Signed-off-by: Qu,Pengfei <pengfei...@intel.com>
---
src/gen9_mfc_hevc.c | 317 +++++++++++++++++++++++++++-------------------------
1 file changed, 164 insertions(+), 153 deletions(-)
diff --git a/src/gen9_mfc_hevc.c b/src/gen9_mfc_hevc.c
index 927e36b..94eb38a 100644
--- a/src/gen9_mfc_hevc.c
+++ b/src/gen9_mfc_hevc.c
@@ -1146,20 +1146,18 @@ gen9_hcpe_hevc_fill_indirect_cu_intra(VADriverContextP
ctx,
int qp, unsigned int *msg,
int ctb_x, int ctb_y,
int mb_x, int mb_y,
- int ctb_width_in_mb, int width_in_ctb,
int num_cu_record, int slice_type)
+ int ctb_width_in_mb, int width_in_ctb,
int num_cu_record, int slice_type,int cu_index,int index)
{
/* here cu == mb, so we use mb address as the cu address */
/* to fill the indirect cu by the vme out */
- static int mb_addr_raster_to_zigzag_64[4][4] = { {0, 1, 4, 5}, {2, 3, 6,
7}, {8, 9, 12, 13}, {10, 11, 14, 15} };
- static int mb_addr_raster_to_zigzag_32[2][2] = { {0, 1}, {2, 3 } };
static int intra_mode_8x8_avc2hevc[9] = {26, 10, 1, 34, 18, 24, 13, 28, 8};
static int intra_mode_16x16_avc2hevc[4] = {26, 10, 1, 34};
struct gen9_hcpe_context *mfc_context = encoder_context->mfc_context;
unsigned char * cu_record_ptr = NULL;
unsigned int * cu_msg = NULL;
int ctb_address = (ctb_y * width_in_ctb + ctb_x) * num_cu_record;
- int mb_address_in_ctb = ((ctb_width_in_mb == 4) ?
mb_addr_raster_to_zigzag_64[mb_x][mb_y] : ((ctb_width_in_mb == 2) ?
mb_addr_raster_to_zigzag_32[mb_x][mb_y] : 0));
- int cu_address = (ctb_address + mb_address_in_ctb) * 16 * 4;
+ int mb_address_in_ctb = 0;
+ int cu_address = (ctb_address + mb_address_in_ctb + cu_index) * 16 * 4;
int zero = 0;
int is_inter = 0;
int intraMbMode = 0;
@@ -1167,6 +1165,9 @@ gen9_hcpe_hevc_fill_indirect_cu_intra(VADriverContextP
ctx,
int intraMode[4];
int inerpred_idc = 0;
int intra_chroma_mode = 5;
+ int cu_size = 1;
+ int tu_size = 0x55;
+ int tu_count = 4;
if (!is_inter) inerpred_idc = 0xff;
@@ -1176,29 +1177,37 @@ gen9_hcpe_hevc_fill_indirect_cu_intra(VADriverContextP
ctx,
if (intraMbMode == AVC_INTRA_16X16) {
cu_part_mode = 0; //2Nx2N
intra_chroma_mode = 5;
+ cu_size = 1;
+ tu_size = 0x55;
+ tu_count = 4;
intraMode[0] = intra_mode_16x16_avc2hevc[msg[1] & 0xf];
intraMode[1] = intra_mode_16x16_avc2hevc[msg[1] & 0xf];
intraMode[2] = intra_mode_16x16_avc2hevc[msg[1] & 0xf];
intraMode[3] = intra_mode_16x16_avc2hevc[msg[1] & 0xf];
} else if (intraMbMode == AVC_INTRA_8X8) {
- cu_part_mode = 3; //NxN
- intra_chroma_mode = 0;
- intraMode[0] = intra_mode_8x8_avc2hevc[msg[1] & 0xf];
- intraMode[1] = intra_mode_8x8_avc2hevc[(msg[1] >> 4) & 0xf];
- intraMode[2] = intra_mode_8x8_avc2hevc[(msg[1] >> 8) & 0xf];
- intraMode[3] = intra_mode_8x8_avc2hevc[(msg[1] >> 12) & 0xf];
+ cu_part_mode = 0; //2Nx2N
+ intra_chroma_mode = 5;
+ cu_size = 0;
+ tu_size = 0;
+ tu_count = 4;
+ intraMode[0] = intra_mode_8x8_avc2hevc[msg[1] >> (index << 2) & 0xf];
+ intraMode[1] = intra_mode_8x8_avc2hevc[msg[1] >> (index << 2) & 0xf];
+ intraMode[2] = intra_mode_8x8_avc2hevc[msg[1] >> (index << 2) & 0xf];
+ intraMode[3] = intra_mode_8x8_avc2hevc[msg[1] >> (index << 2) & 0xf];
} else { // for 4x4 to use 8x8 replace
cu_part_mode = 3; //NxN
intra_chroma_mode = 0;
- intraMode[0] = intra_mode_8x8_avc2hevc[0];
- intraMode[1] = intra_mode_8x8_avc2hevc[0];
- intraMode[2] = intra_mode_8x8_avc2hevc[0];
- intraMode[3] = intra_mode_8x8_avc2hevc[0];
+ cu_size = 0;
+ tu_size = 0;
+ tu_count = 4;
+ intraMode[0] = intra_mode_8x8_avc2hevc[msg[1] >> ((index << 4) + 0) &
0xf];
+ intraMode[1] = intra_mode_8x8_avc2hevc[msg[1] >> ((index << 4) + 4) &
0xf];
+ intraMode[2] = intra_mode_8x8_avc2hevc[msg[1] >> ((index << 4) + 8) &
0xf];
+ intraMode[3] = intra_mode_8x8_avc2hevc[msg[1] >> ((index << 4) + 12) &
0xf];
}
- dri_bo_map(mfc_context->hcp_indirect_cu_object.bo , 1);
cu_record_ptr = (unsigned char
*)mfc_context->hcp_indirect_cu_object.bo->virtual;
/* get the mb info from the vme out */
cu_msg = (unsigned int *)(cu_record_ptr + cu_address);
@@ -1212,7 +1221,7 @@ gen9_hcpe_hevc_fill_indirect_cu_intra(VADriverContextP
ctx,
cu_part_mode << 4 | /* cu_part_mode */
zero << 3 | /* cu_transquant_bypass_flag */
is_inter << 2 | /* cu_pred_mode :intra 1,inter 1*/
- 1 /* cu_size */
+ cu_size /* cu_size */
);
cu_msg[1] = (zero << 30 | /* reserved */
intraMode[3] << 24 | /* intra_mode */
@@ -1260,8 +1269,8 @@ gen9_hcpe_hevc_fill_indirect_cu_intra(VADriverContextP
ctx,
zero /* ref_idx_l0[0] */
);
- cu_msg[11] = 0x55; /* tu_size 00000000 00000000 00000000 10101010 or 0x0*/
- cu_msg[12] = (3 << 28 | /* tu count - 1 */
+ cu_msg[11] = tu_size; /* tu_size 00000000 00000000 00000000 10101010 or
0x0*/
+ cu_msg[12] = ((tu_count - 1) << 28 | /* tu count - 1 */
zero << 16 | /* reserved */
zero /* tu_xform_Yskip[15:0] */
);
@@ -1270,9 +1279,6 @@ gen9_hcpe_hevc_fill_indirect_cu_intra(VADriverContextP
ctx,
);
cu_msg[14] = zero ;
cu_msg[15] = zero ;
-
- dri_bo_unmap(mfc_context->hcp_indirect_cu_object.bo);
-
}
/* here 1 MB = 1CU = 16x16 */
@@ -1283,24 +1289,24 @@ gen9_hcpe_hevc_fill_indirect_cu_inter(VADriverContextP
ctx,
int qp, unsigned int *msg,
int ctb_x, int ctb_y,
int mb_x, int mb_y,
- int ctb_width_in_mb, int width_in_ctb,
int num_cu_record, int slice_type)
+ int ctb_width_in_mb, int width_in_ctb,
int num_cu_record, int slice_type, int cu_index,int index)
{
/* here cu == mb, so we use mb address as the cu address */
/* to fill the indirect cu by the vme out */
- static int mb_addr_raster_to_zigzag_64[4][4] = { {0, 1, 4, 5}, {2, 3, 6,
7}, {8, 9, 12, 13}, {10, 11, 14, 15} };
- static int mb_addr_raster_to_zigzag_32[2][2] = { {0, 1}, {2, 3 } };
-
struct gen9_hcpe_context *mfc_context = encoder_context->mfc_context;
struct gen6_vme_context *vme_context = encoder_context->vme_context;
unsigned char * cu_record_ptr = NULL;
unsigned int * cu_msg = NULL;
int ctb_address = (ctb_y * width_in_ctb + ctb_x) * num_cu_record;
- int mb_address_in_ctb = ((ctb_width_in_mb == 4) ?
mb_addr_raster_to_zigzag_64[mb_x][mb_y] : ((ctb_width_in_mb == 2) ?
mb_addr_raster_to_zigzag_32[mb_x][mb_y] : 0));
- int cu_address = (ctb_address + mb_address_in_ctb) * 16 * 4;
+ int mb_address_in_ctb = 0;
+ int cu_address = (ctb_address + mb_address_in_ctb + cu_index) * 16 * 4;
int zero = 0;
int cu_part_mode = 0;
int submb_pre_mode = 0;
int is_inter = 1;
+ int cu_size = 1;
+ int tu_size = 0x55;
+ int tu_count = 4;
unsigned int *mv_ptr;
{
@@ -1314,7 +1320,6 @@ gen9_hcpe_hevc_fill_indirect_cu_inter(VADriverContextP
ctx,
/* 0/2/4/6/8... ??? l0, 1/3/5/7...: l1 ; now it only support
16x16,16x8,8x16,8x8*/
if ((msg[0] & AVC_INTER_MODE_MASK) == AVC_INTER_16X16) {
- // MV[0] and MV[2] are replicated
mv_ptr[4] = mv_ptr[0];
mv_ptr[5] = mv_ptr[1];
mv_ptr[2] = mv_ptr[0];
@@ -1322,8 +1327,10 @@ gen9_hcpe_hevc_fill_indirect_cu_inter(VADriverContextP
ctx,
mv_ptr[6] = mv_ptr[0];
mv_ptr[7] = mv_ptr[1];
cu_part_mode = 0;
+ cu_size = 1;
+ tu_size = 0x55;
+ tu_count = 4;
} else if ((msg[0] & AVC_INTER_MODE_MASK) == AVC_INTER_8X16) {
- // MV[0] and MV[2] are replicated
mv_ptr[4] = mv_ptr[0];
mv_ptr[5] = mv_ptr[1];
mv_ptr[2] = mv_ptr[8];
@@ -1331,8 +1338,10 @@ gen9_hcpe_hevc_fill_indirect_cu_inter(VADriverContextP
ctx,
mv_ptr[6] = mv_ptr[8];
mv_ptr[7] = mv_ptr[9];
cu_part_mode = 1;
+ cu_size = 1;
+ tu_size = 0x55;
+ tu_count = 4;
} else if ((msg[0] & AVC_INTER_MODE_MASK) == AVC_INTER_16X8) {
- // MV[0] and MV[1] are replicated
mv_ptr[2] = mv_ptr[0];
mv_ptr[3] = mv_ptr[1];
mv_ptr[4] = mv_ptr[16];
@@ -1340,30 +1349,39 @@ gen9_hcpe_hevc_fill_indirect_cu_inter(VADriverContextP
ctx,
mv_ptr[6] = mv_ptr[24];
mv_ptr[7] = mv_ptr[25];
cu_part_mode = 2;
- } else if (((msg[0] & AVC_INTER_MODE_MASK) == AVC_INTER_8X8) &&
- !(msg[1] & SUBMB_SHAPE_MASK)) {
- // Don't touch MV[0] or MV[1]
- mv_ptr[2] = mv_ptr[8];
- mv_ptr[3] = mv_ptr[9];
- mv_ptr[4] = mv_ptr[16];
- mv_ptr[5] = mv_ptr[17];
- mv_ptr[6] = mv_ptr[24];
- mv_ptr[7] = mv_ptr[25];
- cu_part_mode = 3;
- } else {
- // Don't touch MV[0] or MV[1]
- // default use 8x8
- mv_ptr[2] = mv_ptr[8];
- mv_ptr[3] = mv_ptr[9];
- mv_ptr[4] = mv_ptr[16];
- mv_ptr[5] = mv_ptr[17];
- mv_ptr[6] = mv_ptr[24];
- mv_ptr[7] = mv_ptr[25];
- cu_part_mode = 3;
+ cu_size = 1;
+ tu_size = 0x55;
+ tu_count = 4;
+ }else if((msg[0] & AVC_INTER_MODE_MASK) == AVC_INTER_8X8) {
+ mv_ptr[0] = mv_ptr[index * 8 + 0 ];
+ mv_ptr[1] = mv_ptr[index * 8 + 1 ];
+ mv_ptr[2] = mv_ptr[index * 8 + 0 ];
+ mv_ptr[3] = mv_ptr[index * 8 + 1 ];
+ mv_ptr[4] = mv_ptr[index * 8 + 0 ];
+ mv_ptr[5] = mv_ptr[index * 8 + 1 ];
+ mv_ptr[6] = mv_ptr[index * 8 + 0 ];
+ mv_ptr[7] = mv_ptr[index * 8 + 1 ];
+ cu_part_mode = 0;
+ cu_size = 0;
+ tu_size = 0x0;
+ tu_count = 4;
+
+ }else
+ {
+ mv_ptr[4] = mv_ptr[0];
+ mv_ptr[5] = mv_ptr[1];
+ mv_ptr[2] = mv_ptr[0];
+ mv_ptr[3] = mv_ptr[1];
+ mv_ptr[6] = mv_ptr[0];
+ mv_ptr[7] = mv_ptr[1];
+ cu_part_mode = 0;
+ cu_size = 1;
+ tu_size = 0x55;
+ tu_count = 4;
+
}
}
- dri_bo_map(mfc_context->hcp_indirect_cu_object.bo , 1);
cu_record_ptr = (unsigned char
*)mfc_context->hcp_indirect_cu_object.bo->virtual;
/* get the mb info from the vme out */
cu_msg = (unsigned int *)(cu_record_ptr + cu_address);
@@ -1377,7 +1395,7 @@ gen9_hcpe_hevc_fill_indirect_cu_inter(VADriverContextP
ctx,
cu_part_mode << 4 | /* cu_part_mode */
zero << 3 | /* cu_transquant_bypass_flag */
is_inter << 2 | /* cu_pred_mode :intra 1,inter 1*/
- 1 /* cu_size */
+ cu_size /* cu_size */
);
cu_msg[1] = (zero << 30 | /* reserved */
zero << 24 | /* intra_mode */
@@ -1425,8 +1443,8 @@ gen9_hcpe_hevc_fill_indirect_cu_inter(VADriverContextP
ctx,
((vme_context->ref_index_in_mb[0] >> 0) & 0xf) /*
ref_idx_l0[0] */
);
- cu_msg[11] = 0x55; /* tu_size 00000000 00000000 00000000 10101010 or 0x0*/
- cu_msg[12] = (3 << 28 | /* tu count - 1 */
+ cu_msg[11] = tu_size; /* tu_size 00000000 00000000 00000000 10101010 or
0x0*/
+ cu_msg[12] = ((tu_count - 1) << 28 | /* tu count - 1 */
zero << 16 | /* reserved */
zero /* tu_xform_Yskip[15:0] */
);
@@ -1435,105 +1453,12 @@ gen9_hcpe_hevc_fill_indirect_cu_inter(VADriverContextP
ctx,
);
cu_msg[14] = zero ;
cu_msg[15] = zero ;
-
- dri_bo_unmap(mfc_context->hcp_indirect_cu_object.bo);
-
-}
-
-static void
-gen9_hcpe_hevc_vmeout_to_indirect_cu_buffer(VADriverContextP ctx,
- struct encode_state *encode_state,
- struct intel_encoder_context *encoder_context,
- int slice_index)
-{
- /* to do */
- /* to fill the indirect cu by the vme out */
- struct gen9_hcpe_context *mfc_context = encoder_context->mfc_context;
- struct gen6_vme_context *vme_context = encoder_context->vme_context;
- VAEncSequenceParameterBufferHEVC *pSequenceParameter =
(VAEncSequenceParameterBufferHEVC *)encode_state->seq_param_ext->buffer;
- VAEncPictureParameterBufferHEVC *pPicParameter =
(VAEncPictureParameterBufferHEVC *)encode_state->pic_param_ext->buffer;
- VAEncSliceParameterBufferHEVC *pSliceParameter =
(VAEncSliceParameterBufferHEVC
*)encode_state->slice_params_ext[slice_index]->buffer;
- unsigned int *msg = NULL;
- unsigned char *msg_ptr = NULL;
- int qp = pPicParameter->pic_init_qp + pSliceParameter->slice_qp_delta;
- unsigned int rate_control_mode = encoder_context->rate_control_mode;
-
- int slice_type = pSliceParameter->slice_type;
- int is_intra = slice_type == HEVC_SLICE_I;
-
- int log2_cu_size =
pSequenceParameter->log2_min_luma_coding_block_size_minus3 + 3;
- int log2_ctb_size =
pSequenceParameter->log2_diff_max_min_luma_coding_block_size + log2_cu_size;
- int ctb_size = 1 << log2_ctb_size;
- int ctb_width_in_mb = (ctb_size + 15) / 16;
- int num_mb_in_ctb = ctb_width_in_mb * ctb_width_in_mb;
-
- int width_in_ctb = (pSequenceParameter->pic_width_in_luma_samples +
ctb_size - 1) / ctb_size;
-
- int width_in_mbs = (mfc_context->surface_state.width + 15) / 16;
-
- int num_cu_record = 64;
-
- if (log2_ctb_size == 5) num_cu_record = 16;
- else if (log2_ctb_size == 4) num_cu_record = 4;
- else if (log2_ctb_size == 6) num_cu_record = 64;
-
- int i_ctb;
- int ctb_x, ctb_y;
-
- int macroblock_address = 0;
-
- if (rate_control_mode == VA_RC_CBR) {
- qp = mfc_context->bit_rate_control_context[slice_type].QpPrimeY;
- pSliceParameter->slice_qp_delta = qp - pPicParameter->pic_init_qp;
- }
-
- dri_bo_map(vme_context->vme_output.bo , 1);
- msg_ptr = (unsigned char *)vme_context->vme_output.bo->virtual;
-
- for (i_ctb = pSliceParameter->slice_segment_address; i_ctb <
pSliceParameter->slice_segment_address + pSliceParameter->num_ctu_in_slice;
i_ctb++) {
- ctb_x = i_ctb % width_in_ctb;
- ctb_y = i_ctb / width_in_ctb;
-
- int mb_x, mb_y;
- int mb_addr = 0;
- macroblock_address = (i_ctb - ctb_x) * num_mb_in_ctb + ctb_x *
ctb_width_in_mb;
- for (mb_y = 0; mb_y < ctb_width_in_mb; mb_y++) {
- mb_addr = macroblock_address + mb_y * width_in_mbs ;
- for (mb_x = 0; mb_x < ctb_width_in_mb; mb_x++) {
- mb_addr++;
-
- /* get the mb info from the vme out */
- msg = (unsigned int *)(msg_ptr + mb_addr *
vme_context->vme_output.size_block);
-
- /*fill to indirect cu */
- /*to do */
- if (is_intra) {
- /* fill intra cu */
- gen9_hcpe_hevc_fill_indirect_cu_intra(ctx, encode_state,
encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y, ctb_width_in_mb,
width_in_ctb, num_cu_record, slice_type);
- } 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;
- if (intra_rdo < inter_rdo) {
- /* fill intra cu */
- gen9_hcpe_hevc_fill_indirect_cu_intra(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type);
- } else {
- msg += AVC_INTER_MSG_OFFSET;
- /* fill inter cu */
- gen9_hcpe_hevc_fill_indirect_cu_inter(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type);
- }
- }
-
- }
- }
- }
-
- dri_bo_unmap(vme_context->vme_output.bo);
}
#define HEVC_SPLIT_CU_FLAG_64_64
((0x1<<20)|(0xf<<16)|(0x0<<12)|(0x0<<8)|(0x0<<4)|(0x0))
#define HEVC_SPLIT_CU_FLAG_32_32
((0x1<<20)|(0x0<<16)|(0x0<<12)|(0x0<<8)|(0x0<<4)|(0x0))
#define HEVC_SPLIT_CU_FLAG_16_16
((0x0<<20)|(0x0<<16)|(0x0<<12)|(0x0<<8)|(0x0<<4)|(0x0))
+#define HEVC_SPLIT_CU_FLAG_8_8
((0x1<<20)|(0x0<<16)|(0x0<<12)|(0x0<<8)|(0x0<<4)|(0x0))
void
@@ -1646,6 +1571,7 @@ gen9_hcpe_hevc_pipeline_slice_programing(VADriverContextP
ctx,
struct intel_batchbuffer *slice_batch)
{
struct gen9_hcpe_context *mfc_context = encoder_context->mfc_context;
+ struct gen6_vme_context *vme_context = encoder_context->vme_context;
VAEncSequenceParameterBufferHEVC *pSequenceParameter =
(VAEncSequenceParameterBufferHEVC *)encode_state->seq_param_ext->buffer;
VAEncPictureParameterBufferHEVC *pPicParameter =
(VAEncPictureParameterBufferHEVC *)encode_state->pic_param_ext->buffer;
VAEncSliceParameterBufferHEVC *pSliceParameter =
(VAEncSliceParameterBufferHEVC
*)encode_state->slice_params_ext[slice_index]->buffer;
@@ -1666,7 +1592,24 @@
gen9_hcpe_hevc_pipeline_slice_programing(VADriverContextP ctx,
int num_mb_in_ctb = ctb_width_in_mb * ctb_width_in_mb;
int i_ctb, ctb_x, ctb_y;
unsigned int split_coding_unit_flag = 0;
+ int width_in_mbs = (mfc_context->surface_state.width + 15) / 16;
+ int row_pad_flag = (pSequenceParameter->pic_height_in_luma_samples %
ctb_size)> 0 ? 1:0;
+
+ int is_intra = (slice_type == HEVC_SLICE_I);
+ unsigned int *msg = NULL;
+ unsigned char *msg_ptr = NULL;
+ int macroblock_address = 0;
+ int num_cu_record = 64;
+ int cu_count = 1;
+ int tmp_mb_mode = 0;
+ int mb_x = 0, mb_y = 0;
+ int mb_addr = 0;
+ int cu_index = 0;
+ int inter_rdo, intra_rdo;
+ if (log2_ctb_size == 5) num_cu_record = 16;
+ else if (log2_ctb_size == 4) num_cu_record = 4;
+ else if (log2_ctb_size == 6) num_cu_record = 64;
if (rate_control_mode == VA_RC_CBR) {
qp = mfc_context->bit_rate_control_context[slice_type].QpPrimeY;
pSliceParameter->slice_qp_delta = qp - pPicParameter->pic_init_qp;
@@ -1706,16 +1649,85 @@
gen9_hcpe_hevc_pipeline_slice_programing(VADriverContextP ctx,
split_coding_unit_flag = (ctb_width_in_mb == 4) ? HEVC_SPLIT_CU_FLAG_64_64
: ((ctb_width_in_mb == 2) ? HEVC_SPLIT_CU_FLAG_32_32 :
HEVC_SPLIT_CU_FLAG_16_16);
- for (i_ctb = pSliceParameter->slice_segment_address;
- i_ctb < pSliceParameter->slice_segment_address +
pSliceParameter->num_ctu_in_slice; i_ctb++) {
+ dri_bo_map(vme_context->vme_output.bo , 1);
+ msg_ptr = (unsigned char *)vme_context->vme_output.bo->virtual;
+ dri_bo_map(mfc_context->hcp_indirect_cu_object.bo , 1);
+
+ for (i_ctb = pSliceParameter->slice_segment_address;i_ctb <
pSliceParameter->slice_segment_address + pSliceParameter->num_ctu_in_slice;
i_ctb++) {
int last_ctb = (i_ctb == (pSliceParameter->slice_segment_address +
pSliceParameter->num_ctu_in_slice - 1));
+ int ctb_height_in_mb = ctb_width_in_mb;
ctb_x = i_ctb % width_in_ctb;
ctb_y = i_ctb / width_in_ctb;
+ if(ctb_y == (height_in_ctb - 1) && row_pad_flag) ctb_height_in_mb = 1;
+
+ mb_x = 0;
+ mb_y = 0;
+ macroblock_address = (i_ctb - ctb_x) * num_mb_in_ctb + ctb_x *
ctb_width_in_mb;
+ split_coding_unit_flag = ((ctb_width_in_mb == 2) ?
HEVC_SPLIT_CU_FLAG_32_32 : HEVC_SPLIT_CU_FLAG_16_16);
+ cu_count = 1;
+ cu_index = 0;
+ mb_addr = 0;
+ msg = NULL;
+ for (mb_y = 0; mb_y < ctb_height_in_mb; mb_y++)
+ {
+ mb_addr = macroblock_address + mb_y * width_in_mbs ;
+ for (mb_x = 0; mb_x < ctb_width_in_mb; mb_x++)
+ {
+ split_coding_unit_flag = ((ctb_width_in_mb == 2) ?
HEVC_SPLIT_CU_FLAG_32_32 : HEVC_SPLIT_CU_FLAG_16_16);
+ /* get the mb info from the vme out */
+ msg = (unsigned int *)(msg_ptr + mb_addr *
vme_context->vme_output.size_block);
- gen9_hcpe_hevc_pak_object(ctx, ctb_x, ctb_y, last_ctb,
encoder_context, num_mb_in_ctb, split_coding_unit_flag, slice_batch);
+ inter_rdo = msg[AVC_INTER_RDO_OFFSET] & AVC_RDO_MASK;
+ intra_rdo = msg[AVC_INTRA_RDO_OFFSET] & AVC_RDO_MASK;
+ /*fill to indirect cu */
+ /*to do */
+ if (is_intra || intra_rdo < inter_rdo) {
+ /* fill intra cu */
+ tmp_mb_mode = (msg[0] & AVC_INTRA_MODE_MASK) >> 4;
+ if (tmp_mb_mode == AVC_INTRA_16X16) {
+ gen9_hcpe_hevc_fill_indirect_cu_intra(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,0);
+ } else { // for 4x4 to use 8x8 replace
+ gen9_hcpe_hevc_fill_indirect_cu_intra(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,0);
+ gen9_hcpe_hevc_fill_indirect_cu_intra(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,1);
+ gen9_hcpe_hevc_fill_indirect_cu_intra(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,2);
+ gen9_hcpe_hevc_fill_indirect_cu_intra(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,3);
+ if(ctb_width_in_mb == 2)
+ split_coding_unit_flag |= 0x1 << (mb_x + mb_y *
ctb_width_in_mb + 16);
+ else if(ctb_width_in_mb == 1)
+ split_coding_unit_flag |= 0x1 << 20;
+ }
+ } else {
+ msg += AVC_INTER_MSG_OFFSET;
+ /* fill inter cu */
+ tmp_mb_mode = msg[0] & AVC_INTER_MODE_MASK;
+ if (tmp_mb_mode == AVC_INTER_8X8){
+ gen9_hcpe_hevc_fill_indirect_cu_inter(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,0);
+ gen9_hcpe_hevc_fill_indirect_cu_inter(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,1);
+ gen9_hcpe_hevc_fill_indirect_cu_inter(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,2);
+ gen9_hcpe_hevc_fill_indirect_cu_inter(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,3);
+ if(ctb_width_in_mb == 2)
+ split_coding_unit_flag |= 0x1 << (mb_x + mb_y *
ctb_width_in_mb + 16);
+ else if(ctb_width_in_mb == 1)
+ split_coding_unit_flag |= 0x1 << 20;
+
+ }else if(tmp_mb_mode == AVC_INTER_16X16 ||
+ tmp_mb_mode == AVC_INTER_8X16 ||
+ tmp_mb_mode == AVC_INTER_16X8) {
+ gen9_hcpe_hevc_fill_indirect_cu_inter(ctx,
encode_state, encoder_context, qp, msg, ctb_x, ctb_y, mb_x, mb_y,
ctb_width_in_mb, width_in_ctb, num_cu_record, slice_type,cu_index++,0);
+ }
+ }
+ mb_addr++;
+ }
+ }
+ cu_count = cu_index;
+ // PAK object fill accordingly.
+ gen9_hcpe_hevc_pak_object(ctx, ctb_x, ctb_y, last_ctb,
encoder_context, cu_count, split_coding_unit_flag, slice_batch);
}
+ dri_bo_unmap(mfc_context->hcp_indirect_cu_object.bo);
+ dri_bo_unmap(vme_context->vme_output.bo);
+
if (last_slice) {
mfc_context->insert_object(ctx, encoder_context,
tail_data, 2, 8,
@@ -1741,7 +1753,6 @@ gen9_hcpe_hevc_software_batchbuffer(VADriverContextP ctx,
batch_bo = batch->buffer;
for (i = 0; i < encode_state->num_slice_params_ext; i++) {
- gen9_hcpe_hevc_vmeout_to_indirect_cu_buffer(ctx, encode_state,
encoder_context, i);
gen9_hcpe_hevc_pipeline_slice_programing(ctx, encode_state,
encoder_context, i, batch);
}
--
1.9.1
_______________________________________________
Libva mailing list
Libva@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/libva
