buf.h"
 #include "vl_csc.h"
 #include "vl_basic_csc.h"
 
@@ -126,7 +127,8 @@ int vlCreateContext
 
 	vlInitCommon(ctx);
 
-	vlCreateR16SNormMC(pipe, picture_width, picture_height, picture_format, &ctx->render);
+	/*vlCreateR16SNormMC(pipe, picture_width, picture_height, picture_format, &ctx->render);*/
+	vlCreateR16SNormBufferedMC(pipe, picture_width, picture_height, picture_format, &ctx->render);
 	vlCreateBasicCSC(pipe, &ctx->csc);
 
 	*context = ctx;
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
index 80b09a6d1d..3272220ef8 100644
--- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
@@ -57,7 +57,7 @@ struct vlR16SnormMC
 	struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
 };
 
-int vlBegin
+static int vlBegin
 (
 	struct vlRender *render
 )
@@ -312,7 +312,7 @@ static int vlGrabBlocks
 	return 0;
 }
 
-int vlRenderIMacroBlock
+static int vlRenderIMacroBlock
 (
 	struct vlR16SnormMC *mc,
 	enum vlPictureType picture_type,
@@ -370,13 +370,13 @@ int vlRenderIMacroBlock
 	pipe->bind_fs_state(pipe, mc->i_fs);
 
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
-	
+
 	mc->cur_buf++;
 
 	return 0;
 }
 
-int vlRenderPMacroBlock
+static int vlRenderPMacroBlock
 (
 	struct vlR16SnormMC *mc,
 	enum vlPictureType picture_type,
@@ -468,13 +468,13 @@ int vlRenderPMacroBlock
 	pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers);
 
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
-	
+
 	mc->cur_buf++;
 
 	return 0;
 }
 
-int vlRenderBMacroBlock
+static int vlRenderBMacroBlock
 (
 	struct vlR16SnormMC *mc,
 	enum vlPictureType picture_type,
@@ -580,13 +580,13 @@ int vlRenderBMacroBlock
 	pipe->bind_sampler_states(pipe, 5, (void**)mc->samplers);
 
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
-	
+
 	mc->cur_buf++;
 
 	return 0;
 }
 
-int vlRenderMacroBlocksMpeg2R16Snorm
+static int vlRenderMacroBlocksMpeg2R16Snorm
 (
 	struct vlRender *render,
 	struct vlMpeg2MacroBlockBatch *batch,
@@ -702,7 +702,17 @@ int vlRenderMacroBlocksMpeg2R16Snorm
 	return 0;
 }
 
-int vlEnd
+static int vlEnd
+(
+	struct vlRender *render
+)
+{
+	assert(render);
+
+	return 0;
+}
+
+static int vlFlush
 (
 	struct vlRender *render
 )
@@ -712,7 +722,7 @@ int vlEnd
 	return 0;
 }
 
-int vlDestroy
+static int vlDestroy
 (
 	struct vlRender *render
 )
@@ -765,7 +775,7 @@ int vlDestroy
  * Need to be scaled to cover mbW*mbH macroblock pixels and translated into
  * position on target surface.
  */
-const struct vlVertex2f macroblock_verts[24] =
+static const struct vlVertex2f macroblock_verts[24] =
 {
 	{0.0f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.0f},
 	{0.5f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.5f},
@@ -785,7 +795,7 @@ const struct vlVertex2f macroblock_verts[24] =
  * in a bW*(bH*4) texture. First luma block located at 0,0->bW,bH; second at
  * 0,bH->bW,2bH; third at 0,2bH->bW,3bH; fourth at 0,3bH->bW,4bH.
  */
-const struct vlVertex2f macroblock_luma_texcoords[24] =
+static const struct vlVertex2f macroblock_luma_texcoords[24] =
 {
 	{0.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.0f},
 	{1.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.25f},
@@ -804,7 +814,7 @@ const struct vlVertex2f macroblock_luma_texcoords[24] =
  * Represents texcoords for the above for rendering 1 chroma block.
  * Straight forward 0,0->1,1 mapping so we can reuse the MB pos vectors.
  */
-const struct vlVertex2f *macroblock_chroma_420_texcoords = macroblock_verts;
+static const struct vlVertex2f *macroblock_chroma_420_texcoords = macroblock_verts;
 
 /*
  * Represents texcoords for the above for rendering 2 chroma blocks arranged
@@ -812,13 +822,13 @@ const struct vlVertex2f *macroblock_chroma_420_texcoords = macroblock_verts;
  * 0,bH->bW,2bH. We can render this with 0,0->1,1 mapping.
  * Straight forward 0,0->1,1 mapping so we can reuse MB pos vectors.
  */
-const struct vlVertex2f *macroblock_chroma_422_texcoords = macroblock_verts;
+static const struct vlVertex2f *macroblock_chroma_422_texcoords = macroblock_verts;
 
 /*
  * Represents texcoords for the above for rendering 4 chroma blocks.
  * Same case as 4 luma blocks.
  */
-const struct vlVertex2f *macroblock_chroma_444_texcoords = macroblock_luma_texcoords;
+static const struct vlVertex2f *macroblock_chroma_444_texcoords = macroblock_luma_texcoords;
 
 /*
  * Used when rendering P and B macroblocks, multiplier is applied to the A channel,
@@ -826,7 +836,7 @@ const struct vlVertex2f *macroblock_chroma_444_texcoords = macroblock_luma_texco
  * get back the differential. The differential is then added to the samples from the
  * reference surface(s).
  */
-const struct vlFragmentShaderConsts fs_consts =
+static const struct vlFragmentShaderConsts fs_consts =
 {
 	{32767.0f / 255.0f, 32767.0f / 255.0f, 32767.0f / 255.0f, 0.0f},
 	{0.5f, 2.0f, 0.0f, 0.0f}
@@ -2093,7 +2103,7 @@ static int vlCreateFragmentShaderFieldBMB
 	return 0;
 }
 
-int vlCreateDataBufs
+static int vlCreateDataBufs
 (
 	struct vlR16SnormMC *mc
 )
@@ -2319,6 +2329,7 @@ int vlCreateR16SNormMC
 	mc->base.vlBegin = &vlBegin;
 	mc->base.vlRenderMacroBlocksMpeg2 = &vlRenderMacroBlocksMpeg2R16Snorm;
 	mc->base.vlEnd = &vlEnd;
+	mc->base.vlFlush = &vlFlush;
 	mc->base.vlDestroy = &vlDestroy;
 	mc->pipe = pipe;
 	mc->video_width = video_width;
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h
index a6eecf05b6..9842926bf7 100644
--- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h
@@ -1,5 +1,5 @@
-#ifndef vl_mc_h
-#define vl_mc_h
+#ifndef vl_r16snorm_mc_h
+#define vl_r16snorm_mc_h
 
 #include "vl_types.h"
 
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
new file mode 100644
index 0000000000..fc383cb8f6
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
@@ -0,0 +1,2341 @@
+#define VL_INTERNAL
+#include "vl_r16snorm_mc_buf.h"
+#include <assert.h>
+#include <stdlib.h>
+#include <pipe/p_context.h>
+#include <pipe/p_winsys.h>
+#include <pipe/p_screen.h>
+#include <pipe/p_state.h>
+#include <pipe/p_util.h>
+#include <pipe/p_inlines.h>
+#include <tgsi/tgsi_parse.h>
+#include <tgsi/tgsi_build.h>
+#include "vl_render.h"
+#include "vl_shader_build.h"
+#include "vl_surface.h"
+#include "vl_util.h"
+#include "vl_types.h"
+#include "vl_defs.h"
+
+#define NUM_BUF_SETS 4	/* Number of rotating buffer sets to use */
+
+enum vlMacroBlockTypeEx
+{
+	vlMacroBlockExTypeIntra,
+	vlMacroBlockExTypeFwdPredictedFrame,
+	vlMacroBlockExTypeFwdPredictedField,
+	vlMacroBlockExTypeBkwdPredictedFrame,
+	vlMacroBlockExTypeBkwdPredictedField,
+	vlMacroBlockExTypeBiPredictedFrame,
+	vlMacroBlockExTypeBiPredictedField,
+
+	vlNumMacroBlockExTypes
+};
+
+struct vlVertexShaderConsts
+{
+	struct vlVertex4f denorm;
+};
+
+struct vlFragmentShaderConsts
+{
+	struct vlVertex4f multiplier;
+	struct vlVertex4f div;
+};
+
+struct vlR16SnormBufferedMC
+{
+	struct vlRender				base;
+
+	unsigned int				video_width, video_height;
+	enum vlFormat				video_format;
+
+	unsigned int				cur_buf;
+	struct vlSurface			*buffered_surface;
+	struct vlSurface			*past_surface, *future_surface;
+	struct vlVertex2f			surface_tex_inv_size;
+	unsigned int				num_macroblocks[vlNumMacroBlockExTypes];
+	unsigned int				total_num_macroblocks;
+
+	struct pipe_context			*pipe;
+	struct pipe_viewport_state		viewport;
+	struct pipe_framebuffer_state		render_target;
+	struct pipe_sampler_state		*samplers[5];
+	struct pipe_texture			*textures[NUM_BUF_SETS][5];
+	void					*i_vs, *p_vs[2], *b_vs[2];
+	void					*i_fs, *p_fs[2], *b_fs[2];
+	struct pipe_vertex_buffer 		vertex_bufs[NUM_BUF_SETS][vlNumMacroBlockExTypes][3];
+	struct pipe_vertex_element		vertex_elems[5];
+	struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
+};
+
+static int vlBegin
+(
+	struct vlRender *render
+)
+{
+	assert(render);
+
+	return 0;
+}
+
+static int vlGrabFrameCodedBlock(short *src, short *dst, unsigned int dst_pitch)
+{
+	unsigned int y;
+
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		memcpy
+		(
+			dst + y * dst_pitch,
+			src + y * VL_BLOCK_WIDTH,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	return 0;
+}
+
+static int vlGrabFieldCodedBlock(short *src, short *dst, unsigned int dst_pitch)
+{
+	unsigned int y;
+
+	for (y = 0; y < VL_BLOCK_HEIGHT / 2; ++y)
+		memcpy
+		(
+			dst + y * dst_pitch * 2,
+			src + y * VL_BLOCK_WIDTH,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	dst += VL_BLOCK_HEIGHT * dst_pitch;
+
+	for (; y < VL_BLOCK_HEIGHT; ++y)
+		memcpy
+		(
+			dst + y * dst_pitch * 2,
+			src + y * VL_BLOCK_WIDTH,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	return 0;
+}
+
+static int vlGrabNoBlock(short *dst, unsigned int dst_pitch)
+{
+	unsigned int y;
+
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		memset
+		(
+			dst + y * dst_pitch,
+			0,
+			VL_BLOCK_WIDTH * 2
+		);
+
+	return 0;
+}
+
+static int vlGrabBlocks
+(
+	struct vlR16SnormBufferedMC *mc,
+	unsigned int mbx,
+	unsigned int mby,
+	enum vlDCTType dct_type,
+	unsigned int coded_block_pattern,
+	short *blocks
+)
+{
+	struct pipe_surface	*tex_surface;
+	short			*texels;
+	unsigned int		tex_pitch;
+	unsigned int		x, y, tb = 0, sb = 0;
+	unsigned int		mbpx = mbx * VL_MACROBLOCK_WIDTH, mbpy = mby * VL_MACROBLOCK_HEIGHT;
+
+	assert(mc);
+	assert(blocks);
+
+	tex_surface = mc->pipe->screen->get_tex_surface
+	(
+		mc->pipe->screen,
+		mc->textures[mc->cur_buf % NUM_BUF_SETS][0],
+		0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
+	);
+
+	texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE);
+	tex_pitch = tex_surface->stride / tex_surface->block.size;
+
+	texels += mbpy * tex_pitch + mbpx;
+
+	for (y = 0; y < 2; ++y)
+	{
+		for (x = 0; x < 2; ++x, ++tb)
+		{
+			if ((coded_block_pattern >> (5 - tb)) & 1)
+			{
+				short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
+
+				if (dct_type == vlDCTTypeFrameCoded)
+				{
+					vlGrabFrameCodedBlock
+					(
+						cur_block,
+						texels + y * tex_pitch * VL_BLOCK_HEIGHT + x * VL_BLOCK_WIDTH,
+						tex_pitch
+					);
+				}
+				else
+				{
+					vlGrabFieldCodedBlock
+					(
+						cur_block,
+						texels + y * tex_pitch + x * VL_BLOCK_WIDTH,
+						tex_pitch
+					);
+				}
+
+				++sb;
+			}
+			else
+				vlGrabNoBlock(texels + y * tex_pitch * VL_BLOCK_HEIGHT + x * VL_BLOCK_WIDTH, tex_pitch);
+		}
+	}
+
+	pipe_surface_unmap(tex_surface);
+
+	/* TODO: Implement 422, 444 */
+	mbpx >>= 1;
+	mbpy >>= 1;
+
+	for (tb = 0; tb < 2; ++tb)
+	{
+		tex_surface = mc->pipe->screen->get_tex_surface
+		(
+			mc->pipe->screen,
+			mc->textures[mc->cur_buf % NUM_BUF_SETS][tb + 1],
+			0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
+		);
+
+		texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE);
+		tex_pitch = tex_surface->stride / tex_surface->block.size;
+
+		texels += mbpy * tex_pitch + mbpx;
+
+		if ((coded_block_pattern >> (1 - tb)) & 1)
+		{
+			short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
+
+			vlGrabFrameCodedBlock
+			(
+				cur_block,
+				texels,
+				tex_pitch
+			);
+
+			++sb;
+		}
+		else
+			vlGrabNoBlock(texels, tex_pitch);
+
+		pipe_surface_unmap(tex_surface);
+	}
+
+	return 0;
+}
+
+#if 0
+static int vlGrabMacroBlock
+(
+	struct vlR16SnormBufferedMC *mc,
+	struct vlMpeg2MacroBlock *macroblock
+)
+{
+	const struct vlVertex2f	unit =
+	{
+		mc->surface_tex_inv_size.x * VL_MACROBLOCK_WIDTH,
+		mc->surface_tex_inv_size.y * VL_MACROBLOCK_HEIGHT
+	};
+	const struct vlVertex2f half =
+	{
+		mc->surface_tex_inv_size.x * (VL_MACROBLOCK_WIDTH / 2),
+		mc->surface_tex_inv_size.y * (VL_MACROBLOCK_HEIGHT / 2)
+	};
+
+	struct vlVertex2f	*vb;
+	enum vlMacroBlockTypeEx	mb_type_ex;
+	struct vlVertex2f	mo_vec[2];
+	unsigned int		i;
+
+	assert(mc);
+	assert(macroblock);
+
+	switch (macroblock->mb_type)
+	{
+		case vlMacroBlockTypeIntra:
+		{
+			mb_type_ex = vlMacroBlockExTypeIntra;
+			break;
+		}
+		case vlMacroBlockTypeFwdPredicted:
+		{
+			mb_type_ex = macroblock->mo_type == vlMotionTypeFrame ?
+				vlMacroBlockExTypeFwdPredictedFrame : vlMacroBlockExTypeFwdPredictedField;
+			break;
+		}
+		case vlMacroBlockTypeBkwdPredicted:
+		{
+			mb_type_ex = macroblock->mo_type == vlMotionTypeFrame ?
+				vlMacroBlockExTypeBkwdPredictedFrame : vlMacroBlockExTypeBkwdPredictedField;
+			break;
+		}
+		case vlMacroBlockTypeBiPredicted:
+		{
+			mb_type_ex = macroblock->mo_type == vlMotionTypeFrame ?
+				vlMacroBlockExTypeBiPredictedFrame : vlMacroBlockExTypeBiPredictedField;
+			break;
+		}
+		default:
+			assert(0);
+	}
+
+	switch (macroblock->mb_type)
+	{
+		case vlMacroBlockTypeBiPredicted:
+		{
+			vb = (struct vlVertex2f*)mc->pipe->winsys->buffer_map
+			(
+				mc->pipe->winsys,
+				mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_type_ex][2].buffer,
+				PIPE_BUFFER_USAGE_CPU_WRITE
+			) + mc->num_macroblocks[mb_type_ex] * 2 * 24;
+
+			mo_vec[0].x = macroblock->PMV[0][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+			mo_vec[0].y = macroblock->PMV[0][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+			if (macroblock->mo_type == vlMotionTypeFrame)
+			{
+				for (i = 0; i < 24 * 2; i += 2)
+				{
+					vb[i].x = mo_vec[0].x;
+					vb[i].y = mo_vec[0].y;
+				}
+			}
+			else
+			{
+				mo_vec[1].x = macroblock->PMV[1][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+				mo_vec[1].y = macroblock->PMV[1][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+				for (i = 0; i < 24 * 2; i += 2)
+				{
+					vb[i].x = mo_vec[0].x;
+					vb[i].y = mo_vec[0].y;
+					vb[i + 1].x = mo_vec[1].x;
+					vb[i + 1].y = mo_vec[1].y;
+				}
+			}
+
+			mc->pipe->winsys->buffer_unmap(mc->pipe->winsys, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_type_ex][2].buffer);
+
+			/* fall-through */
+		}
+		case vlMacroBlockTypeFwdPredicted:
+		case vlMacroBlockTypeBkwdPredicted:
+		{
+			vb = (struct vlVertex2f*)mc->pipe->winsys->buffer_map
+			(
+				mc->pipe->winsys,
+				mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_type_ex][1].buffer,
+				PIPE_BUFFER_USAGE_CPU_WRITE
+			) + mc->num_macroblocks[mb_type_ex] * 2 * 24;
+
+			if (macroblock->mb_type == vlMacroBlockTypeBkwdPredicted)
+			{
+				mo_vec[0].x = macroblock->PMV[0][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+				mo_vec[0].y = macroblock->PMV[0][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+				if (macroblock->mo_type == vlMotionTypeField)
+				{
+					mo_vec[1].x = macroblock->PMV[1][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+					mo_vec[1].y = macroblock->PMV[1][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+				}
+			}
+			else
+			{
+				mo_vec[0].x = macroblock->PMV[0][0][0] * 0.5f * mc->surface_tex_inv_size.x;
+				mo_vec[0].y = macroblock->PMV[0][0][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+				if (macroblock->mo_type == vlMotionTypeField)
+				{
+					mo_vec[1].x = macroblock->PMV[1][0][0] * 0.5f * mc->surface_tex_inv_size.x;
+					mo_vec[1].y = macroblock->PMV[1][0][1] * 0.5f * mc->surface_tex_inv_size.y;
+				}
+			}
+
+			if (macroblock->mo_type == vlMotionTypeFrame)
+			{
+				for (i = 0; i < 24 * 2; i += 2)
+				{
+					vb[i].x = mo_vec[0].x;
+					vb[i].y = mo_vec[0].y;
+				}
+			}
+			else
+			{
+				for (i = 0; i < 24 * 2; i += 2)
+				{
+					vb[i].x = mo_vec[0].x;
+					vb[i].y = mo_vec[0].y;
+					vb[i + 1].x = mo_vec[1].x;
+					vb[i + 1].y = mo_vec[1].y;
+				}
+			}
+
+			mc->pipe->winsys->buffer_unmap(mc->pipe->winsys, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_type_ex][1].buffer);
+
+			/* fall-through */
+		}
+		case vlMacroBlockTypeIntra:
+		{
+			vb = (struct vlVertex2f*)mc->pipe->winsys->buffer_map
+			(
+				mc->pipe->winsys,
+				mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_type_ex][0].buffer,
+				PIPE_BUFFER_USAGE_CPU_WRITE
+			) + mc->num_macroblocks[mb_type_ex] * 24;
+
+			vb[0].x = macroblock->mbx * unit.x;		vb[0].y = macroblock->mby * unit.y;
+			vb[1].x = macroblock->mbx * unit.x;		vb[1].y = macroblock->mby * unit.y + half.y;
+			vb[2].x = macroblock->mbx * unit.x + half.x;	vb[2].y = macroblock->mby * unit.y;
+
+			vb[3].x = macroblock->mbx * unit.x + half.x;	vb[3].y = macroblock->mby * unit.y;
+			vb[4].x = macroblock->mbx * unit.x;		vb[4].y = macroblock->mby * unit.y + half.y;
+			vb[5].x = macroblock->mbx * unit.x + half.x;	vb[5].y = macroblock->mby * unit.y + half.y;
+
+			vb[6].x = macroblock->mbx * unit.x + half.x;	vb[6].y = macroblock->mby * unit.y;
+			vb[7].x = macroblock->mbx * unit.x + half.x;	vb[7].y = macroblock->mby * unit.y + half.y;
+			vb[8].x = macroblock->mbx * unit.x + unit.x;	vb[8].y = macroblock->mby * unit.y;
+
+			vb[9].x = macroblock->mbx * unit.x + unit.x;	vb[9].y = macroblock->mby * unit.y;
+			vb[10].x = macroblock->mbx * unit.x + half.x;	vb[10].y = macroblock->mby * unit.y + half.y;
+			vb[11].x = macroblock->mbx * unit.x + unit.x;	vb[11].y = macroblock->mby * unit.y + half.y;
+
+			vb[12].x = macroblock->mbx * unit.x;		vb[12].y = macroblock->mby * unit.y + half.y;
+			vb[13].x = macroblock->mbx * unit.x;		vb[13].y = macroblock->mby * unit.y + unit.y;
+			vb[14].x = macroblock->mbx * unit.x + half.x;	vb[14].y = macroblock->mby * unit.y + half.y;
+
+			vb[15].x = macroblock->mbx * unit.x + half.x;	vb[15].y = macroblock->mby * unit.y + half.y;
+			vb[16].x = macroblock->mbx * unit.x;		vb[16].y = macroblock->mby * unit.y + unit.y;
+			vb[17].x = macroblock->mbx * unit.x + half.x;	vb[17].y = macroblock->mby * unit.y + unit.y;
+
+			vb[18].x = macroblock->mbx * unit.x + half.x;	vb[18].y = macroblock->mby * unit.y + half.y;
+			vb[19].x = macroblock->mbx * unit.x + half.x;	vb[19].y = macroblock->mby * unit.y + unit.y;
+			vb[20].x = macroblock->mbx * unit.x + unit.x;	vb[20].y = macroblock->mby * unit.y + half.y;
+
+			vb[21].x = macroblock->mbx * unit.x + unit.x;	vb[21].y = macroblock->mby * unit.y + half.y;
+			vb[22].x = macroblock->mbx * unit.x + half.x;	vb[22].y = macroblock->mby * unit.y + unit.y;
+			vb[23].x = macroblock->mbx * unit.x + unit.x;	vb[23].y = macroblock->mby * unit.y + unit.y;
+
+			mc->pipe->winsys->buffer_unmap(mc->pipe->winsys, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_type_ex][0].buffer);
+
+			break;
+		}
+		default:
+			assert(0);
+	}
+
+	vlGrabBlocks
+	(
+		mc,
+		macroblock->mbx,
+		macroblock->mby,
+		macroblock->dct_type,
+		macroblock->cbp,
+		macroblock->blocks
+	);
+
+	mc->num_macroblocks[mb_type_ex]++;
+	mc->total_num_macroblocks++;
+
+	return 0;
+}
+#else
+static int vlGrabMacroBlock
+(
+	struct vlR16SnormBufferedMC *mc,
+	struct vlMpeg2MacroBlock *macroblock
+)
+{
+	const struct vlVertex2f	unit =
+	{
+		mc->surface_tex_inv_size.x * VL_MACROBLOCK_WIDTH,
+		mc->surface_tex_inv_size.y * VL_MACROBLOCK_HEIGHT
+	};
+	const struct vlVertex2f half =
+	{
+		mc->surface_tex_inv_size.x * (VL_MACROBLOCK_WIDTH / 2),
+		mc->surface_tex_inv_size.y * (VL_MACROBLOCK_HEIGHT / 2)
+	};
+
+	struct vlVertex2f	*vb;
+	unsigned int		mb_buf_id;
+	struct vlVertex2f	mo_vec[2];
+	unsigned int		i;
+
+	assert(mc);
+	assert(macroblock);
+
+	switch (macroblock->mb_type)
+	{
+		case vlMacroBlockTypeIntra:
+		{
+			mb_buf_id = vlMacroBlockExTypeIntra;
+			break;
+		}
+		case vlMacroBlockTypeFwdPredicted:
+		{
+			mb_buf_id = macroblock->mo_type == vlMotionTypeFrame ?
+				vlMacroBlockExTypeFwdPredictedFrame : vlMacroBlockExTypeFwdPredictedField;
+			break;
+		}
+		case vlMacroBlockTypeBkwdPredicted:
+		{
+			mb_buf_id = macroblock->mo_type == vlMotionTypeFrame ?
+				vlMacroBlockExTypeBkwdPredictedFrame : vlMacroBlockExTypeBkwdPredictedField;
+			break;
+		}
+		case vlMacroBlockTypeBiPredicted:
+		{
+			mb_buf_id = macroblock->mo_type == vlMotionTypeFrame ?
+				vlMacroBlockExTypeBiPredictedFrame : vlMacroBlockExTypeBiPredictedField;
+			break;
+		}
+		default:
+			assert(0);
+	}
+
+	vb = (struct vlVertex2f*)mc->pipe->winsys->buffer_map
+	(
+		mc->pipe->winsys,
+		mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_buf_id][0].buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE
+	) + mc->num_macroblocks[mb_buf_id] * 24;
+
+	vb[0].x = macroblock->mbx * unit.x;		vb[0].y = macroblock->mby * unit.y;
+	vb[1].x = macroblock->mbx * unit.x;		vb[1].y = macroblock->mby * unit.y + half.y;
+	vb[2].x = macroblock->mbx * unit.x + half.x;	vb[2].y = macroblock->mby * unit.y;
+
+	vb[3].x = macroblock->mbx * unit.x + half.x;	vb[3].y = macroblock->mby * unit.y;
+	vb[4].x = macroblock->mbx * unit.x;		vb[4].y = macroblock->mby * unit.y + half.y;
+	vb[5].x = macroblock->mbx * unit.x + half.x;	vb[5].y = macroblock->mby * unit.y + half.y;
+
+	vb[6].x = macroblock->mbx * unit.x + half.x;	vb[6].y = macroblock->mby * unit.y;
+	vb[7].x = macroblock->mbx * unit.x + half.x;	vb[7].y = macroblock->mby * unit.y + half.y;
+	vb[8].x = macroblock->mbx * unit.x + unit.x;	vb[8].y = macroblock->mby * unit.y;
+
+	vb[9].x = macroblock->mbx * unit.x + unit.x;	vb[9].y = macroblock->mby * unit.y;
+	vb[10].x = macroblock->mbx * unit.x + half.x;	vb[10].y = macroblock->mby * unit.y + half.y;
+	vb[11].x = macroblock->mbx * unit.x + unit.x;	vb[11].y = macroblock->mby * unit.y + half.y;
+
+	vb[12].x = macroblock->mbx * unit.x;		vb[12].y = macroblock->mby * unit.y + half.y;
+	vb[13].x = macroblock->mbx * unit.x;		vb[13].y = macroblock->mby * unit.y + unit.y;
+	vb[14].x = macroblock->mbx * unit.x + half.x;	vb[14].y = macroblock->mby * unit.y + half.y;
+
+	vb[15].x = macroblock->mbx * unit.x + half.x;	vb[15].y = macroblock->mby * unit.y + half.y;
+	vb[16].x = macroblock->mbx * unit.x;		vb[16].y = macroblock->mby * unit.y + unit.y;
+	vb[17].x = macroblock->mbx * unit.x + half.x;	vb[17].y = macroblock->mby * unit.y + unit.y;
+
+	vb[18].x = macroblock->mbx * unit.x + half.x;	vb[18].y = macroblock->mby * unit.y + half.y;
+	vb[19].x = macroblock->mbx * unit.x + half.x;	vb[19].y = macroblock->mby * unit.y + unit.y;
+	vb[20].x = macroblock->mbx * unit.x + unit.x;	vb[20].y = macroblock->mby * unit.y + half.y;
+
+	vb[21].x = macroblock->mbx * unit.x + unit.x;	vb[21].y = macroblock->mby * unit.y + half.y;
+	vb[22].x = macroblock->mbx * unit.x + half.x;	vb[22].y = macroblock->mby * unit.y + unit.y;
+	vb[23].x = macroblock->mbx * unit.x + unit.x;	vb[23].y = macroblock->mby * unit.y + unit.y;
+
+	mc->pipe->winsys->buffer_unmap(mc->pipe->winsys, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_buf_id][0].buffer);
+
+	if (macroblock->mb_type == vlMacroBlockTypeIntra)
+	{
+		vlGrabBlocks
+		(
+			mc,
+			macroblock->mbx,
+			macroblock->mby,
+			macroblock->dct_type,
+			macroblock->cbp,
+			macroblock->blocks
+		);
+
+		mc->num_macroblocks[mb_buf_id]++;
+		mc->total_num_macroblocks++;
+		return 0;
+	}
+
+	vb = (struct vlVertex2f*)mc->pipe->winsys->buffer_map
+	(
+		mc->pipe->winsys,
+		mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_buf_id][1].buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE
+	) + mc->num_macroblocks[mb_buf_id] * 2 * 24;
+
+	if (macroblock->mb_type == vlMacroBlockTypeFwdPredicted || macroblock->mb_type == vlMacroBlockTypeBiPredicted)
+	{
+		mo_vec[0].x = macroblock->PMV[0][0][0] * 0.5f * mc->surface_tex_inv_size.x;
+		mo_vec[0].y = macroblock->PMV[0][0][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+		if (macroblock->mo_type == vlMotionTypeField)
+		{
+			mo_vec[1].x = macroblock->PMV[1][0][0] * 0.5f * mc->surface_tex_inv_size.x;
+			mo_vec[1].y = macroblock->PMV[1][0][1] * 0.5f * mc->surface_tex_inv_size.y;
+		}
+	}
+	else
+	{
+		mo_vec[0].x = macroblock->PMV[0][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+		mo_vec[0].y = macroblock->PMV[0][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+		if (macroblock->mo_type == vlMotionTypeField)
+		{
+			mo_vec[1].x = macroblock->PMV[1][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+			mo_vec[1].y = macroblock->PMV[1][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+		}
+	}
+
+	if (macroblock->mo_type == vlMotionTypeFrame)
+	{
+		for (i = 0; i < 24 * 2; i += 2)
+		{
+			vb[i].x = mo_vec[0].x;
+			vb[i].y = mo_vec[0].y;
+		}
+	}
+	else
+	{
+		for (i = 0; i < 24 * 2; i += 2)
+		{
+			vb[i].x = mo_vec[0].x;
+			vb[i].y = mo_vec[0].y;
+			vb[i + 1].x = mo_vec[1].x;
+			vb[i + 1].y = mo_vec[1].y;
+		}
+	}
+
+	mc->pipe->winsys->buffer_unmap(mc->pipe->winsys, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_buf_id][1].buffer);
+
+	if (macroblock->mb_type != vlMacroBlockTypeBiPredicted)
+	{
+		vlGrabBlocks
+		(
+			mc,
+			macroblock->mbx,
+			macroblock->mby,
+			macroblock->dct_type,
+			macroblock->cbp,
+			macroblock->blocks
+		);
+
+		mc->num_macroblocks[mb_buf_id]++;
+		mc->total_num_macroblocks++;
+		return 0;
+	}
+
+	vb = (struct vlVertex2f*)mc->pipe->winsys->buffer_map
+	(
+		mc->pipe->winsys,
+		mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_buf_id][2].buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE
+	) + mc->num_macroblocks[mb_buf_id] * 2 * 24;
+
+	mo_vec[0].x = macroblock->PMV[0][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+	mo_vec[0].y = macroblock->PMV[0][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+	if (macroblock->mo_type == vlMotionTypeFrame)
+	{
+		for (i = 0; i < 24 * 2; i += 2)
+		{
+			vb[i].x = mo_vec[0].x;
+			vb[i].y = mo_vec[0].y;
+		}
+	}
+	else
+	{
+		mo_vec[1].x = macroblock->PMV[1][1][0] * 0.5f * mc->surface_tex_inv_size.x;
+		mo_vec[1].y = macroblock->PMV[1][1][1] * 0.5f * mc->surface_tex_inv_size.y;
+
+		for (i = 0; i < 24 * 2; i += 2)
+		{
+			vb[i].x = mo_vec[0].x;
+			vb[i].y = mo_vec[0].y;
+			vb[i + 1].x = mo_vec[1].x;
+			vb[i + 1].y = mo_vec[1].y;
+		}
+	}
+
+	mc->pipe->winsys->buffer_unmap(mc->pipe->winsys, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][mb_buf_id][2].buffer);
+
+	vlGrabBlocks
+	(
+		mc,
+		macroblock->mbx,
+		macroblock->mby,
+		macroblock->dct_type,
+		macroblock->cbp,
+		macroblock->blocks
+	);
+
+	mc->num_macroblocks[mb_buf_id]++;
+	mc->total_num_macroblocks++;
+
+	return 0;
+}
+#endif
+
+static int vlFlush
+(
+	struct vlRender *render
+)
+{
+	struct vlR16SnormBufferedMC	*mc;
+	struct pipe_context		*pipe;
+	struct vlVertexShaderConsts	*vs_consts;
+
+	assert(mc);
+
+	mc = (struct vlR16SnormBufferedMC*)render;
+	pipe = mc->pipe;
+
+	mc->render_target.cbufs[0] = pipe->screen->get_tex_surface
+	(
+		pipe->screen,
+		mc->buffered_surface->texture,
+		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
+	);
+
+	pipe->set_framebuffer_state(pipe, &mc->render_target);
+	pipe->set_viewport_state(pipe, &mc->viewport);
+	vs_consts = pipe->winsys->buffer_map
+	(
+		pipe->winsys,
+		mc->vs_const_buf.buffer,
+		PIPE_BUFFER_USAGE_CPU_WRITE
+	);
+
+	vs_consts->denorm.x = mc->buffered_surface->texture->width[0];
+	vs_consts->denorm.y = mc->buffered_surface->texture->height[0];
+
+	pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer);
+	pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &mc->fs_const_buf);
+
+	if (mc->num_macroblocks[vlMacroBlockExTypeIntra] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 1, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][vlMacroBlockExTypeIntra]);
+		pipe->set_vertex_elements(pipe, 1, mc->vertex_elems);
+		pipe->set_sampler_textures(pipe, 3, mc->textures[mc->cur_buf % NUM_BUF_SETS]);
+		pipe->bind_sampler_states(pipe, 3, (void**)mc->samplers);
+		pipe->bind_vs_state(pipe, mc->i_vs);
+		pipe->bind_fs_state(pipe, mc->i_fs);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, mc->num_macroblocks[vlMacroBlockExTypeIntra] * 24);
+	}
+
+	if (mc->num_macroblocks[vlMacroBlockExTypeFwdPredictedFrame] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][vlMacroBlockExTypeFwdPredictedFrame]);
+		pipe->set_vertex_elements(pipe, 3, mc->vertex_elems);
+		mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->past_surface->texture;
+		pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUF_SETS]);
+		pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers);
+		pipe->bind_vs_state(pipe, mc->p_vs[0]);
+		pipe->bind_fs_state(pipe, mc->p_fs[0]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, mc->num_macroblocks[vlMacroBlockExTypeFwdPredictedFrame] * 24);
+	}
+
+	if (mc->num_macroblocks[vlMacroBlockExTypeFwdPredictedField] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][vlMacroBlockExTypeFwdPredictedField]);
+		pipe->set_vertex_elements(pipe, 3, mc->vertex_elems);
+		mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->past_surface->texture;
+		pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUF_SETS]);
+		pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers);
+		pipe->bind_vs_state(pipe, mc->p_vs[1]);
+		pipe->bind_fs_state(pipe, mc->p_fs[1]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, mc->num_macroblocks[vlMacroBlockExTypeFwdPredictedField] * 24);
+	}
+
+	if (mc->num_macroblocks[vlMacroBlockExTypeBkwdPredictedFrame] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][vlMacroBlockExTypeBkwdPredictedFrame]);
+		pipe->set_vertex_elements(pipe, 3, mc->vertex_elems);
+		mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->future_surface->texture;
+		pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUF_SETS]);
+		pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers);
+		pipe->bind_vs_state(pipe, mc->p_vs[0]);
+		pipe->bind_fs_state(pipe, mc->p_fs[0]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, mc->num_macroblocks[vlMacroBlockExTypeBkwdPredictedFrame] * 24);
+	}
+
+	if (mc->num_macroblocks[vlMacroBlockExTypeBkwdPredictedField] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 2, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][vlMacroBlockExTypeBkwdPredictedField]);
+		pipe->set_vertex_elements(pipe, 3, mc->vertex_elems);
+		mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->future_surface->texture;
+		pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUF_SETS]);
+		pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers);
+		pipe->bind_vs_state(pipe, mc->p_vs[1]);
+		pipe->bind_fs_state(pipe, mc->p_fs[1]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, mc->num_macroblocks[vlMacroBlockExTypeBkwdPredictedField] * 24);
+	}
+
+	if (mc->num_macroblocks[vlMacroBlockExTypeBiPredictedFrame] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 3, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][vlMacroBlockExTypeBiPredictedFrame]);
+		pipe->set_vertex_elements(pipe, 5, mc->vertex_elems);
+		mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->past_surface->texture;
+		mc->textures[mc->cur_buf % NUM_BUF_SETS][4] = mc->future_surface->texture;
+		pipe->set_sampler_textures(pipe, 5, mc->textures[mc->cur_buf % NUM_BUF_SETS]);
+		pipe->bind_sampler_states(pipe, 5, (void**)mc->samplers);
+		pipe->bind_vs_state(pipe, mc->b_vs[0]);
+		pipe->bind_fs_state(pipe, mc->b_fs[0]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, mc->num_macroblocks[vlMacroBlockExTypeBiPredictedFrame] * 24);
+	}
+
+	if (mc->num_macroblocks[vlMacroBlockExTypeBiPredictedField] > 0)
+	{
+		pipe->set_vertex_buffers(pipe, 3, mc->vertex_bufs[mc->cur_buf % NUM_BUF_SETS][vlMacroBlockExTypeBiPredictedField]);
+		pipe->set_vertex_elements(pipe, 5, mc->vertex_elems);
+		mc->textures[mc->cur_buf % NUM_BUF_SETS][3] = mc->past_surface->texture;
+		mc->textures[mc->cur_buf % NUM_BUF_SETS][4] = mc->future_surface->texture;
+		pipe->set_sampler_textures(pipe, 5, mc->textures[mc->cur_buf % NUM_BUF_SETS]);
+		pipe->bind_sampler_states(pipe, 5, (void**)mc->samplers);
+		pipe->bind_vs_state(pipe, mc->b_vs[1]);
+		pipe->bind_fs_state(pipe, mc->b_fs[1]);
+
+		pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, mc->num_macroblocks[vlMacroBlockExTypeBiPredictedField] * 24);
+	}
+
+	memset(mc->num_macroblocks, 0, sizeof(unsigned int) * 7);
+	mc->total_num_macroblocks = 0;
+
+	return 0;
+}
+
+static int vlRenderMacroBlocksMpeg2R16SnormBuffered
+(
+	struct vlRender *render,
+	struct vlMpeg2MacroBlockBatch *batch,
+	struct vlSurface *surface
+)
+{
+	struct vlR16SnormBufferedMC	*mc;
+	unsigned int			i;
+
+	assert(render);
+
+	mc = (struct vlR16SnormBufferedMC*)render;
+
+	if (mc->buffered_surface)
+	{
+		if
+		(
+			mc->buffered_surface != surface /*||
+			mc->past_surface != batch->past_surface ||
+			mc->future_surface != batch->future_surface*/
+		)
+		{
+			vlFlush(&mc->base);
+			mc->buffered_surface = surface;
+			mc->past_surface = batch->past_surface;
+			mc->future_surface = batch->future_surface;
+			mc->surface_tex_inv_size.x = 1.0f / surface->texture->width[0];
+			mc->surface_tex_inv_size.y = 1.0f / surface->texture->height[0];
+		}
+	}
+	else
+	{
+		mc->buffered_surface = surface;
+		mc->past_surface = batch->past_surface;
+		mc->future_surface = batch->future_surface;
+		mc->surface_tex_inv_size.x = 1.0f / surface->texture->width[0];
+		mc->surface_tex_inv_size.y = 1.0f / surface->texture->height[0];
+	}
+
+	for (i = 0; i < batch->num_macroblocks; ++i)
+		vlGrabMacroBlock(mc, &batch->macroblocks[i]);
+
+	return 0;
+}
+
+static int vlEnd
+(
+	struct vlRender *render
+)
+{
+	assert(render);
+
+	return 0;
+}
+
+static int vlDestroy
+(
+	struct vlRender *render
+)
+{
+	struct vlR16SnormBufferedMC	*mc;
+	struct pipe_context		*pipe;
+	unsigned int			g, h, i;
+
+	assert(render);
+
+	mc = (struct vlR16SnormBufferedMC*)render;
+	pipe = mc->pipe;
+
+	for (i = 0; i < 5; ++i)
+		pipe->delete_sampler_state(pipe, mc->samplers[i]);
+
+	for (g = 0; g < NUM_BUF_SETS; ++g)
+		for (h = 0; h < 7; ++h)
+			for (i = 0; i < 3; ++i)
+				pipe->winsys->buffer_destroy(pipe->winsys, mc->vertex_bufs[g][h][i].buffer);
+
+	/* Textures 3 & 4 are not created directly, no need to release them here */
+	for (i = 0; i < NUM_BUF_SETS; ++i)
+	{
+		pipe_texture_release(&mc->textures[i][0]);
+		pipe_texture_release(&mc->textures[i][1]);
+		pipe_texture_release(&mc->textures[i][2]);
+	}
+
+	pipe->delete_vs_state(pipe, mc->i_vs);
+	pipe->delete_fs_state(pipe, mc->i_fs);
+
+	for (i = 0; i < 2; ++i)
+	{
+		pipe->delete_vs_state(pipe, mc->p_vs[i]);
+		pipe->delete_fs_state(pipe, mc->p_fs[i]);
+		pipe->delete_vs_state(pipe, mc->b_vs[i]);
+		pipe->delete_fs_state(pipe, mc->b_fs[i]);
+	}
+
+	pipe->winsys->buffer_destroy(pipe->winsys, mc->vs_const_buf.buffer);
+	pipe->winsys->buffer_destroy(pipe->winsys, mc->fs_const_buf.buffer);
+
+	free(mc);
+
+	return 0;
+}
+
+/*
+ * Muliplier renormalizes block samples from 16 bits to 12 bits.
+ * Divider is used when calculating Y % 2 for choosing top or bottom
+ * field for P or B macroblocks.
+ * TODO: Use immediates.
+ */
+static const struct vlFragmentShaderConsts fs_consts =
+{
+	{32767.0f / 255.0f, 32767.0f / 255.0f, 32767.0f / 255.0f, 0.0f},
+	{0.5f, 2.0f, 0.0f, 0.0f}
+};
+
+static int vlCreateVertexShaderIMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 50;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos, luma & chroma texcoords
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma/chroma texcoords
+	 */
+	for (i = 0; i < 2; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i0		; Move input luma/chroma texcoords to output
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, 0);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->i_vs = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderIMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/* decl i0			; Luma/chroma texcoords */
+	decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, 1, 0, 0, TGSI_INTERPOLATE_LINEAR);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header,max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i0, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i0, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul o0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->i_fs = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFramePMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos, luma/chroma texcoords
+	 * decl i1		; Ref surface top field texcoords
+	 * decl i2		; Ref surface bottom field texcoords (unused, packed in the same stream)
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma/chroma texcoords
+	 * decl o2		; Ref macroblock texcoords
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i0		; Move input luma/chroma texcoords to output
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, 0);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* add o2, i0, i1	; Translate vertex pos by motion vec to form ref macroblock texcoords */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 2, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->p_vs[0] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFieldPMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos, luma/chroma texcoords
+	 * decl i1              ; Ref surface top field texcoords
+	 * decl i2              ; Ref surface bottom field texcoords
+	 */
+	for (i = 0; i < 3; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl c0		; Texcoord denorm coefficients */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma/chroma texcoords
+	 * decl o2		; Top field ref macroblock texcoords
+	 * decl o3		; Bottom field ref macroblock texcoords
+	 * decl o4		; Denormalized vertex pos
+	 */
+	for (i = 0; i < 5; i++)
+	{
+		decl = vl_decl_output((i == 0 || i == 5) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i0		; Move input luma/chroma texcoords to output
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i == 0 ? 0 : i - 1);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * add o2, i0, i1	; Translate vertex pos by motion vec to form top field macroblock texcoords
+	 * add o3, i0, i2	; Translate vertex pos by motion vec to form bottom field macroblock texcoords
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 2, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, i + 1);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul o4, i0, c0	; Denorm vertex pos */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 5, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->p_vs[1] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFramePMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Texcoords for s0, s1, s2
+	 * decl i1			; Texcoords for s3
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for ref surface texture
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i0, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i0, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* tex2d t1, i1, s3		; Read texel from ref macroblock */
+	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 1, TGSI_FILE_SAMPLER, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->p_fs[0] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFieldPMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 200;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Texcoords for s0, s1, s2
+	 * decl i1			; Texcoords for s3
+	 * decl i2			; Texcoords for s3
+	 * decl i3			; Denormalized vertex pos
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
+	 * decl c1			; Constants 1/2 & 2 in .x, .y channels for Y-mod-2 top/bottom field selection
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0-t4 */
+	decl = vl_decl_temps(0, 4);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for ref surface texture
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i0, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i0, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t1, i1, s3		; Read texel from ref macroblock top field
+	 * tex2d t2, i2, s3		; Read texel from ref macroblock bottom field
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 1, TGSI_FILE_SAMPLER, 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* XXX: Pos values off by 0.5? */
+	/* sub t4, i3.y, c1.x		; Sub 0.5 from denormalized pos */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 3, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* floor t3, t3			; Get rid of fractional part */
+	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t3, c1.y		; Multiply by 2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->p_fs[1] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFrameBMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos, luma/chroma texcoords
+	 * decl i1              ; First ref surface top field texcoords
+	 * decl i2              ; First ref surface bottom field texcoords (unused, packed in the same stream)
+	 * decl i3		; Second ref surface top field texcoords
+	 * decl i4		; Second ref surface bottom field texcoords (unused, packed in the same stream)
+	 */
+	for (i = 0; i < 5; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma/chroma texcoords
+	 * decl o2		; First ref macroblock texcoords
+	 * decl o3		; Second ref macroblock texcoords
+	 */
+	for (i = 0; i < 4; i++)
+	{
+		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i0		; Move input luma/chroma texcoords to output
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, 0);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * add o2, i0, i1	; Translate vertex pos by motion vec to form first ref macroblock texcoords
+	 * add o3, i0, i3	; Translate vertex pos by motion vec to form second ref macroblock texcoords
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 2, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, i * 2 + 1);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->b_vs[0] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateVertexShaderFieldBMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	vs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0		; Vertex pos, Luma/chroma texcoords
+	 * decl i1              ; First ref surface top field texcoords
+	 * decl i2              ; First ref surface bottom field texcoords
+	 * decl i3              ; Second ref surface top field texcoords
+	 * decl i4              ; Second ref surface bottom field texcoords
+	 */
+	for (i = 0; i < 5; i++)
+	{
+		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl c0		; Denorm coefficients */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 6);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl o0		; Vertex pos
+	 * decl o1		; Luma/chroma texcoords
+	 * decl o2		; Top field past ref macroblock texcoords
+	 * decl o3		; Bottom field past ref macroblock texcoords
+	 * decl o4		; Top field future ref macroblock texcoords
+	 * decl o5		; Bottom field future ref macroblock texcoords
+	 * decl o6		; Denormalized vertex pos
+	 */
+	for (i = 0; i < 7; i++)
+	{
+		decl = vl_decl_output((i == 0 || i == 7) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* decl t0, t1 */
+	decl = vl_decl_temps(0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * mov o0, i0		; Move input vertex pos to output
+	 * mov o1, i0		; Move input luma/chroma texcoords to output
+	 * mov o2, i1		; Move past top field texcoords to output
+	 * mov o3, i2		; Move past bottom field texcoords to output
+	 * mov o4, i3		; Move future top field texcoords to output
+	 * mov o5, i4		; Move future bottom field texcoords to output
+	 */
+	for (i = 0; i < 6; ++i)
+	{
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, 0);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * add o2, i0, i1	; Translate vertex pos by motion vec to form first top field macroblock texcoords
+	 * add o3, i0, i2	; Translate vertex pos by motion vec to form first bottom field macroblock texcoords
+	 * add o4, i0, i3	; Translate vertex pos by motion vec to form second top field macroblock texcoords
+	 * add o5, i0, i4	; Translate vertex pos by motion vec to form second bottom field macroblock texcoords
+	 */
+	for (i = 0; i < 4; ++i)
+	{
+		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 2, TGSI_FILE_INPUT, 0, TGSI_FILE_INPUT, i + 1);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* mul o6, i0, c0	; Denorm vertex pos */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 6, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	vs.tokens = tokens;
+	mc->b_vs[1] = pipe->create_vs_state(pipe, &vs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFrameBMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 100;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Texcoords for s0, s1, s2
+	 * decl i1			; Texcoords for s3
+	 * decl i2			; Texcoords for s4
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
+	 * decl c1			; Constant 1/2 in .x channel to use as weight to blend past and future texels
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0-t2 */
+	decl = vl_decl_temps(0, 2);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for past ref surface texture
+	 * decl s4			; Sampler for future ref surface texture
+	 */
+	for (i = 0; i < 5; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i0, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i0, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t1, i1, s3		; Read texel from past ref macroblock
+	 * tex2d t2, i2, s4		; Read texel from future ref macroblock
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 1, TGSI_FILE_SAMPLER, i + 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->b_fs[0] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateFragmentShaderFieldBMB
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int		max_tokens = 200;
+
+	struct pipe_context		*pipe;
+	struct pipe_shader_state	fs;
+	struct tgsi_token		*tokens;
+	struct tgsi_header		*header;
+
+	struct tgsi_full_declaration	decl;
+	struct tgsi_full_instruction	inst;
+
+	unsigned int			ti;
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+
+	/* Version */
+	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
+	/* Header */
+	header = (struct tgsi_header*)&tokens[1];
+	*header = tgsi_build_header();
+	/* Processor */
+	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
+
+	ti = 3;
+
+	/*
+	 * decl i0			; Texcoords for s0, s1, s2
+	 * decl i1			; Texcoords for s3
+	 * decl i2			; Texcoords for s3
+	 * decl i3			; Texcoords for s4
+	 * decl i4			; Texcoords for s4
+	 * decl i5			; Denormalized vertex pos
+	 */
+	for (i = 0; i < 6; ++i)
+	{
+		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
+	 * decl c1			; Constants 1/2 & 2 in .x, .y channels to use as weight to blend past and future texels
+	 *				; and for Y-mod-2 top/bottom field selection
+	 */
+	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl o0			; Fragment color */
+	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/* decl t0-t5 */
+	decl = vl_decl_temps(0, 5);
+	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * decl s0			; Sampler for luma texture
+	 * decl s1			; Sampler for chroma Cb texture
+	 * decl s2			; Sampler for chroma Cr texture
+	 * decl s3			; Sampler for past ref surface texture
+	 * decl s4			; Sampler for future ref surface texture
+	 */
+	for (i = 0; i < 5; ++i)
+	{
+		decl = vl_decl_samplers(i, i);
+		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/*
+	 * tex2d t1, i0, s0		; Read texel from luma texture
+	 * mov t0.x, t1.x		; Move luma sample into .x component
+	 * tex2d t1, i0, s1		; Read texel from chroma Cb texture
+	 * mov t0.y, t1.x		; Move Cb sample into .y component
+	 * tex2d t1, i0, s2		; Read texel from chroma Cr texture
+	 * mov t0.z, t1.x		; Move Cr sample into .z component
+	 */
+	for (i = 0; i < 3; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 0, TGSI_FILE_SAMPLER, i);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	}
+
+	/* mul t0, t0, c0		; Rescale texel to correct range */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* XXX: Pos values off by 0.5? */
+	/* sub t4, i5.y, c1.x		; Sub 0.5 from denormalized pos */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 5, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
+	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* floor t3, t3			; Get rid of fractional part */
+	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* mul t3, t3, c1.y		; Multiply by 2 */
+	inst = vl_inst3( TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
+	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
+	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t1, i1, s3		; Read texel from past ref macroblock top field
+	 * tex2d t2, i2, s3		; Read texel from past ref macroblock bottom field
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 1, TGSI_FILE_SAMPLER, 3);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/*
+	 * tex2d t4, i3, s4		; Read texel from future ref macroblock top field
+	 * tex2d t5, i4, s4		; Read texel from future ref macroblock bottom field
+	 */
+	for (i = 0; i < 2; ++i)
+	{
+		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 4, TGSI_FILE_INPUT, i + 3, TGSI_FILE_SAMPLER, 4);
+		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+	}
+
+	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
+	/* lerp t2, t3, t4, t5		; Choose between top and bottom fields based on Y % 2 */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 2, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 5);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
+	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
+	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
+	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	/* end */
+	inst = vl_end();
+	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
+
+	fs.tokens = tokens;
+	mc->b_fs[1] = pipe->create_fs_state(pipe, &fs);
+	free(tokens);
+
+	return 0;
+}
+
+static int vlCreateDataBufs
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	const unsigned int	mbw = align(mc->video_width, VL_MACROBLOCK_WIDTH) / VL_MACROBLOCK_WIDTH;
+	const unsigned int	mbh = align(mc->video_height, VL_MACROBLOCK_HEIGHT) / VL_MACROBLOCK_HEIGHT;
+	const unsigned int	num_mb_per_frame = mbw * mbh;
+
+	struct pipe_context	*pipe;
+	unsigned int		g, h, i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+
+	for (g = 0; g < NUM_BUF_SETS; ++g)
+	{
+		for (h = 0; h < 7; ++h)
+		{
+			/* Create our vertex buffer and vertex buffer element */
+			mc->vertex_bufs[g][h][0].pitch = sizeof(struct vlVertex2f);
+			mc->vertex_bufs[g][h][0].max_index = 24 * num_mb_per_frame - 1;
+			mc->vertex_bufs[g][h][0].buffer_offset = 0;
+			mc->vertex_bufs[g][h][0].buffer = pipe->winsys->buffer_create
+			(
+				pipe->winsys,
+				1,
+				PIPE_BUFFER_USAGE_VERTEX,
+				sizeof(struct vlVertex2f) * 24 * num_mb_per_frame
+			);
+		}
+	}
+
+	/* Position & block luma, block chroma texcoord element */
+	mc->vertex_elems[0].src_offset = 0;
+	mc->vertex_elems[0].vertex_buffer_index = 0;
+	mc->vertex_elems[0].nr_components = 2;
+	mc->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	for (g = 0; g < NUM_BUF_SETS; ++g)
+	{
+		for (h = 0; h < 7; ++h)
+		{
+			for (i = 1; i < 3; ++i)
+			{
+				mc->vertex_bufs[g][h][i].pitch = sizeof(struct vlVertex2f) * 2;
+				mc->vertex_bufs[g][h][i].max_index = 24 * num_mb_per_frame - 1;
+				mc->vertex_bufs[g][h][i].buffer_offset = 0;
+				mc->vertex_bufs[g][h][i].buffer = pipe->winsys->buffer_create
+				(
+					pipe->winsys,
+					1,
+					PIPE_BUFFER_USAGE_VERTEX,
+					sizeof(struct vlVertex2f) * 2 * 24 * num_mb_per_frame
+				);
+			}
+		}
+	}
+
+	/* First ref surface top field texcoord element */
+	mc->vertex_elems[1].src_offset = 0;
+	mc->vertex_elems[1].vertex_buffer_index = 1;
+	mc->vertex_elems[1].nr_components = 2;
+	mc->vertex_elems[1].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* First ref surface bottom field texcoord element */
+	mc->vertex_elems[2].src_offset = sizeof(struct vlVertex2f);
+	mc->vertex_elems[2].vertex_buffer_index = 1;
+	mc->vertex_elems[2].nr_components = 2;
+	mc->vertex_elems[2].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Second ref surface top field texcoord element */
+	mc->vertex_elems[3].src_offset = 0;
+	mc->vertex_elems[3].vertex_buffer_index = 2;
+	mc->vertex_elems[3].nr_components = 2;
+	mc->vertex_elems[3].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Second ref surface bottom field texcoord element */
+	mc->vertex_elems[4].src_offset = sizeof(struct vlVertex2f);
+	mc->vertex_elems[4].vertex_buffer_index = 2;
+	mc->vertex_elems[4].nr_components = 2;
+	mc->vertex_elems[4].src_format = PIPE_FORMAT_R32G32_FLOAT;
+
+	/* Create our constant buffer */
+	mc->vs_const_buf.size = sizeof(struct vlVertexShaderConsts);
+	mc->vs_const_buf.buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_CONSTANT,
+		mc->vs_const_buf.size
+	);
+
+	mc->fs_const_buf.size = sizeof(struct vlFragmentShaderConsts);
+	mc->fs_const_buf.buffer = pipe->winsys->buffer_create
+	(
+		pipe->winsys,
+		1,
+		PIPE_BUFFER_USAGE_CONSTANT,
+		mc->fs_const_buf.size
+	);
+
+	memcpy
+	(
+		pipe->winsys->buffer_map(pipe->winsys, mc->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
+		&fs_consts,
+		sizeof(struct vlFragmentShaderConsts)
+	);
+
+	pipe->winsys->buffer_unmap(pipe->winsys, mc->fs_const_buf.buffer);
+
+	return 0;
+}
+
+static int vlInit
+(
+	struct vlR16SnormBufferedMC *mc
+)
+{
+	struct pipe_context		*pipe;
+	struct pipe_sampler_state	sampler;
+	struct pipe_texture		template;
+	unsigned int			filters[5];
+	unsigned int			i;
+
+	assert(mc);
+
+	pipe = mc->pipe;
+
+	/* For MC we render to textures, which are rounded up to nearest POT */
+	mc->viewport.scale[0] = vlRoundUpPOT(mc->video_width);
+	mc->viewport.scale[1] = vlRoundUpPOT(mc->video_height);
+	mc->viewport.scale[2] = 1;
+	mc->viewport.scale[3] = 1;
+	mc->viewport.translate[0] = 0;
+	mc->viewport.translate[1] = 0;
+	mc->viewport.translate[2] = 0;
+	mc->viewport.translate[3] = 0;
+
+	mc->render_target.width = vlRoundUpPOT(mc->video_width);
+	mc->render_target.height = vlRoundUpPOT(mc->video_height);
+	mc->render_target.num_cbufs = 1;
+	/* FB for MC stage is a vlSurface created by the user, set at render time */
+	mc->render_target.zsbuf = NULL;
+
+	filters[0] = PIPE_TEX_FILTER_NEAREST;
+	filters[1] = mc->video_format == vlFormatYCbCr444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
+	filters[2] = mc->video_format == vlFormatYCbCr444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
+	filters[3] = PIPE_TEX_FILTER_LINEAR;
+	filters[4] = PIPE_TEX_FILTER_LINEAR;
+
+	for (i = 0; i < 5; ++i)
+	{
+		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_img_filter = filters[i];
+		sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
+		sampler.mag_img_filter = filters[i];
+		sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
+		sampler.compare_func = PIPE_FUNC_ALWAYS;
+		sampler.normalized_coords = 1;
+		/*sampler.prefilter = ;*/
+		/*sampler.shadow_ambient = ;*/
+		/*sampler.lod_bias = ;*/
+		sampler.min_lod = 0;
+		/*sampler.max_lod = ;*/
+		/*sampler.border_color[i] = ;*/
+		/*sampler.max_anisotropy = ;*/
+		mc->samplers[i] = pipe->create_sampler_state(pipe, &sampler);
+	}
+
+	memset(&template, 0, sizeof(struct pipe_texture));
+	template.target = PIPE_TEXTURE_2D;
+	template.format = PIPE_FORMAT_R16_SNORM;
+	template.last_level = 0;
+	template.width[0] = vlRoundUpPOT(mc->video_width);
+	template.height[0] = vlRoundUpPOT(mc->video_height);
+	template.depth[0] = 1;
+	template.compressed = 0;
+	pf_get_block(template.format, &template.block);
+
+	for (i = 0; i < NUM_BUF_SETS; ++i)
+		mc->textures[i][0] = pipe->screen->texture_create(pipe->screen, &template);
+
+	if (mc->video_format == vlFormatYCbCr420)
+	{
+		template.width[0] = vlRoundUpPOT(mc->video_width / 2);
+		template.height[0] = vlRoundUpPOT(mc->video_height / 2);
+	}
+	else if (mc->video_format == vlFormatYCbCr422)
+		template.height[0] = vlRoundUpPOT(mc->video_height / 2);
+
+	for (i = 0; i < NUM_BUF_SETS; ++i)
+	{
+		mc->textures[i][1] = pipe->screen->texture_create(pipe->screen, &template);
+		mc->textures[i][2] = pipe->screen->texture_create(pipe->screen, &template);
+	}
+
+	/* textures[3] & textures[4] are assigned from vlSurfaces for P and B macroblocks at render time */
+
+	vlCreateVertexShaderIMB(mc);
+	vlCreateFragmentShaderIMB(mc);
+	vlCreateVertexShaderFramePMB(mc);
+	vlCreateVertexShaderFieldPMB(mc);
+	vlCreateFragmentShaderFramePMB(mc);
+	vlCreateFragmentShaderFieldPMB(mc);
+	vlCreateVertexShaderFrameBMB(mc);
+	vlCreateVertexShaderFieldBMB(mc);
+	vlCreateFragmentShaderFrameBMB(mc);
+	vlCreateFragmentShaderFieldBMB(mc);
+	vlCreateDataBufs(mc);
+
+	return 0;
+}
+
+int vlCreateR16SNormBufferedMC
+(
+	struct pipe_context *pipe,
+	unsigned int video_width,
+	unsigned int video_height,
+	enum vlFormat video_format,
+	struct vlRender **render
+)
+{
+	struct vlR16SnormBufferedMC *mc;
+
+	assert(pipe);
+	assert(render);
+
+	mc = calloc(1, sizeof(struct vlR16SnormBufferedMC));
+
+	mc->base.vlBegin = &vlBegin;
+	mc->base.vlRenderMacroBlocksMpeg2 = &vlRenderMacroBlocksMpeg2R16SnormBuffered;
+	mc->base.vlEnd = &vlEnd;
+	mc->base.vlFlush = &vlFlush;
+	mc->base.vlDestroy = &vlDestroy;
+	mc->pipe = pipe;
+	mc->video_width = video_width;
+	mc->video_height = video_height;
+
+	mc->cur_buf = 0;
+	mc->buffered_surface = NULL;
+	mc->past_surface = NULL;
+	mc->future_surface = NULL;
+	memset(mc->num_macroblocks, 0, sizeof(unsigned int) * 7);
+	mc->total_num_macroblocks = 0;
+
+	vlInit(mc);
+
+	*render = &mc->base;
+
+	return 0;
+}
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h
new file mode 100644
index 0000000000..30f67db3e7
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.h
@@ -0,0 +1,18 @@
+#ifndef vl_r16snorm_mc_buf_h
+#define vl_r16snorm_mc_buf_h
+
+#include "vl_types.h"
+
+struct pipe_context;
+struct vlRender;
+
+int vlCreateR16SNormBufferedMC
+(
+	struct pipe_context *pipe,
+	unsigned int video_width,
+	unsigned int video_height,
+	enum vlFormat video_format,
+	struct vlRender **render
+);
+
+#endif
diff --git a/src/gallium/state_trackers/g3dvl/vl_render.h b/src/gallium/state_trackers/g3dvl/vl_render.h
index 63016b5cbe..166030b498 100644
--- a/src/gallium/state_trackers/g3dvl/vl_render.h
+++ b/src/gallium/state_trackers/g3dvl/vl_render.h
@@ -24,6 +24,11 @@ struct vlRender
 		struct vlRender *render
 	);
 
+	int (*vlFlush)
+	(
+		struct vlRender *render
+	);
+
 	int (*vlDestroy)
 	(
 		struct vlRender *render
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c
index ffc8122172..687fd1ec29 100644
--- a/src/gallium/state_trackers/g3dvl/vl_surface.c
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.c
@@ -113,6 +113,8 @@ int vlPutPicture
 	assert(surface);
 	assert(surface->context);
 
+	surface->context->render->vlFlush(surface->context->render);
+
 	csc = surface->context->csc;
 	pipe = surface->context->pipe;
 
-- 
cgit v1.2.3


From db1021a37c29a60c70ce294077680ca3e98a6460 Mon Sep 17 00:00:00 2001
From: Younes Manton <younes.m@gmail.com>
Date: Sat, 20 Dec 2008 16:31:29 -0500
Subject: g3dvl: Get rid of old unbuffered motion compensation code.

---
 src/gallium/state_trackers/g3dvl/Makefile         |    2 +-
 src/gallium/state_trackers/g3dvl/vl_context.c     |    2 -
 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c | 2344 ---------------------
 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h |   18 -
 4 files changed, 1 insertion(+), 2365 deletions(-)
 delete mode 100644 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
 delete mode 100644 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h

(limited to 'src/gallium/state_trackers/g3dvl/Makefile')

diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
index 4f7a953484..84a0b2c6d8 100644
--- a/src/gallium/state_trackers/g3dvl/Makefile
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -1,6 +1,6 @@
 TARGET		= libg3dvl.a
 OBJECTS		= vl_display.o vl_screen.o vl_context.o vl_surface.o vl_shader_build.o vl_util.o vl_basic_csc.o	\
-		  vl_r16snorm_mc.o vl_r16snorm_mc_buf.o
+		  vl_r16snorm_mc_buf.o
 GALLIUMDIR	= ../..
 
 CFLAGS		+= -g -Wall -fPIC -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary -I${GALLIUMDIR}/winsys/g3dvl
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c
index fe107e406d..5b7bb73b39 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.c
+++ b/src/gallium/state_trackers/g3dvl/vl_context.c
@@ -5,7 +5,6 @@
 #include <pipe/p_context.h>
 #include <pipe/p_state.h>
 #include "vl_render.h"
-#include "vl_r16snorm_mc.h"
 #include "vl_r16snorm_mc_buf.h"
 #include "vl_csc.h"
 #include "vl_basic_csc.h"
@@ -127,7 +126,6 @@ int vlCreateContext
 
 	vlInitCommon(ctx);
 
-	/*vlCreateR16SNormMC(pipe, picture_width, picture_height, picture_format, &ctx->render);*/
 	vlCreateR16SNormBufferedMC(pipe, picture_width, picture_height, picture_format, &ctx->render);
 	vlCreateBasicCSC(pipe, &ctx->csc);
 
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
deleted file mode 100644
index 3272220ef8..0000000000
--- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.c
+++ /dev/null
@@ -1,2344 +0,0 @@
-#define VL_INTERNAL
-#include "vl_r16snorm_mc.h"
-#include <assert.h>
-#include <stdlib.h>
-#include <pipe/p_context.h>
-#include <pipe/p_winsys.h>
-#include <pipe/p_state.h>
-#include <pipe/p_inlines.h>
-#include <tgsi/tgsi_parse.h>
-#include <tgsi/tgsi_build.h>
-#include "vl_render.h"
-#include "vl_shader_build.h"
-#include "vl_surface.h"
-#include "vl_util.h"
-#include "vl_types.h"
-#include "vl_defs.h"
-
-#define NUM_BUFS 4	/* Number of rotating buffers to use */
-
-struct vlVertexShaderConsts
-{
-	/*struct vlVertex4f scale;
-	struct vlVertex4f denorm;*/
-	struct vlVertex4f	scale;
-	struct vlVertex4f	mb_pos_trans;
-	struct vlVertex4f	denorm;
-	struct
-	{
-		struct vlVertex4f	top_field;
-		struct vlVertex4f	bottom_field;
-	} mb_tc_trans[2];
-};
-
-struct vlFragmentShaderConsts
-{
-	struct vlVertex4f multiplier;
-	struct vlVertex4f div;
-};
-
-struct vlR16SnormMC
-{
-	struct vlRender				base;
-
-	unsigned int				video_width, video_height;
-	enum vlFormat				video_format;
-	unsigned int				cur_buf;
-
-	struct pipe_context			*pipe;
-	struct pipe_viewport_state		viewport;
-	struct pipe_framebuffer_state		render_target;
-	struct pipe_sampler_state		*samplers[5];
-	struct pipe_texture			*textures[NUM_BUFS][5];
-	void					*i_vs, *p_vs[2], *b_vs[2];
-	void					*i_fs, *p_fs[2], *b_fs[2];
-	struct pipe_vertex_buffer 		vertex_bufs[3];
-	struct pipe_vertex_element		vertex_elems[3];
-	struct pipe_constant_buffer		vs_const_buf, fs_const_buf;
-};
-
-static int vlBegin
-(
-	struct vlRender *render
-)
-{
-	struct vlR16SnormMC	*mc;
-	struct pipe_context	*pipe;
-
-	assert(render);
-
-	mc = (struct vlR16SnormMC*)render;
-	pipe = mc->pipe;
-
-	/* Frame buffer set in vlRender*Macroblock() */
-	/* Shaders, samplers, textures set in vlRender*Macroblock() */
-	pipe->set_vertex_buffers(pipe, 3, mc->vertex_bufs);
-	pipe->set_vertex_elements(pipe, 3, mc->vertex_elems);
-	pipe->set_viewport_state(pipe, &mc->viewport);
-	pipe->set_constant_buffer(pipe, PIPE_SHADER_VERTEX, 0, &mc->vs_const_buf);
-	pipe->set_constant_buffer(pipe, PIPE_SHADER_FRAGMENT, 0, &mc->fs_const_buf);
-
-	return 0;
-}
-
-/*static int vlGrabMacroBlock
-(
-	struct vlR16SnormMC *mc,
-	struct vlMpeg2MacroBlock *macroblock
-)
-{
-	assert(mc);
-	assert(macroblock);
-
-
-
-	return 0;
-}*/
-
-/*#define DO_IDCT*/
-
-#ifdef DO_IDCT
-static int vlTransformBlock(short *src, short *dst, short bias)
-{
-	static const float basis[8][8] =
-	{
-		{0.3536,   0.4904,   0.4619,   0.4157,   0.3536,   0.2778,   0.1913,   0.0975},
-		{0.3536,   0.4157,   0.1913,  -0.0975,  -0.3536,  -0.4904,  -0.4619,  -0.2778},
-		{0.3536,   0.2778,  -0.1913,  -0.4904,  -0.3536,   0.0975,   0.4619,   0.4157},
-		{0.3536,   0.0975,  -0.4619,  -0.2778,   0.3536,   0.4157,  -0.1913,  -0.4904},
-		{0.3536,  -0.0975,  -0.4619,   0.2778,   0.3536,  -0.4157,  -0.1913,   0.4904},
-		{0.3536,  -0.2778,  -0.1913,   0.4904,  -0.3536,  -0.0975,   0.4619,  -0.4157},
-		{0.3536,  -0.4157,   0.1913,   0.0975,  -0.3536,   0.4904,  -0.4619,   0.2778},
-		{0.3536,  -0.4904,   0.4619,  -0.4157,   0.3536,  -0.2778,   0.1913,  -0.0975}
-	};
-
-	unsigned int	x, y;
-	short		tmp[64];
-
-	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
-		for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-			tmp[y * VL_BLOCK_WIDTH + x] = (short)
-			(
-				src[y * VL_BLOCK_WIDTH + 0] * basis[x][0] +
-				src[y * VL_BLOCK_WIDTH + 1] * basis[x][1] +
-				src[y * VL_BLOCK_WIDTH + 2] * basis[x][2] +
-				src[y * VL_BLOCK_WIDTH + 3] * basis[x][3] +
-				src[y * VL_BLOCK_WIDTH + 4] * basis[x][4] +
-				src[y * VL_BLOCK_WIDTH + 5] * basis[x][5] +
-				src[y * VL_BLOCK_WIDTH + 6] * basis[x][6] +
-				src[y * VL_BLOCK_WIDTH + 7] * basis[x][7]
-			);
-
-	for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-		for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
-		{
-			dst[y * VL_BLOCK_WIDTH + x] = bias + (short)
-			(
-				tmp[0 * VL_BLOCK_WIDTH + x] * basis[y][0] +
-				tmp[1 * VL_BLOCK_WIDTH + x] * basis[y][1] +
-				tmp[2 * VL_BLOCK_WIDTH + x] * basis[y][2] +
-				tmp[3 * VL_BLOCK_WIDTH + x] * basis[y][3] +
-				tmp[4 * VL_BLOCK_WIDTH + x] * basis[y][4] +
-				tmp[5 * VL_BLOCK_WIDTH + x] * basis[y][5] +
-				tmp[6 * VL_BLOCK_WIDTH + x] * basis[y][6] +
-				tmp[7 * VL_BLOCK_WIDTH + x] * basis[y][7]
-			);
-			if (dst[y * VL_BLOCK_WIDTH + x] > 255)
-				dst[y * VL_BLOCK_WIDTH + x] = 255;
-			else if (bias > 0 && dst[y * VL_BLOCK_WIDTH + x] < 0)
-				dst[y * VL_BLOCK_WIDTH + x] = 0;
-		}
-	return 0;
-}
-#endif
-
-static int vlGrabFrameCodedBlock(short *src, short *dst, unsigned int dst_pitch)
-{
-	unsigned int y;
-
-	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
-		memcpy
-		(
-			dst + y * dst_pitch,
-			src + y * VL_BLOCK_WIDTH,
-			VL_BLOCK_WIDTH * 2
-		);
-
-	return 0;
-}
-
-static int vlGrabFieldCodedBlock(short *src, short *dst, unsigned int dst_pitch)
-{
-	unsigned int y;
-
-	for (y = 0; y < VL_BLOCK_HEIGHT / 2; ++y)
-		memcpy
-		(
-			dst + y * dst_pitch * 2,
-			src + y * VL_BLOCK_WIDTH,
-			VL_BLOCK_WIDTH * 2
-		);
-
-	dst += VL_BLOCK_HEIGHT * dst_pitch;
-
-	for (; y < VL_BLOCK_HEIGHT; ++y)
-		memcpy
-		(
-			dst + y * dst_pitch * 2,
-			src + y * VL_BLOCK_WIDTH,
-			VL_BLOCK_WIDTH * 2
-		);
-
-	return 0;
-}
-
-static int vlGrabNoBlock(short *dst, unsigned int dst_pitch)
-{
-	unsigned int y;
-
-	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
-		memset
-		(
-			dst + y * dst_pitch,
-			0,
-			VL_BLOCK_WIDTH * 2
-		);
-
-	return 0;
-}
-
-enum vlSampleType
-{
-	vlSampleTypeFull,
-	vlSampleTypeDiff
-};
-
-static int vlGrabBlocks
-(
-	struct vlR16SnormMC *mc,
-	unsigned int coded_block_pattern,
-	enum vlDCTType dct_type,
-	enum vlSampleType sample_type,
-	short *blocks
-)
-{
-	struct pipe_surface	*tex_surface;
-	short			*texels;
-	unsigned int		tex_pitch;
-	unsigned int		tb, sb = 0;
-
-	assert(mc);
-	assert(blocks);
-
-	tex_surface = mc->pipe->screen->get_tex_surface
-	(
-		mc->pipe->screen,
-		mc->textures[mc->cur_buf % NUM_BUFS][0],
-		0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
-	);
-
-	texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE);
-	tex_pitch = tex_surface->stride / tex_surface->block.size;
-
-	for (tb = 0; tb < 4; ++tb)
-	{
-		if ((coded_block_pattern >> (5 - tb)) & 1)
-		{
-			short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
-
-#ifdef DO_IDCT
-			vlTransformBlock(cur_block, cur_block, sample_type == vlSampleTypeFull ? 128 : 0);
-#endif
-
-			if (dct_type == vlDCTTypeFrameCoded)
-				vlGrabFrameCodedBlock
-				(
-					cur_block,
-					texels + tb * tex_pitch * VL_BLOCK_HEIGHT,
-					tex_pitch
-				);
-			else
-				vlGrabFieldCodedBlock
-				(
-					cur_block,
-					texels + (tb % 2) * tex_pitch * VL_BLOCK_HEIGHT + (tb / 2) * tex_pitch,
-					tex_pitch
-				);
-
-			++sb;
-		}
-		else
-			vlGrabNoBlock(texels + tb * tex_pitch * VL_BLOCK_HEIGHT, tex_pitch);
-	}
-
-	pipe_surface_unmap(tex_surface);
-
-	/* TODO: Implement 422, 444 */
-	for (tb = 0; tb < 2; ++tb)
-	{
-		tex_surface = mc->pipe->screen->get_tex_surface
-		(
-			mc->pipe->screen,
-			mc->textures[mc->cur_buf % NUM_BUFS][tb + 1],
-			0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
-		);
-
-		texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE);
-		tex_pitch = tex_surface->stride / tex_surface->block.size;
-
-		if ((coded_block_pattern >> (1 - tb)) & 1)
-		{
-			short *cur_block = blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT;
-
-#ifdef DO_IDCT
-			vlTransformBlock(cur_block, cur_block, sample_type == vlSampleTypeFull ? 128 : 0);
-#endif
-
-			vlGrabFrameCodedBlock
-			(
-				cur_block,
-				texels,
-				tex_pitch
-			);
-
-			++sb;
-		}
-		else
-			vlGrabNoBlock(texels, tex_pitch);
-
-		pipe_surface_unmap(tex_surface);
-	}
-
-	return 0;
-}
-
-static int vlRenderIMacroBlock
-(
-	struct vlR16SnormMC *mc,
-	enum vlPictureType picture_type,
-	enum vlFieldOrder field_order,
-	unsigned int mbx,
-	unsigned int mby,
-	unsigned int coded_block_pattern,
-	enum vlDCTType dct_type,
-	short *blocks,
-	struct vlSurface *surface
-)
-{
-	struct pipe_context		*pipe;
-	struct vlVertexShaderConsts	*vs_consts;
-
-	assert(blocks);
-	assert(surface);
-
-	/* TODO: Implement interlaced rendering */
-	if (picture_type != vlPictureTypeFrame)
-		return 0;
-
-	vlGrabBlocks(mc, coded_block_pattern, dct_type, vlSampleTypeFull, blocks);
-
-	pipe = mc->pipe;
-
-	vs_consts = pipe->winsys->buffer_map
-	(
-		pipe->winsys,
-		mc->vs_const_buf.buffer,
-		PIPE_BUFFER_USAGE_CPU_WRITE
-	);
-
-	vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->texture->width[0];
-	vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->texture->height[0];
-	vs_consts->scale.z = 1.0f;
-	vs_consts->scale.w = 1.0f;
-	vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->texture->width[0];
-	vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->texture->height[0];
-	vs_consts->mb_pos_trans.z = 0.0f;
-	vs_consts->mb_pos_trans.w = 0.0f;
-
-	pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer);
-
-	mc->render_target.cbufs[0] = pipe->screen->get_tex_surface
-	(
-		pipe->screen,
-		surface->texture,
-		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
-	);
-	pipe->set_framebuffer_state(pipe, &mc->render_target);
-	pipe->set_sampler_textures(pipe, 3, mc->textures[mc->cur_buf % NUM_BUFS]);
-	pipe->bind_sampler_states(pipe, 3, (void**)mc->samplers);
-	pipe->bind_vs_state(pipe, mc->i_vs);
-	pipe->bind_fs_state(pipe, mc->i_fs);
-
-	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
-
-	mc->cur_buf++;
-
-	return 0;
-}
-
-static int vlRenderPMacroBlock
-(
-	struct vlR16SnormMC *mc,
-	enum vlPictureType picture_type,
-	enum vlFieldOrder field_order,
-	unsigned int mbx,
-	unsigned int mby,
-	enum vlMotionType mc_type,
-	short top_x,
-	short top_y,
-	short bottom_x,
-	short bottom_y,
-	unsigned int coded_block_pattern,
-	enum vlDCTType dct_type,
-	short *blocks,
-	struct vlSurface *ref_surface,
-	struct vlSurface *surface
-)
-{
-	struct pipe_context		*pipe;
-	struct vlVertexShaderConsts	*vs_consts;
-
-	assert(motion_vectors);
-	assert(blocks);
-	assert(ref_surface);
-	assert(surface);
-
-	/* TODO: Implement interlaced rendering */
-	if (picture_type != vlPictureTypeFrame)
-		return 0;
-	/* TODO: Implement other MC types */
-	if (mc_type != vlMotionTypeFrame && mc_type != vlMotionTypeField)
-		return 0;
-
-	vlGrabBlocks(mc, coded_block_pattern, dct_type, vlSampleTypeDiff, blocks);
-
-	pipe = mc->pipe;
-
-	vs_consts = pipe->winsys->buffer_map
-	(
-		pipe->winsys,
-		mc->vs_const_buf.buffer,
-		PIPE_BUFFER_USAGE_CPU_WRITE
-	);
-
-	vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->texture->width[0];
-	vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->texture->height[0];
-	vs_consts->scale.z = 1.0f;
-	vs_consts->scale.w = 1.0f;
-	vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->texture->width[0];
-	vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->texture->height[0];
-	vs_consts->mb_pos_trans.z = 0.0f;
-	vs_consts->mb_pos_trans.w = 0.0f;
-	vs_consts->mb_tc_trans[0].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + top_x * 0.5f) / (float)surface->texture->width[0];
-	vs_consts->mb_tc_trans[0].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + top_y * 0.5f) / (float)surface->texture->height[0];
-	vs_consts->mb_tc_trans[0].top_field.z = 0.0f;
-	vs_consts->mb_tc_trans[0].top_field.w = 0.0f;
-
-	if (mc_type == vlMotionTypeField)
-	{
-		vs_consts->denorm.x = (float)surface->texture->width[0];
-		vs_consts->denorm.y = (float)surface->texture->height[0];
-
-		vs_consts->mb_tc_trans[0].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + bottom_x * 0.5f) / (float)surface->texture->width[0];
-		vs_consts->mb_tc_trans[0].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + bottom_y * 0.5f) / (float)surface->texture->height[0];
-		vs_consts->mb_tc_trans[0].bottom_field.z = 0.0f;
-		vs_consts->mb_tc_trans[0].bottom_field.w = 0.0f;
-
-		pipe->bind_vs_state(pipe, mc->p_vs[1]);
-		pipe->bind_fs_state(pipe, mc->p_fs[1]);
-	}
-	else
-	{
-		pipe->bind_vs_state(pipe, mc->p_vs[0]);
-		pipe->bind_fs_state(pipe, mc->p_fs[0]);
-	}
-
-	pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer);
-
-	mc->render_target.cbufs[0] = pipe->screen->get_tex_surface
-	(
-		pipe->screen,
-		surface->texture,
-		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
-	);
-	pipe->set_framebuffer_state(pipe, &mc->render_target);
-
-	mc->textures[mc->cur_buf % NUM_BUFS][3] = ref_surface->texture;
-	pipe->set_sampler_textures(pipe, 4, mc->textures[mc->cur_buf % NUM_BUFS]);
-	pipe->bind_sampler_states(pipe, 4, (void**)mc->samplers);
-
-	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
-
-	mc->cur_buf++;
-
-	return 0;
-}
-
-static int vlRenderBMacroBlock
-(
-	struct vlR16SnormMC *mc,
-	enum vlPictureType picture_type,
-	enum vlFieldOrder field_order,
-	unsigned int mbx,
-	unsigned int mby,
-	enum vlMotionType mc_type,
-	short top_past_x,
-	short top_past_y,
-	short bottom_past_x,
-	short bottom_past_y,
-	short top_future_x,
-	short top_future_y,
-	short bottom_future_x,
-	short bottom_future_y,
-	unsigned int coded_block_pattern,
-	enum vlDCTType dct_type,
-	short *blocks,
-	struct vlSurface *past_surface,
-	struct vlSurface *future_surface,
-	struct vlSurface *surface
-)
-{
-	struct pipe_context		*pipe;
-	struct vlVertexShaderConsts	*vs_consts;
-
-	assert(motion_vectors);
-	assert(blocks);
-	assert(ref_surface);
-	assert(surface);
-
-	/* TODO: Implement interlaced rendering */
-	if (picture_type != vlPictureTypeFrame)
-		return 0;
-	/* TODO: Implement other MC types */
-	if (mc_type != vlMotionTypeFrame && mc_type != vlMotionTypeField)
-		return 0;
-
-	vlGrabBlocks(mc, coded_block_pattern, dct_type, vlSampleTypeDiff, blocks);
-
-	pipe = mc->pipe;
-
-	vs_consts = pipe->winsys->buffer_map
-	(
-		pipe->winsys,
-		mc->vs_const_buf.buffer,
-		PIPE_BUFFER_USAGE_CPU_WRITE
-	);
-
-	vs_consts->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->texture->width[0];
-	vs_consts->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->texture->height[0];
-	vs_consts->scale.z = 1.0f;
-	vs_consts->scale.w = 1.0f;
-	vs_consts->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->texture->width[0];
-	vs_consts->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->texture->height[0];
-	vs_consts->mb_pos_trans.z = 0.0f;
-	vs_consts->mb_pos_trans.w = 0.0f;
-	vs_consts->mb_tc_trans[0].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + top_past_x * 0.5f) / (float)surface->texture->width[0];
-	vs_consts->mb_tc_trans[0].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + top_past_y * 0.5f) / (float)surface->texture->height[0];
-	vs_consts->mb_tc_trans[0].top_field.z = 0.0f;
-	vs_consts->mb_tc_trans[0].top_field.w = 0.0f;
-	vs_consts->mb_tc_trans[1].top_field.x = (mbx * VL_MACROBLOCK_WIDTH + top_future_x * 0.5f) / (float)surface->texture->width[0];
-	vs_consts->mb_tc_trans[1].top_field.y = (mby * VL_MACROBLOCK_HEIGHT + top_future_y * 0.5f) / (float)surface->texture->height[0];
-	vs_consts->mb_tc_trans[1].top_field.z = 0.0f;
-	vs_consts->mb_tc_trans[1].top_field.w = 0.0f;
-
-	if (mc_type == vlMotionTypeField)
-	{
-		vs_consts->denorm.x = (float)surface->texture->width[0];
-		vs_consts->denorm.y = (float)surface->texture->height[0];
-
-		vs_consts->mb_tc_trans[0].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + bottom_past_x * 0.5f) / (float)surface->texture->width[0];
-		vs_consts->mb_tc_trans[0].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + bottom_past_y * 0.5f) / (float)surface->texture->height[0];
-		vs_consts->mb_tc_trans[0].bottom_field.z = 0.0f;
-		vs_consts->mb_tc_trans[0].bottom_field.w = 0.0f;
-		vs_consts->mb_tc_trans[1].bottom_field.x = (mbx * VL_MACROBLOCK_WIDTH + bottom_future_x * 0.5f) / (float)surface->texture->width[0];
-		vs_consts->mb_tc_trans[1].bottom_field.y = (mby * VL_MACROBLOCK_HEIGHT + bottom_future_y * 0.5f) / (float)surface->texture->height[0];
-		vs_consts->mb_tc_trans[1].bottom_field.z = 0.0f;
-		vs_consts->mb_tc_trans[1].bottom_field.w = 0.0f;
-
-		pipe->bind_vs_state(pipe, mc->b_vs[1]);
-		pipe->bind_fs_state(pipe, mc->b_fs[1]);
-	}
-	else
-	{
-		pipe->bind_vs_state(pipe, mc->b_vs[0]);
-		pipe->bind_fs_state(pipe, mc->b_fs[0]);
-	}
-
-	pipe->winsys->buffer_unmap(pipe->winsys, mc->vs_const_buf.buffer);
-
-	mc->render_target.cbufs[0] = pipe->screen->get_tex_surface
-	(
-		pipe->screen,
-		surface->texture,
-		0, 0, 0, PIPE_BUFFER_USAGE_GPU_READ | PIPE_BUFFER_USAGE_GPU_WRITE
-	);
-	pipe->set_framebuffer_state(pipe, &mc->render_target);
-
-	mc->textures[mc->cur_buf % NUM_BUFS][3] = past_surface->texture;
-	mc->textures[mc->cur_buf % NUM_BUFS][4] = future_surface->texture;
-	pipe->set_sampler_textures(pipe, 5, mc->textures[mc->cur_buf % NUM_BUFS]);
-	pipe->bind_sampler_states(pipe, 5, (void**)mc->samplers);
-
-	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
-
-	mc->cur_buf++;
-
-	return 0;
-}
-
-static int vlRenderMacroBlocksMpeg2R16Snorm
-(
-	struct vlRender *render,
-	struct vlMpeg2MacroBlockBatch *batch,
-	struct vlSurface *surface
-)
-{
-	struct vlR16SnormMC	*mc;
-	unsigned int		i;
-
-	assert(render);
-
-	mc = (struct vlR16SnormMC*)render;
-
-	/*for (i = 0; i < batch->num_macroblocks; ++i)
-		vlGrabMacroBlock(batch->macroblocks[i]);*/
-
-	for (i = 0; i < batch->num_macroblocks; ++i)
-	{
-		switch (batch->macroblocks[i].mb_type)
-		{
-			case vlMacroBlockTypeIntra:
-			{
-				vlRenderIMacroBlock
-				(
-					mc,
-					batch->picture_type,
-					batch->field_order,
-					batch->macroblocks[i].mbx,
-					batch->macroblocks[i].mby,
-					batch->macroblocks[i].cbp,
-					batch->macroblocks[i].dct_type,
-					batch->macroblocks[i].blocks,
-					surface
-				);
-				break;
-			}
-			case vlMacroBlockTypeFwdPredicted:
-			{
-				vlRenderPMacroBlock
-				(
-					mc,
-					batch->picture_type,
-					batch->field_order,
-					batch->macroblocks[i].mbx,
-					batch->macroblocks[i].mby,
-					batch->macroblocks[i].mo_type,
-					batch->macroblocks[i].PMV[0][0][0],
-					batch->macroblocks[i].PMV[0][0][1],
-					batch->macroblocks[i].PMV[1][0][0],
-					batch->macroblocks[i].PMV[1][0][1],
-					batch->macroblocks[i].cbp,
-					batch->macroblocks[i].dct_type,
-					batch->macroblocks[i].blocks,
-					batch->past_surface,
-					surface
-				);
-				break;
-			}
-			case vlMacroBlockTypeBkwdPredicted:
-			{
-				vlRenderPMacroBlock
-				(
-					mc,
-					batch->picture_type,
-					batch->field_order,
-					batch->macroblocks[i].mbx,
-					batch->macroblocks[i].mby,
-					batch->macroblocks[i].mo_type,
-					batch->macroblocks[i].PMV[0][1][0],
-					batch->macroblocks[i].PMV[0][1][1],
-					batch->macroblocks[i].PMV[1][1][0],
-					batch->macroblocks[i].PMV[1][1][1],
-					batch->macroblocks[i].cbp,
-					batch->macroblocks[i].dct_type,
-					batch->macroblocks[i].blocks,
-					batch->future_surface,
-					surface
-				);
-				break;
-			}
-			case vlMacroBlockTypeBiPredicted:
-			{
-				vlRenderBMacroBlock
-				(
-					mc,
-					batch->picture_type,
-					batch->field_order,
-					batch->macroblocks[i].mbx,
-					batch->macroblocks[i].mby,
-					batch->macroblocks[i].mo_type,
-					batch->macroblocks[i].PMV[0][0][0],
-					batch->macroblocks[i].PMV[0][0][1],
-					batch->macroblocks[i].PMV[1][0][0],
-					batch->macroblocks[i].PMV[1][0][1],
-					batch->macroblocks[i].PMV[0][1][0],
-					batch->macroblocks[i].PMV[0][1][1],
-					batch->macroblocks[i].PMV[1][1][0],
-					batch->macroblocks[i].PMV[1][1][1],
-					batch->macroblocks[i].cbp,
-					batch->macroblocks[i].dct_type,
-					batch->macroblocks[i].blocks,
-					batch->past_surface,
-					batch->future_surface,
-					surface
-				);
-				break;
-			}
-			default:
-				assert(0);
-		}
-	}
-
-	return 0;
-}
-
-static int vlEnd
-(
-	struct vlRender *render
-)
-{
-	assert(render);
-
-	return 0;
-}
-
-static int vlFlush
-(
-	struct vlRender *render
-)
-{
-	assert(render);
-
-	return 0;
-}
-
-static int vlDestroy
-(
-	struct vlRender *render
-)
-{
-	struct vlR16SnormMC	*mc;
-	struct pipe_context	*pipe;
-	unsigned int		i;
-
-	assert(render);
-
-	mc = (struct vlR16SnormMC*)render;
-	pipe = mc->pipe;
-
-	for (i = 0; i < 5; ++i)
-		pipe->delete_sampler_state(pipe, mc->samplers[i]);
-
-	for (i = 0; i < 3; ++i)
-		pipe->winsys->buffer_destroy(pipe->winsys, mc->vertex_bufs[i].buffer);
-
-	/* Textures 3 & 4 are not created directly, no need to release them here */
-	for (i = 0; i < NUM_BUFS; ++i)
-	{
-		pipe_texture_release(&mc->textures[i][0]);
-		pipe_texture_release(&mc->textures[i][1]);
-		pipe_texture_release(&mc->textures[i][2]);
-	}
-
-	pipe->delete_vs_state(pipe, mc->i_vs);
-	pipe->delete_fs_state(pipe, mc->i_fs);
-
-	for (i = 0; i < 2; ++i)
-	{
-		pipe->delete_vs_state(pipe, mc->p_vs[i]);
-		pipe->delete_fs_state(pipe, mc->p_fs[i]);
-		pipe->delete_vs_state(pipe, mc->b_vs[i]);
-		pipe->delete_fs_state(pipe, mc->b_fs[i]);
-	}
-
-	pipe->winsys->buffer_destroy(pipe->winsys, mc->vs_const_buf.buffer);
-	pipe->winsys->buffer_destroy(pipe->winsys, mc->fs_const_buf.buffer);
-
-	free(mc);
-
-	return 0;
-}
-
-/*
- * Represents 8 triangles (4 quads, 1 per block) in noormalized coords
- * that render a macroblock.
- * Need to be scaled to cover mbW*mbH macroblock pixels and translated into
- * position on target surface.
- */
-static const struct vlVertex2f macroblock_verts[24] =
-{
-	{0.0f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.0f},
-	{0.5f, 0.0f}, {0.0f, 0.5f}, {0.5f, 0.5f},
-
-	{0.5f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.0f},
-	{1.0f, 0.0f}, {0.5f, 0.5f}, {1.0f, 0.5f},
-
-	{0.0f, 0.5f}, {0.0f, 1.0f}, {0.5f, 0.5f},
-	{0.5f, 0.5f}, {0.0f, 1.0f}, {0.5f, 1.0f},
-
-	{0.5f, 0.5f}, {0.5f, 1.0f}, {1.0f, 0.5f},
-	{1.0f, 0.5f}, {0.5f, 1.0f}, {1.0f, 1.0f}
-};
-
-/*
- * Represents texcoords for the above for rendering 4 luma blocks arranged
- * in a bW*(bH*4) texture. First luma block located at 0,0->bW,bH; second at
- * 0,bH->bW,2bH; third at 0,2bH->bW,3bH; fourth at 0,3bH->bW,4bH.
- */
-static const struct vlVertex2f macroblock_luma_texcoords[24] =
-{
-	{0.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.0f},
-	{1.0f, 0.0f}, {0.0f, 0.25f}, {1.0f, 0.25f},
-
-	{0.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.25f},
-	{1.0f, 0.25f}, {0.0f, 0.5f}, {1.0f, 0.5f},
-
-	{0.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.5f},
-	{1.0f, 0.5f}, {0.0f, 0.75f}, {1.0f, 0.75f},
-
-	{0.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 0.75f},
-	{1.0f, 0.75f}, {0.0f, 1.0f}, {1.0f, 1.0f}
-};
-
-/*
- * Represents texcoords for the above for rendering 1 chroma block.
- * Straight forward 0,0->1,1 mapping so we can reuse the MB pos vectors.
- */
-static const struct vlVertex2f *macroblock_chroma_420_texcoords = macroblock_verts;
-
-/*
- * Represents texcoords for the above for rendering 2 chroma blocks arranged
- * in a bW*(bH*2) texture. First chroma block located at 0,0->bW,bH; second at
- * 0,bH->bW,2bH. We can render this with 0,0->1,1 mapping.
- * Straight forward 0,0->1,1 mapping so we can reuse MB pos vectors.
- */
-static const struct vlVertex2f *macroblock_chroma_422_texcoords = macroblock_verts;
-
-/*
- * Represents texcoords for the above for rendering 4 chroma blocks.
- * Same case as 4 luma blocks.
- */
-static const struct vlVertex2f *macroblock_chroma_444_texcoords = macroblock_luma_texcoords;
-
-/*
- * Used when rendering P and B macroblocks, multiplier is applied to the A channel,
- * which is then added to the L channel, then the bias is subtracted from that to
- * get back the differential. The differential is then added to the samples from the
- * reference surface(s).
- */
-static const struct vlFragmentShaderConsts fs_consts =
-{
-	{32767.0f / 255.0f, 32767.0f / 255.0f, 32767.0f / 255.0f, 0.0f},
-	{0.5f, 2.0f, 0.0f, 0.0f}
-};
-
-static int vlCreateVertexShaderIMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 50;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* decl t0 */
-	decl = vl_decl_temps(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * mov o1, i1		; Move input luma texcoords to output
-	 * mov o2, i2		; Move input chroma texcoords to output
-	 */
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	vs.tokens = tokens;
-	mc->i_vs = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateFragmentShaderIMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 100;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl t0, t1 */
-	decl = vl_decl_temps(0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header,max_tokens - ti);
-	}
-
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	}
-
-	/* mul o0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	mc->i_fs = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateVertexShaderFramePMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 100;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 * decl c2		; Unused
-	 * decl c3		; Translation vector to move ref macroblock texcoords into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 3);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 * decl o3		; Ref macroblock texcoords
-	 */
-	for (i = 0; i < 4; i++)
-	{
-		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* decl t0 */
-	decl = vl_decl_temps(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * mov o1, i1		; Move input luma texcoords to output
-	 * mov o2, i2		; Move input chroma texcoords to output
-	 */
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* add o3, t0, c3	; Translate rect into position on ref macroblock */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	vs.tokens = tokens;
-	mc->p_vs[0] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateVertexShaderFieldPMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 100;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration
-		(
-			&decl,
-			&tokens[ti],
-			header,
-			max_tokens - ti
-		);
-	}
-
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 * decl c2		; Denorm coefficients
-	 * decl c3		; Translation vector to move top field ref macroblock texcoords into position
-	 * decl c4		; Translation vector to move bottom field ref macroblock texcoords into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 4);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 * decl o3		; Top field ref macroblock texcoords
-	 * decl o4		; Bottom field ref macroblock texcoords
-	 * decl o5		; Denormalized vertex pos
-	 */
-	for (i = 0; i < 6; i++)
-	{
-		decl = vl_decl_output((i == 0 || i == 5) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* decl t0, t1 */
-	decl = vl_decl_temps(0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add t1, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* mov o0, t1		; Move vertex pos to output */
-	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	mov o1, i1		; Move input luma texcoords to output
-	mov o2, i2		; Move input chroma texcoords to output
-	*/
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* add o3, t0, c3	; Translate top field rect into position on ref macroblock
-	   add o4, t0, c4	; Translate bottom field rect into position on ref macroblock */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* mul o5, t1, c2	; Denorm vertex pos */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 5, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	vs.tokens = tokens;
-	mc->p_vs[1] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateFragmentShaderFramePMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 100;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 * decl i2			; Texcoords for s3
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl t0, t1 */
-	decl = vl_decl_temps(0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 * decl s3			; Sampler for ref surface texture
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	}
-
-	/* mul t0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* tex2d t1, i2, s3		; Read texel from ref macroblock */
-	inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, 2, TGSI_FILE_SAMPLER, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, t1		; Add ref and differential to form final output */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	mc->p_fs[0] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateFragmentShaderFieldPMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 200;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 * decl i2			; Texcoords for s3
-	 * decl i3			; Texcoords for s3
-	 * decl i4			; Denormalized vertex pos
-	 */
-	for (i = 0; i < 5; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
-	 * decl c1			; Constants 1/2 & 2 in .x, .y channels for Y-mod-2 top/bottom field selection
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl t0-t4 */
-	decl = vl_decl_temps(0, 4);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 * decl s3			; Sampler for ref surface texture
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	}
-
-	/* mul t0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * tex2d t1, i2, s3		; Read texel from ref macroblock top field
-	 * tex2d t2, i3, s3		; Read texel from ref macroblock bottom field
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* XXX: Pos values off by 0.5? */
-	/* sub t4, i4.y, c1.x		; Sub 0.5 from denormalized pos */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 4, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* floor t3, t3			; Get rid of fractional part */
-	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* mul t3, t3, c1.y		; Multiply by 2 */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
-	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, t1		; Add ref and differential to form final output */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	mc->p_fs[1] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateVertexShaderFrameBMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 100;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 * decl c2		; Unused
-	 * decl c3		; Translation vector to move past ref macroblock texcoords into position
-	 * decl c4		; Unused
-	 * decl c5		; Translation vector to move future ref macroblock texcoords into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 5);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 * decl o3		; Past ref macroblock texcoords
-	 * decl o4		; Future ref macroblock texcoords
-	 */
-	for (i = 0; i < 5; i++)
-	{
-		decl = vl_decl_output(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* decl t0 */
-	decl = vl_decl_temps(0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * mov o1, i1		; Move input luma texcoords to output
-	 * mov o2, i2		; Move input chroma texcoords to output
-	 */
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* add o3, t0, c3	; Translate rect into position on past ref macroblock
-	   add o4, t0, c5	; Translate rect into position on future ref macroblock */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i * 2 + 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	vs.tokens = tokens;
-	mc->b_vs[0] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateVertexShaderFieldBMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 100;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	vs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_VERTEX, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0		; Vertex pos
-	 * decl i1		; Luma texcoords
-	 * decl i2		; Chroma texcoords
-	 */
-	for (i = 0; i < 3; i++)
-	{
-		decl = vl_decl_input(i == 0 ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * decl c0		; Scaling vector to scale unit rect to macroblock size
-	 * decl c1		; Translation vector to move macroblock into position
-	 * decl c2		; Denorm coefficients
-	 * decl c3		; Translation vector to move top field past ref macroblock texcoords into position
-	 * decl c4		; Translation vector to move bottom field past ref macroblock texcoords into position
-	 * decl c5		; Translation vector to move top field future ref macroblock texcoords into position
-	 * decl c6		; Translation vector to move bottom field future ref macroblock texcoords into position
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 6);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl o0		; Vertex pos
-	 * decl o1		; Luma texcoords
-	 * decl o2		; Chroma texcoords
-	 * decl o3		; Top field past ref macroblock texcoords
-	 * decl o4		; Bottom field past ref macroblock texcoords
-	 * decl o5		; Top field future ref macroblock texcoords
-	 * decl o6		; Bottom field future ref macroblock texcoords
-	 * decl o7		; Denormalized vertex pos
-	 */
-	for (i = 0; i < 8; i++)
-	{
-		decl = vl_decl_output((i == 0 || i == 7) ? TGSI_SEMANTIC_POSITION : TGSI_SEMANTIC_GENERIC, i, i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* decl t0, t1 */
-	decl = vl_decl_temps(0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* mul t0, i0, c0	; Scale unit rect to normalized MB size */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_INPUT, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add t1, t0, c1	; Translate rect into position */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* mov o0, t1		; Move vertex pos to output */
-	inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * mov o1, i1		; Move input luma texcoords to output
-	 * mov o2, i2		; Move input chroma texcoords to output
-	 */
-	for (i = 1; i < 3; ++i)
-	{
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_OUTPUT, i, TGSI_FILE_INPUT, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * add o3, t0, c3	; Translate top field rect into position on past ref macroblock
-	 * add o4, t0, c4	; Translate bottom field rect into position on past ref macroblock
-	 * add o5, t0, c5	; Translate top field rect into position on future ref macroblock
-	 * add o6, t0, c6	; Translate bottom field rect into position on future ref macroblock
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, i + 3, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, i + 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* mul o7, t1, c2	; Denorm vertex pos */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_OUTPUT, 7, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 2);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	vs.tokens = tokens;
-	mc->b_vs[1] = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateFragmentShaderFrameBMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 100;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 * decl i2			; Texcoords for s3
-	 * decl i3			; Texcoords for s4
-	 */
-	for (i = 0; i < 4; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
-	 * decl c1			; Constant 1/2 in .x channel to use as weight to blend past and future texels
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl t0-t2 */
-	decl = vl_decl_temps(0, 2);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 * decl s3			; Sampler for past ref surface texture
-	 * decl s4			; Sampler for future ref surface texture
-	 */
-	for (i = 0; i < 5; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	}
-
-	/* mul t0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * tex2d t1, i2, s3		; Read texel from past ref macroblock
-	 * tex2d t2, i3, s4		; Read texel from future ref macroblock
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, i + 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	mc->b_fs[0] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateFragmentShaderFieldBMB
-(
-	struct vlR16SnormMC *mc
-)
-{
-	const unsigned int		max_tokens = 200;
-
-	struct pipe_context		*pipe;
-	struct pipe_shader_state	fs;
-	struct tgsi_token		*tokens;
-	struct tgsi_header		*header;
-
-	struct tgsi_full_declaration	decl;
-	struct tgsi_full_instruction	inst;
-
-	unsigned int			ti;
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
-
-	/* Version */
-	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
-	/* Header */
-	header = (struct tgsi_header*)&tokens[1];
-	*header = tgsi_build_header();
-	/* Processor */
-	*(struct tgsi_processor*)&tokens[2] = tgsi_build_processor(TGSI_PROCESSOR_FRAGMENT, header);
-
-	ti = 3;
-
-	/*
-	 * decl i0			; Texcoords for s0
-	 * decl i1			; Texcoords for s1, s2
-	 * decl i2			; Texcoords for s3
-	 * decl i3			; Texcoords for s3
-	 * decl i4			; Texcoords for s4
-	 * decl i5			; Texcoords for s4
-	 * decl i6			; Denormalized vertex pos
-	 */
-	for (i = 0; i < 7; ++i)
-	{
-		decl = vl_decl_interpolated_input(TGSI_SEMANTIC_GENERIC, i + 1, i, i, TGSI_INTERPOLATE_LINEAR);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * decl c0			; Scaling factor, rescales 16-bit snorm to 9-bit snorm
-	 * decl c1			; Constants 1/2 & 2 in .x, .y channels to use as weight to blend past and future texels
-	 *				; and for Y-mod-2 top/bottom field selection
-	 */
-	decl = vl_decl_constants(TGSI_SEMANTIC_GENERIC, 0, 0, 1);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl o0			; Fragment color */
-	decl = vl_decl_output(TGSI_SEMANTIC_COLOR, 0, 0, 0);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/* decl t0-t5 */
-	decl = vl_decl_temps(0, 5);
-	ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * decl s0			; Sampler for luma texture
-	 * decl s1			; Sampler for chroma Cb texture
-	 * decl s2			; Sampler for chroma Cr texture
-	 * decl s3			; Sampler for past ref surface texture
-	 * decl s4			; Sampler for future ref surface texture
-	 */
-	for (i = 0; i < 5; ++i)
-	{
-		decl = vl_decl_samplers(i, i);
-		ti += tgsi_build_full_declaration(&decl, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/*
-	 * tex2d t1, i0, s0		; Read texel from luma texture
-	 * mov t0.x, t1.x		; Move luma sample into .x component
-	 * tex2d t1, i1, s1		; Read texel from chroma Cb texture
-	 * mov t0.y, t1.x		; Move Cb sample into .y component
-	 * tex2d t1, i1, s2		; Read texel from chroma Cr texture
-	 * mov t0.z, t1.x		; Move Cr sample into .z component
-	 */
-	for (i = 0; i < 3; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_INPUT, i > 0 ? 1 : 0, TGSI_FILE_SAMPLER, i);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-		inst = vl_inst2(TGSI_OPCODE_MOV, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-		inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-		inst.FullDstRegisters[0].DstRegister.WriteMask = TGSI_WRITEMASK_X << i;
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	}
-
-	/* mul t0, t0, c0		; Rescale texel to correct range */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_CONSTANT, 0);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* XXX: Pos values off by 0.5? */
-	/* sub t4, i6.y, c1.x		; Sub 0.5 from denormalized pos */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_INPUT, 6, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* mul t3, t4, c1.x		; Multiply pos Y-coord by 1/2 */
-	inst = vl_inst3(TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* floor t3, t3			; Get rid of fractional part */
-	inst = vl_inst2(TGSI_OPCODE_FLOOR, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* mul t3, t3, c1.y		; Multiply by 2 */
-	inst = vl_inst3( TGSI_OPCODE_MUL, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_CONSTANT, 1);
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleX = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleY = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleZ = TGSI_SWIZZLE_Y;
-	inst.FullSrcRegisters[1].SrcRegister.SwizzleW = TGSI_SWIZZLE_Y;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* sub t3, t4, t3		; Subtract from original Y to get Y % 2 */
-	inst = vl_inst3(TGSI_OPCODE_SUB, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 3);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * tex2d t1, i2, s3		; Read texel from past ref macroblock top field
-	 * tex2d t2, i3, s3		; Read texel from past ref macroblock bottom field
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 1, TGSI_FILE_INPUT, i + 2, TGSI_FILE_SAMPLER, 3);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
-	/* lerp t1, t3, t1, t2		; Choose between top and bottom fields based on Y % 2 */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/*
-	 * tex2d t4, i4, s4		; Read texel from future ref macroblock top field
-	 * tex2d t5, i5, s4		; Read texel from future ref macroblock bottom field
-	 */
-	for (i = 0; i < 2; ++i)
-	{
-		inst = vl_tex(TGSI_TEXTURE_2D, TGSI_FILE_TEMPORARY, i + 4, TGSI_FILE_INPUT, i + 4, TGSI_FILE_SAMPLER, 4);
-		ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-	}
-
-	/* TODO: Move to conditional tex fetch on t3 instead of lerp */
-	/* lerp t2, t3, t4, t5		; Choose between top and bottom fields based on Y % 2 */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 2, TGSI_FILE_TEMPORARY, 3, TGSI_FILE_TEMPORARY, 4, TGSI_FILE_TEMPORARY, 5);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* lerp t1, c1.x, t1, t2	; Blend past and future texels */
-	inst = vl_inst4(TGSI_OPCODE_LERP, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_CONSTANT, 1, TGSI_FILE_TEMPORARY, 1, TGSI_FILE_TEMPORARY, 2);
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleX = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleY = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleZ = TGSI_SWIZZLE_X;
-	inst.FullSrcRegisters[0].SrcRegister.SwizzleW = TGSI_SWIZZLE_X;
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* add o0, t0, t1		; Add past/future ref and differential to form final output */
-	inst = vl_inst3(TGSI_OPCODE_ADD, TGSI_FILE_OUTPUT, 0, TGSI_FILE_TEMPORARY, 0, TGSI_FILE_TEMPORARY, 1);
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	/* end */
-	inst = vl_end();
-	ti += tgsi_build_full_instruction(&inst, &tokens[ti], header, max_tokens - ti);
-
-	fs.tokens = tokens;
-	mc->b_fs[1] = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
-
-	return 0;
-}
-
-static int vlCreateDataBufs
-(
-	struct vlR16SnormMC *mc
-)
-{
-	struct pipe_context	*pipe;
-	unsigned int		i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-
-	/* Create our vertex buffer and vertex buffer element */
-	mc->vertex_bufs[0].pitch = sizeof(struct vlVertex2f);
-	mc->vertex_bufs[0].max_index = 23;
-	mc->vertex_bufs[0].buffer_offset = 0;
-	mc->vertex_bufs[0].buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_VERTEX,
-		sizeof(struct vlVertex2f) * 24
-	);
-
-	mc->vertex_elems[0].src_offset = 0;
-	mc->vertex_elems[0].vertex_buffer_index = 0;
-	mc->vertex_elems[0].nr_components = 2;
-	mc->vertex_elems[0].src_format = PIPE_FORMAT_R32G32_FLOAT;
-
-	/* Create our texcoord buffers and texcoord buffer elements */
-	for (i = 1; i < 3; ++i)
-	{
-		mc->vertex_bufs[i].pitch = sizeof(struct vlVertex2f);
-		mc->vertex_bufs[i].max_index = 23;
-		mc->vertex_bufs[i].buffer_offset = 0;
-		mc->vertex_bufs[i].buffer = pipe->winsys->buffer_create
-		(
-			pipe->winsys,
-			1,
-			PIPE_BUFFER_USAGE_VERTEX,
-			sizeof(struct vlVertex2f) * 24
-		);
-
-		mc->vertex_elems[i].src_offset = 0;
-		mc->vertex_elems[i].vertex_buffer_index = i;
-		mc->vertex_elems[i].nr_components = 2;
-		mc->vertex_elems[i].src_format = PIPE_FORMAT_R32G32_FLOAT;
-	}
-
-	/* Fill buffers */
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, mc->vertex_bufs[0].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		macroblock_verts,
-		sizeof(struct vlVertex2f) * 24
-	);
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, mc->vertex_bufs[1].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		macroblock_luma_texcoords,
-		sizeof(struct vlVertex2f) * 24
-	);
-	/* TODO: Accomodate 422, 444 */
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, mc->vertex_bufs[2].buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		macroblock_chroma_420_texcoords,
-		sizeof(struct vlVertex2f) * 24
-	);
-
-	for (i = 0; i < 3; ++i)
-		pipe->winsys->buffer_unmap(pipe->winsys, mc->vertex_bufs[i].buffer);
-
-	/* Create our constant buffer */
-	mc->vs_const_buf.size = sizeof(struct vlVertexShaderConsts);
-	mc->vs_const_buf.buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_CONSTANT,
-		mc->vs_const_buf.size
-	);
-
-	mc->fs_const_buf.size = sizeof(struct vlFragmentShaderConsts);
-	mc->fs_const_buf.buffer = pipe->winsys->buffer_create
-	(
-		pipe->winsys,
-		1,
-		PIPE_BUFFER_USAGE_CONSTANT,
-		mc->fs_const_buf.size
-	);
-
-	memcpy
-	(
-		pipe->winsys->buffer_map(pipe->winsys, mc->fs_const_buf.buffer, PIPE_BUFFER_USAGE_CPU_WRITE),
-		&fs_consts,
-		sizeof(struct vlFragmentShaderConsts)
-	);
-
-	pipe->winsys->buffer_unmap(pipe->winsys, mc->fs_const_buf.buffer);
-
-	return 0;
-}
-
-static int vlInit
-(
-	struct vlR16SnormMC *mc
-)
-{
-	struct pipe_context		*pipe;
-	struct pipe_sampler_state	sampler;
-	struct pipe_texture		template;
-	unsigned int			filters[5];
-	unsigned int			i;
-
-	assert(mc);
-
-	pipe = mc->pipe;
-
-	/* For MC we render to textures, which are rounded up to nearest POT */
-	mc->viewport.scale[0] = vlRoundUpPOT(mc->video_width);
-	mc->viewport.scale[1] = vlRoundUpPOT(mc->video_height);
-	mc->viewport.scale[2] = 1;
-	mc->viewport.scale[3] = 1;
-	mc->viewport.translate[0] = 0;
-	mc->viewport.translate[1] = 0;
-	mc->viewport.translate[2] = 0;
-	mc->viewport.translate[3] = 0;
-
-	mc->render_target.width = vlRoundUpPOT(mc->video_width);
-	mc->render_target.height = vlRoundUpPOT(mc->video_height);
-	mc->render_target.num_cbufs = 1;
-	/* FB for MC stage is a vlSurface, set in vlSetRenderSurface() */
-	mc->render_target.zsbuf = NULL;
-
-	filters[0] = PIPE_TEX_FILTER_NEAREST;
-	filters[1] = mc->video_format == vlFormatYCbCr444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
-	filters[2] = mc->video_format == vlFormatYCbCr444 ? PIPE_TEX_FILTER_NEAREST : PIPE_TEX_FILTER_LINEAR;
-	filters[3] = PIPE_TEX_FILTER_LINEAR;
-	filters[4] = PIPE_TEX_FILTER_LINEAR;
-
-	for (i = 0; i < 5; ++i)
-	{
-		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_img_filter = filters[i];
-		sampler.min_mip_filter = PIPE_TEX_MIPFILTER_NONE;
-		sampler.mag_img_filter = filters[i];
-		sampler.compare_mode = PIPE_TEX_COMPARE_NONE;
-		sampler.compare_func = PIPE_FUNC_ALWAYS;
-		sampler.normalized_coords = 1;
-		/*sampler.prefilter = ;*/
-		/*sampler.shadow_ambient = ;*/
-		/*sampler.lod_bias = ;*/
-		sampler.min_lod = 0;
-		/*sampler.max_lod = ;*/
-		/*sampler.border_color[i] = ;*/
-		/*sampler.max_anisotropy = ;*/
-		mc->samplers[i] = pipe->create_sampler_state(pipe, &sampler);
-	}
-
-	memset(&template, 0, sizeof(struct pipe_texture));
-	template.target = PIPE_TEXTURE_2D;
-	template.format = PIPE_FORMAT_R16_SNORM;
-	template.last_level = 0;
-	template.width[0] = 8;
-	template.height[0] = 8 * 4;
-	template.depth[0] = 1;
-	template.compressed = 0;
-	pf_get_block(template.format, &template.block);
-
-	for (i = 0; i < NUM_BUFS; ++i)
-		mc->textures[i][0] = pipe->screen->texture_create(pipe->screen, &template);
-
-	if (mc->video_format == vlFormatYCbCr420)
-		template.height[0] = 8;
-	else if (mc->video_format == vlFormatYCbCr422)
-		template.height[0] = 8 * 2;
-	else if (mc->video_format == vlFormatYCbCr444)
-		template.height[0] = 8 * 4;
-	else
-		assert(0);
-
-	for (i = 0; i < NUM_BUFS; ++i)
-	{
-		mc->textures[i][1] = pipe->screen->texture_create(pipe->screen, &template);
-		mc->textures[i][2] = pipe->screen->texture_create(pipe->screen, &template);
-	}
-
-	/* textures[3] & textures[4] are assigned from vlSurfaces for P and B macroblocks at render time */
-
-	vlCreateVertexShaderIMB(mc);
-	vlCreateFragmentShaderIMB(mc);
-	vlCreateVertexShaderFramePMB(mc);
-	vlCreateVertexShaderFieldPMB(mc);
-	vlCreateFragmentShaderFramePMB(mc);
-	vlCreateFragmentShaderFieldPMB(mc);
-	vlCreateVertexShaderFrameBMB(mc);
-	vlCreateVertexShaderFieldBMB(mc);
-	vlCreateFragmentShaderFrameBMB(mc);
-	vlCreateFragmentShaderFieldBMB(mc);
-	vlCreateDataBufs(mc);
-
-	return 0;
-}
-
-int vlCreateR16SNormMC
-(
-	struct pipe_context *pipe,
-	unsigned int video_width,
-	unsigned int video_height,
-	enum vlFormat video_format,
-	struct vlRender **render
-)
-{
-	struct vlR16SnormMC *mc;
-
-	assert(pipe);
-	assert(render);
-
-	mc = calloc(1, sizeof(struct vlR16SnormMC));
-
-	mc->base.vlBegin = &vlBegin;
-	mc->base.vlRenderMacroBlocksMpeg2 = &vlRenderMacroBlocksMpeg2R16Snorm;
-	mc->base.vlEnd = &vlEnd;
-	mc->base.vlFlush = &vlFlush;
-	mc->base.vlDestroy = &vlDestroy;
-	mc->pipe = pipe;
-	mc->video_width = video_width;
-	mc->video_height = video_height;
-	mc->cur_buf = 0;
-
-	vlInit(mc);
-
-	*render = &mc->base;
-
-	return 0;
-}
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h
deleted file mode 100644
index 9842926bf7..0000000000
--- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc.h
+++ /dev/null
@@ -1,18 +0,0 @@
-#ifndef vl_r16snorm_mc_h
-#define vl_r16snorm_mc_h
-
-#include "vl_types.h"
-
-struct pipe_context;
-struct vlRender;
-
-int vlCreateR16SNormMC
-(
-	struct pipe_context *pipe,
-	unsigned int video_width,
-	unsigned int video_height,
-	enum vlFormat video_format,
-	struct vlRender **render
-);
-
-#endif
-- 
cgit v1.2.3


From 1e9c3efcc783cee46268cc227234ed118f0cc08b Mon Sep 17 00:00:00 2001
From: Younes Manton <younes.m@gmail.com>
Date: Sat, 20 Dec 2008 17:09:55 -0500
Subject: g3dvl: Use Gallium MALLOC wrappers.

---
 src/gallium/state_trackers/g3dvl/Makefile             |  5 ++++-
 src/gallium/state_trackers/g3dvl/vl_basic_csc.c       | 14 +++++++-------
 src/gallium/state_trackers/g3dvl/vl_context.c         |  6 +++---
 src/gallium/state_trackers/g3dvl/vl_display.c         |  6 +++---
 src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c | 10 +++++-----
 src/gallium/state_trackers/g3dvl/vl_screen.c          |  6 +++---
 src/gallium/state_trackers/g3dvl/vl_surface.c         |  6 +++---
 src/libXvMC/Makefile                                  |  2 +-
 src/libXvMC/block.c                                   | 10 +++++-----
 9 files changed, 34 insertions(+), 31 deletions(-)

(limited to 'src/gallium/state_trackers/g3dvl/Makefile')

diff --git a/src/gallium/state_trackers/g3dvl/Makefile b/src/gallium/state_trackers/g3dvl/Makefile
index 84a0b2c6d8..cddfca54fe 100644
--- a/src/gallium/state_trackers/g3dvl/Makefile
+++ b/src/gallium/state_trackers/g3dvl/Makefile
@@ -3,7 +3,10 @@ OBJECTS		= vl_display.o vl_screen.o vl_context.o vl_surface.o vl_shader_build.o
 		  vl_r16snorm_mc_buf.o
 GALLIUMDIR	= ../..
 
-CFLAGS		+= -g -Wall -fPIC -I${GALLIUMDIR}/include -I${GALLIUMDIR}/auxiliary -I${GALLIUMDIR}/winsys/g3dvl
+CFLAGS		+= -g -Wall -Werror=implicit-function-declaration -fPIC	\
+		   -I${GALLIUMDIR}/include				\
+		   -I${GALLIUMDIR}/auxiliary				\
+		   -I${GALLIUMDIR}/winsys/g3dvl				\
 
 #############################################
 
diff --git a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
index 626d23cd46..3ce93cf49d 100644
--- a/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
+++ b/src/gallium/state_trackers/g3dvl/vl_basic_csc.c
@@ -1,13 +1,13 @@
 #define VL_INTERNAL
 #include "vl_basic_csc.h"
 #include <assert.h>
-#include <stdlib.h>
 #include <pipe/p_context.h>
 #include <pipe/p_winsys.h>
 #include <pipe/p_state.h>
 #include <pipe/p_inlines.h>
 #include <tgsi/tgsi_parse.h>
 #include <tgsi/tgsi_build.h>
+#include <util/u_memory.h>
 #include "vl_csc.h"
 #include "vl_surface.h"
 #include "vl_shader_build.h"
@@ -237,7 +237,7 @@ static int vlDestroy
 	pipe->winsys->buffer_destroy(pipe->winsys, basic_csc->vs_const_buf.buffer);
 	pipe->winsys->buffer_destroy(pipe->winsys, basic_csc->fs_const_buf.buffer);
 
-	free(basic_csc);
+	FREE(basic_csc);
 
 	return 0;
 }
@@ -369,7 +369,7 @@ static int vlCreateVertexShader
 	assert(context);
 
 	pipe = csc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+	tokens = (struct tgsi_token*)MALLOC(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
@@ -430,7 +430,7 @@ static int vlCreateVertexShader
 
 	vs.tokens = tokens;
 	csc->vertex_shader = pipe->create_vs_state(pipe, &vs);
-	free(tokens);
+	FREE(tokens);
 
 	return 0;
 }
@@ -456,7 +456,7 @@ static int vlCreateFragmentShader
 	assert(context);
 
 	pipe = csc->pipe;
-	tokens = (struct tgsi_token*)malloc(max_tokens * sizeof(struct tgsi_token));
+	tokens = (struct tgsi_token*)MALLOC(max_tokens * sizeof(struct tgsi_token));
 
 	/* Version */
 	*(struct tgsi_version*)&tokens[0] = tgsi_build_version();
@@ -517,7 +517,7 @@ static int vlCreateFragmentShader
 
 	fs.tokens = tokens;
 	csc->fragment_shader = pipe->create_fs_state(pipe, &fs);
-	free(tokens);
+	FREE(tokens);
 
 	return 0;
 }
@@ -691,7 +691,7 @@ int vlCreateBasicCSC
 	assert(pipe);
 	assert(csc);
 
-	basic_csc = calloc(1, sizeof(struct vlBasicCSC));
+	basic_csc = CALLOC_STRUCT(vlBasicCSC);
 
 	if (!basic_csc)
 		return 1;
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c
index 5b7bb73b39..fbea1363d8 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.c
+++ b/src/gallium/state_trackers/g3dvl/vl_context.c
@@ -1,9 +1,9 @@
 #define VL_INTERNAL
 #include "vl_context.h"
 #include <assert.h>
-#include <stdlib.h>
 #include <pipe/p_context.h>
 #include <pipe/p_state.h>
+#include <util/u_memory.h>
 #include "vl_render.h"
 #include "vl_r16snorm_mc_buf.h"
 #include "vl_csc.h"
@@ -111,7 +111,7 @@ int vlCreateContext
 	assert(context);
 	assert(pipe);
 
-	ctx = calloc(1, sizeof(struct vlContext));
+	ctx = CALLOC_STRUCT(vlContext);
 
 	if (!ctx)
 		return 1;
@@ -152,7 +152,7 @@ int vlDestroyContext
 	context->pipe->delete_rasterizer_state(context->pipe, context->raster);
 	context->pipe->delete_depth_stencil_alpha_state(context->pipe, context->dsa);
 
-	free(context);
+	FREE(context);
 
 	return 0;
 }
diff --git a/src/gallium/state_trackers/g3dvl/vl_display.c b/src/gallium/state_trackers/g3dvl/vl_display.c
index af80faa7f5..dce06de758 100644
--- a/src/gallium/state_trackers/g3dvl/vl_display.c
+++ b/src/gallium/state_trackers/g3dvl/vl_display.c
@@ -1,7 +1,7 @@
 #define VL_INTERNAL
 #include "vl_display.h"
 #include <assert.h>
-#include <stdlib.h>
+#include <util/u_memory.h>
 
 int vlCreateDisplay
 (
@@ -14,7 +14,7 @@ int vlCreateDisplay
 	assert(native_display);
 	assert(display);
 
-	dpy = calloc(1, sizeof(struct vlDisplay));
+	dpy = CALLOC_STRUCT(vlDisplay);
 
 	if (!dpy)
 		return 1;
@@ -32,7 +32,7 @@ int vlDestroyDisplay
 {
 	assert(display);
 
-	free(display);
+	FREE(display);
 
 	return 0;
 }
diff --git a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
index a31f5c58f4..fcea899ef9 100644
--- a/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
+++ b/src/gallium/state_trackers/g3dvl/vl_r16snorm_mc_buf.c
@@ -1,7 +1,6 @@
 #define VL_INTERNAL
 #include "vl_r16snorm_mc_buf.h"
 #include <assert.h>
-#include <stdlib.h>
 #include <pipe/p_context.h>
 #include <pipe/p_winsys.h>
 #include <pipe/p_screen.h>
@@ -10,6 +9,7 @@
 #include <tgsi/tgsi_parse.h>
 #include <tgsi/tgsi_build.h>
 #include <util/u_math.h>
+#include <util/u_memory.h>
 #include "vl_render.h"
 #include "vl_shader_build.h"
 #include "vl_surface.h"
@@ -869,8 +869,8 @@ static int vlDestroy
 	pipe->winsys->buffer_destroy(pipe->winsys, mc->vs_const_buf.buffer);
 	pipe->winsys->buffer_destroy(pipe->winsys, mc->fs_const_buf.buffer);
 
-	free(mc->macroblocks);
-	free(mc);
+	FREE(mc->macroblocks);
+	FREE(mc);
 
 	return 0;
 }
@@ -1007,7 +1007,7 @@ static int vlCreateDataBufs
 
 	pipe->winsys->buffer_unmap(pipe->winsys, mc->fs_const_buf.buffer);
 
-	mc->macroblocks = malloc(sizeof(struct vlMpeg2MacroBlock) * mc->macroblocks_per_picture);
+	mc->macroblocks = MALLOC(sizeof(struct vlMpeg2MacroBlock) * mc->macroblocks_per_picture);
 
 	return 0;
 }
@@ -1133,7 +1133,7 @@ int vlCreateR16SNormBufferedMC
 	assert(pipe);
 	assert(render);
 
-	mc = calloc(1, sizeof(struct vlR16SnormBufferedMC));
+	mc = CALLOC_STRUCT(vlR16SnormBufferedMC);
 
 	mc->base.vlBegin = &vlBegin;
 	mc->base.vlRenderMacroBlocksMpeg2 = &vlRenderMacroBlocksMpeg2R16SnormBuffered;
diff --git a/src/gallium/state_trackers/g3dvl/vl_screen.c b/src/gallium/state_trackers/g3dvl/vl_screen.c
index 484f63b0d4..ade8643a66 100644
--- a/src/gallium/state_trackers/g3dvl/vl_screen.c
+++ b/src/gallium/state_trackers/g3dvl/vl_screen.c
@@ -1,7 +1,7 @@
 #define VL_INTERNAL
 #include "vl_screen.h"
 #include <assert.h>
-#include <stdlib.h>
+#include <util/u_memory.h>
 
 int vlCreateScreen
 (
@@ -17,7 +17,7 @@ int vlCreateScreen
 	assert(pscreen);
 	assert(vl_screen);
 
-	scrn = calloc(1, sizeof(struct vlScreen));
+	scrn = CALLOC_STRUCT(vlScreen);
 
 	if (!scrn)
 		return 1;
@@ -37,7 +37,7 @@ int vlDestroyScreen
 {
 	assert(screen);
 
-	free(screen);
+	FREE(screen);
 
 	return 0;
 }
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c
index 076bd40d41..911469f966 100644
--- a/src/gallium/state_trackers/g3dvl/vl_surface.c
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.c
@@ -1,11 +1,11 @@
 #define VL_INTERNAL
 #include "vl_surface.h"
 #include <assert.h>
-#include <stdlib.h>
 #include <string.h>
 #include <pipe/p_screen.h>
 #include <pipe/p_state.h>
 #include <pipe/p_inlines.h>
+#include <util/u_memory.h>
 #include <vl_winsys.h>
 #include "vl_screen.h"
 #include "vl_context.h"
@@ -28,7 +28,7 @@ int vlCreateSurface
 	assert(screen);
 	assert(surface);
 
-	sfc = calloc(1, sizeof(struct vlSurface));
+	sfc = CALLOC_STRUCT(vlSurface);
 
 	if (!sfc)
 		return 1;
@@ -64,7 +64,7 @@ int vlDestroySurface
 	assert(surface);
 
 	pipe_texture_release(&surface->texture);
-	free(surface);
+	FREE(surface);
 
 	return 0;
 }
diff --git a/src/libXvMC/Makefile b/src/libXvMC/Makefile
index 4de26e9620..7badcfd264 100644
--- a/src/libXvMC/Makefile
+++ b/src/libXvMC/Makefile
@@ -8,7 +8,7 @@ ifeq (${DRIVER}, softpipe)
 OBJECTS         += ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
 endif
 
-CFLAGS	+= -g -fPIC -Wall				\
+CFLAGS	+= -g -fPIC -Wall -Werror=implicit-function-declaration	\
 	   -I${GALLIUMDIR}/state_trackers/g3dvl		\
 	   -I${GALLIUMDIR}/winsys/g3dvl			\
 	   -I${GALLIUMDIR}/include			\
diff --git a/src/libXvMC/block.c b/src/libXvMC/block.c
index 328b035576..b38a89be09 100644
--- a/src/libXvMC/block.c
+++ b/src/libXvMC/block.c
@@ -1,7 +1,7 @@
 #include <assert.h>
-#include <stdlib.h>
 #include <X11/Xlib.h>
 #include <X11/extensions/XvMC.h>
+#include <util/u_memory.h>
 #include <vl_display.h>
 #include <vl_screen.h>
 #include <vl_context.h>
@@ -26,7 +26,7 @@ Status XvMCCreateBlocks(Display *display, XvMCContext *context, unsigned int num
 
 	blocks->context_id = context->context_id;
 	blocks->num_blocks = num_blocks;
-	blocks->blocks = malloc(BLOCK_SIZE * num_blocks);
+	blocks->blocks = MALLOC(BLOCK_SIZE * num_blocks);
 	/* Since we don't have a VL type for blocks, set privData to the display so we can catch mismatches */
 	blocks->privData = display;
 
@@ -38,7 +38,7 @@ Status XvMCDestroyBlocks(Display *display, XvMCBlockArray *blocks)
 	assert(display);
 	assert(blocks);
 	assert(display == blocks->privData);
-	free(blocks->blocks);
+	FREE(blocks->blocks);
 
 	return Success;
 }
@@ -61,7 +61,7 @@ Status XvMCCreateMacroBlocks(Display *display, XvMCContext *context, unsigned in
 
 	blocks->context_id = context->context_id;
 	blocks->num_blocks = num_blocks;
-	blocks->macro_blocks = malloc(sizeof(XvMCMacroBlock) * num_blocks);
+	blocks->macro_blocks = MALLOC(sizeof(XvMCMacroBlock) * num_blocks);
 	/* Since we don't have a VL type for blocks, set privData to the display so we can catch mismatches */
 	blocks->privData = display;
 
@@ -73,7 +73,7 @@ Status XvMCDestroyMacroBlocks(Display *display, XvMCMacroBlockArray *blocks)
 	assert(display);
 	assert(blocks);
 	assert(display == blocks->privData);
-	free(blocks->macro_blocks);
+	FREE(blocks->macro_blocks);
 
 	return Success;
 }
-- 
cgit v1.2.3