/**
* @file
- * H.264 / AVC / MPEG4 part10 codec.
+ * H.264 / AVC / MPEG-4 part10 codec.
* @author Michael Niedermayer <michaelni@gmx.at>
*/
av_fast_malloc(&sl->bipred_scratchpad, &sl->bipred_scratchpad_allocated, 16 * 6 * alloc_size);
// edge emu needs blocksize + filter length - 1
- // (= 21x21 for h264)
+ // (= 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],
h->next_outputed_poc = h1->next_outputed_poc;
memcpy(h->mmco, h1->mmco, sizeof(h->mmco));
- h->mmco_index = h1->mmco_index;
+ h->nb_mmco = h1->nb_mmco;
h->mmco_reset = h1->mmco_reset;
h->long_ref_count = h1->long_ref_count;
h->short_ref_count = h1->short_ref_count;
return 0;
if (!h->droppable) {
- err = ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
+ err = ff_h264_execute_ref_pic_marking(h, h->mmco, h->nb_mmco);
h->poc.prev_poc_msb = h->poc.poc_msb;
h->poc.prev_poc_lsb = h->poc.poc_lsb;
}
pic->mmco_reset = 0;
pic->recovered = 0;
+ pic->f->pict_type = h->slice_ctx[0].slice_type;
+
if (CONFIG_ERROR_RESILIENCE && h->enable_er)
ff_er_frame_start(&h->slice_ctx[0].er);
h->next_output_pic = NULL;
+ h->postpone_filter = 0;
+
assert(h->cur_pic_ptr->long_ref == 0);
return 0;
static int h264_slice_header_init(H264Context *h)
{
const SPS *sps = h->ps.sps;
- int nb_slices = (HAVE_THREADS &&
- h->avctx->active_thread_type & FF_THREAD_SLICE) ?
- h->avctx->thread_count : 1;
int i, ret;
ff_set_sar(h->avctx, sps->sar);
sps->chroma_format_idc);
ff_videodsp_init(&h->vdsp, sps->bit_depth_luma);
- if (nb_slices > H264_MAX_THREADS || (nb_slices > h->mb_height && h->mb_height)) {
- int max_slices;
- if (h->mb_height)
- max_slices = FFMIN(H264_MAX_THREADS, h->mb_height);
- else
- max_slices = H264_MAX_THREADS;
- av_log(h->avctx, AV_LOG_WARNING, "too many threads/slices %d,"
- " reducing to %d\n", nb_slices, max_slices);
- nb_slices = max_slices;
- }
- h->slice_context_count = nb_slices;
-
if (!HAVE_THREADS || !(h->avctx->active_thread_type & FF_THREAD_SLICE)) {
ret = ff_h264_slice_context_init(h, &h->slice_ctx[0]);
if (ret < 0) {
return ret;
}
} else {
- for (i = 0; i < h->slice_context_count; i++) {
+ for (i = 0; i < h->nb_slice_ctx; i++) {
H264SliceContext *sl = &h->slice_ctx[i];
sl->h264 = h;
return 0;
}
-/**
- * Decode a slice header.
- * This will (re)intialize the decoder and call h264_frame_start() as needed.
- *
- * @param h h264context
- *
- * @return 0 if okay, <0 if an error occurred, 1 if decoding must not be multithreaded
- */
-int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl)
+static int h264_slice_header_parse(H264Context *h, H264SliceContext *sl)
{
const SPS *sps;
const PPS *pps;
- unsigned int first_mb_in_slice;
unsigned int pps_id;
int ret;
- unsigned int slice_type, tmp, i, j;
+ unsigned int slice_type, tmp, i;
int last_pic_structure, last_pic_droppable;
int needs_reinit = 0;
int field_pic_flag, bottom_field_flag;
int frame_num, droppable, picture_structure;
int mb_aff_frame = 0;
- first_mb_in_slice = get_ue_golomb(&sl->gb);
+ sl->first_mb_addr = get_ue_golomb(&sl->gb);
- if (first_mb_in_slice == 0) { // FIXME better field boundary detection
+ 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 (slice_type > 9) {
av_log(h->avctx, AV_LOG_ERROR,
"slice type %d too large at %d\n",
- slice_type, first_mb_in_slice);
+ slice_type, sl->first_mb_addr);
return AVERROR_INVALIDDATA;
}
if (slice_type > 4) {
return AVERROR_INVALIDDATA;
}
- // to make a few old functions happy, it's wrong though
- if (!h->setup_finished)
- h->pict_type = sl->slice_type;
-
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);
ret = ff_generate_sliding_window_mmcos(h, 1);
if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE))
return ret;
- ret = ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
+ ret = ff_h264_execute_ref_pic_marking(h, h->mmco, h->nb_mmco);
if (ret < 0 && (h->avctx->err_recognition & AV_EF_EXPLODE))
return ret;
/* Error concealment: If a ref is missing, copy the previous ref
}
}
- assert(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");
- return AVERROR_INVALIDDATA;
- }
- sl->resync_mb_x = sl->mb_x = first_mb_in_slice % h->mb_width;
- sl->resync_mb_y = sl->mb_y = (first_mb_in_slice / h->mb_width) <<
- 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);
-
if (h->picture_structure == PICT_FRAME) {
h->curr_pic_num = h->poc.frame_num;
h->max_pic_num = 1 << sps->log2_max_frame_num;
}
}
- if (!h->setup_finished)
- ff_h264_init_poc(h->cur_pic_ptr->field_poc, &h->cur_pic_ptr->poc,
- sps, &h->poc, h->picture_structure, h->nal_ref_idc);
-
if (pps->redundant_pic_cnt_present)
sl->redundant_pic_count = get_ue_golomb(&sl->gb);
}
}
+ sl->pwt.use_weight = 0;
+ for (i = 0; i < 2; i++) {
+ sl->pwt.luma_weight_flag[i] = 0;
+ sl->pwt.chroma_weight_flag[i] = 0;
+ }
if ((pps->weighted_pred && sl->slice_type_nos == AV_PICTURE_TYPE_P) ||
(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);
- else if (pps->weighted_bipred_idc == 2 &&
- sl->slice_type_nos == AV_PICTURE_TYPE_B) {
- implicit_weight_table(h, sl, -1);
- } else {
- sl->pwt.use_weight = 0;
- for (i = 0; i < 2; i++) {
- sl->pwt.luma_weight_flag[i] = 0;
- sl->pwt.chroma_weight_flag[i] = 0;
- }
- }
// If frame-mt is enabled, only update mmco tables for the first slice
- // in a field. Subsequent slices can temporarily clobber h->mmco_index
+ // in a field. Subsequent slices can temporarily clobber h->nb_mmco
// or h->mmco, which will cause ref list mix-ups and decoding errors
// further down the line. This may break decoding if the first slice is
// corrupt, thus we only do this if frame-mt is enabled.
return AVERROR_INVALIDDATA;
}
- if (FRAME_MBAFF(h)) {
- ff_h264_fill_mbaff_ref_list(sl);
-
- if (pps->weighted_bipred_idc == 2 && sl->slice_type_nos == AV_PICTURE_TYPE_B) {
- implicit_weight_table(h, sl, 0);
- implicit_weight_table(h, sl, 1);
- }
- }
-
- if (sl->slice_type_nos == AV_PICTURE_TYPE_B && !sl->direct_spatial_mv_pred)
- ff_h264_direct_dist_scale_factor(h, sl);
- ff_h264_direct_ref_list_init(h, sl);
-
if (sl->slice_type_nos != AV_PICTURE_TYPE_I && pps->cabac) {
tmp = get_ue_golomb_31(&sl->gb);
if (tmp > 2) {
}
}
+ return 0;
+}
+
+/**
+ * Decode a slice header.
+ * This will (re)initialize the decoder and call h264_frame_start() as needed.
+ *
+ * @param h h264context
+ *
+ * @return 0 if okay, <0 if an error occurred
+ */
+int ff_h264_decode_slice_header(H264Context *h, H264SliceContext *sl)
+{
+ int i, j, ret = 0;
+
+ ret = h264_slice_header_parse(h, sl);
+ if (ret < 0)
+ return ret;
+
+ assert(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");
+ return AVERROR_INVALIDDATA;
+ }
+ sl->resync_mb_x = sl->mb_x = sl->first_mb_addr % h->mb_width;
+ sl->resync_mb_y = sl->mb_y = (sl->first_mb_addr / h->mb_width) <<
+ 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);
+
+ if (!h->setup_finished)
+ ff_h264_init_poc(h->cur_pic_ptr->field_poc, &h->cur_pic_ptr->poc,
+ h->ps.sps, &h->poc, h->picture_structure, h->nal_ref_idc);
+
+ ret = ff_h264_build_ref_list(h, sl);
+ if (ret < 0)
+ return ret;
+
+ if (h->ps.pps->weighted_bipred_idc == 2 &&
+ sl->slice_type_nos == AV_PICTURE_TYPE_B) {
+ implicit_weight_table(h, sl, -1);
+ if (FRAME_MBAFF(h)) {
+ implicit_weight_table(h, sl, 0);
+ implicit_weight_table(h, sl, 1);
+ }
+ }
+
+ if (sl->slice_type_nos == AV_PICTURE_TYPE_B && !sl->direct_spatial_mv_pred)
+ ff_h264_direct_dist_scale_factor(h, sl);
+ ff_h264_direct_ref_list_init(h, sl);
+
if (h->avctx->skip_loop_filter >= AVDISCARD_ALL ||
(h->avctx->skip_loop_filter >= AVDISCARD_NONKEY &&
sl->slice_type_nos != AV_PICTURE_TYPE_I) ||
h->nal_ref_idc == 0))
sl->deblocking_filter = 0;
- if (sl->deblocking_filter == 1 && h->max_contexts > 1) {
+ if (sl->deblocking_filter == 1 && h->nb_slice_ctx > 1) {
if (h->avctx->flags2 & AV_CODEC_FLAG2_FAST) {
/* Cheat slightly for speed:
* Do not bother to deblock across slices. */
sl->deblocking_filter = 2;
} else {
- h->max_contexts = 1;
- if (!h->single_decode_warning) {
- av_log(h->avctx, AV_LOG_INFO,
- "Cannot parallelize deblocking type 1, decoding such frames in sequential order\n");
- h->single_decode_warning = 1;
- }
- if (sl != h->slice_ctx) {
- av_log(h->avctx, AV_LOG_ERROR,
- "Deblocking switched inside frame.\n");
- return 1;
- }
+ h->postpone_filter = 1;
}
}
sl->qp_thresh = 15 -
FFMIN(sl->slice_alpha_c0_offset, sl->slice_beta_offset) -
FFMAX3(0,
- pps->chroma_qp_index_offset[0],
- pps->chroma_qp_index_offset[1]) +
- 6 * (sps->bit_depth_luma - 8);
+ h->ps.pps->chroma_qp_index_offset[0],
+ h->ps.pps->chroma_qp_index_offset[1]) +
+ 6 * (h->ps.sps->bit_depth_luma - 8);
sl->slice_num = ++h->current_slice;
if (sl->slice_num >= MAX_SLICES) {
for (j = 0; j < 2; j++) {
int id_list[16];
- int *ref2frm = sl->ref2frm[sl->slice_num & (MAX_SLICES - 1)][j];
+ int *ref2frm = h->ref2frm[sl->slice_num & (MAX_SLICES - 1)][j];
for (i = 0; i < 16; i++) {
id_list[i] = 60;
if (j < sl->list_count && i < sl->ref_count[j] &&
if (h->avctx->debug & FF_DEBUG_PICT_INFO) {
av_log(h->avctx, AV_LOG_DEBUG,
- "slice:%d %s mb:%d %c%s%s pps:%u frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d:%d:%d weight:%d%s %s\n",
+ "slice:%d %s mb:%d %c%s%s frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d:%d:%d weight:%d%s %s\n",
sl->slice_num,
(h->picture_structure == PICT_FRAME ? "F" : h->picture_structure == PICT_TOP_FIELD ? "T" : "B"),
- first_mb_in_slice,
+ sl->mb_y * h->mb_width + sl->mb_x,
av_get_picture_type_char(sl->slice_type),
sl->slice_type_fixed ? " fix" : "",
h->nal_unit_type == NAL_IDR_SLICE ? " IDR" : "",
- pps_id, h->poc.frame_num,
+ h->poc.frame_num,
h->cur_pic_ptr->field_poc[0],
h->cur_pic_ptr->field_poc[1],
sl->ref_count[0], sl->ref_count[1],
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] = sl->ref2frm[h->slice_table[top_xy] & (MAX_SLICES - 1)][0] + (MB_MBAFF(sl) ? 20 : 2);
+ int (*ref2frm)[64] = h->ref2frm[h->slice_table[top_xy] & (MAX_SLICES - 1)][0] + (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]];
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] = sl->ref2frm[h->slice_table[left_xy[LTOP]] & (MAX_SLICES - 1)][0] + (MB_MBAFF(sl) ? 20 : 2);
+ int (*ref2frm)[64] = h->ref2frm[h->slice_table[left_xy[LTOP]] & (MAX_SLICES - 1)][0] + (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]);
{
int8_t *ref = &h->cur_pic.ref_index[list][4 * mb_xy];
- int (*ref2frm)[64] = sl->ref2frm[sl->slice_num & (MAX_SLICES - 1)][0] + (MB_MBAFF(sl) ? 20 : 2);
+ 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;
AV_WN32A(&ref_cache[0 * 8], ref01);
top_xy = mb_xy - (h->mb_stride << MB_FIELD(sl));
- /* Wow, what a mess, why didn't they simplify the interlacing & intra
- * stuff, I can't imagine that these complex rules are worth it. */
-
left_xy[LBOT] = left_xy[LTOP] = mb_xy - 1;
if (FRAME_MBAFF(h)) {
const int left_mb_field_flag = IS_INTERLACED(h->cur_pic.mb_type[mb_xy - 1]);
const int pixel_shift = h->pixel_shift;
const int block_h = 16 >> h->chroma_y_shift;
+ if (h->postpone_filter)
+ return;
+
if (sl->deblocking_filter) {
for (mb_x = start_x; mb_x < end_x; mb_x++)
for (mb_y = end_mb_y - FRAME_MBAFF(h); mb_y <= end_mb_y; mb_y++) {
int mb_xy, mb_type;
mb_xy = sl->mb_xy = mb_x + mb_y * h->mb_stride;
- sl->slice_num = h->slice_table[mb_xy];
mb_type = h->cur_pic.mb_type[mb_xy];
- sl->list_count = h->list_counts[mb_xy];
if (FRAME_MBAFF(h))
sl->mb_mbaff =
H264SliceContext *sl = arg;
const H264Context *h = sl->h264;
int lf_x_start = sl->mb_x;
+ int orig_deblock = sl->deblocking_filter;
int ret;
sl->linesize = h->cur_pic_ptr->f->linesize[0];
sl->mb_skip_run = -1;
+ if (h->postpone_filter)
+ sl->deblocking_filter = 0;
+
sl->is_complex = FRAME_MBAFF(h) || h->picture_structure != PICT_FRAME ||
avctx->codec_id != AV_CODEC_ID_H264 ||
(CONFIG_GRAY && (h->flags & AV_CODEC_FLAG_GRAY));
sl->mb_y, ER_MB_END);
if (sl->mb_x >= lf_x_start)
loop_filter(h, sl, lf_x_start, sl->mb_x + 1);
- return 0;
+ goto finish;
}
if (ret < 0 || sl->cabac.bytestream > sl->cabac.bytestream_end + 2) {
av_log(h->avctx, AV_LOG_ERROR,
sl->mb_y, ER_MB_END);
if (sl->mb_x > lf_x_start)
loop_filter(h, sl, lf_x_start, sl->mb_x);
- return 0;
+ 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);
- return 0;
+ 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);
if (sl->mb_x > lf_x_start)
loop_filter(h, sl, lf_x_start, sl->mb_x);
- return 0;
+ goto finish;
} else {
er_add_slice(sl, sl->resync_mb_x, sl->resync_mb_y, sl->mb_x,
sl->mb_y, ER_MB_ERROR);
}
}
}
+
+finish:
+ sl->deblocking_filter = orig_deblock;
+ return 0;
}
/**
int ret;
h->slice_ctx[0].next_slice_idx = h->mb_width * h->mb_height;
+ h->postpone_filter = 0;
ret = decode_slice(avctx, &h->slice_ctx[0]);
h->mb_y = h->slice_ctx[0].mb_y;
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 (h->postpone_filter) {
+ h->postpone_filter = 0;
+
+ for (i = 0; i < context_count; i++) {
+ int y_end, x_end;
+
+ sl = &h->slice_ctx[i];
+ y_end = FFMIN(sl->mb_y + 1, h->mb_height);
+ x_end = (sl->mb_y >= h->mb_height) ? h->mb_width : sl->mb_x;
+
+ for (j = sl->resync_mb_y; j < y_end; j += 1 + FIELD_OR_MBAFF_PICTURE(h)) {
+ sl->mb_y = j;
+ loop_filter(h, sl, j > sl->resync_mb_y ? 0 : sl->resync_mb_x,
+ j == y_end - 1 ? x_end : h->mb_width);
+ }
+ }
+ }
}
return 0;
projects / ffmpeg / blobdiff