diff options
Diffstat (limited to 'libavcodec/h264_slice.c')
-rw-r--r-- | libavcodec/h264_slice.c | 627 |
1 files changed, 458 insertions, 169 deletions
diff --git a/libavcodec/h264_slice.c b/libavcodec/h264_slice.c index 031f8cc7d8..522a36cb89 100644 --- a/libavcodec/h264_slice.c +++ b/libavcodec/h264_slice.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 */ @@ -43,14 +43,14 @@ #include "rectangle.h" #include "thread.h" -static const uint8_t field_scan[16] = { +static const uint8_t field_scan[16+1] = { 0 + 0 * 4, 0 + 1 * 4, 1 + 0 * 4, 0 + 2 * 4, 0 + 3 * 4, 1 + 1 * 4, 1 + 2 * 4, 1 + 3 * 4, 2 + 0 * 4, 2 + 1 * 4, 2 + 2 * 4, 2 + 3 * 4, 3 + 0 * 4, 3 + 1 * 4, 3 + 2 * 4, 3 + 3 * 4, }; -static const uint8_t field_scan8x8[64] = { +static const uint8_t field_scan8x8[64+1] = { 0 + 0 * 8, 0 + 1 * 8, 0 + 2 * 8, 1 + 0 * 8, 1 + 1 * 8, 0 + 3 * 8, 0 + 4 * 8, 1 + 2 * 8, 2 + 0 * 8, 1 + 3 * 8, 0 + 5 * 8, 0 + 6 * 8, @@ -69,7 +69,7 @@ static const uint8_t field_scan8x8[64] = { 7 + 4 * 8, 7 + 5 * 8, 7 + 6 * 8, 7 + 7 * 8, }; -static const uint8_t field_scan8x8_cavlc[64] = { +static const uint8_t field_scan8x8_cavlc[64+1] = { 0 + 0 * 8, 1 + 1 * 8, 2 + 0 * 8, 0 + 7 * 8, 2 + 2 * 8, 2 + 3 * 8, 2 + 4 * 8, 3 + 3 * 8, 3 + 4 * 8, 4 + 3 * 8, 4 + 4 * 8, 5 + 3 * 8, @@ -89,7 +89,7 @@ static const uint8_t field_scan8x8_cavlc[64] = { }; // zigzag_scan8x8_cavlc[i] = zigzag_scan8x8[(i/4) + 16*(i%4)] -static const uint8_t zigzag_scan8x8_cavlc[64] = { +static const uint8_t zigzag_scan8x8_cavlc[64+1] = { 0 + 0 * 8, 1 + 1 * 8, 1 + 2 * 8, 2 + 2 * 8, 4 + 1 * 8, 0 + 5 * 8, 3 + 3 * 8, 7 + 0 * 8, 3 + 4 * 8, 1 + 7 * 8, 5 + 3 * 8, 6 + 3 * 8, @@ -131,9 +131,9 @@ static int alloc_scratch_buffers(H264SliceContext *sl, int linesize) // (= 21x21 for H.264) av_fast_malloc(&sl->edge_emu_buffer, &sl->edge_emu_buffer_allocated, alloc_size * 2 * 21); - av_fast_malloc(&sl->top_borders[0], &sl->top_borders_allocated[0], + av_fast_mallocz(&sl->top_borders[0], &sl->top_borders_allocated[0], h->mb_width * 16 * 3 * sizeof(uint8_t) * 2); - av_fast_malloc(&sl->top_borders[1], &sl->top_borders_allocated[1], + av_fast_mallocz(&sl->top_borders[1], &sl->top_borders_allocated[1], h->mb_width * 16 * 3 * sizeof(uint8_t) * 2); if (!sl->bipred_scratchpad || !sl->edge_emu_buffer || @@ -192,6 +192,10 @@ static int alloc_picture(H264Context *h, H264Picture *pic) if (ret < 0) goto fail; + pic->crop = h->ps.sps->crop; + pic->crop_top = h->ps.sps->crop_top; + pic->crop_left= h->ps.sps->crop_left; + if (h->avctx->hwaccel) { const AVHWAccel *hwaccel = h->avctx->hwaccel; av_assert0(!pic->hwaccel_picture_private); @@ -202,6 +206,18 @@ static int alloc_picture(H264Context *h, H264Picture *pic) pic->hwaccel_picture_private = pic->hwaccel_priv_buf->data; } } + if (CONFIG_GRAY && !h->avctx->hwaccel && h->flags & AV_CODEC_FLAG_GRAY && pic->f->data[2]) { + int h_chroma_shift, v_chroma_shift; + av_pix_fmt_get_chroma_sub_sample(pic->f->format, + &h_chroma_shift, &v_chroma_shift); + + for(i=0; i<AV_CEIL_RSHIFT(pic->f->height, v_chroma_shift); i++) { + memset(pic->f->data[1] + pic->f->linesize[1]*i, + 0x80, AV_CEIL_RSHIFT(pic->f->width, h_chroma_shift)); + memset(pic->f->data[2] + pic->f->linesize[2]*i, + 0x80, AV_CEIL_RSHIFT(pic->f->width, h_chroma_shift)); + } + } if (!h->qscale_table_pool) { ret = init_table_pools(h); @@ -254,40 +270,13 @@ static int find_unused_picture(H264Context *h) return i; } -static int initialize_cur_frame(H264Context *h) -{ - H264Picture *cur; - int ret; - - release_unused_pictures(h, 1); - ff_h264_unref_picture(h, &h->cur_pic); - h->cur_pic_ptr = NULL; - - ret = find_unused_picture(h); - if (ret < 0) { - av_log(h->avctx, AV_LOG_ERROR, "no frame buffer available\n"); - return ret; - } - cur = &h->DPB[ret]; - - ret = alloc_picture(h, cur); - if (ret < 0) - return ret; - - ret = ff_h264_ref_picture(h, &h->cur_pic, cur); - if (ret < 0) - return ret; - h->cur_pic_ptr = cur; - - return 0; -} -#define IN_RANGE(a, b, size) (((a) >= (b)) && ((a) < ((b) + (size)))) +#define IN_RANGE(a, b, size) (((void*)(a) >= (void*)(b)) && ((void*)(a) < (void*)((b) + (size)))) #define REBASE_PICTURE(pic, new_ctx, old_ctx) \ - ((pic && pic >= old_ctx->DPB && \ - pic < old_ctx->DPB + H264_MAX_PICTURE_COUNT) ? \ - &new_ctx->DPB[pic - old_ctx->DPB] : NULL) + (((pic) && (pic) >= (old_ctx)->DPB && \ + (pic) < (old_ctx)->DPB + H264_MAX_PICTURE_COUNT) ? \ + &(new_ctx)->DPB[(pic) - (old_ctx)->DPB] : NULL) static void copy_picture_range(H264Picture **to, H264Picture **from, int count, H264Context *new_base, @@ -296,10 +285,9 @@ static void copy_picture_range(H264Picture **to, H264Picture **from, int count, int i; for (i = 0; i < count; i++) { - assert((IN_RANGE(from[i], old_base, sizeof(*old_base)) || - IN_RANGE(from[i], old_base->DPB, - sizeof(H264Picture) * H264_MAX_PICTURE_COUNT) || - !from[i])); + av_assert1(!from[i] || + IN_RANGE(from[i], old_base, 1) || + IN_RANGE(from[i], old_base->DPB, H264_MAX_PICTURE_COUNT)); to[i] = REBASE_PICTURE(from[i], new_base, old_base); } } @@ -314,11 +302,12 @@ int ff_h264_update_thread_context(AVCodecContext *dst, int need_reinit = 0; int i, ret; - if (dst == src || !h1->context_initialized) + if (dst == src) return 0; - if (!h1->ps.sps) - return AVERROR_INVALIDDATA; + // We can't fail if SPS isn't set at it breaks current skip_frame code + //if (!h1->ps.sps) + // return AVERROR_INVALIDDATA; if (inited && (h->width != h1->width || @@ -332,6 +321,9 @@ int ff_h264_update_thread_context(AVCodecContext *dst, need_reinit = 1; } + /* copy block_offset since frame_start may not be called */ + memcpy(h->block_offset, h1->block_offset, sizeof(h->block_offset)); + // SPS/PPS for (i = 0; i < FF_ARRAY_ELEMS(h->ps.sps_list); i++) { av_buffer_unref(&h->ps.sps_list[i]); @@ -350,7 +342,22 @@ int ff_h264_update_thread_context(AVCodecContext *dst, } } - h->ps.sps = h1->ps.sps; + av_buffer_unref(&h->ps.pps_ref); + av_buffer_unref(&h->ps.sps_ref); + h->ps.pps = NULL; + h->ps.sps = NULL; + if (h1->ps.pps_ref) { + h->ps.pps_ref = av_buffer_ref(h1->ps.pps_ref); + if (!h->ps.pps_ref) + return AVERROR(ENOMEM); + h->ps.pps = (const PPS*)h->ps.pps_ref->data; + } + if (h1->ps.sps_ref) { + h->ps.sps_ref = av_buffer_ref(h1->ps.sps_ref); + if (!h->ps.sps_ref) + return AVERROR(ENOMEM); + h->ps.sps = (SPS*)h->ps.sps_ref->data; + } if (need_reinit || !inited) { h->width = h1->width; @@ -361,11 +368,12 @@ int ff_h264_update_thread_context(AVCodecContext *dst, h->mb_stride = h1->mb_stride; h->b_stride = h1->b_stride; - if ((err = h264_slice_header_init(h)) < 0) { - av_log(h->avctx, AV_LOG_ERROR, "h264_slice_header_init() failed"); - return err; + if (h->context_initialized || h1->context_initialized) { + if ((err = h264_slice_header_init(h)) < 0) { + av_log(h->avctx, AV_LOG_ERROR, "h264_slice_header_init() failed"); + return err; + } } - /* copy block_offset since frame_start may not be called */ memcpy(h->block_offset, h1->block_offset, sizeof(h->block_offset)); } @@ -378,6 +386,9 @@ int ff_h264_update_thread_context(AVCodecContext *dst, h->first_field = h1->first_field; h->picture_structure = h1->picture_structure; h->droppable = h1->droppable; + h->backup_width = h1->backup_width; + h->backup_height = h1->backup_height; + h->backup_pix_fmt = h1->backup_pix_fmt; for (i = 0; i < H264_MAX_PICTURE_COUNT; i++) { ff_h264_unref_picture(h, &h->DPB[i]); @@ -401,12 +412,14 @@ int ff_h264_update_thread_context(AVCodecContext *dst, // extradata/NAL handling h->is_avc = h1->is_avc; h->nal_length_size = h1->nal_length_size; + h->sei.unregistered.x264_build = h1->sei.unregistered.x264_build; memcpy(&h->poc, &h1->poc, sizeof(h->poc)); h->curr_pic_num = h1->curr_pic_num; h->max_pic_num = h1->max_pic_num; + memcpy(h->default_ref, h1->default_ref, sizeof(h->default_ref)); memcpy(h->short_ref, h1->short_ref, sizeof(h->short_ref)); memcpy(h->long_ref, h1->long_ref, sizeof(h->long_ref)); memcpy(h->delayed_pic, h1->delayed_pic, sizeof(h->delayed_pic)); @@ -426,6 +439,8 @@ int ff_h264_update_thread_context(AVCodecContext *dst, copy_picture_range(h->delayed_pic, h1->delayed_pic, MAX_DELAYED_PIC_COUNT + 2, h, h1); + h->frame_recovered = h1->frame_recovered; + if (!h->cur_pic_ptr) return 0; @@ -438,7 +453,6 @@ int ff_h264_update_thread_context(AVCodecContext *dst, h->poc.prev_frame_num = h->poc.frame_num; h->recovery_frame = h1->recovery_frame; - h->frame_recovered = h1->frame_recovered; return err; } @@ -448,12 +462,28 @@ static int h264_frame_start(H264Context *h) H264Picture *pic; int i, ret; const int pixel_shift = h->pixel_shift; + int c[4] = { + 1<<(h->ps.sps->bit_depth_luma-1), + 1<<(h->ps.sps->bit_depth_chroma-1), + 1<<(h->ps.sps->bit_depth_chroma-1), + -1 + }; - ret = initialize_cur_frame(h); - if (ret < 0) - return ret; + if (!ff_thread_can_start_frame(h->avctx)) { + av_log(h->avctx, AV_LOG_ERROR, "Attempt to start a frame outside SETUP state\n"); + return -1; + } + + release_unused_pictures(h, 1); + h->cur_pic_ptr = NULL; + + i = find_unused_picture(h); + if (i < 0) { + av_log(h->avctx, AV_LOG_ERROR, "no frame buffer available\n"); + return i; + } + pic = &h->DPB[i]; - pic = h->cur_pic_ptr; pic->reference = h->droppable ? 0 : h->picture_structure; pic->f->coded_picture_number = h->coded_picture_number++; pic->field_picture = h->picture_structure != PICT_FRAME; @@ -466,11 +496,39 @@ static int h264_frame_start(H264Context *h) pic->f->key_frame = 0; pic->mmco_reset = 0; pic->recovered = 0; + pic->invalid_gap = 0; + pic->sei_recovery_frame_cnt = h->sei.recovery_point.recovery_frame_cnt; pic->f->pict_type = h->slice_ctx[0].slice_type; - if (CONFIG_ERROR_RESILIENCE && h->enable_er) + if ((ret = alloc_picture(h, pic)) < 0) + return ret; + if(!h->frame_recovered && !h->avctx->hwaccel +#if FF_API_CAP_VDPAU + && !(h->avctx->codec->capabilities & AV_CODEC_CAP_HWACCEL_VDPAU) +#endif + ) + ff_color_frame(pic->f, c); + + h->cur_pic_ptr = pic; + ff_h264_unref_picture(h, &h->cur_pic); + if (CONFIG_ERROR_RESILIENCE) { + ff_h264_set_erpic(&h->slice_ctx[0].er.cur_pic, NULL); + } + + if ((ret = ff_h264_ref_picture(h, &h->cur_pic, h->cur_pic_ptr)) < 0) + return ret; + + for (i = 0; i < h->nb_slice_ctx; i++) { + h->slice_ctx[i].linesize = h->cur_pic_ptr->f->linesize[0]; + h->slice_ctx[i].uvlinesize = h->cur_pic_ptr->f->linesize[1]; + } + + if (CONFIG_ERROR_RESILIENCE && h->enable_er) { ff_er_frame_start(&h->slice_ctx[0].er); + ff_h264_set_erpic(&h->slice_ctx[0].er.last_pic, NULL); + ff_h264_set_erpic(&h->slice_ctx[0].er.next_pic, NULL); + } for (i = 0; i < 16; i++) { h->block_offset[i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 4 * pic->f->linesize[0] * ((scan8[i] - scan8[0]) >> 3); @@ -483,11 +541,6 @@ static int h264_frame_start(H264Context *h) h->block_offset[48 + 32 + i] = (4 * ((scan8[i] - scan8[0]) & 7) << pixel_shift) + 8 * pic->f->linesize[1] * ((scan8[i] - scan8[0]) >> 3); } - /* Some macroblocks can be accessed before they're available in case - * of lost slices, MBAFF or threading. */ - memset(h->slice_table, -1, - (h->mb_height * h->mb_stride - 1) * sizeof(*h->slice_table)); - /* We mark the current picture as non-reference after allocating it, so * that if we break out due to an error it can be released automatically * in the next ff_mpv_frame_start(). @@ -623,7 +676,7 @@ static void implicit_weight_table(const H264Context *h, H264SliceContext *sl, in cur_poc = h->cur_pic_ptr->field_poc[h->picture_structure - 1]; } if (sl->ref_count[0] == 1 && sl->ref_count[1] == 1 && !FRAME_MBAFF(h) && - sl->ref_list[0][0].poc + sl->ref_list[1][0].poc == 2 * cur_poc) { + sl->ref_list[0][0].poc + (int64_t)sl->ref_list[1][0].poc == 2 * cur_poc) { sl->pwt.use_weight = 0; sl->pwt.use_weight_chroma = 0; return; @@ -644,7 +697,7 @@ static void implicit_weight_table(const H264Context *h, H264SliceContext *sl, in sl->pwt.chroma_log2_weight_denom = 5; for (ref0 = ref_start; ref0 < ref_count0; ref0++) { - int poc0 = sl->ref_list[0][ref0].poc; + int64_t poc0 = sl->ref_list[0][ref0].poc; for (ref1 = ref_start; ref1 < ref_count1; ref1++) { int w = 32; if (!sl->ref_list[0][ref0].parent->long_ref && !sl->ref_list[1][ref1].parent->long_ref) { @@ -675,13 +728,13 @@ static void init_scan_tables(H264Context *h) { int i; for (i = 0; i < 16; i++) { -#define TRANSPOSE(x) (x >> 2) | ((x << 2) & 0xF) +#define TRANSPOSE(x) ((x) >> 2) | (((x) << 2) & 0xF) h->zigzag_scan[i] = TRANSPOSE(ff_zigzag_scan[i]); h->field_scan[i] = TRANSPOSE(field_scan[i]); #undef TRANSPOSE } for (i = 0; i < 64; i++) { -#define TRANSPOSE(x) (x >> 3) | ((x & 7) << 3) +#define TRANSPOSE(x) ((x) >> 3) | (((x) & 7) << 3) h->zigzag_scan8x8[i] = TRANSPOSE(ff_zigzag_direct[i]); h->zigzag_scan8x8_cavlc[i] = TRANSPOSE(zigzag_scan8x8_cavlc[i]); h->field_scan8x8[i] = TRANSPOSE(field_scan8x8[i]); @@ -689,31 +742,33 @@ static void init_scan_tables(H264Context *h) #undef TRANSPOSE } if (h->ps.sps->transform_bypass) { // FIXME same ugly - h->zigzag_scan_q0 = ff_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 , ff_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 )); } } -static enum AVPixelFormat get_pixel_format(H264Context *h) +static enum AVPixelFormat get_pixel_format(H264Context *h, int force_callback) { #define HWACCEL_MAX (CONFIG_H264_DXVA2_HWACCEL + \ CONFIG_H264_D3D11VA_HWACCEL + \ CONFIG_H264_VAAPI_HWACCEL + \ (CONFIG_H264_VDA_HWACCEL * 2) + \ + CONFIG_H264_VIDEOTOOLBOX_HWACCEL + \ CONFIG_H264_VDPAU_HWACCEL) enum AVPixelFormat pix_fmts[HWACCEL_MAX + 2], *fmt = pix_fmts; const enum AVPixelFormat *choices = pix_fmts; + int i; switch (h->ps.sps->bit_depth_luma) { case 9: @@ -738,6 +793,28 @@ static enum AVPixelFormat get_pixel_format(H264Context *h) else *fmt++ = AV_PIX_FMT_YUV420P10; break; + case 12: + if (CHROMA444(h)) { + if (h->avctx->colorspace == AVCOL_SPC_RGB) { + *fmt++ = AV_PIX_FMT_GBRP12; + } else + *fmt++ = AV_PIX_FMT_YUV444P12; + } else if (CHROMA422(h)) + *fmt++ = AV_PIX_FMT_YUV422P12; + else + *fmt++ = AV_PIX_FMT_YUV420P12; + break; + case 14: + if (CHROMA444(h)) { + if (h->avctx->colorspace == AVCOL_SPC_RGB) { + *fmt++ = AV_PIX_FMT_GBRP14; + } else + *fmt++ = AV_PIX_FMT_YUV444P14; + } else if (CHROMA422(h)) + *fmt++ = AV_PIX_FMT_YUV422P14; + else + *fmt++ = AV_PIX_FMT_YUV420P14; + break; case 8: #if CONFIG_H264_VDPAU_HWACCEL *fmt++ = AV_PIX_FMT_VDPAU; @@ -768,6 +845,9 @@ static enum AVPixelFormat get_pixel_format(H264Context *h) *fmt++ = AV_PIX_FMT_VDA_VLD; *fmt++ = AV_PIX_FMT_VDA; #endif +#if CONFIG_H264_VIDEOTOOLBOX_HWACCEL + *fmt++ = AV_PIX_FMT_VIDEOTOOLBOX; +#endif if (h->avctx->codec->pix_fmts) choices = h->avctx->codec->pix_fmts; else if (h->avctx->color_range == AVCOL_RANGE_JPEG) @@ -784,7 +864,10 @@ static enum AVPixelFormat get_pixel_format(H264Context *h) *fmt = AV_PIX_FMT_NONE; - return ff_get_format(h->avctx, choices); + for (i=0; choices[i] != AV_PIX_FMT_NONE; i++) + if (choices[i] == h->avctx->pix_fmt && !force_callback) + return choices[i]; + return ff_thread_get_format(h->avctx, choices); } /* export coded and cropped frame dimensions to AVCodecContext */ @@ -793,10 +876,15 @@ static int init_dimensions(H264Context *h) SPS *sps = h->ps.sps; int width = h->width - (sps->crop_right + sps->crop_left); int height = h->height - (sps->crop_top + sps->crop_bottom); + av_assert0(sps->crop_right + sps->crop_left < (unsigned)h->width); + av_assert0(sps->crop_top + sps->crop_bottom < (unsigned)h->height); /* handle container cropping */ if (FFALIGN(h->avctx->width, 16) == FFALIGN(width, 16) && - FFALIGN(h->avctx->height, 16) == FFALIGN(height, 16)) { + FFALIGN(h->avctx->height, 16) == FFALIGN(height, 16) && + h->avctx->width <= width && + h->avctx->height <= height + ) { width = h->avctx->width; height = h->avctx->height; } @@ -840,7 +928,7 @@ static int h264_slice_header_init(H264Context *h) if (h->sei.unregistered.x264_build < 44U) den *= 2; av_reduce(&h->avctx->framerate.den, &h->avctx->framerate.num, - sps->num_units_in_tick, den, 1 << 30); + sps->num_units_in_tick * h->avctx->ticks_per_frame, den, 1 << 30); } ff_h264_free_tables(h); @@ -852,16 +940,32 @@ static int h264_slice_header_init(H264Context *h) ret = ff_h264_alloc_tables(h); if (ret < 0) { av_log(h->avctx, AV_LOG_ERROR, "Could not allocate memory\n"); - return ret; + goto fail; } - if (sps->bit_depth_luma < 8 || sps->bit_depth_luma > 10) { +#if FF_API_CAP_VDPAU + if (h->avctx->codec && + h->avctx->codec->capabilities & AV_CODEC_CAP_HWACCEL_VDPAU && + (sps->bit_depth_luma != 8 || sps->chroma_format_idc > 1)) { + av_log(h->avctx, AV_LOG_ERROR, + "VDPAU decoding does not support video colorspace.\n"); + ret = AVERROR_INVALIDDATA; + goto fail; + } +#endif + + if (sps->bit_depth_luma < 8 || sps->bit_depth_luma > 14 || + sps->bit_depth_luma == 11 || sps->bit_depth_luma == 13 + ) { av_log(h->avctx, AV_LOG_ERROR, "Unsupported bit depth %d\n", sps->bit_depth_luma); - return AVERROR_INVALIDDATA; + ret = AVERROR_INVALIDDATA; + goto fail; } + h->cur_bit_depth_luma = h->avctx->bits_per_raw_sample = sps->bit_depth_luma; + h->cur_chroma_format_idc = sps->chroma_format_idc; h->pixel_shift = sps->bit_depth_luma > 8; h->chroma_format_idc = sps->chroma_format_idc; h->bit_depth_luma = sps->bit_depth_luma; @@ -878,7 +982,7 @@ static int h264_slice_header_init(H264Context *h) ret = ff_h264_slice_context_init(h, &h->slice_ctx[0]); if (ret < 0) { av_log(h->avctx, AV_LOG_ERROR, "context_init() failed.\n"); - return ret; + goto fail; } } else { for (i = 0; i < h->nb_slice_ctx; i++) { @@ -891,7 +995,7 @@ static int h264_slice_header_init(H264Context *h) if ((ret = ff_h264_slice_context_init(h, sl)) < 0) { av_log(h->avctx, AV_LOG_ERROR, "context_init() failed.\n"); - return ret; + goto fail; } } } @@ -899,6 +1003,21 @@ static int h264_slice_header_init(H264Context *h) h->context_initialized = 1; return 0; +fail: + ff_h264_free_tables(h); + h->context_initialized = 0; + return ret; +} + +static enum AVPixelFormat non_j_pixfmt(enum AVPixelFormat a) +{ + switch (a) { + case AV_PIX_FMT_YUVJ420P: return AV_PIX_FMT_YUV420P; + case AV_PIX_FMT_YUVJ422P: return AV_PIX_FMT_YUV422P; + case AV_PIX_FMT_YUVJ444P: return AV_PIX_FMT_YUV444P; + default: + return a; + } } static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) @@ -910,19 +1029,51 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) int ret; unsigned int slice_type, tmp, i; int last_pic_structure, last_pic_droppable; + int must_reinit; int needs_reinit = 0; int field_pic_flag, bottom_field_flag; + int first_slice = sl == h->slice_ctx && !h->current_slice; int frame_num, droppable, picture_structure; - int mb_aff_frame = 0; + int mb_aff_frame, last_mb_aff_frame; - first_mb_in_slice = get_ue_golomb(&sl->gb); + if (first_slice) + av_assert0(!h->setup_finished); + + first_mb_in_slice = get_ue_golomb_long(&sl->gb); if (first_mb_in_slice == 0) { // FIXME better field boundary detection - if (h->current_slice && h->cur_pic_ptr && FIELD_PICTURE(h)) { - ff_h264_field_end(h, sl, 1); + if (h->current_slice) { + if (h->setup_finished) { + av_log(h->avctx, AV_LOG_ERROR, "Too many fields\n"); + return AVERROR_INVALIDDATA; + } + if (h->max_contexts > 1) { + if (!h->single_decode_warning) { + av_log(h->avctx, AV_LOG_WARNING, "Cannot decode multiple access units as slice threads\n"); + h->single_decode_warning = 1; + } + h->max_contexts = 1; + return SLICE_SINGLETHREAD; + } + + if (h->cur_pic_ptr && FIELD_PICTURE(h) && h->first_field) { + ret = ff_h264_field_end(h, h->slice_ctx, 1); + h->current_slice = 0; + if (ret < 0) + return ret; + } else if (h->cur_pic_ptr && !FIELD_PICTURE(h) && !h->first_field && h->nal_unit_type == NAL_IDR_SLICE) { + av_log(h, AV_LOG_WARNING, "Broken frame packetizing\n"); + ret = ff_h264_field_end(h, h->slice_ctx, 1); + h->current_slice = 0; + ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, 0); + ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, 1); + h->cur_pic_ptr = NULL; + if (ret < 0) + return ret; + } else + return AVERROR_INVALIDDATA; } - h->current_slice = 0; if (!h->first_field) { if (h->cur_pic_ptr && !h->droppable) { ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, @@ -932,6 +1083,9 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) } } + if (!h->current_slice) + av_assert0(sl == h->slice_ctx); + slice_type = get_ue_golomb_31(&sl->gb); if (slice_type > 9) { av_log(h->avctx, AV_LOG_ERROR, @@ -955,6 +1109,17 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) return AVERROR_INVALIDDATA; } + if (h->current_slice == 0 && !h->first_field) { + if ( + (h->avctx->skip_frame >= AVDISCARD_NONREF && !h->nal_ref_idc) || + (h->avctx->skip_frame >= AVDISCARD_BIDIR && sl->slice_type_nos == AV_PICTURE_TYPE_B) || + (h->avctx->skip_frame >= AVDISCARD_NONINTRA && sl->slice_type_nos != AV_PICTURE_TYPE_I) || + (h->avctx->skip_frame >= AVDISCARD_NONKEY && h->nal_unit_type != NAL_IDR_SLICE && h->sei.recovery_point.recovery_frame_cnt < 0) || + h->avctx->skip_frame >= AVDISCARD_ALL) { + return SLICE_SKIPED; + } + } + pps_id = get_ue_golomb(&sl->gb); if (pps_id >= MAX_PPS_COUNT) { av_log(h->avctx, AV_LOG_ERROR, "pps_id %u out of range\n", pps_id); @@ -966,22 +1131,52 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) pps_id); return AVERROR_INVALIDDATA; } - if (!h->setup_finished) { - h->ps.pps = (const PPS*)h->ps.pps_list[pps_id]->data; - } else if (h->ps.pps != (const PPS*)h->ps.pps_list[pps_id]->data) { - av_log(h->avctx, AV_LOG_ERROR, "PPS changed between slices\n"); - return AVERROR_INVALIDDATA; - } - if (!h->ps.sps_list[h->ps.pps->sps_id]) { + pps = (const PPS*)h->ps.pps_list[pps_id]->data; + + if (!h->ps.sps_list[pps->sps_id]) { av_log(h->avctx, AV_LOG_ERROR, "non-existing SPS %u referenced\n", - h->ps.pps->sps_id); + pps->sps_id); return AVERROR_INVALIDDATA; } + if (first_slice) { + av_buffer_unref(&h->ps.pps_ref); + h->ps.pps = NULL; + h->ps.pps_ref = av_buffer_ref(h->ps.pps_list[pps_id]); + if (!h->ps.pps_ref) + return AVERROR(ENOMEM); + h->ps.pps = (const PPS*)h->ps.pps_ref->data; + } else { + if (h->ps.pps->sps_id != pps->sps_id || + h->ps.pps->transform_8x8_mode != pps->transform_8x8_mode /*|| + (h->setup_finished && h->ps.pps != pps)*/) { + av_log(h->avctx, AV_LOG_ERROR, "PPS changed between slices\n"); + return AVERROR_INVALIDDATA; + } + } + if (h->ps.sps != (const SPS*)h->ps.sps_list[h->ps.pps->sps_id]->data) { - h->ps.sps = (SPS*)h->ps.sps_list[h->ps.pps->sps_id]->data; + if (!first_slice) { + av_log(h->avctx, AV_LOG_ERROR, + "SPS changed in the middle of the frame\n"); + return AVERROR_INVALIDDATA; + } + + av_buffer_unref(&h->ps.sps_ref); + h->ps.sps = NULL; + h->ps.sps_ref = av_buffer_ref(h->ps.sps_list[h->ps.pps->sps_id]); + if (!h->ps.sps_ref) + return AVERROR(ENOMEM); + h->ps.sps = (const SPS*)h->ps.sps_ref->data; + + if (h->mb_width != h->ps.sps->mb_width || + h->mb_height != h->ps.sps->mb_height * (2 - h->ps.sps->frame_mbs_only_flag) || + h->cur_bit_depth_luma != h->ps.sps->bit_depth_luma || + h->cur_chroma_format_idc != h->ps.sps->chroma_format_idc + ) + needs_reinit = 1; if (h->bit_depth_luma != h->ps.sps->bit_depth_luma || h->chroma_format_idc != h->ps.sps->chroma_format_idc) @@ -991,15 +1186,26 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) pps = h->ps.pps; sps = h->ps.sps; + must_reinit = (h->context_initialized && + ( 16*sps->mb_width != h->avctx->coded_width + || 16*sps->mb_height * (2 - sps->frame_mbs_only_flag) != h->avctx->coded_height + || h->cur_bit_depth_luma != sps->bit_depth_luma + || h->cur_chroma_format_idc != sps->chroma_format_idc + || h->mb_width != sps->mb_width + || h->mb_height != sps->mb_height * (2 - sps->frame_mbs_only_flag) + )); + if (h->avctx->pix_fmt == AV_PIX_FMT_NONE + || (non_j_pixfmt(h->avctx->pix_fmt) != non_j_pixfmt(get_pixel_format(h, 0)))) + must_reinit = 1; + + if (first_slice && av_cmp_q(sps->sar, h->avctx->sample_aspect_ratio)) + must_reinit = 1; + if (!h->setup_finished) { h->avctx->profile = ff_h264_get_profile(sps); h->avctx->level = sps->level_idc; h->avctx->refs = sps->ref_frame_count; - if (h->mb_width != sps->mb_width || - h->mb_height != sps->mb_height * (2 - sps->frame_mbs_only_flag)) - needs_reinit = 1; - h->mb_width = sps->mb_width; h->mb_height = sps->mb_height * (2 - sps->frame_mbs_only_flag); h->mb_num = h->mb_width * h->mb_height; @@ -1017,8 +1223,8 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) return ret; if (sps->video_signal_type_present_flag) { - h->avctx->color_range = sps->full_range ? AVCOL_RANGE_JPEG - : AVCOL_RANGE_MPEG; + h->avctx->color_range = sps->full_range > 0 ? AVCOL_RANGE_JPEG + : AVCOL_RANGE_MPEG; if (sps->colour_description_present_flag) { if (h->avctx->colorspace != sps->colorspace) needs_reinit = 1; @@ -1029,7 +1235,8 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) } } - if (h->context_initialized && needs_reinit) { + if (h->context_initialized && + (must_reinit || needs_reinit)) { h->context_initialized = 0; if (sl != h->slice_ctx) { av_log(h->avctx, AV_LOG_ERROR, @@ -1041,14 +1248,16 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) return AVERROR_INVALIDDATA; } + av_assert1(first_slice); + ff_h264_flush_change(h); - if ((ret = get_pixel_format(h)) < 0) + if ((ret = get_pixel_format(h, 1)) < 0) return ret; h->avctx->pix_fmt = ret; av_log(h->avctx, AV_LOG_INFO, "Reinit context to %dx%d, " - "pix_fmt: %d\n", h->width, h->height, h->avctx->pix_fmt); + "pix_fmt: %s\n", h->width, h->height, av_get_pix_fmt_name(h->avctx->pix_fmt)); if ((ret = h264_slice_header_init(h)) < 0) { av_log(h->avctx, AV_LOG_ERROR, @@ -1063,7 +1272,7 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) return AVERROR_PATCHWELCOME; } - if ((ret = get_pixel_format(h)) < 0) + if ((ret = get_pixel_format(h, 1)) < 0) return ret; h->avctx->pix_fmt = ret; @@ -1075,11 +1284,20 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) } frame_num = get_bits(&sl->gb, sps->log2_max_frame_num); + if (!first_slice) { + if (h->poc.frame_num != frame_num) { + av_log(h->avctx, AV_LOG_ERROR, "Frame num change from %d to %d\n", + h->poc.frame_num, frame_num); + return AVERROR_INVALIDDATA; + } + } + if (!h->setup_finished) h->poc.frame_num = frame_num; sl->mb_mbaff = 0; - + mb_aff_frame = 0; + last_mb_aff_frame = h->mb_aff_frame; last_pic_structure = h->picture_structure; last_pic_droppable = h->droppable; @@ -1087,7 +1305,12 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) if (sps->frame_mbs_only_flag) { picture_structure = PICT_FRAME; } else { + if (!h->ps.sps->direct_8x8_inference_flag && slice_type == AV_PICTURE_TYPE_B) { + av_log(h->avctx, AV_LOG_ERROR, "This stream was generated by a broken encoder, invalid 8x8 inference\n"); + return -1; + } field_pic_flag = get_bits1(&sl->gb); + if (field_pic_flag) { bottom_field_flag = get_bits1(&sl->gb); picture_structure = PICT_TOP_FIELD + bottom_field_flag; @@ -1096,16 +1319,11 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) mb_aff_frame = sps->mb_aff; } } - if (!h->setup_finished) { - h->droppable = droppable; - h->picture_structure = picture_structure; - h->mb_aff_frame = mb_aff_frame; - } - sl->mb_field_decoding_flag = h->picture_structure != PICT_FRAME; - if (h->current_slice != 0) { + if (h->current_slice) { if (last_pic_structure != picture_structure || - last_pic_droppable != droppable) { + last_pic_droppable != droppable || + last_mb_aff_frame != mb_aff_frame) { av_log(h->avctx, AV_LOG_ERROR, "Changing field mode (%d -> %d) between slices is not allowed\n", last_pic_structure, h->picture_structure); @@ -1116,7 +1334,16 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) h->current_slice + 1); return AVERROR_INVALIDDATA; } - } else { + } + + if (!h->setup_finished) { + h->droppable = droppable; + h->picture_structure = picture_structure; + h->mb_aff_frame = mb_aff_frame; + } + sl->mb_field_decoding_flag = picture_structure != PICT_FRAME; + + if (h->current_slice == 0) { /* Shorten frame num gaps so we don't have to allocate reference * frames just to throw them away */ if (h->poc.frame_num != h->poc.prev_frame_num) { @@ -1139,17 +1366,23 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) * Here, we're using that to see if we should mark previously * decode frames as "finished". * We have to do that before the "dummy" in-between frame allocation, - * since that can modify s->current_picture_ptr. */ + * since that can modify h->cur_pic_ptr. */ if (h->first_field) { - assert(h->cur_pic_ptr); - assert(h->cur_pic_ptr->f->buf[0]); + av_assert0(h->cur_pic_ptr); + av_assert0(h->cur_pic_ptr->f->buf[0]); assert(h->cur_pic_ptr->reference != DELAYED_PIC_REF); + /* Mark old field/frame as completed */ + if (h->cur_pic_ptr->tf.owner == h->avctx) { + ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, + last_pic_structure == PICT_BOTTOM_FIELD); + } + /* figure out if we have a complementary field pair */ if (!FIELD_PICTURE(h) || h->picture_structure == last_pic_structure) { /* Previous field is unmatched. Don't display it, but let it * remain for reference if marked as such. */ - if (!last_pic_droppable && last_pic_structure != PICT_FRAME) { + if (last_pic_structure != PICT_FRAME) { ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, last_pic_structure == PICT_TOP_FIELD); } @@ -1159,7 +1392,7 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) * different frame_nums. Consider this field first in * pair. Throw away previous field except for reference * purposes. */ - if (!last_pic_droppable && last_pic_structure != PICT_FRAME) { + if (last_pic_structure != PICT_FRAME) { ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, last_pic_structure == PICT_TOP_FIELD); } @@ -1186,12 +1419,15 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) } } - while (h->poc.frame_num != h->poc.prev_frame_num && + while (h->poc.frame_num != h->poc.prev_frame_num && !h->first_field && h->poc.frame_num != (h->poc.prev_frame_num + 1) % (1 << sps->log2_max_frame_num)) { H264Picture *prev = h->short_ref_count ? h->short_ref[0] : NULL; av_log(h->avctx, AV_LOG_DEBUG, "Frame num gap %d %d\n", h->poc.frame_num, h->poc.prev_frame_num); - ret = initialize_cur_frame(h); + if (!sps->gaps_in_frame_num_allowed_flag) + for(i=0; i<FF_ARRAY_ELEMS(h->last_pocs); i++) + h->last_pocs[i] = INT_MIN; + ret = h264_frame_start(h); if (ret < 0) { h->first_field = 0; return ret; @@ -1200,6 +1436,7 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) h->poc.prev_frame_num++; h->poc.prev_frame_num %= 1 << sps->log2_max_frame_num; h->cur_pic_ptr->frame_num = h->poc.prev_frame_num; + h->cur_pic_ptr->invalid_gap = !sps->gaps_in_frame_num_allowed_flag; ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, 0); ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, 1); ret = ff_generate_sliding_window_mmcos(h, 1); @@ -1225,8 +1462,8 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) (const uint8_t **)prev->f->data, prev->f->linesize, prev->f->format, - h->mb_width * 16, - h->mb_height * 16); + prev->f->width, + prev->f->height); h->short_ref[0]->poc = prev->poc + 2; } h->short_ref[0]->frame_num = h->poc.prev_frame_num; @@ -1237,18 +1474,22 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) * We're using that to see whether to continue decoding in that * frame, or to allocate a new one. */ if (h->first_field) { - assert(h->cur_pic_ptr); - assert(h->cur_pic_ptr->f->buf[0]); + av_assert0(h->cur_pic_ptr); + av_assert0(h->cur_pic_ptr->f->buf[0]); assert(h->cur_pic_ptr->reference != DELAYED_PIC_REF); /* figure out if we have a complementary field pair */ if (!FIELD_PICTURE(h) || h->picture_structure == last_pic_structure) { /* Previous field is unmatched. Don't display it, but let it * remain for reference if marked as such. */ + h->missing_fields ++; h->cur_pic_ptr = NULL; h->first_field = FIELD_PICTURE(h); } else { + h->missing_fields = 0; if (h->cur_pic_ptr->frame_num != h->poc.frame_num) { + ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX, + h->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. */ @@ -1272,9 +1513,18 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) } else { release_unused_pictures(h, 0); } + /* Some macroblocks can be accessed before they're available in case + * of lost slices, MBAFF or threading. */ + if (FIELD_PICTURE(h)) { + for(i = (h->picture_structure == PICT_BOTTOM_FIELD); i<h->mb_height; i++) + memset(h->slice_table + i*h->mb_stride, -1, (h->mb_stride - (i+1==h->mb_height)) * sizeof(*h->slice_table)); + } else { + memset(h->slice_table, -1, + (h->mb_height * h->mb_stride - 1) * sizeof(*h->slice_table)); + } } - assert(h->mb_num == h->mb_width * h->mb_height); + av_assert1(h->mb_num == h->mb_width * h->mb_height); if (first_mb_in_slice << FIELD_OR_MBAFF_PICTURE(h) >= h->mb_num || first_mb_in_slice >= h->mb_num) { av_log(h->avctx, AV_LOG_ERROR, "first_mb_in_slice overflow\n"); @@ -1285,7 +1535,7 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) FIELD_OR_MBAFF_PICTURE(h); if (h->picture_structure == PICT_BOTTOM_FIELD) sl->resync_mb_y = sl->mb_y = sl->mb_y + 1; - assert(sl->mb_y < h->mb_height); + av_assert1(sl->mb_y < h->mb_height); if (h->picture_structure == PICT_FRAME) { h->curr_pic_num = h->poc.frame_num; @@ -1296,7 +1546,7 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) } if (h->nal_unit_type == NAL_IDR_SLICE) - get_ue_golomb(&sl->gb); /* idr_pic_id */ + get_ue_golomb_long(&sl->gb); /* idr_pic_id */ if (sps->poc_type == 0) { int poc_lsb = get_bits(&sl->gb, sps->log2_max_poc_lsb); @@ -1337,7 +1587,7 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) ret = ff_h264_parse_ref_count(&sl->list_count, sl->ref_count, &sl->gb, pps, sl->slice_type_nos, - h->picture_structure); + h->picture_structure, h->avctx); if (ret < 0) return ret; @@ -1353,7 +1603,7 @@ static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl) (pps->weighted_bipred_idc == 1 && sl->slice_type_nos == AV_PICTURE_TYPE_B)) ff_h264_pred_weight_table(&sl->gb, sps, sl->ref_count, - sl->slice_type_nos, &sl->pwt); + sl->slice_type_nos, &sl->pwt, h->avctx); else if (pps->weighted_bipred_idc == 2 && sl->slice_type_nos == AV_PICTURE_TYPE_B) { implicit_weight_table(h, sl, -1); @@ -1461,6 +1711,8 @@ int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl) if (h->avctx->skip_loop_filter >= AVDISCARD_ALL || (h->avctx->skip_loop_filter >= AVDISCARD_NONKEY && + h->nal_unit_type != NAL_IDR_SLICE) || + (h->avctx->skip_loop_filter >= AVDISCARD_NONINTRA && sl->slice_type_nos != AV_PICTURE_TYPE_I) || (h->avctx->skip_loop_filter >= AVDISCARD_BIDIR && sl->slice_type_nos == AV_PICTURE_TYPE_B) || @@ -1468,7 +1720,7 @@ int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl) h->nal_ref_idc == 0)) sl->deblocking_filter = 0; - if (sl->deblocking_filter == 1 && h->nb_slice_ctx > 1) { + if (sl->deblocking_filter == 1 && h->max_contexts > 1) { if (h->avctx->flags2 & AV_CODEC_FLAG2_FAST) { /* Cheat slightly for speed: * Do not bother to deblock across slices. */ @@ -1485,9 +1737,14 @@ int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl) 6 * (h->ps.sps->bit_depth_luma - 8); sl->slice_num = ++h->current_slice; - if (sl->slice_num >= MAX_SLICES) { - av_log(h->avctx, AV_LOG_ERROR, - "Too many slices, increase MAX_SLICES and recompile\n"); + + if (sl->slice_num) + h->slice_row[(sl->slice_num-1)&(MAX_SLICES-1)]= sl->resync_mb_y; + if ( h->slice_row[sl->slice_num&(MAX_SLICES-1)] + 3 >= sl->resync_mb_y + && h->slice_row[sl->slice_num&(MAX_SLICES-1)] <= sl->resync_mb_y + && sl->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(h->avctx, AV_LOG_WARNING, "Possibly too many slices (%d >= %d), increase MAX_SLICES and recompile if there are artifacts\n", sl->slice_num, MAX_SLICES); } for (j = 0; j < 2; j++) { @@ -1580,12 +1837,12 @@ static av_always_inline void fill_filter_caches_inter(const 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(sl) ? 20 : 2); + const int *ref2frm = h->ref2frm[h->slice_table[top_xy] & (MAX_SLICES - 1)][list] + (MB_MBAFF(sl) ? 20 : 2); AV_COPY128(mv_dst - 1 * 8, h->cur_pic.motion_val[list][b_xy + 0]); ref_cache[0 - 1 * 8] = - ref_cache[1 - 1 * 8] = ref2frm[list][h->cur_pic.ref_index[list][b8_xy + 0]]; + ref_cache[1 - 1 * 8] = ref2frm[h->cur_pic.ref_index[list][b8_xy + 0]]; ref_cache[2 - 1 * 8] = - ref_cache[3 - 1 * 8] = ref2frm[list][h->cur_pic.ref_index[list][b8_xy + 1]]; + ref_cache[3 - 1 * 8] = ref2frm[h->cur_pic.ref_index[list][b8_xy + 1]]; } else { AV_ZERO128(mv_dst - 1 * 8); AV_WN32A(&ref_cache[0 - 1 * 8], ((LIST_NOT_USED) & 0xFF) * 0x01010101u); @@ -1595,15 +1852,15 @@ static av_always_inline void fill_filter_caches_inter(const 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(sl) ? 20 : 2); + const int *ref2frm = h->ref2frm[h->slice_table[left_xy[LTOP]] & (MAX_SLICES - 1)][list] + (MB_MBAFF(sl) ? 20 : 2); AV_COPY32(mv_dst - 1 + 0, h->cur_pic.motion_val[list][b_xy + b_stride * 0]); AV_COPY32(mv_dst - 1 + 8, h->cur_pic.motion_val[list][b_xy + b_stride * 1]); AV_COPY32(mv_dst - 1 + 16, h->cur_pic.motion_val[list][b_xy + b_stride * 2]); AV_COPY32(mv_dst - 1 + 24, h->cur_pic.motion_val[list][b_xy + b_stride * 3]); ref_cache[-1 + 0] = - ref_cache[-1 + 8] = ref2frm[list][h->cur_pic.ref_index[list][b8_xy + 2 * 0]]; + ref_cache[-1 + 8] = ref2frm[h->cur_pic.ref_index[list][b8_xy + 2 * 0]]; ref_cache[-1 + 16] = - ref_cache[-1 + 24] = ref2frm[list][h->cur_pic.ref_index[list][b8_xy + 2 * 1]]; + ref_cache[-1 + 24] = ref2frm[h->cur_pic.ref_index[list][b8_xy + 2 * 1]]; } else { AV_ZERO32(mv_dst - 1 + 0); AV_ZERO32(mv_dst - 1 + 8); @@ -1628,9 +1885,9 @@ static av_always_inline void fill_filter_caches_inter(const H264Context *h, { int8_t *ref = &h->cur_pic.ref_index[list][4 * mb_xy]; - int (*ref2frm)[64] = h->ref2frm[sl->slice_num & (MAX_SLICES - 1)][0] + (MB_MBAFF(sl) ? 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; + const int *ref2frm = h->ref2frm[sl->slice_num & (MAX_SLICES - 1)][list] + (MB_MBAFF(sl) ? 20 : 2); + uint32_t ref01 = (pack16to32(ref2frm[ref[0]], ref2frm[ref[1]]) & 0x00FF00FF) * 0x0101; + uint32_t ref23 = (pack16to32(ref2frm[ref[2]], ref2frm[ref[3]]) & 0x00FF00FF) * 0x0101; AV_WN32A(&ref_cache[0 * 8], ref01); AV_WN32A(&ref_cache[1 * 8], ref01); AV_WN32A(&ref_cache[2 * 8], ref23); @@ -1898,7 +2155,7 @@ static void decode_finish_row(const H264Context *h, H264SliceContext *sl) ff_h264_draw_horiz_band(h, sl, top, height); - if (h->droppable) + if (h->droppable || sl->h264->slice_ctx[0].er.error_occurred) return; ff_thread_report_progress(&h->cur_pic_ptr->tf, top + height - 1, @@ -1909,15 +2166,14 @@ static void er_add_slice(H264SliceContext *sl, int startx, int starty, int endx, int endy, int status) { -#if CONFIG_ERROR_RESILIENCE - ERContext *er = &sl->er; - if (!sl->h264->enable_er) return; - er->ref_count = sl->ref_count[0]; - ff_er_add_slice(er, startx, starty, endx, endy, status); -#endif + if (CONFIG_ERROR_RESILIENCE) { + ERContext *er = &sl->h264->slice_ctx[0].er; + + ff_er_add_slice(er, startx, starty, endx, endy, status); + } } static int decode_slice(struct AVCodecContext *avctx, void *arg) @@ -1937,6 +2193,8 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) sl->mb_skip_run = -1; + av_assert0(h->block_offset[15] == (4 * ((scan8[15] - scan8[0]) & 7) << h->pixel_shift) + 4 * sl->linesize * ((scan8[15] - scan8[0]) >> 3)); + if (h->postpone_filter) sl->deblocking_filter = 0; @@ -1944,24 +2202,37 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) avctx->codec_id != AV_CODEC_ID_H264 || (CONFIG_GRAY && (h->flags & AV_CODEC_FLAG_GRAY)); + if (!(h->avctx->active_thread_type & FF_THREAD_SLICE) && h->picture_structure == PICT_FRAME && h->slice_ctx[0].er.error_status_table) { + const int start_i = av_clip(sl->resync_mb_x + sl->resync_mb_y * h->mb_width, 0, h->mb_num - 1); + if (start_i) { + int prev_status = h->slice_ctx[0].er.error_status_table[h->slice_ctx[0].er.mb_index2xy[start_i - 1]]; + prev_status &= ~ VP_START; + if (prev_status != (ER_MV_END | ER_DC_END | ER_AC_END)) + h->slice_ctx[0].er.error_occurred = 1; + } + } + if (h->ps.pps->cabac) { /* realign */ align_get_bits(&sl->gb); /* init cabac */ - ff_init_cabac_decoder(&sl->cabac, + ret = ff_init_cabac_decoder(&sl->cabac, sl->gb.buffer + get_bits_count(&sl->gb) / 8, (get_bits_left(&sl->gb) + 7) / 8); + if (ret < 0) + return ret; ff_h264_init_cabac_states(h, sl); for (;;) { // START_TIMER int ret, eos; - if (sl->mb_x + sl->mb_y * h->mb_width >= sl->next_slice_idx) { av_log(h->avctx, AV_LOG_ERROR, "Slice overlaps with next at %d\n", sl->next_slice_idx); + er_add_slice(sl, sl->resync_mb_x, sl->resync_mb_y, sl->mb_x, + sl->mb_y, ER_MB_ERROR); return AVERROR_INVALIDDATA; } @@ -1991,9 +2262,11 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) loop_filter(h, sl, lf_x_start, sl->mb_x + 1); goto finish; } - if (ret < 0 || sl->cabac.bytestream > sl->cabac.bytestream_end + 2) { + if (sl->cabac.bytestream > sl->cabac.bytestream_end + 2 ) + av_log(h->avctx, AV_LOG_DEBUG, "bytestream overread %"PTRDIFF_SPECIFIER"\n", sl->cabac.bytestream_end - sl->cabac.bytestream); + if (ret < 0 || sl->cabac.bytestream > sl->cabac.bytestream_end + 4) { av_log(h->avctx, AV_LOG_ERROR, - "error while decoding MB %d %d, bytestream %td\n", + "error while decoding MB %d %d, bytestream %"PTRDIFF_SPECIFIER"\n", sl->mb_x, sl->mb_y, sl->cabac.bytestream_end - sl->cabac.bytestream); er_add_slice(sl, sl->resync_mb_x, sl->resync_mb_y, sl->mb_x, @@ -2030,6 +2303,8 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) if (sl->mb_x + sl->mb_y * h->mb_width >= sl->next_slice_idx) { av_log(h->avctx, AV_LOG_ERROR, "Slice overlaps with next at %d\n", sl->next_slice_idx); + er_add_slice(sl, sl->resync_mb_x, sl->resync_mb_y, sl->mb_x, + sl->mb_y, ER_MB_ERROR); return AVERROR_INVALIDDATA; } @@ -2070,14 +2345,15 @@ static int decode_slice(struct AVCodecContext *avctx, void *arg) ff_tlog(h->avctx, "slice end %d %d\n", get_bits_count(&sl->gb), sl->gb.size_in_bits); - if (get_bits_left(&sl->gb) == 0) { + if ( get_bits_left(&sl->gb) == 0 + || get_bits_left(&sl->gb) > 0 && !(h->avctx->err_recognition & AV_EF_AGGRESSIVE)) { er_add_slice(sl, sl->resync_mb_x, sl->resync_mb_y, sl->mb_x - 1, sl->mb_y, ER_MB_END); goto finish; } else { er_add_slice(sl, sl->resync_mb_x, sl->resync_mb_y, - sl->mb_x - 1, sl->mb_y, ER_MB_END); + sl->mb_x, sl->mb_y, ER_MB_END); return AVERROR_INVALIDDATA; } @@ -2122,7 +2398,15 @@ int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count) H264SliceContext *sl; int i, j; - if (h->avctx->hwaccel) + av_assert0(context_count && h->slice_ctx[context_count - 1].mb_y < h->mb_height); + + h->slice_ctx[0].next_slice_idx = INT_MAX; + + if (h->avctx->hwaccel +#if FF_API_CAP_VDPAU + || h->avctx->codec->capabilities & AV_CODEC_CAP_HWACCEL_VDPAU +#endif + ) return 0; if (context_count == 1) { int ret; @@ -2134,12 +2418,15 @@ int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count) h->mb_y = h->slice_ctx[0].mb_y; return ret; } else { + av_assert0(context_count > 0); for (i = 0; i < context_count; i++) { int next_slice_idx = h->mb_width * h->mb_height; int slice_idx; sl = &h->slice_ctx[i]; - sl->er.error_count = 0; + if (CONFIG_ERROR_RESILIENCE) { + sl->er.error_count = 0; + } /* make sure none of those slices overlap */ slice_idx = sl->mb_y * h->mb_width + sl->mb_x; @@ -2160,8 +2447,10 @@ int ff_h264_execute_decode_slices(H264Context *h, unsigned context_count) /* pull back stuff from slices to master context */ sl = &h->slice_ctx[context_count - 1]; h->mb_y = sl->mb_y; - for (i = 1; i < context_count; i++) - h->slice_ctx[0].er.error_count += h->slice_ctx[i].er.error_count; + if (CONFIG_ERROR_RESILIENCE) { + for (i = 1; i < context_count; i++) + h->slice_ctx[0].er.error_count += h->slice_ctx[i].er.error_count; + } if (h->postpone_filter) { h->postpone_filter = 0; |