summaryrefslogtreecommitdiff
diff options
context:
space:
mode:
authorZhao Yakui <yakui.zhao@intel.com>2013-08-12 15:13:23 +0800
committerXiang, Haihao <haihao.xiang@intel.com>2013-09-06 15:31:21 +0800
commita27920e84023e9be75997db6b61bbe94f36f2963 (patch)
tree26fba243b55351526a73c00b6cf8ca32f9df89f4
parent679529ebebe889afa4ffb3b7459218b41804bfac (diff)
Rewrite inter-frame shader for MPEG2 encoding on HSW to follow MPEG2 spec
Now the MPEG2/H264 uses the same mode/motion vector prediction shader. But the MV search region of mpeg2 is different with that on H264, which causes that the wrong mode/motion vector prediction is used for MPEG2. Signed-off-by: Zhao Yakui <yakui.zhao@intel.com> (cherry picked from commit 6842f08aa375b5942cee4b9d06421609c212895a)
-rw-r--r--src/gen6_mfc_common.c15
-rw-r--r--src/gen6_vme.h8
-rw-r--r--src/gen75_vme.c3
-rw-r--r--src/shaders/vme/Makefile.am6
-rw-r--r--src/shaders/vme/mpeg2_inter_haswell.asm696
-rw-r--r--src/shaders/vme/mpeg2_inter_haswell.g75a3
-rw-r--r--src/shaders/vme/mpeg2_inter_haswell.g75b278
-rw-r--r--src/shaders/vme/vme75_mpeg2.inc3
8 files changed, 1008 insertions, 4 deletions
diff --git a/src/gen6_mfc_common.c b/src/gen6_mfc_common.c
index c2ec080..b77916d 100644
--- a/src/gen6_mfc_common.c
+++ b/src/gen6_mfc_common.c
@@ -1111,3 +1111,18 @@ intel_mfc_avc_ref_idx_state(VADriverContextP ctx,
}
ADVANCE_BCS_BATCH(batch);
}
+
+
+void intel_vme_mpeg2_state_setup(VADriverContextP ctx,
+ struct encode_state *encode_state,
+ struct intel_encoder_context *encoder_context)
+{
+ struct gen6_vme_context *vme_context = encoder_context->vme_context;
+ uint32_t *vme_state_message = (uint32_t *)(vme_context->vme_state_message);
+ VAEncSequenceParameterBufferMPEG2 *seq_param = (VAEncSequenceParameterBufferMPEG2 *)encode_state->seq_param_ext->buffer;
+ int width_in_mbs = ALIGN(seq_param->picture_width, 16) / 16;
+ int height_in_mbs = ALIGN(seq_param->picture_height, 16) / 16;
+
+ vme_state_message[MPEG2_PIC_WIDTH_HEIGHT] = (height_in_mbs << 16) |
+ width_in_mbs;
+}
diff --git a/src/gen6_vme.h b/src/gen6_vme.h
index 17f199e..9d7acdf 100644
--- a/src/gen6_vme.h
+++ b/src/gen6_vme.h
@@ -87,6 +87,9 @@ struct gen6_vme_context
unsigned int vme_kernel_sum;
};
+#define MPEG2_PIC_WIDTH_HEIGHT 30
+
+
Bool gen75_vme_context_init(VADriverContextP ctx, struct intel_encoder_context *encoder_context);
extern void intel_vme_update_mbmv_cost(VADriverContextP ctx,
@@ -137,4 +140,9 @@ gen7_vme_walker_fill_vme_batchbuffer(VADriverContextP ctx,
extern void
gen7_vme_scoreboard_init(VADriverContextP ctx, struct gen6_vme_context *vme_context);
+extern void
+intel_vme_mpeg2_state_setup(VADriverContextP ctx,
+ struct encode_state *encode_state,
+ struct intel_encoder_context *encoder_context);
+
#endif /* _GEN6_VME_H_ */
diff --git a/src/gen75_vme.c b/src/gen75_vme.c
index 231ffc6..41b8564 100644
--- a/src/gen75_vme.c
+++ b/src/gen75_vme.c
@@ -113,7 +113,7 @@ static const uint32_t gen75_vme_mpeg2_intra_frame[][4] = {
};
static const uint32_t gen75_vme_mpeg2_inter_frame[][4] = {
-#include "shaders/vme/mpeg2_inter_frame_haswell.g75b"
+#include "shaders/vme/mpeg2_inter_haswell.g75b"
};
static const uint32_t gen75_vme_mpeg2_batchbuffer[][4] = {
@@ -952,6 +952,7 @@ gen75_vme_mpeg2_prepare(VADriverContextP ctx,
gen75_vme_mpeg2_surface_setup(ctx, encode_state, slice_param->is_intra_slice, encoder_context);
gen75_vme_interface_setup(ctx, encode_state, encoder_context);
gen75_vme_vme_state_setup(ctx, encode_state, slice_param->is_intra_slice, encoder_context);
+ intel_vme_mpeg2_state_setup(ctx, encode_state, encoder_context);
gen75_vme_constant_setup(ctx, encode_state, encoder_context);
/*Programing media pipeline*/
diff --git a/src/shaders/vme/Makefile.am b/src/shaders/vme/Makefile.am
index e3c401d..27a1828 100644
--- a/src/shaders/vme/Makefile.am
+++ b/src/shaders/vme/Makefile.am
@@ -1,6 +1,6 @@
VME_CORE = batchbuffer.asm intra_frame.asm inter_frame.asm
VME7_CORE = batchbuffer.asm intra_frame_ivb.asm inter_frame_ivb.asm inter_bframe_ivb..asm
-VME75_CORE = batchbuffer.asm intra_frame_haswell.asm inter_frame_haswell.asm inter_bframe_haswell.asm
+VME75_CORE = batchbuffer.asm intra_frame_haswell.asm inter_frame_haswell.asm inter_bframe_haswell.asm mpeg2_inter_haswell.asm
INTEL_G6B = batchbuffer.g6b intra_frame.g6b inter_frame.g6b
INTEL_G6A = batchbuffer.g6a intra_frame.g6a inter_frame.g6a
@@ -12,8 +12,8 @@ INTEL_G7A = batchbuffer.g7a intra_frame.g7a inter_frame.g7a mpeg2_inter_frame.g7
INTEL_GEN7_INC = batchbuffer.inc vme.inc vme7_mpeg2.inc vme7.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 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_G75B = batchbuffer.g75b intra_frame_haswell.g75b inter_frame_haswell.g75b mpeg2_inter_frame_haswell.g75b inter_bframe_haswell.g75b mpeg2_inter_haswell.g75b
+INTEL_G75A = batchbuffer.g75a intra_frame_haswell.g75a inter_frame_haswell.g75a mpeg2_inter_frame_haswell.g75a inter_bframe_haswell.g75a mpeg2_inter_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/mpeg2_inter_haswell.asm b/src/shaders/vme/mpeg2_inter_haswell.asm
new file mode 100644
index 0000000..17e4843
--- /dev/null
+++ b/src/shaders/vme/mpeg2_inter_haswell.asm
@@ -0,0 +1,696 @@
+/*
+ * 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.
+ * Author : Zhao Yakui <yakui.zhao@intel.com>
+ */
+// Modual name: mpeg2_inter_haswell.asm
+//
+// Make MPEG2 inter predition estimation for Inter-frame on Haswell
+//
+
+//
+// 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
+ */
+__INTER_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) 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 */
+
+shl (2) pic_ref.0<1>:uw r4.24<2,2,1>:uw 4:uw {align1};
+mov (2) pic_ref.16<1>:uw r4.20<2,2,1>:uw {align1};
+mov (8) mb_mvp_ref.0<1>:ud 0:ud {align1};
+mov (8) mb_ref_win.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 2 oword (16bytes), 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_2,
+ OBR_BIND_IDX,
+ OBR_WRITE_COMMIT_CATEGORY,
+ OBR_HEADER_PRESENT
+ )
+ mlen 1
+ rlen 1
+ {align1};
+/* TODO: RefID is required after multi-references are added */
+/* MV */
+mov (2) mba_result.4<1>:ud mb_mv0.0<2,2,1>:ud {align1};
+mov (1) mba_result.16<1>:w MB_PRED_FLAG {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, neither 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_2,
+ OBR_BIND_IDX,
+ OBR_WRITE_COMMIT_CATEGORY,
+ OBR_HEADER_PRESENT
+ )
+ mlen 1
+ rlen 1
+ {align1};
+/* TODO: RefID is required after multi-references are added */
+mov (2) mbb_result.4<1>:ud mb_mv0.0<2,2,1>:ud {align1};
+mov (1) mbb_result.16<1>:w MB_PRED_FLAG {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_2,
+ OBR_BIND_IDX,
+ OBR_WRITE_COMMIT_CATEGORY,
+ OBR_HEADER_PRESENT
+ )
+ mlen 1
+ rlen 1
+ {align1};
+/* TODO: RefID is required after multi-references are added */
+/* Forward MV */
+mov (2) mbc_result.4<1>:ud mb_mv0.0<2,2,1>:ud {align1};
+mov (1) mbc_result.16<1>:w MB_PRED_FLAG {align1};
+
+jmpi (1) mb_mvp_start;
+mbd_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_D:uw {align1};
+(f0.0) jmpi (1) mb_mvp_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_2,
+ OBR_BIND_IDX,
+ OBR_WRITE_COMMIT_CATEGORY,
+ OBR_HEADER_PRESENT
+ )
+ mlen 1
+ rlen 1
+ {align1};
+
+/* TODO: RefID is required after multi-references are added */
+
+/* Forward MV */
+mov (2) mbc_result.4<1>:ud mb_mv0.0<2,2,1>:ud {align1};
+mov (1) mbc_result.16<1>:w MB_PRED_FLAG {align1};
+
+mb_mvp_start:
+/*TODO: Add the skip prediction */
+/* Check whether both MB B and C are inavailable */
+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 (1) mbb_result.4<1>:ud mba_result.4<0,1,0>:ud {align1};
+(f0.0) mov (1) mbc_result.4<1>:ud mba_result.4<0,1,0>:ud {align1};
+(f0.0) mov (1) mbb_result.20<1>:uw mba_result.20<0,1,0>:uw {align1};
+(f0.0) mov (1) mbc_result.20<1>:uw mba_result.20<0,1,0>:uw {align1};
+(f0.0) mov (1) mb_mvp_ref.0<1>:ud mba_result.4<0,1,0>:ud {align1};
+(-f0.0) mov (1) mb_mvp_ref.0<1>:ud 0:ud {align1};
+jmpi (1) __mb_hwdep_end;
+
+mb_median_start:
+/* 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) __mb_hwdep_end;
+
+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};
+
+__mb_hwdep_end:
+
+/* Calibrate the ref window for MPEG2 */
+mov (1) vme_m0.0<1>:W -16:W {align1};
+mov (1) vme_m0.2<1>:W -12:W {align1};
+
+mov (1) INPUT_ARG0.0<1>:ud vme_m0.0<0,1,0>:ud {align1};
+mov (1) INPUT_ARG0.8<1>:ud vme_m0.8<0,1,0>:ud {align1};
+mov (8) INPUT_ARG1.0<1>:ud pic_ref.0<8,8,1>:ud {align1};
+
+SAVE_RET {align1};
+jmpi (1) ref_boundary_check;
+mov (2) vme_m0.0<1>:w RET_ARG<2,2,1>:w {align1};
+
+/* 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};
+
+/* the neighbour pixel is zero for MPEG2 Intra-prediction */
+
+/* m4 */
+mov (8) vme_msg_4<1>:UD 0: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};
+mov (1) tmp_reg0.0<1>:UW INTRA16_DC_PRED:UW {align1};
+mov (1) vme_msg_4.4<1>:ub tmp_reg0.0<0,1,0>:UB {align1};
+
+/* m5 */
+mov (8) vme_msg_5<1>:UD 0x0:UD {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};
+
+
+/* m6 */
+mov (8) vme_msg_6.0<1>:UD 0:Ud {align1};
+
+/*
+ * SIC VME message
+ */
+/* m0 */
+mov (8) vme_msg_0.0<1>:UD vme_m0.0<8,8,1>:UD {align1};
+
+/* Disable Intra8x8/Intra4x4 Intra-prediction */
+/* m1 */
+mov (8) vme_m1.0<1>:ud 0x0:UD {align1};
+
+mov (1) intra_flag<1>:UW 0x0:UW {align1} ;
+mov (1) tmp_reg0.0<1>:uw LUMA_INTRA_8x8_DISABLE:uw {align1};
+add (1) tmp_reg0.0<1>:uw tmp_reg0.0<0,1,0>:uw LUMA_INTRA_4x4_DISABLE:uw {align1};
+mov (1) intra_part_mask_ub<1>:UB tmp_reg0.0<0,1,0>:ub {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};
+
+/* Enable DC HAAR component when calculating HARR SATD block */
+mov (1) tmp_reg0.0<1>:UW DC_HARR_ENABLE: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_SINGLE + INTER_PART_MASK + INTER_SAD_HAAR:UD {align1}; /* 16x16 Source, harr */
+mov (1) vme_m0.22<1>:UW REF_REGION_SIZE {align1}; /* Reference Width&Height, 48x40 */
+
+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 START_CENTER + SEARCH_PATH_LEN:UD {align1};
+/* Set the MV cost center */
+mov (1) vme_m1.16<1>:ud 0:ud {align1};
+mov (1) vme_m1.20<1>:ud 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 0x01010101:UD {align1};
+mov (1) vme_msg_3.4<1>:UD 0x10010101:UD {align1};
+mov (1) vme_msg_3.8<1>:UD 0x0F0F0F0F:UD {align1};
+mov (1) vme_msg_3.12<1>:UD 0x100F0F0F:UD {align1};
+mov (1) vme_msg_3.16<1>:UD 0x01010101:UD {align1};
+mov (1) vme_msg_3.20<1>:UD 0x10010101:UD {align1};
+mov (1) vme_msg_3.24<1>:UD 0x0F0F0F0F:UD {align1};
+mov (1) vme_msg_3.28<1>:UD 0x100F0F0F:UD {align1};
+
+mov (1) vme_msg_4.0<1>:UD 0x01010101:UD {align1};
+mov (1) vme_msg_4.4<1>:UD 0x10010101:UD {align1};
+mov (1) vme_msg_4.8<1>:UD 0x0F0F0F0F:UD {align1};
+mov (1) vme_msg_4.12<1>:UD 0x000F0F0F:UD {align1};
+
+mov (4) 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};
+
+mov (1) vme_m0.12<1>:UD INTER_SAD_HAAR + SUB_PEL_MODE_HALF + FBR_BME_DISABLE:UD {align1}; /* 16x16 Source, 1/2 pixel, harr, BME disable */
+
+/* Bilinear filter */
+mov (1) tmp_reg0.0<1>:uw 0x04:uw {align1};
+add (1) vme_m1.30<1>:ub vme_m1.30<0,1,0>:ub tmp_reg0.0<0,1,0>:ub {align1};
+
+mov (8) vme_msg_0.0<1>:UD vme_m0.0<8,8,1>:UD {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_2,
+ OBW_BIND_IDX,
+ OBW_WRITE_COMMIT_CATEGORY,
+ OBW_HEADER_PRESENT
+ )
+ mlen 2
+ 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};
+
+/* Issue message fence so that the previous write message is committed */
+send (16)
+ mb_ind
+ mb_wb.0<1>:ud
+ NULL
+ data_port(
+ OBR_CACHE_TYPE,
+ OBR_MESSAGE_FENCE,
+ OBR_MF_COMMIT,
+ OBR_BIND_IDX,
+ OBR_WRITE_COMMIT_CATEGORY,
+ OBR_HEADER_PRESENT
+ )
+ mlen 1
+ rlen 1
+ {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 ;
+
+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};
+
+nop;
+nop;
+
+ref_boundary_check:
+
+/* The left/up coordinate of reference window */
+add (2) TEMP_VAR0.0<1>:w INPUT_ARG0.8<2,2,1>:w INPUT_ARG0.0<2,2,1>:w {align1};
+/* The right/bottom coordinate of reference window */
+add (1) TEMP_VAR0.16<1>:w TEMP_VAR0.0<0,1,0>:w 48:w {align1};
+add (1) TEMP_VAR0.18<1>:w TEMP_VAR0.2<0,1,0>:w 40:w {align1};
+x_left_cmp:
+ cmp.l.f0.0 (1) null:w TEMP_VAR0.0<0,1,0>:w 0:w {align1};
+ (-f0.0) jmpi (1) x_right_cmp;
+ (f0.0) mov (1) TEMP_VAR0.0<1>:w 0:w {align1};
+ jmpi (1) y_top_cmp;
+x_right_cmp:
+ cmp.g.f0.0 (1) null:w TEMP_VAR0.16<0,1,0>:w INPUT_ARG1.0<0,1,0>:w {align1};
+ (-f0.0) jmpi (1) y_top_cmp;
+ (f0.0) add (1) TEMP_VAR0.0<1>:w INPUT_ARG1.0<0,1,0>:w -48:w {align1};
+y_top_cmp:
+ cmp.l.f0.0 (1) null:w TEMP_VAR0.2<0,1,0>:w 0:w {align1};
+ (-f0.0) jmpi (1) y_bottom_cmp;
+ (f0.0) mov (1) TEMP_VAR0.2<1>:w 0:w {align1};
+ jmpi (1) y_bottom_end;
+y_bottom_cmp:
+ cmp.g.f0.0 (1) null:w TEMP_VAR0.18<0,1,0>:w INPUT_ARG1.2<0,1,0>:w {align1};
+ (f0.0) add (1) TEMP_VAR0.2<1>:w INPUT_ARG1.2<0,1,0>:w -40:w {align1};
+
+y_bottom_end:
+mul (2) TEMP_VAR1.0<1>:w INPUT_ARG0.8<2,2,1>:w -1:w {align1};
+add (2) RET_ARG<1>:w TEMP_VAR0.0<2,2,1>:w TEMP_VAR1.0<2,2,1>:w {align1};
+ RETURN {align1};
+nop;
+nop;
diff --git a/src/shaders/vme/mpeg2_inter_haswell.g75a b/src/shaders/vme/mpeg2_inter_haswell.g75a
new file mode 100644
index 0000000..355812c
--- /dev/null
+++ b/src/shaders/vme/mpeg2_inter_haswell.g75a
@@ -0,0 +1,3 @@
+#include "vme75.inc"
+#include "vme75_mpeg2.inc"
+#include "mpeg2_inter_haswell.asm"
diff --git a/src/shaders/vme/mpeg2_inter_haswell.g75b b/src/shaders/vme/mpeg2_inter_haswell.g75b
new file mode 100644
index 0000000..b3e67e7
--- /dev/null
+++ b/src/shaders/vme/mpeg2_inter_haswell.g75b
@@ -0,0 +1,278 @@
+ { 0x00800001, 0x24000061, 0x00000000, 0x00000000 },
+ { 0x00800001, 0x24400061, 0x00000000, 0x00000000 },
+ { 0x00800001, 0x24800061, 0x00000000, 0x00000000 },
+ { 0x00800001, 0x24c00061, 0x00000000, 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 },
+ { 0x00200009, 0x2a402d29, 0x00450098, 0x00040004 },
+ { 0x00200001, 0x2a500129, 0x00450094, 0x00000000 },
+ { 0x00600001, 0x2ac00061, 0x00000000, 0x00000000 },
+ { 0x00600001, 0x2a800061, 0x00000000, 0x00000000 },
+ { 0x01000005, 0x20002d28, 0x000000a6, 0x00040004 },
+ { 0x00010020, 0x34001c00, 0x00001400, 0x00000740 },
+ { 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, 0x000000f0 },
+ { 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, 0x00000040 },
+ { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 },
+ { 0x0a800031, 0x2ba01ca1, 0x00000b40, 0x02180203 },
+ { 0x00200001, 0x2ae40021, 0x00450ba0, 0x00000000 },
+ { 0x00000001, 0x2af001ed, 0x00000000, 0x00010001 },
+ { 0x00600001, 0x2b400061, 0x00000000, 0x00000000 },
+ { 0x01000005, 0x20002e28, 0x000000a5, 0x00100010 },
+ { 0x00210001, 0x2b1401ed, 0x00000000, 0xffffffff },
+ { 0x00210001, 0x2b3401ed, 0x00000000, 0xffffffff },
+ { 0x00010020, 0x34001c00, 0x00001400, 0x00000350 },
+ { 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, 0x00000040 },
+ { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 },
+ { 0x0a800031, 0x2ba01ca1, 0x00000b40, 0x02180203 },
+ { 0x00200001, 0x2b040021, 0x00450ba0, 0x00000000 },
+ { 0x00000001, 0x2b1001ed, 0x00000000, 0x00010001 },
+ { 0x00600001, 0x2b400061, 0x00000000, 0x00000000 },
+ { 0x01000005, 0x20002e28, 0x000000a5, 0x00080008 },
+ { 0x00010020, 0x34001c00, 0x00001400, 0x00000110 },
+ { 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, 0x00000170 },
+ { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 },
+ { 0x0a800031, 0x2ba01ca1, 0x00000b40, 0x02180203 },
+ { 0x00200001, 0x2b240021, 0x00450ba0, 0x00000000 },
+ { 0x00000001, 0x2b3001ed, 0x00000000, 0x00010001 },
+ { 0x00000020, 0x34001c00, 0x00001400, 0x00000120 },
+ { 0x00600001, 0x2b400061, 0x00000000, 0x00000000 },
+ { 0x01000005, 0x20002e28, 0x000000a5, 0x00040004 },
+ { 0x00010020, 0x34001c00, 0x00001400, 0x000000f0 },
+ { 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, 0x00000040 },
+ { 0x00000040, 0x2b480c21, 0x00000b48, 0x00000003 },
+ { 0x0a800031, 0x2ba01cb1, 0x00000b40, 0x02180203 },
+ { 0x00200001, 0x2b240021, 0x00450ba0, 0x00000000 },
+ { 0x00000001, 0x2b3001ed, 0x00000000, 0x00010001 },
+ { 0x00000040, 0x240014a5, 0x00000b00, 0x00000b20 },
+ { 0x01000010, 0x20001ca4, 0x00000400, 0x00000000 },
+ { 0x00110020, 0x34001c00, 0x00001400, 0x00000080 },
+ { 0x02000010, 0x20001ca4, 0x00000ae0, 0x00000001 },
+ { 0x00010001, 0x2b040021, 0x00000ae4, 0x00000000 },
+ { 0x00010001, 0x2b240021, 0x00000ae4, 0x00000000 },
+ { 0x00010001, 0x2b140129, 0x00000af4, 0x00000000 },
+ { 0x00010001, 0x2b340129, 0x00000af4, 0x00000000 },
+ { 0x00010001, 0x2ac00021, 0x00000ae4, 0x00000000 },
+ { 0x00110001, 0x2ac00061, 0x00000000, 0x00000000 },
+ { 0x00000020, 0x34001c00, 0x00001400, 0x00000190 },
+ { 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, 0x00000720 },
+ { 0x00000001, 0x2ac001ad, 0x00000fe4, 0x00000000 },
+ { 0x00000001, 0x2fa001ad, 0x00000ae6, 0x00000000 },
+ { 0x00000001, 0x2fa401ad, 0x00000b06, 0x00000000 },
+ { 0x00000001, 0x2fa801ad, 0x00000b26, 0x00000000 },
+ { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+ { 0x00000020, 0x34001c00, 0x00001400, 0x000006c0 },
+ { 0x00000001, 0x2ac201ad, 0x00000fe4, 0x00000000 },
+ { 0x00000001, 0x244001ed, 0x00000000, 0xfff0fff0 },
+ { 0x00000001, 0x244201ed, 0x00000000, 0xfff4fff4 },
+ { 0x00000001, 0x2fa00021, 0x00000440, 0x00000000 },
+ { 0x00000001, 0x2fa80021, 0x00000448, 0x00000000 },
+ { 0x00600001, 0x2fc00021, 0x008d0a40, 0x00000000 },
+ { 0x00000040, 0x2fe00c01, 0x00001400, 0x00000020 },
+ { 0x00000020, 0x34001c00, 0x00001400, 0x00000760 },
+ { 0x00200001, 0x244001ad, 0x00450fe4, 0x00000000 },
+ { 0x00600001, 0x25600021, 0x008d0020, 0x00000000 },
+ { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 },
+ { 0x00600001, 0x28600061, 0x00000000, 0x00000000 },
+ { 0x00600001, 0x28800061, 0x00000000, 0x00000000 },
+ { 0x00000001, 0x24000169, 0x00000000, 0x00010001 },
+ { 0x00000001, 0x28850231, 0x00000400, 0x00000000 },
+ { 0x00000001, 0x24000169, 0x00000000, 0x00bb00bb },
+ { 0x00000001, 0x28840231, 0x00000400, 0x00000000 },
+ { 0x00600001, 0x28a00061, 0x00000000, 0x00000000 },
+ { 0x00000001, 0x28b00061, 0x00000000, 0x11111111 },
+ { 0x00000001, 0x28bc0061, 0x00000000, 0x00010101 },
+ { 0x00600001, 0x28c00061, 0x00000000, 0x00000000 },
+ { 0x00600001, 0x28000021, 0x008d0440, 0x00000000 },
+ { 0x00600001, 0x24600061, 0x00000000, 0x00000000 },
+ { 0x00000001, 0x247c0169, 0x00000000, 0x00000000 },
+ { 0x00000001, 0x24000169, 0x00000000, 0x00020002 },
+ { 0x00000040, 0x24002d29, 0x00000400, 0x00040004 },
+ { 0x00000001, 0x247c0231, 0x00000400, 0x00000000 },
+ { 0x00000001, 0x247d0231, 0x000000a5, 0x00000000 },
+ { 0x00000001, 0x24000169, 0x00000000, 0x00000000 },
+ { 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, 0x7e200000 },
+ { 0x00000001, 0x24560169, 0x00000000, 0x28302830 },
+ { 0x00000001, 0x24440021, 0x00000440, 0x00000000 },
+ { 0x00600001, 0x28000021, 0x008d0440, 0x00000000 },
+ { 0x00000001, 0x24600061, 0x00000000, 0x00000002 },
+ { 0x00000001, 0x24640231, 0x0000009c, 0x00000000 },
+ { 0x00000001, 0x24680061, 0x00000000, 0x30003030 },
+ { 0x00000001, 0x24700061, 0x00000000, 0x00000000 },
+ { 0x00000001, 0x24740061, 0x00000000, 0x00000000 },
+ { 0x00600001, 0x28200021, 0x008d0460, 0x00000000 },
+ { 0x00600001, 0x28400021, 0x008d0560, 0x00000000 },
+ { 0x00000001, 0x28600061, 0x00000000, 0x01010101 },
+ { 0x00000001, 0x28640061, 0x00000000, 0x10010101 },
+ { 0x00000001, 0x28680061, 0x00000000, 0x0f0f0f0f },
+ { 0x00000001, 0x286c0061, 0x00000000, 0x100f0f0f },
+ { 0x00000001, 0x28700061, 0x00000000, 0x01010101 },
+ { 0x00000001, 0x28740061, 0x00000000, 0x10010101 },
+ { 0x00000001, 0x28780061, 0x00000000, 0x0f0f0f0f },
+ { 0x00000001, 0x287c0061, 0x00000000, 0x100f0f0f },
+ { 0x00000001, 0x28800061, 0x00000000, 0x01010101 },
+ { 0x00000001, 0x28840061, 0x00000000, 0x10010101 },
+ { 0x00000001, 0x28880061, 0x00000000, 0x0f0f0f0f },
+ { 0x00000001, 0x288c0061, 0x00000000, 0x000f0f0f },
+ { 0x00400001, 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, 0x00241000 },
+ { 0x00000001, 0x24000169, 0x00000000, 0x00040004 },
+ { 0x00000040, 0x247e4631, 0x0000047e, 0x00000400 },
+ { 0x00600001, 0x28000021, 0x008d0440, 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, 0x040a0203 },
+ { 0x00000040, 0x24880c21, 0x00000488, 0x00000008 },
+ { 0x00600001, 0x28000021, 0x008d0480, 0x00000000 },
+ { 0x00600001, 0x28200021, 0x008d0240, 0x00000000 },
+ { 0x0a800031, 0x20001cac, 0x00000800, 0x040a0203 },
+ { 0x0a800031, 0x2b601ca1, 0x00000b40, 0x0219e003 },
+ { 0x00600001, 0x2e000021, 0x008d0000, 0x00000000 },
+ { 0x07800031, 0x24001ca8, 0x00000e00, 0x82000010 },
+ { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+ { 0x0000007e, 0x00000000, 0x00000000, 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 },
+ { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+ { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+ { 0x00200040, 0x2f6035ad, 0x00450fa8, 0x00450fa0 },
+ { 0x00000040, 0x2f703dad, 0x00000f60, 0x00300030 },
+ { 0x00000040, 0x2f723dad, 0x00000f62, 0x00280028 },
+ { 0x05000010, 0x20003dac, 0x00000f60, 0x00000000 },
+ { 0x00110020, 0x34001c00, 0x00001400, 0x00000020 },
+ { 0x00010001, 0x2f6001ed, 0x00000000, 0x00000000 },
+ { 0x00000020, 0x34001c00, 0x00001400, 0x00000030 },
+ { 0x03000010, 0x200035ac, 0x00000f70, 0x00000fc0 },
+ { 0x00110020, 0x34001c00, 0x00001400, 0x00000010 },
+ { 0x00010040, 0x2f603dad, 0x00000fc0, 0xffd0ffd0 },
+ { 0x05000010, 0x20003dac, 0x00000f62, 0x00000000 },
+ { 0x00110020, 0x34001c00, 0x00001400, 0x00000020 },
+ { 0x00010001, 0x2f6201ed, 0x00000000, 0x00000000 },
+ { 0x00000020, 0x34001c00, 0x00001400, 0x00000020 },
+ { 0x03000010, 0x200035ac, 0x00000f72, 0x00000fc2 },
+ { 0x00010040, 0x2f623dad, 0x00000fc2, 0xffd8ffd8 },
+ { 0x00200041, 0x2f803dad, 0x00450fa8, 0xffffffff },
+ { 0x00200040, 0x2fe435ad, 0x00450f60, 0x00450f80 },
+ { 0x00000001, 0x34000020, 0x00000fe0, 0x00000000 },
+ { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
+ { 0x0000007e, 0x00000000, 0x00000000, 0x00000000 },
diff --git a/src/shaders/vme/vme75_mpeg2.inc b/src/shaders/vme/vme75_mpeg2.inc
index 9b877ac..c8e42cc 100644
--- a/src/shaders/vme/vme75_mpeg2.inc
+++ b/src/shaders/vme/vme75_mpeg2.inc
@@ -16,3 +16,6 @@
*/
define(`INTER_PART_MASK', `0x7e000000')
+define(`mpeg2_ref', `r83')
+define(`pic_ref', `r82')
+define(`INTRA16_DC_PRED', `0xBB')