g3dvl: Support for field and frame based MC for progressive pictures.

MC support for frame and field based motion prediction. Also various bug
fixes, clean up.

9 files changed, 1977 insertions, 231 deletions

diff --git a/src/gallium/state_trackers/g3dvl/tests/.gitignore b/src/gallium/state_trackers/g3dvl/tests/.gitignore
index 939666da9ab..9b1ec4e2122 100644
--- a/src/gallium/state_trackers/g3dvl/tests/.gitignore
+++ b/src/gallium/state_trackers/g3dvl/tests/.gitignore
@@ -2,5 +2,5 @@ test_context
 test_surface
 test_i_rendering
 test_p_rendering
+test_pf_rendering
 test_b_rendering
-
diff --git a/src/gallium/state_trackers/g3dvl/tests/Makefile b/src/gallium/state_trackers/g3dvl/tests/Makefile
index 8f983593c3a..45cefa2e570 100644
--- a/src/gallium/state_trackers/g3dvl/tests/Makefile
+++ b/src/gallium/state_trackers/g3dvl/tests/Makefile
@@ -20,7 +20,7 @@ LIBS	+= -lg3dvl -lsoftpipe -ldraw -ltgsi -ltranslate -lrtasm -lcso_cache -lutil
 
 .PHONY	= all clean
 
-all: test_context test_surface test_i_rendering test_p_rendering test_b_rendering
+all: test_context test_surface test_i_rendering test_p_rendering test_pf_rendering test_b_rendering
 
 test_context: test_context.o ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
 	$(CC) ${LDFLAGS} -o $@ $^ ${LIBS}
@@ -34,9 +34,12 @@ test_i_rendering: test_i_rendering.o ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
 test_p_rendering: test_p_rendering.o ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
 	$(CC) ${LDFLAGS} -o $@ $^ ${LIBS}
 
+test_pf_rendering: test_pf_rendering.o ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
+	$(CC) ${LDFLAGS} -o $@ $^ ${LIBS}
+
 test_b_rendering: test_b_rendering.o ${GALLIUMDIR}/winsys/g3dvl/xsp_winsys.o
 	$(CC) ${LDFLAGS} -o $@ $^ ${LIBS}
 
 clean:
-	rm -rf *.o test_context test_surface test_i_rendering test_p_rendering test_b_rendering
+	rm -rf *.o test_context test_surface test_i_rendering test_p_rendering test_pf_rendering test_b_rendering
 
diff --git a/src/gallium/state_trackers/g3dvl/tests/test_pf_rendering.c b/src/gallium/state_trackers/g3dvl/tests/test_pf_rendering.c
new file mode 100644
index 00000000000..43586fc553c
--- /dev/null
+++ b/src/gallium/state_trackers/g3dvl/tests/test_pf_rendering.c
@@ -0,0 +1,214 @@
+#include <stdio.h>
+#include <X11/Xlib.h>
+#include <vl_context.h>
+#include <vl_surface.h>
+#include <xsp_winsys.h>
+
+static const unsigned short ycbcr16x16_420[8*8*6] =
+{
+	0x00A5,0x00A5,0x00A5,0x0072,0x00A5,0x0072,0x0072,0x0072,
+	0x0072,0x00A5,0x0072,0x0072,0x00A5,0x0072,0x0072,0x0072,
+	0x0072,0x00A5,0x0072,0x0072,0x00A5,0x00A5,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x00A5,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x0072,0x00A5,0x00A5,
+	0x0072,0x0072,0x0072,0x0072,0x0072,0x00A5,0x00A5,0x00A5,
+	0x0072,0x0072,0x0072,0x0072,0x00A5,0x00A5,0x00A5,0x00A5,
+	
+	0x004F,0x004F,0x004F,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,
+	0x00B2,0x00B2,0x004F,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,
+	0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,
+	0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,
+	0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,
+	0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,
+	0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x004F,
+	0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x00B2,0x004F,0x004F,
+	
+	0x003E,0x003E,0x003E,0x0060,0x0060,0x0060,0x0060,0x0060,
+	0x0060,0x0060,0x003E,0x0060,0x0060,0x0060,0x0060,0x0060,
+	0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,
+	0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,
+	0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,
+	0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,
+	0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x003E,
+	0x0060,0x0060,0x0060,0x0060,0x0060,0x0060,0x003E,0x003E
+};
+
+static const signed short ycbcr16x16_420_2[8*8*6] =
+{
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0,
+	  0,  0,  0,  0,  0,  0,  0,  0
+};
+
+int main(int argc, char **argv)
+{
+	const unsigned int	video_width = 32, video_height = 32;
+	const unsigned int	window_width = video_width * 2, window_height = video_height * 2;
+	int			quit = 0;
+	Display			*display;
+	Window			root, window;
+	Pixmap			framebuffer;
+	XEvent			event;
+	struct pipe_context	*pipe;
+	struct VL_CONTEXT	*ctx;
+	struct VL_SURFACE	*sfc, *ref_sfc;
+	struct VL_MOTION_VECTOR	motion_vector =
+	{
+		{0, 0}, {32, 32}
+	};
+	
+	display = XOpenDisplay(NULL);
+	root = XDefaultRootWindow(display);
+	window = XCreateSimpleWindow(display, root, 0, 0, window_width, window_height, 0, 0, 0);
+	framebuffer = XCreatePixmap(display, root, window_width, window_height, 24);
+	
+	XSelectInput(display, window, ExposureMask | KeyPressMask);
+	XMapWindow(display, window);
+	XSync(display, 0);
+	
+	pipe = create_pipe_context(display);
+	vlCreateContext(display, pipe, video_width, video_height, VL_FORMAT_YCBCR_420, &ctx);
+	vlCreateSurface(ctx, &sfc);
+	vlCreateSurface(ctx, &ref_sfc);
+	
+	vlRenderIMacroBlock(VL_FRAME_PICTURE, VL_FIELD_FIRST, 0, 0, 0x3F, VL_DCT_FRAME_CODED, (short*)ycbcr16x16_420, ref_sfc);
+	vlRenderIMacroBlock(VL_FRAME_PICTURE, VL_FIELD_FIRST, 1, 0, 0x3F, VL_DCT_FRAME_CODED, (short*)ycbcr16x16_420, ref_sfc);
+	vlRenderIMacroBlock(VL_FRAME_PICTURE, VL_FIELD_FIRST, 0, 1, 0x3F, VL_DCT_FRAME_CODED, (short*)ycbcr16x16_420, ref_sfc);
+	vlRenderIMacroBlock(VL_FRAME_PICTURE, VL_FIELD_FIRST, 1, 1, 0x3F, VL_DCT_FRAME_CODED, (short*)ycbcr16x16_420, ref_sfc);
+	vlRenderPMacroBlock
+	(
+		VL_FRAME_PICTURE,
+		VL_FIELD_FIRST,
+		0,
+		0,
+		VL_FIELD_MC,
+		&motion_vector,
+		0x3F,
+		VL_DCT_FRAME_CODED,
+		(short*)ycbcr16x16_420_2,
+		ref_sfc,
+		sfc
+	);
+	vlPutSurface(sfc, framebuffer, 0, 0, video_width, video_height, 0, 0, window_width, window_height, VL_FRAME_PICTURE);
+	
+	puts("Press any key to continue...");
+	
+	while (!quit)
+	{
+		XNextEvent(display, &event);
+		switch (event.type)
+		{
+			case Expose:
+			{
+				XCopyArea
+				(
+					display,
+					framebuffer,
+					window,
+					XDefaultGC(display, XDefaultScreen(display)),
+					0,
+					0,
+					window_width,
+					window_height,
+					0,
+					0
+				);
+				break;
+			}
+			case KeyPress:
+			{
+				quit = 1;
+				break;
+			}
+		}
+	}
+	
+	vlDestroySurface(sfc);
+	vlDestroySurface(ref_sfc);
+	vlDestroyContext(ctx);
+	
+	XFreePixmap(display, framebuffer);
+	XDestroyWindow(display, window);
+	XCloseDisplay(display);
+	
+	return 0;
+}
+
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.c b/src/gallium/state_trackers/g3dvl/vl_context.c
index 3b9afabbb84..d2b1ad7948e 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.c
+++ b/src/gallium/state_trackers/g3dvl/vl_context.c
@@ -36,8 +36,8 @@ static int vlCreateVertexShaderIMC(struct VL_CONTEXT *context)
 	const unsigned int		semantic_names[3] =
 					{
 						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,
-						TGSI_SEMANTIC_GENERIC,
+						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;
@@ -353,7 +353,7 @@ static int vlCreateFragmentShaderIMC(struct VL_CONTEXT *context)
 	return 0;
 }
 
-static int vlCreateVertexShaderPMC(struct VL_CONTEXT *context)
+static int vlCreateVertexShaderFramePMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 100;
 	const unsigned int		num_input_attribs = 3;
@@ -361,15 +361,15 @@ static int vlCreateVertexShaderPMC(struct VL_CONTEXT *context)
 	const unsigned int		input_semantic_names[3] =
 					{
 						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,
-						TGSI_SEMANTIC_GENERIC
+						TGSI_SEMANTIC_GENERIC,	/* Luma texcoords */
+						TGSI_SEMANTIC_GENERIC	/* Chroma texcoords */
 					};
 	const unsigned int		output_semantic_names[4] =
 					{
 						TGSI_SEMANTIC_POSITION,
-						TGSI_SEMANTIC_GENERIC,
-						TGSI_SEMANTIC_GENERIC,
-						TGSI_SEMANTIC_GENERIC
+						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};
@@ -430,14 +430,15 @@ static int vlCreateVertexShaderPMC(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] translates the ref surface texcoords to the ref macroblock */
+	   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 = 2;
+	decl.u.DeclarationRange.Last = 3;
 	ti += tgsi_build_full_declaration
 	(
 		&decl,
@@ -526,7 +527,7 @@ static int vlCreateVertexShaderPMC(struct VL_CONTEXT *context)
 		);
 	}
 
-	/* add o3, t0, c2	; Translate texcoords into position */
+	/* add o3, t0, c3	; Translate texcoords into position */
 	inst = tgsi_default_full_instruction();
 	inst.Instruction.Opcode = TGSI_OPCODE_ADD;
 	inst.Instruction.NumDstRegs = 1;
@@ -536,6 +537,264 @@ static int vlCreateVertexShaderPMC(struct VL_CONTEXT *context)
 	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
+	);
+
+	/* 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
+	);
+	
+	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;
+	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;
+	
+	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 */
+	processor = (struct tgsi_processor*)&tokens[2];
+	*processor = tgsi_build_processor(proc_type, header);
+
+	ti = 3;
+
+	/* Declare inputs (pos, texcoords) */
+	for (i = 0; i < num_input_attribs; 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
+		);
+	}
+	
+	/* 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
+	);
+
+	/* Declare outputs (pos, texcoords) */
+	for (i = 0; i < num_output_attribs; 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
+		);
+	}
+	
+	/* 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 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
+	);
+	
+	/* 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
+	);
+	
+	/*
+	mov o1, i1		; Move luma & chroma texcoords to output
+	mov o2, i2
+	*/
+	for (i = 1; i < num_output_attribs - 1; ++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
+		);
+	}
+
+	/* add o3, t0, c3	; Translate top field texcoords into position
+	   add o4, t0, c4	; Translate bottom field texcoords into position */
+	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
+		);
+	}
+	
+	/* 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
 	(
@@ -560,14 +819,14 @@ static int vlCreateVertexShaderPMC(struct VL_CONTEXT *context)
 	
 	vs.tokens = tokens;
 	
-	context->states.mc.p_vs = pipe->create_vs_state(pipe, &vs);
+	context->states.mc.p_vs[1] = pipe->create_vs_state(pipe, &vs);
 	
 	free(tokens);
 	
 	return 0;
 }
 
-static int vlCreateFragmentShaderPMC(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;
@@ -837,14 +1096,432 @@ static int vlCreateFragmentShaderPMC(struct VL_CONTEXT *context)
 
 	fs.tokens = tokens;
 	
-	context->states.mc.p_fs = pipe->create_fs_state(pipe, &fs);
+	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;
+	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;
+	
+	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 */
+	processor = (struct tgsi_processor*)&tokens[2];
+	*processor = tgsi_build_processor(proc_type, 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 */
+	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
+		);
+	}
+	
+	/* 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
+	);
+	
+	/* Declare samplers */
+	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
+		);
+	}
+	
+	/*
+	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
+	*/
+	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.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
+		);
+		
+		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.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
+		);
+	}
+	
+	/* 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
+	);
+	
+	/* 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
+	);
+	
+	/* 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
+	);
+	
+	/* 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
+		);
+	}
+	
+	/* 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;
+	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
+	);
+	
+	/* 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;
+	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 = 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
+	);
+	
+	/* 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;
+	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 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
+	);
+	
+	/* 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
+	);
+	
+	/* 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
+	);
+
+	/* 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
+	);
+
+	fs.tokens = tokens;
+	
+	context->states.mc.p_fs[1] = pipe->create_fs_state(pipe, &fs);
 	
 	free(tokens);
 	
 	return 0;
 }
 
-static int vlCreateVertexShaderBMC(struct VL_CONTEXT *context)
+static int vlCreateVertexShaderFrameBMC(struct VL_CONTEXT *context)
 {
 	const unsigned int		max_tokens = 100;
 	const unsigned int		num_input_attribs = 3;
@@ -922,15 +1599,17 @@ static int vlCreateVertexShaderBMC(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] translates the past surface texcoords to the ref macroblock
-	   C[3] translates the future surface texcoords to the ref macroblock */
+	   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 = 3;
+	decl.u.DeclarationRange.Last = 5;
 	ti += tgsi_build_full_declaration
 	(
 		&decl,
@@ -1019,8 +1698,8 @@ static int vlCreateVertexShaderBMC(struct VL_CONTEXT *context)
 		);
 	}
 	
-	/* add o3, t0, c2	; Translate past surface texcoords into position
-	   add o4, t0, c3	; Repeat for future surface texcoords */
+	/* add o3, t0, c3	; Translate past surface texcoords into position
+	   add o4, t0, c5	; Repeat for future surface texcoords */
 	for (i = 0; i < 2; ++i)
 	{
 		inst = tgsi_default_full_instruction();
@@ -1032,7 +1711,252 @@ static int vlCreateVertexShaderBMC(struct VL_CONTEXT *context)
 		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;
+		inst.FullSrcRegisters[1].SrcRegister.Index = 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
+	);
+	
+	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;
+	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;
+	
+	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 */
+	processor = (struct tgsi_processor*)&tokens[2];
+	*processor = tgsi_build_processor(proc_type, header);
+
+	ti = 3;
+
+	/* Declare inputs (pos, texcoords) */
+	for (i = 0; i < num_input_attribs; 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
+		);
+	}
+	
+	/* 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
+	);
+
+	/* Declare outputs (pos, texcoords) */
+	for (i = 0; i < num_output_attribs; 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
+		);
+	}
+	
+	/* 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 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
+	);
+	
+	/* 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
+	);
+	
+	/*
+	mov o1, i1		; Move luma & chroma texcoords to output
+	mov o2, i2
+	*/
+	for (i = 1; i < num_output_attribs - 1; ++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
+		);
+	}
+
+	/* 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 */
+	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,
@@ -1041,6 +1965,25 @@ static int vlCreateVertexShaderBMC(struct VL_CONTEXT *context)
 			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
+	);
 
 	/* END */
 	inst = tgsi_default_full_instruction();
@@ -1057,14 +2000,14 @@ static int vlCreateVertexShaderBMC(struct VL_CONTEXT *context)
 	
 	vs.tokens = tokens;
 	
-	context->states.mc.b_vs = pipe->create_vs_state(pipe, &vs);
+	context->states.mc.b_vs[1] = pipe->create_vs_state(pipe, &vs);
 	
 	free(tokens);
 	
 	return 0;
 }
 
-static int vlCreateFragmentShaderBMC(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;
@@ -1123,14 +2066,15 @@ static int vlCreateFragmentShaderBMC(struct VL_CONTEXT *context)
 	
 	/* 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*/
+	   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 = 1;
+	decl.u.DeclarationRange.Last = 2;
 	ti += tgsi_build_full_declaration
 	(
 		&decl,
@@ -1304,7 +2248,259 @@ static int vlCreateFragmentShaderBMC(struct VL_CONTEXT *context)
 		);
 	}
 	
-	/* add t0, t0, t1	; Add past and differential to form partial output */
+	/* 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.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
+	);
+	
+	/* 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
+	);
+
+	/* 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
+	);
+
+	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;
+	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;
+	
+	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 */
+	processor = (struct tgsi_processor*)&tokens[2];
+	*processor = tgsi_build_processor(proc_type, 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 */
+	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
+		);
+	}
+	
+	/* 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
+	);
+	
+	/* Declare samplers */
+	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
+		);
+	}
+	
+	/*
+	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
+	*/
+	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.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
+		);
+		
+		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.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
+		);
+	}
+	
+	/* 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
+	);
+	
+	/* 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;
@@ -1323,7 +2519,251 @@ static int vlCreateFragmentShaderBMC(struct VL_CONTEXT *context)
 		max_tokens - ti
 	);
 	
-	/* add o0, t0, t2	; Add future and differential to form final output */
+	/* 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
+	);
+	
+	/* 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;
+	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
+	);
+	
+	/* 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;
+	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 = 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
+	);
+	
+	/* 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;
+	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 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
+	);
+	
+	/* tex2d t1, i2, s3	; Read texel from past macroblock top field
+	   tex2d t2, i3, s3	; Read texel from past 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
+		);
+	}
+	
+	/* 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
+	);
+	
+	/* tex2d t4, i4, s4	; Read texel from future macroblock top field
+	   tex2d t5, i5, s4	; Read texel from future 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
+		);
+	}
+	
+	/* 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
+	);
+	
+	/* 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;
+	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
+	);
+	
+	/* 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;
@@ -1333,7 +2773,7 @@ static int vlCreateFragmentShaderBMC(struct VL_CONTEXT *context)
 	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 = 2;
+	inst.FullSrcRegisters[1].SrcRegister.Index = 1;
 	ti += tgsi_build_full_instruction
 	(
 		&inst,
@@ -1357,7 +2797,7 @@ static int vlCreateFragmentShaderBMC(struct VL_CONTEXT *context)
 
 	fs.tokens = tokens;
 	
-	context->states.mc.b_fs = pipe->create_fs_state(pipe, &fs);
+	context->states.mc.b_fs[1] = pipe->create_fs_state(pipe, &fs);
 	
 	free(tokens);
 	
@@ -1491,7 +2931,6 @@ static int vlInitMC(struct VL_CONTEXT *context)
 	context->states.mc.render_target.height = 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.cbufs[0] = ;*/
 	context->states.mc.render_target.zsbuf = NULL;
 	
 	filters[0] = PIPE_TEX_FILTER_NEAREST;
@@ -1530,6 +2969,7 @@ static int vlInitMC(struct VL_CONTEXT *context)
 	template.depth[0] = 1;
 	template.compressed = 0;
 	template.cpp = 2;
+	
 	context->states.mc.textures[0] = pipe->screen->texture_create(pipe->screen, &template);
 	
 	if (context->video_format == VL_FORMAT_YCBCR_420)
@@ -1548,10 +2988,14 @@ static int vlInitMC(struct VL_CONTEXT *context)
 	
 	vlCreateVertexShaderIMC(context);
 	vlCreateFragmentShaderIMC(context);
-	vlCreateVertexShaderPMC(context);
-	vlCreateFragmentShaderPMC(context);
-	vlCreateVertexShaderBMC(context);
-	vlCreateFragmentShaderBMC(context);
+	vlCreateVertexShaderFramePMC(context);
+	vlCreateVertexShaderFieldPMC(context);
+	vlCreateFragmentShaderFramePMC(context);
+	vlCreateFragmentShaderFieldPMC(context);
+	vlCreateVertexShaderFrameBMC(context);
+	vlCreateVertexShaderFieldBMC(context);
+	vlCreateFragmentShaderFrameBMC(context);
+	vlCreateFragmentShaderFieldBMC(context);
 	vlCreateDataBufsMC(context);
 	
 	return 0;
@@ -1575,10 +3019,14 @@ static int vlDestroyMC(struct VL_CONTEXT *context)
 	
 	context->pipe->delete_vs_state(context->pipe, context->states.mc.i_vs);
 	context->pipe->delete_fs_state(context->pipe, context->states.mc.i_fs);
-	context->pipe->delete_vs_state(context->pipe, context->states.mc.p_vs);
-	context->pipe->delete_fs_state(context->pipe, context->states.mc.p_fs);
-	context->pipe->delete_vs_state(context->pipe, context->states.mc.b_vs);
-	context->pipe->delete_fs_state(context->pipe, context->states.mc.b_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);
@@ -1982,7 +3430,7 @@ static int vlCreateDataBufsCSC(struct VL_CONTEXT *context)
 	
 	/*
 	TODO: Refactor this into a seperate function,
-	allow changing the csc matrix at runtime to switch between regular & full versions
+	allow changing the CSC matrix at runtime to switch between regular & full versions
 	*/
 	memcpy
 	(
@@ -2166,7 +3614,7 @@ static int vlDestroy(struct VL_CONTEXT *context)
 {
 	assert(context);
 	
-	/* Must unbind shaders before we can delete them for some reason */
+	/* 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);
 	
diff --git a/src/gallium/state_trackers/g3dvl/vl_context.h b/src/gallium/state_trackers/g3dvl/vl_context.h
index f26a4c5b6a7..8a123180734 100644
--- a/src/gallium/state_trackers/g3dvl/vl_context.h
+++ b/src/gallium/state_trackers/g3dvl/vl_context.h
@@ -34,8 +34,8 @@ struct VL_CONTEXT
 			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, *b_vs;
-			struct pipe_shader_state		*i_fs, *p_fs, *b_fs;
+			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;
diff --git a/src/gallium/state_trackers/g3dvl/vl_data.c b/src/gallium/state_trackers/g3dvl/vl_data.c
index 27893aee95e..7e6ee8ac12f 100644
--- a/src/gallium/state_trackers/g3dvl/vl_data.c
+++ b/src/gallium/state_trackers/g3dvl/vl_data.c
@@ -87,7 +87,8 @@ const struct VL_TEXCOORD2F *vl_surface_texcoords = (const struct VL_TEXCOORD2F*)
 const struct VL_MC_FS_CONSTS vl_mc_fs_consts =
 {
 	{256.0f, 256.0f, 256.0f, 0.0f},
-	{256.0f / 255.0f, 256.0f / 255.0f, 256.0f / 255.0f, 0.0f}
+	{256.0f / 255.0f, 256.0f / 255.0f, 256.0f / 255.0f, 0.0f},
+	{0.5f, 2.0f, 0.0f, 0.0f}
 };
 
 /*
diff --git a/src/gallium/state_trackers/g3dvl/vl_surface.c b/src/gallium/state_trackers/g3dvl/vl_surface.c
index 6451e54953d..d2220d7abf5 100644
--- a/src/gallium/state_trackers/g3dvl/vl_surface.c
+++ b/src/gallium/state_trackers/g3dvl/vl_surface.c
@@ -8,6 +8,85 @@
 #include "vl_context.h"
 #include "vl_defs.h"
 
+static int vlGrabFrameCodedFullBlock(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 vlGrabFrameCodedDiffBlock(short *src, short *dst, unsigned int dst_pitch)
+{
+	unsigned int x, y;
+	
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		for (x = 0; x < VL_BLOCK_WIDTH; ++x)
+			dst[y * dst_pitch + x] = src[y * VL_BLOCK_WIDTH + x] + 0x100;
+	
+	return 0;
+}
+
+static int vlGrabFieldCodedFullBlock(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 vlGrabFieldCodedDiffBlock(short *src, short *dst, unsigned int dst_pitch)
+{
+	unsigned int x, y;
+	
+	for (y = 0; y < VL_BLOCK_HEIGHT / 2; ++y)
+		for (x = 0; x < VL_BLOCK_WIDTH; ++x)
+			dst[y * dst_pitch * 2 + x] = src[y * VL_BLOCK_WIDTH + x] + 0x100;
+	
+	dst += VL_BLOCK_HEIGHT * dst_pitch;
+	
+	for (; y < VL_BLOCK_HEIGHT; ++y)
+		for (x = 0; x < VL_BLOCK_WIDTH; ++x)
+			dst[y * dst_pitch * 2 + x] = src[y * VL_BLOCK_WIDTH + x] + 0x100;
+	
+	return 0;
+}
+
+static int vlGrabNoBlock(short *dst, unsigned int dst_pitch)
+{
+	unsigned int x, y;
+	
+	for (y = 0; y < VL_BLOCK_HEIGHT; ++y)
+		for (x = 0; x < VL_BLOCK_WIDTH; ++x)
+			dst[y * dst_pitch + x] = 0x100;
+	
+	return 0;
+}
+
 static int vlGrabBlocks
 (
 	struct VL_CONTEXT *context,
@@ -19,7 +98,7 @@ static int vlGrabBlocks
 {
 	struct pipe_surface	*tex_surface;
 	short			*texels;
-	unsigned int		b, x, y, y2;
+	unsigned int		tb, sb = 0;
 	
 	assert(context);
 	assert(blocks);
@@ -33,134 +112,81 @@ static int vlGrabBlocks
 	
 	texels = pipe_surface_map(tex_surface, PIPE_BUFFER_USAGE_CPU_WRITE);
 	
-	for (b = 0; b < 4; ++b)
+	for (tb = 0; tb < 4; ++tb)
 	{
-		if ((coded_block_pattern >> b) & 1)
+		if ((coded_block_pattern >> (5 - tb)) & 1)
 		{
 			if (dct_type == VL_DCT_FRAME_CODED)
-			{
 				if (sample_type == VL_FULL_SAMPLE)
-				{
-					for (y = VL_BLOCK_HEIGHT * b; y < VL_BLOCK_HEIGHT * (b + 1); ++y)
-						memcpy
-						(
-							texels + y * tex_surface->pitch,
-							blocks + y * VL_BLOCK_WIDTH,
-							VL_BLOCK_WIDTH * 2
-						);
-				}
+					vlGrabFrameCodedFullBlock
+					(
+						blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT,
+						texels + tb * tex_surface->pitch * VL_BLOCK_HEIGHT,
+						tex_surface->pitch
+					);
 				else
-				{
-					for (y = VL_BLOCK_HEIGHT * b; y < VL_BLOCK_HEIGHT * (b + 1); ++y)
-						for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-							texels[y * tex_surface->pitch + x] =
-							blocks[y * VL_BLOCK_WIDTH + x] + 0x100;
-				}
-			}
+					vlGrabFrameCodedDiffBlock
+					(
+						blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT,
+						texels + tb * tex_surface->pitch * VL_BLOCK_HEIGHT,
+						tex_surface->pitch
+					);
 			else
-			{
 				if (sample_type == VL_FULL_SAMPLE)
-				{
-					for
+					vlGrabFieldCodedFullBlock
 					(
-						y = VL_BLOCK_HEIGHT * (b % 2), y2 = VL_BLOCK_HEIGHT * b;
-						y < VL_BLOCK_HEIGHT * ((b % 2) + 1);
-						y += 2, ++y2
-					)
-						memcpy
-						(
-							texels + y * tex_surface->pitch,
-							blocks + y2 * VL_BLOCK_WIDTH,
-							VL_BLOCK_WIDTH * 2
-						);
-					for
-					(
-						y = VL_BLOCK_HEIGHT * ((b % 2) + 2);
-						y < VL_BLOCK_HEIGHT * (((b % 2) + 2) + 1);
-						y += 2, ++y2
-					)
-						memcpy
-						(
-							texels + y * tex_surface->pitch,
-							blocks + y2 * VL_BLOCK_WIDTH,
-							VL_BLOCK_WIDTH * 2
-						);
-				}
+						blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT,
+						texels + (tb % 2) * tex_surface->pitch * VL_BLOCK_HEIGHT + (tb / 2) * tex_surface->pitch,
+						tex_surface->pitch
+					);
 				else
-				{
-					for
+					vlGrabFieldCodedDiffBlock
 					(
-						y = VL_BLOCK_HEIGHT * (b % 2), y2 = VL_BLOCK_HEIGHT * b;
-						y < VL_BLOCK_HEIGHT * ((b % 2) + 1);
-						y += 2, ++y2
-					)
-						for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-							texels[y * tex_surface->pitch + x] =
-							blocks[y2 * VL_BLOCK_WIDTH + x] + 0x100;
-					for
-					(
-						y = VL_BLOCK_HEIGHT * ((b % 2) + 2);
-						y < VL_BLOCK_HEIGHT * (((b % 2) + 2) + 1);
-						y += 2, ++y2
-					)
-						for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-							texels[y * tex_surface->pitch + x] =
-							blocks[y2 * VL_BLOCK_WIDTH + x] + 0x100;
-				}
-			}
+						blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT,
+						texels + (tb % 2) * tex_surface->pitch * VL_BLOCK_HEIGHT + (tb / 2) * tex_surface->pitch,
+						tex_surface->pitch
+					);
+			++sb;
 		}
 		else
-		{
-			for (y = VL_BLOCK_HEIGHT * b; y < VL_BLOCK_HEIGHT * (b + 1); ++y)
-			{
-				for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-					texels[y * tex_surface->pitch + x] = 0x100;
-			}
-		}
+			vlGrabNoBlock(texels + tb * tex_surface->pitch * VL_BLOCK_HEIGHT, tex_surface->pitch);
 	}
 	
 	pipe_surface_unmap(tex_surface);
 	
 	/* TODO: Implement 422, 444 */
-	for (b = 0; b < 2; ++b)
+	for (tb = 0; tb < 2; ++tb)
 	{
 		tex_surface = context->pipe->screen->get_tex_surface
-		(
-			context->pipe->screen,
-			context->states.mc.textures[b + 1],
-			0, 0, 0, PIPE_BUFFER_USAGE_CPU_WRITE
-		);
+			(
+				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);
 		
-		if ((coded_block_pattern >> (b + 4)) & 1)
-		{
+		if ((coded_block_pattern >> (1 - tb)) & 1)
+		{			
 			if (sample_type == VL_FULL_SAMPLE)
-			{
-				for (y = 0; y < tex_surface->height; ++y)
-					memcpy
-					(
-						texels + y * tex_surface->pitch,
-						blocks + VL_BLOCK_SIZE * (b + 4) + y * VL_BLOCK_WIDTH,
-						VL_BLOCK_WIDTH * 2
-					);
-			}
+				vlGrabFrameCodedFullBlock
+				(
+					blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT,
+					texels,
+					tex_surface->pitch
+				);
 			else
-			{
-				for (y = 0; y < tex_surface->height; ++y)
-					for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-						texels[y * tex_surface->pitch + x] =
-						blocks[VL_BLOCK_SIZE * (b + 4) + y * VL_BLOCK_WIDTH + x] + 0x100;
-			}
+				vlGrabFrameCodedDiffBlock
+				(
+					blocks + sb * VL_BLOCK_WIDTH * VL_BLOCK_HEIGHT,
+					texels,
+					tex_surface->pitch
+				);
+			
+			++sb;
 		}
 		else
-		{
-			for (y = 0; y < tex_surface->height; ++y)
-			{
-				for (x = 0; x < VL_BLOCK_WIDTH; ++x)
-					texels[y * tex_surface->pitch + x] = 0x100;
-			}
-		}
+			vlGrabNoBlock(texels, tex_surface->pitch);
 		
 		pipe_surface_unmap(tex_surface);
 	}
@@ -229,41 +255,35 @@ int vlRenderIMacroBlock
 )
 {
 	struct pipe_context	*pipe;
-	struct VL_MC_VS_CONSTS	*vscbdata;
+	struct VL_MC_VS_CONSTS	*vs_consts;
 	
 	assert(blocks);
 	assert(surface);
 	
 	/* TODO: Implement interlaced rendering */
-	/*assert(picture_type == VL_FRAME_PICTURE);*/
 	if (picture_type != VL_FRAME_PICTURE)
-	{
-		/*fprintf(stderr, "field picture (I) unimplemented, ignoring\n");*/
 		return 0;
-	}
 	
 	pipe = surface->context->pipe;
 	
-	vscbdata = pipe->winsys->buffer_map
+	vs_consts = pipe->winsys->buffer_map
 	(
 		pipe->winsys,
 		surface->context->states.mc.vs_const_buf.buffer,
 		PIPE_BUFFER_USAGE_CPU_WRITE
 	);
 	
-	vscbdata->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->width;
-	vscbdata->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->height;
-	vscbdata->scale.z = 1.0f;
-	vscbdata->scale.w = 1.0f;
-	vscbdata->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->width;
-	vscbdata->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->height;
-	vscbdata->mb_pos_trans.z = 0.0f;
-	vscbdata->mb_pos_trans.w = 0.0f;
+	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);
 	
-	vlGrabBlocks(surface->context, coded_block_pattern, dct_type, VL_FULL_SAMPLE, blocks);
-	
 	surface->context->states.mc.render_target.cbufs[0] = pipe->screen->get_tex_surface
 	(
 		pipe->screen,
@@ -276,6 +296,8 @@ int vlRenderIMacroBlock
 	pipe->bind_vs_state(pipe, surface->context->states.mc.i_vs);
 	pipe->bind_fs_state(pipe, surface->context->states.mc.i_fs);
 	
+	vlGrabBlocks(surface->context, coded_block_pattern, dct_type, VL_FULL_SAMPLE, blocks);
+	
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
 	
 	return 0;
@@ -297,7 +319,7 @@ int vlRenderPMacroBlock
 )
 {
 	struct pipe_context	*pipe;
-	struct VL_MC_VS_CONSTS	*vscbdata;
+	struct VL_MC_VS_CONSTS	*vs_consts;
 	
 	assert(motion_vectors);
 	assert(blocks);
@@ -305,46 +327,55 @@ int vlRenderPMacroBlock
 	assert(surface);
 	
 	/* TODO: Implement interlaced rendering */
-	/*assert(picture_type == VL_FRAME_PICTURE);*/
 	if (picture_type != VL_FRAME_PICTURE)
-	{
-		/*fprintf(stderr, "field picture (P) unimplemented, ignoring\n");*/
 		return 0;
-	}
-	/* TODO: Implement field based motion compensation */
-	/*assert(mc_type == VL_FRAME_MC);*/
-	if (mc_type != VL_FRAME_MC)
-	{
-		/*fprintf(stderr, "field MC (P) unimplemented, ignoring\n");*/
+	/* TODO: Implement other MC types */
+	if (mc_type != VL_FRAME_MC && mc_type != VL_FIELD_MC)
 		return 0;
-	}
 	
 	pipe = surface->context->pipe;
 	
-	vscbdata = pipe->winsys->buffer_map
+	vs_consts = pipe->winsys->buffer_map
 	(
 		pipe->winsys,
 		surface->context->states.mc.vs_const_buf.buffer,
 		PIPE_BUFFER_USAGE_CPU_WRITE
 	);
 	
-	vscbdata->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->width;
-	vscbdata->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->height;
-	vscbdata->scale.z = 1.0f;
-	vscbdata->scale.w = 1.0f;
-	vscbdata->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->width;
-	vscbdata->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->height;
-	vscbdata->mb_pos_trans.z = 0.0f;
-	vscbdata->mb_pos_trans.w = 0.0f;
-	vscbdata->mb_tc_trans[0].x = (mbx * VL_MACROBLOCK_WIDTH + motion_vector->top_field.x * 0.5f) / (float)surface->width;
-	vscbdata->mb_tc_trans[0].y = (mby * VL_MACROBLOCK_HEIGHT + motion_vector->top_field.y * 0.5f) / (float)surface->height;
-	vscbdata->mb_tc_trans[0].z = 0.0f;
-	vscbdata->mb_tc_trans[0].w = 0.0f;
+	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);
 	
-	vlGrabBlocks(surface->context, coded_block_pattern, dct_type, VL_DIFFERENCE_SAMPLE, blocks);
-	
 	surface->context->states.mc.render_target.cbufs[0] = pipe->screen->get_tex_surface
 	(
 		pipe->screen,
@@ -356,8 +387,8 @@ int vlRenderPMacroBlock
 	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->bind_vs_state(pipe, surface->context->states.mc.p_vs);
-	pipe->bind_fs_state(pipe, surface->context->states.mc.p_fs);
+	
+	vlGrabBlocks(surface->context, coded_block_pattern, dct_type, VL_DIFFERENCE_SAMPLE, blocks);
 	
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
 	
@@ -381,7 +412,7 @@ int vlRenderBMacroBlock
 )
 {
 	struct pipe_context	*pipe;
-	struct VL_MC_VS_CONSTS	*vscbdata;
+	struct VL_MC_VS_CONSTS	*vs_consts;
 	
 	assert(motion_vectors);
 	assert(blocks);
@@ -389,50 +420,63 @@ int vlRenderBMacroBlock
 	assert(surface);
 	
 	/* TODO: Implement interlaced rendering */
-	/*assert(picture_type == VL_FRAME_PICTURE);*/
 	if (picture_type != VL_FRAME_PICTURE)
-	{
-		/*fprintf(stderr, "field picture (B) unimplemented, ignoring\n");*/
 		return 0;
-	}
-	/* TODO: Implement field based motion compensation */
-	/*assert(mc_type == VL_FRAME_MC);*/
-	if (mc_type != VL_FRAME_MC)
-	{
-		/*fprintf(stderr, "field MC (B) unimplemented, ignoring\n");*/
+	/* TODO: Implement other MC types */
+	if (mc_type != VL_FRAME_MC && mc_type != VL_FIELD_MC)
 		return 0;
-	}
 	
 	pipe = surface->context->pipe;
 	
-	vscbdata = pipe->winsys->buffer_map
+	vs_consts = pipe->winsys->buffer_map
 	(
 		pipe->winsys,
 		surface->context->states.mc.vs_const_buf.buffer,
 		PIPE_BUFFER_USAGE_CPU_WRITE
 	);
 	
-	vscbdata->scale.x = VL_MACROBLOCK_WIDTH / (float)surface->width;
-	vscbdata->scale.y = VL_MACROBLOCK_HEIGHT / (float)surface->height;
-	vscbdata->scale.z = 1.0f;
-	vscbdata->scale.w = 1.0f;
-	vscbdata->mb_pos_trans.x = (mbx * VL_MACROBLOCK_WIDTH) / (float)surface->width;
-	vscbdata->mb_pos_trans.y = (mby * VL_MACROBLOCK_HEIGHT) / (float)surface->height;
-	vscbdata->mb_pos_trans.z = 0.0f;
-	vscbdata->mb_pos_trans.w = 0.0f;
-	vscbdata->mb_tc_trans[0].x = (mbx * VL_MACROBLOCK_WIDTH + motion_vector[0].top_field.x * 0.5f) / (float)surface->width;
-	vscbdata->mb_tc_trans[0].y = (mby * VL_MACROBLOCK_HEIGHT + motion_vector[0].top_field.y * 0.5f) / (float)surface->height;
-	vscbdata->mb_tc_trans[0].z = 0.0f;
-	vscbdata->mb_tc_trans[0].w = 0.0f;
-	vscbdata->mb_tc_trans[1].x = (mbx * VL_MACROBLOCK_WIDTH + motion_vector[1].top_field.x * 0.5f) / (float)surface->width;
-	vscbdata->mb_tc_trans[1].y = (mby * VL_MACROBLOCK_HEIGHT + motion_vector[1].top_field.y * 0.5f) / (float)surface->height;
-	vscbdata->mb_tc_trans[1].z = 0.0f;
-	vscbdata->mb_tc_trans[1].w = 0.0f;
+	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);
 	
-	vlGrabBlocks(surface->context, coded_block_pattern, dct_type, VL_DIFFERENCE_SAMPLE, blocks);
-	
 	surface->context->states.mc.render_target.cbufs[0] = pipe->screen->get_tex_surface
 	(
 		pipe->screen,
@@ -445,8 +489,8 @@ int vlRenderBMacroBlock
 	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->bind_vs_state(pipe, surface->context->states.mc.b_vs);
-	pipe->bind_fs_state(pipe, surface->context->states.mc.b_fs);
+	
+	vlGrabBlocks(surface->context, coded_block_pattern, dct_type, VL_DIFFERENCE_SAMPLE, blocks);
 	
 	pipe->draw_arrays(pipe, PIPE_PRIM_TRIANGLES, 0, 24);
 	
@@ -513,7 +557,7 @@ int vlPutSurface
 			destw,
 			desth,
 			PIPE_FORMAT_A8R8G8B8_UNORM,
-			/*XXX: SoftPipe doesn't change GPU usage to CPU like it does for textures */
+			/* 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
 		);
diff --git a/src/gallium/state_trackers/g3dvl/vl_types.h b/src/gallium/state_trackers/g3dvl/vl_types.h
index 7040b745036..97753699db6 100644
--- a/src/gallium/state_trackers/g3dvl/vl_types.h
+++ b/src/gallium/state_trackers/g3dvl/vl_types.h
@@ -36,7 +36,9 @@ enum VL_SAMPLE_TYPE
 enum VL_MC_TYPE
 {
 	VL_FIELD_MC,
-	VL_FRAME_MC
+	VL_FRAME_MC,
+	VL_DUAL_PRIME_MC,
+	VL_16x8_MC = VL_FRAME_MC
 };
 
 struct VL_VERTEX4F
@@ -58,13 +60,19 @@ struct VL_MC_VS_CONSTS
 {
 	struct VL_VERTEX4F	scale;
 	struct VL_VERTEX4F	mb_pos_trans;
-	struct VL_VERTEX4F	mb_tc_trans[2];
+	struct VL_VERTEX4F	denorm;
+	struct
+	{
+		struct VL_VERTEX4F	top_field;
+		struct VL_VERTEX4F	bottom_field;
+	} mb_tc_trans[2];
 };
 
 struct VL_MC_FS_CONSTS
 {
 	struct VL_VERTEX4F	multiplier;
 	struct VL_VERTEX4F	bias;
+	struct VL_VERTEX4F	y_divider;
 };
 
 struct VL_CSC_FS_CONSTS
diff --git a/src/libXvMC/surface.c b/src/libXvMC/surface.c
index e4602d82042..5656895650e 100644
--- a/src/libXvMC/surface.c
+++ b/src/libXvMC/surface.c
@@ -32,6 +32,34 @@ static enum VL_PICTURE PictureToVL(int xvmc_pic)
 	return vl_pic;
 }
 
+static enum VL_MC_TYPE 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;
+		}
+		case XVMC_PREDICTION_FIELD:
+		{
+			vl_mc_type = VL_FIELD_MC;
+			break;
+		}
+		case XVMC_PREDICTION_DUAL_PRIME:
+		{
+			vl_mc_type = VL_DUAL_PRIME_MC;
+			break;
+		}
+		default:
+			assert(0);
+	}
+	
+	return vl_mc_type;
+}
+
 Status XvMCCreateSurface(Display *display, XvMCContext *context, XvMCSurface *surface)
 {
 	struct VL_CONTEXT *vl_ctx;
@@ -155,7 +183,7 @@ Status XvMCRenderSurface
 				flags == XVMC_SECOND_FIELD ? VL_FIELD_SECOND : VL_FIELD_FIRST,
 				macroblocks->macro_blocks[i].x,
 				macroblocks->macro_blocks[i].y,
-				macroblocks->macro_blocks[i].motion_type == XVMC_PREDICTION_FRAME ? VL_FIELD_MC : VL_FRAME_MC,
+				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,
@@ -188,7 +216,7 @@ Status XvMCRenderSurface
 				flags == XVMC_SECOND_FIELD ? VL_FIELD_SECOND : VL_FIELD_FIRST,
 				macroblocks->macro_blocks[i].x,
 				macroblocks->macro_blocks[i].y,
-				macroblocks->macro_blocks[i].motion_type == XVMC_PREDICTION_FRAME ? VL_FIELD_MC : VL_FRAME_MC,
+				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,
@@ -233,7 +261,7 @@ Status XvMCRenderSurface
 				flags == XVMC_SECOND_FIELD ? VL_FIELD_SECOND : VL_FIELD_FIRST,
 				macroblocks->macro_blocks[i].x,
 				macroblocks->macro_blocks[i].y,
-				macroblocks->macro_blocks[i].motion_type == XVMC_PREDICTION_FRAME ? VL_FIELD_MC : VL_FRAME_MC,
+				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,