Add the bidirectional MV prediction for B-frame on Haswell

Now the B-frame only supports the forward prediction. In order to
get the better prediction result, the bidirection prediction is added
for B-frame. At the same time the MV prediction is also added.

Signed-off-by: Zhao Yakui <yakui.zhao@intel.com>

8 files changed, 1299 insertions, 15 deletions

diff --git a/src/gen6_mfc_common.c b/src/gen6_mfc_common.c
index 2ece62bf..1cab68cf 100644
--- a/src/gen6_mfc_common.c
+++ b/src/gen6_mfc_common.c
@@ -789,7 +789,7 @@ void intel_vme_update_mbmv_cost(VADriverContextP ctx,
 		vme_state_message[MODE_INTER_8X8] = 0x4a;
 		vme_state_message[MODE_INTER_8X4] = 0x4a;
 		vme_state_message[MODE_INTER_4X4] = 0x4a;
-		vme_state_message[MODE_INTER_BWD] = 0x4a;
+		vme_state_message[MODE_INTER_BWD] = 0x2a;
 		return;	
 	}
 	m_costf = lambda * 10;
@@ -836,7 +836,7 @@ void intel_vme_update_mbmv_cost(VADriverContextP ctx,
 		m_costf = lambda * 6.5;
 		m_cost = m_costf;
 		vme_state_message[MODE_INTER_4X4] = intel_format_lutvalue(m_cost, 0x6f);
-		m_costf = lambda * 2.5;
+		m_costf = lambda * 1.5;
 		m_cost = m_costf;
 		vme_state_message[MODE_INTER_BWD] = intel_format_lutvalue(m_cost, 0x6f);
 	}
diff --git a/src/gen75_mfc.c b/src/gen75_mfc.c
index 184fcb7a..dd8bf715 100644
--- a/src/gen75_mfc.c
+++ b/src/gen75_mfc.c
@@ -259,11 +259,13 @@ gen75_mfc_avc_img_state(VADriverContextP ctx, struct encode_state *encode_state,
     BEGIN_BCS_BATCH(batch, 16);
 
     OUT_BCS_BATCH(batch, MFX_AVC_IMG_STATE | (16 - 2));
+	/*DW1. MB setting of frame */
     OUT_BCS_BATCH(batch,
                   ((width_in_mbs * height_in_mbs) & 0xFFFF));
     OUT_BCS_BATCH(batch, 
                   ((height_in_mbs - 1) << 16) | 
                   ((width_in_mbs - 1) << 0));
+	/* DW3 QP setting */
     OUT_BCS_BATCH(batch, 
                   (0 << 24) |	/* Second Chroma QP Offset */
                   (0 << 16) |	/* Chroma QP Offset */
@@ -289,17 +291,22 @@ gen75_mfc_avc_img_state(VADriverContextP ctx, struct encode_state *encode_state,
                   (1 << 2)  |   /* Frame MB only flag */
                   (0 << 1)  |   /* MBAFF mode is in active */
                   (0 << 0));    /* Field picture flag */
+	/* DW5 Trellis quantization */
     OUT_BCS_BATCH(batch, 0);    /* Mainly about MB rate control and debug, just ignoring */
     OUT_BCS_BATCH(batch,        /* Inter and Intra Conformance Max size limit */
                   (0xBB8 << 16) |       /* InterMbMaxSz */
                   (0xEE8) );            /* IntraMbMaxSz */
     OUT_BCS_BATCH(batch, 0);            /* Reserved */
+	/* DW8. QP delta */
     OUT_BCS_BATCH(batch, 0);            /* Slice QP Delta for bitrate control */
-    OUT_BCS_BATCH(batch, 0);            /* Slice QP Delta for bitrate control */	
+    OUT_BCS_BATCH(batch, 0);            /* Slice QP Delta for bitrate control */
+	/* DW10. Bit setting for MB */	
     OUT_BCS_BATCH(batch, 0x8C000000);
     OUT_BCS_BATCH(batch, 0x00010000);
+	/* DW12. */
     OUT_BCS_BATCH(batch, 0);
-    OUT_BCS_BATCH(batch, 0);
+    OUT_BCS_BATCH(batch, 0x02010100);
+	/* DW14. For short format */
     OUT_BCS_BATCH(batch, 0);
     OUT_BCS_BATCH(batch, 0);
 
@@ -893,6 +900,7 @@ gen75_mfc_avc_slice_state(VADriverContextP ctx,
     int bit_rate_control_target, maxQpN, maxQpP;
     unsigned char correct[6], grow, shrink;
     int i;
+    int bslice = 0;
     int weighted_pred_idc = 0;
     unsigned int luma_log2_weight_denom = slice_param->luma_log2_weight_denom;
     unsigned int chroma_log2_weight_denom = slice_param->chroma_log2_weight_denom;
@@ -910,6 +918,7 @@ gen75_mfc_avc_slice_state(VADriverContextP ctx,
         weighted_pred_idc = pic_param->pic_fields.bits.weighted_pred_flag;
     } else if (slice_type == SLICE_TYPE_B) {
         weighted_pred_idc = pic_param->pic_fields.bits.weighted_bipred_idc;
+	bslice = 1;
 
         if (weighted_pred_idc == 2) {
             /* 8.4.3 - Derivation process for prediction weights (8-279) */
@@ -938,7 +947,7 @@ gen75_mfc_avc_slice_state(VADriverContextP ctx,
         OUT_BCS_BATCH(batch, 0);			/*no reference frames and pred_weight_table*/
     } else {
         OUT_BCS_BATCH(batch,
-                      (1 << 16) |        	        /*1 reference frame*/
+                      (1 << 16) | (bslice << 24) |        	        /*1 reference frame*/
                       (chroma_log2_weight_denom << 8) |
                       (luma_log2_weight_denom << 0));
     }
diff --git a/src/gen75_vme.c b/src/gen75_vme.c
index 60eb19a1..b53155d9 100644
--- a/src/gen75_vme.c
+++ b/src/gen75_vme.c
@@ -55,6 +55,7 @@
 
 #define VME_INTRA_SHADER        0
 #define VME_INTER_SHADER        1
+#define VME_BINTER_SHADER	3
 #define VME_BATCHBUFFER         2
 
 #define CURBE_ALLOCATION_SIZE   37              /* in 256-bit */
@@ -75,6 +76,10 @@ static const uint32_t gen75_vme_inter_frame[][4] = {
 #include "shaders/vme/inter_frame_haswell.g75b"
 };
 
+static const uint32_t gen75_vme_inter_bframe[][4] = {
+#include "shaders/vme/inter_bframe_haswell.g75b"
+};
+
 static const uint32_t gen75_vme_batchbuffer[][4] = {
 #include "shaders/vme/batchbuffer.g75b"
 };
@@ -101,6 +106,13 @@ static struct i965_kernel gen75_vme_kernels[] = {
         sizeof(gen75_vme_batchbuffer),
         NULL
     },
+    {
+        "VME inter BFrame",
+        VME_BINTER_SHADER,
+        gen75_vme_inter_bframe,
+        sizeof(gen75_vme_inter_bframe),
+        NULL
+    }
 };
 
 static const uint32_t gen75_vme_mpeg2_intra_frame[][4] = {
@@ -305,7 +317,7 @@ static VAStatus gen75_vme_interface_setup(VADriverContextP ctx,
     assert(bo->virtual);
     desc = bo->virtual;
 
-    for (i = 0; i < GEN6_VME_KERNEL_NUMBER; i++) {
+    for (i = 0; i < vme_context->vme_kernel_sum; i++) {
         struct i965_kernel *kernel;
         kernel = &vme_context->gpe_context.kernels[i];
         assert(sizeof(*desc) == 32);
@@ -738,6 +750,7 @@ static void gen75_vme_pipeline_programing(VADriverContextP ctx,
     int is_intra = pSliceParameter->slice_type == SLICE_TYPE_I;
     int width_in_mbs = pSequenceParameter->picture_width_in_mbs;
     int height_in_mbs = pSequenceParameter->picture_height_in_mbs;
+    int kernel_shader;
     bool allow_hwscore = true;
     int s;
 
@@ -748,19 +761,29 @@ static void gen75_vme_pipeline_programing(VADriverContextP ctx,
 		break;
 	}
     }
-
+	if ((pSliceParameter->slice_type == SLICE_TYPE_I) ||
+		(pSliceParameter->slice_type == SLICE_TYPE_I)) {
+		kernel_shader = VME_INTRA_SHADER;
+	} else if ((pSliceParameter->slice_type == SLICE_TYPE_P) ||
+		(pSliceParameter->slice_type == SLICE_TYPE_SP)) {
+		kernel_shader = VME_INTER_SHADER;
+	} else {
+		kernel_shader = VME_BINTER_SHADER;
+		if (!allow_hwscore)
+			kernel_shader = VME_INTER_SHADER;
+	}
     if (allow_hwscore)
 	gen75_vme_walker_fill_vme_batchbuffer(ctx, 
                                   encode_state,
                                   width_in_mbs, height_in_mbs,
-                                  is_intra ? VME_INTRA_SHADER : VME_INTER_SHADER,
+                                  kernel_shader,
                                   pPicParameter->pic_fields.bits.transform_8x8_mode_flag,
                                   encoder_context);
     else
 	gen75_vme_fill_vme_batchbuffer(ctx, 
                                    encode_state,
                                    width_in_mbs, height_in_mbs,
-                                   is_intra ? VME_INTRA_SHADER : VME_INTER_SHADER,
+                                   kernel_shader,
                                    pPicParameter->pic_fields.bits.transform_8x8_mode_flag,
                                    encoder_context);
 
@@ -1122,6 +1145,7 @@ Bool gen75_vme_context_init(VADriverContextP ctx, struct intel_encoder_context *
 {
     struct gen6_vme_context *vme_context = calloc(1, sizeof(struct gen6_vme_context));
     struct i965_kernel *vme_kernel_list = NULL;
+	int i965_kernel_num;
 
     switch (encoder_context->profile) {
     case VAProfileH264Baseline:
@@ -1129,13 +1153,14 @@ Bool gen75_vme_context_init(VADriverContextP ctx, struct intel_encoder_context *
     case VAProfileH264High:
         vme_kernel_list = gen75_vme_kernels;
         encoder_context->vme_pipeline = gen75_vme_pipeline;
-        
+       	i965_kernel_num = sizeof(gen75_vme_kernels) / sizeof(struct i965_kernel); 
         break;
 
     case VAProfileMPEG2Simple:
     case VAProfileMPEG2Main:
         vme_kernel_list = gen75_vme_mpeg2_kernels;
         encoder_context->vme_pipeline = gen75_vme_mpeg2_pipeline;
+       	i965_kernel_num = sizeof(gen75_vme_mpeg2_kernels) / sizeof(struct i965_kernel); 
 
         break;
 
@@ -1145,7 +1170,7 @@ Bool gen75_vme_context_init(VADriverContextP ctx, struct intel_encoder_context *
 
         break;
     }
-
+	vme_context->vme_kernel_sum = i965_kernel_num;
     vme_context->gpe_context.surface_state_binding_table.length = (SURFACE_STATE_PADDED_SIZE + sizeof(unsigned int)) * MAX_MEDIA_SURFACES_GEN6;
 
     vme_context->gpe_context.idrt.max_entries = MAX_INTERFACE_DESC_GEN6;
@@ -1181,7 +1206,7 @@ Bool gen75_vme_context_init(VADriverContextP ctx, struct intel_encoder_context *
     i965_gpe_load_kernels(ctx,
                           &vme_context->gpe_context,
                           vme_kernel_list,
-                          GEN6_VME_KERNEL_NUMBER);
+                          i965_kernel_num);
     vme_context->vme_surface2_setup = gen7_gpe_surface2_setup;
     vme_context->vme_media_rw_surface_setup = gen7_gpe_media_rw_surface_setup;
     vme_context->vme_buffer_suface_setup = gen7_gpe_buffer_suface_setup;
diff --git a/src/shaders/vme/Makefile.am b/src/shaders/vme/Makefile.am
index 5ccf6d22..cfe8f5f3 100644
--- a/src/shaders/vme/Makefile.am
+++ b/src/shaders/vme/Makefile.am
@@ -1,5 +1,5 @@
 VME_CORE	= batchbuffer.asm intra_frame.asm inter_frame.asm
-VME75_CORE	= batchbuffer.asm intra_frame_haswell.asm inter_frame_haswell.asm
+VME75_CORE	= batchbuffer.asm intra_frame_haswell.asm inter_frame_haswell.asm inter_bframe_haswell.asm
 
 INTEL_G6B	= batchbuffer.g6b intra_frame.g6b inter_frame.g6b
 INTEL_G6A	= batchbuffer.g6a intra_frame.g6a inter_frame.g6a
@@ -11,8 +11,8 @@ INTEL_G7A	= batchbuffer.g7a intra_frame.g7a inter_frame.g7a mpeg2_inter_frame.g7
 INTEL_GEN7_INC	= batchbuffer.inc vme.inc vme7_mpeg2.inc
 INTEL_GEN7_ASM	= $(INTEL_G7A:%.g7a=%.gen7.asm)
 
-INTEL_G75B	= batchbuffer.g75b intra_frame_haswell.g75b inter_frame_haswell.g75b mpeg2_inter_frame_haswell.g75b
-INTEL_G75A	= batchbuffer.g75a intra_frame_haswell.g75a inter_frame_haswell.g75a mpeg2_inter_frame_haswell.g75a
+INTEL_G75B	= batchbuffer.g75b intra_frame_haswell.g75b inter_frame_haswell.g75b mpeg2_inter_frame_haswell.g75b inter_bframe_haswell.g75b
+INTEL_G75A	= batchbuffer.g75a intra_frame_haswell.g75a inter_frame_haswell.g75a mpeg2_inter_frame_haswell.g75a inter_bframe_haswell.g75a
 INTEL_GEN75_INC	= batchbuffer.inc vme75.inc vme75_mpeg2.inc
 INTEL_GEN75_ASM	= $(INTEL_G75A:%.g75a=%.gen75.asm)
 
diff --git a/src/shaders/vme/inter_bframe_haswell.asm b/src/shaders/vme/inter_bframe_haswell.asm
new file mode 100644
index 00000000..9ab35d32
--- /dev/null
+++ b/src/shaders/vme/inter_bframe_haswell.asm
@@ -0,0 +1,828 @@
+/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * This program is licensed under the terms and conditions of the
+ * Eclipse Public License (EPL), version 1.0.  The full text of the EPL is at
+ * http://www.opensource.org/licenses/eclipse-1.0.php.
+ * Authors: Zhao Yakui <yakui.zhao@intel.com>
+ */
+// Modual name: Inter_bframe_haswell.asm
+//
+// Make intra predition estimation for Inter frame for B-frame
+//
+
+//
+//  Now, begin source code....
+//
+
+#define SAVE_RET	add (1) RETURN_REG<1>:ud   ip:ud	32:ud
+#define	RETURN		mov (1)	ip:ud	RETURN_REG<0,1,0>:ud
+
+/*
+ * __START
+ */
+__INTRA_START:
+mov  (16) tmp_reg0.0<1>:UD      0x0:UD {align1};
+mov  (16) tmp_reg2.0<1>:UD      0x0:UD {align1};
+mov  (16) tmp_reg4.0<1>:UD      0x0:UD {align1} ;
+mov  (16) tmp_reg6.0<1>:UD      0x0:UD {align1} ;
+
+shl  (2) read0_header.0<1>:D    orig_xy_ub<2,2,1>:UB 4:UW {align1};    /* (x, y) * 16 */
+add  (1) read0_header.0<1>:D    read0_header.0<0,1,0>:D -8:W {align1};     /* X offset */
+add  (1) read0_header.4<1>:D    read0_header.4<0,1,0>:D -1:W {align1};     /* Y offset */ 
+mov  (1) read0_header.8<1>:UD   BLOCK_32X1 {align1};
+mov  (1) read0_header.20<1>:UB  thread_id_ub {align1};                  /* dispatch id */
+
+shl  (2) read1_header.0<1>:D    orig_xy_ub<2,2,1>:UB 4:UW {align1};    /* (x, y) * 16 */
+add  (1) read1_header.0<1>:D    read1_header.0<0,1,0>:D -4:W {align1};     /* X offset */
+mov  (1) read1_header.8<1>:UD   BLOCK_4X16 {align1};
+mov  (1) read1_header.20<1>:UB  thread_id_ub {align1};                  /* dispatch id */
+        
+shl  (2) vme_m0.8<1>:UW         orig_xy_ub<2,2,1>:UB 4:UW {align1};    /* (x, y) * 16 */
+mov  (1) vme_m0.20<1>:UB        thread_id_ub {align1};                  /* dispatch id */
+
+mul  (1) obw_m0.8<1>:UD         w_in_mb_uw<0,1,0>:UW orig_y_ub<0,1,0>:UB {align1};
+add  (1) obw_m0.8<1>:UD         obw_m0.8<0,1,0>:UD orig_x_ub<0,1,0>:UB {align1};
+mul  (1) obw_m0.8<1>:UD         obw_m0.8<0,1,0>:UD 24:UD {align1};
+mov  (1) obw_m0.20<1>:UB        thread_id_ub {align1};                  /* dispatch id */
+        
+/*
+ * Media Read Message -- fetch Luma neighbor edge pixels 
+ */
+/* ROW */
+mov  (8) msg_reg0.0<1>:UD       read0_header.0<8,8,1>:UD {align1};        
+send (8) msg_ind INEP_ROW<1>:UB null read(BIND_IDX_INEP, 0, 0, 4) mlen 1 rlen 1 {align1};
+
+/* COL */
+mov  (8) msg_reg0.0<1>:UD       read1_header.0<8,8,1>:UD {align1};                
+send (8) msg_ind INEP_COL0<1>:UB null read(BIND_IDX_INEP, 0, 0, 4) mlen 1 rlen 2 {align1};
+        
+/*
+ * Media Read Message -- fetch Chroma neighbor edge pixels 
+ */
+/* ROW */
+shl  (2) read0_header.0<1>:D    orig_xy_ub<2,2,1>:UB 3:UW {align1};    /* x * 16 , y * 8 */
+mul  (1) read0_header.0<1>:D    read0_header.0<0,1,0>:D  2:W {align1};
+add  (1) read0_header.0<1>:D    read0_header.0<0,1,0>:D -8:W {align1};     /* X offset */
+add  (1) read0_header.4<1>:D    read0_header.4<0,1,0>:D -1:W {align1};     /* Y offset */ 
+mov  (8) msg_reg0.0<1>:UD       read0_header.0<8,8,1>:UD {align1};        
+send (8) msg_ind CHROMA_ROW<1>:UB null read(BIND_IDX_CBCR, 0, 0, 4) mlen 1 rlen 1 {align1};
+
+/* COL */
+shl  (2) read1_header.0<1>:D    orig_xy_ub<2,2,1>:UB 3:UW {align1};    /* x * 16, y * 8 */
+mul  (1) read1_header.0<1>:D    read1_header.0<0,1,0>:D  2:W {align1};
+add  (1) read1_header.0<1>:D    read1_header.0<0,1,0>:D -4:W {align1};     /* X offset */
+mov  (1) read1_header.8<1>:UD   BLOCK_8X4 {align1};
+mov  (8) msg_reg0.0<1>:UD       read1_header.0<8,8,1>:UD {align1};                
+send (8) msg_ind CHROMA_COL<1>:UB null read(BIND_IDX_CBCR, 0, 0, 4) mlen 1 rlen 1 {align1};
+
+mov  (8) vme_m1.0<1>:ud		0:ud		{align1};
+mov  (8) mb_mvp_ref.0<1>:ud	0:ud		{align1};
+and.z.f0.0 (1)		null:uw	mb_hwdep<0,1,0>:uw		0x04:uw   {align1};
+(f0.0) jmpi (1) __mb_hwdep_end;
+
+/* read back the data for MB A */
+/* the layout of MB result is: rx.0(Available). rx.4(MVa), rX.8(MVb), rX.16(Pred_L0 flag),
+*  rX.18 (Pred_L1 flag), rX.20(Forward reference ID), rX.22(Backwared reference ID)
+*/
+mov  (8) mba_result.0<1>:ud	0x0:ud		{align1};
+mov  (8) mbb_result.0<1>:ud	0x0:ud		{align1};
+mov  (8) mbc_result.0<1>:ud	0x0:ud		{align1};
+mba_start:
+mov  (8) mb_msg0.0<1>:ud	0:ud		{align1};
+and.z.f0.0 (1)		null:uw	input_mb_intra_ub<0,1,0>:ub	INTRA_PRED_AVAIL_FLAG_AE:uw   {align1};
+/* MB A doesn't exist. Zero MV. mba_flag is zero and ref ID = -1 */
+(f0.0)  mov  (2)    	mba_result.20<1>:w	-1:w	{align1};
+(f0.0)  jmpi (1)	mbb_start;
+mov  (1) mba_result.0<1>:d	MB_AVAIL		{align1};	
+mov  (2) tmp_reg0.0<1>:UW	orig_xy_ub<2,2,1>:UB	{align1};
+add  (1) tmp_reg0.0<1>:w	tmp_reg0.0<0,1,0>:w	-1:w	{align1};
+mul  (1) mb_msg0.8<1>:UD       w_in_mb_uw<0,1,0>:UW tmp_reg0.2<0,1,0>:UW {align1};
+add  (1) mb_msg0.8<1>:UD       mb_msg0.8<0,1,0>:UD   tmp_reg0.0<0,1,0>:uw {align1};
+mul  (1) mb_msg0.8<1>:UD       mb_msg0.8<0,1,0>:UD 24:UD {align1};
+mov  (1) mb_msg0.20<1>:UB        thread_id_ub {align1};                  /* dispatch id */
+
+/* bind index 3, read 4 oword (64bytes), msg type: 0(OWord Block Read) */
+send (16)
+        mb_ind
+        mb_wb.0<1>:ud
+	NULL
+        data_port(
+                OBR_CACHE_TYPE,
+                OBR_MESSAGE_TYPE,
+                OBR_CONTROL_4,
+                OBR_BIND_IDX,
+                OBR_WRITE_COMMIT_CATEGORY,
+                OBR_HEADER_PRESENT
+        )
+        mlen 1
+        rlen 2
+        {align1};
+
+/* TODO: RefID is required after multi-references are added */
+cmp.l.f0.0 (1)		null:w	mb_intra_wb.16<0,1,0>:uw	mb_inter_wb.8<0,1,0>:uw {align1};
+(f0.0)   mov (2)	mba_result.20<1>:w			-1:w	{align1};
+(f0.0)   jmpi	(1)	mbb_start;
+
+add   (1) mb_msg0.8<1>:UD	mb_msg0.8<0,1,0>:ud	3:ud {align1};
+/* Read MV for MB A */
+/* bind index 3, read 8 oword (128bytes), msg type: 0(OWord Block Read) */
+send (16)
+        mb_ind
+        mb_mv0.0<1>:ud
+	NULL
+        data_port(
+                OBR_CACHE_TYPE,
+                OBR_MESSAGE_TYPE,
+                OBR_CONTROL_8,
+                OBR_BIND_IDX,
+                OBR_WRITE_COMMIT_CATEGORY,
+                OBR_HEADER_PRESENT
+        )
+        mlen 1
+        rlen 4
+        {align1};
+/* TODO: RefID is required after multi-references are added */
+/* MV */
+mov	(2)	mba_result.4<1>:ud		mb_mv1.8<2,2,1>:ud	{align1};
+mov  	(2)	mba_result.20<1>:w		-1:w		{align1};
+mov	(1)	INPUT_ARG0.0<1>:ud	mb_inter_wb.4<0,1,0>:ud	{align1};
+mov	(1)	INPUT_ARG0.4<1>:ud	mb_inter_wb.0<0,1,0>:ud	{align1};
+mov	(1)	INPUT_ARG0.8<1>:ud	INTER_BLOCK1:ud	{align1};
+SAVE_RET	{align1};
+jmpi	(1)	mb_pred_func;
+mov 	(1)	mb_pred_mode.0<1>:uw	RET_ARG<0,1,0>:uw	{align1};
+cmp.e.f0.0 (1)	null:uw		mb_pred_mode.0<0,1,0>:uw	PRED_L0 {align1};
+(f0.0)	mov   (1)	mba_result.16<1>:uw		MB_PRED_FLAG		{align1};
+(f0.0)	mov   (1)	mba_result.20<1>:w		0:w		{align1};
+(f0.0)  jmpi	(1) mbb_start;
+cmp.e.f0.0 (1)	null:uw		mb_pred_mode.0<0,1,0>:uw	PRED_L1 {align1};
+(f0.0)	mov   (1)	mba_result.18<1>:uw		MB_PRED_FLAG		{align1};
+(f0.0)	mov   (1)	mba_result.22<1>:w		0:w		{align1};
+(f0.0)  jmpi	(1) mbb_start;
+mov   (2)	mba_result.16<1>:uw		MB_PRED_FLAG		{align1};
+mov   (2)	mba_result.20<1>:w		0:w		{align1};
+
+mbb_start:
+mov  (8) mb_msg0.0<1>:ud	0:ud		{align1};
+and.z.f0.0 (1)		null:uw	input_mb_intra_ub<0,1,0>:ub	INTRA_PRED_AVAIL_FLAG_B:uw   {align1};
+/* MB B doesn't exist. Zero MV. mba_flag is zero */
+/* If MB B doesn't exist, neight of MB C nor D exists */
+(f0.0)  mov  (2)    	mbb_result.20<1>:w	-1:w		{align1};
+(f0.0)  mov  (2)    	mbc_result.20<1>:w	-1:w		{align1};
+(f0.0)  jmpi (1)	mb_mvp_start;
+mov  (1) mbb_result.0<1>:d	MB_AVAIL		{align1};	
+mov  (2) tmp_reg0.0<1>:UW	orig_xy_ub<2,2,1>:UB	{align1};
+add  (1) tmp_reg0.2<1>:w	tmp_reg0.2<0,1,0>:w	-1:w	{align1};
+mul  (1) mb_msg0.8<1>:UD       w_in_mb_uw<0,1,0>:UW tmp_reg0.2<0,1,0>:UW {align1};
+add  (1) mb_msg0.8<1>:UD       mb_msg0.8<0,1,0>:UD   tmp_reg0.0<0,1,0>:uw {align1};
+mul  (1) mb_msg0.8<1>:UD       mb_msg0.8<0,1,0>:UD 24:UD {align1};
+mov  (1) mb_msg0.20<1>:UB        thread_id_ub {align1};                  /* dispatch id */
+
+/* bind index 3, read 4 oword (64bytes), msg type: 0(OWord Block Read) */
+send (16)
+        mb_ind
+        mb_wb.0<1>:ud
+	NULL
+        data_port(
+                OBR_CACHE_TYPE,
+                OBR_MESSAGE_TYPE,
+                OBR_CONTROL_4,
+                OBR_BIND_IDX,
+                OBR_WRITE_COMMIT_CATEGORY,
+                OBR_HEADER_PRESENT
+        )
+        mlen 1
+        rlen 2
+        {align1};
+
+/* TODO: RefID is required after multi-references are added */
+cmp.l.f0.0 (1)		null:w	mb_intra_wb.16<0,1,0>:uw	mb_inter_wb.8<0,1,0>:uw {align1};
+(f0.0)   mov (2)	mbb_result.20<1>:w			-1:w	{align1};
+(f0.0)   jmpi	(1)	mbc_start;
+add   (1) mb_msg0.8<1>:UD	mb_msg0.8<0,1,0>:ud	3:ud {align1};
+/* Read MV for MB B */
+/* bind index 3, read 8 oword (128bytes), msg type: 0(OWord Block Read) */
+send (16)
+        mb_ind
+        mb_mv0.0<1>:ud
+	NULL
+        data_port(
+                OBR_CACHE_TYPE,
+                OBR_MESSAGE_TYPE,
+                OBR_CONTROL_8,
+                OBR_BIND_IDX,
+                OBR_WRITE_COMMIT_CATEGORY,
+                OBR_HEADER_PRESENT
+        )
+        mlen 1
+        rlen 4
+        {align1};
+/* TODO: RefID is required after multi-references are added */
+mov	   (2)		mbb_result.4<1>:ud		mb_mv2.16<2,2,1>:ud	{align1};
+mov	   (2)		mbb_result.20<1>:w		-1:w	{align1};
+mov	(1)	INPUT_ARG0.0<1>:ud	mb_inter_wb.4<0,1,0>:ud	{align1};
+mov	(1)	INPUT_ARG0.4<1>:ud	mb_inter_wb.0<0,1,0>:ud	{align1};
+mov	(1)	INPUT_ARG0.8<1>:ud	INTER_BLOCK2:ud	{align1};
+SAVE_RET	{align1};
+jmpi	(1)	mb_pred_func;
+mov 	(1)	mb_pred_mode.0<1>:uw	RET_ARG<0,1,0>:uw	{align1};
+cmp.e.f0.0 (1)	null:uw		mb_pred_mode.0<0,1,0>:uw	PRED_L0 {align1};
+(f0.0)	mov   (1)	mbb_result.16<1>:uw		MB_PRED_FLAG		{align1};
+(f0.0)	mov   (1)	mbb_result.20<1>:w		0:w		{align1};
+(f0.0)  jmpi	(1) mbc_start;
+cmp.e.f0.0 (1)	null:uw		mb_pred_mode.0<0,1,0>:uw	PRED_L1 {align1};
+(f0.0)	mov   (1)	mbb_result.18<1>:uw		MB_PRED_FLAG		{align1};
+(f0.0)	mov   (1)	mbb_result.22<1>:w		0:w		{align1};
+(f0.0)  jmpi	(1) mbc_start;
+mov   (2)	mbb_result.16<1>:uw		MB_PRED_FLAG		{align1};
+mov   (2)	mbb_result.20<1>:w		0:w		{align1};
+
+mbc_start:
+mov  (8) mb_msg0.0<1>:ud	0:ud		{align1};
+and.z.f0.0 (1)		null:uw	input_mb_intra_ub<0,1,0>:ub	INTRA_PRED_AVAIL_FLAG_C:uw   {align1};
+/* MB C doesn't exist. Zero MV. mba_flag is zero */
+/* Based on h264 spec the MB D will be replaced if MB C doesn't exist */
+(f0.0)  jmpi (1)	mbd_start;
+mov  (1) mbc_result.0<1>:d	MB_AVAIL		{align1};	
+mov  (2) tmp_reg0.0<1>:UW	orig_xy_ub<2,2,1>:UB	{align1};
+add  (1) tmp_reg0.2<1>:w	tmp_reg0.2<0,1,0>:w	-1:w	{align1};
+add  (1) tmp_reg0.0<1>:w	tmp_reg0.0<0,1,0>:w	1:w	{align1};
+mul  (1) mb_msg0.8<1>:UD       w_in_mb_uw<0,1,0>:UW tmp_reg0.2<0,1,0>:UW {align1};
+add  (1) mb_msg0.8<1>:UD       mb_msg0.8<0,1,0>:UD   tmp_reg0.0<0,1,0>:uw {align1};
+mul  (1) mb_msg0.8<1>:UD       mb_msg0.8<0,1,0>:UD 24:UD {align1};
+mov  (1) mb_msg0.20<1>:UB        thread_id_ub {align1};                  /* dispatch id */
+
+/* bind index 3, read 4 oword (64bytes), msg type: 0(OWord Block Read) */
+send (16)
+        mb_ind
+        mb_wb.0<1>:ud
+	NULL
+        data_port(
+                OBR_CACHE_TYPE,
+                OBR_MESSAGE_TYPE,
+                OBR_CONTROL_4,
+                OBR_BIND_IDX,
+                OBR_WRITE_COMMIT_CATEGORY,
+                OBR_HEADER_PRESENT
+        )
+        mlen 1
+        rlen 2
+        {align1};
+
+/* TODO: RefID is required after multi-references are added */
+cmp.l.f0.0 (1)		null:w	mb_intra_wb.16<0,1,0>:uw	mb_inter_wb.8<0,1,0>:uw {align1};
+(f0.0)   mov (2)	mbc_result.20<1>:w			-1:w	{align1};
+(f0.0)   jmpi	(1)	mb_mvp_start;
+add   (1) mb_msg0.8<1>:UD	mb_msg0.8<0,1,0>:ud	3:ud {align1};
+/* Read MV for MB C */
+/* bind index 3, read 8 oword (128bytes), msg type: 0(OWord Block Read) */
+send (16)
+        mb_ind
+        mb_mv0.0<1>:ud
+	NULL
+        data_port(
+                OBR_CACHE_TYPE,
+                OBR_MESSAGE_TYPE,
+                OBR_CONTROL_8,
+                OBR_BIND_IDX,
+                OBR_WRITE_COMMIT_CATEGORY,
+                OBR_HEADER_PRESENT
+        )
+        mlen 1
+        rlen 4
+        {align1};
+/* TODO: RefID is required after multi-references are added */
+/* Forward MV */
+mov	   (2)		mbc_result.4<1>:ud		mb_mv2.16<2,2,1>:ud	{align1};
+mov	   (2)		mbc_result.20<1>:w		-1:w	{align1};
+mov	(1)	INPUT_ARG0.0<1>:ud	mb_inter_wb.4<0,1,0>:ud	{align1};
+mov	(1)	INPUT_ARG0.4<1>:ud	mb_inter_wb.0<0,1,0>:ud	{align1};
+mov	(1)	INPUT_ARG0.8<1>:ud	INTER_BLOCK2:ud	{align1};
+SAVE_RET	{align1};
+jmpi	(1)	mb_pred_func;
+mov 	(1)	mb_pred_mode.0<1>:uw	RET_ARG<0,1,0>:uw	{align1};
+cmp.e.f0.0 (1)	null:uw		mb_pred_mode.0<0,1,0>:uw	PRED_L0 {align1};
+(f0.0)	mov    (1) mbc_result.16<1>:uw		MB_PRED_FLAG		{align1};
+(f0.0)	mov    (1) mbc_result.20<1>:w		0:w		{align1};
+(f0.0)  jmpi   (1) mb_mvp_start;
+cmp.e.f0.0 (1)	null:uw		mb_pred_mode.0<0,1,0>:uw	PRED_L1 {align1};
+(f0.0)	mov   (1)	mbc_result.18<1>:uw		MB_PRED_FLAG		{align1};
+(f0.0)	mov    (1) mbc_result.22<1>:w		0:w		{align1};
+(f0.0)  jmpi  (1) mb_mvp_start;
+mov  (2)  mbc_result.16<1>:uw		MB_PRED_FLAG		{align1};
+mov  (2)  mbc_result.20<1>:w		0:w	{align1};
+
+jmpi   (1)    mb_mvp_start;
+mbd_start:
+mov  (1) mbc_result.0<1>:d	MB_AVAIL		{align1};	
+mov  (2) tmp_reg0.0<1>:UW	orig_xy_ub<2,2,1>:UB	{align1};
+add  (2) tmp_reg0.0<1>:w	tmp_reg0.0<2,2,1>:w	-1:w	{align1};
+mul  (1) mb_msg0.8<1>:UD       w_in_mb_uw<0,1,0>:UW tmp_reg0.2<0,1,0>:UW {align1};
+add  (1) mb_msg0.8<1>:UD       mb_msg0.8<0,1,0>:UD   tmp_reg0.0<0,1,0>:uw {align1};
+mul  (1) mb_msg0.8<1>:UD       mb_msg0.8<0,1,0>:UD 24:UD {align1};
+mov  (1) mb_msg0.20<1>:UB        thread_id_ub {align1};                  /* dispatch id */
+
+/* bind index 3, read 4 oword (64bytes), msg type: 0(OWord Block Read) */
+send (16)
+        mb_ind
+        mb_wb.0<1>:ud
+	NULL
+        data_port(
+                OBR_CACHE_TYPE,
+                OBR_MESSAGE_TYPE,
+                OBR_CONTROL_4,
+                OBR_BIND_IDX,
+                OBR_WRITE_COMMIT_CATEGORY,
+                OBR_HEADER_PRESENT
+        )
+        mlen 1
+        rlen 2
+        {align1};
+
+cmp.l.f0.0 (1)		null:w	mb_intra_wb.16<0,1,0>:uw	mb_inter_wb.8<0,1,0>:uw {align1};
+(f0.0)   mov (2)	mbc_result.20<1>:w			-1:w	{align1};
+(f0.0)   jmpi	(1)	mb_mvp_start;
+
+add   (1) mb_msg0.8<1>:UD	mb_msg0.8<0,1,0>:ud	3:ud {align1};
+/* Read MV for MB D */
+/* bind index 3, read 8 oword (128bytes), msg type: 0(OWord Block Read) */
+send (16)
+        mb_ind
+        mb_mv0.0<1>:ub
+	NULL
+        data_port(
+                OBR_CACHE_TYPE,
+                OBR_MESSAGE_TYPE,
+                OBR_CONTROL_8,
+                OBR_BIND_IDX,
+                OBR_WRITE_COMMIT_CATEGORY,
+                OBR_HEADER_PRESENT
+        )
+        mlen 1
+        rlen 4
+        {align1};
+
+/* TODO: RefID is required after multi-references are added */
+
+/* Forward MV */
+mov	   (2)		mbc_result.4<1>:ud		mb_mv3.24<2,2,1>:ud	{align1};
+mov	   (2)		mbc_result.20<1>:w		-1:w	{align1};
+mov	(1)	INPUT_ARG0.0<1>:ud	mb_inter_wb.4<0,1,0>:ud	{align1};
+mov	(1)	INPUT_ARG0.4<1>:ud	mb_inter_wb.0<0,1,0>:ud	{align1};
+mov	(1)	INPUT_ARG0.8<1>:ud	INTER_BLOCK3:ud	{align1};
+SAVE_RET	{align1};
+jmpi	(1)	mb_pred_func;
+mov 	(1)	mb_pred_mode.0<1>:uw	RET_ARG<0,1,0>:uw	{align1};
+mov	   (1)	mbc_result.18<1>:w		MB_PRED_FLAG		{align1};
+cmp.e.f0.0 (1)	null:uw		mb_pred_mode.0<0,1,0>:uw	PRED_L0 {align1};
+(f0.0)	mov    (1) mbc_result.16<1>:uw		MB_PRED_FLAG		{align1};
+(f0.0)	mov    (1) mbc_result.20<1>:w		0:w		{align1};
+(f0.0)  jmpi	(1) mb_mvp_start;
+cmp.e.f0.0 (1)	null:uw		mb_pred_mode.0<0,1,0>:uw	PRED_L1 {align1};
+(f0.0)	mov   (1)	mbc_result.18<1>:uw		MB_PRED_FLAG		{align1};
+(f0.0)	mov    (1) mbc_result.22<1>:w		0:w		{align1};
+(f0.0)  jmpi  (1) mb_mvp_start;
+mov  (2)  mbc_result.16<1>:uw		MB_PRED_FLAG		{align1};
+mov  (2)  mbc_result.20<1>:w		0:w	{align1};
+	
+mb_mvp_start:
+/*TODO: Add the skip prediction */
+/* Check whether both MB and C are invailable */
+add	(1)	tmp_reg0.0<1>:d		mbb_result.0<0,1,0>:d	mbc_result.0<0,1,0>:d	{align1};
+cmp.z.f0.0 (1)	null:d			tmp_reg0.0<0,1,0>:d	0:d	{align1};
+(-f0.0)	jmpi (1)	mb_median_start;
+cmp.nz.f0.0 (1)	null:d	mba_result.0<0,1,0>:d		1:d		{align1};
+(f0.0)	mov	(2)	mbb_result.4<1>:ud		mba_result.4<2,2,1>:ud	{align1};	
+(f0.0)	mov	(2)	mbc_result.4<1>:ud		mba_result.4<2,2,1>:ud	{align1};	
+(f0.0)	mov	(2)	mbb_result.20<1>:uw		mba_result.20<2,2,1>:uw	{align1};	
+(f0.0)	mov	(2)	mbc_result.20<1>:uw		mba_result.20<2,2,1>:uw	{align1};	
+(f0.0)  mov     (2)	mb_mvp_ref.0<1>:ud		mba_result.4<2,2,1>:ud	{align1};
+(-f0.0) mov	(2)	mb_mvp_ref.0<1>:ud		0:ud			{align1};
+jmpi	(1)	__mb_hwdep_end;
+	
+mb_median_start:
+/* forward_MVP */
+/* check whether only one neighbour MB has the same ref ID with the current MB */
+mov (8)	tmp_reg0.0<1>:ud		0:ud		{align1};
+cmp.z.f0.0	(1)	null:d	mba_result.20<1>:w	0:w	{align1};
+(f0.0)	add	(1)	tmp_reg0.0<1>:w		tmp_reg0.0<1>:w	1:w	{align1};
+(f0.0)	mov	(1)	tmp_reg0.4<1>:ud	mba_result.4<0,1,0>:ud	{align1};
+cmp.z.f0.0	(1)	null:d	mbb_result.20<1>:w	0:w	{align1};
+(f0.0)	add	(1)	tmp_reg0.0<1>:w		tmp_reg0.0<1>:w	1:w	{align1};
+(f0.0)	mov	(1)	tmp_reg0.4<1>:ud	mbb_result.4<0,1,0>:ud	{align1};
+cmp.z.f0.0	(1)	null:d	mbc_result.20<1>:w	0:w	{align1};
+(f0.0)	add	(1)	tmp_reg0.0<1>:w		tmp_reg0.0<1>:w	1:w	{align1};
+(f0.0)	mov	(1)	tmp_reg0.4<1>:ud	mbc_result.4<0,1,0>:ud	{align1};
+cmp.e.f0.0	(1)	null:d	tmp_reg0.0<1>:w	 1:w	{align1};
+(f0.0)	mov	(1)     mb_mvp_ref.0<1>:ud	tmp_reg0.4<0,1,0>:ud	{align1};
+(f0.0)	jmpi (1)  mvp_backward;
+
+mov	(1)	INPUT_ARG0.0<1>:w	mba_result.4<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.4<1>:w	mbb_result.4<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.8<1>:w	mbc_result.4<0,1,0>:w	{align1};
+SAVE_RET	{align1};
+ jmpi	(1)	word_imedian;
+mov	(1)	mb_mvp_ref.0<1>:w		RET_ARG<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.0<1>:w	mba_result.6<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.4<1>:w	mbb_result.6<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.8<1>:w	mbc_result.6<0,1,0>:w	{align1};
+SAVE_RET	{align1};
+jmpi	(1)	word_imedian; 
+mov	(1)	mb_mvp_ref.2<1>:w		RET_ARG<0,1,0>:w	{align1};
+
+
+mvp_backward:
+/* check whether only one neighbour MB has the same ref ID with the current MB */
+mov (8)	tmp_reg0.0<1>:ud		0:ud		{align1};
+cmp.z.f0.0	(1)	null:d	mba_result.22<1>:w	0:w	{align1};
+(f0.0)	add	(1)	tmp_reg0.0<1>:w		tmp_reg0.0<1>:w	1:w	{align1};
+(f0.0)	mov	(1)	tmp_reg0.4<1>:ud	mba_result.8<0,1,0>:ud	{align1};
+cmp.z.f0.0	(1)	null:d	mbb_result.22<1>:w	0:w	{align1};
+(f0.0)	add	(1)	tmp_reg0.0<1>:w		tmp_reg0.0<1>:w	1:w	{align1};
+(f0.0)	mov	(1)	tmp_reg0.4<1>:ud	mbb_result.8<0,1,0>:ud	{align1};
+cmp.z.f0.0	(1)	null:d	mbc_result.20<1>:w	0:w	{align1};
+(f0.0)	add	(1)	tmp_reg0.0<1>:w		tmp_reg0.0<1>:w	1:w	{align1};
+(f0.0)	mov	(1)	tmp_reg0.4<1>:ud	mbc_result.8<0,1,0>:ud	{align1};
+cmp.e.f0.0	(1)	null:d	tmp_reg0.0<1>:w	 1:w	{align1};
+(f0.0)	mov	(1)     mb_mvp_ref.4<1>:ud	tmp_reg0.4<0,1,0>:ud	{align1};
+(f0.0)	jmpi (1) __mb_hwdep_end;
+
+mov	(1)	INPUT_ARG0.0<1>:w	mba_result.8<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.4<1>:w	mbb_result.8<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.8<1>:w	mbc_result.8<0,1,0>:w	{align1};
+SAVE_RET	{align1};
+ jmpi	(1)	word_imedian;
+mov	(1)	mb_mvp_ref.4<1>:w		RET_ARG<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.0<1>:w	mba_result.10<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.4<1>:w	mbb_result.10<0,1,0>:w	{align1};
+mov	(1)	INPUT_ARG0.8<1>:w	mbc_result.10<0,1,0>:w	{align1};
+SAVE_RET	{align1};
+jmpi	(1)	word_imedian; 
+mov	(1)	mb_mvp_ref.6<1>:w		RET_ARG<0,1,0>:w	{align1};
+
+__mb_hwdep_end:
+/* m2, get the MV/Mb cost passed from constant buffer when
+spawning thread by MEDIA_OBJECT */       
+mov (8) vme_m2<1>:UD            r1.0<8,8,1>:UD {align1};
+
+mov (8) vme_msg_2<1>:UD		vme_m2.0<8,8,1>:UD {align1};
+
+/* m3 */
+mov (8) vme_msg_3<1>:UD		0x0:UD {align1};	        
+
+/* m4 */
+mov  (1) INEP_ROW.0<1>:UD       0x0:UD {align1};
+and  (1) INEP_ROW.4<1>:UD       INEP_ROW.4<0,1,0>:UD            0xFF000000:UD {align1};
+mov  (8) vme_msg_4<1>:UD         INEP_ROW.0<8,8,1>:UD {align1};
+
+/* m5 */        
+mov  (8) vme_msg_5<1>:UD         0x0:UD {align1};
+mov (16) vme_msg_5.0<1>:UB       INEP_COL0.3<32,8,4>:UB {align1};
+mov  (1) vme_msg_5.16<1>:UD      INTRA_PREDICTORE_MODE {align1};
+
+/* the penalty for Intra mode */
+mov  (1) vme_msg_5.28<1>:UD	0x010101:UD {align1};
+mov  (1) vme_msg_5.20<1>:UW      CHROMA_ROW.6<0,1,0>:UW {align1};
+
+
+/* m6 */
+
+mov  (4) vme_msg_6.16<1>:UD      CHROMA_ROW.8<4,4,1>:UD {align1};
+mov  (8) vme_msg_6.0<1>:UW       CHROMA_COL.2<16,8,2>:UW {align1};
+
+/*
+ * SIC VME message
+ */
+/* m0 */        
+mov  (8) vme_msg_0.0<1>:UD      vme_m0.0<8,8,1>:UD {align1};
+mov  (1) tmp_reg0.0<1>:UW	LUMA_INTRA_MODE:UW {align1};
+/* Use the Luma mode */
+mov  (1) vme_msg_4.5<1>:UB	tmp_reg0.0<0,1,0>:UB {align1};
+
+/* m1 */
+mov  (1) intra_flag<1>:UW       0x0:UW {align1}                     ;
+and.z.f0.0 (1) null<1>:UW transform_8x8_ub<0,1,0>:UB 1:UW {align1};
+(f0.0) mov  (1) intra_part_mask_ub<1>:UB  LUMA_INTRA_8x8_DISABLE {align1};
+
+/* assign MB intra struct from the thread payload*/
+mov (1) mb_intra_struct_ub<1>:UB input_mb_intra_ub<0,1,0>:UB {align1}; 
+
+/* Disable DC HAAR component when calculating HARR SATD block */
+mov  (1) tmp_reg0.0<1>:UW	DC_HARR_DISABLE:UW		{align1};
+mov  (1) vme_m1.30<1>:UB	tmp_reg0.0<0,1,0>:UB  {align1};
+
+mov  (1) vme_m0.12<1>:UD        INTRA_SAD_HAAR:UD {align1};    /* 16x16 Source, Intra_harr */
+/* m0 */        
+mov  (8) vme_msg_0.0<1>:UD      vme_m0.0<8,8,1>:UD {align1};
+mov  (8) vme_msg_1<1>:UD        vme_m1.0<8,8,1>:UD {align1};
+
+/* after verification it will be passed by using payload */
+send (8)
+        vme_msg_ind
+        vme_wb<1>:UD
+        null
+        cre(
+                BIND_IDX_VME,
+                VME_SIC_MESSAGE_TYPE
+        )
+        mlen sic_vme_msg_length
+        rlen vme_wb_length
+        {align1};
+/*
+ * Oword Block Write message
+ */
+mov  (8) msg_reg0.0<1>:UD       obw_m0<8,8,1>:UD {align1};
+        
+mov  (1) msg_reg1.0<1>:UD       vme_wb.0<0,1,0>:UD      {align1};
+mov  (1) msg_reg1.4<1>:UD       vme_wb.16<0,1,0>:UD     {align1};
+mov  (1) msg_reg1.8<1>:UD       vme_wb.20<0,1,0>:UD     {align1};
+mov  (1) msg_reg1.12<1>:UD      vme_wb.24<0,1,0>:UD     {align1};
+
+/* Distortion, Intra (17-16), */
+mov  (1) msg_reg1.16<1>:UW      vme_wb.12<0,1,0>:UW     {align1};
+
+mov  (1) msg_reg1.20<1>:UD      vme_wb.8<0,1,0>:UD     {align1};
+/* VME clock counts */
+mov  (1) msg_reg1.24<1>:UD      vme_wb.28<0,1,0>:UD     {align1};
+
+mov  (1) msg_reg1.28<1>:UD      obw_m0.8<0,1,0>:UD     {align1};
+
+/* bind index 3, write 2 oword (32bytes), msg type: 8(OWord Block Write) */
+send (16)
+        msg_ind
+        obw_wb
+        null
+        data_port(
+                OBW_CACHE_TYPE,
+                OBW_MESSAGE_TYPE,
+                OBW_CONTROL_2,
+                OBW_BIND_IDX,
+                OBW_WRITE_COMMIT_CATEGORY,
+                OBW_HEADER_PRESENT
+        )
+        mlen 2
+        rlen obw_wb_length
+        {align1};
+
+/* IME search */
+mov  (1) vme_m0.12<1>:UD        SEARCH_CTRL_DUAL_REFERENCE + INTER_PART_MASK + INTER_SAD_HAAR:UD {align1};    /* 16x16 Source, harr */
+mov  (1) vme_m0.22<1>:UW        DREF_REGION_SIZE {align1};        
+/* Dual Reference Width&Height,32x32 */
+
+mov  (1) vme_m0.0<1>:UD		vme_m0.8<0,1,0>:UD      {align1};
+
+/* Reference = (x-8,y-8)-(x+8,y+8) */
+add  (1) vme_m0.0<1>:W          vme_m0.0<0,1,0>:W -8:W {align1};		
+add  (1) vme_m0.2<1>:W          vme_m0.2<0,1,0>:W -8:W {align1};
+
+mov  (1) vme_m0.0<1>:W		-8:W			{align1};
+mov  (1) vme_m0.2<1>:W		-8:W			{align1};
+
+mov  (1) vme_m0.4<1>:UD		vme_m0.0<0,1,0>:UD	{align1};
+
+mov  (8) vme_msg_0.0<1>:UD      vme_m0.0<8,8,1>:UD {align1};
+
+mov  (1) vme_m1.0<1>:UD         ADAPTIVE_SEARCH_ENABLE:ud {align1} ;
+/* the Max MV number is passed by constant buffer */
+mov  (1) vme_m1.4<1>:UB         r4.28<0,1,0>:UB {align1};          
+mov  (1) vme_m1.8<1>:UD         DSTART_CENTER + DSEARCH_PATH_LEN:UD {align1};
+/* Set the MV cost center */ 
+mov  (1) vme_m1.16<1>:ud	mb_mvp_ref.0<0,1,0>:ud	{align1};
+mov  (1) vme_m1.20<1>:ud	mb_mvp_ref.4<0,1,0>:ud	{align1};
+mov  (8) vme_msg_1.0<1>:UD      vme_m1.0<8,8,1>:UD {align1};
+
+mov (8) vme_msg_2<1>:UD		vme_m2.0<8,8,1>:UD {align1};
+/* M3/M4 search path */
+
+mov  (1) vme_msg_3.0<1>:UD	0x10010101:UD {align1};
+mov  (1) vme_msg_3.4<1>:UD	0x100F0F0F:UD {align1};
+mov  (1) vme_msg_3.8<1>:UD	0x10010101:UD {align1};
+mov  (1) vme_msg_3.12<1>:UD	0x000F0F0F:UD {align1};
+
+mov  (4) vme_msg_3.16<1>:UD	0x0:UD {align1};
+mov  (8) vme_msg_4.16<1>:UD	0x0:UD {align1};
+
+send (8)
+        vme_msg_ind
+        vme_wb<1>:UD
+        null
+        vme(
+                BIND_IDX_VME,
+                0,
+                0,
+                VME_IME_MESSAGE_TYPE
+        )
+        mlen ime_vme_msg_length
+        rlen vme_wb_length {align1};
+
+/* Set Macroblock-shape/mode for FBR */
+
+mov  (1) vme_m2.20<1>:UD	0x0:UD {align1};
+mov  (1) vme_m2.21<1>:UB	vme_wb.25<0,1,0>:UB	{align1};
+mov  (1) vme_m2.22<1>:UB	vme_wb.26<0,1,0>:UB	{align1};
+
+and  (1) tmp_reg0.0<1>:UW	vme_wb.0<0,1,0>:UW	0x03:UW {align1};
+mov  (1) vme_m2.20<1>:UB	tmp_reg0.0<0,1,0>:UB    {align1};
+
+/* Send FBR message into CRE */
+
+mov  (8) vme_msg_3.0<1>:UD       vme_wb1.0<8,8,1>:UD {align1};
+mov  (8) vme_msg_4.0<1>:ud       vme_wb2.0<8,8,1>:ud {align1};
+mov  (8) vme_msg_5.0<1>:ud       vme_wb3.0<8,8,1>:ud {align1};
+mov  (8) vme_msg_6.0<1>:ud       vme_wb4.0<8,8,1>:ud {align1};                
+
+ /* 16x16 Source, 1/4 pixel, harr, BME ENABLE */
+mov  (1) vme_m0.12<1>:UD	INTER_SAD_HAAR + SUB_PEL_MODE_QUARTER + FBR_BME_ENABLE:UD {align1};
+   
+mov  (8) vme_msg_0.0<1>:UD	vme_m0.0<8,8,1>:UD  {align1};
+
+mov  (1) tmp_reg0.0<1>:uw	BI_WEIGHT	{align1};
+mov  (1) vme_m1.6<1>:UB		tmp_reg0.0<0,1,0>:ub	{align1};
+mov  (8) vme_msg_1.0<1>:UD	vme_m1.0<8,8,1>:UD  {align1};
+
+mov  (8) vme_msg_2.0<1>:UD		vme_m2.0<8,8,1>:UD	{align1};
+
+/* after verification it will be passed by using payload */
+send (8)
+        vme_msg_ind
+        vme_wb<1>:UD
+        null
+        cre(
+                BIND_IDX_VME,
+                VME_FBR_MESSAGE_TYPE
+        )
+        mlen fbr_vme_msg_length
+        rlen vme_wb_length
+        {align1};
+
+add  (1) obw_m0.8<1>:UD         obw_m0.8<0,1,0>:UD 0x02:UD {align1};
+mov  (8) msg_reg0.0<1>:UD       obw_m0<8,8,1>:UD {align1};
+/* write FME info */
+mov  (1) msg_reg1.0<1>:UD       vme_wb.0<0,1,0>:UD      {align1};
+
+mov  (1) msg_reg1.4<1>:UD       vme_wb.24<0,1,0>:UD     {align1};
+/* Inter distortion of FME */
+mov  (1) msg_reg1.8<1>:UD       vme_wb.8<0,1,0>:UD     {align1};
+
+mov  (1) msg_reg1.12<1>:UD	vme_m2.20<0,1,0>:UD {align1};
+
+/* bind index 3, write  oword (16bytes), msg type: 8(OWord Block Write) */
+send (16)
+        msg_ind
+        obw_wb
+        null
+        data_port(
+                OBW_CACHE_TYPE,
+                OBW_MESSAGE_TYPE,
+                OBW_CONTROL_0,
+                OBW_BIND_IDX,
+                OBW_WRITE_COMMIT_CATEGORY,
+                OBW_HEADER_PRESENT
+        )
+        mlen 2
+        rlen obw_wb_length
+        {align1};
+
+/* Write FME/BME MV */
+add  (1) obw_m0.8<1>:UD         obw_m0.8<0,1,0>:UD 0x01:UD {align1};
+mov  (8) msg_reg0.0<1>:UD       obw_m0.0<8,8,1>:UD {align1};
+
+
+mov  (8) msg_reg1.0<1>:UD       vme_wb1.0<8,8,1>:UD {align1};
+mov  (8) msg_reg2.0<1>:ud       vme_wb2.0<8,8,1>:ud {align1};
+mov  (8) msg_reg3.0<1>:ud       vme_wb3.0<8,8,1>:ud {align1};
+mov  (8) msg_reg4.0<1>:ud       vme_wb4.0<8,8,1>:ud {align1};                
+/* bind index 3, write  8 oword (128 bytes), msg type: 8(OWord Block Write) */
+send (16)
+        msg_ind
+        obw_wb
+        null
+        data_port(
+                OBW_CACHE_TYPE,
+                OBW_MESSAGE_TYPE,
+                OBW_CONTROL_8,
+                OBW_BIND_IDX,
+                OBW_WRITE_COMMIT_CATEGORY,
+                OBW_HEADER_PRESENT
+        )
+        mlen 5
+        rlen obw_wb_length
+        {align1};
+
+/* Write FME/BME RefID */
+add  (1) obw_m0.8<1>:UD         obw_m0.8<0,1,0>:UD 0x08:UD {align1};
+mov  (8) msg_reg0.0<1>:UD       obw_m0<8,8,1>:UD {align1};
+
+mov  (8) msg_reg1.0<1>:UD	vme_wb6.0<8,8,1>:UD {align1};
+
+/* bind index 3, write 2 oword (32bytes), msg type: 8(OWord Block Write) */
+send (16)
+        msg_ind
+        obw_wb
+        null
+        data_port(
+                OBW_CACHE_TYPE,
+                OBW_MESSAGE_TYPE,
+                OBW_CONTROL_2,
+                OBW_BIND_IDX,
+                OBW_WRITE_COMMIT_CATEGORY,
+                OBW_HEADER_PRESENT
+        )
+        mlen 2
+        rlen obw_wb_length
+        {align1};
+
+__EXIT: 
+/*
+ * kill thread
+ */        
+mov  (8) ts_msg_reg0<1>:UD         r0<8,8,1>:UD {align1};
+send (16) ts_msg_ind acc0<1>UW null thread_spawner(0, 0, 1) mlen 1 rlen 0 {align1 EOT};
+
+
+	nop		;
+	nop		;
+/* Compare three word data to get the min value */
+word_imin:
+	cmp.le.f0.0 (1)		null:w		INPUT_ARG0.0<0,1,0>:w	INPUT_ARG0.4<0,1,0>:w {align1};
+	(f0.0) mov  (1)		TEMP_VAR0.0<1>:w INPUT_ARG0.0<0,1,0>:w			  {align1};
+	(-f0.0) mov (1)		TEMP_VAR0.0<1>:w INPUT_ARG0.4<0,1,0>:w			  {align1};
+	cmp.le.f0.0 (1)		null:w		TEMP_VAR0.0<0,1,0>:w	INPUT_ARG0.8<0,1,0>:w {align1};
+	(f0.0) mov  (1)		RET_ARG<1>:w TEMP_VAR0.0<0,1,0>:w			  {align1};
+	(-f0.0) mov (1)		RET_ARG<1>:w INPUT_ARG0.8<0,1,0>:w			  {align1};
+	RETURN		{align1};	
+	
+/* Compare three word data to get the max value */
+word_imax:
+	cmp.ge.f0.0 (1)		null:w		INPUT_ARG0.0<0,1,0>:w	INPUT_ARG0.4<0,1,0>:w {align1};
+	(f0.0) mov  (1)		TEMP_VAR0.0<1>:w INPUT_ARG0.0<0,1,0>:w			  {align1};
+	(-f0.0) mov (1)		TEMP_VAR0.0<1>:w INPUT_ARG0.4<0,1,0>:w			  {align1};
+	cmp.ge.f0.0 (1)		null:w		TEMP_VAR0.0<0,1,0>:w	INPUT_ARG0.8<0,1,0>:w {align1};
+	(f0.0) mov  (1)		RET_ARG<1>:w TEMP_VAR0.0<0,1,0>:w			  {align1};
+	(-f0.0) mov (1)		RET_ARG<1>:w INPUT_ARG0.8<0,1,0>:w			  {align1};
+	RETURN		{align1};	
+	
+word_imedian:
+	cmp.ge.f0.0 (1) null:w INPUT_ARG0.0<0,1,0>:w INPUT_ARG0.4<0,1,0>:w {align1};
+	(f0.0)	jmpi (1) cmp_a_ge_b;
+	cmp.ge.f0.0 (1) null:w INPUT_ARG0.0<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1};
+	(f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.0<0,1,0>:w {align1};
+	(f0.0) jmpi (1) cmp_end;
+	cmp.ge.f0.0 (1) null:w INPUT_ARG0.4<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1};
+	(f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.8<0,1,0>:w {align1};
+	(-f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.4<0,1,0>:w {align1};
+	jmpi (1) cmp_end;
+cmp_a_ge_b:
+	cmp.ge.f0.0 (1) null:w INPUT_ARG0.4<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1};
+	(f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.4<0,1,0>:w {align1};
+	(f0.0) jmpi (1) cmp_end;
+	cmp.ge.f0.0 (1) null:w INPUT_ARG0.0<0,1,0>:w INPUT_ARG0.8<0,1,0>:w {align1};
+	(f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.8<0,1,0>:w {align1};
+	(-f0.0) mov (1) RET_ARG<1>:w INPUT_ARG0.0<0,1,0>:w {align1};
+cmp_end:
+ 	RETURN	{align1};
+
+mb_pred_func:
+        mov	(8)	TEMP_VAR0.0<1>:ud	0:ud	{align1};
+	mov     (1)	TEMP_VAR0.0<1>:ub	INPUT_ARG0.2<0,1,0>:ub	{align1};
+	and	(1)   TEMP_VAR0.4<1>:uw   INPUT_ARG0.4<0,1,0>:uw	INTER_MASK:uw	{align1};
+	/* INTER16x16 mode. The bit1-0 is the prediction mode */
+	cmp.e.f0.0 (1) null:uw	TEMP_VAR0.4<1>:uw	INTER_16X16MODE:uw	{align1};
+	(f0.0)	and (1) RET_ARG<1>:uw	TEMP_VAR0.0<0,1,0>:uw	PRED_MASK {align1};
+	(f0.0)  jmpi (1) end_mb_pred;
+	/* Check whether it is INTER8x8 mode. */
+	cmp.e.f0.0 (1) null:uw	TEMP_VAR0.4<1>:uw	INTER_8X8MODE:uw	{align1};
+	(f0.0)	jmpi (1) mb_pred_func_8;
+	
+	/* Check whether it is INTER16x8 mode. */
+	cmp.e.f0.0 (1) null:uw	TEMP_VAR0.4<1>:uw	INTER_16X8MODE:uw	{align1};
+	(f0.0)	jmpi (1) mb_pred_func_168;
+mb_pred_func_816:	
+	/* Block 0/2 uses the bit1-0. Block 1/3 uses the bit3-2 */
+	mov	(1) TEMP_VAR0.8<1>:uw INPUT_ARG0.8<0,1,0>:uw	{align1};
+	and.z.f0.0 (1)	null:uw	TEMP_VAR0.8<1>:uw	INTER_BLOCK1:uw	{align1};
+	(f0.0)	and	(1) RET_ARG<1>:uw	TEMP_VAR0.0<0,1,0>:uw	PRED_MASK {align1};
+	(f0.0)  jmpi (1) end_mb_pred;
+	shr	(1)  TEMP_VAR0.16<1>:uw TEMP_VAR0.0<0,1,0>:uw	2:uw {align1};
+	and	(1)  RET_ARG<1>:uw	TEMP_VAR0.16<1>:uw	PRED_MASK  {align1};	
+	jmpi	(1) end_mb_pred;
+
+mb_pred_func_168:
+	/* Block 0/1 uses the bit1-0. Block 2/3 uses the bit3-2 */
+	mov	(1) TEMP_VAR0.8<1>:uw INPUT_ARG0.8<0,1,0>:uw	{align1};
+	cmp.l.f0.0 (1)	null:uw	TEMP_VAR0.8<1>:uw	INTER_BLOCK2:uw	{align1};
+	(f0.0)	and  (1) RET_ARG<1>:uw	TEMP_VAR0.0<0,1,0>:uw	PRED_MASK {align1};
+	(f0.0)  jmpi (1) end_mb_pred;
+	shr	(1)  TEMP_VAR0.16<1>:uw TEMP_VAR0.0<0,1,0>:uw	2:uw {align1};
+	and	(1)  RET_ARG<1>:uw	TEMP_VAR0.16<1>:uw	PRED_MASK  {align1};	
+	jmpi	(1) end_mb_pred;
+
+mb_pred_func_8:
+	/* 8X8 mode. Every block uses two bits as the prediction mode. */
+	mul     (1)  TEMP_VAR0.8<1>:uw  INPUT_ARG0.8<0,1,0>:uw	2:uw {align1};
+	shr	(1)  TEMP_VAR0.16<1>:uw TEMP_VAR0.0<0,1,0>:uw	TEMP_VAR0.8<0,1,0>:uw {align1};
+	and	(1)  RET_ARG<1>:uw	TEMP_VAR0.16<1>:uw	PRED_MASK  {align1};	
+end_mb_pred:
+ 	RETURN	{align1};
+	
diff --git a/src/shaders/vme/inter_bframe_haswell.g75a b/src/shaders/vme/inter_bframe_haswell.g75a
new file mode 100644
index 00000000..b5eca7e3
--- /dev/null
+++ b/src/shaders/vme/inter_bframe_haswell.g75a
@@ -0,0 +1,2 @@
+#include "vme75.inc"
+#include "inter_bframe_haswell.asm"
diff --git a/src/shaders/vme/inter_bframe_haswell.g75b b/src/shaders/vme/inter_bframe_haswell.g75b
new file mode 100644
index 00000000..6327a320
--- /dev/null
+++ b/src/shaders/vme/inter_bframe_haswell.g75b
@@ -0,0 +1,397 @@
+   { 0x00800001, 0x24000061, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x24400061, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x24800061, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x24c00061, 0x00000000, 0x00000000 },
+   { 0x00200009, 0x24002e25, 0x004500a0, 0x00040004 },
+   { 0x00000040, 0x24003ca5, 0x00000400, 0xfff8fff8 },
+   { 0x00000040, 0x24043ca5, 0x00000404, 0xffffffff },
+   { 0x00000001, 0x240800e1, 0x00000000, 0x0000001f },
+   { 0x00000001, 0x24140231, 0x00000014, 0x00000000 },
+   { 0x00200009, 0x24202e25, 0x004500a0, 0x00040004 },
+   { 0x00000040, 0x24203ca5, 0x00000420, 0xfffcfffc },
+   { 0x00000001, 0x242800e1, 0x00000000, 0x000f0003 },
+   { 0x00000001, 0x24340231, 0x00000014, 0x00000000 },
+   { 0x00200009, 0x24482e29, 0x004500a0, 0x00040004 },
+   { 0x00000001, 0x24540231, 0x00000014, 0x00000000 },
+   { 0x00000041, 0x24884521, 0x000000a2, 0x000000a1 },
+   { 0x00000040, 0x24884421, 0x00000488, 0x000000a0 },
+   { 0x00000041, 0x24880c21, 0x00000488, 0x00000018 },
+   { 0x00000001, 0x24940231, 0x00000014, 0x00000000 },
+   { 0x00600001, 0x28000021, 0x008d0400, 0x00000000 },
+   { 0x04600031, 0x23801cb1, 0x00000800, 0x02190004 },
+   { 0x00600001, 0x28000021, 0x008d0420, 0x00000000 },
+   { 0x04600031, 0x23a01cb1, 0x00000800, 0x02290004 },
+   { 0x00200009, 0x24002e25, 0x004500a0, 0x00030003 },
+   { 0x00000041, 0x24003ca5, 0x00000400, 0x00020002 },
+   { 0x00000040, 0x24003ca5, 0x00000400, 0xfff8fff8 },
+   { 0x00000040, 0x24043ca5, 0x00000404, 0xffffffff },
+   { 0x00600001, 0x28000021, 0x008d0400, 0x00000000 },
+   { 0x04600031, 0x26001cb1, 0x00000800, 0x02190006 },
+   { 0x00200009, 0x24202e25, 0x004500a0, 0x00030003 },
+   { 0x00000041, 0x24203ca5, 0x00000420, 0x00020002 },
+   { 0x00000040, 0x24203ca5, 0x00000420, 0xfffcfffc },
+   { 0x00000001, 0x242800e1, 0x00000000, 0x00070003 },
+   { 0x00600001, 0x28000021, 0x008d0420, 0x00000000 },
+   { 0x04600031, 0x26201cb1, 0x00000800, 0x02190006 },
+   { 0x00600001, 0x24600061, 0x00000000, 0x00000000 },
+   { 0x00600001, 0x2ac00061, 0x00000000, 0x00000000 },
+   { 0x01000005, 0x20002d28, 0x000000a6, 0x00040004 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000cb0 },
+   { 0x00600001, 0x2ae00061, 0x00000000, 0x00000000 },
+   { 0x00600001, 0x2b000061, 0x00000000, 0x00000000 },
+   { 0x00600001, 0x2b200061, 0x00000000, 0x00000000 },
+   { 0x00600001, 0x2b400061, 0x00000000, 0x00000000 },
+   { 0x01000005, 0x20002e28, 0x000000a5, 0x00600060 },
+   { 0x00210001, 0x2af401ed, 0x00000000, 0xffffffff },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x000001f0 },
+   { 0x00000001, 0x2ae000e5, 0x00000000, 0x00000001 },
+   { 0x00200001, 0x24000229, 0x004500a0, 0x00000000 },
+   { 0x00000040, 0x24003dad, 0x00000400, 0xffffffff },
+   { 0x00000041, 0x2b482521, 0x000000a2, 0x00000402 },
+   { 0x00000040, 0x2b482421, 0x00000b48, 0x00000400 },
+   { 0x00000041, 0x2b480c21, 0x00000b48, 0x00000018 },
+   { 0x00000001, 0x2b540231, 0x00000014, 0x00000000 },
+   { 0x0a800031, 0x2b601ca1, 0x00000b40, 0x02280303 },
+   { 0x05000010, 0x2000252c, 0x00000b70, 0x00000b88 },
+   { 0x00210001, 0x2af401ed, 0x00000000, 0xffffffff },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000140 },
+   { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 },
+   { 0x0a800031, 0x2ba01ca1, 0x00000b40, 0x02480403 },
+   { 0x00200001, 0x2ae40021, 0x00450bc8, 0x00000000 },
+   { 0x00200001, 0x2af401ed, 0x00000000, 0xffffffff },
+   { 0x00000001, 0x2fa00021, 0x00000b84, 0x00000000 },
+   { 0x00000001, 0x2fa40021, 0x00000b80, 0x00000000 },
+   { 0x00000001, 0x2fa80061, 0x00000000, 0x00000001 },
+   { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x000012f0 },
+   { 0x00000001, 0x2aa00129, 0x00000fe4, 0x00000000 },
+   { 0x01000010, 0x20002d28, 0x00000aa0, 0x00000000 },
+   { 0x00010001, 0x2af001e9, 0x00000000, 0x00010001 },
+   { 0x00010001, 0x2af401ed, 0x00000000, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000060 },
+   { 0x01000010, 0x20002d28, 0x00000aa0, 0x00010001 },
+   { 0x00010001, 0x2af201e9, 0x00000000, 0x00010001 },
+   { 0x00010001, 0x2af601ed, 0x00000000, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000020 },
+   { 0x00200001, 0x2af001e9, 0x00000000, 0x00010001 },
+   { 0x00200001, 0x2af401ed, 0x00000000, 0x00000000 },
+   { 0x00600001, 0x2b400061, 0x00000000, 0x00000000 },
+   { 0x01000005, 0x20002e28, 0x000000a5, 0x00100010 },
+   { 0x00210001, 0x2b1401ed, 0x00000000, 0xffffffff },
+   { 0x00210001, 0x2b3401ed, 0x00000000, 0xffffffff },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000630 },
+   { 0x00000001, 0x2b0000e5, 0x00000000, 0x00000001 },
+   { 0x00200001, 0x24000229, 0x004500a0, 0x00000000 },
+   { 0x00000040, 0x24023dad, 0x00000402, 0xffffffff },
+   { 0x00000041, 0x2b482521, 0x000000a2, 0x00000402 },
+   { 0x00000040, 0x2b482421, 0x00000b48, 0x00000400 },
+   { 0x00000041, 0x2b480c21, 0x00000b48, 0x00000018 },
+   { 0x00000001, 0x2b540231, 0x00000014, 0x00000000 },
+   { 0x0a800031, 0x2b601ca1, 0x00000b40, 0x02280303 },
+   { 0x05000010, 0x2000252c, 0x00000b70, 0x00000b88 },
+   { 0x00210001, 0x2b1401ed, 0x00000000, 0xffffffff },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000140 },
+   { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 },
+   { 0x0a800031, 0x2ba01ca1, 0x00000b40, 0x02480403 },
+   { 0x00200001, 0x2b040021, 0x00450bf0, 0x00000000 },
+   { 0x00200001, 0x2b1401ed, 0x00000000, 0xffffffff },
+   { 0x00000001, 0x2fa00021, 0x00000b84, 0x00000000 },
+   { 0x00000001, 0x2fa40021, 0x00000b80, 0x00000000 },
+   { 0x00000001, 0x2fa80061, 0x00000000, 0x00000002 },
+   { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x000010b0 },
+   { 0x00000001, 0x2aa00129, 0x00000fe4, 0x00000000 },
+   { 0x01000010, 0x20002d28, 0x00000aa0, 0x00000000 },
+   { 0x00010001, 0x2b1001e9, 0x00000000, 0x00010001 },
+   { 0x00010001, 0x2b1401ed, 0x00000000, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000060 },
+   { 0x01000010, 0x20002d28, 0x00000aa0, 0x00010001 },
+   { 0x00010001, 0x2b1201e9, 0x00000000, 0x00010001 },
+   { 0x00010001, 0x2b1601ed, 0x00000000, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000020 },
+   { 0x00200001, 0x2b1001e9, 0x00000000, 0x00010001 },
+   { 0x00200001, 0x2b1401ed, 0x00000000, 0x00000000 },
+   { 0x00600001, 0x2b400061, 0x00000000, 0x00000000 },
+   { 0x01000005, 0x20002e28, 0x000000a5, 0x00080008 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000210 },
+   { 0x00000001, 0x2b2000e5, 0x00000000, 0x00000001 },
+   { 0x00200001, 0x24000229, 0x004500a0, 0x00000000 },
+   { 0x00000040, 0x24023dad, 0x00000402, 0xffffffff },
+   { 0x00000040, 0x24003dad, 0x00000400, 0x00010001 },
+   { 0x00000041, 0x2b482521, 0x000000a2, 0x00000402 },
+   { 0x00000040, 0x2b482421, 0x00000b48, 0x00000400 },
+   { 0x00000041, 0x2b480c21, 0x00000b48, 0x00000018 },
+   { 0x00000001, 0x2b540231, 0x00000014, 0x00000000 },
+   { 0x0a800031, 0x2b601ca1, 0x00000b40, 0x02280303 },
+   { 0x05000010, 0x2000252c, 0x00000b70, 0x00000b88 },
+   { 0x00210001, 0x2b3401ed, 0x00000000, 0xffffffff },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000350 },
+   { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 },
+   { 0x0a800031, 0x2ba01ca1, 0x00000b40, 0x02480403 },
+   { 0x00200001, 0x2b240021, 0x00450bf0, 0x00000000 },
+   { 0x00200001, 0x2b3401ed, 0x00000000, 0xffffffff },
+   { 0x00000001, 0x2fa00021, 0x00000b84, 0x00000000 },
+   { 0x00000001, 0x2fa40021, 0x00000b80, 0x00000000 },
+   { 0x00000001, 0x2fa80061, 0x00000000, 0x00000002 },
+   { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x00000e80 },
+   { 0x00000001, 0x2aa00129, 0x00000fe4, 0x00000000 },
+   { 0x01000010, 0x20002d28, 0x00000aa0, 0x00000000 },
+   { 0x00010001, 0x2b3001e9, 0x00000000, 0x00010001 },
+   { 0x00010001, 0x2b3401ed, 0x00000000, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000270 },
+   { 0x01000010, 0x20002d28, 0x00000aa0, 0x00010001 },
+   { 0x00010001, 0x2b3201e9, 0x00000000, 0x00010001 },
+   { 0x00010001, 0x2b3601ed, 0x00000000, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000230 },
+   { 0x00200001, 0x2b3001e9, 0x00000000, 0x00010001 },
+   { 0x00200001, 0x2b3401ed, 0x00000000, 0x00000000 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x00000200 },
+   { 0x00000001, 0x2b2000e5, 0x00000000, 0x00000001 },
+   { 0x00200001, 0x24000229, 0x004500a0, 0x00000000 },
+   { 0x00200040, 0x24003dad, 0x00450400, 0xffffffff },
+   { 0x00000041, 0x2b482521, 0x000000a2, 0x00000402 },
+   { 0x00000040, 0x2b482421, 0x00000b48, 0x00000400 },
+   { 0x00000041, 0x2b480c21, 0x00000b48, 0x00000018 },
+   { 0x00000001, 0x2b540231, 0x00000014, 0x00000000 },
+   { 0x0a800031, 0x2b601ca1, 0x00000b40, 0x02280303 },
+   { 0x05000010, 0x2000252c, 0x00000b70, 0x00000b88 },
+   { 0x00210001, 0x2b3401ed, 0x00000000, 0xffffffff },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000150 },
+   { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 },
+   { 0x0a800031, 0x2ba01cb1, 0x00000b40, 0x02480403 },
+   { 0x00200001, 0x2b240021, 0x00450c18, 0x00000000 },
+   { 0x00200001, 0x2b3401ed, 0x00000000, 0xffffffff },
+   { 0x00000001, 0x2fa00021, 0x00000b84, 0x00000000 },
+   { 0x00000001, 0x2fa40021, 0x00000b80, 0x00000000 },
+   { 0x00000001, 0x2fa80061, 0x00000000, 0x00000003 },
+   { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x00000c80 },
+   { 0x00000001, 0x2aa00129, 0x00000fe4, 0x00000000 },
+   { 0x00000001, 0x2b3201ed, 0x00000000, 0x00010001 },
+   { 0x01000010, 0x20002d28, 0x00000aa0, 0x00000000 },
+   { 0x00010001, 0x2b3001e9, 0x00000000, 0x00010001 },
+   { 0x00010001, 0x2b3401ed, 0x00000000, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000060 },
+   { 0x01000010, 0x20002d28, 0x00000aa0, 0x00010001 },
+   { 0x00010001, 0x2b3201e9, 0x00000000, 0x00010001 },
+   { 0x00010001, 0x2b3601ed, 0x00000000, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000020 },
+   { 0x00200001, 0x2b3001e9, 0x00000000, 0x00010001 },
+   { 0x00200001, 0x2b3401ed, 0x00000000, 0x00000000 },
+   { 0x00000040, 0x240014a5, 0x00000b00, 0x00000b20 },
+   { 0x01000010, 0x20001ca4, 0x00000400, 0x00000000 },
+   { 0x00110020, 0x34001c00, 0x00001400, 0x00000080 },
+   { 0x02000010, 0x20001ca4, 0x00000ae0, 0x00000001 },
+   { 0x00210001, 0x2b040021, 0x00450ae4, 0x00000000 },
+   { 0x00210001, 0x2b240021, 0x00450ae4, 0x00000000 },
+   { 0x00210001, 0x2b140129, 0x00450af4, 0x00000000 },
+   { 0x00210001, 0x2b340129, 0x00450af4, 0x00000000 },
+   { 0x00210001, 0x2ac00021, 0x00450ae4, 0x00000000 },
+   { 0x00310001, 0x2ac00061, 0x00000000, 0x00000000 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x00000320 },
+   { 0x00600001, 0x24000061, 0x00000000, 0x00000000 },
+   { 0x01000010, 0x20003da4, 0x00200af4, 0x00000000 },
+   { 0x00010040, 0x24003dad, 0x00200400, 0x00010001 },
+   { 0x00010001, 0x24040021, 0x00000ae4, 0x00000000 },
+   { 0x01000010, 0x20003da4, 0x00200b14, 0x00000000 },
+   { 0x00010040, 0x24003dad, 0x00200400, 0x00010001 },
+   { 0x00010001, 0x24040021, 0x00000b04, 0x00000000 },
+   { 0x01000010, 0x20003da4, 0x00200b34, 0x00000000 },
+   { 0x00010040, 0x24003dad, 0x00200400, 0x00010001 },
+   { 0x00010001, 0x24040021, 0x00000b24, 0x00000000 },
+   { 0x01000010, 0x20003da4, 0x00200400, 0x00010001 },
+   { 0x00010001, 0x2ac00021, 0x00000404, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x000000c0 },
+   { 0x00000001, 0x2fa001ad, 0x00000ae4, 0x00000000 },
+   { 0x00000001, 0x2fa401ad, 0x00000b04, 0x00000000 },
+   { 0x00000001, 0x2fa801ad, 0x00000b24, 0x00000000 },
+   { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x000008f0 },
+   { 0x00000001, 0x2ac001ad, 0x00000fe4, 0x00000000 },
+   { 0x00000001, 0x2fa001ad, 0x00000ae6, 0x00000000 },
+   { 0x00000001, 0x2fa401ad, 0x00000b06, 0x00000000 },
+   { 0x00000001, 0x2fa801ad, 0x00000b26, 0x00000000 },
+   { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x00000890 },
+   { 0x00000001, 0x2ac201ad, 0x00000fe4, 0x00000000 },
+   { 0x00600001, 0x24000061, 0x00000000, 0x00000000 },
+   { 0x01000010, 0x20003da4, 0x00200af6, 0x00000000 },
+   { 0x00010040, 0x24003dad, 0x00200400, 0x00010001 },
+   { 0x00010001, 0x24040021, 0x00000ae8, 0x00000000 },
+   { 0x01000010, 0x20003da4, 0x00200b16, 0x00000000 },
+   { 0x00010040, 0x24003dad, 0x00200400, 0x00010001 },
+   { 0x00010001, 0x24040021, 0x00000b08, 0x00000000 },
+   { 0x01000010, 0x20003da4, 0x00200b34, 0x00000000 },
+   { 0x00010040, 0x24003dad, 0x00200400, 0x00010001 },
+   { 0x00010001, 0x24040021, 0x00000b28, 0x00000000 },
+   { 0x01000010, 0x20003da4, 0x00200400, 0x00010001 },
+   { 0x00010001, 0x2ac40021, 0x00000404, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x000000c0 },
+   { 0x00000001, 0x2fa001ad, 0x00000ae8, 0x00000000 },
+   { 0x00000001, 0x2fa401ad, 0x00000b08, 0x00000000 },
+   { 0x00000001, 0x2fa801ad, 0x00000b28, 0x00000000 },
+   { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x00000760 },
+   { 0x00000001, 0x2ac401ad, 0x00000fe4, 0x00000000 },
+   { 0x00000001, 0x2fa001ad, 0x00000aea, 0x00000000 },
+   { 0x00000001, 0x2fa401ad, 0x00000b0a, 0x00000000 },
+   { 0x00000001, 0x2fa801ad, 0x00000b2a, 0x00000000 },
+   { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x00000700 },
+   { 0x00000001, 0x2ac601ad, 0x00000fe4, 0x00000000 },
+   { 0x00600001, 0x25600021, 0x008d0020, 0x00000000 },
+   { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 },
+   { 0x00600001, 0x28600061, 0x00000000, 0x00000000 },
+   { 0x00000001, 0x23800061, 0x00000000, 0x00000000 },
+   { 0x00000005, 0x23840c21, 0x00000384, 0xff000000 },
+   { 0x00600001, 0x28800021, 0x008d0380, 0x00000000 },
+   { 0x00600001, 0x28a00061, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x28a00231, 0x00cf03a3, 0x00000000 },
+   { 0x00000001, 0x28b00061, 0x00000000, 0x11111111 },
+   { 0x00000001, 0x28bc0061, 0x00000000, 0x00010101 },
+   { 0x00000001, 0x28b40129, 0x00000606, 0x00000000 },
+   { 0x00400001, 0x28d00021, 0x00690608, 0x00000000 },
+   { 0x00600001, 0x28c00129, 0x00ae0622, 0x00000000 },
+   { 0x00600001, 0x28000021, 0x008d0440, 0x00000000 },
+   { 0x00000001, 0x24000169, 0x00000000, 0x00010001 },
+   { 0x00000001, 0x28850231, 0x00000400, 0x00000000 },
+   { 0x00000001, 0x247c0169, 0x00000000, 0x00000000 },
+   { 0x01000005, 0x20002e28, 0x000000a4, 0x00010001 },
+   { 0x00010001, 0x247c00f1, 0x00000000, 0x00000002 },
+   { 0x00000001, 0x247d0231, 0x000000a5, 0x00000000 },
+   { 0x00000001, 0x24000169, 0x00000000, 0x00200020 },
+   { 0x00000001, 0x247e0231, 0x00000400, 0x00000000 },
+   { 0x00000001, 0x244c0061, 0x00000000, 0x00800000 },
+   { 0x00600001, 0x28000021, 0x008d0440, 0x00000000 },
+   { 0x00600001, 0x28200021, 0x008d0460, 0x00000000 },
+   { 0x0d600031, 0x21801ca1, 0x00000800, 0x0e782000 },
+   { 0x00600001, 0x28000021, 0x008d0480, 0x00000000 },
+   { 0x00000001, 0x28200021, 0x00000180, 0x00000000 },
+   { 0x00000001, 0x28240021, 0x00000190, 0x00000000 },
+   { 0x00000001, 0x28280021, 0x00000194, 0x00000000 },
+   { 0x00000001, 0x282c0021, 0x00000198, 0x00000000 },
+   { 0x00000001, 0x28300129, 0x0000018c, 0x00000000 },
+   { 0x00000001, 0x28340021, 0x00000188, 0x00000000 },
+   { 0x00000001, 0x28380021, 0x0000019c, 0x00000000 },
+   { 0x00000001, 0x283c0021, 0x00000488, 0x00000000 },
+   { 0x0a800031, 0x20001cac, 0x00000800, 0x040a0203 },
+   { 0x00000001, 0x244c0061, 0x00000000, 0x00200700 },
+   { 0x00000001, 0x24560169, 0x00000000, 0x20202020 },
+   { 0x00000001, 0x24400021, 0x00000448, 0x00000000 },
+   { 0x00000040, 0x24403dad, 0x00000440, 0xfff8fff8 },
+   { 0x00000040, 0x24423dad, 0x00000442, 0xfff8fff8 },
+   { 0x00000001, 0x244001ed, 0x00000000, 0xfff8fff8 },
+   { 0x00000001, 0x244201ed, 0x00000000, 0xfff8fff8 },
+   { 0x00000001, 0x24440021, 0x00000440, 0x00000000 },
+   { 0x00600001, 0x28000021, 0x008d0440, 0x00000000 },
+   { 0x00000001, 0x24600061, 0x00000000, 0x00000002 },
+   { 0x00000001, 0x24640231, 0x0000009c, 0x00000000 },
+   { 0x00000001, 0x24680061, 0x00000000, 0x00001212 },
+   { 0x00000001, 0x24700021, 0x00000ac0, 0x00000000 },
+   { 0x00000001, 0x24740021, 0x00000ac4, 0x00000000 },
+   { 0x00600001, 0x28200021, 0x008d0460, 0x00000000 },
+   { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 },
+   { 0x00000001, 0x28600061, 0x00000000, 0x10010101 },
+   { 0x00000001, 0x28640061, 0x00000000, 0x100f0f0f },
+   { 0x00000001, 0x28680061, 0x00000000, 0x10010101 },
+   { 0x00000001, 0x286c0061, 0x00000000, 0x000f0f0f },
+   { 0x00400001, 0x28700061, 0x00000000, 0x00000000 },
+   { 0x00600001, 0x28900061, 0x00000000, 0x00000000 },
+   { 0x08600031, 0x21801ca1, 0x00000800, 0x0a784000 },
+   { 0x00000001, 0x25740061, 0x00000000, 0x00000000 },
+   { 0x00000001, 0x25750231, 0x00000199, 0x00000000 },
+   { 0x00000001, 0x25760231, 0x0000019a, 0x00000000 },
+   { 0x00000005, 0x24002d29, 0x00000180, 0x00030003 },
+   { 0x00000001, 0x25740231, 0x00000400, 0x00000000 },
+   { 0x00600001, 0x28600021, 0x008d01a0, 0x00000000 },
+   { 0x00600001, 0x28800021, 0x008d01c0, 0x00000000 },
+   { 0x00600001, 0x28a00021, 0x008d01e0, 0x00000000 },
+   { 0x00600001, 0x28c00021, 0x008d0200, 0x00000000 },
+   { 0x00000001, 0x244c0061, 0x00000000, 0x00203000 },
+   { 0x00600001, 0x28000021, 0x008d0440, 0x00000000 },
+   { 0x00000001, 0x24000169, 0x00000000, 0x00200020 },
+   { 0x00000001, 0x24660231, 0x00000400, 0x00000000 },
+   { 0x00600001, 0x28200021, 0x008d0460, 0x00000000 },
+   { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 },
+   { 0x0d600031, 0x21801ca1, 0x00000800, 0x0e786000 },
+   { 0x00000040, 0x24880c21, 0x00000488, 0x00000002 },
+   { 0x00600001, 0x28000021, 0x008d0480, 0x00000000 },
+   { 0x00000001, 0x28200021, 0x00000180, 0x00000000 },
+   { 0x00000001, 0x28240021, 0x00000198, 0x00000000 },
+   { 0x00000001, 0x28280021, 0x00000188, 0x00000000 },
+   { 0x00000001, 0x282c0021, 0x00000574, 0x00000000 },
+   { 0x0a800031, 0x20001cac, 0x00000800, 0x040a0003 },
+   { 0x00000040, 0x24880c21, 0x00000488, 0x00000001 },
+   { 0x00600001, 0x28000021, 0x008d0480, 0x00000000 },
+   { 0x00600001, 0x28200021, 0x008d01a0, 0x00000000 },
+   { 0x00600001, 0x28400021, 0x008d01c0, 0x00000000 },
+   { 0x00600001, 0x28600021, 0x008d01e0, 0x00000000 },
+   { 0x00600001, 0x28800021, 0x008d0200, 0x00000000 },
+   { 0x0a800031, 0x20001cac, 0x00000800, 0x0a0a0403 },
+   { 0x00000040, 0x24880c21, 0x00000488, 0x00000008 },
+   { 0x00600001, 0x28000021, 0x008d0480, 0x00000000 },
+   { 0x00600001, 0x28200021, 0x008d0240, 0x00000000 },
+   { 0x0a800031, 0x20001cac, 0x00000800, 0x040a0203 },
+   { 0x00600001, 0x2e000021, 0x008d0000, 0x00000000 },
+   { 0x07800031, 0x24001ca8, 0x00000e00, 0x82000010 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+   { 0x06000010, 0x200035ac, 0x00000fa0, 0x00000fa4 },
+   { 0x00010001, 0x2f6001ad, 0x00000fa0, 0x00000000 },
+   { 0x00110001, 0x2f6001ad, 0x00000fa4, 0x00000000 },
+   { 0x06000010, 0x200035ac, 0x00000f60, 0x00000fa8 },
+   { 0x00010001, 0x2fe401ad, 0x00000f60, 0x00000000 },
+   { 0x00110001, 0x2fe401ad, 0x00000fa8, 0x00000000 },
+   { 0x00000001, 0x34000020, 0x00000fe0, 0x00000000 },
+   { 0x04000010, 0x200035ac, 0x00000fa0, 0x00000fa4 },
+   { 0x00010001, 0x2f6001ad, 0x00000fa0, 0x00000000 },
+   { 0x00110001, 0x2f6001ad, 0x00000fa4, 0x00000000 },
+   { 0x04000010, 0x200035ac, 0x00000f60, 0x00000fa8 },
+   { 0x00010001, 0x2fe401ad, 0x00000f60, 0x00000000 },
+   { 0x00110001, 0x2fe401ad, 0x00000fa8, 0x00000000 },
+   { 0x00000001, 0x34000020, 0x00000fe0, 0x00000000 },
+   { 0x04000010, 0x200035ac, 0x00000fa0, 0x00000fa4 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000070 },
+   { 0x04000010, 0x200035ac, 0x00000fa0, 0x00000fa8 },
+   { 0x00010001, 0x2fe401ad, 0x00000fa0, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x000000a0 },
+   { 0x04000010, 0x200035ac, 0x00000fa4, 0x00000fa8 },
+   { 0x00010001, 0x2fe401ad, 0x00000fa8, 0x00000000 },
+   { 0x00110001, 0x2fe401ad, 0x00000fa4, 0x00000000 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x00000060 },
+   { 0x04000010, 0x200035ac, 0x00000fa4, 0x00000fa8 },
+   { 0x00010001, 0x2fe401ad, 0x00000fa4, 0x00000000 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000030 },
+   { 0x04000010, 0x200035ac, 0x00000fa0, 0x00000fa8 },
+   { 0x00010001, 0x2fe401ad, 0x00000fa8, 0x00000000 },
+   { 0x00110001, 0x2fe401ad, 0x00000fa0, 0x00000000 },
+   { 0x00000001, 0x34000020, 0x00000fe0, 0x00000000 },
+   { 0x00600001, 0x2f600061, 0x00000000, 0x00000000 },
+   { 0x00000001, 0x2f600231, 0x00000fa2, 0x00000000 },
+   { 0x00000005, 0x2f642d29, 0x00000fa4, 0x00030003 },
+   { 0x01000010, 0x20002d28, 0x00200f64, 0x00000000 },
+   { 0x00010005, 0x2fe42d29, 0x00000f60, 0x00030003 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000150 },
+   { 0x01000010, 0x20002d28, 0x00200f64, 0x00030003 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000100 },
+   { 0x01000010, 0x20002d28, 0x00200f64, 0x00010001 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000070 },
+   { 0x00000001, 0x2f680129, 0x00000fa8, 0x00000000 },
+   { 0x01000005, 0x20002d28, 0x00200f68, 0x00010001 },
+   { 0x00010005, 0x2fe42d29, 0x00000f60, 0x00030003 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x000000d0 },
+   { 0x00000008, 0x2f702d29, 0x00000f60, 0x00020002 },
+   { 0x00000005, 0x2fe42d29, 0x00200f70, 0x00030003 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x000000a0 },
+   { 0x00000001, 0x2f680129, 0x00000fa8, 0x00000000 },
+   { 0x05000010, 0x20002d28, 0x00200f68, 0x00020002 },
+   { 0x00010005, 0x2fe42d29, 0x00000f60, 0x00030003 },
+   { 0x00010020, 0x34001c00, 0x00001400, 0x00000060 },
+   { 0x00000008, 0x2f702d29, 0x00000f60, 0x00020002 },
+   { 0x00000005, 0x2fe42d29, 0x00200f70, 0x00030003 },
+   { 0x00000020, 0x34001c00, 0x00001400, 0x00000030 },
+   { 0x00000041, 0x2f682d29, 0x00000fa8, 0x00020002 },
+   { 0x00000008, 0x2f702529, 0x00000f60, 0x00000f68 },
+   { 0x00000005, 0x2fe42d29, 0x00200f70, 0x00030003 },
+   { 0x00000001, 0x34000020, 0x00000fe0, 0x00000000 },
diff --git a/src/shaders/vme/vme75.inc b/src/shaders/vme/vme75.inc
index a65a4eed..5a121b94 100644
--- a/src/shaders/vme/vme75.inc
+++ b/src/shaders/vme/vme75.inc
@@ -294,6 +294,7 @@ define(`mb_hwdep',           `r5.6')
 define(`MB_AVAIL',              `1:d')
 define(`MB_PRED_FLAG',          `1:w')
 
+define(`mb_pred_mode',		`r85')
 define(`mb_mvp_ref',		`r86')
 define(`mba_result',		`r87')
 define(`mbb_result',		`r88')
@@ -309,3 +310,25 @@ define(`mb_mv2',		`r95')
 define(`mb_mv3',		`r96')
 define(`mb_ref',		`r97')
 
+define(`DREF_REGION_SIZE',               `0x2020:UW')
+define(`PRED_L0',               `0x0':uw)
+define(`PRED_L1',               `0x1':uw)
+define(`PRED_BI',               `0x2':uw)
+define(`PRED_DIRECT',           `0x3':uw)
+define(`PRED_MASK',             `0x3':uw)
+
+/* The MAX search len per reference is 16 */
+define(`DSEARCH_PATH_LEN',               `0x00001212')
+define(`BI_WEIGHT',             `0x20':uw)
+define(`DSTART_CENTER',                  `0x00000000')
+define(`INTER_MASK',               	`0x03')
+define(`INTER_16X16MODE',		`0x0')
+define(`INTER_16X8MODE',		`0x01')
+define(`INTER_8X16MODE',		`0x02')
+define(`INTER_8X8MODE',			`0x03')
+define(`INTER_BLOCK0',			`0x0')
+define(`INTER_BLOCK1',			`0x1')
+define(`INTER_BLOCK2',			`0x2')
+define(`INTER_BLOCK3',			`0x3')
+define(`INTER_16X8MODE',		`0x01')
+define(`INTER_8X16MODE',		`0x02')