diff --git a/libavcodec/dxva2_h264.c b/libavcodec/dxva2_h264.c index cb033a2e92..dbc5087619 100644 --- a/libavcodec/dxva2_h264.c +++ b/libavcodec/dxva2_h264.c @@ -276,7 +276,7 @@ static void fill_slice_long(AVCodecContext *avctx, DXVA_Slice_H264_Long *slice, } } slice->slice_qs_delta = 0; /* XXX not implemented by Libav */ - slice->slice_qp_delta = h->qscale - h->pps.init_qp; + slice->slice_qp_delta = sl->qscale - h->pps.init_qp; slice->redundant_pic_cnt = h->redundant_pic_count; if (h->slice_type == AV_PICTURE_TYPE_B) slice->direct_spatial_mv_pred_flag = h->direct_spatial_mv_pred; diff --git a/libavcodec/h264.h b/libavcodec/h264.h index a8e1101aa5..1820ca5990 100644 --- a/libavcodec/h264.h +++ b/libavcodec/h264.h @@ -299,6 +299,9 @@ typedef struct H264Picture { typedef struct H264SliceContext { struct H264Context *h264; + int qscale; + int chroma_qp[2]; // QPc + // Weighted pred stuff int use_weight; int use_weight_chroma; @@ -332,7 +335,6 @@ typedef struct H264Context { int nb_slice_ctx; int pixel_shift; ///< 0 for 8-bit H264, 1 for high-bit-depth H264 - int chroma_qp[2]; // QPc int qp_thresh; ///< QP threshold to skip loopfilter @@ -341,7 +343,6 @@ typedef struct H264Context { ptrdiff_t linesize, uvlinesize; int chroma_x_shift, chroma_y_shift; - int qscale; int droppable; int coded_picture_number; int low_delay; @@ -804,15 +805,15 @@ void ff_h264_decode_init_vlc(void); * Decode a macroblock * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error */ -int ff_h264_decode_mb_cavlc(H264Context *h); +int ff_h264_decode_mb_cavlc(H264Context *h, H264SliceContext *sl); /** * Decode a CABAC coded macroblock * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR on error */ -int ff_h264_decode_mb_cabac(H264Context *h); +int ff_h264_decode_mb_cabac(H264Context *h, H264SliceContext *sl); -void ff_h264_init_cabac_states(H264Context *h); +void ff_h264_init_cabac_states(H264Context *h, H264SliceContext *sl); void h264_init_dequant_tables(H264Context *h); @@ -820,10 +821,10 @@ void ff_h264_direct_dist_scale_factor(H264Context *const h); void ff_h264_direct_ref_list_init(H264Context *const h); void ff_h264_pred_direct_motion(H264Context *const h, int *mb_type); -void ff_h264_filter_mb_fast(H264Context *h, int mb_x, int mb_y, +void ff_h264_filter_mb_fast(H264Context *h, H264SliceContext *sl, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize); -void ff_h264_filter_mb(H264Context *h, int mb_x, int mb_y, +void ff_h264_filter_mb(H264Context *h, H264SliceContext *sl, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize); diff --git a/libavcodec/h264_cabac.c b/libavcodec/h264_cabac.c index 0ad8ac0bde..5699bceec1 100644 --- a/libavcodec/h264_cabac.c +++ b/libavcodec/h264_cabac.c @@ -1261,10 +1261,11 @@ static const int8_t cabac_context_init_PB[3][1024][2] = } }; -void ff_h264_init_cabac_states(H264Context *h) { +void ff_h264_init_cabac_states(H264Context *h, H264SliceContext *sl) +{ int i; const int8_t (*tab)[2]; - const int slice_qp = av_clip(h->qscale - 6*(h->sps.bit_depth_luma-8), 0, 51); + const int slice_qp = av_clip(sl->qscale - 6*(h->sps.bit_depth_luma-8), 0, 51); if( h->slice_type_nos == AV_PICTURE_TYPE_I ) tab = cabac_context_init_I; else tab = cabac_context_init_PB[h->cabac_init_idc]; @@ -1830,12 +1831,14 @@ static av_always_inline void decode_cabac_residual_nondc(H264Context *h, decode_cabac_residual_nondc_internal( h, block, cat, n, scantable, qmul, max_coeff ); } -static av_always_inline void decode_cabac_luma_residual( H264Context *h, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p ) +static av_always_inline void decode_cabac_luma_residual(H264Context *h, H264SliceContext *sl, + const uint8_t *scan, const uint8_t *scan8x8, + int pixel_shift, int mb_type, int cbp, int p) { static const uint8_t ctx_cat[4][3] = {{0,6,10},{1,7,11},{2,8,12},{5,9,13}}; const uint32_t *qmul; int i8x8, i4x4; - int qscale = p == 0 ? h->qscale : h->chroma_qp[p-1]; + int qscale = p == 0 ? sl->qscale : sl->chroma_qp[p - 1]; if( IS_INTRA16x16( mb_type ) ) { AV_ZERO128(h->mb_luma_dc[p]+0); AV_ZERO128(h->mb_luma_dc[p]+8); @@ -1880,7 +1883,8 @@ static av_always_inline void decode_cabac_luma_residual( H264Context *h, const u * Decode a macroblock. * @return 0 if OK, ER_AC_ERROR / ER_DC_ERROR / ER_MV_ERROR if an error is noticed */ -int ff_h264_decode_mb_cabac(H264Context *h) { +int ff_h264_decode_mb_cabac(H264Context *h, H264SliceContext *sl) +{ int mb_xy; int mb_type, partition_count, cbp = 0; int dct8x8_allowed= h->pps.transform_8x8_mode; @@ -1906,7 +1910,7 @@ int ff_h264_decode_mb_cabac(H264Context *h) { h->mb_mbaff = h->mb_field_decoding_flag = decode_cabac_field_decoding_flag(h); } - decode_mb_skip(h); + decode_mb_skip(h, sl); h->cbp_table[mb_xy] = 0; h->chroma_pred_mode_table[mb_xy] = 0; @@ -2323,11 +2327,11 @@ decode_intra_mb: const uint32_t *qmul; if(IS_INTERLACED(mb_type)){ - scan8x8= h->qscale ? h->field_scan8x8 : h->field_scan8x8_q0; - scan= h->qscale ? h->field_scan : h->field_scan_q0; + scan8x8 = sl->qscale ? h->field_scan8x8 : h->field_scan8x8_q0; + scan = sl->qscale ? h->field_scan : h->field_scan_q0; }else{ - scan8x8= h->qscale ? h->zigzag_scan8x8 : h->zigzag_scan8x8_q0; - scan= h->qscale ? h->zigzag_scan : h->zigzag_scan_q0; + scan8x8 = sl->qscale ? h->zigzag_scan8x8 : h->zigzag_scan8x8_q0; + scan = sl->qscale ? h->zigzag_scan : h->zigzag_scan_q0; } // decode_cabac_mb_dqp @@ -2350,20 +2354,20 @@ decode_intra_mb: else val= -((val + 1)>>1); h->last_qscale_diff = val; - h->qscale += val; - if(((unsigned)h->qscale) > max_qp){ - if(h->qscale<0) h->qscale+= max_qp+1; - else h->qscale-= max_qp+1; + sl->qscale += val; + if (((unsigned)sl->qscale) > max_qp){ + if (sl->qscale < 0) sl->qscale += max_qp + 1; + else sl->qscale -= max_qp + 1; } - h->chroma_qp[0] = get_chroma_qp(h, 0, h->qscale); - h->chroma_qp[1] = get_chroma_qp(h, 1, h->qscale); + sl->chroma_qp[0] = get_chroma_qp(h, 0, sl->qscale); + sl->chroma_qp[1] = get_chroma_qp(h, 1, sl->qscale); }else h->last_qscale_diff=0; - decode_cabac_luma_residual(h, scan, scan8x8, pixel_shift, mb_type, cbp, 0); + decode_cabac_luma_residual(h, sl, scan, scan8x8, pixel_shift, mb_type, cbp, 0); if (CHROMA444(h)) { - decode_cabac_luma_residual(h, scan, scan8x8, pixel_shift, mb_type, cbp, 1); - decode_cabac_luma_residual(h, scan, scan8x8, pixel_shift, mb_type, cbp, 2); + decode_cabac_luma_residual(h, sl, scan, scan8x8, pixel_shift, mb_type, cbp, 1); + decode_cabac_luma_residual(h, sl, scan, scan8x8, pixel_shift, mb_type, cbp, 2); } else if (CHROMA422(h)) { if( cbp&0x30 ){ int c; @@ -2377,7 +2381,7 @@ decode_intra_mb: int c, i, i8x8; for( c = 0; c < 2; c++ ) { int16_t *mb = h->mb + (16*(16 + 16*c) << pixel_shift); - qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][h->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 (i = 0; i < 4; i++) { const int index = 16 + 16 * c + 8*i8x8 + i; @@ -2400,7 +2404,7 @@ decode_intra_mb: if( cbp&0x20 ) { int c, i; for( c = 0; c < 2; c++ ) { - qmul = h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][h->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++ ) { const int index = 16 + 16 * c + i; decode_cabac_residual_nondc(h, h->mb + (16*index << pixel_shift), 4, index, scan + 1, qmul, 15); @@ -2418,7 +2422,7 @@ decode_intra_mb: h->last_qscale_diff = 0; } - h->cur_pic.qscale_table[mb_xy] = h->qscale; + h->cur_pic.qscale_table[mb_xy] = sl->qscale; write_back_non_zero_count(h); return 0; diff --git a/libavcodec/h264_cavlc.c b/libavcodec/h264_cavlc.c index 0ab0355143..0c517f104c 100644 --- a/libavcodec/h264_cavlc.c +++ b/libavcodec/h264_cavlc.c @@ -623,9 +623,9 @@ static int decode_residual(H264Context *h, GetBitContext *gb, int16_t *block, in return 0; } -static av_always_inline int decode_luma_residual(H264Context *h, GetBitContext *gb, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p){ +static av_always_inline int decode_luma_residual(H264Context *h, H264SliceContext *sl, GetBitContext *gb, const uint8_t *scan, const uint8_t *scan8x8, int pixel_shift, int mb_type, int cbp, int p){ int i4x4, i8x8; - int qscale = p == 0 ? h->qscale : h->chroma_qp[p-1]; + int qscale = p == 0 ? sl->qscale : sl->chroma_qp[p - 1]; if(IS_INTRA16x16(mb_type)){ AV_ZERO128(h->mb_luma_dc[p]+0); AV_ZERO128(h->mb_luma_dc[p]+8); @@ -689,7 +689,8 @@ static av_always_inline int decode_luma_residual(H264Context *h, GetBitContext * } } -int ff_h264_decode_mb_cavlc(H264Context *h){ +int ff_h264_decode_mb_cavlc(H264Context *h, H264SliceContext *sl) +{ int mb_xy; int partition_count; unsigned int mb_type, cbp; @@ -711,7 +712,7 @@ int ff_h264_decode_mb_cavlc(H264Context *h){ if(h->mb_skip_run==0) h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&h->gb); } - decode_mb_skip(h); + decode_mb_skip(h, sl); return 0; } } @@ -1082,38 +1083,38 @@ decode_intra_mb: const int max_qp = 51 + 6*(h->sps.bit_depth_luma-8); if(IS_INTERLACED(mb_type)){ - scan8x8= h->qscale ? h->field_scan8x8_cavlc : h->field_scan8x8_cavlc_q0; - scan= h->qscale ? h->field_scan : h->field_scan_q0; + scan8x8 = sl->qscale ? h->field_scan8x8_cavlc : h->field_scan8x8_cavlc_q0; + scan = sl->qscale ? h->field_scan : h->field_scan_q0; }else{ - scan8x8= h->qscale ? h->zigzag_scan8x8_cavlc : h->zigzag_scan8x8_cavlc_q0; - scan= h->qscale ? h->zigzag_scan : h->zigzag_scan_q0; + scan8x8 = sl->qscale ? h->zigzag_scan8x8_cavlc : h->zigzag_scan8x8_cavlc_q0; + scan = sl->qscale ? h->zigzag_scan : h->zigzag_scan_q0; } dquant= get_se_golomb(&h->gb); - h->qscale += dquant; + sl->qscale += dquant; - if(((unsigned)h->qscale) > max_qp){ - if(h->qscale<0) h->qscale+= max_qp+1; - else h->qscale-= max_qp+1; - if(((unsigned)h->qscale) > max_qp){ + if (((unsigned)sl->qscale) > max_qp){ + if (sl->qscale < 0) sl->qscale += max_qp + 1; + else sl->qscale -= max_qp+1; + if (((unsigned)sl->qscale) > max_qp){ av_log(h->avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, h->mb_x, h->mb_y); return -1; } } - h->chroma_qp[0]= get_chroma_qp(h, 0, h->qscale); - h->chroma_qp[1]= get_chroma_qp(h, 1, h->qscale); + sl->chroma_qp[0] = get_chroma_qp(h, 0, sl->qscale); + sl->chroma_qp[1] = get_chroma_qp(h, 1, sl->qscale); - if( (ret = decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 0)) < 0 ){ + if ((ret = decode_luma_residual(h, sl, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 0)) < 0 ) { return -1; } h->cbp_table[mb_xy] |= ret << 12; if (CHROMA444(h)) { - if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 1) < 0 ){ + if (decode_luma_residual(h, sl, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 1) < 0 ) { return -1; } - if( decode_luma_residual(h, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 2) < 0 ){ + if (decode_luma_residual(h, sl, gb, scan, scan8x8, pixel_shift, mb_type, cbp, 2) < 0 ) { return -1; } } else if (CHROMA422(h)) { @@ -1128,7 +1129,7 @@ decode_intra_mb: if(cbp&0x20){ 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)][h->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); for (i8x8 = 0; i8x8 < 2; i8x8++) { for (i4x4 = 0; i4x4 < 4; i4x4++) { @@ -1153,7 +1154,7 @@ decode_intra_mb: if(cbp&0x20){ 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)][h->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++){ const int index= 16 + 16*chroma_idx + i4x4; if( decode_residual(h, gb, h->mb + (16*index << pixel_shift), index, scan + 1, qmul, 15) < 0){ @@ -1171,7 +1172,7 @@ decode_intra_mb: fill_rectangle(&h->non_zero_count_cache[scan8[16]], 4, 4, 8, 0, 1); fill_rectangle(&h->non_zero_count_cache[scan8[32]], 4, 4, 8, 0, 1); } - h->cur_pic.qscale_table[mb_xy] = h->qscale; + h->cur_pic.qscale_table[mb_xy] = sl->qscale; write_back_non_zero_count(h); return 0; diff --git a/libavcodec/h264_loopfilter.c b/libavcodec/h264_loopfilter.c index a15130cb98..f6457286a1 100644 --- a/libavcodec/h264_loopfilter.c +++ b/libavcodec/h264_loopfilter.c @@ -415,10 +415,14 @@ static av_always_inline void h264_filter_mb_fast_internal(H264Context *h, } } -void ff_h264_filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize) { +void ff_h264_filter_mb_fast(H264Context *h, H264SliceContext *sl, + int mb_x, int mb_y, uint8_t *img_y, + uint8_t *img_cb, uint8_t *img_cr, + unsigned int linesize, unsigned int uvlinesize) +{ assert(!FRAME_MBAFF(h)); if(!h->h264dsp.h264_loop_filter_strength || h->pps.chroma_qp_diff) { - ff_h264_filter_mb(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize); + ff_h264_filter_mb(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize); return; } @@ -462,7 +466,14 @@ static int check_mv(H264Context *h, long b_idx, long bn_idx, int mvy_limit){ return v; } -static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize, int mb_xy, int mb_type, int mvy_limit, int first_vertical_edge_done, int a, int b, int chroma, int dir) { +static av_always_inline void filter_mb_dir(H264Context *h, H264SliceContext *sl, + int mb_x, int mb_y, + uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, + unsigned int linesize, unsigned int uvlinesize, + int mb_xy, int mb_type, int mvy_limit, + int first_vertical_edge_done, int a, int b, + int chroma, int dir) +{ int edge; int chroma_qp_avg[2]; int chroma444 = CHROMA444(h); @@ -518,8 +529,8 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u tprintf(h->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize); { int i; for (i = 0; i < 4; i++) tprintf(h->avctx, " bS[%d]:%d", i, bS[i]); tprintf(h->avctx, "\n"); } filter_mb_edgeh( &img_y[j*linesize], tmp_linesize, bS, qp, a, b, h, 0 ); - chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.qscale_table[mbn_xy]) + 1) >> 1; - chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.qscale_table[mbn_xy]) + 1) >> 1; + chroma_qp_avg[0] = (sl->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.qscale_table[mbn_xy]) + 1) >> 1; + chroma_qp_avg[1] = (sl->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.qscale_table[mbn_xy]) + 1) >> 1; if (chroma) { if (chroma444) { filter_mb_edgeh (&img_cb[j*uvlinesize], tmp_uvlinesize, bS, chroma_qp_avg[0], a, b, h, 0); @@ -581,8 +592,8 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u if(bS[0]+bS[1]+bS[2]+bS[3]){ qp = (h->cur_pic.qscale_table[mb_xy] + h->cur_pic.qscale_table[mbm_xy] + 1) >> 1; tprintf(h->avctx, "filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize); - chroma_qp_avg[0] = (h->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.qscale_table[mbm_xy]) + 1) >> 1; - chroma_qp_avg[1] = (h->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.qscale_table[mbm_xy]) + 1) >> 1; + chroma_qp_avg[0] = (sl->chroma_qp[0] + get_chroma_qp(h, 0, h->cur_pic.qscale_table[mbm_xy]) + 1) >> 1; + chroma_qp_avg[1] = (sl->chroma_qp[1] + get_chroma_qp(h, 1, h->cur_pic.qscale_table[mbm_xy]) + 1) >> 1; if( dir == 0 ) { filter_mb_edgev( &img_y[0], linesize, bS, qp, a, b, h, 1 ); if (chroma) { @@ -668,11 +679,11 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u filter_mb_edgev( &img_y[4*edge << h->pixel_shift], linesize, bS, qp, a, b, h, 0 ); if (chroma) { if (chroma444) { - filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0); - filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0); + filter_mb_edgev ( &img_cb[4*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); + filter_mb_edgev ( &img_cr[4*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); } else if( (edge&1) == 0 ) { - filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0); - filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0); + filter_mb_edgecv( &img_cb[2*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); + filter_mb_edgecv( &img_cr[2*edge << h->pixel_shift], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); } } } else { @@ -680,18 +691,18 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u if (deblock_edge) filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0); if (chroma) { - filter_mb_edgech(&img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0); - filter_mb_edgech(&img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0); + filter_mb_edgech(&img_cb[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); + filter_mb_edgech(&img_cr[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); } } else { filter_mb_edgeh(&img_y[4*edge*linesize], linesize, bS, qp, a, b, h, 0); if (chroma) { if (chroma444) { - filter_mb_edgeh (&img_cb[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0); - filter_mb_edgeh (&img_cr[4*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0); + filter_mb_edgeh (&img_cb[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); + filter_mb_edgeh (&img_cr[4*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); } else if ((edge&1) == 0) { - filter_mb_edgech(&img_cb[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[0], a, b, h, 0); - filter_mb_edgech(&img_cr[2*edge*uvlinesize], uvlinesize, bS, h->chroma_qp[1], a, b, h, 0); + filter_mb_edgech(&img_cb[2*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[0], a, b, h, 0); + filter_mb_edgech(&img_cr[2*edge*uvlinesize], uvlinesize, bS, sl->chroma_qp[1], a, b, h, 0); } } } @@ -699,7 +710,11 @@ static av_always_inline void filter_mb_dir(H264Context *h, int mb_x, int mb_y, u } } -void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize) { +void ff_h264_filter_mb(H264Context *h, H264SliceContext *sl, + int mb_x, int mb_y, + uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, + unsigned int linesize, unsigned int uvlinesize) +{ const int mb_xy= mb_x + mb_y*h->mb_stride; const int mb_type = h->cur_pic.mb_type[mb_xy]; const int mvy_limit = IS_INTERLACED(mb_type) ? 2 : 4; @@ -817,13 +832,13 @@ void ff_h264_filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint { int dir; for (dir = 0; dir < 2; dir++) - filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, + filter_mb_dir(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, dir ? 0 : first_vertical_edge_done, a, b, chroma, dir); } #else - filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, a, b, chroma, 0); - filter_mb_dir(h, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, a, b, chroma, 1); + filter_mb_dir(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, first_vertical_edge_done, a, b, chroma, 0); + filter_mb_dir(h, sl, mb_x, mb_y, img_y, img_cb, img_cr, linesize, uvlinesize, mb_xy, mb_type, mvy_limit, 0, a, b, chroma, 1); #endif } diff --git a/libavcodec/h264_mb.c b/libavcodec/h264_mb.c index b2899b62a7..86beb19b37 100644 --- a/libavcodec/h264_mb.c +++ b/libavcodec/h264_mb.c @@ -605,6 +605,7 @@ static av_always_inline void dctcoef_set(int16_t *mb, int high_bit_depth, } static av_always_inline void hl_decode_mb_predict_luma(H264Context *h, + H264SliceContext *sl, int mb_type, int is_h264, int simple, int transform_bypass, @@ -616,7 +617,7 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h, void (*idct_add)(uint8_t *dst, int16_t *block, int stride); void (*idct_dc_add)(uint8_t *dst, int16_t *block, int stride); int i; - int qscale = p == 0 ? h->qscale : h->chroma_qp[p - 1]; + int qscale = p == 0 ? sl->qscale : sl->chroma_qp[p - 1]; block_offset += 16 * p; if (IS_INTRA4x4(mb_type)) { if (IS_8x8DCT(mb_type)) { @@ -720,7 +721,8 @@ static av_always_inline void hl_decode_mb_predict_luma(H264Context *h, } } -static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, int mb_type, +static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, H264SliceContext *sl, + int mb_type, int is_h264, int simple, int transform_bypass, int pixel_shift, @@ -784,7 +786,7 @@ static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, int mb_type, // FIXME benchmark weird rule, & below uint8_t *const ptr = dest_y + block_offset[i]; ff_svq3_add_idct_c(ptr, h->mb + i * 16 + p * 256, linesize, - h->qscale, IS_INTRA(mb_type) ? 1 : 0); + sl->qscale, IS_INTRA(mb_type) ? 1 : 0); } } } @@ -807,7 +809,7 @@ void ff_h264_hl_decode_mb(H264Context *h, H264SliceContext *sl) const int mb_xy = h->mb_xy; const int mb_type = h->cur_pic.mb_type[mb_xy]; int is_complex = CONFIG_SMALL || h->is_complex || - IS_INTRA_PCM(mb_type) || h->qscale == 0; + IS_INTRA_PCM(mb_type) || sl->qscale == 0; if (CHROMA444(h)) { if (is_complex || h->pixel_shift) diff --git a/libavcodec/h264_mb_template.c b/libavcodec/h264_mb_template.c index 262ad0c2f6..4805f69055 100644 --- a/libavcodec/h264_mb_template.c +++ b/libavcodec/h264_mb_template.c @@ -50,7 +50,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h, H264SliceContext *sl) int linesize, uvlinesize /*dct_offset*/; int i, j; int *block_offset = &h->block_offset[0]; - const int transform_bypass = !SIMPLE && (h->qscale == 0 && h->sps.transform_bypass); + const int transform_bypass = !SIMPLE && (sl->qscale == 0 && h->sps.transform_bypass); /* is_h264 should always be true if SVQ3 is disabled. */ const int is_h264 = !CONFIG_SVQ3_DECODER || SIMPLE || h->avctx->codec_id == AV_CODEC_ID_H264; void (*idct_add)(uint8_t *dst, int16_t *block, int stride); @@ -167,7 +167,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h, H264SliceContext *sl) h->hpc.pred8x8[h->chroma_pred_mode](dest_cr, uvlinesize); } - hl_decode_mb_predict_luma(h, mb_type, is_h264, SIMPLE, + hl_decode_mb_predict_luma(h, sl, mb_type, is_h264, SIMPLE, transform_bypass, PIXEL_SHIFT, block_offset, linesize, dest_y, 0); @@ -190,7 +190,7 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h, H264SliceContext *sl) } } - hl_decode_mb_idct_luma(h, mb_type, is_h264, SIMPLE, transform_bypass, + hl_decode_mb_idct_luma(h, sl, mb_type, is_h264, SIMPLE, transform_bypass, PIXEL_SHIFT, block_offset, linesize, dest_y, 0); if ((SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) && @@ -231,11 +231,11 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h, H264SliceContext *sl) if (is_h264) { int qp[2]; if (chroma422) { - qp[0] = h->chroma_qp[0] + 3; - qp[1] = h->chroma_qp[1] + 3; + qp[0] = sl->chroma_qp[0] + 3; + qp[1] = sl->chroma_qp[1] + 3; } else { - qp[0] = h->chroma_qp[0]; - qp[1] = h->chroma_qp[1]; + qp[0] = sl->chroma_qp[0]; + qp[1] = sl->chroma_qp[1]; } if (h->non_zero_count_cache[scan8[CHROMA_DC_BLOCK_INDEX + 0]]) h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16 * 16 * 1 << PIXEL_SHIFT), @@ -248,16 +248,16 @@ static av_noinline void FUNC(hl_decode_mb)(H264Context *h, H264SliceContext *sl) h->non_zero_count_cache); } else if (CONFIG_SVQ3_DECODER) { h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16 * 16 * 1, - h->dequant4_coeff[IS_INTRA(mb_type) ? 1 : 4][h->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->dequant4_coeff[IS_INTRA(mb_type) ? 2 : 5][h->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 (i = j * 16; i < j * 16 + 4; i++) if (h->non_zero_count_cache[scan8[i]] || h->mb[i * 16]) { uint8_t *const ptr = dest[j - 1] + block_offset[i]; ff_svq3_add_idct_c(ptr, h->mb + i * 16, uvlinesize, - ff_h264_chroma_qp[0][h->qscale + 12] - 12, 2); + ff_h264_chroma_qp[0][sl->qscale + 12] - 12, 2); } } } @@ -282,7 +282,7 @@ static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h, H264SliceContext int linesize; int i, j, p; int *block_offset = &h->block_offset[0]; - const int transform_bypass = !SIMPLE && (h->qscale == 0 && h->sps.transform_bypass); + const int transform_bypass = !SIMPLE && (sl->qscale == 0 && h->sps.transform_bypass); const int plane_count = (SIMPLE || !CONFIG_GRAY || !(h->flags & CODEC_FLAG_GRAY)) ? 3 : 1; for (p = 0; p < plane_count; p++) { @@ -347,7 +347,7 @@ static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h, H264SliceContext linesize, 1, 1, SIMPLE, PIXEL_SHIFT); for (p = 0; p < plane_count; p++) - hl_decode_mb_predict_luma(h, mb_type, 1, SIMPLE, + hl_decode_mb_predict_luma(h, sl, mb_type, 1, SIMPLE, transform_bypass, PIXEL_SHIFT, block_offset, linesize, dest[p], p); @@ -363,7 +363,7 @@ static av_noinline void FUNC(hl_decode_mb_444)(H264Context *h, H264SliceContext } for (p = 0; p < plane_count; p++) - hl_decode_mb_idct_luma(h, mb_type, 1, SIMPLE, transform_bypass, + hl_decode_mb_idct_luma(h, sl, mb_type, 1, SIMPLE, transform_bypass, PIXEL_SHIFT, block_offset, linesize, dest[p], p); } diff --git a/libavcodec/h264_mvpred.h b/libavcodec/h264_mvpred.h index f7c38879cd..e89e91efd6 100644 --- a/libavcodec/h264_mvpred.h +++ b/libavcodec/h264_mvpred.h @@ -794,7 +794,7 @@ static void fill_decode_caches(H264Context *h, int mb_type) /** * decodes a P_SKIP or B_SKIP macroblock */ -static void av_unused decode_mb_skip(H264Context *h) +static void av_unused decode_mb_skip(H264Context *h, H264SliceContext *sl) { const int mb_xy = h->mb_xy; int mb_type = 0; @@ -822,7 +822,7 @@ static void av_unused decode_mb_skip(H264Context *h) write_back_motion(h, mb_type); h->cur_pic.mb_type[mb_xy] = mb_type; - h->cur_pic.qscale_table[mb_xy] = h->qscale; + h->cur_pic.qscale_table[mb_xy] = sl->qscale; h->slice_table[mb_xy] = h->slice_num; h->prev_mb_skipped = 1; } diff --git a/libavcodec/h264_slice.c b/libavcodec/h264_slice.c index d6ae9e7ca8..4e0384b3c6 100644 --- a/libavcodec/h264_slice.c +++ b/libavcodec/h264_slice.c @@ -1663,9 +1663,9 @@ int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl, H264Contex av_log(h->avctx, AV_LOG_ERROR, "QP %u out of range\n", tmp); return AVERROR_INVALIDDATA; } - h->qscale = tmp; - h->chroma_qp[0] = get_chroma_qp(h, 0, h->qscale); - h->chroma_qp[1] = get_chroma_qp(h, 1, h->qscale); + sl->qscale = tmp; + sl->chroma_qp[0] = get_chroma_qp(h, 0, sl->qscale); + sl->chroma_qp[1] = get_chroma_qp(h, 1, sl->qscale); // FIXME qscale / qp ... stuff if (h->slice_type == AV_PICTURE_TYPE_SP) get_bits1(&h->gb); /* sp_for_switch_flag */ @@ -1790,7 +1790,7 @@ int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl, H264Contex h->cur_pic_ptr->field_poc[0], h->cur_pic_ptr->field_poc[1], h->ref_count[0], h->ref_count[1], - h->qscale, + sl->qscale, h->deblocking_filter, h->slice_alpha_c0_offset, h->slice_beta_offset, sl->use_weight, @@ -2048,7 +2048,7 @@ static int fill_filter_caches(H264Context *h, int mb_type) return 0; } -static void loop_filter(H264Context *h, int start_x, int end_x) +static void loop_filter(H264Context *h, H264SliceContext *sl, int start_x, int end_x) { uint8_t *dest_y, *dest_cb, *dest_cr; int linesize, uvlinesize, mb_x, mb_y; @@ -2098,14 +2098,14 @@ static void loop_filter(H264Context *h, int start_x, int end_x) uvlinesize, 0); if (fill_filter_caches(h, mb_type)) continue; - h->chroma_qp[0] = get_chroma_qp(h, 0, h->cur_pic.qscale_table[mb_xy]); - h->chroma_qp[1] = get_chroma_qp(h, 1, h->cur_pic.qscale_table[mb_xy]); + sl->chroma_qp[0] = get_chroma_qp(h, 0, h->cur_pic.qscale_table[mb_xy]); + sl->chroma_qp[1] = get_chroma_qp(h, 1, h->cur_pic.qscale_table[mb_xy]); if (FRAME_MBAFF(h)) { - ff_h264_filter_mb(h, mb_x, mb_y, dest_y, dest_cb, dest_cr, + ff_h264_filter_mb(h, sl, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize); } else { - ff_h264_filter_mb_fast(h, mb_x, mb_y, dest_y, dest_cb, + ff_h264_filter_mb_fast(h, sl, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize); } } @@ -2113,8 +2113,8 @@ static void loop_filter(H264Context *h, int start_x, int end_x) h->slice_type = old_slice_type; h->mb_x = end_x; h->mb_y = end_mb_y - FRAME_MBAFF(h); - h->chroma_qp[0] = get_chroma_qp(h, 0, h->qscale); - h->chroma_qp[1] = get_chroma_qp(h, 1, h->qscale); + sl->chroma_qp[0] = get_chroma_qp(h, 0, sl->qscale); + sl->chroma_qp[1] = get_chroma_qp(h, 1, sl->qscale); } static void predict_field_decoding_flag(H264Context *h) @@ -2193,11 +2193,11 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) h->gb.buffer + get_bits_count(&h->gb) / 8, (get_bits_left(&h->gb) + 7) / 8); - ff_h264_init_cabac_states(h); + ff_h264_init_cabac_states(h, sl); for (;;) { // START_TIMER - int ret = ff_h264_decode_mb_cabac(h); + int ret = ff_h264_decode_mb_cabac(h, sl); int eos; // STOP_TIMER("decode_mb_cabac") @@ -2208,7 +2208,7 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) if (ret >= 0 && FRAME_MBAFF(h)) { h->mb_y++; - ret = ff_h264_decode_mb_cabac(h); + ret = ff_h264_decode_mb_cabac(h, sl); if (ret >= 0) ff_h264_hl_decode_mb(h, sl); @@ -2221,7 +2221,7 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) er_add_slice(h, h->resync_mb_x, h->resync_mb_y, h->mb_x - 1, h->mb_y, ER_MB_END); if (h->mb_x >= lf_x_start) - loop_filter(h, lf_x_start, h->mb_x + 1); + loop_filter(h, sl, lf_x_start, h->mb_x + 1); return 0; } if (ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 2) { @@ -2235,7 +2235,7 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) } if (++h->mb_x >= h->mb_width) { - loop_filter(h, lf_x_start, h->mb_x); + loop_filter(h, sl, lf_x_start, h->mb_x); h->mb_x = lf_x_start = 0; decode_finish_row(h); ++h->mb_y; @@ -2252,13 +2252,13 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) er_add_slice(h, h->resync_mb_x, h->resync_mb_y, h->mb_x - 1, h->mb_y, ER_MB_END); if (h->mb_x > lf_x_start) - loop_filter(h, lf_x_start, h->mb_x); + loop_filter(h, sl, lf_x_start, h->mb_x); return 0; } } } else { for (;;) { - int ret = ff_h264_decode_mb_cavlc(h); + int ret = ff_h264_decode_mb_cavlc(h, sl); if (ret >= 0) ff_h264_hl_decode_mb(h, sl); @@ -2266,7 +2266,7 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) // FIXME optimal? or let mb_decode decode 16x32 ? if (ret >= 0 && FRAME_MBAFF(h)) { h->mb_y++; - ret = ff_h264_decode_mb_cavlc(h); + ret = ff_h264_decode_mb_cavlc(h, sl); if (ret >= 0) ff_h264_hl_decode_mb(h, sl); @@ -2282,7 +2282,7 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) } if (++h->mb_x >= h->mb_width) { - loop_filter(h, lf_x_start, h->mb_x); + loop_filter(h, sl, lf_x_start, h->mb_x); h->mb_x = lf_x_start = 0; decode_finish_row(h); ++h->mb_y; @@ -2317,7 +2317,7 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) er_add_slice(h, h->resync_mb_x, h->resync_mb_y, h->mb_x - 1, h->mb_y, ER_MB_END); if (h->mb_x > lf_x_start) - loop_filter(h, lf_x_start, h->mb_x); + loop_filter(h, sl, lf_x_start, h->mb_x); return 0; } else { diff --git a/libavcodec/svq3.c b/libavcodec/svq3.c index aa2911d04d..e6e4b0df7a 100644 --- a/libavcodec/svq3.c +++ b/libavcodec/svq3.c @@ -481,6 +481,7 @@ static inline int svq3_mc_dir(SVQ3Context *s, int size, int mode, static int svq3_decode_mb(SVQ3Context *s, unsigned int mb_type) { H264Context *h = &s->h; + H264SliceContext *sl = &h->slice_ctx[0]; int i, j, k, m, dir, mode; int cbp = 0; uint32_t vlc; @@ -697,10 +698,10 @@ static int svq3_decode_mb(SVQ3Context *s, unsigned int mb_type) } if (IS_INTRA16x16(mb_type) || (h->pict_type != AV_PICTURE_TYPE_I && s->adaptive_quant && cbp)) { - h->qscale += svq3_get_se_golomb(&h->gb); + sl->qscale += svq3_get_se_golomb(&h->gb); - if (h->qscale > 31u) { - av_log(h->avctx, AV_LOG_ERROR, "qscale:%d\n", h->qscale); + if (sl->qscale > 31u) { + av_log(h->avctx, AV_LOG_ERROR, "qscale:%d\n", sl->qscale); return -1; } } @@ -716,7 +717,7 @@ static int svq3_decode_mb(SVQ3Context *s, unsigned int mb_type) if (cbp) { const int index = IS_INTRA16x16(mb_type) ? 1 : 0; - const int type = ((h->qscale < 24 && IS_INTRA4x4(mb_type)) ? 2 : 1); + const int type = ((sl->qscale < 24 && IS_INTRA4x4(mb_type)) ? 2 : 1); for (i = 0; i < 4; i++) if ((cbp & (1 << i))) { @@ -772,6 +773,7 @@ static int svq3_decode_slice_header(AVCodecContext *avctx) { SVQ3Context *s = avctx->priv_data; H264Context *h = &s->h; + H264SliceContext *sl = &h->slice_ctx[0]; const int mb_xy = h->mb_xy; int i, header; unsigned slice_id; @@ -826,7 +828,7 @@ static int svq3_decode_slice_header(AVCodecContext *avctx) } h->slice_num = get_bits(&h->gb, 8); - h->qscale = get_bits(&h->gb, 5); + sl->qscale = get_bits(&h->gb, 5); s->adaptive_quant = get_bits1(&h->gb); /* unknown fields */ @@ -891,7 +893,7 @@ static av_cold int svq3_decode_init(AVCodecContext *avctx) avctx->pix_fmt = AV_PIX_FMT_YUVJ420P; avctx->color_range = AVCOL_RANGE_JPEG; - h->chroma_qp[0] = h->chroma_qp[1] = 4; + h->slice_ctx[0].chroma_qp[0] = h->slice_ctx[0].chroma_qp[1] = 4; h->chroma_x_shift = h->chroma_y_shift = 1; s->halfpel_flag = 1; @@ -1194,7 +1196,7 @@ static int svq3_decode_frame(AVCodecContext *avctx, void *data, "%c hpel:%d, tpel:%d aqp:%d qp:%d, slice_num:%02X\n", av_get_picture_type_char(h->pict_type), s->halfpel_flag, s->thirdpel_flag, - s->adaptive_quant, h->qscale, h->slice_num); + s->adaptive_quant, h->slice_ctx[0].qscale, h->slice_num); if (avctx->skip_frame >= AVDISCARD_NONREF && h->pict_type == AV_PICTURE_TYPE_B || avctx->skip_frame >= AVDISCARD_NONKEY && h->pict_type != AV_PICTURE_TYPE_I || diff --git a/libavcodec/vaapi_h264.c b/libavcodec/vaapi_h264.c index d910dba822..d88b548144 100644 --- a/libavcodec/vaapi_h264.c +++ b/libavcodec/vaapi_h264.c @@ -334,7 +334,7 @@ static int vaapi_h264_decode_slice(AVCodecContext *avctx, slice_param->num_ref_idx_l0_active_minus1 = h->list_count > 0 ? h->ref_count[0] - 1 : 0; slice_param->num_ref_idx_l1_active_minus1 = h->list_count > 1 ? h->ref_count[1] - 1 : 0; slice_param->cabac_init_idc = h->cabac_init_idc; - slice_param->slice_qp_delta = h->qscale - h->pps.init_qp; + slice_param->slice_qp_delta = sl->qscale - h->pps.init_qp; slice_param->disable_deblocking_filter_idc = h->deblocking_filter < 2 ? !h->deblocking_filter : h->deblocking_filter; slice_param->slice_alpha_c0_offset_div2 = h->slice_alpha_c0_offset / 2; slice_param->slice_beta_offset_div2 = h->slice_beta_offset / 2;