falkTX
/
FFmpeg
mirror of https://github.com/falkTX/FFmpeg.git
1
0
Fork 0
Code Issues 0 Releases 338 Wiki Activity
Browse Source

h264: move mb[_{padding,luma_dc}] into the per-slice context

tags/n2.7
Anton Khirnov 10 years ago
parent
b063582e0c
commit
bf03a878a7
7 changed files with 78 additions and 74 deletions
Unified View
Diff Options
Show Stats Download Patch File Download Diff File
  1. +7
    -4
      libavcodec/h264.h
  2. +12
    -12
      libavcodec/h264_cabac.c
  3. +12
    -12
      libavcodec/h264_cavlc.c
  4. +24
    -24
      libavcodec/h264_mb.c
  5. +13
    -13
      libavcodec/h264_mb_template.c
  6. +4
    -3
      libavcodec/h264_slice.c
  7. +6
    -6
      libavcodec/svq3.c

+ 7
- 4
libavcodec/h264.h View File

@@ -389,6 +389,13 @@ typedef struct H264SliceContext {
DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8]; DECLARE_ALIGNED(8, int8_t, ref_cache)[2][5 * 8];


DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4]; DECLARE_ALIGNED(8, uint16_t, sub_mb_type)[4];

///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2];
DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
///< as mb is addressed by scantable[i] and scantable is uint8_t we can either
///< check that i is not too large or ensure that there is some unused stuff after mb
int16_t mb_padding[256 * 2];
} H264SliceContext; } H264SliceContext;


/** /**
@@ -469,10 +476,6 @@ typedef struct H264Context {
GetBitContext *intra_gb_ptr; GetBitContext *intra_gb_ptr;
GetBitContext *inter_gb_ptr; GetBitContext *inter_gb_ptr;


DECLARE_ALIGNED(16, int16_t, mb)[16 * 48 * 2]; ///< as a dct coeffecient is int32_t in high depth, we need to reserve twice the space.
DECLARE_ALIGNED(16, int16_t, mb_luma_dc)[3][16 * 2];
int16_t mb_padding[256 * 2]; ///< as mb is addressed by scantable[i] and scantable is uint8_t we can either check that i is not too large or ensure that there is some unused stuff after mb

/** /**
* Cabac * Cabac
*/ */


+ 12
- 12
libavcodec/h264_cabac.c View File

@@ -1857,17 +1857,17 @@ static av_always_inline void decode_cabac_luma_residual(H264Context *h, H264Slic
int i8x8, i4x4; int i8x8, i4x4;
int qscale = p == 0 ? sl->qscale : sl->chroma_qp[p - 1]; int qscale = p == 0 ? sl->qscale : sl->chroma_qp[p - 1];
if( IS_INTRA16x16( mb_type ) ) { if( IS_INTRA16x16( mb_type ) ) {
AV_ZERO128(h->mb_luma_dc[p]+0);
AV_ZERO128(h->mb_luma_dc[p]+8);
AV_ZERO128(h->mb_luma_dc[p]+16);
AV_ZERO128(h->mb_luma_dc[p]+24);
decode_cabac_residual_dc(h, sl, h->mb_luma_dc[p], ctx_cat[0][p], LUMA_DC_BLOCK_INDEX+p, scan, 16);
AV_ZERO128(sl->mb_luma_dc[p]+0);
AV_ZERO128(sl->mb_luma_dc[p]+8);
AV_ZERO128(sl->mb_luma_dc[p]+16);
AV_ZERO128(sl->mb_luma_dc[p]+24);
decode_cabac_residual_dc(h, sl, sl->mb_luma_dc[p], ctx_cat[0][p], LUMA_DC_BLOCK_INDEX+p, scan, 16);


if( cbp&15 ) { if( cbp&15 ) {
qmul = h->dequant4_coeff[p][qscale]; qmul = h->dequant4_coeff[p][qscale];
for( i4x4 = 0; i4x4 < 16; i4x4++ ) { for( i4x4 = 0; i4x4 < 16; i4x4++ ) {
const int index = 16*p + i4x4; const int index = 16*p + i4x4;
decode_cabac_residual_nondc(h, sl, h->mb + (16*index << pixel_shift), ctx_cat[1][p], index, scan + 1, qmul, 15);
decode_cabac_residual_nondc(h, sl, sl->mb + (16*index << pixel_shift), ctx_cat[1][p], index, scan + 1, qmul, 15);
} }
} else { } else {
fill_rectangle(&sl->non_zero_count_cache[scan8[16*p]], 4, 4, 8, 0, 1); fill_rectangle(&sl->non_zero_count_cache[scan8[16*p]], 4, 4, 8, 0, 1);
@@ -1878,14 +1878,14 @@ static av_always_inline void decode_cabac_luma_residual(H264Context *h, H264Slic
if( cbp & (1<<i8x8) ) { if( cbp & (1<<i8x8) ) {
if( IS_8x8DCT(mb_type) ) { if( IS_8x8DCT(mb_type) ) {
const int index = 16*p + 4*i8x8; const int index = 16*p + 4*i8x8;
decode_cabac_residual_nondc(h, sl, h->mb + (16*index << pixel_shift), ctx_cat[3][p], index,
decode_cabac_residual_nondc(h, sl, sl->mb + (16*index << pixel_shift), ctx_cat[3][p], index,
scan8x8, h->dequant8_coeff[cqm][qscale], 64); scan8x8, h->dequant8_coeff[cqm][qscale], 64);
} else { } else {
qmul = h->dequant4_coeff[cqm][qscale]; qmul = h->dequant4_coeff[cqm][qscale];
for( i4x4 = 0; i4x4 < 4; i4x4++ ) { for( i4x4 = 0; i4x4 < 4; i4x4++ ) {
const int index = 16*p + 4*i8x8 + i4x4; const int index = 16*p + 4*i8x8 + i4x4;
//START_TIMER //START_TIMER
decode_cabac_residual_nondc(h, sl, h->mb + (16*index << pixel_shift), ctx_cat[2][p], index, scan, qmul, 16);
decode_cabac_residual_nondc(h, sl, sl->mb + (16*index << pixel_shift), ctx_cat[2][p], index, scan, qmul, 16);
//STOP_TIMER("decode_residual") //STOP_TIMER("decode_residual")
} }
} }
@@ -2389,7 +2389,7 @@ decode_intra_mb:
if( cbp&0x30 ){ if( cbp&0x30 ){
int c; int c;
for (c = 0; c < 2; c++) for (c = 0; c < 2; c++)
decode_cabac_residual_dc_422(h, sl, h->mb + ((256 + 16*16*c) << pixel_shift), 3,
decode_cabac_residual_dc_422(h, sl, sl->mb + ((256 + 16*16*c) << pixel_shift), 3,
CHROMA_DC_BLOCK_INDEX + c, CHROMA_DC_BLOCK_INDEX + c,
chroma422_dc_scan, 8); chroma422_dc_scan, 8);
} }
@@ -2397,7 +2397,7 @@ decode_intra_mb:
if( cbp&0x20 ) { if( cbp&0x20 ) {
int c, i, i8x8; int c, i, i8x8;
for( c = 0; c < 2; c++ ) { for( c = 0; c < 2; c++ ) {
int16_t *mb = h->mb + (16*(16 + 16*c) << pixel_shift);
int16_t *mb = sl->mb + (16*(16 + 16*c) << pixel_shift);
qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[c]]; qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[c]];
for (i8x8 = 0; i8x8 < 2; i8x8++) { for (i8x8 = 0; i8x8 < 2; i8x8++) {
for (i = 0; i < 4; i++) { for (i = 0; i < 4; i++) {
@@ -2415,7 +2415,7 @@ decode_intra_mb:
if( cbp&0x30 ){ if( cbp&0x30 ){
int c; int c;
for (c = 0; c < 2; c++) for (c = 0; c < 2; c++)
decode_cabac_residual_dc(h, sl, h->mb + ((256 + 16*16*c) << pixel_shift), 3, CHROMA_DC_BLOCK_INDEX+c, chroma_dc_scan, 4);
decode_cabac_residual_dc(h, sl, sl->mb + ((256 + 16*16*c) << pixel_shift), 3, CHROMA_DC_BLOCK_INDEX+c, chroma_dc_scan, 4);
} }


if( cbp&0x20 ) { if( cbp&0x20 ) {
@@ -2424,7 +2424,7 @@ decode_intra_mb:
qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[c]]; qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[c]];
for( i = 0; i < 4; i++ ) { for( i = 0; i < 4; i++ ) {
const int index = 16 + 16 * c + i; const int index = 16 + 16 * c + i;
decode_cabac_residual_nondc(h, sl, h->mb + (16*index << pixel_shift), 4, index, scan + 1, qmul, 15);
decode_cabac_residual_nondc(h, sl, sl->mb + (16*index << pixel_shift), 4, index, scan + 1, qmul, 15);
} }
} }
} else { } else {


+ 12
- 12
libavcodec/h264_cavlc.c View File

@@ -632,11 +632,11 @@ static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContex
int i4x4, i8x8; int i4x4, i8x8;
int qscale = p == 0 ? sl->qscale : sl->chroma_qp[p - 1]; int qscale = p == 0 ? sl->qscale : sl->chroma_qp[p - 1];
if(IS_INTRA16x16(mb_type)){ if(IS_INTRA16x16(mb_type)){
AV_ZERO128(h->mb_luma_dc[p]+0);
AV_ZERO128(h->mb_luma_dc[p]+8);
AV_ZERO128(h->mb_luma_dc[p]+16);
AV_ZERO128(h->mb_luma_dc[p]+24);
if( decode_residual(h, sl, h->intra_gb_ptr, h->mb_luma_dc[p], LUMA_DC_BLOCK_INDEX+p, scan, NULL, 16) < 0){
AV_ZERO128(sl->mb_luma_dc[p]+0);
AV_ZERO128(sl->mb_luma_dc[p]+8);
AV_ZERO128(sl->mb_luma_dc[p]+16);
AV_ZERO128(sl->mb_luma_dc[p]+24);
if( decode_residual(h, sl, h->intra_gb_ptr, sl->mb_luma_dc[p], LUMA_DC_BLOCK_INDEX+p, scan, NULL, 16) < 0){
return -1; //FIXME continue if partitioned and other return -1 too return -1; //FIXME continue if partitioned and other return -1 too
} }


@@ -646,7 +646,7 @@ static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContex
for(i8x8=0; i8x8<4; i8x8++){ for(i8x8=0; i8x8<4; i8x8++){
for(i4x4=0; i4x4<4; i4x4++){ for(i4x4=0; i4x4<4; i4x4++){
const int index= i4x4 + 4*i8x8 + p*16; const int index= i4x4 + 4*i8x8 + p*16;
if( decode_residual(h, sl, h->intra_gb_ptr, h->mb + (16*index << pixel_shift),
if( decode_residual(h, sl, h->intra_gb_ptr, sl->mb + (16*index << pixel_shift),
index, scan + 1, h->dequant4_coeff[p][qscale], 15) < 0 ){ index, scan + 1, h->dequant4_coeff[p][qscale], 15) < 0 ){
return -1; return -1;
} }
@@ -664,7 +664,7 @@ static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContex
for(i8x8=0; i8x8<4; i8x8++){ for(i8x8=0; i8x8<4; i8x8++){
if(cbp & (1<<i8x8)){ if(cbp & (1<<i8x8)){
if(IS_8x8DCT(mb_type)){ if(IS_8x8DCT(mb_type)){
int16_t *buf = &h->mb[64*i8x8+256*p << pixel_shift];
int16_t *buf = &sl->mb[64*i8x8+256*p << pixel_shift];
uint8_t *nnz; uint8_t *nnz;
for(i4x4=0; i4x4<4; i4x4++){ for(i4x4=0; i4x4<4; i4x4++){
const int index= i4x4 + 4*i8x8 + p*16; const int index= i4x4 + 4*i8x8 + p*16;
@@ -678,7 +678,7 @@ static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContex
}else{ }else{
for(i4x4=0; i4x4<4; i4x4++){ for(i4x4=0; i4x4<4; i4x4++){
const int index= i4x4 + 4*i8x8 + p*16; const int index= i4x4 + 4*i8x8 + p*16;
if( decode_residual(h, sl, gb, h->mb + (16*index << pixel_shift), index,
if( decode_residual(h, sl, gb, sl->mb + (16*index << pixel_shift), index,
scan, h->dequant4_coeff[cqm][qscale], 16) < 0 ){ scan, h->dequant4_coeff[cqm][qscale], 16) < 0 ){
return -1; return -1;
} }
@@ -1125,7 +1125,7 @@ decode_intra_mb:
} else if (CHROMA422(h)) { } else if (CHROMA422(h)) {
if(cbp&0x30){ if(cbp&0x30){
for(chroma_idx=0; chroma_idx<2; chroma_idx++) for(chroma_idx=0; chroma_idx<2; chroma_idx++)
if (decode_residual(h, sl, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift),
if (decode_residual(h, sl, gb, sl->mb + ((256 + 16*16*chroma_idx) << pixel_shift),
CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma422_dc_scan, CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma422_dc_scan,
NULL, 8) < 0) { NULL, 8) < 0) {
return -1; return -1;
@@ -1135,7 +1135,7 @@ decode_intra_mb:
if(cbp&0x20){ if(cbp&0x20){
for(chroma_idx=0; chroma_idx<2; chroma_idx++){ for(chroma_idx=0; chroma_idx<2; chroma_idx++){
const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[chroma_idx]]; const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[chroma_idx]];
int16_t *mb = h->mb + (16*(16 + 16*chroma_idx) << pixel_shift);
int16_t *mb = sl->mb + (16*(16 + 16*chroma_idx) << pixel_shift);
for (i8x8 = 0; i8x8 < 2; i8x8++) { for (i8x8 = 0; i8x8 < 2; i8x8++) {
for (i4x4 = 0; i4x4 < 4; i4x4++) { for (i4x4 = 0; i4x4 < 4; i4x4++) {
const int index = 16 + 16*chroma_idx + 8*i8x8 + i4x4; const int index = 16 + 16*chroma_idx + 8*i8x8 + i4x4;
@@ -1152,7 +1152,7 @@ decode_intra_mb:
} else /* yuv420 */ { } else /* yuv420 */ {
if(cbp&0x30){ if(cbp&0x30){
for(chroma_idx=0; chroma_idx<2; chroma_idx++) for(chroma_idx=0; chroma_idx<2; chroma_idx++)
if( decode_residual(h, sl, gb, h->mb + ((256 + 16*16*chroma_idx) << pixel_shift), CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma_dc_scan, NULL, 4) < 0){
if( decode_residual(h, sl, gb, sl->mb + ((256 + 16*16*chroma_idx) << pixel_shift), CHROMA_DC_BLOCK_INDEX+chroma_idx, chroma_dc_scan, NULL, 4) < 0){
return -1; return -1;
} }
} }
@@ -1162,7 +1162,7 @@ decode_intra_mb:
const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[chroma_idx]]; const uint32_t *qmul = h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][sl->chroma_qp[chroma_idx]];
for(i4x4=0; i4x4<4; i4x4++){ for(i4x4=0; i4x4<4; i4x4++){
const int index= 16 + 16*chroma_idx + i4x4; const int index= 16 + 16*chroma_idx + i4x4;
if( decode_residual(h, sl, gb, h->mb + (16*index << pixel_shift), index, scan + 1, qmul, 15) < 0){
if( decode_residual(h, sl, gb, sl->mb + (16*index << pixel_shift), index, scan + 1, qmul, 15) < 0){
return -1; return -1;
} }
} }


+ 24
- 24
libavcodec/h264_mb.c View File

@@ -638,16 +638,16 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h,
uint8_t *const ptr = dest_y + block_offset[i]; uint8_t *const ptr = dest_y + block_offset[i];
const int dir = sl->intra4x4_pred_mode_cache[scan8[i]]; const int dir = sl->intra4x4_pred_mode_cache[scan8[i]];
if (transform_bypass && h->sps.profile_idc == 244 && dir <= 1) { if (transform_bypass && h->sps.profile_idc == 244 && dir <= 1) {
h->hpc.pred8x8l_add[dir](ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize);
h->hpc.pred8x8l_add[dir](ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
} else { } else {
const int nnz = sl->non_zero_count_cache[scan8[i + p * 16]]; const int nnz = sl->non_zero_count_cache[scan8[i + p * 16]];
h->hpc.pred8x8l[dir](ptr, (sl->topleft_samples_available << i) & 0x8000, h->hpc.pred8x8l[dir](ptr, (sl->topleft_samples_available << i) & 0x8000,
(sl->topright_samples_available << i) & 0x4000, linesize); (sl->topright_samples_available << i) & 0x4000, linesize);
if (nnz) { if (nnz) {
if (nnz == 1 && dctcoef_get(h->mb, pixel_shift, i * 16 + p * 256))
idct_dc_add(ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize);
if (nnz == 1 && dctcoef_get(sl->mb, pixel_shift, i * 16 + p * 256))
idct_dc_add(ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
else else
idct_add(ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize);
idct_add(ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
} }
} }
} }
@@ -664,7 +664,7 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h,
const int dir = sl->intra4x4_pred_mode_cache[scan8[i]]; const int dir = sl->intra4x4_pred_mode_cache[scan8[i]];


if (transform_bypass && h->sps.profile_idc == 244 && dir <= 1) { if (transform_bypass && h->sps.profile_idc == 244 && dir <= 1) {
h->hpc.pred4x4_add[dir](ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize);
h->hpc.pred4x4_add[dir](ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
} else { } else {
uint8_t *topright; uint8_t *topright;
int nnz, tr; int nnz, tr;
@@ -689,12 +689,12 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h,
nnz = sl->non_zero_count_cache[scan8[i + p * 16]]; nnz = sl->non_zero_count_cache[scan8[i + p * 16]];
if (nnz) { if (nnz) {
if (is_h264) { if (is_h264) {
if (nnz == 1 && dctcoef_get(h->mb, pixel_shift, i * 16 + p * 256))
idct_dc_add(ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize);
if (nnz == 1 && dctcoef_get(sl->mb, pixel_shift, i * 16 + p * 256))
idct_dc_add(ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
else else
idct_add(ptr, h->mb + (i * 16 + p * 256 << pixel_shift), linesize);
idct_add(ptr, sl->mb + (i * 16 + p * 256 << pixel_shift), linesize);
} else if (CONFIG_SVQ3_DECODER) } else if (CONFIG_SVQ3_DECODER)
ff_svq3_add_idct_c(ptr, h->mb + i * 16 + p * 256, linesize, qscale, 0);
ff_svq3_add_idct_c(ptr, sl->mb + i * 16 + p * 256, linesize, qscale, 0);
} }
} }
} }
@@ -704,8 +704,8 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h,
if (is_h264) { if (is_h264) {
if (sl->non_zero_count_cache[scan8[LUMA_DC_BLOCK_INDEX + p]]) { if (sl->non_zero_count_cache[scan8[LUMA_DC_BLOCK_INDEX + p]]) {
if (!transform_bypass) if (!transform_bypass)
h->h264dsp.h264_luma_dc_dequant_idct(h->mb + (p * 256 << pixel_shift),
h->mb_luma_dc[p],
h->h264dsp.h264_luma_dc_dequant_idct(sl->mb + (p * 256 << pixel_shift),
sl->mb_luma_dc[p],
h->dequant4_coeff[p][qscale][0]); h->dequant4_coeff[p][qscale][0]);
else { else {
static const uint8_t dc_mapping[16] = { static const uint8_t dc_mapping[16] = {
@@ -715,15 +715,15 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h,
10 * 16, 11 * 16, 14 * 16, 15 * 16 10 * 16, 11 * 16, 14 * 16, 15 * 16
}; };
for (i = 0; i < 16; i++) for (i = 0; i < 16; i++)
dctcoef_set(h->mb + (p * 256 << pixel_shift),
dctcoef_set(sl->mb + (p * 256 << pixel_shift),
pixel_shift, dc_mapping[i], pixel_shift, dc_mapping[i],
dctcoef_get(h->mb_luma_dc[p],
dctcoef_get(sl->mb_luma_dc[p],
pixel_shift, i)); pixel_shift, i));
} }
} }
} else if (CONFIG_SVQ3_DECODER) } else if (CONFIG_SVQ3_DECODER)
ff_svq3_luma_dc_dequant_idct_c(h->mb + p * 256,
h->mb_luma_dc[p], qscale);
ff_svq3_luma_dc_dequant_idct_c(sl->mb + p * 256,
sl->mb_luma_dc[p], qscale);
} }
} }


@@ -747,19 +747,19 @@ static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, H264SliceCon
(sl->intra16x16_pred_mode == VERT_PRED8x8 || (sl->intra16x16_pred_mode == VERT_PRED8x8 ||
sl->intra16x16_pred_mode == HOR_PRED8x8)) { sl->intra16x16_pred_mode == HOR_PRED8x8)) {
h->hpc.pred16x16_add[sl->intra16x16_pred_mode](dest_y, block_offset, h->hpc.pred16x16_add[sl->intra16x16_pred_mode](dest_y, block_offset,
h->mb + (p * 256 << pixel_shift),
sl->mb + (p * 256 << pixel_shift),
linesize); linesize);
} else { } else {
for (i = 0; i < 16; i++) for (i = 0; i < 16; i++)
if (sl->non_zero_count_cache[scan8[i + p * 16]] || if (sl->non_zero_count_cache[scan8[i + p * 16]] ||
dctcoef_get(h->mb, pixel_shift, i * 16 + p * 256))
dctcoef_get(sl->mb, pixel_shift, i * 16 + p * 256))
h->h264dsp.h264_add_pixels4_clear(dest_y + block_offset[i], h->h264dsp.h264_add_pixels4_clear(dest_y + block_offset[i],
h->mb + (i * 16 + p * 256 << pixel_shift),
sl->mb + (i * 16 + p * 256 << pixel_shift),
linesize); linesize);
} }
} else { } else {
h->h264dsp.h264_idct_add16intra(dest_y, block_offset, h->h264dsp.h264_idct_add16intra(dest_y, block_offset,
h->mb + (p * 256 << pixel_shift),
sl->mb + (p * 256 << pixel_shift),
linesize, linesize,
sl->non_zero_count_cache + p * 5 * 8); sl->non_zero_count_cache + p * 5 * 8);
} }
@@ -771,27 +771,27 @@ static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, H264SliceCon
for (i = 0; i < 16; i += di) for (i = 0; i < 16; i += di)
if (sl->non_zero_count_cache[scan8[i + p * 16]]) if (sl->non_zero_count_cache[scan8[i + p * 16]])
idct_add(dest_y + block_offset[i], idct_add(dest_y + block_offset[i],
h->mb + (i * 16 + p * 256 << pixel_shift),
sl->mb + (i * 16 + p * 256 << pixel_shift),
linesize); linesize);
} else { } else {
if (IS_8x8DCT(mb_type)) if (IS_8x8DCT(mb_type))
h->h264dsp.h264_idct8_add4(dest_y, block_offset, h->h264dsp.h264_idct8_add4(dest_y, block_offset,
h->mb + (p * 256 << pixel_shift),
sl->mb + (p * 256 << pixel_shift),
linesize, linesize,
sl->non_zero_count_cache + p * 5 * 8); sl->non_zero_count_cache + p * 5 * 8);
else else
h->h264dsp.h264_idct_add16(dest_y, block_offset, h->h264dsp.h264_idct_add16(dest_y, block_offset,
h->mb + (p * 256 << pixel_shift),
sl->mb + (p * 256 << pixel_shift),
linesize, linesize,
sl->non_zero_count_cache + p * 5 * 8); sl->non_zero_count_cache + p * 5 * 8);
} }
} }
} else if (CONFIG_SVQ3_DECODER) { } else if (CONFIG_SVQ3_DECODER) {
for (i = 0; i < 16; i++) for (i = 0; i < 16; i++)
if (sl->non_zero_count_cache[scan8[i + p * 16]] || h->mb[i * 16 + p * 256]) {
if (sl->non_zero_count_cache[scan8[i + p * 16]] || sl->mb[i * 16 + p * 256]) {
// FIXME benchmark weird rule, & below // FIXME benchmark weird rule, & below
uint8_t *const ptr = dest_y + block_offset[i]; uint8_t *const ptr = dest_y + block_offset[i];
ff_svq3_add_idct_c(ptr, h->mb + i * 16 + p * 256, linesize,
ff_svq3_add_idct_c(ptr, sl->mb + i * 16 + p * 256, linesize,
sl->qscale, IS_INTRA(mb_type) ? 1 : 0); sl->qscale, IS_INTRA(mb_type) ? 1 : 0);
} }
} }


+ 13
- 13
libavcodec/h264_mb_template.c View File

@@ -202,27 +202,27 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h, H264SliceContext *sl)
sl->chroma_pred_mode == HOR_PRED8x8)) { sl->chroma_pred_mode == HOR_PRED8x8)) {
h->hpc.pred8x8_add[sl->chroma_pred_mode](dest[0], h->hpc.pred8x8_add[sl->chroma_pred_mode](dest[0],
block_offset + 16, block_offset + 16,
h->mb + (16 * 16 * 1 << PIXEL_SHIFT),
sl->mb + (16 * 16 * 1 << PIXEL_SHIFT),
uvlinesize); uvlinesize);
h->hpc.pred8x8_add[sl->chroma_pred_mode](dest[1], h->hpc.pred8x8_add[sl->chroma_pred_mode](dest[1],
block_offset + 32, block_offset + 32,
h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
sl->mb + (16 * 16 * 2 << PIXEL_SHIFT),
uvlinesize); uvlinesize);
} else { } else {
idct_add = h->h264dsp.h264_add_pixels4_clear; idct_add = h->h264dsp.h264_add_pixels4_clear;
for (j = 1; j < 3; j++) { for (j = 1; j < 3; j++) {
for (i = j * 16; i < j * 16 + 4; i++) for (i = j * 16; i < j * 16 + 4; i++)
if (sl->non_zero_count_cache[scan8[i]] || if (sl->non_zero_count_cache[scan8[i]] ||
dctcoef_get(h->mb, PIXEL_SHIFT, i * 16))
dctcoef_get(sl->mb, PIXEL_SHIFT, i * 16))
idct_add(dest[j - 1] + block_offset[i], idct_add(dest[j - 1] + block_offset[i],
h->mb + (i * 16 << PIXEL_SHIFT),
sl->mb + (i * 16 << PIXEL_SHIFT),
uvlinesize); uvlinesize);
if (chroma422) { if (chroma422) {
for (i = j * 16 + 4; i < j * 16 + 8; i++) for (i = j * 16 + 4; i < j * 16 + 8; i++)
if (sl->non_zero_count_cache[scan8[i + 4]] || if (sl->non_zero_count_cache[scan8[i + 4]] ||
dctcoef_get(h->mb, PIXEL_SHIFT, i * 16))
dctcoef_get(sl->mb, PIXEL_SHIFT, i * 16))
idct_add(dest[j - 1] + block_offset[i + 4], idct_add(dest[j - 1] + block_offset[i + 4],
h->mb + (i * 16 << PIXEL_SHIFT),
sl->mb + (i * 16 << PIXEL_SHIFT),
uvlinesize); uvlinesize);
} }
} }
@@ -238,24 +238,24 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h, H264SliceContext *sl)
qp[1] = sl->chroma_qp[1]; qp[1] = sl->chroma_qp[1];
} }
if (sl->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 0]]) if (sl->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 0]])
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 1 << PIXEL_SHIFT),
h->h264dsp.h264_chroma_dc_dequant_idct(sl->mb + (16 * 16 * 1 << PIXEL_SHIFT),
h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][qp[0]][0]); h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][qp[0]][0]);
if (sl->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 1]]) if (sl->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 1]])
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 2 << PIXEL_SHIFT),
h->h264dsp.h264_chroma_dc_dequant_idct(sl->mb + (16 * 16 * 2 << PIXEL_SHIFT),
h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][qp[1]][0]); h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][qp[1]][0]);
h->h264dsp.h264_idct_add8(dest, block_offset, h->h264dsp.h264_idct_add8(dest, block_offset,
h->mb, uvlinesize,
sl->mb, uvlinesize,
sl->non_zero_count_cache); sl->non_zero_count_cache);
} else if (CONFIG_SVQ3_DECODER) { } else if (CONFIG_SVQ3_DECODER) {
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 1,
h->h264dsp.h264_chroma_dc_dequant_idct(sl->mb + 16 * 16 * 1,
h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][sl->chroma_qp[0]][0]); h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][sl->chroma_qp[0]][0]);
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 2,
h->h264dsp.h264_chroma_dc_dequant_idct(sl->mb + 16 * 16 * 2,
h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][sl->chroma_qp[1]][0]); h->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][sl->chroma_qp[1]][0]);
for (j = 1; j < 3; j++) { for (j = 1; j < 3; j++) {
for (i = j * 16; i < j * 16 + 4; i++) for (i = j * 16; i < j * 16 + 4; i++)
if (sl->non_zero_count_cache[scan8[i]] || h->mb[i * 16]) {
if (sl->non_zero_count_cache[scan8[i]] || sl->mb[i * 16]) {
uint8_t *const ptr = dest[j - 1] + block_offset[i]; uint8_t *const ptr = dest[j - 1] + block_offset[i];
ff_svq3_add_idct_c(ptr, h->mb + i * 16,
ff_svq3_add_idct_c(ptr, sl->mb + i * 16,
uvlinesize, uvlinesize,
ff_h264_chroma_qp[0][sl->qscale + 12] - 12, 2); ff_h264_chroma_qp[0][sl->qscale + 12] - 12, 2);
} }


+ 4
- 3
libavcodec/h264_slice.c View File

@@ -498,9 +498,6 @@ int ff_h264_update_thread_context(AVCodecContext *dst,
memset(h->sps_buffers, 0, sizeof(h->sps_buffers)); memset(h->sps_buffers, 0, sizeof(h->sps_buffers));
memset(h->pps_buffers, 0, sizeof(h->pps_buffers)); memset(h->pps_buffers, 0, sizeof(h->pps_buffers));
memset(&h->er, 0, sizeof(h->er)); memset(&h->er, 0, sizeof(h->er));
memset(&h->mb, 0, sizeof(h->mb));
memset(&h->mb_luma_dc, 0, sizeof(h->mb_luma_dc));
memset(&h->mb_padding, 0, sizeof(h->mb_padding));
h->context_initialized = 0; h->context_initialized = 0;


memset(&h->cur_pic, 0, sizeof(h->cur_pic)); memset(&h->cur_pic, 0, sizeof(h->cur_pic));
@@ -509,6 +506,10 @@ int ff_h264_update_thread_context(AVCodecContext *dst,


h->slice_ctx = orig_slice_ctx; h->slice_ctx = orig_slice_ctx;


memset(&h->slice_ctx[0].mb, 0, sizeof(h->slice_ctx[0].mb));
memset(&h->slice_ctx[0].mb_luma_dc, 0, sizeof(h->slice_ctx[0].mb_luma_dc));
memset(&h->slice_ctx[0].mb_padding, 0, sizeof(h->slice_ctx[0].mb_padding));

h->avctx = dst; h->avctx = dst;
h->DPB = NULL; h->DPB = NULL;
h->qscale_table_pool = NULL; h->qscale_table_pool = NULL;


+ 6
- 6
libavcodec/svq3.c View File

@@ -707,9 +707,9 @@ static int svq3_decode_mb(SVQ3Context *s, unsigned int mb_type)
} }
} }
if (IS_INTRA16x16(mb_type)) { if (IS_INTRA16x16(mb_type)) {
AV_ZERO128(h->mb_luma_dc[0] + 0);
AV_ZERO128(h->mb_luma_dc[0] + 8);
if (svq3_decode_block(&h->gb, h->mb_luma_dc[0], 0, 1)) {
AV_ZERO128(sl->mb_luma_dc[0] + 0);
AV_ZERO128(sl->mb_luma_dc[0] + 8);
if (svq3_decode_block(&h->gb, sl->mb_luma_dc[0], 0, 1)) {
av_log(h->avctx, AV_LOG_ERROR, av_log(h->avctx, AV_LOG_ERROR,
"error while decoding intra luma dc\n"); "error while decoding intra luma dc\n");
return -1; return -1;
@@ -728,7 +728,7 @@ static int svq3_decode_mb(SVQ3Context *s, unsigned int mb_type)
: (4 * i + j); : (4 * i + j);
sl->non_zero_count_cache[scan8[k]] = 1; sl->non_zero_count_cache[scan8[k]] = 1;


if (svq3_decode_block(&h->gb, &h->mb[16 * k], index, type)) {
if (svq3_decode_block(&h->gb, &sl->mb[16 * k], index, type)) {
av_log(h->avctx, AV_LOG_ERROR, av_log(h->avctx, AV_LOG_ERROR,
"error while decoding block\n"); "error while decoding block\n");
return -1; return -1;
@@ -738,7 +738,7 @@ static int svq3_decode_mb(SVQ3Context *s, unsigned int mb_type)


if ((cbp & 0x30)) { if ((cbp & 0x30)) {
for (i = 1; i < 3; ++i) for (i = 1; i < 3; ++i)
if (svq3_decode_block(&h->gb, &h->mb[16 * 16 * i], 0, 3)) {
if (svq3_decode_block(&h->gb, &sl->mb[16 * 16 * i], 0, 3)) {
av_log(h->avctx, AV_LOG_ERROR, av_log(h->avctx, AV_LOG_ERROR,
"error while decoding chroma dc block\n"); "error while decoding chroma dc block\n");
return -1; return -1;
@@ -750,7 +750,7 @@ static int svq3_decode_mb(SVQ3Context *s, unsigned int mb_type)
k = 16 * i + j; k = 16 * i + j;
sl->non_zero_count_cache[scan8[k]] = 1; sl->non_zero_count_cache[scan8[k]] = 1;


if (svq3_decode_block(&h->gb, &h->mb[16 * k], 1, 1)) {
if (svq3_decode_block(&h->gb, &sl->mb[16 * k], 1, 1)) {
av_log(h->avctx, AV_LOG_ERROR, av_log(h->avctx, AV_LOG_ERROR,
"error while decoding chroma ac block\n"); "error while decoding chroma ac block\n");
return -1; return -1;


Write Preview
Loading…
Cancel
Save