diff options
Diffstat (limited to 'libavcodec/h264.c')
| -rw-r--r-- | libavcodec/h264.c | 321 |
1 files changed, 219 insertions, 102 deletions
diff --git a/libavcodec/h264.c b/libavcodec/h264.c index b1204c0..071a6ad 100644 --- a/libavcodec/h264.c +++ b/libavcodec/h264.c @@ -2,20 +2,20 @@ * H.26L/H.264/AVC/JVT/14496-10/... decoder * Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at> * - * This file is part of Libav. + * This file is part of FFmpeg. * - * Libav is free software; you can redistribute it and/or + * FFmpeg is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation; either * version 2.1 of the License, or (at your option) any later version. * - * Libav is distributed in the hope that it will be useful, + * FFmpeg is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public - * License along with Libav; if not, write to the Free Software + * License along with FFmpeg; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA */ @@ -97,12 +97,9 @@ int ff_h264_check_intra4x4_pred_mode(H264Context *h){ } return 0; -} //FIXME cleanup like ff_h264_check_intra_pred_mode +} //FIXME cleanup like check_intra_pred_mode -/** - * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks. - */ -int ff_h264_check_intra_pred_mode(H264Context *h, int mode){ +static int check_intra_pred_mode(H264Context *h, int mode, int is_chroma){ MpegEncContext * const s = &h->s; static const int8_t top [7]= {LEFT_DC_PRED8x8, 1,-1,-1}; static const int8_t left[7]= { TOP_DC_PRED8x8,-1, 2,-1,DC_128_PRED8x8}; @@ -122,7 +119,7 @@ int ff_h264_check_intra_pred_mode(H264Context *h, int mode){ if((h->left_samples_available&0x8080) != 0x8080){ mode= left[ mode ]; - if(h->left_samples_available&0x8080){ //mad cow disease mode, aka MBAFF + constrained_intra_pred + if(is_chroma && (h->left_samples_available&0x8080)){ //mad cow disease mode, aka MBAFF + constrained_intra_pred mode= ALZHEIMER_DC_L0T_PRED8x8 + (!(h->left_samples_available&0x8000)) + 2*(mode == DC_128_PRED8x8); } if(mode<0){ @@ -134,6 +131,23 @@ int ff_h264_check_intra_pred_mode(H264Context *h, int mode){ return mode; } +/** + * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks. + */ +int ff_h264_check_intra16x16_pred_mode(H264Context *h, int mode) +{ + return check_intra_pred_mode(h, mode, 0); +} + +/** + * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks. + */ +int ff_h264_check_intra_chroma_pred_mode(H264Context *h, int mode) +{ + return check_intra_pred_mode(h, mode, 1); +} + + const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length){ int i, si, di; uint8_t *dst; @@ -174,20 +188,28 @@ const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_l i-= RS; } - if(i>=length-1){ //no escaped 0 - *dst_length= length; - *consumed= length+1; //+1 for the header - return src; - } - bufidx = h->nal_unit_type == NAL_DPC ? 1 : 0; // use second escape buffer for inter data - av_fast_malloc(&h->rbsp_buffer[bufidx], &h->rbsp_buffer_size[bufidx], length+FF_INPUT_BUFFER_PADDING_SIZE); + si=h->rbsp_buffer_size[bufidx]; + av_fast_malloc(&h->rbsp_buffer[bufidx], &h->rbsp_buffer_size[bufidx], length+FF_INPUT_BUFFER_PADDING_SIZE+MAX_MBPAIR_SIZE); dst= h->rbsp_buffer[bufidx]; + if(si != h->rbsp_buffer_size[bufidx]) + memset(dst + length, 0, FF_INPUT_BUFFER_PADDING_SIZE+MAX_MBPAIR_SIZE); if (dst == NULL){ return NULL; } + if(i>=length-1){ //no escaped 0 + *dst_length= length; + *consumed= length+1; //+1 for the header + if(h->s.avctx->flags2 & CODEC_FLAG2_FAST){ + return src; + }else{ + memcpy(dst, src, length); + return dst; + } + } + //printf("decoding esc\n"); memcpy(dst, src, i); si=di=i; @@ -671,7 +693,7 @@ static inline void prefetch_motion(H264Context *h, int list, int pixel_shift, in s->dsp.prefetch(src[1]+off, s->linesize, 4); s->dsp.prefetch(src[2]+off, s->linesize, 4); }else{ - off= ((mx>>1) << pixel_shift) + ((my>>1) + (s->mb_x&7))*s->uvlinesize + (64 << pixel_shift); + off= (((mx>>1)+64)<<pixel_shift) + ((my>>1) + (s->mb_x&7))*s->uvlinesize; s->dsp.prefetch(src[1]+off, src[2]-src[1], 2); } } @@ -942,7 +964,7 @@ static void clone_tables(H264Context *dst, H264Context *src, int i){ dst->list_counts = src->list_counts; dst->s.obmc_scratchpad = NULL; - ff_h264_pred_init(&dst->hpc, src->s.codec_id, src->sps.bit_depth_luma); + ff_h264_pred_init(&dst->hpc, src->s.codec_id, src->sps.bit_depth_luma, src->sps.chroma_format_idc); } /** @@ -970,30 +992,39 @@ static av_cold void common_init(H264Context *h){ s->height = s->avctx->height; s->codec_id= s->avctx->codec->id; - ff_h264dsp_init(&h->h264dsp, 8); - ff_h264_pred_init(&h->hpc, s->codec_id, 8); + s->avctx->bits_per_raw_sample = 8; + h->cur_chroma_format_idc = 1; + + ff_h264dsp_init(&h->h264dsp, + s->avctx->bits_per_raw_sample, h->cur_chroma_format_idc); + ff_h264_pred_init(&h->hpc, s->codec_id, + s->avctx->bits_per_raw_sample, h->cur_chroma_format_idc); h->dequant_coeff_pps= -1; s->unrestricted_mv=1; s->decode=1; //FIXME + s->dsp.dct_bits = 16; dsputil_init(&s->dsp, s->avctx); // needed so that idct permutation is known early memset(h->pps.scaling_matrix4, 16, 6*16*sizeof(uint8_t)); memset(h->pps.scaling_matrix8, 16, 2*64*sizeof(uint8_t)); } -int ff_h264_decode_extradata(H264Context *h) +int ff_h264_decode_extradata(H264Context *h, const uint8_t *buf, int size) { AVCodecContext *avctx = h->s.avctx; - if(avctx->extradata[0] == 1){ + if(!buf || size <= 0) + return -1; + + if(buf[0] == 1){ int i, cnt, nalsize; - unsigned char *p = avctx->extradata; + const unsigned char *p = buf; h->is_avc = 1; - if(avctx->extradata_size < 7) { + if(size < 7) { av_log(avctx, AV_LOG_ERROR, "avcC too short\n"); return -1; } @@ -1005,7 +1036,7 @@ int ff_h264_decode_extradata(H264Context *h) p += 6; for (i = 0; i < cnt; i++) { nalsize = AV_RB16(p) + 2; - if (p - avctx->extradata + nalsize > avctx->extradata_size) + if(nalsize > size - (p-buf)) return -1; if(decode_nal_units(h, p, nalsize) < 0) { av_log(avctx, AV_LOG_ERROR, "Decoding sps %d from avcC failed\n", i); @@ -1017,7 +1048,7 @@ int ff_h264_decode_extradata(H264Context *h) cnt = *(p++); // Number of pps for (i = 0; i < cnt; i++) { nalsize = AV_RB16(p) + 2; - if (p - avctx->extradata + nalsize > avctx->extradata_size) + if(nalsize > size - (p-buf)) return -1; if (decode_nal_units(h, p, nalsize) < 0) { av_log(avctx, AV_LOG_ERROR, "Decoding pps %d from avcC failed\n", i); @@ -1026,10 +1057,10 @@ int ff_h264_decode_extradata(H264Context *h) p += nalsize; } // Now store right nal length size, that will be use to parse all other nals - h->nal_length_size = (avctx->extradata[4] & 0x03) + 1; + h->nal_length_size = (buf[4] & 0x03) + 1; } else { h->is_avc = 0; - if(decode_nal_units(h, avctx->extradata, avctx->extradata_size) < 0) + if(decode_nal_units(h, buf, size) < 0) return -1; } return 0; @@ -1073,7 +1104,7 @@ av_cold int ff_h264_decode_init(AVCodecContext *avctx){ } if(avctx->extradata_size > 0 && avctx->extradata && - ff_h264_decode_extradata(h)) + ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size)) return -1; if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames < h->sps.num_reorder_frames){ @@ -1197,6 +1228,7 @@ static int decode_update_thread_context(AVCodecContext *dst, const AVCodecContex copy_picture_range(h->delayed_pic, h1->delayed_pic, MAX_DELAYED_PIC_COUNT+2, s, s1); h->last_slice_type = h1->last_slice_type; + h->sync = h1->sync; if(!s->current_picture_ptr) return 0; @@ -1383,7 +1415,7 @@ static void decode_postinit(H264Context *h, int setup_finished){ pics = 0; while(h->delayed_pic[pics]) pics++; - assert(pics <= MAX_DELAYED_PIC_COUNT); + av_assert0(pics <= MAX_DELAYED_PIC_COUNT); h->delayed_pic[pics++] = cur; if (cur->f.reference == 0) @@ -1428,15 +1460,21 @@ static void decode_postinit(H264Context *h, int setup_finished){ av_log(s->avctx, AV_LOG_DEBUG, "no picture\n"); } + if (h->next_output_pic && h->next_output_pic->sync) { + h->sync |= 2; + } + if (setup_finished) ff_thread_finish_setup(s->avctx); } -static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int chroma444, int simple){ +static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int simple){ MpegEncContext * const s = &h->s; uint8_t *top_border; int top_idx = 1; const int pixel_shift = h->pixel_shift; + int chroma444 = CHROMA444; + int chroma422 = CHROMA422; src_y -= linesize; src_cb -= uvlinesize; @@ -1460,6 +1498,14 @@ static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, ui AV_COPY128(top_border+16, src_cb + 15*uvlinesize); AV_COPY128(top_border+32, src_cr + 15*uvlinesize); } + } else if(chroma422){ + if (pixel_shift) { + AV_COPY128(top_border+32, src_cb + 15*uvlinesize); + AV_COPY128(top_border+48, src_cr + 15*uvlinesize); + } else { + AV_COPY64(top_border+16, src_cb + 15*uvlinesize); + AV_COPY64(top_border+24, src_cr + 15*uvlinesize); + } } else { if (pixel_shift) { AV_COPY128(top_border+32, src_cb+7*uvlinesize); @@ -1495,6 +1541,14 @@ static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, ui AV_COPY128(top_border+16, src_cb + 16*linesize); AV_COPY128(top_border+32, src_cr + 16*linesize); } + } else if(chroma422) { + if (pixel_shift) { + AV_COPY128(top_border+32, src_cb+16*uvlinesize); + AV_COPY128(top_border+48, src_cr+16*uvlinesize); + } else { + AV_COPY64(top_border+16, src_cb+16*uvlinesize); + AV_COPY64(top_border+24, src_cr+16*uvlinesize); + } } else { if (pixel_shift) { AV_COPY128(top_border+32, src_cb+8*uvlinesize); @@ -1773,10 +1827,11 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i /* is_h264 should always be true if SVQ3 is disabled. */ const int is_h264 = !CONFIG_SVQ3_DECODER || simple || s->codec_id == CODEC_ID_H264; void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride); + const int block_h = 16>>s->chroma_y_shift; dest_y = s->current_picture.f.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16; - dest_cb = s->current_picture.f.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * 8; - dest_cr = s->current_picture.f.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * 8; + dest_cb = s->current_picture.f.data[1] + (mb_x << pixel_shift)*8 + mb_y * s->uvlinesize * block_h; + dest_cr = s->current_picture.f.data[2] + (mb_x << pixel_shift)*8 + mb_y * s->uvlinesize * block_h; s->dsp.prefetch(dest_y + (s->mb_x&3)*4*s->linesize + (64 << pixel_shift), s->linesize, 4); s->dsp.prefetch(dest_cb + (s->mb_x&7)*s->uvlinesize + (64 << pixel_shift), dest_cr - dest_cb, 2); @@ -1789,8 +1844,8 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i block_offset = &h->block_offset[48]; if(mb_y&1){ //FIXME move out of this function? dest_y -= s->linesize*15; - dest_cb-= s->uvlinesize*7; - dest_cr-= s->uvlinesize*7; + dest_cb-= s->uvlinesize*(block_h-1); + dest_cr-= s->uvlinesize*(block_h-1); } if(FRAME_MBAFF) { int list; @@ -1816,8 +1871,8 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i } if (!simple && IS_INTRA_PCM(mb_type)) { + const int bit_depth = h->sps.bit_depth_luma; if (pixel_shift) { - const int bit_depth = h->sps.bit_depth_luma; int j; GetBitContext gb; init_get_bits(&gb, (uint8_t*)h->mb, 384*bit_depth); @@ -1831,23 +1886,18 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i if (!h->sps.chroma_format_idc) { for (i = 0; i < 8; i++) { uint16_t *tmp_cb = (uint16_t*)(dest_cb + i*uvlinesize); - for (j = 0; j < 8; j++) { - tmp_cb[j] = 1 << (bit_depth - 1); - } - } - for (i = 0; i < 8; i++) { uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize); for (j = 0; j < 8; j++) { - tmp_cr[j] = 1 << (bit_depth - 1); + tmp_cb[j] = tmp_cr[j] = 1 << (bit_depth - 1); } } } else { - for (i = 0; i < 8; i++) { + for (i = 0; i < block_h; i++) { uint16_t *tmp_cb = (uint16_t*)(dest_cb + i*uvlinesize); for (j = 0; j < 8; j++) tmp_cb[j] = get_bits(&gb, bit_depth); } - for (i = 0; i < 8; i++) { + for (i = 0; i < block_h; i++) { uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize); for (j = 0; j < 8; j++) tmp_cr[j] = get_bits(&gb, bit_depth); @@ -1860,14 +1910,14 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i } if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){ if (!h->sps.chroma_format_idc) { - for (i = 0; i < 8; i++) { - memset(dest_cb + i*uvlinesize, 128, 8); - memset(dest_cr + i*uvlinesize, 128, 8); + for (i=0; i<8; i++) { + memset(dest_cb+ i*uvlinesize, 1 << (bit_depth - 1), 8); + memset(dest_cr+ i*uvlinesize, 1 << (bit_depth - 1), 8); } } else { - for (i = 0; i < 8; i++) { - memcpy(dest_cb + i*uvlinesize, h->mb + 128 + i*4, 8); - memcpy(dest_cr + i*uvlinesize, h->mb + 160 + i*4, 8); + for (i=0; i<block_h; i++) { + memcpy(dest_cb+ i*uvlinesize, h->mb + 128 + i*4, 8); + memcpy(dest_cr+ i*uvlinesize, h->mb + 160 + i*4, 8); } } } @@ -1913,14 +1963,24 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i } }else{ if(is_h264){ + int qp[2]; + if (CHROMA422) { + qp[0] = h->chroma_qp[0]+3; + qp[1] = h->chroma_qp[1]+3; + } else { + qp[0] = h->chroma_qp[0]; + qp[1] = h->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), h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]); + h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*1 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][qp[0]][0]); if(h->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->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]); + h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*2 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][qp[1]][0]); h->h264dsp.h264_idct_add8(dest, block_offset, h->mb, uvlinesize, h->non_zero_count_cache); - }else{ + } +#if CONFIG_SVQ3_DECODER + else{ 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->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]); for(j=1; j<3; j++){ @@ -1932,6 +1992,7 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i } } } +#endif } } } @@ -2224,7 +2285,7 @@ static void idr(H264Context *h){ static void flush_dpb(AVCodecContext *avctx){ H264Context *h= avctx->priv_data; int i; - for(i=0; i<MAX_DELAYED_PIC_COUNT; i++) { + for(i=0; i<=MAX_DELAYED_PIC_COUNT; i++) { if(h->delayed_pic[i]) h->delayed_pic[i]->f.reference = 0; h->delayed_pic[i]= NULL; @@ -2237,6 +2298,8 @@ static void flush_dpb(AVCodecContext *avctx){ h->s.first_field= 0; ff_h264_reset_sei(h); ff_mpeg_flush(avctx); + h->recovery_frame= -1; + h->sync= 0; } static int init_poc(H264Context *h){ @@ -2555,22 +2618,27 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ h->b_stride= s->mb_width*4; + s->chroma_y_shift = h->sps.chroma_format_idc <= 1; // 400 uses yuv420p + s->width = 16*s->mb_width - (2>>CHROMA444)*FFMIN(h->sps.crop_right, (8<<CHROMA444)-1); if(h->sps.frame_mbs_only_flag) - s->height= 16*s->mb_height - (2>>CHROMA444)*FFMIN(h->sps.crop_bottom, (8<<CHROMA444)-1); + s->height= 16*s->mb_height - (1<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1); else - s->height= 16*s->mb_height - (4>>CHROMA444)*FFMIN(h->sps.crop_bottom, (8<<CHROMA444)-1); + s->height= 16*s->mb_height - (2<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1); if (s->context_initialized && ( s->width != s->avctx->width || s->height != s->avctx->height + || s->avctx->bits_per_raw_sample != h->sps.bit_depth_luma + || h->cur_chroma_format_idc != h->sps.chroma_format_idc || av_cmp_q(h->sps.sar, s->avctx->sample_aspect_ratio))) { if(h != h0) { - av_log_missing_feature(s->avctx, "Width/height changing with threads is", 0); + av_log_missing_feature(s->avctx, "Width/height/bit depth/chroma idc changing with threads is", 0); return -1; // width / height changed during parallelized decoding } free_tables(h, 0); flush_dpb(s->avctx); MPV_common_end(s); + h->list_count = 0; } if (!s->context_initialized) { if (h != h0) { @@ -2582,8 +2650,27 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ s->avctx->sample_aspect_ratio= h->sps.sar; av_assert0(s->avctx->sample_aspect_ratio.den); + if (s->avctx->bits_per_raw_sample != h->sps.bit_depth_luma || + h->cur_chroma_format_idc != h->sps.chroma_format_idc) { + if (h->sps.bit_depth_luma >= 8 && h->sps.bit_depth_luma <= 10 && + (h->sps.bit_depth_luma != 9 || !CHROMA422)) { + s->avctx->bits_per_raw_sample = h->sps.bit_depth_luma; + h->cur_chroma_format_idc = h->sps.chroma_format_idc; + h->pixel_shift = h->sps.bit_depth_luma > 8; + + ff_h264dsp_init(&h->h264dsp, h->sps.bit_depth_luma, h->sps.chroma_format_idc); + ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma, h->sps.chroma_format_idc); + s->dsp.dct_bits = h->sps.bit_depth_luma > 8 ? 32 : 16; + dsputil_init(&s->dsp, s->avctx); + } else { + av_log(s->avctx, AV_LOG_DEBUG, "Unsupported bit depth: %d chroma_idc: %d\n", + h->sps.bit_depth_luma, h->sps.chroma_format_idc); + return -1; + } + } + if(h->sps.video_signal_type_present_flag){ - s->avctx->color_range = h->sps.full_range ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG; + s->avctx->color_range = h->sps.full_range>0 ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG; if(h->sps.colour_description_present_flag){ s->avctx->color_primaries = h->sps.color_primaries; s->avctx->color_trc = h->sps.color_trc; @@ -2601,14 +2688,28 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ switch (h->sps.bit_depth_luma) { case 9 : - s->avctx->pix_fmt = CHROMA444 ? PIX_FMT_YUV444P9 : PIX_FMT_YUV420P9; + if (CHROMA444) + s->avctx->pix_fmt = PIX_FMT_YUV444P9; + else + s->avctx->pix_fmt = PIX_FMT_YUV420P9; break; case 10 : - s->avctx->pix_fmt = CHROMA444 ? PIX_FMT_YUV444P10 : PIX_FMT_YUV420P10; + if (CHROMA444) + s->avctx->pix_fmt = PIX_FMT_YUV444P10; + else if (CHROMA422) + s->avctx->pix_fmt = PIX_FMT_YUV422P10; + else + s->avctx->pix_fmt = PIX_FMT_YUV420P10; break; default: if (CHROMA444){ s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ444P : PIX_FMT_YUV444P; + if (s->avctx->colorspace == AVCOL_SPC_RGB) { + s->avctx->pix_fmt = PIX_FMT_GBR24P; + av_log(h->s.avctx, AV_LOG_DEBUG, "Detected GBR colorspace.\n"); + } + }else if (CHROMA422) { + s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ422P : PIX_FMT_YUV422P; }else{ s->avctx->pix_fmt = s->avctx->get_format(s->avctx, s->avctx->codec->pix_fmts ? @@ -2830,6 +2931,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ h->ref_count[1]= h->pps.ref_count[1]; if(h->slice_type_nos != AV_PICTURE_TYPE_I){ + unsigned max= (16<<(s->picture_structure != PICT_FRAME))-1; if(h->slice_type_nos == AV_PICTURE_TYPE_B){ h->direct_spatial_mv_pred= get_bits1(&s->gb); } @@ -2840,18 +2942,18 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ if(h->slice_type_nos==AV_PICTURE_TYPE_B) h->ref_count[1]= get_ue_golomb(&s->gb) + 1; - if(h->ref_count[0]-1 > 32-1 || h->ref_count[1]-1 > 32-1){ - av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow\n"); - h->ref_count[0]= h->ref_count[1]= 1; - return -1; - } + } + if(h->ref_count[0]-1 > max || h->ref_count[1]-1 > max){ + av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow\n"); + h->ref_count[0]= h->ref_count[1]= 1; + return -1; } if(h->slice_type_nos == AV_PICTURE_TYPE_B) h->list_count= 2; else h->list_count= 1; }else - h->list_count= 0; + h->ref_count[1]= h->ref_count[0]= h->list_count= 0; if(!default_ref_list_done){ ff_h264_fill_default_ref_list(h); @@ -2985,8 +3087,14 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ h0->last_slice_type = slice_type; h->slice_num = ++h0->current_slice; - if(h->slice_num >= MAX_SLICES){ - av_log(s->avctx, AV_LOG_ERROR, "Too many slices, increase MAX_SLICES and recompile\n"); + + if(h->slice_num) + h0->slice_row[(h->slice_num-1)&(MAX_SLICES-1)]= s->resync_mb_y; + if ( h0->slice_row[h->slice_num&(MAX_SLICES-1)] + 3 >= s->resync_mb_y + && h0->slice_row[h->slice_num&(MAX_SLICES-1)] <= s->resync_mb_y + && h->slice_num >= MAX_SLICES) { + //in case of ASO this check needs to be updated depending on how we decide to assign slice numbers in this case + av_log(s->avctx, AV_LOG_WARNING, "Possibly too many slices (%d >= %d), increase MAX_SLICES and recompile if there are artifacts\n", h->slice_num, MAX_SLICES); } for(j=0; j<2; j++){ @@ -3272,6 +3380,7 @@ static void loop_filter(H264Context *h, int start_x, int end_x){ const int end_mb_y= s->mb_y + FRAME_MBAFF; const int old_slice_type= h->slice_type; const int pixel_shift = h->pixel_shift; + const int block_h = 16>>s->chroma_y_shift; if(h->deblocking_filter) { for(mb_x= start_x; mb_x<end_x; mb_x++){ @@ -3288,8 +3397,8 @@ static void loop_filter(H264Context *h, int start_x, int end_x){ s->mb_x= mb_x; s->mb_y= mb_y; dest_y = s->current_picture.f.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16; - dest_cb = s->current_picture.f.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444); - dest_cr = s->current_picture.f.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444); + dest_cb = s->current_picture.f.data[1] + (mb_x << pixel_shift)*(8<<CHROMA444) + mb_y * s->uvlinesize * block_h; + dest_cr = s->current_picture.f.data[2] + (mb_x << pixel_shift)*(8<<CHROMA444) + mb_y * s->uvlinesize * block_h; //FIXME simplify above if (MB_FIELD) { @@ -3297,14 +3406,14 @@ static void loop_filter(H264Context *h, int start_x, int end_x){ uvlinesize = h->mb_uvlinesize = s->uvlinesize * 2; if(mb_y&1){ //FIXME move out of this function? dest_y -= s->linesize*15; - dest_cb-= s->uvlinesize*((8 << CHROMA444)-1); - dest_cr-= s->uvlinesize*((8 << CHROMA444)-1); + dest_cb-= s->uvlinesize*(block_h-1); + dest_cr-= s->uvlinesize*(block_h-1); } } else { linesize = h->mb_linesize = s->linesize; uvlinesize = h->mb_uvlinesize = s->uvlinesize; } - backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, CHROMA444, 0); + backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0); if(fill_filter_caches(h, mb_type)) continue; h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mb_xy]); @@ -3475,7 +3584,8 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg){ if(s->mb_y >= s->mb_height){ tprintf(s->avctx, "slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits); - if(get_bits_count(&s->gb) == s->gb.size_in_bits ) { + if( get_bits_count(&s->gb) == s->gb.size_in_bits + || get_bits_count(&s->gb) < s->gb.size_in_bits && s->avctx->error_recognition < FF_ER_AGGRESSIVE) { ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask); return 0; @@ -3525,6 +3635,7 @@ static int execute_decode_slices(H264Context *h, int context_count){ hx = h->thread_context[i]; hx->s.error_recognition = avctx->error_recognition; hx->s.error_count = 0; + hx->x264_build= h->x264_build; } avctx->execute(avctx, (void *)decode_slice, @@ -3612,13 +3723,13 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){ s->workaround_bugs |= FF_BUG_TRUNCATED; if(!(s->workaround_bugs & FF_BUG_TRUNCATED)){ - while(ptr[dst_length - 1] == 0 && dst_length > 0) + while(dst_length > 0 && ptr[dst_length - 1] == 0) dst_length--; } bit_length= !dst_length ? 0 : (8*dst_length - ff_h264_decode_rbsp_trailing(h, ptr + dst_length - 1)); if(s->avctx->debug&FF_DEBUG_STARTCODE){ - av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d at %d/%d length %d\n", hx->nal_unit_type, buf_index, buf_size, dst_length); + av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d/%d at %d/%d length %d\n", hx->nal_unit_type, hx->nal_ref_idc, buf_index, buf_size, dst_length); } if (h->is_avc && (nalsize != consumed) && nalsize){ @@ -3668,9 +3779,22 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){ if((err = decode_slice_header(hx, h))) break; + if (h->sei_recovery_frame_cnt >= 0 && h->recovery_frame < 0) { + h->recovery_frame = (h->frame_num + h->sei_recovery_frame_cnt) % + (1 << h->sps.log2_max_frame_num); + } + s->current_picture_ptr->f.key_frame |= - (hx->nal_unit_type == NAL_IDR_SLICE) || - (h->sei_recovery_frame_cnt >= 0); + (hx->nal_unit_type == NAL_IDR_SLICE); + + if (h->recovery_frame == h->frame_num) { + h->sync |= 1; + h->recovery_frame = -1; + } + + h->sync |= !!s->current_picture_ptr->f.key_frame; + h->sync |= 3*!!(s->flags2 & CODEC_FLAG2_SHOW_ALL); + s->current_picture_ptr->sync = h->sync; if (h->current_slice == 1) { if(!(s->flags2 & CODEC_FLAG2_CHUNKS)) { @@ -3733,7 +3857,11 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){ break; case NAL_SPS: init_get_bits(&s->gb, ptr, bit_length); - ff_h264_decode_seq_parameter_set(h); + if(ff_h264_decode_seq_parameter_set(h) < 0 && h->is_avc && (nalsize != consumed) && nalsize){ + av_log(h->s.avctx, AV_LOG_DEBUG, "SPS decoding failure, trying alternative mode\n"); + init_get_bits(&s->gb, &buf[buf_index + 1 - consumed], 8*nalsize); + ff_h264_decode_seq_parameter_set(h); + } if (s->flags& CODEC_FLAG_LOW_DELAY || (h->sps.bitstream_restriction_flag && !h->sps.num_reorder_frames)) @@ -3741,21 +3869,6 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){ if(avctx->has_b_frames < 2) avctx->has_b_frames= !s->low_delay; - - if (avctx->bits_per_raw_sample != h->sps.bit_depth_luma) { - if (h->sps.bit_depth_luma >= 8 && h->sps.bit_depth_luma <= 10) { - avctx->bits_per_raw_sample = h->sps.bit_depth_luma; - h->pixel_shift = h->sps.bit_depth_luma > 8; - - ff_h264dsp_init(&h->h264dsp, h->sps.bit_depth_luma); - ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma); - s->dsp.dct_bits = h->sps.bit_depth_luma > 8 ? 32 : 16; - dsputil_init(&s->dsp, s->avctx); - } else { - av_log(avctx, AV_LOG_ERROR, "Unsupported bit depth: %d\n", h->sps.bit_depth_luma); - return -1; - } - } break; case NAL_PPS: init_get_bits(&s->gb, ptr, bit_length); @@ -3849,6 +3962,8 @@ static int decode_frame(AVCodecContext *avctx, return 0; } + if(h->is_avc && buf_size >= 9 && AV_RB32(buf)==0x0164001F && buf[5] && buf[8]==0x67) + return ff_h264_decode_extradata(h, buf, buf_size); buf_index=decode_nal_units(h, buf, buf_size); if(buf_index < 0) @@ -3872,13 +3987,12 @@ static int decode_frame(AVCodecContext *avctx, field_end(h, 0); - if (!h->next_output_pic) { - /* Wait for second field. */ - *data_size = 0; - - } else { - *data_size = sizeof(AVFrame); - *pict = *(AVFrame*)h->next_output_pic; + *data_size = 0; /* Wait for second field. */ + if (h->next_output_pic && h->next_output_pic->sync) { + if(h->sync>1 || h->next_output_pic->f.pict_type != AV_PICTURE_TYPE_B){ + *data_size = sizeof(AVFrame); + *pict = *(AVFrame*)h->next_output_pic; + } } } @@ -3913,6 +4027,7 @@ static inline void fill_mb_avail(H264Context *h){ #undef random #define COUNT 8000 #define SIZE (COUNT*40) +extern AVCodec ff_h264_decoder; int main(void){ int i; uint8_t temp[SIZE]; @@ -3922,6 +4037,8 @@ int main(void){ DSPContext dsp; AVCodecContext avctx; + avcodec_get_context_defaults3(&avctx, &ff_h264_decoder); + dsputil_init(&dsp, &avctx); init_put_bits(&pb, temp, SIZE); |
