Commit 857d7194 authored by Michael Niedermayer's avatar Michael Niedermayer

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  h264: add a pointer for weighted prediction temporary buffer

Conflicts:
	libavcodec/h264.c
	libavcodec/h264.h
Merged-by: 's avatarMichael Niedermayer <michaelni@gmx.at>
parents 6777aa63 a394959b
...@@ -671,9 +671,9 @@ static av_always_inline void mc_part_weighted(H264Context *h, int n, int square, ...@@ -671,9 +671,9 @@ static av_always_inline void mc_part_weighted(H264Context *h, int n, int square,
if (list0 && list1) { if (list0 && list1) {
/* don't optimize for luma-only case, since B-frames usually /* don't optimize for luma-only case, since B-frames usually
* use implicit weights => chroma too. */ * use implicit weights => chroma too. */
uint8_t *tmp_cb = s->obmc_scratchpad; uint8_t *tmp_cb = h->bipred_scratchpad;
uint8_t *tmp_cr = s->obmc_scratchpad + (16 << pixel_shift); uint8_t *tmp_cr = h->bipred_scratchpad + (16 << pixel_shift);
uint8_t *tmp_y = s->obmc_scratchpad + 16 * h->mb_uvlinesize; uint8_t *tmp_y = h->bipred_scratchpad + 16 * h->mb_uvlinesize;
int refn0 = h->ref_cache[0][scan8[n]]; int refn0 = h->ref_cache[0][scan8[n]];
int refn1 = h->ref_cache[1][scan8[n]]; int refn1 = h->ref_cache[1][scan8[n]];
...@@ -790,7 +790,7 @@ static void free_tables(H264Context *h, int free_rbsp) ...@@ -790,7 +790,7 @@ static void free_tables(H264Context *h, int free_rbsp)
continue; continue;
av_freep(&hx->top_borders[1]); av_freep(&hx->top_borders[1]);
av_freep(&hx->top_borders[0]); av_freep(&hx->top_borders[0]);
av_freep(&hx->s.obmc_scratchpad); av_freep(&hx->bipred_scratchpad);
if (free_rbsp) { if (free_rbsp) {
av_freep(&hx->rbsp_buffer[1]); av_freep(&hx->rbsp_buffer[1]);
av_freep(&hx->rbsp_buffer[0]); av_freep(&hx->rbsp_buffer[0]);
...@@ -915,8 +915,6 @@ int ff_h264_alloc_tables(H264Context *h) ...@@ -915,8 +915,6 @@ int ff_h264_alloc_tables(H264Context *h)
h->mb2br_xy[mb_xy] = 8 * (FMO ? mb_xy : (mb_xy % (2 * s->mb_stride))); h->mb2br_xy[mb_xy] = 8 * (FMO ? mb_xy : (mb_xy % (2 * s->mb_stride)));
} }
s->obmc_scratchpad = NULL;
if (!h->dequant4_coeff[0]) if (!h->dequant4_coeff[0])
init_dequant_tables(h); init_dequant_tables(h);
...@@ -944,7 +942,7 @@ static void clone_tables(H264Context *dst, H264Context *src, int i) ...@@ -944,7 +942,7 @@ static void clone_tables(H264Context *dst, H264Context *src, int i)
dst->mvd_table[1] = src->mvd_table[1] + i * 8 * 2 * s->mb_stride; dst->mvd_table[1] = src->mvd_table[1] + i * 8 * 2 * s->mb_stride;
dst->direct_table = src->direct_table; dst->direct_table = src->direct_table;
dst->list_counts = src->list_counts; dst->list_counts = src->list_counts;
dst->s.obmc_scratchpad = NULL; dst->bipred_scratchpad = NULL;
ff_h264_pred_init(&dst->hpc, src->s.codec_id, src->sps.bit_depth_luma, ff_h264_pred_init(&dst->hpc, src->s.codec_id, src->sps.bit_depth_luma,
src->sps.chroma_format_idc); src->sps.chroma_format_idc);
} }
...@@ -1214,15 +1212,13 @@ static int decode_update_thread_context(AVCodecContext *dst, ...@@ -1214,15 +1212,13 @@ static int decode_update_thread_context(AVCodecContext *dst,
} }
context_init(h); context_init(h);
/* frame_start may not be called for the next thread (if it's decoding
* a bottom field) so this has to be allocated here */
h->s.obmc_scratchpad = av_malloc(16 * 6 * s->linesize);
} }
for (i = 0; i < 2; i++) { for (i = 0; i < 2; i++) {
h->rbsp_buffer[i] = NULL; h->rbsp_buffer[i] = NULL;
h->rbsp_buffer_size[i] = 0; h->rbsp_buffer_size[i] = 0;
} }
h->bipred_scratchpad = NULL;
h->thread_context[0] = h; h->thread_context[0] = h;
...@@ -1230,6 +1226,11 @@ static int decode_update_thread_context(AVCodecContext *dst, ...@@ -1230,6 +1226,11 @@ static int decode_update_thread_context(AVCodecContext *dst,
s->dsp.clear_blocks(h->mb + (24 * 16 << h->pixel_shift)); s->dsp.clear_blocks(h->mb + (24 * 16 << h->pixel_shift));
} }
/* frame_start may not be called for the next thread (if it's decoding
* a bottom field) so this has to be allocated here */
if (!h->bipred_scratchpad)
h->bipred_scratchpad = av_malloc(16 * 6 * s->linesize);
// extradata/NAL handling // extradata/NAL handling
h->is_avc = h1->is_avc; h->is_avc = h1->is_avc;
...@@ -1321,8 +1322,8 @@ int ff_h264_frame_start(H264Context *h) ...@@ -1321,8 +1322,8 @@ int ff_h264_frame_start(H264Context *h)
/* can't be in alloc_tables because linesize isn't known there. /* can't be in alloc_tables because linesize isn't known there.
* FIXME: redo bipred weight to not require extra buffer? */ * FIXME: redo bipred weight to not require extra buffer? */
for (i = 0; i < s->slice_context_count; i++) for (i = 0; i < s->slice_context_count; i++)
if (h->thread_context[i] && !h->thread_context[i]->s.obmc_scratchpad) if (h->thread_context[i] && !h->thread_context[i]->bipred_scratchpad)
h->thread_context[i]->s.obmc_scratchpad = av_malloc(16 * 6 * s->linesize); h->thread_context[i]->bipred_scratchpad = av_malloc(16 * 6 * s->linesize);
/* Some macroblocks can be accessed before they're available in case /* Some macroblocks can be accessed before they're available in case
* of lost slices, MBAFF or threading. */ * of lost slices, MBAFF or threading. */
...@@ -2642,6 +2643,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0) ...@@ -2642,6 +2643,7 @@ static int decode_slice_header(H264Context *h, H264Context *h0)
"Could not allocate memory for h264\n"); "Could not allocate memory for h264\n");
return AVERROR(ENOMEM); return AVERROR(ENOMEM);
} }
h->bipred_scratchpad = NULL;
if (!HAVE_THREADS || !(s->avctx->active_thread_type & FF_THREAD_SLICE)) { if (!HAVE_THREADS || !(s->avctx->active_thread_type & FF_THREAD_SLICE)) {
if (context_init(h) < 0) { if (context_init(h) < 0) {
......
...@@ -595,6 +595,7 @@ typedef struct H264Context { ...@@ -595,6 +595,7 @@ typedef struct H264Context {
int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs int initial_cpb_removal_delay[32]; ///< Initial timestamps for CPBs
int cur_chroma_format_idc; int cur_chroma_format_idc;
uint8_t *bipred_scratchpad;
int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low int16_t slice_row[MAX_SLICES]; ///< to detect when MAX_SLICES is too low
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment