[Libva] [PATCH 10/14] ROI:enable on gen8 and gen9

Pengfei Qu Pengfei.Qu at intel.com
Thu Jun 30 02:33:07 UTC 2016


v1:
ROI enable on gen8 and gen9
Enable GPU to construct GPU command under ROI scenario

Signed-off-by: Zhao Yakui <yakui.zhao at intel.com>
Signed-off-by: Pengfei Qu <Pengfei.Qu at intel.com>
---
 src/gen6_mfc_common.c                |  20 ++--
 src/gen8_mfc.c                       |  55 +++++++---
 src/gen8_vme.c                       |  23 +++-
 src/gen9_mfc.c                       |  35 +++---
 src/gen9_vme.c                       |  42 +++++++-
 src/i965_drv_video.c                 |  12 ++-
 src/shaders/vme/Makefile.am          |  10 +-
 src/shaders/vme/intra_frame_gen8.asm |  21 +++-
 src/shaders/vme/intra_frame_gen8.g8b |   7 +-
 src/shaders/vme/intra_frame_gen9.g9b |   7 +-
 src/shaders/vme/mpeg2_intra_gen8.asm | 201 +++++++++++++++++++++++++++++++++++
 src/shaders/vme/mpeg2_intra_gen8.g8a |   2 +
 src/shaders/vme/mpeg2_intra_gen8.g8b |  72 +++++++++++++
 src/shaders/vme/mpeg2_intra_gen9.g9a |   2 +
 src/shaders/vme/mpeg2_intra_gen9.g9b |  72 +++++++++++++
 src/shaders/vme/vme8.inc             |   3 +
 16 files changed, 527 insertions(+), 57 deletions(-)
 create mode 100644 src/shaders/vme/mpeg2_intra_gen8.asm
 create mode 100644 src/shaders/vme/mpeg2_intra_gen8.g8a
 create mode 100644 src/shaders/vme/mpeg2_intra_gen8.g8b
 create mode 100644 src/shaders/vme/mpeg2_intra_gen9.g9a
 create mode 100644 src/shaders/vme/mpeg2_intra_gen9.g9b

diff --git a/src/gen6_mfc_common.c b/src/gen6_mfc_common.c
index 876ba41..45b0e8b 100644
--- a/src/gen6_mfc_common.c
+++ b/src/gen6_mfc_common.c
@@ -1857,6 +1857,7 @@ intel_h264_enc_roi_cbr(VADriverContextP ctx,
      * qp_delta set by user is added to base_qp, which is then clapped by
      * [base_qp-min_delta, base_qp+max_delta].
      */
+    pMiscParamROI->roi_flags.bits.roi_value_is_qp_delta = 1;
     assert (pMiscParamROI->roi_flags.bits.roi_value_is_qp_delta);
 
     sum_roi = 0.0f;
@@ -1937,6 +1938,7 @@ intel_h264_enc_roi_config(VADriverContextP ctx,
 {
     char *qp_ptr;
     int i, j;
+    struct i965_driver_data *i965 = i965_driver_data(ctx);
     VAEncMiscParameterBuffer* pMiscParamROI;
     VAEncMiscParameterBufferROI *pParamROI;
     VAEncROI *region_roi;
@@ -1946,12 +1948,13 @@ intel_h264_enc_roi_config(VADriverContextP ctx,
     int width_in_mbs = pSequenceParameter->picture_width_in_mbs;
     int height_in_mbs = pSequenceParameter->picture_height_in_mbs;
 
-	int row_start, row_end, col_start, col_end;
-   int num_roi;
+    int row_start, row_end, col_start, col_end;
+    int num_roi;
 
     vme_context->roi_enabled = 0;
     encoder_context->soft_batch_force = 0;
-	/* Restriction: Disable ROI when multi-slice is enabled */
+
+    /* Restriction: Disable ROI when multi-slice is enabled */
     if (!encoder_context->context_roi || (encode_state->num_slice_params_ext > 1))
         return;
 
@@ -1965,7 +1968,8 @@ intel_h264_enc_roi_config(VADriverContextP ctx,
     /* check whether number of ROI is correct */
     num_roi = (pParamROI->num_roi > I965_MAX_NUM_ROI_REGIONS) ? I965_MAX_NUM_ROI_REGIONS : pParamROI->num_roi;
 
-    vme_context->roi_enabled = 1;
+    if(num_roi > 0)
+        vme_context->roi_enabled = 1;
 
     if ((vme_context->saved_width_mbs !=  width_in_mbs) ||
         (vme_context->saved_height_mbs != height_in_mbs)) {
@@ -2031,9 +2035,13 @@ intel_h264_enc_roi_config(VADriverContextP ctx,
          */
         vme_context->roi_enabled = 0;
     }
-	if (vme_context->roi_enabled)
+    if (vme_context->roi_enabled)
         encoder_context->soft_batch_force = 1;
-	return;
+
+    if( IS_GEN8(i965->intel.device_info) || IS_GEN9(i965->intel.device_info) )
+        encoder_context->soft_batch_force = 0;
+
+    return;
 }
 
 /* HEVC */
diff --git a/src/gen8_mfc.c b/src/gen8_mfc.c
index 1f8e57b..186bb9b 100644
--- a/src/gen8_mfc.c
+++ b/src/gen8_mfc.c
@@ -939,8 +939,6 @@ gen8_mfc_avc_slice_state(VADriverContextP ctx,
 #define    AVC_INTER_MV_OFFSET     48
 #define    AVC_RDO_MASK            0xFFFF
 
-#if MFC_SOFTWARE_BATCH
-
 static int
 gen8_mfc_avc_pak_object_intra(VADriverContextP ctx, int x, int y, int end_mb,
                               int qp,unsigned int *msg,
@@ -1114,6 +1112,7 @@ gen8_mfc_avc_pipeline_slice_programing(VADriverContextP ctx,
     int slice_type = intel_avc_enc_slice_type_fixup(pSliceParameter->slice_type);
     int is_intra = slice_type == SLICE_TYPE_I;
     int qp_slice;
+    int qp_mb;
 
     qp_slice = qp;
     if (rate_control_mode == VA_RC_CBR) {
@@ -1156,20 +1155,24 @@ gen8_mfc_avc_pipeline_slice_programing(VADriverContextP ctx,
         x = i % width_in_mbs;
         y = i / width_in_mbs;
         msg = (unsigned int *) (msg_ptr + i * vme_context->vme_output.size_block);
+        if (vme_context->roi_enabled) {
+            qp_mb = *(vme_context->qp_per_mb + i);
+        } else
+            qp_mb = qp;
 
         if (is_intra) {
             assert(msg);
-            gen8_mfc_avc_pak_object_intra(ctx, x, y, last_mb, qp, msg, encoder_context, 0, 0, slice_batch);
+            gen8_mfc_avc_pak_object_intra(ctx, x, y, last_mb, qp_mb, msg, encoder_context, 0, 0, slice_batch);
         } else {
 	    int inter_rdo, intra_rdo;
 	    inter_rdo = msg[AVC_INTER_RDO_OFFSET] & AVC_RDO_MASK;
 	    intra_rdo = msg[AVC_INTRA_RDO_OFFSET] & AVC_RDO_MASK;
 	    offset = i * vme_context->vme_output.size_block + AVC_INTER_MV_OFFSET;
 	    if (intra_rdo < inter_rdo) { 
-                gen8_mfc_avc_pak_object_intra(ctx, x, y, last_mb, qp, msg, encoder_context, 0, 0, slice_batch);
+                gen8_mfc_avc_pak_object_intra(ctx, x, y, last_mb, qp_mb, msg, encoder_context, 0, 0, slice_batch);
             } else {
 		msg += AVC_INTER_MSG_OFFSET;
-                gen8_mfc_avc_pak_object_inter(ctx, x, y, last_mb, qp, msg, offset, encoder_context, 0, 0, pSliceParameter->slice_type, slice_batch);
+                gen8_mfc_avc_pak_object_inter(ctx, x, y, last_mb, qp_mb, msg, offset, encoder_context, 0, 0, pSliceParameter->slice_type, slice_batch);
             }
         }
     }
@@ -1217,7 +1220,6 @@ gen8_mfc_avc_software_batchbuffer(VADriverContextP ctx,
     return batch_bo;
 }
 
-#else
 
 static void
 gen8_mfc_batchbuffer_surfaces_input(VADriverContextP ctx,
@@ -1374,6 +1376,8 @@ gen8_mfc_avc_batchbuffer_slice_command(VADriverContextP ctx,
     int last_mb, slice_end_x, slice_end_y;
     int remaining_mb = total_mbs;
     uint32_t fwd_ref , bwd_ref, mb_flag;
+    char tmp_qp;
+    int number_roi_mbs, max_mb_cmds, i;
 
     last_mb = slice_param->macroblock_address + total_mbs - 1;
     slice_end_x = last_mb % width_in_mbs;
@@ -1401,13 +1405,32 @@ gen8_mfc_avc_batchbuffer_slice_command(VADriverContextP ctx,
 	number_mb_cmds = width_in_mbs;
     }
 
+    max_mb_cmds = number_mb_cmds;
+
     do {
-	if (number_mb_cmds >= remaining_mb) {
-		number_mb_cmds = remaining_mb;
-	}
 	mb_x = (slice_param->macroblock_address + starting_offset) % width_in_mbs;
 	mb_y = (slice_param->macroblock_address + starting_offset) / width_in_mbs;
 
+        number_mb_cmds = max_mb_cmds;
+        if (vme_context->roi_enabled) {
+
+            number_roi_mbs = 1;
+            tmp_qp = *(vme_context->qp_per_mb + starting_offset);
+            for (i = 1; i < max_mb_cmds; i++) {
+                if (tmp_qp != *(vme_context->qp_per_mb + starting_offset + i))
+                    break;
+
+                number_roi_mbs++;
+            }
+
+            number_mb_cmds = number_roi_mbs;
+            qp = tmp_qp;
+        }
+
+        if (number_mb_cmds >= remaining_mb) {
+            number_mb_cmds = remaining_mb;
+        }
+
         gen8_mfc_batchbuffer_emit_object_command(batch,
 						  mb_flag,
                                                   head_offset,
@@ -1585,8 +1608,6 @@ gen8_mfc_avc_hardware_batchbuffer(VADriverContextP ctx,
     return mfc_context->aux_batchbuffer_surface.bo;
 }
 
-#endif
-
 static void
 gen8_mfc_avc_pipeline_programing(VADriverContextP ctx,
                                  struct encode_state *encode_state,
@@ -1601,11 +1622,11 @@ gen8_mfc_avc_pipeline_programing(VADriverContextP ctx,
         return; 
     }
 
-#if MFC_SOFTWARE_BATCH
-    slice_batch_bo = gen8_mfc_avc_software_batchbuffer(ctx, encode_state, encoder_context);
-#else
-    slice_batch_bo = gen8_mfc_avc_hardware_batchbuffer(ctx, encode_state, encoder_context);
-#endif
+    if (encoder_context->soft_batch_force)
+        slice_batch_bo = gen8_mfc_avc_software_batchbuffer(ctx, encode_state, encoder_context);
+    else
+        slice_batch_bo = gen8_mfc_avc_hardware_batchbuffer(ctx, encode_state, encoder_context);
+
 
     // begin programing
     intel_batchbuffer_start_atomic_bcs(batch, 0x4000); 
@@ -1639,6 +1660,8 @@ gen8_mfc_avc_encode_picture(VADriverContextP ctx,
     unsigned int rate_control_mode = encoder_context->rate_control_mode;
     int current_frame_bits_size;
     int sts;
+
+    encoder_context->soft_batch_force = 0;
  
     for (;;) {
         gen8_mfc_init(ctx, encode_state, encoder_context);
diff --git a/src/gen8_vme.c b/src/gen8_vme.c
index 07dc462..c529a5b 100644
--- a/src/gen8_vme.c
+++ b/src/gen8_vme.c
@@ -96,7 +96,7 @@ static struct i965_kernel gen8_vme_kernels[] = {
 };
 
 static const uint32_t gen8_vme_mpeg2_intra_frame[][4] = {
-#include "shaders/vme/intra_frame_gen8.g8b"
+#include "shaders/vme/mpeg2_intra_gen8.g8b"
 };
 
 static const uint32_t gen8_vme_mpeg2_inter_frame[][4] = {
@@ -562,6 +562,17 @@ gen8_vme_fill_vme_batchbuffer(VADriverContextP ctx,
     int mb_x = 0, mb_y = 0;
     int i, s;
     unsigned int *command_ptr;
+    struct gen6_mfc_context *mfc_context = encoder_context->mfc_context;
+    VAEncPictureParameterBufferH264 *pic_param = (VAEncPictureParameterBufferH264 *)encode_state->pic_param_ext->buffer;
+    VAEncSliceParameterBufferH264 *slice_param = (VAEncSliceParameterBufferH264 *)encode_state->slice_params_ext[0]->buffer;
+    int qp;
+    int slice_type = intel_avc_enc_slice_type_fixup(slice_param->slice_type);
+    int qp_mb, qp_index;
+
+    if (encoder_context->rate_control_mode == VA_RC_CQP)
+        qp = pic_param->pic_init_qp + slice_param->slice_qp_delta;
+    else
+        qp = mfc_context->bit_rate_control_context[slice_type].QpPrimeY;
 
     dri_bo_map(vme_context->vme_batchbuffer.bo, 1);
     command_ptr = vme_context->vme_batchbuffer.bo->virtual;
@@ -599,7 +610,7 @@ gen8_vme_fill_vme_batchbuffer(VADriverContextP ctx,
 	    if ((i == mb_width) && slice_mb_x) {
 		mb_intra_ub &= ~(INTRA_PRED_AVAIL_FLAG_D);
 	    }
-            *command_ptr++ = (CMD_MEDIA_OBJECT | (8 - 2));
+            *command_ptr++ = (CMD_MEDIA_OBJECT | (9 - 2));
             *command_ptr++ = kernel;
             *command_ptr++ = 0;
             *command_ptr++ = 0;
@@ -609,6 +620,13 @@ gen8_vme_fill_vme_batchbuffer(VADriverContextP ctx,
             /*inline data */
             *command_ptr++ = (mb_width << 16 | mb_y << 8 | mb_x);
             *command_ptr++ = ((encoder_context->quality_level << 24) | (1 << 16) | transform_8x8_mode_flag | (mb_intra_ub << 8));
+            /* qp occupies one byte */
+            if (vme_context->roi_enabled) {
+                qp_index = mb_y * mb_width + mb_x;
+                qp_mb = *(vme_context->qp_per_mb + qp_index);
+            } else
+                qp_mb = qp;
+            *command_ptr++ = qp_mb;
 
             *command_ptr++ = CMD_MEDIA_STATE_FLUSH;
             *command_ptr++ = 0;
@@ -725,6 +743,7 @@ static VAStatus gen8_vme_prepare(VADriverContextP ctx,
 
     intel_vme_update_mbmv_cost(ctx, encode_state, encoder_context);
     intel_h264_initialize_mbmv_cost(ctx, encode_state, encoder_context);
+    intel_h264_enc_roi_config(ctx, encode_state, encoder_context);
 
     /*Setup all the memory object*/
     gen8_vme_surface_setup(ctx, encode_state, is_intra, encoder_context);
diff --git a/src/gen9_mfc.c b/src/gen9_mfc.c
index 6cca579..109e697 100644
--- a/src/gen9_mfc.c
+++ b/src/gen9_mfc.c
@@ -47,8 +47,6 @@
 #define SURFACE_STATE_OFFSET(index)             (SURFACE_STATE_PADDED_SIZE * index)
 #define BINDING_TABLE_OFFSET(index)             (SURFACE_STATE_OFFSET(MAX_MEDIA_SURFACES_GEN6) + sizeof(unsigned int) * index)
 
-#define MFC_SOFTWARE_HASWELL	1
-
 #define B0_STEP_REV		2
 #define IS_STEPPING_BPLUS(i965)	((i965->intel.revision) >= B0_STEP_REV)
 
@@ -856,8 +854,6 @@ gen9_mfc_avc_slice_state(VADriverContextP ctx,
 }
 
 
-#ifdef MFC_SOFTWARE_HASWELL
-
 static int
 gen9_mfc_avc_pak_object_intra(VADriverContextP ctx, int x, int y, int end_mb,
                               int qp,unsigned int *msg,
@@ -1037,6 +1033,7 @@ gen9_mfc_avc_pipeline_slice_programing(VADriverContextP ctx,
     int slice_type = intel_avc_enc_slice_type_fixup(pSliceParameter->slice_type);
     int is_intra = slice_type == SLICE_TYPE_I;
     int qp_slice;
+    int qp_mb;
 
     qp_slice = qp;
     if (rate_control_mode == VA_RC_CBR) {
@@ -1080,19 +1077,24 @@ gen9_mfc_avc_pipeline_slice_programing(VADriverContextP ctx,
         y = i / width_in_mbs;
         msg = (unsigned int *) (msg_ptr + i * vme_context->vme_output.size_block);
 
+        if (vme_context->roi_enabled) {
+            qp_mb = *(vme_context->qp_per_mb + i);
+        } else
+            qp_mb = qp;
+
         if (is_intra) {
             assert(msg);
-            gen9_mfc_avc_pak_object_intra(ctx, x, y, last_mb, qp, msg, encoder_context, 0, 0, slice_batch);
+            gen9_mfc_avc_pak_object_intra(ctx, x, y, last_mb, qp_mb, msg, encoder_context, 0, 0, slice_batch);
         } else {
 	    int inter_rdo, intra_rdo;
 	    inter_rdo = msg[AVC_INTER_RDO_OFFSET] & AVC_RDO_MASK;
 	    intra_rdo = msg[AVC_INTRA_RDO_OFFSET] & AVC_RDO_MASK;
 	    offset = i * vme_context->vme_output.size_block + AVC_INTER_MV_OFFSET;
 	    if (intra_rdo < inter_rdo) {
-                gen9_mfc_avc_pak_object_intra(ctx, x, y, last_mb, qp, msg, encoder_context, 0, 0, slice_batch);
+                gen9_mfc_avc_pak_object_intra(ctx, x, y, last_mb, qp_mb, msg, encoder_context, 0, 0, slice_batch);
             } else {
 		msg += AVC_INTER_MSG_OFFSET;
-                gen9_mfc_avc_pak_object_inter(ctx, x, y, last_mb, qp, msg, offset, encoder_context, 0, 0, pSliceParameter->slice_type, slice_batch);
+                gen9_mfc_avc_pak_object_inter(ctx, x, y, last_mb, qp_mb, msg, offset, encoder_context, 0, 0, pSliceParameter->slice_type, slice_batch);
             }
         }
     }
@@ -1142,8 +1144,6 @@ gen9_mfc_avc_software_batchbuffer(VADriverContextP ctx,
     return batch_bo;
 }
 
-#else
-
 static void
 gen9_mfc_batchbuffer_surfaces_input(VADriverContextP ctx,
                                     struct encode_state *encode_state,
@@ -1537,7 +1537,6 @@ gen9_mfc_avc_hardware_batchbuffer(VADriverContextP ctx,
     return mfc_context->mfc_batchbuffer_surface.bo;
 }
 
-#endif
 
 static void
 gen9_mfc_avc_pipeline_programing(VADriverContextP ctx,
@@ -1553,11 +1552,11 @@ gen9_mfc_avc_pipeline_programing(VADriverContextP ctx,
         return;
     }
 
-#ifdef MFC_SOFTWARE_HASWELL
-    slice_batch_bo = gen9_mfc_avc_software_batchbuffer(ctx, encode_state, encoder_context);
-#else
-    slice_batch_bo = gen9_mfc_avc_hardware_batchbuffer(ctx, encode_state, encoder_context);
-#endif
+    if (encoder_context->soft_batch_force)
+        slice_batch_bo = gen9_mfc_avc_software_batchbuffer(ctx, encode_state, encoder_context);
+    else
+        slice_batch_bo = gen9_mfc_avc_hardware_batchbuffer(ctx, encode_state, encoder_context);
+
 
     // begin programing
     intel_batchbuffer_start_atomic_bcs(batch, 0x4000);
@@ -1705,12 +1704,12 @@ Bool gen9_mfc_context_init(VADriverContextP ctx, struct intel_encoder_context *e
 {
     struct gen6_mfc_context *mfc_context = NULL;
 
-#if MFC_SOFTWARE_HASWELL
+
     if ((encoder_context->codec == CODEC_H264) ||
         (encoder_context->codec == CODEC_H264_MVC)) {
-        return gen8_mfc_context_init(ctx, encoder_context);
+            return gen8_mfc_context_init(ctx, encoder_context);
     }
-#endif
+
 
     if ((encoder_context->codec == CODEC_VP8) ||
         (encoder_context->codec == CODEC_MPEG2))
diff --git a/src/gen9_vme.c b/src/gen9_vme.c
index 5f9b796..6fa41dd 100644
--- a/src/gen9_vme.c
+++ b/src/gen9_vme.c
@@ -96,7 +96,7 @@ static struct i965_kernel gen9_vme_kernels[] = {
 };
 
 static const uint32_t gen9_vme_mpeg2_intra_frame[][4] = {
-#include "shaders/vme/intra_frame_gen9.g9b"
+#include "shaders/vme/mpeg2_intra_gen9.g9b"
 };
 
 static const uint32_t gen9_vme_mpeg2_inter_frame[][4] = {
@@ -352,6 +352,9 @@ gen9_vme_surface_setup(VADriverContextP ctx,
     /* VME output */
     gen9_vme_avc_output_buffer_setup(ctx, encode_state, 3, encoder_context);
     gen9_vme_avc_output_vme_batchbuffer_setup(ctx, encode_state, 5, encoder_context);
+    intel_h264_setup_cost_surface(ctx, encode_state, encoder_context,
+                                  BINDING_TABLE_OFFSET(INTEL_COST_TABLE_OFFSET),
+                                  SURFACE_STATE_OFFSET(INTEL_COST_TABLE_OFFSET));
 
     return VA_STATUS_SUCCESS;
 }
@@ -603,6 +606,17 @@ gen9_vme_fill_vme_batchbuffer(VADriverContextP ctx,
     int mb_x = 0, mb_y = 0;
     int i, s;
     unsigned int *command_ptr;
+    struct gen6_mfc_context *mfc_context = encoder_context->mfc_context;
+    VAEncPictureParameterBufferH264 *pic_param = (VAEncPictureParameterBufferH264 *)encode_state->pic_param_ext->buffer;
+    VAEncSliceParameterBufferH264 *slice_param = (VAEncSliceParameterBufferH264 *)encode_state->slice_params_ext[0]->buffer;
+    int qp;
+    int slice_type = intel_avc_enc_slice_type_fixup(slice_param->slice_type);
+    int qp_mb, qp_index;
+
+    if (encoder_context->rate_control_mode == VA_RC_CQP)
+        qp = pic_param->pic_init_qp + slice_param->slice_qp_delta;
+    else
+        qp = mfc_context->bit_rate_control_context[slice_type].QpPrimeY;
 
     dri_bo_map(vme_context->vme_batchbuffer.bo, 1);
     command_ptr = vme_context->vme_batchbuffer.bo->virtual;
@@ -640,7 +654,7 @@ gen9_vme_fill_vme_batchbuffer(VADriverContextP ctx,
             if ((i == mb_width) && slice_mb_x) {
                 mb_intra_ub &= ~(INTRA_PRED_AVAIL_FLAG_D);
             }
-            *command_ptr++ = (CMD_MEDIA_OBJECT | (8 - 2));
+            *command_ptr++ = (CMD_MEDIA_OBJECT | (9 - 2));
             *command_ptr++ = kernel;
             *command_ptr++ = 0;
             *command_ptr++ = 0;
@@ -650,6 +664,13 @@ gen9_vme_fill_vme_batchbuffer(VADriverContextP ctx,
             /*inline data */
             *command_ptr++ = (mb_width << 16 | mb_y << 8 | mb_x);
             *command_ptr++ = ((encoder_context->quality_level << 24) | (1 << 16) | transform_8x8_mode_flag | (mb_intra_ub << 8));
+            /* qp occupies one byte */
+            if (vme_context->roi_enabled) {
+                qp_index = mb_y * mb_width + mb_x;
+                qp_mb = *(vme_context->qp_per_mb + qp_index);
+            } else
+                qp_mb = qp;
+            *command_ptr++ = qp_mb;
 
             *command_ptr++ = CMD_MEDIA_STATE_FLUSH;
             *command_ptr++ = 0;
@@ -767,6 +788,8 @@ static VAStatus gen9_vme_prepare(VADriverContextP ctx,
     }
 
     intel_vme_update_mbmv_cost(ctx, encode_state, encoder_context);
+    intel_h264_initialize_mbmv_cost(ctx, encode_state, encoder_context);
+    intel_h264_enc_roi_config(ctx, encode_state, encoder_context);
 
     /*Setup all the memory object*/
     gen9_vme_surface_setup(ctx, encode_state, is_intra, encoder_context);
@@ -1776,6 +1799,21 @@ gen9_vme_context_destroy(void *context)
         vme_context->vme_state_message = NULL;
     }
 
+    if (vme_context->i_qp_cost_table) {
+        dri_bo_unreference(vme_context->i_qp_cost_table);
+        vme_context->i_qp_cost_table = NULL;
+    }
+
+    if (vme_context->p_qp_cost_table) {
+        dri_bo_unreference(vme_context->p_qp_cost_table);
+        vme_context->p_qp_cost_table = NULL;
+    }
+
+    if (vme_context->b_qp_cost_table) {
+        dri_bo_unreference(vme_context->b_qp_cost_table);
+        vme_context->b_qp_cost_table = NULL;
+    }
+
     free(vme_context);
 }
 
diff --git a/src/i965_drv_video.c b/src/i965_drv_video.c
index 9812dff..90ff146 100644
--- a/src/i965_drv_video.c
+++ b/src/i965_drv_video.c
@@ -1053,11 +1053,13 @@ i965_GetConfigAttributes(VADriverContextP ctx,
                 (profile == VAProfileH264ConstrainedBaseline ||
                  profile == VAProfileH264Main ||
                  profile == VAProfileH264High ) &&
-                 (IS_GEN7(i965->intel.device_info))) {
-                    VAConfigAttribValEncROI *roi_config = (VAConfigAttribValEncROI *)&(attrib_list[i].value);
-                    roi_config->bits.num_roi_regions = I965_MAX_NUM_ROI_REGIONS;
-                    roi_config->bits.roi_rc_priority_support = 0;
-                    roi_config->bits.roi_rc_qp_delat_support = 1;
+                 (IS_GEN7(i965->intel.device_info)||
+                 IS_GEN8(i965->intel.device_info)||
+                 IS_GEN9(i965->intel.device_info))){
+                     VAConfigAttribValEncROI *roi_config = (VAConfigAttribValEncROI *)&(attrib_list[i].value);
+                     roi_config->bits.num_roi_regions = I965_MAX_NUM_ROI_REGIONS;
+                     roi_config->bits.roi_rc_priority_support = 0;
+                     roi_config->bits.roi_rc_qp_delat_support = 1;
             }
 
             break;
diff --git a/src/shaders/vme/Makefile.am b/src/shaders/vme/Makefile.am
index 83d337a..71c93de 100644
--- a/src/shaders/vme/Makefile.am
+++ b/src/shaders/vme/Makefile.am
@@ -1,7 +1,7 @@
 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 mpeg2_inter_ivb.asm
 VME75_CORE	= batchbuffer.asm intra_frame_haswell.asm inter_frame_haswell.asm inter_bframe_haswell.asm mpeg2_inter_haswell.asm
-VME8_CORE	= intra_frame_gen8.asm	inter_frame_gen8.asm inter_bframe_gen8.asm mpeg2_inter_gen8.asm vp8_intra_frame_gen8.asm vp8_inter_frame_gen8.asm
+VME8_CORE	= intra_frame_gen8.asm	inter_frame_gen8.asm inter_bframe_gen8.asm mpeg2_intra_gen8.asm mpeg2_inter_gen8.asm vp8_intra_frame_gen8.asm vp8_inter_frame_gen8.asm
 VME9_CORE	= $(VME8_CORE)
 
 INTEL_G6B	= batchbuffer.g6b intra_frame.g6b inter_frame.g6b
@@ -20,14 +20,14 @@ INTEL_GEN75_INC	= batchbuffer.inc vme75.inc vme75_mpeg2.inc
 INTEL_GEN75_ASM	= $(INTEL_G75A:%.g75a=%.gen75.asm)
 
 
-INTEL_G8B	= intra_frame_gen8.g8b inter_frame_gen8.g8b inter_bframe_gen8.g8b mpeg2_inter_gen8.g8b vp8_intra_frame_gen8.g8b vp8_inter_frame_gen8.g8b
-INTEL_G8A	= intra_frame_gen8.g8a inter_frame_gen8.g8a inter_bframe_gen8.g8a mpeg2_inter_gen8.g8a vp8_intra_frame_gen8.g8a vp8_inter_frame_gen8.g8a
+INTEL_G8B	= intra_frame_gen8.g8b inter_frame_gen8.g8b inter_bframe_gen8.g8b mpeg2_intra_gen8.g8b mpeg2_inter_gen8.g8b vp8_intra_frame_gen8.g8b vp8_inter_frame_gen8.g8b
+INTEL_G8A	= intra_frame_gen8.g8a inter_frame_gen8.g8a inter_bframe_gen8.g8a mpeg2_intra_gen8.g8a mpeg2_inter_gen8.g8a vp8_intra_frame_gen8.g8a vp8_inter_frame_gen8.g8a
 INTEL_GEN8_INC	= vme8.inc vme75_mpeg2.inc
 INTEL_GEN8_ASM	= $(INTEL_G8A:%.g8a=%.gen8.asm)
 
 
-INTEL_G9B	= intra_frame_gen9.g9b inter_frame_gen9.g9b inter_bframe_gen9.g9b mpeg2_inter_gen9.g9b vp8_intra_frame_gen9.g9b vp8_inter_frame_gen9.g9b
-INTEL_G9A	= intra_frame_gen9.g9a inter_frame_gen9.g9a inter_bframe_gen9.g9a mpeg2_inter_gen9.g9a vp8_intra_frame_gen9.g9a vp8_inter_frame_gen9.g9a
+INTEL_G9B	= intra_frame_gen9.g9b inter_frame_gen9.g9b inter_bframe_gen9.g9b mpeg2_intra_gen9.g9b mpeg2_inter_gen9.g9b vp8_intra_frame_gen9.g9b vp8_inter_frame_gen9.g9b
+INTEL_G9A	= intra_frame_gen9.g9a inter_frame_gen9.g9a inter_bframe_gen9.g9a mpeg2_intra_gen9.g9a mpeg2_inter_gen9.g9a vp8_intra_frame_gen9.g9a vp8_inter_frame_gen9.g9a
 INTEL_GEN9_INC	= $(INTEL_GEN8_INC)
 INTEL_GEN9_ASM	= $(INTEL_G9A:%.g9a=%.gen9.asm)
 
diff --git a/src/shaders/vme/intra_frame_gen8.asm b/src/shaders/vme/intra_frame_gen8.asm
index 15b260e..ee40ec9 100644
--- a/src/shaders/vme/intra_frame_gen8.asm
+++ b/src/shaders/vme/intra_frame_gen8.asm
@@ -89,9 +89,28 @@ 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) msg_reg0.0<1>:ud   0:ud {align1};
+mov  (1) msg_reg0.20<1>:UB  thread_id_ub {align1};                  /* dispatch id */
+mov  (1) tmp_reg0.0<1>:UD   qp_ub<0,1,0>:ub {align1};
+mul  (1) msg_reg0.8<1>:ud tmp_reg0.0<0,1,0>:ud 2:ud {align1};
+send (16)
+        msg_ind
+        vme_cost_wb.0<1>:ud
+        null
+        data_port(
+                OBR_CACHE_TYPE,
+                OBR_MESSAGE_TYPE,
+                OBR_CONTROL_2,
+                BIND_IDX_COST,
+                OBR_WRITE_COMMIT_CATEGORY,
+                OBR_HEADER_PRESENT
+        )
+        mlen 1
+        rlen 1
+        {align1};
 /* m2, get the MV/Mb cost passed by constant buffer 
 when creating EU thread by MEDIA_OBJECT */       
-mov (8) vme_msg_2<1>:UD         r1.0<8,8,1>:UD {align1};
+mov (8) vme_msg_2<1>:UD         vme_cost_wb<8,8,1>:UD {align1};
 
 /* m3. This is changed for FWD/BWD cost center */
 mov (8) vme_msg_3<1>:UD		0x0:UD {align1};	        
diff --git a/src/shaders/vme/intra_frame_gen8.g8b b/src/shaders/vme/intra_frame_gen8.g8b
index 56c7283..f028f2d 100644
--- a/src/shaders/vme/intra_frame_gen8.g8b
+++ b/src/shaders/vme/intra_frame_gen8.g8b
@@ -33,7 +33,12 @@
    { 0x00000001, 0x24280e08, 0x08000000, 0x00070003 },
    { 0x00600001, 0x28000208, 0x008d0420, 0x00000000 },
    { 0x04600031, 0x26200a88, 0x0e000800, 0x02190006 },
-   { 0x00600001, 0x28400208, 0x008d0020, 0x00000000 },
+   { 0x00600001, 0x28000608, 0x00000000, 0x00000000 },
+   { 0x00000001, 0x28142288, 0x00000014, 0x00000000 },
+   { 0x00000001, 0x24002208, 0x000000a8, 0x00000000 },
+   { 0x00000041, 0x28080208, 0x06000400, 0x00000002 },
+   { 0x0a800031, 0x25a00a08, 0x0e000800, 0x02180208 },
+   { 0x00600001, 0x28400208, 0x008d05a0, 0x00000000 },
    { 0x00600001, 0x28600608, 0x00000000, 0x00000000 },
    { 0x00600001, 0x28800608, 0x00000000, 0x00000000 },
    { 0x00000001, 0x23800608, 0x00000000, 0x00000000 },
diff --git a/src/shaders/vme/intra_frame_gen9.g9b b/src/shaders/vme/intra_frame_gen9.g9b
index 63d7455..23b10ad 100644
--- a/src/shaders/vme/intra_frame_gen9.g9b
+++ b/src/shaders/vme/intra_frame_gen9.g9b
@@ -33,7 +33,12 @@
    { 0x00000001, 0x24280e08, 0x08000000, 0x00070003 },
    { 0x00600001, 0x28000208, 0x008d0420, 0x00000000 },
    { 0x0c600031, 0x26200a88, 0x06000800, 0x02190006 },
-   { 0x00600001, 0x28400208, 0x008d0020, 0x00000000 },
+   { 0x00600001, 0x28000608, 0x00000000, 0x00000000 },
+   { 0x00000001, 0x28142288, 0x00000014, 0x00000000 },
+   { 0x00000001, 0x24002208, 0x000000a8, 0x00000000 },
+   { 0x00000041, 0x28080208, 0x06000400, 0x00000002 },
+   { 0x0a800031, 0x25a00a08, 0x06000800, 0x02180208 },
+   { 0x00600001, 0x28400208, 0x008d05a0, 0x00000000 },
    { 0x00600001, 0x28600608, 0x00000000, 0x00000000 },
    { 0x00600001, 0x28800608, 0x00000000, 0x00000000 },
    { 0x00000001, 0x23800608, 0x00000000, 0x00000000 },
diff --git a/src/shaders/vme/mpeg2_intra_gen8.asm b/src/shaders/vme/mpeg2_intra_gen8.asm
new file mode 100644
index 0000000..fbf030b
--- /dev/null
+++ b/src/shaders/vme/mpeg2_intra_gen8.asm
@@ -0,0 +1,201 @@
+/*
+ * Copyright © <2010>, Intel Corporation.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL PRECISION INSIGHT AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ */
+// Modual name: IntraFrame_gen8.asm
+//
+// Make intra predition estimation for Intra frame on Gen8
+//
+
+//
+//  Now, begin source code....
+//
+
+/*
+ * __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 0x02: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};
+
+/* m2, get the MV/Mb cost passed by constant buffer
+when creating EU thread by MEDIA_OBJECT */
+mov (8) vme_msg_2<1>:UD         r1.0<8,8,1>:UD {align1};
+
+/* m3. This is changed for FWD/BWD cost center */
+mov (8) vme_msg_3<1>:UD		0x0:UD {align1};
+
+/* m4.*/
+mov (8) vme_msg_4<1>:ud		0x0:ud	{align1};
+
+/* m5 */
+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_5<1>:UD         INEP_ROW.0<8,8,1>:UD {align1};
+
+mov  (1) tmp_reg0.0<1>:UW	LUMA_CHROMA_MODE:UW {align1};
+/* Use the Luma mode */
+mov  (1) vme_msg_5.5<1>:UB	tmp_reg0.0<0,1,0>:UB {align1};
+
+/* m6 */
+mov  (8) vme_msg_6<1>:UD         0x0:UD {align1};
+mov (16) vme_msg_6.0<1>:UB       INEP_COL0.3<32,8,4>:UB {align1};
+mov  (1) vme_msg_6.16<1>:UD      INTRA_PREDICTORE_MODE {align1};
+
+/* the penalty for Intra mode */
+mov  (1) vme_msg_6.28<1>:UD	0x010101:UD {align1};
+mov  (1) vme_msg_6.20<1>:UW      CHROMA_ROW.6<0,1,0>:UW {align1};
+
+
+/* m7 */
+
+mov  (4) vme_msg_7.16<1>:UD      CHROMA_ROW.8<4,4,1>:UD {align1};
+mov  (8) vme_msg_7.0<1>:UW       CHROMA_COL.2<16,8,2>:UW {align1};
+
+/*
+ * VME message
+ */
+
+/* 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  (8) vme_msg_1<1>:UD        vme_m1.0<8,8,1>:UD {align1};
+/* m0 */
+/* 16x16 Source, Intra_harr */
+add  (1) vme_m0.12<1>:UD        vme_m0.12<0,1,0>:ud	INTRA_SAD_HAAR:UD {align1};
+mov  (8) vme_msg_0<1>:UD        vme_m0.0<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};
+
+__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};
diff --git a/src/shaders/vme/mpeg2_intra_gen8.g8a b/src/shaders/vme/mpeg2_intra_gen8.g8a
new file mode 100644
index 0000000..7abd8b9
--- /dev/null
+++ b/src/shaders/vme/mpeg2_intra_gen8.g8a
@@ -0,0 +1,2 @@
+#include "vme8.inc"
+#include "mpeg2_intra_gen8.asm"
diff --git a/src/shaders/vme/mpeg2_intra_gen8.g8b b/src/shaders/vme/mpeg2_intra_gen8.g8b
new file mode 100644
index 0000000..56c7283
--- /dev/null
+++ b/src/shaders/vme/mpeg2_intra_gen8.g8b
@@ -0,0 +1,72 @@
+   { 0x00800001, 0x24000608, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x24400608, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x24800608, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x24c00608, 0x00000000, 0x00000000 },
+   { 0x00200009, 0x24002228, 0x164500a0, 0x00040004 },
+   { 0x00000040, 0x24000a28, 0x1e000400, 0xfff8fff8 },
+   { 0x00000040, 0x24040a28, 0x1e000404, 0xffffffff },
+   { 0x00000001, 0x24080e08, 0x08000000, 0x0000001f },
+   { 0x00000001, 0x24142288, 0x00000014, 0x00000000 },
+   { 0x00200009, 0x24202228, 0x164500a0, 0x00040004 },
+   { 0x00000040, 0x24200a28, 0x1e000420, 0xfffcfffc },
+   { 0x00000001, 0x24280e08, 0x08000000, 0x000f0003 },
+   { 0x00000001, 0x24342288, 0x00000014, 0x00000000 },
+   { 0x00200009, 0x24482248, 0x164500a0, 0x00040004 },
+   { 0x00000001, 0x24542288, 0x00000014, 0x00000000 },
+   { 0x00000041, 0x24881208, 0x220000a2, 0x000000a1 },
+   { 0x00000040, 0x24880208, 0x22000488, 0x000000a0 },
+   { 0x00000041, 0x24880208, 0x06000488, 0x00000002 },
+   { 0x00000001, 0x24942288, 0x00000014, 0x00000000 },
+   { 0x00600001, 0x28000208, 0x008d0400, 0x00000000 },
+   { 0x04600031, 0x23800a88, 0x0e000800, 0x02190004 },
+   { 0x00600001, 0x28000208, 0x008d0420, 0x00000000 },
+   { 0x04600031, 0x23a00a88, 0x0e000800, 0x02290004 },
+   { 0x00200009, 0x24002228, 0x164500a0, 0x00030003 },
+   { 0x00000041, 0x24000a28, 0x1e000400, 0x00020002 },
+   { 0x00000040, 0x24000a28, 0x1e000400, 0xfff8fff8 },
+   { 0x00000040, 0x24040a28, 0x1e000404, 0xffffffff },
+   { 0x00600001, 0x28000208, 0x008d0400, 0x00000000 },
+   { 0x04600031, 0x26000a88, 0x0e000800, 0x02190006 },
+   { 0x00200009, 0x24202228, 0x164500a0, 0x00030003 },
+   { 0x00000041, 0x24200a28, 0x1e000420, 0x00020002 },
+   { 0x00000040, 0x24200a28, 0x1e000420, 0xfffcfffc },
+   { 0x00000001, 0x24280e08, 0x08000000, 0x00070003 },
+   { 0x00600001, 0x28000208, 0x008d0420, 0x00000000 },
+   { 0x04600031, 0x26200a88, 0x0e000800, 0x02190006 },
+   { 0x00600001, 0x28400208, 0x008d0020, 0x00000000 },
+   { 0x00600001, 0x28600608, 0x00000000, 0x00000000 },
+   { 0x00600001, 0x28800608, 0x00000000, 0x00000000 },
+   { 0x00000001, 0x23800608, 0x00000000, 0x00000000 },
+   { 0x00000005, 0x23840208, 0x06000384, 0xff000000 },
+   { 0x00600001, 0x28a00208, 0x008d0380, 0x00000000 },
+   { 0x00000001, 0x24001648, 0x10000000, 0x00000000 },
+   { 0x00000001, 0x28a52288, 0x00000400, 0x00000000 },
+   { 0x00600001, 0x28c00608, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x28c02288, 0x00cf03a3, 0x00000000 },
+   { 0x00000001, 0x28d00608, 0x00000000, 0x11111111 },
+   { 0x00000001, 0x28dc0608, 0x00000000, 0x00010101 },
+   { 0x00000001, 0x28d41248, 0x00000606, 0x00000000 },
+   { 0x00400001, 0x28f00208, 0x00690608, 0x00000000 },
+   { 0x00600001, 0x28e01248, 0x00ae0622, 0x00000000 },
+   { 0x00000001, 0x247c1648, 0x10000000, 0x00000000 },
+   { 0x01000005, 0x20002240, 0x160000a4, 0x00010001 },
+   { 0x00010001, 0x247c0e88, 0x08000000, 0x00000002 },
+   { 0x00000001, 0x247d2288, 0x000000a5, 0x00000000 },
+   { 0x00000001, 0x24001648, 0x10000000, 0x00200020 },
+   { 0x00000001, 0x247e2288, 0x00000400, 0x00000000 },
+   { 0x00600001, 0x28200208, 0x008d0460, 0x00000000 },
+   { 0x00000040, 0x244c0208, 0x0600044c, 0x00800000 },
+   { 0x00600001, 0x28000208, 0x008d0440, 0x00000000 },
+   { 0x0d600031, 0x21800a08, 0x0e000800, 0x10782000 },
+   { 0x00600001, 0x28000208, 0x008d0480, 0x00000000 },
+   { 0x00000001, 0x28200208, 0x00000180, 0x00000000 },
+   { 0x00000001, 0x28240208, 0x00000190, 0x00000000 },
+   { 0x00000001, 0x28280208, 0x00000194, 0x00000000 },
+   { 0x00000001, 0x282c0208, 0x00000198, 0x00000000 },
+   { 0x00000001, 0x28301248, 0x0000018c, 0x00000000 },
+   { 0x00000001, 0x28340208, 0x00000188, 0x00000000 },
+   { 0x00000001, 0x28380208, 0x0000019c, 0x00000000 },
+   { 0x00000001, 0x283c0208, 0x00000488, 0x00000000 },
+   { 0x0a800031, 0x20000a60, 0x0e000800, 0x040a0203 },
+   { 0x00600001, 0x2e000208, 0x008d0000, 0x00000000 },
+   { 0x07800031, 0x24000a40, 0x0e000e00, 0x82000010 },
diff --git a/src/shaders/vme/mpeg2_intra_gen9.g9a b/src/shaders/vme/mpeg2_intra_gen9.g9a
new file mode 100644
index 0000000..7abd8b9
--- /dev/null
+++ b/src/shaders/vme/mpeg2_intra_gen9.g9a
@@ -0,0 +1,2 @@
+#include "vme8.inc"
+#include "mpeg2_intra_gen8.asm"
diff --git a/src/shaders/vme/mpeg2_intra_gen9.g9b b/src/shaders/vme/mpeg2_intra_gen9.g9b
new file mode 100644
index 0000000..63d7455
--- /dev/null
+++ b/src/shaders/vme/mpeg2_intra_gen9.g9b
@@ -0,0 +1,72 @@
+   { 0x00800001, 0x24000608, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x24400608, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x24800608, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x24c00608, 0x00000000, 0x00000000 },
+   { 0x00200009, 0x24002228, 0x164500a0, 0x00040004 },
+   { 0x00000040, 0x24000a28, 0x1e000400, 0xfff8fff8 },
+   { 0x00000040, 0x24040a28, 0x1e000404, 0xffffffff },
+   { 0x00000001, 0x24080e08, 0x08000000, 0x0000001f },
+   { 0x00000001, 0x24142288, 0x00000014, 0x00000000 },
+   { 0x00200009, 0x24202228, 0x164500a0, 0x00040004 },
+   { 0x00000040, 0x24200a28, 0x1e000420, 0xfffcfffc },
+   { 0x00000001, 0x24280e08, 0x08000000, 0x000f0003 },
+   { 0x00000001, 0x24342288, 0x00000014, 0x00000000 },
+   { 0x00200009, 0x24482248, 0x164500a0, 0x00040004 },
+   { 0x00000001, 0x24542288, 0x00000014, 0x00000000 },
+   { 0x00000041, 0x24881208, 0x220000a2, 0x000000a1 },
+   { 0x00000040, 0x24880208, 0x22000488, 0x000000a0 },
+   { 0x00000041, 0x24880208, 0x06000488, 0x00000002 },
+   { 0x00000001, 0x24942288, 0x00000014, 0x00000000 },
+   { 0x00600001, 0x28000208, 0x008d0400, 0x00000000 },
+   { 0x0c600031, 0x23800a88, 0x06000800, 0x02190004 },
+   { 0x00600001, 0x28000208, 0x008d0420, 0x00000000 },
+   { 0x0c600031, 0x23a00a88, 0x06000800, 0x02290004 },
+   { 0x00200009, 0x24002228, 0x164500a0, 0x00030003 },
+   { 0x00000041, 0x24000a28, 0x1e000400, 0x00020002 },
+   { 0x00000040, 0x24000a28, 0x1e000400, 0xfff8fff8 },
+   { 0x00000040, 0x24040a28, 0x1e000404, 0xffffffff },
+   { 0x00600001, 0x28000208, 0x008d0400, 0x00000000 },
+   { 0x0c600031, 0x26000a88, 0x06000800, 0x02190006 },
+   { 0x00200009, 0x24202228, 0x164500a0, 0x00030003 },
+   { 0x00000041, 0x24200a28, 0x1e000420, 0x00020002 },
+   { 0x00000040, 0x24200a28, 0x1e000420, 0xfffcfffc },
+   { 0x00000001, 0x24280e08, 0x08000000, 0x00070003 },
+   { 0x00600001, 0x28000208, 0x008d0420, 0x00000000 },
+   { 0x0c600031, 0x26200a88, 0x06000800, 0x02190006 },
+   { 0x00600001, 0x28400208, 0x008d0020, 0x00000000 },
+   { 0x00600001, 0x28600608, 0x00000000, 0x00000000 },
+   { 0x00600001, 0x28800608, 0x00000000, 0x00000000 },
+   { 0x00000001, 0x23800608, 0x00000000, 0x00000000 },
+   { 0x00000005, 0x23840208, 0x06000384, 0xff000000 },
+   { 0x00600001, 0x28a00208, 0x008d0380, 0x00000000 },
+   { 0x00000001, 0x24001648, 0x10000000, 0x00000000 },
+   { 0x00000001, 0x28a52288, 0x00000400, 0x00000000 },
+   { 0x00600001, 0x28c00608, 0x00000000, 0x00000000 },
+   { 0x00800001, 0x28c02288, 0x00cf03a3, 0x00000000 },
+   { 0x00000001, 0x28d00608, 0x00000000, 0x11111111 },
+   { 0x00000001, 0x28dc0608, 0x00000000, 0x00010101 },
+   { 0x00000001, 0x28d41248, 0x00000606, 0x00000000 },
+   { 0x00400001, 0x28f00208, 0x00690608, 0x00000000 },
+   { 0x00600001, 0x28e01248, 0x00ae0622, 0x00000000 },
+   { 0x00000001, 0x247c1648, 0x10000000, 0x00000000 },
+   { 0x01000005, 0x20002240, 0x160000a4, 0x00010001 },
+   { 0x00010001, 0x247c0e88, 0x08000000, 0x00000002 },
+   { 0x00000001, 0x247d2288, 0x000000a5, 0x00000000 },
+   { 0x00000001, 0x24001648, 0x10000000, 0x00200020 },
+   { 0x00000001, 0x247e2288, 0x00000400, 0x00000000 },
+   { 0x00600001, 0x28200208, 0x008d0460, 0x00000000 },
+   { 0x00000040, 0x244c0208, 0x0600044c, 0x00800000 },
+   { 0x00600001, 0x28000208, 0x008d0440, 0x00000000 },
+   { 0x0d600031, 0x21800a08, 0x06000800, 0x10782000 },
+   { 0x00600001, 0x28000208, 0x008d0480, 0x00000000 },
+   { 0x00000001, 0x28200208, 0x00000180, 0x00000000 },
+   { 0x00000001, 0x28240208, 0x00000190, 0x00000000 },
+   { 0x00000001, 0x28280208, 0x00000194, 0x00000000 },
+   { 0x00000001, 0x282c0208, 0x00000198, 0x00000000 },
+   { 0x00000001, 0x28301248, 0x0000018c, 0x00000000 },
+   { 0x00000001, 0x28340208, 0x00000188, 0x00000000 },
+   { 0x00000001, 0x28380208, 0x0000019c, 0x00000000 },
+   { 0x00000001, 0x283c0208, 0x00000488, 0x00000000 },
+   { 0x0a800031, 0x20000a60, 0x06000800, 0x040a0203 },
+   { 0x00600001, 0x2e000208, 0x008d0000, 0x00000000 },
+   { 0x07800031, 0x24000a40, 0x06000e00, 0x82000010 },
diff --git a/src/shaders/vme/vme8.inc b/src/shaders/vme/vme8.inc
index 5f05ef2..17b0fce 100644
--- a/src/shaders/vme/vme8.inc
+++ b/src/shaders/vme/vme8.inc
@@ -59,6 +59,7 @@ define(`BIND_IDX_VME_REF0',             `1')
 define(`BIND_IDX_VME_REF1',             `2')
 define(`BIND_IDX_OUTPUT',               `3')
 define(`BIND_IDX_INEP',                 `4')
+define(`BIND_IDX_COST',                 `8')
 
 define(`SUB_PEL_MODE_INTEGER',          `0x00000000')
 define(`SUB_PEL_MODE_HALF',             `0x00001000')
@@ -158,6 +159,7 @@ define(`transform_8x8_ub',      `inline_reg0.4')
 define(`input_mb_intra_ub',     `inline_reg0.5')
 define(`num_macroblocks',       `inline_reg0.6')
 define(`quality_level_ub',      `inline_reg0.7')
+define(`qp_ub',                 `inline_reg0.8')
 
 /*
  * GRF 6~11 -- reserved
@@ -242,6 +244,7 @@ define(`tmp_uw7',               `tmp_rega.14')
 
 define(`vme_m2',                `r43')
 define(`vme_m3',                `r44')
+define(`vme_cost_wb',                `r45')
 /*
  * MRF registers
  */        
-- 
2.7.4



More information about the Libva mailing list