X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavcodec%2Fh264.c;h=c07d10afa89bda43b23d7212f71457ffaf2bda0f;hb=0114c571d4c8cc1036850ced924683709390681a;hp=890dd223f34356df9dbf08d2c083cf439677b97c;hpb=0becb07842b57ea225ddf0726de33b5f8e669297;p=ffmpeg diff --git a/libavcodec/h264.c b/libavcodec/h264.c index 890dd223f34..c07d10afa89 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 * - * 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 */ +#define UNCHECKED_BITSTREAM_READER 1 + #include "libavutil/imgutils.h" +#include "libavutil/opt.h" #include "internal.h" #include "cabac.h" #include "cabac_functions.h" @@ -50,13 +53,17 @@ const uint16_t ff_h264_mb_sizes[4] = { 256, 384, 512, 768 }; static const uint8_t rem6[QP_MAX_NUM + 1] = { 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, - 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, + 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, + 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, + 0, 1, 2, 3, }; static const uint8_t div6[QP_MAX_NUM + 1] = { 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, - 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 10, 10, 10, 10, + 7, 7, 7, 7, 7, 7, 8, 8, 8, 8, 8, 8, 9, 9, 9, 9, 9, 9, 10, 10, 10, + 10,10,10,11,11,11,11,11,11,12,12,12,12,12,12,13,13,13, 13, 13, 13, + 14,14,14,14, }; static const enum PixelFormat hwaccel_pixfmt_list_h264_jpeg_420[] = { @@ -67,6 +74,12 @@ static const enum PixelFormat hwaccel_pixfmt_list_h264_jpeg_420[] = { PIX_FMT_NONE }; +int avpriv_h264_has_num_reorder_frames(AVCodecContext *avctx) +{ + H264Context *h = avctx->priv_data; + return h ? h->sps.num_reorder_frames : 0; +} + /** * Check if the top & left blocks are available if needed and * change the dc mode so it only uses the available blocks. @@ -175,58 +188,72 @@ const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, src++; length--; +#define STARTCODE_TEST \ + if (i + 2 < length && src[i + 1] == 0 && src[i + 2] <= 3) { \ + if (src[i + 2] != 3) { \ + /* startcode, so we must be past the end */ \ + length = i; \ + } \ + break; \ + } #if HAVE_FAST_UNALIGNED +#define FIND_FIRST_ZERO \ + if (i > 0 && !src[i]) \ + i--; \ + while (src[i]) \ + i++ #if HAVE_FAST_64BIT -#define RS 7 for (i = 0; i + 1 < length; i += 9) { if (!((~AV_RN64A(src + i) & (AV_RN64A(src + i) - 0x0100010001000101ULL)) & 0x8000800080008080ULL)) + continue; + FIND_FIRST_ZERO; + STARTCODE_TEST; + i -= 7; + } #else -#define RS 3 for (i = 0; i + 1 < length; i += 5) { if (!((~AV_RN32A(src + i) & (AV_RN32A(src + i) - 0x01000101U)) & 0x80008080U)) -#endif continue; - if (i > 0 && !src[i]) - i--; - while (src[i]) - i++; + FIND_FIRST_ZERO; + STARTCODE_TEST; + i -= 3; + } +#endif #else -#define RS 0 for (i = 0; i + 1 < length; i += 2) { if (src[i]) continue; if (i > 0 && src[i - 1] == 0) i--; -#endif - if (i + 2 < length && src[i + 1] == 0 && src[i + 2] <= 3) { - if (src[i + 2] != 3) { - /* startcode, so we must be past the end */ - length = i; - } - break; - } - i -= RS; - } - - if (i >= length - 1) { // no escaped 0 - *dst_length = length; - *consumed = length + 1; // +1 for the header - return src; + STARTCODE_TEST; } +#endif // use second escape buffer for inter data bufidx = h->nal_unit_type == NAL_DPC ? 1 : 0; - 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_padded_malloc(&h->rbsp_buffer[bufidx], &h->rbsp_buffer_size[bufidx], length+MAX_MBPAIR_SIZE); dst = h->rbsp_buffer[bufidx]; 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; @@ -714,33 +741,6 @@ static av_always_inline void mc_part_weighted(H264Context *h, int n, int square, } } -static av_always_inline void mc_part(H264Context *h, int n, int square, - int height, int delta, - uint8_t *dest_y, uint8_t *dest_cb, - uint8_t *dest_cr, - int x_offset, int y_offset, - qpel_mc_func *qpix_put, - h264_chroma_mc_func chroma_put, - qpel_mc_func *qpix_avg, - h264_chroma_mc_func chroma_avg, - h264_weight_func *weight_op, - h264_biweight_func *weight_avg, - int list0, int list1, - int pixel_shift, int chroma_idc) -{ - if ((h->use_weight == 2 && list0 && list1 && - (h->implicit_weight[h->ref_cache[0][scan8[n]]][h->ref_cache[1][scan8[n]]][h->s.mb_y & 1] != 32)) || - h->use_weight == 1) - mc_part_weighted(h, n, square, height, delta, dest_y, dest_cb, dest_cr, - x_offset, y_offset, qpix_put, chroma_put, - weight_op[0], weight_op[1], weight_avg[0], - weight_avg[1], list0, list1, pixel_shift, chroma_idc); - else - mc_part_std(h, n, square, height, delta, dest_y, dest_cb, dest_cr, - x_offset, y_offset, qpix_put, chroma_put, qpix_avg, - chroma_avg, list0, list1, pixel_shift, chroma_idc); -} - static av_always_inline void prefetch_motion(H264Context *h, int list, int pixel_shift, int chroma_idc) { @@ -760,154 +760,12 @@ static av_always_inline void prefetch_motion(H264Context *h, int list, 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)<>1) + (s->mb_x&7))*s->uvlinesize; s->dsp.prefetch(src[1] + off, src[2] - src[1], 2); } } } -static av_always_inline void hl_motion(H264Context *h, uint8_t *dest_y, - uint8_t *dest_cb, uint8_t *dest_cr, - qpel_mc_func(*qpix_put)[16], - h264_chroma_mc_func(*chroma_put), - qpel_mc_func(*qpix_avg)[16], - h264_chroma_mc_func(*chroma_avg), - h264_weight_func *weight_op, - h264_biweight_func *weight_avg, - int pixel_shift, int chroma_idc) -{ - MpegEncContext *const s = &h->s; - const int mb_xy = h->mb_xy; - const int mb_type = s->current_picture.f.mb_type[mb_xy]; - - assert(IS_INTER(mb_type)); - - if (HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME)) - await_references(h); - prefetch_motion(h, 0, pixel_shift, chroma_idc); - - if (IS_16X16(mb_type)) { - mc_part(h, 0, 1, 16, 0, dest_y, dest_cb, dest_cr, 0, 0, - qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0], - weight_op, weight_avg, - IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), - pixel_shift, chroma_idc); - } else if (IS_16X8(mb_type)) { - mc_part(h, 0, 0, 8, 8 << pixel_shift, dest_y, dest_cb, dest_cr, 0, 0, - qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0], - weight_op, weight_avg, - IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), - pixel_shift, chroma_idc); - mc_part(h, 8, 0, 8, 8 << pixel_shift, dest_y, dest_cb, dest_cr, 0, 4, - qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0], - weight_op, weight_avg, - IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), - pixel_shift, chroma_idc); - } else if (IS_8X16(mb_type)) { - mc_part(h, 0, 0, 16, 8 * h->mb_linesize, dest_y, dest_cb, dest_cr, 0, 0, - qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], - &weight_op[1], &weight_avg[1], - IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1), - pixel_shift, chroma_idc); - mc_part(h, 4, 0, 16, 8 * h->mb_linesize, dest_y, dest_cb, dest_cr, 4, 0, - qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], - &weight_op[1], &weight_avg[1], - IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1), - pixel_shift, chroma_idc); - } else { - int i; - - assert(IS_8X8(mb_type)); - - for (i = 0; i < 4; i++) { - const int sub_mb_type = h->sub_mb_type[i]; - const int n = 4 * i; - int x_offset = (i & 1) << 2; - int y_offset = (i & 2) << 1; - - if (IS_SUB_8X8(sub_mb_type)) { - mc_part(h, n, 1, 8, 0, dest_y, dest_cb, dest_cr, - x_offset, y_offset, - qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1], - &weight_op[1], &weight_avg[1], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - } else if (IS_SUB_8X4(sub_mb_type)) { - mc_part(h, n, 0, 4, 4 << pixel_shift, dest_y, dest_cb, dest_cr, - x_offset, y_offset, - qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1], - &weight_op[1], &weight_avg[1], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - mc_part(h, n + 2, 0, 4, 4 << pixel_shift, - dest_y, dest_cb, dest_cr, x_offset, y_offset + 2, - qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1], - &weight_op[1], &weight_avg[1], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - } else if (IS_SUB_4X8(sub_mb_type)) { - mc_part(h, n, 0, 8, 4 * h->mb_linesize, - dest_y, dest_cb, dest_cr, x_offset, y_offset, - qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], - &weight_op[2], &weight_avg[2], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - mc_part(h, n + 1, 0, 8, 4 * h->mb_linesize, - dest_y, dest_cb, dest_cr, x_offset + 2, y_offset, - qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], - &weight_op[2], &weight_avg[2], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - } else { - int j; - assert(IS_SUB_4X4(sub_mb_type)); - for (j = 0; j < 4; j++) { - int sub_x_offset = x_offset + 2 * (j & 1); - int sub_y_offset = y_offset + (j & 2); - mc_part(h, n + j, 1, 4, 0, - dest_y, dest_cb, dest_cr, sub_x_offset, sub_y_offset, - qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2], - &weight_op[2], &weight_avg[2], - IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1), - pixel_shift, chroma_idc); - } - } - } - } - - prefetch_motion(h, 1, pixel_shift, chroma_idc); -} - -static av_always_inline void hl_motion_420(H264Context *h, uint8_t *dest_y, - uint8_t *dest_cb, uint8_t *dest_cr, - qpel_mc_func(*qpix_put)[16], - h264_chroma_mc_func(*chroma_put), - qpel_mc_func(*qpix_avg)[16], - h264_chroma_mc_func(*chroma_avg), - h264_weight_func *weight_op, - h264_biweight_func *weight_avg, - int pixel_shift) -{ - hl_motion(h, dest_y, dest_cb, dest_cr, qpix_put, chroma_put, - qpix_avg, chroma_avg, weight_op, weight_avg, pixel_shift, 1); -} - -static av_always_inline void hl_motion_422(H264Context *h, uint8_t *dest_y, - uint8_t *dest_cb, uint8_t *dest_cr, - qpel_mc_func(*qpix_put)[16], - h264_chroma_mc_func(*chroma_put), - qpel_mc_func(*qpix_avg)[16], - h264_chroma_mc_func(*chroma_avg), - h264_weight_func *weight_op, - h264_biweight_func *weight_avg, - int pixel_shift) -{ - hl_motion(h, dest_y, dest_cb, dest_cr, qpix_put, chroma_put, - qpix_avg, chroma_avg, weight_op, weight_avg, pixel_shift, 2); -} - static void free_tables(H264Context *h, int free_rbsp) { int i; @@ -1019,7 +877,7 @@ 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 = s->mb_stride * 2 * 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, @@ -1126,12 +984,18 @@ 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; /* needed so that IDCT permutation is known early */ ff_dsputil_init(&s->dsp, s->avctx); @@ -1139,17 +1003,20 @@ static av_cold void common_init(H264Context *h) 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; } @@ -1161,7 +1028,7 @@ 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, @@ -1174,7 +1041,7 @@ 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, @@ -1184,13 +1051,13 @@ int ff_h264_decode_extradata(H264Context *h) p += nalsize; } // Now store right nal length size, that will be used 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) @@ -1225,17 +1092,20 @@ 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) { + if (avctx->codec_id == AV_CODEC_ID_H264) { if (avctx->ticks_per_frame == 1) s->avctx->time_base.den *= 2; avctx->ticks_per_frame = 2; } if (avctx->extradata_size > 0 && avctx->extradata && - ff_h264_decode_extradata(h)) + ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size) < 0) { + ff_h264_free_context(h); return -1; + } if (h->sps.bitstream_restriction_flag && s->avctx->has_b_frames < h->sps.num_reorder_frames) { @@ -1302,7 +1172,7 @@ static int decode_update_thread_context(AVCodecContext *dst, int inited = s->context_initialized, err; int i; - if (dst == src || !s1->context_initialized) + if (dst == src) return 0; err = ff_mpeg_update_thread_context(dst, src); @@ -1322,12 +1192,19 @@ static int decode_update_thread_context(AVCodecContext *dst, sizeof(H264Context) - sizeof(MpegEncContext)); 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; @@ -1335,10 +1212,6 @@ static int decode_update_thread_context(AVCodecContext *dst, 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)); } @@ -1382,6 +1255,7 @@ static int decode_update_thread_context(AVCodecContext *dst, 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; @@ -1414,6 +1288,7 @@ 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); @@ -1449,7 +1324,7 @@ int ff_h264_frame_start(H264Context *h) * SVQ3 as well as most other codecs have only last/next/current and thus * get released even with set reference, besides SVQ3 and others do not * mark frames as reference later "naturally". */ - if (s->codec_id != CODEC_ID_SVQ3) + if (s->codec_id != AV_CODEC_ID_SVQ3) s->current_picture_ptr->f.reference = 0; s->current_picture_ptr->field_poc[0] = @@ -1476,7 +1351,6 @@ 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; @@ -1560,6 +1434,8 @@ 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 */ @@ -1576,79 +1452,51 @@ 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_VERBOSE, "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 = 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; + 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->delayed_pic[0]->f.key_frame || h->delayed_pic[0]->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 (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++; - } + out_of_order = out->poc < h->next_outputed_poc; - if (pics > s->avctx->has_b_frames) { + if (out_of_order || pics > s->avctx->has_b_frames) { out->f.reference &= ~DELAYED_PIC_REF; // 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 @@ -1656,28 +1504,18 @@ static void decode_postinit(H264Context *h, int setup_finished) 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) @@ -1744,7 +1582,7 @@ static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, } top_border = h->top_borders[top_idx][s->mb_x]; - /* There are two lines saved, the line above the the top macroblock + /* There are two lines saved, the line above the top macroblock * of a pair, and the line above the bottom macroblock. */ AV_COPY128(top_border, src_y + 16 * linesize); if (pixel_shift) @@ -2077,373 +1915,17 @@ static av_always_inline void hl_decode_mb_idct_luma(H264Context *h, int mb_type, } } -static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, - int pixel_shift) -{ - MpegEncContext *const s = &h->s; - const int mb_x = s->mb_x; - const int mb_y = s->mb_y; - const int mb_xy = h->mb_xy; - const int mb_type = s->current_picture.f.mb_type[mb_xy]; - uint8_t *dest_y, *dest_cb, *dest_cr; - int linesize, uvlinesize /*dct_offset*/; - int i, j; - int *block_offset = &h->block_offset[0]; - const int transform_bypass = !simple && (s->qscale == 0 && h->sps.transform_bypass); - /* 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; - const int chroma422 = CHROMA422; - - 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) * 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); - - h->list_counts[mb_xy] = h->list_count; - - if (!simple && MB_FIELD) { - linesize = h->mb_linesize = s->linesize * 2; - uvlinesize = h->mb_uvlinesize = s->uvlinesize * 2; - 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 * (block_h - 1); - dest_cr -= s->uvlinesize * (block_h - 1); - } - if (FRAME_MBAFF) { - int list; - for (list = 0; list < h->list_count; list++) { - if (!USES_LIST(mb_type, list)) - continue; - if (IS_16X16(mb_type)) { - int8_t *ref = &h->ref_cache[list][scan8[0]]; - fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (s->mb_y & 1), 1); - } else { - for (i = 0; i < 16; i += 4) { - int ref = h->ref_cache[list][scan8[i]]; - if (ref >= 0) - fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, - 8, (16 + ref) ^ (s->mb_y & 1), 1); - } - } - } - } - } else { - linesize = h->mb_linesize = s->linesize; - uvlinesize = h->mb_uvlinesize = s->uvlinesize; - // dct_offset = s->linesize * 16; - } - - if (!simple && IS_INTRA_PCM(mb_type)) { - if (pixel_shift) { - const int bit_depth = h->sps.bit_depth_luma; - int j; - GetBitContext gb; - init_get_bits(&gb, (uint8_t *)h->mb, - ff_h264_mb_sizes[h->sps.chroma_format_idc] * bit_depth); - - for (i = 0; i < 16; i++) { - uint16_t *tmp_y = (uint16_t *)(dest_y + i * linesize); - for (j = 0; j < 16; j++) - tmp_y[j] = get_bits(&gb, bit_depth); - } - if (simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) { - 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); - } - } else { - 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 < 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); - } - } - } - } else { - for (i = 0; i < 16; i++) - memcpy(dest_y + i * linesize, (uint8_t *)h->mb + i * 16, 16); - 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); - } - } else { - uint8_t *src_cb = (uint8_t *)h->mb + 256; - uint8_t *src_cr = (uint8_t *)h->mb + 256 + block_h * 8; - for (i = 0; i < block_h; i++) { - memcpy(dest_cb + i * uvlinesize, src_cb + i * 8, 8); - memcpy(dest_cr + i * uvlinesize, src_cr + i * 8, 8); - } - } - } - } - } else { - if (IS_INTRA(mb_type)) { - if (h->deblocking_filter) - xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, - uvlinesize, 1, 0, simple, pixel_shift); - - if (simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) { - h->hpc.pred8x8[h->chroma_pred_mode](dest_cb, uvlinesize); - h->hpc.pred8x8[h->chroma_pred_mode](dest_cr, uvlinesize); - } - - hl_decode_mb_predict_luma(h, mb_type, is_h264, simple, - transform_bypass, pixel_shift, - block_offset, linesize, dest_y, 0); - - if (h->deblocking_filter) - xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, - uvlinesize, 0, 0, simple, pixel_shift); - } else if (is_h264) { - if (chroma422) { - hl_motion_422(h, dest_y, dest_cb, dest_cr, - s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, - s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, - h->h264dsp.weight_h264_pixels_tab, - h->h264dsp.biweight_h264_pixels_tab, - pixel_shift); - } else { - hl_motion_420(h, dest_y, dest_cb, dest_cr, - s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, - s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, - h->h264dsp.weight_h264_pixels_tab, - h->h264dsp.biweight_h264_pixels_tab, - pixel_shift); - } - } - - hl_decode_mb_idct_luma(h, mb_type, is_h264, simple, transform_bypass, - pixel_shift, block_offset, linesize, dest_y, 0); - - if ((simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) && - (h->cbp & 0x30)) { - uint8_t *dest[2] = { dest_cb, dest_cr }; - if (transform_bypass) { - if (IS_INTRA(mb_type) && h->sps.profile_idc == 244 && - (h->chroma_pred_mode == VERT_PRED8x8 || - h->chroma_pred_mode == HOR_PRED8x8)) { - h->hpc.pred8x8_add[h->chroma_pred_mode](dest[0], - block_offset + 16, - h->mb + (16 * 16 * 1 << pixel_shift), - uvlinesize); - h->hpc.pred8x8_add[h->chroma_pred_mode](dest[1], - block_offset + 32, - h->mb + (16 * 16 * 2 << pixel_shift), - uvlinesize); - } else { - idct_add = s->dsp.add_pixels4; - for (j = 1; j < 3; j++) { - for (i = j * 16; i < j * 16 + 4; i++) - if (h->non_zero_count_cache[scan8[i]] || - dctcoef_get(h->mb, pixel_shift, i * 16)) - idct_add(dest[j - 1] + block_offset[i], - h->mb + (i * 16 << pixel_shift), - uvlinesize); - if (chroma422) { - for (i = j * 16 + 4; i < j * 16 + 8; i++) - if (h->non_zero_count_cache[scan8[i + 4]] || - dctcoef_get(h->mb, pixel_shift, i * 16)) - idct_add(dest[j - 1] + block_offset[i + 4], - h->mb + (i * 16 << pixel_shift), - uvlinesize); - } - } - } - } 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][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][qp[1]][0]); - h->h264dsp.h264_idct_add8(dest, block_offset, - h->mb, uvlinesize, - 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->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++) { - 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][s->qscale + 12] - 12, 2); - } - } - } - } - } - } - if (h->cbp || IS_INTRA(mb_type)) { - s->dsp.clear_blocks(h->mb); - s->dsp.clear_blocks(h->mb + (24 * 16 << pixel_shift)); - } -} - -static av_always_inline void hl_decode_mb_444_internal(H264Context *h, - int simple, - int pixel_shift) -{ - MpegEncContext *const s = &h->s; - const int mb_x = s->mb_x; - const int mb_y = s->mb_y; - const int mb_xy = h->mb_xy; - const int mb_type = s->current_picture.f.mb_type[mb_xy]; - uint8_t *dest[3]; - int linesize; - int i, j, p; - int *block_offset = &h->block_offset[0]; - const int transform_bypass = !simple && (s->qscale == 0 && h->sps.transform_bypass); - const int plane_count = (simple || !CONFIG_GRAY || !(s->flags & CODEC_FLAG_GRAY)) ? 3 : 1; - - for (p = 0; p < plane_count; p++) { - dest[p] = s->current_picture.f.data[p] + - ((mb_x << pixel_shift) + mb_y * s->linesize) * 16; - s->dsp.prefetch(dest[p] + (s->mb_x & 3) * 4 * s->linesize + (64 << pixel_shift), - s->linesize, 4); - } +#define BITS 8 +#define SIMPLE 1 +#include "h264_mb_template.c" - h->list_counts[mb_xy] = h->list_count; - - if (!simple && MB_FIELD) { - linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize * 2; - block_offset = &h->block_offset[48]; - if (mb_y & 1) // FIXME move out of this function? - for (p = 0; p < 3; p++) - dest[p] -= s->linesize * 15; - if (FRAME_MBAFF) { - int list; - for (list = 0; list < h->list_count; list++) { - if (!USES_LIST(mb_type, list)) - continue; - if (IS_16X16(mb_type)) { - int8_t *ref = &h->ref_cache[list][scan8[0]]; - fill_rectangle(ref, 4, 4, 8, (16 + *ref) ^ (s->mb_y & 1), 1); - } else { - for (i = 0; i < 16; i += 4) { - int ref = h->ref_cache[list][scan8[i]]; - if (ref >= 0) - fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, - 8, (16 + ref) ^ (s->mb_y & 1), 1); - } - } - } - } - } else { - linesize = h->mb_linesize = h->mb_uvlinesize = s->linesize; - } +#undef BITS +#define BITS 16 +#include "h264_mb_template.c" - if (!simple && IS_INTRA_PCM(mb_type)) { - if (pixel_shift) { - const int bit_depth = h->sps.bit_depth_luma; - GetBitContext gb; - init_get_bits(&gb, (uint8_t *)h->mb, 768 * bit_depth); - - for (p = 0; p < plane_count; p++) - for (i = 0; i < 16; i++) { - uint16_t *tmp = (uint16_t *)(dest[p] + i * linesize); - for (j = 0; j < 16; j++) - tmp[j] = get_bits(&gb, bit_depth); - } - } else { - for (p = 0; p < plane_count; p++) - for (i = 0; i < 16; i++) - memcpy(dest[p] + i * linesize, - (uint8_t *)h->mb + p * 256 + i * 16, 16); - } - } else { - if (IS_INTRA(mb_type)) { - if (h->deblocking_filter) - xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, - linesize, 1, 1, simple, pixel_shift); - - for (p = 0; p < plane_count; p++) - hl_decode_mb_predict_luma(h, mb_type, 1, simple, - transform_bypass, pixel_shift, - block_offset, linesize, dest[p], p); - - if (h->deblocking_filter) - xchg_mb_border(h, dest[0], dest[1], dest[2], linesize, - linesize, 0, 1, simple, pixel_shift); - } else { - hl_motion(h, dest[0], dest[1], dest[2], - s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab, - s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab, - h->h264dsp.weight_h264_pixels_tab, - h->h264dsp.biweight_h264_pixels_tab, pixel_shift, 3); - } - - for (p = 0; p < plane_count; p++) - hl_decode_mb_idct_luma(h, mb_type, 1, simple, transform_bypass, - pixel_shift, block_offset, linesize, - dest[p], p); - } - if (h->cbp || IS_INTRA(mb_type)) { - s->dsp.clear_blocks(h->mb); - s->dsp.clear_blocks(h->mb + (24 * 16 << pixel_shift)); - } -} - -/** - * Process a macroblock; this case avoids checks for expensive uncommon cases. - */ -#define hl_decode_mb_simple(sh, bits) \ -static void hl_decode_mb_simple_ ## bits(H264Context *h) \ -{ \ - hl_decode_mb_internal(h, 1, sh); \ -} - -hl_decode_mb_simple(0, 8) -hl_decode_mb_simple(1, 16) - -/** - * Process a macroblock; this handles edge cases, such as interlacing. - */ -static av_noinline void hl_decode_mb_complex(H264Context *h) -{ - hl_decode_mb_internal(h, 0, h->pixel_shift); -} - -static av_noinline void hl_decode_mb_444_complex(H264Context *h) -{ - hl_decode_mb_444_internal(h, 0, h->pixel_shift); -} - -static av_noinline void hl_decode_mb_444_simple(H264Context *h) -{ - hl_decode_mb_444_internal(h, 1, 0); -} +#undef SIMPLE +#define SIMPLE 0 +#include "h264_mb_template.c" void ff_h264_hl_decode_mb(H264Context *h) { @@ -2456,7 +1938,7 @@ void ff_h264_hl_decode_mb(H264Context *h) if (is_complex || h->pixel_shift) hl_decode_mb_444_complex(h); else - hl_decode_mb_444_simple(h); + hl_decode_mb_444_simple_8(h); } else if (is_complex) { hl_decode_mb_complex(h); } else if (h->pixel_shift) { @@ -2600,11 +2082,14 @@ static void implicit_weight_table(H264Context *h, int field) */ 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 */ @@ -2612,21 +2097,22 @@ 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) @@ -2729,19 +2215,19 @@ static void init_scan_tables(H264Context *h) #undef T } if (h->sps.transform_bypass) { // FIXME same ugly - h->zigzag_scan_q0 = zigzag_scan; - h->zigzag_scan8x8_q0 = ff_zigzag_direct; - h->zigzag_scan8x8_cavlc_q0 = zigzag_scan8x8_cavlc; - h->field_scan_q0 = field_scan; - h->field_scan8x8_q0 = field_scan8x8; - h->field_scan8x8_cavlc_q0 = field_scan8x8_cavlc; + memcpy(h->zigzag_scan_q0 , zigzag_scan , sizeof(h->zigzag_scan_q0 )); + memcpy(h->zigzag_scan8x8_q0 , ff_zigzag_direct , sizeof(h->zigzag_scan8x8_q0 )); + memcpy(h->zigzag_scan8x8_cavlc_q0 , zigzag_scan8x8_cavlc , sizeof(h->zigzag_scan8x8_cavlc_q0)); + memcpy(h->field_scan_q0 , field_scan , sizeof(h->field_scan_q0 )); + memcpy(h->field_scan8x8_q0 , field_scan8x8 , sizeof(h->field_scan8x8_q0 )); + memcpy(h->field_scan8x8_cavlc_q0 , field_scan8x8_cavlc , sizeof(h->field_scan8x8_cavlc_q0 )); } else { - h->zigzag_scan_q0 = h->zigzag_scan; - h->zigzag_scan8x8_q0 = h->zigzag_scan8x8; - h->zigzag_scan8x8_cavlc_q0 = h->zigzag_scan8x8_cavlc; - h->field_scan_q0 = h->field_scan; - h->field_scan8x8_q0 = h->field_scan8x8; - h->field_scan8x8_cavlc_q0 = h->field_scan8x8_cavlc; + memcpy(h->zigzag_scan_q0 , h->zigzag_scan , sizeof(h->zigzag_scan_q0 )); + memcpy(h->zigzag_scan8x8_q0 , h->zigzag_scan8x8 , sizeof(h->zigzag_scan8x8_q0 )); + memcpy(h->zigzag_scan8x8_cavlc_q0 , h->zigzag_scan8x8_cavlc , sizeof(h->zigzag_scan8x8_cavlc_q0)); + memcpy(h->field_scan_q0 , h->field_scan , sizeof(h->field_scan_q0 )); + memcpy(h->field_scan8x8_q0 , h->field_scan8x8 , sizeof(h->field_scan8x8_q0 )); + memcpy(h->field_scan8x8_cavlc_q0 , h->field_scan8x8_cavlc , sizeof(h->field_scan8x8_cavlc_q0 )); } } @@ -2877,6 +2363,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0) unsigned int slice_type, tmp, i, j; int default_ref_list_done = 0; int last_pic_structure, last_pic_dropable; + int must_reinit; /* FIXME: 2tap qpel isn't implemented for high bit depth. */ if ((s->avctx->flags2 & CODEC_FLAG2_FAST) && @@ -2888,7 +2375,7 @@ 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) { @@ -2932,7 +2419,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0) pps_id = get_ue_golomb(&s->gb); if (pps_id >= MAX_PPS_COUNT) { - av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n"); + av_log(h->s.avctx, AV_LOG_ERROR, "pps_id %d out of range\n", pps_id); return -1; } if (!h0->pps_buffers[pps_id]) { @@ -2955,6 +2442,19 @@ static int decode_slice_header(H264Context *h, H264Context *h0) s->avctx->level = h->sps.level_idc; s->avctx->refs = h->sps.ref_frame_count; + must_reinit = (s->context_initialized && + ( 16*h->sps.mb_width != s->avctx->coded_width + || 16*h->sps.mb_height * (2 - h->sps.frame_mbs_only_flag) != 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(must_reinit && (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 AVERROR_PATCHWELCOME; // width / height changed during parallelized decoding + } + s->mb_width = h->sps.mb_width; s->mb_height = h->sps.mb_height * (2 - h->sps.frame_mbs_only_flag); @@ -2962,23 +2462,15 @@ 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 || - 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); - return -1; // width / height changed during parallelized decoding - } + if(must_reinit) { free_tables(h, 0); flush_dpb(s->avctx); ff_MPV_common_end(s); + h->list_count = 0; + h->current_slice = 0; } if (!s->context_initialized) { if (h != h0) { @@ -2986,13 +2478,43 @@ static int decode_slice_header(H264Context *h, H264Context *h0) "Cannot (re-)initialize context during parallel decoding.\n"); return -1; } - - avcodec_set_dimensions(s->avctx, s->width, s->height); + if( FFALIGN(s->avctx->width , 16 ) == s->width + && FFALIGN(s->avctx->height, 16*(2 - h->sps.frame_mbs_only_flag)) == s->height + && !h->sps.crop_right && !h->sps.crop_bottom + && (s->avctx->width != s->width || s->avctx->height && s->height) + ) { + av_log(h->s.avctx, AV_LOG_DEBUG, "Using externally provided dimensions\n"); + s->avctx->coded_width = s->width; + s->avctx->coded_height = s->height; + } else{ + avcodec_set_dimensions(s->avctx, s->width, s->height); + s->avctx->width -= (2>>CHROMA444)*FFMIN(h->sps.crop_right, (8<avctx->height -= (1<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 <= 14 && h->sps.bit_depth_luma != 11 && h->sps.bit_depth_luma != 13 && + (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; + ff_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 + 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; @@ -3032,13 +2554,38 @@ static int decode_slice_header(H264Context *h, H264Context *h0) else s->avctx->pix_fmt = PIX_FMT_YUV420P10; break; - case 8: + case 12: + if (CHROMA444) { + if (s->avctx->colorspace == AVCOL_SPC_RGB) { + s->avctx->pix_fmt = PIX_FMT_GBRP12; + } else + s->avctx->pix_fmt = PIX_FMT_YUV444P12; + } else if (CHROMA422) + s->avctx->pix_fmt = PIX_FMT_YUV422P12; + else + s->avctx->pix_fmt = PIX_FMT_YUV420P12; + break; + case 14: if (CHROMA444) { if (s->avctx->colorspace == AVCOL_SPC_RGB) { - s->avctx->pix_fmt = PIX_FMT_GBRP; + s->avctx->pix_fmt = PIX_FMT_GBRP14; } else + s->avctx->pix_fmt = PIX_FMT_YUV444P14; + } else if (CHROMA422) + s->avctx->pix_fmt = PIX_FMT_YUV422P14; + else + s->avctx->pix_fmt = PIX_FMT_YUV420P14; + break; + case 8: + 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 (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; @@ -3089,6 +2636,7 @@ 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); } @@ -3117,6 +2665,10 @@ 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 { @@ -3139,7 +2691,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0) } else { /* 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; int max_frame_num = 1 << h->sps.log2_max_frame_num; @@ -3221,7 +2773,7 @@ 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", @@ -3270,6 +2822,8 @@ static int decode_slice_header(H264Context *h, H264Context *h0) s0->first_field = FIELD_PICTURE; } else { 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 the previous field had different frame_nums. * Consider this field first in pair. Throw away previous * one except for reference purposes. */ @@ -3282,7 +2836,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0) } } else { /* Frame or first field in a potentially complementary pair */ - assert(!s0->current_picture_ptr); +// assert(!s0->current_picture_ptr); s0->first_field = FIELD_PICTURE; } @@ -3347,7 +2901,8 @@ 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) { - int max_refs = s->picture_structure == PICT_FRAME ? 16 : 32; + unsigned max[2]; + max[0] = max[1] = s->picture_structure == PICT_FRAME ? 15 : 31; if (h->slice_type_nos == AV_PICTURE_TYPE_B) h->direct_spatial_mv_pred = get_bits1(&s->gb); @@ -3357,10 +2912,13 @@ static int decode_slice_header(H264Context *h, H264Context *h0) h->ref_count[0] = get_ue_golomb(&s->gb) + 1; if (h->slice_type_nos == AV_PICTURE_TYPE_B) h->ref_count[1] = get_ue_golomb(&s->gb) + 1; + else + // full range is spec-ok in this case, even for frames + max[1] = 31; } - if (h->ref_count[0] > max_refs || h->ref_count[1] > max_refs) { - av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow\n"); + if (h->ref_count[0]-1 > max[0] || h->ref_count[1]-1 > max[1]){ + av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow %u > %u or %u > %u\n", h->ref_count[0]-1, max[0], h->ref_count[1]-1, max[1]); h->ref_count[0] = h->ref_count[1] = 1; return AVERROR_INVALIDDATA; } @@ -3370,7 +2928,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0) 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); @@ -3511,9 +3069,14 @@ 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++) { @@ -3613,7 +3176,7 @@ static av_always_inline void fill_filter_caches_inter(H264Context *h, if (USES_LIST(top_type, list)) { const int b_xy = h->mb2b_xy[top_xy] + 3 * b_stride; const int b8_xy = 4 * top_xy + 2; - int (*ref2frm)[64] = h->ref2frm[h->slice_table[top_xy] & (MAX_SLICES - 1)][0] + (MB_MBAFF ? 20 : 2); + int (*ref2frm)[64] = (void*)(h->ref2frm[h->slice_table[top_xy] & (MAX_SLICES - 1)][0] + (MB_MBAFF ? 20 : 2)); AV_COPY128(mv_dst - 1 * 8, s->current_picture.f.motion_val[list][b_xy + 0]); ref_cache[0 - 1 * 8] = ref_cache[1 - 1 * 8] = ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 0]]; @@ -3628,7 +3191,7 @@ static av_always_inline void fill_filter_caches_inter(H264Context *h, if (USES_LIST(left_type[LTOP], list)) { const int b_xy = h->mb2b_xy[left_xy[LTOP]] + 3; const int b8_xy = 4 * left_xy[LTOP] + 1; - int (*ref2frm)[64] = h->ref2frm[h->slice_table[left_xy[LTOP]] & (MAX_SLICES - 1)][0] + (MB_MBAFF ? 20 : 2); + int (*ref2frm)[64] =(void*)( h->ref2frm[h->slice_table[left_xy[LTOP]] & (MAX_SLICES - 1)][0] + (MB_MBAFF ? 20 : 2)); AV_COPY32(mv_dst - 1 + 0, s->current_picture.f.motion_val[list][b_xy + b_stride * 0]); AV_COPY32(mv_dst - 1 + 8, s->current_picture.f.motion_val[list][b_xy + b_stride * 1]); AV_COPY32(mv_dst - 1 + 16, s->current_picture.f.motion_val[list][b_xy + b_stride * 2]); @@ -3661,7 +3224,7 @@ static av_always_inline void fill_filter_caches_inter(H264Context *h, { int8_t *ref = &s->current_picture.f.ref_index[list][4 * mb_xy]; - int (*ref2frm)[64] = h->ref2frm[h->slice_num & (MAX_SLICES - 1)][0] + (MB_MBAFF ? 20 : 2); + int (*ref2frm)[64] = (void*)(h->ref2frm[h->slice_num & (MAX_SLICES - 1)][0] + (MB_MBAFF ? 20 : 2)); uint32_t ref01 = (pack16to32(ref2frm[list][ref[0]], ref2frm[list][ref[1]]) & 0x00FF00FF) * 0x0101; uint32_t ref23 = (pack16to32(ref2frm[list][ref[2]], ref2frm[list][ref[3]]) & 0x00FF00FF) * 0x0101; AV_WN32A(&ref_cache[0 * 8], ref01); @@ -3956,7 +3519,7 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) s->mb_skip_run = -1; h->is_complex = FRAME_MBAFF || s->picture_structure != PICT_FRAME || - s->codec_id != CODEC_ID_H264 || + s->codec_id != AV_CODEC_ID_H264 || (CONFIG_GRAY && (s->flags & CODEC_FLAG_GRAY)); if (h->pps.cabac) { @@ -4000,7 +3563,9 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) loop_filter(h, lf_x_start, s->mb_x + 1); return 0; } - if (ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 2) { + if (h->cabac.bytestream > h->cabac.bytestream_end + 2 ) + av_log(h->s.avctx, AV_LOG_DEBUG, "bytestream overread %td\n", h->cabac.bytestream_end - h->cabac.bytestream); + if (ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 4) { av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d, bytestream (%td)\n", s->mb_x, s->mb_y, @@ -4071,7 +3636,8 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) tprintf(s->avctx, "slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits); - if (get_bits_left(&s->gb) == 0) { + if ( get_bits_left(&s->gb) == 0 + || get_bits_left(&s->gb) > 0 && !(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); @@ -4132,6 +3698,7 @@ 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, h->thread_context, @@ -4162,6 +3729,10 @@ 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; @@ -4227,16 +3798,14 @@ 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 - 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) av_log(h->s.avctx, AV_LOG_DEBUG, @@ -4275,7 +3844,7 @@ again: 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"); + "Invalid mix of idr and non-idr slices\n"); buf_index = -1; goto end; } @@ -4289,9 +3858,24 @@ again: 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)) @@ -4348,6 +3932,9 @@ again: init_get_bits(&hx->inter_gb, ptr, bit_length); hx->inter_gb_ptr = &hx->inter_gb; + av_log(h->s.avctx, AV_LOG_ERROR, "Partitioned H.264 support is incomplete\n"); + return AVERROR_PATCHWELCOME; + if (hx->redundant_pic_count == 0 && hx->intra_gb_ptr && hx->s.data_partitioning && @@ -4366,12 +3953,13 @@ again: break; case NAL_SPS: init_get_bits(&s->gb, ptr, bit_length); - if (ff_h264_decode_seq_parameter_set(h) < 0 && - h->is_avc && (nalsize != consumed) && nalsize) { + 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, try parsing the coomplete NAL\n"); - init_get_bits(&s->gb, buf + buf_index + 1 - consumed, - 8 * (nalsize - 1)); + "SPS decoding failure, trying again with the complete NAL\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 - 1)); ff_h264_decode_seq_parameter_set(h); } @@ -4379,32 +3967,8 @@ again: (h->sps.bitstream_restriction_flag && !h->sps.num_reorder_frames)) s->low_delay = 1; - 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; - ff_dsputil_init(&s->dsp, s->avctx); - } else { - av_log(avctx, AV_LOG_ERROR, - "Unsupported bit depth: %d\n", - h->sps.bit_depth_luma); - buf_index = -1; - goto end; - } - } break; case NAL_PPS: init_get_bits(&s->gb, ptr, bit_length); @@ -4477,15 +4041,15 @@ static int decode_frame(AVCodecContext *avctx, void *data, 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; @@ -4512,19 +4076,42 @@ out: 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; + const 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; } @@ -4536,10 +4123,9 @@ out: field_end(h, 0); - if (!h->next_output_pic) { - /* Wait for second field. */ - *data_size = 0; - } else { + /* Wait for second field. */ + *data_size = 0; + if (h->next_output_pic && (h->next_output_pic->sync || h->sync>1)) { *data_size = sizeof(AVFrame); *pict = h->next_output_pic->f; } @@ -4570,6 +4156,7 @@ static av_cold int 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); ff_MPV_common_end(s); @@ -4596,10 +4183,30 @@ 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, - .id = CODEC_ID_H264, + .id = AV_CODEC_ID_H264, .priv_data_size = sizeof(H264Context), .init = ff_h264_decode_init, .close = h264_decode_end, @@ -4612,13 +4219,14 @@ 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 AVCodec ff_h264_vdpau_decoder = { .name = "h264_vdpau", .type = AVMEDIA_TYPE_VIDEO, - .id = CODEC_ID_H264, + .id = AV_CODEC_ID_H264, .priv_data_size = sizeof(H264Context), .init = ff_h264_decode_init, .close = h264_decode_end, @@ -4629,5 +4237,6 @@ AVCodec ff_h264_vdpau_decoder = { .pix_fmts = (const enum PixelFormat[]) { PIX_FMT_VDPAU_H264, PIX_FMT_NONE}, .profiles = NULL_IF_CONFIG_SMALL(profiles), + .priv_class = &h264_vdpau_class, }; #endif