summaryrefslogtreecommitdiff
path: root/libavcodec/h264_slice.c
diff options
context:
space:
mode:
Diffstat (limited to 'libavcodec/h264_slice.c')
-rw-r--r--libavcodec/h264_slice.c753
1 files changed, 508 insertions, 245 deletions
diff --git a/libavcodec/h264_slice.c b/libavcodec/h264_slice.c
index 491039626e..c6f40e59c5 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
*/
@@ -47,14 +47,14 @@
#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,
@@ -73,7 +73,7 @@ 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,
@@ -93,7 +93,7 @@ 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,
@@ -135,9 +135,9 @@ 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 ||
@@ -196,6 +196,10 @@ 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);
@@ -206,6 +210,18 @@ 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);
@@ -258,40 +274,13 @@ 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,
@@ -300,10 +289,9 @@ 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);
}
}
@@ -318,11 +306,12 @@ 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 ||
@@ -336,6 +325,9 @@ 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]);
@@ -354,7 +346,22 @@ 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 = (const SPS*)h->ps.sps_ref->data;
+ }
if (need_reinit || !inited) {
h->width = h1->width;
@@ -365,9 +372,11 @@ 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 */
@@ -383,6 +392,9 @@ int ff_h264_update_thread_context(AVCodecContext *dst,
h->picture_structure = h1->picture_structure;
h->mb_aff_frame = h1->mb_aff_frame;
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]);
@@ -406,9 +418,11 @@ 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));
+ 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));
@@ -429,6 +443,8 @@ 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;
@@ -441,7 +457,6 @@ 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;
}
@@ -451,12 +466,28 @@ 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;
@@ -469,11 +500,39 @@ 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);
@@ -486,11 +545,6 @@ 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().
@@ -628,7 +682,7 @@ 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;
@@ -649,7 +703,7 @@ 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) {
@@ -680,13 +734,13 @@ 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]);
@@ -694,31 +748,33 @@ 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:
@@ -743,6 +799,28 @@ 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;
@@ -773,6 +851,9 @@ 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)
@@ -789,40 +870,31 @@ 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 */
static int init_dimensions(H264Context *h)
{
- SPS *sps = h->ps.sps;
+ const SPS *sps = (const 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;
}
- if (width <= 0 || height <= 0) {
- av_log(h->avctx, AV_LOG_ERROR, "Invalid cropped dimensions: %dx%d.\n",
- width, height);
- if (h->avctx->err_recognition & AV_EF_EXPLODE)
- return AVERROR_INVALIDDATA;
-
- av_log(h->avctx, AV_LOG_WARNING, "Ignoring cropping information.\n");
- sps->crop_bottom =
- sps->crop_top =
- sps->crop_right =
- sps->crop_left =
- sps->crop = 0;
-
- width = h->width;
- height = h->height;
- }
-
h->avctx->coded_width = h->width;
h->avctx->coded_height = h->height;
h->avctx->width = width;
@@ -845,7 +917,7 @@ 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);
@@ -857,16 +929,32 @@ 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;
@@ -883,7 +971,7 @@ 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++) {
@@ -896,7 +984,7 @@ 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;
}
}
}
@@ -904,16 +992,51 @@ 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_init_ps(H264Context *h, const H264SliceContext *sl)
+static int h264_init_ps(H264Context *h, const H264SliceContext *sl, int first_slice)
{
const SPS *sps;
- int needs_reinit = 0, ret;
+ int needs_reinit = 0, must_reinit, ret;
+
+ 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[sl->pps_id]);
+ if (!h->ps.pps_ref)
+ return AVERROR(ENOMEM);
+ h->ps.pps = (const PPS*)h->ps.pps_ref->data;
+ }
- h->ps.pps = (const PPS*)h->ps.pps_list[sl->pps_id]->data;
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;
+ 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)
@@ -921,43 +1044,57 @@ static int h264_init_ps(H264Context *h, const H264SliceContext *sl)
}
sps = h->ps.sps;
- h->avctx->profile = ff_h264_get_profile(sps);
- h->avctx->level = sps->level_idc;
- h->avctx->refs = sps->ref_frame_count;
+ 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->mb_width != sps->mb_width ||
- h->mb_height != sps->mb_height * (2 - sps->frame_mbs_only_flag))
- needs_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;
- 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;
- h->mb_stride = h->mb_width + 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;
+ h->mb_stride = h->mb_width + 1;
- h->b_stride = h->mb_width * 4;
+ h->b_stride = h->mb_width * 4;
- h->chroma_y_shift = sps->chroma_format_idc <= 1; // 400 uses yuv420p
+ h->chroma_y_shift = sps->chroma_format_idc <= 1; // 400 uses yuv420p
- h->width = 16 * h->mb_width;
- h->height = 16 * h->mb_height;
+ h->width = 16 * h->mb_width;
+ h->height = 16 * h->mb_height;
- ret = init_dimensions(h);
- if (ret < 0)
- return ret;
+ ret = init_dimensions(h);
+ if (ret < 0)
+ return ret;
- if (sps->video_signal_type_present_flag) {
- h->avctx->color_range = sps->full_range ? AVCOL_RANGE_JPEG
- : AVCOL_RANGE_MPEG;
- if (sps->colour_description_present_flag) {
- if (h->avctx->colorspace != sps->colorspace)
- needs_reinit = 1;
- h->avctx->color_primaries = sps->color_primaries;
- h->avctx->color_trc = sps->color_trc;
- h->avctx->colorspace = sps->colorspace;
+ if (sps->video_signal_type_present_flag) {
+ 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;
+ h->avctx->color_primaries = sps->color_primaries;
+ h->avctx->color_trc = sps->color_trc;
+ h->avctx->colorspace = sps->colorspace;
+ }
}
}
- if (!h->context_initialized || needs_reinit) {
+ if (!h->context_initialized || must_reinit || needs_reinit) {
+ int flush_changes = h->context_initialized;
h->context_initialized = 0;
if (sl != h->slice_ctx) {
av_log(h->avctx, AV_LOG_ERROR,
@@ -969,14 +1106,17 @@ static int h264_init_ps(H264Context *h, const H264SliceContext *sl)
return AVERROR_INVALIDDATA;
}
- ff_h264_flush_change(h);
+ av_assert1(first_slice);
+
+ if (flush_changes)
+ 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_VERBOSE, "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,
@@ -1045,7 +1185,7 @@ static int h264_export_frame_props(H264Context *h)
/* Derive top_field_first from field pocs. */
cur->f->top_field_first = cur->field_poc[0] < cur->field_poc[1];
} else {
- if (cur->f->interlaced_frame || sps->pic_struct_present_flag) {
+ if (sps->pic_struct_present_flag) {
/* Use picture timing SEI information. Even if it is a
* information of a past frame, better than nothing. */
if (h->sei.picture_timing.pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM ||
@@ -1053,6 +1193,10 @@ static int h264_export_frame_props(H264Context *h)
cur->f->top_field_first = 1;
else
cur->f->top_field_first = 0;
+ } else if (cur->f->interlaced_frame) {
+ /* Default to top field first when pic_struct_present_flag
+ * is not set but interlaced frame detected */
+ cur->f->top_field_first = 1;
} else {
/* Most likely progressive */
cur->f->top_field_first = 0;
@@ -1060,16 +1204,13 @@ static int h264_export_frame_props(H264Context *h)
}
if (h->sei.frame_packing.present &&
- h->sei.frame_packing.arrangement_type >= 0 &&
- h->sei.frame_packing.arrangement_type <= 6 &&
+ h->sei.frame_packing.frame_packing_arrangement_type <= 6 &&
h->sei.frame_packing.content_interpretation_type > 0 &&
h->sei.frame_packing.content_interpretation_type < 3) {
H264SEIFramePacking *fp = &h->sei.frame_packing;
AVStereo3D *stereo = av_stereo3d_create_side_data(cur->f);
- if (!stereo)
- return AVERROR(ENOMEM);
-
- switch (fp->arrangement_type) {
+ if (stereo) {
+ switch (fp->frame_packing_arrangement_type) {
case 0:
stereo->type = AV_STEREO3D_CHECKERBOARD;
break;
@@ -1080,7 +1221,7 @@ static int h264_export_frame_props(H264Context *h)
stereo->type = AV_STEREO3D_LINES;
break;
case 3:
- if (fp->quincunx_subsampling)
+ if (fp->quincunx_sampling_flag)
stereo->type = AV_STEREO3D_SIDEBYSIDE_QUINCUNX;
else
stereo->type = AV_STEREO3D_SIDEBYSIDE;
@@ -1098,6 +1239,7 @@ static int h264_export_frame_props(H264Context *h)
if (fp->content_interpretation_type == 2)
stereo->flags = AV_STEREO3D_FLAG_INVERT;
+ }
}
if (h->sei.display_orientation.present &&
@@ -1109,22 +1251,21 @@ static int h264_export_frame_props(H264Context *h)
AVFrameSideData *rotation = av_frame_new_side_data(cur->f,
AV_FRAME_DATA_DISPLAYMATRIX,
sizeof(int32_t) * 9);
- if (!rotation)
- return AVERROR(ENOMEM);
-
- av_display_rotation_set((int32_t *)rotation->data, angle);
- av_display_matrix_flip((int32_t *)rotation->data,
- o->hflip, o->vflip);
+ if (rotation) {
+ av_display_rotation_set((int32_t *)rotation->data, angle);
+ av_display_matrix_flip((int32_t *)rotation->data,
+ o->hflip, o->vflip);
+ }
}
if (h->sei.afd.present) {
AVFrameSideData *sd = av_frame_new_side_data(cur->f, AV_FRAME_DATA_AFD,
sizeof(uint8_t));
- if (!sd)
- return AVERROR(ENOMEM);
- *sd->data = h->sei.afd.active_format_description;
- h->sei.afd.present = 0;
+ if (sd) {
+ *sd->data = h->sei.afd.active_format_description;
+ h->sei.afd.present = 0;
+ }
}
if (h->sei.a53_caption.a53_caption) {
@@ -1132,12 +1273,11 @@ static int h264_export_frame_props(H264Context *h)
AVFrameSideData *sd = av_frame_new_side_data(cur->f,
AV_FRAME_DATA_A53_CC,
a53->a53_caption_size);
- if (!sd)
- return AVERROR(ENOMEM);
-
- memcpy(sd->data, a53->a53_caption, a53->a53_caption_size);
+ if (sd)
+ memcpy(sd->data, a53->a53_caption, a53->a53_caption_size);
av_freep(&a53->a53_caption);
a53->a53_caption_size = 0;
+ h->avctx->properties |= FF_CODEC_PROPERTY_CLOSED_CAPTIONS;
}
return 0;
@@ -1148,13 +1288,14 @@ static int h264_export_frame_props(H264Context *h)
* or a second field in a pair and does the necessary setup.
*/
static int h264_field_start(H264Context *h, const H264SliceContext *sl,
- const H2645NAL *nal)
+ const H2645NAL *nal, int first_slice)
{
+ int i;
const SPS *sps;
int last_pic_structure, last_pic_droppable, ret;
- ret = h264_init_ps(h, sl);
+ ret = h264_init_ps(h, sl, first_slice);
if (ret < 0)
return ret;
@@ -1193,17 +1334,23 @@ static int h264_field_start(H264Context *h, const 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);
}
@@ -1213,7 +1360,7 @@ static int h264_field_start(H264Context *h, const 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);
}
@@ -1240,12 +1387,15 @@ static int h264_field_start(H264Context *h, const 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;
@@ -1254,6 +1404,7 @@ static int h264_field_start(H264Context *h, const 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);
@@ -1278,8 +1429,8 @@ static int h264_field_start(H264Context *h, const 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;
@@ -1290,18 +1441,22 @@ static int h264_field_start(H264Context *h, const 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. */
@@ -1325,6 +1480,15 @@ static int h264_field_start(H264Context *h, const 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));
+ }
ff_h264_init_poc(h->cur_pic_ptr->field_poc, &h->cur_pic_ptr->poc,
h->ps.sps, &h->poc, h->picture_structure, nal->ref_idc);
@@ -1344,30 +1508,28 @@ static int h264_field_start(H264Context *h, const H264SliceContext *sl,
return ret;
}
- if (h->avctx->hwaccel) {
- ret = h->avctx->hwaccel->start_frame(h->avctx, NULL, 0);
- if (ret < 0)
- return ret;
- }
-
return 0;
}
-static int h264_slice_header_parse(H264SliceContext *sl, const H2645NAL *nal,
- const H264ParamSets *ps, AVCodecContext *avctx)
+static int h264_slice_header_parse(const H264Context *h, H264SliceContext *sl,
+ const H2645NAL *nal)
{
const SPS *sps;
const PPS *pps;
int ret;
unsigned int slice_type, tmp, i;
int field_pic_flag, bottom_field_flag;
- int droppable, picture_structure;
+ int first_slice = sl == h->slice_ctx && !h->current_slice;
+ int picture_structure;
+
+ if (first_slice)
+ av_assert0(!h->setup_finished);
- sl->first_mb_addr = get_ue_golomb(&sl->gb);
+ sl->first_mb_addr = get_ue_golomb_long(&sl->gb);
slice_type = get_ue_golomb_31(&sl->gb);
if (slice_type > 9) {
- av_log(avctx, AV_LOG_ERROR,
+ av_log(h->avctx, AV_LOG_ERROR,
"slice type %d too large at %d\n",
slice_type, sl->first_mb_addr);
return AVERROR_INVALIDDATA;
@@ -1384,38 +1546,48 @@ static int h264_slice_header_parse(H264SliceContext *sl, const H2645NAL *nal,
if (nal->type == H264_NAL_IDR_SLICE &&
sl->slice_type_nos != AV_PICTURE_TYPE_I) {
- av_log(avctx, AV_LOG_ERROR, "A non-intra slice in an IDR NAL unit.\n");
+ av_log(h->avctx, AV_LOG_ERROR, "A non-intra slice in an IDR NAL unit.\n");
return AVERROR_INVALIDDATA;
}
sl->pps_id = get_ue_golomb(&sl->gb);
if (sl->pps_id >= MAX_PPS_COUNT) {
- av_log(avctx, AV_LOG_ERROR, "pps_id %u out of range\n", sl->pps_id);
+ av_log(h->avctx, AV_LOG_ERROR, "pps_id %u out of range\n", sl->pps_id);
return AVERROR_INVALIDDATA;
}
- if (!ps->pps_list[sl->pps_id]) {
- av_log(avctx, AV_LOG_ERROR,
+ if (!h->ps.pps_list[sl->pps_id]) {
+ av_log(h->avctx, AV_LOG_ERROR,
"non-existing PPS %u referenced\n",
sl->pps_id);
return AVERROR_INVALIDDATA;
}
- pps = (const PPS*)ps->pps_list[sl->pps_id]->data;
+ pps = (const PPS*)h->ps.pps_list[sl->pps_id]->data;
- if (!ps->sps_list[pps->sps_id]) {
- av_log(avctx, AV_LOG_ERROR,
+ if (!h->ps.sps_list[pps->sps_id]) {
+ av_log(h->avctx, AV_LOG_ERROR,
"non-existing SPS %u referenced\n", pps->sps_id);
return AVERROR_INVALIDDATA;
}
- sps = (const SPS*)ps->sps_list[pps->sps_id]->data;
+ sps = (const SPS*)h->ps.sps_list[pps->sps_id]->data;
sl->frame_num = get_bits(&sl->gb, sps->log2_max_frame_num);
+ if (!first_slice) {
+ if (h->poc.frame_num != sl->frame_num) {
+ av_log(h->avctx, AV_LOG_ERROR, "Frame num change from %d to %d\n",
+ h->poc.frame_num, sl->frame_num);
+ return AVERROR_INVALIDDATA;
+ }
+ }
sl->mb_mbaff = 0;
- droppable = nal->ref_idc == 0;
if (sps->frame_mbs_only_flag) {
picture_structure = PICT_FRAME;
} else {
+ if (!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);
@@ -1436,7 +1608,7 @@ static int h264_slice_header_parse(H264SliceContext *sl, const H2645NAL *nal,
}
if (nal->type == H264_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) {
sl->poc_lsb = get_bits(&sl->gb, sps->log2_max_poc_lsb);
@@ -1461,12 +1633,12 @@ static int h264_slice_header_parse(H264SliceContext *sl, const H2645NAL *nal,
ret = ff_h264_parse_ref_count(&sl->list_count, sl->ref_count,
&sl->gb, pps, sl->slice_type_nos,
- picture_structure);
+ picture_structure, h->avctx);
if (ret < 0)
return ret;
if (sl->slice_type_nos != AV_PICTURE_TYPE_I) {
- ret = ff_h264_decode_ref_pic_list_reordering(sl, avctx);
+ ret = ff_h264_decode_ref_pic_list_reordering(sl, h->avctx);
if (ret < 0) {
sl->ref_count[1] = sl->ref_count[0] = 0;
return ret;
@@ -1482,19 +1654,19 @@ static int h264_slice_header_parse(H264SliceContext *sl, const H2645NAL *nal,
(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);
sl->explicit_ref_marking = 0;
if (nal->ref_idc) {
- ret = ff_h264_decode_ref_pic_marking(sl, &sl->gb, nal, avctx);
- if (ret < 0 && (avctx->err_recognition & AV_EF_EXPLODE))
+ ret = ff_h264_decode_ref_pic_marking(sl, &sl->gb, nal, h->avctx);
+ if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE))
return AVERROR_INVALIDDATA;
}
if (sl->slice_type_nos != AV_PICTURE_TYPE_I && pps->cabac) {
tmp = get_ue_golomb_31(&sl->gb);
if (tmp > 2) {
- av_log(avctx, AV_LOG_ERROR, "cabac_init_idc %u overflow\n", tmp);
+ av_log(h->avctx, AV_LOG_ERROR, "cabac_init_idc %u overflow\n", tmp);
return AVERROR_INVALIDDATA;
}
sl->cabac_init_idc = tmp;
@@ -1503,7 +1675,7 @@ static int h264_slice_header_parse(H264SliceContext *sl, const H2645NAL *nal,
sl->last_qscale_diff = 0;
tmp = pps->init_qp + get_se_golomb(&sl->gb);
if (tmp > 51 + 6 * (sps->bit_depth_luma - 8)) {
- av_log(avctx, AV_LOG_ERROR, "QP %u out of range\n", tmp);
+ av_log(h->avctx, AV_LOG_ERROR, "QP %u out of range\n", tmp);
return AVERROR_INVALIDDATA;
}
sl->qscale = tmp;
@@ -1522,7 +1694,7 @@ static int h264_slice_header_parse(H264SliceContext *sl, const H2645NAL *nal,
if (pps->deblocking_filter_parameters_present) {
tmp = get_ue_golomb_31(&sl->gb);
if (tmp > 2) {
- av_log(avctx, AV_LOG_ERROR,
+ av_log(h->avctx, AV_LOG_ERROR,
"deblocking_filter_idc %u out of range\n", tmp);
return AVERROR_INVALIDDATA;
}
@@ -1537,7 +1709,7 @@ static int h264_slice_header_parse(H264SliceContext *sl, const H2645NAL *nal,
sl->slice_alpha_c0_offset < -12 ||
sl->slice_beta_offset > 12 ||
sl->slice_beta_offset < -12) {
- av_log(avctx, AV_LOG_ERROR,
+ av_log(h->avctx, AV_LOG_ERROR,
"deblocking filter parameters %d %d out of range\n",
sl->slice_alpha_c0_offset, sl->slice_beta_offset);
return AVERROR_INVALIDDATA;
@@ -1560,8 +1732,9 @@ int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl,
const H2645NAL *nal)
{
int i, j, ret = 0;
+ int first_slice = sl == h->slice_ctx && !h->current_slice;
- ret = h264_slice_header_parse(sl, nal, &h->ps, h->avctx);
+ ret = h264_slice_header_parse(h, sl, nal);
if (ret < 0)
return ret;
@@ -1569,35 +1742,86 @@ int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl,
if (sl->redundant_pic_count > 0)
return 0;
- if (!h->setup_finished) {
- if (sl->first_mb_addr == 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 (sl->first_mb_addr == 0 || !h->current_slice) {
+ if (h->setup_finished) {
+ av_log(h->avctx, AV_LOG_ERROR, "Too many fields\n");
+ 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,
- h->picture_structure == PICT_BOTTOM_FIELD);
+ if (sl->first_mb_addr == 0) { // FIXME better field boundary detection
+ if (h->current_slice) {
+ 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 == H264_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;
+ }
+
+ if (!h->first_field) {
+ if (h->cur_pic_ptr && !h->droppable) {
+ ff_thread_report_progress(&h->cur_pic_ptr->tf, INT_MAX,
+ h->picture_structure == PICT_BOTTOM_FIELD);
}
+ h->cur_pic_ptr = NULL;
}
+ }
- if (h->current_slice == 0) {
- ret = h264_field_start(h, sl, nal);
- if (ret < 0)
- return ret;
+ if (!h->current_slice)
+ av_assert0(sl == h->slice_ctx);
+
+ 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 != H264_NAL_IDR_SLICE && h->sei.recovery_point.recovery_frame_cnt < 0) ||
+ h->avctx->skip_frame >= AVDISCARD_ALL) {
+ return SLICE_SKIPED;
}
}
- if (h->current_slice > 0) {
- if (h->ps.pps != (const PPS*)h->ps.pps_list[sl->pps_id]->data) {
+ if (!first_slice) {
+ const PPS *pps = (const PPS*)h->ps.pps_list[sl->pps_id]->data;
+
+ 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) {
+ av_log(h->avctx, AV_LOG_ERROR,
+ "SPS changed in the middle of the frame\n");
+ return AVERROR_INVALIDDATA;
+ }
+ }
+ if (h->current_slice == 0) {
+ ret = h264_field_start(h, sl, nal, first_slice);
+ if (ret < 0)
+ return ret;
+ } else {
if (h->picture_structure != sl->picture_structure ||
h->droppable != (nal->ref_idc == 0)) {
av_log(h->avctx, AV_LOG_ERROR,
@@ -1617,7 +1841,7 @@ int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl,
return AVERROR_INVALIDDATA;
}
- assert(h->mb_num == h->mb_width * h->mb_height);
+ av_assert1(h->mb_num == h->mb_width * h->mb_height);
if (sl->first_mb_addr << FIELD_OR_MBAFF_PICTURE(h) >= h->mb_num ||
sl->first_mb_addr >= h->mb_num) {
av_log(h->avctx, AV_LOG_ERROR, "first_mb_in_slice overflow\n");
@@ -1628,7 +1852,7 @@ int ff_h264_decode_slice_header(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);
ret = ff_h264_build_ref_list(h, sl);
if (ret < 0)
@@ -1649,6 +1873,8 @@ 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 != H264_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) ||
@@ -1656,7 +1882,7 @@ int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl,
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. */
@@ -1673,9 +1899,14 @@ 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++) {
@@ -2086,7 +2317,7 @@ 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,
@@ -2097,15 +2328,14 @@ 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)
@@ -2125,30 +2355,45 @@ 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;
sl->is_complex = FRAME_MBAFF(h) || h->picture_structure != PICT_FRAME ||
(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;
}
@@ -2178,9 +2423,11 @@ 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,
@@ -2217,6 +2464,8 @@ 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;
}
@@ -2257,14 +2506,15 @@ 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;
}
@@ -2309,7 +2559,15 @@ 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;
@@ -2321,12 +2579,15 @@ 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;
@@ -2347,8 +2608,10 @@ 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;
generated by cgit v1.2.3 (git 2.39.1) at 2024-07-17 20:21:10 +0000