diff options
Diffstat (limited to 'libavcodec/h264.c')
-rw-r--r-- | libavcodec/h264.c | 452 |
1 files changed, 262 insertions, 190 deletions
diff --git a/libavcodec/h264.c b/libavcodec/h264.c index a80183b4a1..b9057ad122 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 */ @@ -25,7 +25,10 @@ * @author Michael Niedermayer <michaelni@gmx.at> */ +#define UNCHECKED_BITSTREAM_READER 1 + #include "libavutil/imgutils.h" +#include "libavutil/opt.h" #include "internal.h" #include "cabac.h" #include "cabac_functions.h" @@ -99,12 +102,8 @@ 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 -/** - * Check if the top & left blocks are available if needed and - * change the dc mode so it only uses the available blocks. - */ int ff_h264_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}; @@ -137,6 +136,7 @@ int ff_h264_check_intra_pred_mode(H264Context *h, int mode, int is_chroma){ return mode; } + 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; @@ -177,20 +177,28 @@ 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; @@ -709,7 +717,7 @@ prefetch_motion(H264Context *h, int list, int pixel_shift, int chroma_idc) 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); } } @@ -944,7 +952,7 @@ static void init_dequant_tables(H264Context *h){ int ff_h264_alloc_tables(H264Context *h){ MpegEncContext * const s = &h->s; const int big_mb_num= s->mb_stride * (s->mb_height+1); - const int row_mb_num= 2*s->mb_stride*s->avctx->thread_count; + const int row_mb_num= 2*s->mb_stride*FFMAX(s->avctx->thread_count, 1); int x,y; FF_ALLOCZ_OR_GOTO(h->s.avctx, h->intra4x4_pred_mode, row_mb_num * 8 * sizeof(uint8_t), fail) @@ -1031,29 +1039,38 @@ 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, 1); - ff_h264_pred_init(&h->hpc, s->codec_id, 8, 1); + 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->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; } @@ -1065,7 +1082,7 @@ 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); @@ -1077,7 +1094,7 @@ 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); @@ -1086,13 +1103,13 @@ 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; + return size; } av_cold int ff_h264_decode_init(AVCodecContext *avctx){ @@ -1126,6 +1143,7 @@ av_cold int ff_h264_decode_init(AVCodecContext *avctx){ for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) h->last_pocs[i] = INT_MIN; h->prev_poc_msb= 1<<16; + h->prev_frame_num= -1; h->x264_build = -1; ff_h264_reset_sei(h); if(avctx->codec_id == CODEC_ID_H264){ @@ -1136,7 +1154,7 @@ 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)<0) return -1; if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames < h->sps.num_reorder_frames){ @@ -1190,7 +1208,7 @@ static int decode_update_thread_context(AVCodecContext *dst, const AVCodecContex int inited = s->context_initialized, err; int i; - if(dst == src || !s1->context_initialized) return 0; + if(dst == src) return 0; err = ff_mpeg_update_thread_context(dst, src); if(err) return err; @@ -1206,12 +1224,19 @@ static int decode_update_thread_context(AVCodecContext *dst, const AVCodecContex memcpy(&h->s + 1, &h1->s + 1, sizeof(H264Context) - sizeof(MpegEncContext)); //copy all fields after MpegEnc memset(h->sps_buffers, 0, sizeof(h->sps_buffers)); memset(h->pps_buffers, 0, sizeof(h->pps_buffers)); + + if (s1->context_initialized) { if (ff_h264_alloc_tables(h) < 0) { av_log(dst, AV_LOG_ERROR, "Could not allocate memory for h264\n"); return AVERROR(ENOMEM); } context_init(h); + // frame_start may not be called for the next thread (if it's decoding a bottom field) + // so this has to be allocated here + h->s.obmc_scratchpad = av_malloc(16*6*s->linesize); + } + for(i=0; i<2; i++){ h->rbsp_buffer[i] = NULL; h->rbsp_buffer_size[i] = 0; @@ -1219,10 +1244,6 @@ static int decode_update_thread_context(AVCodecContext *dst, const AVCodecContex h->thread_context[0] = h; - // frame_start may not be called for the next thread (if it's decoding a bottom field) - // so this has to be allocated here - h->s.obmc_scratchpad = av_malloc(16*6*s->linesize); - s->dsp.clear_blocks(h->mb); s->dsp.clear_blocks(h->mb+(24*16<<h->pixel_shift)); } @@ -1261,6 +1282,7 @@ 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; @@ -1291,6 +1313,7 @@ int ff_h264_frame_start(H264Context *h){ * See decode_nal_units(). */ s->current_picture_ptr->f.key_frame = 0; + s->current_picture_ptr->sync = 0; s->current_picture_ptr->mmco_reset= 0; assert(s->linesize && s->uvlinesize); @@ -1349,7 +1372,6 @@ static void decode_postinit(H264Context *h, int setup_finished){ Picture *out = s->current_picture_ptr; Picture *cur = s->current_picture_ptr; int i, pics, out_of_order, out_idx; - int invalid = 0, cnt = 0; s->current_picture_ptr->f.qscale_type = FF_QSCALE_TYPE_H264; s->current_picture_ptr->f.pict_type = s->pict_type; @@ -1428,6 +1450,8 @@ static void decode_postinit(H264Context *h, int setup_finished){ } } + cur->mmco_reset = h->mmco_reset; + h->mmco_reset = 0; //FIXME do something with unavailable reference frames /* Sort B-frames into display order */ @@ -1444,100 +1468,64 @@ static void decode_postinit(H264Context *h, int setup_finished){ s->low_delay= 0; } + for (i = 0; 1; i++) { + if(i == MAX_DELAYED_PIC_COUNT || cur->poc < h->last_pocs[i]){ + if(i) + h->last_pocs[i-1] = cur->poc; + break; + } else if(i) { + h->last_pocs[i-1]= h->last_pocs[i]; + } + } + out_of_order = MAX_DELAYED_PIC_COUNT - i; + if( cur->f.pict_type == AV_PICTURE_TYPE_B + || (h->last_pocs[MAX_DELAYED_PIC_COUNT-2] > INT_MIN && h->last_pocs[MAX_DELAYED_PIC_COUNT-1] - h->last_pocs[MAX_DELAYED_PIC_COUNT-2] > 2)) + out_of_order = FFMAX(out_of_order, 1); + if(s->avctx->has_b_frames < out_of_order && !h->sps.bitstream_restriction_flag){ + av_log(s->avctx, AV_LOG_WARNING, "Increasing reorder buffer to %d\n", out_of_order); + s->avctx->has_b_frames = out_of_order; + s->low_delay = 0; + } + 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) cur->f.reference = DELAYED_PIC_REF; - /* Frame reordering. This code takes pictures from coding order and sorts - * them by their incremental POC value into display order. It supports POC - * gaps, MMCO reset codes and random resets. - * A "display group" can start either with a IDR frame (f.key_frame = 1), - * and/or can be closed down with a MMCO reset code. In sequences where - * there is no delay, we can't detect that (since the frame was already - * output to the user), so we also set h->mmco_reset to detect the MMCO - * reset code. - * FIXME: if we detect insufficient delays (as per s->avctx->has_b_frames), - * we increase the delay between input and output. All frames affected by - * the lag (e.g. those that should have been output before another frame - * that we already returned to the user) will be dropped. This is a bug - * that we will fix later. */ - for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) { - cnt += out->poc < h->last_pocs[i]; - invalid += out->poc == INT_MIN; - } - if (!h->mmco_reset && !cur->f.key_frame && cnt + invalid == MAX_DELAYED_PIC_COUNT && cnt > 0) { - h->mmco_reset = 2; - if (pics > 1) - h->delayed_pic[pics - 2]->mmco_reset = 2; - } - if (h->mmco_reset || cur->f.key_frame) { - for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) - h->last_pocs[i] = INT_MIN; - cnt = 0; - invalid = MAX_DELAYED_PIC_COUNT; - } out = h->delayed_pic[0]; out_idx = 0; - for (i = 1; i < MAX_DELAYED_PIC_COUNT && h->delayed_pic[i] && - !h->delayed_pic[i-1]->mmco_reset && !h->delayed_pic[i]->f.key_frame; i++) - { + for (i = 1; h->delayed_pic[i] && !h->delayed_pic[i]->f.key_frame && !h->delayed_pic[i]->mmco_reset; i++) if(h->delayed_pic[i]->poc < out->poc){ out = h->delayed_pic[i]; out_idx = i; } - } - if (s->avctx->has_b_frames == 0 && (h->delayed_pic[0]->f.key_frame || h->mmco_reset)) - h->next_outputed_poc = INT_MIN; - out_of_order = !out->f.key_frame && !h->mmco_reset && (out->poc < h->next_outputed_poc); + if (s->avctx->has_b_frames == 0 && (h->delayed_pic[0]->f.key_frame || h->delayed_pic[0]->mmco_reset)) + h->next_outputed_poc= INT_MIN; + out_of_order = out->poc < h->next_outputed_poc; - if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames >= h->sps.num_reorder_frames) - { } - else if (out_of_order && pics-1 == s->avctx->has_b_frames && - s->avctx->has_b_frames < MAX_DELAYED_PIC_COUNT) { - if (invalid + cnt < MAX_DELAYED_PIC_COUNT) { - s->avctx->has_b_frames = FFMAX(s->avctx->has_b_frames, cnt); - } - s->low_delay = 0; - } else if (s->low_delay && - ((h->next_outputed_poc != INT_MIN && out->poc > h->next_outputed_poc + 2) || - cur->f.pict_type == AV_PICTURE_TYPE_B)) { - s->low_delay = 0; - s->avctx->has_b_frames++; - } - - if(pics > s->avctx->has_b_frames){ + if(out_of_order || pics > s->avctx->has_b_frames){ out->f.reference &= ~DELAYED_PIC_REF; out->owner2 = s; // for frame threading, the owner must be the second field's thread // or else the first thread can release the picture and reuse it unsafely for(i=out_idx; h->delayed_pic[i]; i++) h->delayed_pic[i] = h->delayed_pic[i+1]; } - memmove(h->last_pocs, &h->last_pocs[1], sizeof(*h->last_pocs) * (MAX_DELAYED_PIC_COUNT - 1)); - h->last_pocs[MAX_DELAYED_PIC_COUNT - 1] = cur->poc; if(!out_of_order && pics > s->avctx->has_b_frames){ h->next_output_pic = out; - if (out->mmco_reset) { - if (out_idx > 0) { - h->next_outputed_poc = out->poc; - h->delayed_pic[out_idx - 1]->mmco_reset = out->mmco_reset; - } else { - h->next_outputed_poc = INT_MIN; - } - } else { - if (out_idx == 0 && pics > 1 && h->delayed_pic[0]->f.key_frame) { - h->next_outputed_poc = INT_MIN; - } else { - h->next_outputed_poc = out->poc; - } - } - h->mmco_reset = 0; + if (out_idx == 0 && h->delayed_pic[0] && (h->delayed_pic[0]->f.key_frame || h->delayed_pic[0]->mmco_reset)) { + h->next_outputed_poc = INT_MIN; + } else + h->next_outputed_poc = out->poc; }else{ - av_log(s->avctx, AV_LOG_DEBUG, "no picture\n"); + av_log(s->avctx, AV_LOG_DEBUG, "no picture %s\n", out_of_order ? "ooo" : ""); + } + + if (h->next_output_pic && h->next_output_pic->sync) { + h->sync |= 2; } if (setup_finished) @@ -1577,7 +1565,7 @@ static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, AV_COPY128(top_border+16, src_cb + 15*uvlinesize); AV_COPY128(top_border+32, src_cr + 15*uvlinesize); } - } else if(chroma422) { + } else if(chroma422){ if (pixel_shift) { AV_COPY128(top_border+32, src_cb + 15*uvlinesize); AV_COPY128(top_border+48, src_cr + 15*uvlinesize); @@ -1952,8 +1940,8 @@ 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); @@ -1967,14 +1955,9 @@ static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, i if (!h->sps.chroma_format_idc) { 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] = 1 << (bit_depth - 1); - } - } - 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] = 1 << (bit_depth - 1); + tmp_cb[j] = tmp_cr[j] = 1 << (bit_depth - 1); } } } else { @@ -1996,12 +1979,12 @@ 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 < block_h; 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 < block_h; i++) { + 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); } @@ -2373,32 +2356,36 @@ static void implicit_weight_table(H264Context *h, int field){ * instantaneous decoder refresh. */ static void idr(H264Context *h){ + int i; ff_h264_remove_all_refs(h); h->prev_frame_num= 0; h->prev_frame_num_offset= 0; - h->prev_poc_msb= + h->prev_poc_msb= 1<<16; h->prev_poc_lsb= 0; + for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) + h->last_pocs[i] = INT_MIN; } /* forget old pics after a seek */ 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; } - for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) - h->last_pocs[i] = INT_MIN; h->outputed_poc=h->next_outputed_poc= INT_MIN; h->prev_interlaced_frame = 1; idr(h); + h->prev_frame_num= -1; if(h->s.current_picture_ptr) h->s.current_picture_ptr->f.reference = 0; 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){ @@ -2653,7 +2640,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ s->me.qpel_avg= s->dsp.avg_h264_qpel_pixels_tab; } - first_mb_in_slice= get_ue_golomb(&s->gb); + first_mb_in_slice= get_ue_golomb_long(&s->gb); if(first_mb_in_slice == 0){ //FIXME better field boundary detection if(h0->current_slice && FIELD_PICTURE){ @@ -2719,35 +2706,55 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ 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 - (1<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1); - else - s->height= 16*s->mb_height - (2<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1); + s->width = 16*s->mb_width; + s->height= 16*s->mb_height; if (s->context_initialized - && ( s->width != s->avctx->width || s->height != s->avctx->height + && ( s->width != s->avctx->coded_width || s->height != s->avctx->coded_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); + if(h != h0 || (s->avctx->active_thread_type & FF_THREAD_FRAME)) { + 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) { av_log(h->s.avctx, AV_LOG_ERROR, "Cannot (re-)initialize context during parallel decoding.\n"); return -1; } - avcodec_set_dimensions(s->avctx, s->width, s->height); + s->avctx->width -= (2>>CHROMA444)*FFMIN(h->sps.crop_right, (8<<CHROMA444)-1); + s->avctx->height -= (1<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1) * (2 - h->sps.frame_mbs_only_flag); 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_ERROR, "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; @@ -2788,10 +2795,13 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ 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_GBRP; - } else - s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ444P : PIX_FMT_YUV444P; + s->avctx->pix_fmt = PIX_FMT_GBR24P; + av_log(h->s.avctx, AV_LOG_DEBUG, "Detected GBR colorspace.\n"); + } else if (s->avctx->colorspace == AVCOL_SPC_YCGCO) { + av_log(h->s.avctx, AV_LOG_WARNING, "Detected unsupported YCgCo colorspace.\n"); + } } else if (CHROMA422) { s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ422P : PIX_FMT_YUV422P; }else{ @@ -2834,6 +2844,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ c->sps = h->sps; c->pps = h->pps; c->pixel_shift = h->pixel_shift; + c->cur_chroma_format_idc = h->cur_chroma_format_idc; init_scan_tables(c); clone_tables(c, h, i); } @@ -2854,6 +2865,10 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ if(h->sps.frame_mbs_only_flag){ s->picture_structure= PICT_FRAME; }else{ + if(!h->sps.direct_8x8_inference_flag && slice_type == AV_PICTURE_TYPE_B){ + av_log(h->s.avctx, AV_LOG_ERROR, "This stream was generated by a broken encoder, invalid 8x8 inference\n"); + return -1; + } if(get_bits1(&s->gb)) { //field_pic_flag s->picture_structure= PICT_TOP_FIELD + get_bits1(&s->gb); //bottom_field_flag } else { @@ -2865,7 +2880,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ if(h0->current_slice == 0){ // Shorten frame num gaps so we don't have to allocate reference frames just to throw them away - if(h->frame_num != h->prev_frame_num) { + if(h->frame_num != h->prev_frame_num && h->prev_frame_num >= 0) { int unwrap_prev_frame_num = h->prev_frame_num, max_frame_num = 1<<h->sps.log2_max_frame_num; if (unwrap_prev_frame_num > h->frame_num) unwrap_prev_frame_num -= max_frame_num; @@ -2879,7 +2894,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ } } - while(h->frame_num != h->prev_frame_num && + while(h->frame_num != h->prev_frame_num && h->prev_frame_num >= 0 && h->frame_num != (h->prev_frame_num+1)%(1<<h->sps.log2_max_frame_num)){ Picture *prev = h->short_ref_count ? h->short_ref[0] : NULL; av_log(h->s.avctx, AV_LOG_DEBUG, "Frame num gap %d %d\n", h->frame_num, h->prev_frame_num); @@ -2927,11 +2942,11 @@ static int decode_slice_header(H264Context *h, H264Context *h0){ s0->first_field = FIELD_PICTURE; } else { - if (h->nal_ref_idc && - s0->current_picture_ptr->f.reference && - s0->current_picture_ptr->frame_num != h->frame_num) { + if (s0->current_picture_ptr->frame_num != h->frame_num) { + ff_thread_report_progress((AVFrame*)s0->current_picture_ptr, INT_MAX, + s0->picture_structure==PICT_BOTTOM_FIELD); /* - * This and previous field were reference, but had + * This and previous field had * different frame_nums. Consider this field first in * pair. Throw away previous field except for reference * purposes. @@ -3015,6 +3030,7 @@ 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); } @@ -3025,18 +3041,18 @@ 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); @@ -3170,8 +3186,14 @@ 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++){ @@ -3661,7 +3683,8 @@ 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->err_recognition & AV_EF_AGGRESSIVE)) { ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, ER_MB_END&part_mask); return 0; @@ -3711,6 +3734,7 @@ static int execute_decode_slices(H264Context *h, int context_count){ hx = h->thread_context[i]; hx->s.err_recognition = avctx->err_recognition; hx->s.error_count = 0; + hx->x264_build= h->x264_build; } avctx->execute(avctx, decode_slice, @@ -3741,7 +3765,12 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){ int nals_needed=0; ///< number of NALs that need decoding before the next frame thread starts int nal_index; + h->nal_unit_type= 0; + + if(!s->slice_context_count) + s->slice_context_count= 1; h->max_contexts = s->slice_context_count; + if(!(s->flags2 & CODEC_FLAG2_CHUNKS)){ h->current_slice = 0; if (!s->first_field) @@ -3798,19 +3827,16 @@ 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 pass %d\n", hx->nal_unit_type, hx->nal_ref_idc, buf_index, buf_size, dst_length, pass); } if (h->is_avc && (nalsize != consumed) && nalsize){ - // set trailing bits in the last partial byte to zero - if (bit_length & 7) - ptr[bit_length >> 3] = ptr[bit_length >> 3] & (0xff << 8 - (bit_length & 7)); av_log(h->s.avctx, AV_LOG_DEBUG, "AVC: Consumed only %d bytes instead of %d\n", consumed, nalsize); } @@ -3844,7 +3870,7 @@ static int decode_nal_units(H264Context *h, const uint8_t *buf, int buf_size){ switch(hx->nal_unit_type){ case NAL_IDR_SLICE: if (h->nal_unit_type != NAL_IDR_SLICE) { - av_log(h->s.avctx, AV_LOG_ERROR, "Invalid mix of idr and non-idr slices"); + av_log(h->s.avctx, AV_LOG_ERROR, "Invalid mix of idr and non-idr slices\n"); return -1; } idr(h); // FIXME ensure we don't lose some frames if there is reordering @@ -3857,9 +3883,24 @@ 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) & ((1 << h->sps.log2_max_frame_num)-1)) > h->sei_recovery_frame_cnt)) { + 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) { + s->current_picture_ptr->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)) { @@ -3922,7 +3963,12 @@ 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 : 1)){ + av_log(h->s.avctx, AV_LOG_DEBUG, "SPS decoding failure, trying alternative mode\n"); + if(h->is_avc) av_assert0(next_avc - buf_index + consumed == nalsize); + init_get_bits(&s->gb, &buf[buf_index + 1 - consumed], 8*(next_avc - buf_index + consumed)); + ff_h264_decode_seq_parameter_set(h); + } if (s->flags& CODEC_FLAG_LOW_DELAY || (h->sps.bitstream_restriction_flag && !h->sps.num_reorder_frames)) @@ -3930,23 +3976,6 @@ 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 || - h->cur_chroma_format_idc != h->sps.chroma_format_idc) { - if (h->sps.bit_depth_luma >= 8 && h->sps.bit_depth_luma <= 10) { - 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(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); @@ -4009,15 +4038,15 @@ static int decode_frame(AVCodecContext *avctx, MpegEncContext *s = &h->s; AVFrame *pict = data; int buf_index = 0; + Picture *out; + int i, out_idx; s->flags= avctx->flags; s->flags2= avctx->flags2; /* end of stream, output what is still in the buffers */ - out: if (buf_size == 0) { - Picture *out; - int i, out_idx; + out: s->current_picture_ptr = NULL; @@ -4040,19 +4069,42 @@ static int decode_frame(AVCodecContext *avctx, return buf_index; } + if(h->is_avc && buf_size >= 9 && buf[0]==1 && buf[2]==0 && (buf[4]&0xFC)==0xFC && (buf[5]&0x1F) && buf[8]==0x67){ + int cnt= buf[5]&0x1f; + uint8_t *p= buf+6; + while(cnt--){ + int nalsize= AV_RB16(p) + 2; + if(nalsize > buf_size - (p-buf) || p[2]!=0x67) + goto not_extra; + p += nalsize; + } + cnt = *(p++); + if(!cnt) + goto not_extra; + while(cnt--){ + int nalsize= AV_RB16(p) + 2; + if(nalsize > buf_size - (p-buf) || p[2]!=0x68) + goto not_extra; + p += nalsize; + } + + return ff_h264_decode_extradata(h, buf, buf_size); + } +not_extra: buf_index=decode_nal_units(h, buf, buf_size); if(buf_index < 0) return -1; if (!s->current_picture_ptr && h->nal_unit_type == NAL_END_SEQUENCE) { - buf_size = 0; + av_assert0(buf_index <= buf_size); goto out; } if(!(s->flags2 & CODEC_FLAG2_CHUNKS) && !s->current_picture_ptr){ - if (avctx->skip_frame >= AVDISCARD_NONREF) - return 0; + if (avctx->skip_frame >= AVDISCARD_NONREF || + buf_size >= 4 && !memcmp("Q264", buf, 4)) + return buf_size; av_log(avctx, AV_LOG_ERROR, "no frame!\n"); return -1; } @@ -4063,13 +4115,10 @@ 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 || h->sync>1)) { + *data_size = sizeof(AVFrame); + *pict = *(AVFrame*)h->next_output_pic; } } @@ -4117,6 +4166,7 @@ av_cold int ff_h264_decode_end(AVCodecContext *avctx) H264Context *h = avctx->priv_data; MpegEncContext *s = &h->s; + ff_h264_remove_all_refs(h); ff_h264_free_context(h); MPV_common_end(s); @@ -4143,6 +4193,26 @@ static const AVProfile profiles[] = { { FF_PROFILE_UNKNOWN }, }; +static const AVOption h264_options[] = { + {"is_avc", "is avc", offsetof(H264Context, is_avc), FF_OPT_TYPE_INT, {.dbl = 0}, 0, 1, 0}, + {"nal_length_size", "nal_length_size", offsetof(H264Context, nal_length_size), FF_OPT_TYPE_INT, {.dbl = 0}, 0, 4, 0}, + {NULL} +}; + +static const AVClass h264_class = { + "H264 Decoder", + av_default_item_name, + h264_options, + LIBAVUTIL_VERSION_INT, +}; + +static const AVClass h264_vdpau_class = { + "H264 VDPAU Decoder", + av_default_item_name, + h264_options, + LIBAVUTIL_VERSION_INT, +}; + AVCodec ff_h264_decoder = { .name = "h264", .type = AVMEDIA_TYPE_VIDEO, @@ -4158,6 +4228,7 @@ AVCodec ff_h264_decoder = { .init_thread_copy = ONLY_IF_THREADS_ENABLED(decode_init_thread_copy), .update_thread_context = ONLY_IF_THREADS_ENABLED(decode_update_thread_context), .profiles = NULL_IF_CONFIG_SMALL(profiles), + .priv_class = &h264_class, }; #if CONFIG_H264_VDPAU_DECODER @@ -4174,5 +4245,6 @@ AVCodec ff_h264_vdpau_decoder = { .long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (VDPAU acceleration)"), .pix_fmts = (const enum PixelFormat[]){PIX_FMT_VDPAU_H264, PIX_FMT_NONE}, .profiles = NULL_IF_CONFIG_SMALL(profiles), + .priv_class = &h264_vdpau_class, }; #endif |