* 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
*/
*/
#include "libavutil/imgutils.h"
+#include "libavutil/opt.h"
#include "internal.h"
#include "dsputil.h"
#include "avcodec.h"
static const enum PixelFormat hwaccel_pixfmt_list_h264_jpeg_420[] = {
PIX_FMT_DXVA2_VLD,
PIX_FMT_VAAPI_VLD,
+ PIX_FMT_VDA_VLD,
PIX_FMT_YUVJ420P,
PIX_FMT_NONE
};
}
return 0;
-} //FIXME cleanup like ff_h264_check_intra_pred_mode
+} //FIXME cleanup like check_intra_pred_mode
-/**
- * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
- */
-int ff_h264_check_intra_pred_mode(H264Context *h, int mode){
+static int check_intra_pred_mode(H264Context *h, int mode, int is_chroma){
MpegEncContext * const s = &h->s;
static const int8_t top [7]= {LEFT_DC_PRED8x8, 1,-1,-1};
static const int8_t left[7]= { TOP_DC_PRED8x8,-1, 2,-1,DC_128_PRED8x8};
if((h->left_samples_available&0x8080) != 0x8080){
mode= left[ mode ];
- if(h->left_samples_available&0x8080){ //mad cow disease mode, aka MBAFF + constrained_intra_pred
+ if(is_chroma && (h->left_samples_available&0x8080)){ //mad cow disease mode, aka MBAFF + constrained_intra_pred
mode= ALZHEIMER_DC_L0T_PRED8x8 + (!(h->left_samples_available&0x8000)) + 2*(mode == DC_128_PRED8x8);
}
if(mode<0){
return mode;
}
+/**
+ * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
+ */
+int ff_h264_check_intra16x16_pred_mode(H264Context *h, int mode)
+{
+ return check_intra_pred_mode(h, mode, 0);
+}
+
+/**
+ * checks if the top & left blocks are available if needed & changes the dc mode so it only uses the available blocks.
+ */
+int ff_h264_check_intra_chroma_pred_mode(H264Context *h, int mode)
+{
+ return check_intra_pred_mode(h, mode, 1);
+}
+
+
const uint8_t *ff_h264_decode_nal(H264Context *h, const uint8_t *src, int *dst_length, int *consumed, int length){
int i, si, di;
uint8_t *dst;
i-= RS;
}
- if(i>=length-1){ //no escaped 0
- *dst_length= length;
- *consumed= length+1; //+1 for the header
- return src;
- }
-
bufidx = h->nal_unit_type == NAL_DPC ? 1 : 0; // use second escape buffer for inter data
- av_fast_malloc(&h->rbsp_buffer[bufidx], &h->rbsp_buffer_size[bufidx], length+FF_INPUT_BUFFER_PADDING_SIZE);
+ si=h->rbsp_buffer_size[bufidx];
+ av_fast_malloc(&h->rbsp_buffer[bufidx], &h->rbsp_buffer_size[bufidx], length+FF_INPUT_BUFFER_PADDING_SIZE+MAX_MBPAIR_SIZE);
dst= h->rbsp_buffer[bufidx];
+ if(si != h->rbsp_buffer_size[bufidx])
+ memset(dst + length, 0, FF_INPUT_BUFFER_PADDING_SIZE+MAX_MBPAIR_SIZE);
if (dst == NULL){
return NULL;
}
+ if(i>=length-1){ //no escaped 0
+ *dst_length= length;
+ *consumed= length+1; //+1 for the header
+ if(h->s.avctx->flags2 & CODEC_FLAG2_FAST){
+ return src;
+ }else{
+ memcpy(dst, src, length);
+ return dst;
+ }
+ }
+
//printf("decoding esc\n");
memcpy(dst, src, i);
si=di=i;
}
#endif
-static inline void mc_dir_part(H264Context *h, Picture *pic, int n, int square, int chroma_height, int delta, int list,
- uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
- int src_x_offset, int src_y_offset,
- qpel_mc_func *qpix_op, h264_chroma_mc_func chroma_op,
- int pixel_shift, int chroma444){
+static av_always_inline void
+mc_dir_part(H264Context *h, Picture *pic, int n, int square,
+ int height, int delta, int list,
+ uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
+ int src_x_offset, int src_y_offset,
+ qpel_mc_func *qpix_op, h264_chroma_mc_func chroma_op,
+ int pixel_shift, int chroma_idc)
+{
MpegEncContext * const s = &h->s;
const int mx= h->mv_cache[list][ scan8[n] ][0] + src_x_offset*8;
int my= h->mv_cache[list][ scan8[n] ][1] + src_y_offset*8;
const int full_my= my>>2;
const int pic_width = 16*s->mb_width;
const int pic_height = 16*s->mb_height >> MB_FIELD;
+ int ysh;
if(mx&7) extra_width -= 3;
if(my&7) extra_height -= 3;
|| full_my < 0-extra_height
|| full_mx + 16/*FIXME*/ > pic_width + extra_width
|| full_my + 16/*FIXME*/ > pic_height + extra_height){
- s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_y - (2 << pixel_shift) - 2*h->mb_linesize, h->mb_linesize, 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, pic_width, pic_height);
+ s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_y - (2 << pixel_shift) - 2*h->mb_linesize, h->mb_linesize,
+ 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, pic_width, pic_height);
src_y= s->edge_emu_buffer + (2 << pixel_shift) + 2*h->mb_linesize;
emu=1;
}
if(CONFIG_GRAY && s->flags&CODEC_FLAG_GRAY) return;
- if(chroma444){
+ if(chroma_idc == 3 /* yuv444 */){
src_cb = pic->f.data[1] + offset;
if(emu){
s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cb - (2 << pixel_shift) - 2*h->mb_linesize, h->mb_linesize,
return;
}
- if(MB_FIELD){
+ ysh = 3 - (chroma_idc == 2 /* yuv422 */);
+ if(chroma_idc == 1 /* yuv420 */ && MB_FIELD){
// chroma offset when predicting from a field of opposite parity
my += 2 * ((s->mb_y & 1) - (pic->f.reference - 1));
emu |= (my>>3) < 0 || (my>>3) + 8 >= (pic_height>>1);
}
- src_cb = pic->f.data[1] + ((mx >> 3) << pixel_shift) + (my >> 3) * h->mb_uvlinesize;
- src_cr = pic->f.data[2] + ((mx >> 3) << pixel_shift) + (my >> 3) * h->mb_uvlinesize;
+
+ src_cb = pic->f.data[1] + ((mx >> 3) << pixel_shift) + (my >> ysh) * h->mb_uvlinesize;
+ src_cr = pic->f.data[2] + ((mx >> 3) << pixel_shift) + (my >> ysh) * h->mb_uvlinesize;
if(emu){
- s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cb, h->mb_uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), pic_width>>1, pic_height>>1);
+ s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cb, h->mb_uvlinesize,
+ 9, 8 * chroma_idc + 1, (mx >> 3), (my >> ysh),
+ pic_width >> 1, pic_height >> (chroma_idc == 1 /* yuv420 */));
src_cb= s->edge_emu_buffer;
}
- chroma_op(dest_cb, src_cb, h->mb_uvlinesize, chroma_height, mx&7, my&7);
+ chroma_op(dest_cb, src_cb, h->mb_uvlinesize, height >> (chroma_idc == 1 /* yuv420 */),
+ mx&7, (my << (chroma_idc == 2 /* yuv422 */)) &7);
if(emu){
- s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cr, h->mb_uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), pic_width>>1, pic_height>>1);
+ s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cr, h->mb_uvlinesize,
+ 9, 8 * chroma_idc + 1, (mx >> 3), (my >> ysh),
+ pic_width >> 1, pic_height >> (chroma_idc == 1 /* yuv420 */));
src_cr= s->edge_emu_buffer;
}
- chroma_op(dest_cr, src_cr, h->mb_uvlinesize, chroma_height, mx&7, my&7);
+ chroma_op(dest_cr, src_cr, h->mb_uvlinesize, height >> (chroma_idc == 1 /* yuv420 */),
+ mx&7, (my << (chroma_idc == 2 /* yuv422 */)) &7);
}
-static inline void mc_part_std(H264Context *h, int n, int square, int chroma_height, int delta,
- uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
- int x_offset, int y_offset,
- qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
- qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg,
- int list0, int list1, int pixel_shift, int chroma444){
+static av_always_inline void
+mc_part_std(H264Context *h, int n, int square, int height, int delta,
+ uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
+ int x_offset, int y_offset,
+ qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
+ qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg,
+ int list0, int list1, int pixel_shift, int chroma_idc)
+{
MpegEncContext * const s = &h->s;
qpel_mc_func *qpix_op= qpix_put;
h264_chroma_mc_func chroma_op= chroma_put;
dest_y += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize;
- if(chroma444){
+ if (chroma_idc == 3 /* yuv444 */) {
dest_cb += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize;
dest_cr += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize;
- }else{
+ } else if (chroma_idc == 2 /* yuv422 */) {
+ dest_cb += ( x_offset << pixel_shift) + 2*y_offset*h->mb_uvlinesize;
+ dest_cr += ( x_offset << pixel_shift) + 2*y_offset*h->mb_uvlinesize;
+ } else /* yuv420 */ {
dest_cb += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize;
dest_cr += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize;
}
if(list0){
Picture *ref= &h->ref_list[0][ h->ref_cache[0][ scan8[n] ] ];
- mc_dir_part(h, ref, n, square, chroma_height, delta, 0,
+ mc_dir_part(h, ref, n, square, height, delta, 0,
dest_y, dest_cb, dest_cr, x_offset, y_offset,
- qpix_op, chroma_op, pixel_shift, chroma444);
+ qpix_op, chroma_op, pixel_shift, chroma_idc);
qpix_op= qpix_avg;
chroma_op= chroma_avg;
if(list1){
Picture *ref= &h->ref_list[1][ h->ref_cache[1][ scan8[n] ] ];
- mc_dir_part(h, ref, n, square, chroma_height, delta, 1,
+ mc_dir_part(h, ref, n, square, height, delta, 1,
dest_y, dest_cb, dest_cr, x_offset, y_offset,
- qpix_op, chroma_op, pixel_shift, chroma444);
+ qpix_op, chroma_op, pixel_shift, chroma_idc);
}
}
-static inline void mc_part_weighted(H264Context *h, int n, int square, int chroma_height, int delta,
- uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
- int x_offset, int y_offset,
- qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
- h264_weight_func luma_weight_op, h264_weight_func chroma_weight_op,
- h264_biweight_func luma_weight_avg, h264_biweight_func chroma_weight_avg,
- int list0, int list1, int pixel_shift, int chroma444){
+static av_always_inline void
+mc_part_weighted(H264Context *h, int n, int square, int height, int delta,
+ uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
+ int x_offset, int y_offset,
+ qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
+ h264_weight_func luma_weight_op, h264_weight_func chroma_weight_op,
+ h264_biweight_func luma_weight_avg, h264_biweight_func chroma_weight_avg,
+ int list0, int list1, int pixel_shift, int chroma_idc){
MpegEncContext * const s = &h->s;
+ int chroma_height;
dest_y += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize;
- if(chroma444){
+ if (chroma_idc == 3 /* yuv444 */) {
+ chroma_height = height;
chroma_weight_avg = luma_weight_avg;
chroma_weight_op = luma_weight_op;
dest_cb += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize;
dest_cr += (2*x_offset << pixel_shift) + 2*y_offset*h->mb_linesize;
- }else{
+ } else if (chroma_idc == 2 /* yuv422 */) {
+ chroma_height = height;
+ dest_cb += ( x_offset << pixel_shift) + 2*y_offset*h->mb_uvlinesize;
+ dest_cr += ( x_offset << pixel_shift) + 2*y_offset*h->mb_uvlinesize;
+ } else /* yuv420 */ {
+ chroma_height = height >> 1;
dest_cb += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize;
dest_cr += ( x_offset << pixel_shift) + y_offset*h->mb_uvlinesize;
}
int refn0 = h->ref_cache[0][ scan8[n] ];
int refn1 = h->ref_cache[1][ scan8[n] ];
- mc_dir_part(h, &h->ref_list[0][refn0], n, square, chroma_height, delta, 0,
+ mc_dir_part(h, &h->ref_list[0][refn0], n, square, height, delta, 0,
dest_y, dest_cb, dest_cr,
- x_offset, y_offset, qpix_put, chroma_put, pixel_shift, chroma444);
- mc_dir_part(h, &h->ref_list[1][refn1], n, square, chroma_height, delta, 1,
+ x_offset, y_offset, qpix_put, chroma_put,
+ pixel_shift, chroma_idc);
+ mc_dir_part(h, &h->ref_list[1][refn1], n, square, height, delta, 1,
tmp_y, tmp_cb, tmp_cr,
- x_offset, y_offset, qpix_put, chroma_put, pixel_shift, chroma444);
+ x_offset, y_offset, qpix_put, chroma_put,
+ pixel_shift, chroma_idc);
if(h->use_weight == 2){
int weight0 = h->implicit_weight[refn0][refn1][s->mb_y&1];
int weight1 = 64 - weight0;
- luma_weight_avg( dest_y, tmp_y, h-> mb_linesize, 5, weight0, weight1, 0);
- chroma_weight_avg(dest_cb, tmp_cb, h->mb_uvlinesize, 5, weight0, weight1, 0);
- chroma_weight_avg(dest_cr, tmp_cr, h->mb_uvlinesize, 5, weight0, weight1, 0);
+ luma_weight_avg( dest_y, tmp_y, h-> mb_linesize,
+ height, 5, weight0, weight1, 0);
+ chroma_weight_avg(dest_cb, tmp_cb, h->mb_uvlinesize,
+ chroma_height, 5, weight0, weight1, 0);
+ chroma_weight_avg(dest_cr, tmp_cr, h->mb_uvlinesize,
+ chroma_height, 5, weight0, weight1, 0);
}else{
- luma_weight_avg(dest_y, tmp_y, h->mb_linesize, h->luma_log2_weight_denom,
+ luma_weight_avg(dest_y, tmp_y, h->mb_linesize, height, h->luma_log2_weight_denom,
h->luma_weight[refn0][0][0] , h->luma_weight[refn1][1][0],
h->luma_weight[refn0][0][1] + h->luma_weight[refn1][1][1]);
- chroma_weight_avg(dest_cb, tmp_cb, h->mb_uvlinesize, h->chroma_log2_weight_denom,
+ chroma_weight_avg(dest_cb, tmp_cb, h->mb_uvlinesize, chroma_height, h->chroma_log2_weight_denom,
h->chroma_weight[refn0][0][0][0] , h->chroma_weight[refn1][1][0][0],
h->chroma_weight[refn0][0][0][1] + h->chroma_weight[refn1][1][0][1]);
- chroma_weight_avg(dest_cr, tmp_cr, h->mb_uvlinesize, h->chroma_log2_weight_denom,
+ chroma_weight_avg(dest_cr, tmp_cr, h->mb_uvlinesize, chroma_height, h->chroma_log2_weight_denom,
h->chroma_weight[refn0][0][1][0] , h->chroma_weight[refn1][1][1][0],
h->chroma_weight[refn0][0][1][1] + h->chroma_weight[refn1][1][1][1]);
}
int list = list1 ? 1 : 0;
int refn = h->ref_cache[list][ scan8[n] ];
Picture *ref= &h->ref_list[list][refn];
- mc_dir_part(h, ref, n, square, chroma_height, delta, list,
+ mc_dir_part(h, ref, n, square, height, delta, list,
dest_y, dest_cb, dest_cr, x_offset, y_offset,
- qpix_put, chroma_put, pixel_shift, chroma444);
+ qpix_put, chroma_put, pixel_shift, chroma_idc);
- luma_weight_op(dest_y, h->mb_linesize, h->luma_log2_weight_denom,
+ luma_weight_op(dest_y, h->mb_linesize, height, h->luma_log2_weight_denom,
h->luma_weight[refn][list][0], h->luma_weight[refn][list][1]);
if(h->use_weight_chroma){
- chroma_weight_op(dest_cb, h->mb_uvlinesize, h->chroma_log2_weight_denom,
+ chroma_weight_op(dest_cb, h->mb_uvlinesize, chroma_height, h->chroma_log2_weight_denom,
h->chroma_weight[refn][list][0][0], h->chroma_weight[refn][list][0][1]);
- chroma_weight_op(dest_cr, h->mb_uvlinesize, h->chroma_log2_weight_denom,
+ chroma_weight_op(dest_cr, h->mb_uvlinesize, chroma_height, h->chroma_log2_weight_denom,
h->chroma_weight[refn][list][1][0], h->chroma_weight[refn][list][1][1]);
}
}
}
-static inline void mc_part(H264Context *h, int n, int square, int chroma_height, int delta,
- uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
- int x_offset, int y_offset,
- qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
- qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg,
- h264_weight_func *weight_op, h264_biweight_func *weight_avg,
- int list0, int list1, int pixel_shift, int chroma444){
+static av_always_inline void
+mc_part(H264Context *h, int n, int square, int height, int delta,
+ uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
+ int x_offset, int y_offset,
+ qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
+ qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg,
+ h264_weight_func *weight_op, h264_biweight_func *weight_avg,
+ int list0, int list1, int pixel_shift, int chroma_idc)
+{
if((h->use_weight==2 && list0 && list1
&& (h->implicit_weight[ h->ref_cache[0][scan8[n]] ][ h->ref_cache[1][scan8[n]] ][h->s.mb_y&1] != 32))
|| h->use_weight==1)
- mc_part_weighted(h, n, square, chroma_height, delta, dest_y, dest_cb, dest_cr,
+ mc_part_weighted(h, n, square, height, delta, dest_y, dest_cb, dest_cr,
x_offset, y_offset, qpix_put, chroma_put,
- weight_op[0], weight_op[3], weight_avg[0],
- weight_avg[3], list0, list1, pixel_shift, chroma444);
+ weight_op[0], weight_op[1], weight_avg[0],
+ weight_avg[1], list0, list1, pixel_shift, chroma_idc);
else
- mc_part_std(h, n, square, chroma_height, delta, dest_y, dest_cb, dest_cr,
+ mc_part_std(h, n, square, height, delta, dest_y, dest_cb, dest_cr,
x_offset, y_offset, qpix_put, chroma_put, qpix_avg,
- chroma_avg, list0, list1, pixel_shift, chroma444);
+ chroma_avg, list0, list1, pixel_shift, chroma_idc);
}
-static inline void prefetch_motion(H264Context *h, int list, int pixel_shift, int chroma444){
+static av_always_inline void
+prefetch_motion(H264Context *h, int list, int pixel_shift, int chroma_idc)
+{
/* fetch pixels for estimated mv 4 macroblocks ahead
* optimized for 64byte cache lines */
MpegEncContext * const s = &h->s;
uint8_t **src = h->ref_list[list][refn].f.data;
int off= (mx << pixel_shift) + (my + (s->mb_x&3)*4)*h->mb_linesize + (64 << pixel_shift);
s->dsp.prefetch(src[0]+off, s->linesize, 4);
- if(chroma444){
+ if (chroma_idc == 3 /* yuv444 */) {
s->dsp.prefetch(src[1]+off, s->linesize, 4);
s->dsp.prefetch(src[2]+off, s->linesize, 4);
}else{
- off= ((mx>>1) << pixel_shift) + ((my>>1) + (s->mb_x&7))*s->uvlinesize + (64 << pixel_shift);
+ off= (((mx>>1)+64)<<pixel_shift) + ((my>>1) + (s->mb_x&7))*s->uvlinesize;
s->dsp.prefetch(src[1]+off, src[2]-src[1], 2);
}
}
qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put),
qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg),
h264_weight_func *weight_op, h264_biweight_func *weight_avg,
- int pixel_shift, int chroma444){
+ int pixel_shift, int chroma_idc)
+{
MpegEncContext * const s = &h->s;
const int mb_xy= h->mb_xy;
const int mb_type = s->current_picture.f.mb_type[mb_xy];
assert(IS_INTER(mb_type));
- if(HAVE_PTHREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME))
+ if(HAVE_THREADS && (s->avctx->active_thread_type & FF_THREAD_FRAME))
await_references(h);
- prefetch_motion(h, 0, pixel_shift, chroma444);
+ prefetch_motion(h, 0, pixel_shift, chroma_idc);
if(IS_16X16(mb_type)){
- mc_part(h, 0, 1, 8, 0, dest_y, dest_cb, dest_cr, 0, 0,
+ mc_part(h, 0, 1, 16, 0, dest_y, dest_cb, dest_cr, 0, 0,
qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0],
weight_op, weight_avg,
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1),
- pixel_shift, chroma444);
+ pixel_shift, chroma_idc);
}else if(IS_16X8(mb_type)){
- mc_part(h, 0, 0, 4, 8 << pixel_shift, dest_y, dest_cb, dest_cr, 0, 0,
+ mc_part(h, 0, 0, 8, 8 << pixel_shift, dest_y, dest_cb, dest_cr, 0, 0,
qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0],
- &weight_op[1], &weight_avg[1],
+ weight_op, weight_avg,
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1),
- pixel_shift, chroma444);
- mc_part(h, 8, 0, 4, 8 << pixel_shift, dest_y, dest_cb, dest_cr, 0, 4,
+ pixel_shift, chroma_idc);
+ mc_part(h, 8, 0, 8, 8 << pixel_shift, dest_y, dest_cb, dest_cr, 0, 4,
qpix_put[1], chroma_put[0], qpix_avg[1], chroma_avg[0],
- &weight_op[1], &weight_avg[1],
+ weight_op, weight_avg,
IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1),
- pixel_shift, chroma444);
+ pixel_shift, chroma_idc);
}else if(IS_8X16(mb_type)){
- mc_part(h, 0, 0, 8, 8*h->mb_linesize, dest_y, dest_cb, dest_cr, 0, 0,
+ mc_part(h, 0, 0, 16, 8*h->mb_linesize, dest_y, dest_cb, dest_cr, 0, 0,
qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
- &weight_op[2], &weight_avg[2],
+ &weight_op[1], &weight_avg[1],
IS_DIR(mb_type, 0, 0), IS_DIR(mb_type, 0, 1),
- pixel_shift, chroma444);
- mc_part(h, 4, 0, 8, 8*h->mb_linesize, dest_y, dest_cb, dest_cr, 4, 0,
+ pixel_shift, chroma_idc);
+ mc_part(h, 4, 0, 16, 8*h->mb_linesize, dest_y, dest_cb, dest_cr, 4, 0,
qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
- &weight_op[2], &weight_avg[2],
+ &weight_op[1], &weight_avg[1],
IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1),
- pixel_shift, chroma444);
+ pixel_shift, chroma_idc);
}else{
int i;
int y_offset= (i&2)<<1;
if(IS_SUB_8X8(sub_mb_type)){
- mc_part(h, n, 1, 4, 0, dest_y, dest_cb, dest_cr, x_offset, y_offset,
+ mc_part(h, n, 1, 8, 0, dest_y, dest_cb, dest_cr, x_offset, y_offset,
qpix_put[1], chroma_put[1], qpix_avg[1], chroma_avg[1],
- &weight_op[3], &weight_avg[3],
+ &weight_op[1], &weight_avg[1],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
- pixel_shift, chroma444);
+ pixel_shift, chroma_idc);
}else if(IS_SUB_8X4(sub_mb_type)){
- mc_part(h, n , 0, 2, 4 << pixel_shift, dest_y, dest_cb, dest_cr, x_offset, y_offset,
+ mc_part(h, n , 0, 4, 4 << pixel_shift, dest_y, dest_cb, dest_cr, x_offset, y_offset,
qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1],
- &weight_op[4], &weight_avg[4],
+ &weight_op[1], &weight_avg[1],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
- pixel_shift, chroma444);
- mc_part(h, n+2, 0, 2, 4 << pixel_shift, dest_y, dest_cb, dest_cr, x_offset, y_offset+2,
+ pixel_shift, chroma_idc);
+ mc_part(h, n+2, 0, 4, 4 << pixel_shift, dest_y, dest_cb, dest_cr, x_offset, y_offset+2,
qpix_put[2], chroma_put[1], qpix_avg[2], chroma_avg[1],
- &weight_op[4], &weight_avg[4],
+ &weight_op[1], &weight_avg[1],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
- pixel_shift, chroma444);
+ pixel_shift, chroma_idc);
}else if(IS_SUB_4X8(sub_mb_type)){
- mc_part(h, n , 0, 4, 4*h->mb_linesize, dest_y, dest_cb, dest_cr, x_offset, y_offset,
+ mc_part(h, n , 0, 8, 4*h->mb_linesize, dest_y, dest_cb, dest_cr, x_offset, y_offset,
qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
- &weight_op[5], &weight_avg[5],
+ &weight_op[2], &weight_avg[2],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
- pixel_shift, chroma444);
- mc_part(h, n+1, 0, 4, 4*h->mb_linesize, dest_y, dest_cb, dest_cr, x_offset+2, y_offset,
+ pixel_shift, chroma_idc);
+ mc_part(h, n+1, 0, 8, 4*h->mb_linesize, dest_y, dest_cb, dest_cr, x_offset+2, y_offset,
qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
- &weight_op[5], &weight_avg[5],
+ &weight_op[2], &weight_avg[2],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
- pixel_shift, chroma444);
+ pixel_shift, chroma_idc);
}else{
int j;
assert(IS_SUB_4X4(sub_mb_type));
for(j=0; j<4; j++){
int sub_x_offset= x_offset + 2*(j&1);
int sub_y_offset= y_offset + (j&2);
- mc_part(h, n+j, 1, 2, 0, dest_y, dest_cb, dest_cr, sub_x_offset, sub_y_offset,
+ mc_part(h, n+j, 1, 4, 0, dest_y, dest_cb, dest_cr, sub_x_offset, sub_y_offset,
qpix_put[2], chroma_put[2], qpix_avg[2], chroma_avg[2],
- &weight_op[6], &weight_avg[6],
+ &weight_op[2], &weight_avg[2],
IS_DIR(sub_mb_type, 0, 0), IS_DIR(sub_mb_type, 0, 1),
- pixel_shift, chroma444);
+ pixel_shift, chroma_idc);
}
}
}
}
- prefetch_motion(h, 1, pixel_shift, chroma444);
+ prefetch_motion(h, 1, pixel_shift, chroma_idc);
+}
+
+static av_always_inline void
+hl_motion_420(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
+ qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put),
+ qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg),
+ h264_weight_func *weight_op, h264_biweight_func *weight_avg,
+ int pixel_shift)
+{
+ hl_motion(h, dest_y, dest_cb, dest_cr, qpix_put, chroma_put,
+ qpix_avg, chroma_avg, weight_op, weight_avg, pixel_shift, 1);
+}
+
+static av_always_inline void
+hl_motion_422(H264Context *h, uint8_t *dest_y, uint8_t *dest_cb, uint8_t *dest_cr,
+ qpel_mc_func (*qpix_put)[16], h264_chroma_mc_func (*chroma_put),
+ qpel_mc_func (*qpix_avg)[16], h264_chroma_mc_func (*chroma_avg),
+ h264_weight_func *weight_op, h264_biweight_func *weight_avg,
+ int pixel_shift)
+{
+ hl_motion(h, dest_y, dest_cb, dest_cr, qpix_put, chroma_put,
+ qpix_avg, chroma_avg, weight_op, weight_avg, pixel_shift, 2);
}
static void free_tables(H264Context *h, int free_rbsp){
dst->list_counts = src->list_counts;
dst->s.obmc_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);
}
/**
s->height = s->avctx->height;
s->codec_id= s->avctx->codec->id;
- ff_h264dsp_init(&h->h264dsp, 8);
- ff_h264_pred_init(&h->hpc, s->codec_id, 8);
+ s->avctx->bits_per_raw_sample = 8;
+ h->cur_chroma_format_idc = 1;
+
+ ff_h264dsp_init(&h->h264dsp,
+ s->avctx->bits_per_raw_sample, h->cur_chroma_format_idc);
+ ff_h264_pred_init(&h->hpc, s->codec_id,
+ s->avctx->bits_per_raw_sample, h->cur_chroma_format_idc);
h->dequant_coeff_pps= -1;
s->unrestricted_mv=1;
- s->decode=1; //FIXME
+ s->dsp.dct_bits = 16;
dsputil_init(&s->dsp, s->avctx); // needed so that idct permutation is known early
memset(h->pps.scaling_matrix4, 16, 6*16*sizeof(uint8_t));
memset(h->pps.scaling_matrix8, 16, 2*64*sizeof(uint8_t));
}
-int ff_h264_decode_extradata(H264Context *h)
+int ff_h264_decode_extradata(H264Context *h, const uint8_t *buf, int size)
{
AVCodecContext *avctx = h->s.avctx;
- if(avctx->extradata[0] == 1){
+ if(!buf || size <= 0)
+ return -1;
+
+ if(buf[0] == 1){
int i, cnt, nalsize;
- unsigned char *p = avctx->extradata;
+ const unsigned char *p = buf;
h->is_avc = 1;
- if(avctx->extradata_size < 7) {
+ if(size < 7) {
av_log(avctx, AV_LOG_ERROR, "avcC too short\n");
return -1;
}
p += 6;
for (i = 0; i < cnt; i++) {
nalsize = AV_RB16(p) + 2;
+ if(nalsize > size - (p-buf))
+ return -1;
if(decode_nal_units(h, p, nalsize) < 0) {
av_log(avctx, AV_LOG_ERROR, "Decoding sps %d from avcC failed\n", i);
return -1;
cnt = *(p++); // Number of pps
for (i = 0; i < cnt; i++) {
nalsize = AV_RB16(p) + 2;
+ if(nalsize > size - (p-buf))
+ return -1;
if (decode_nal_units(h, p, nalsize) < 0) {
av_log(avctx, AV_LOG_ERROR, "Decoding pps %d from avcC failed\n", i);
return -1;
p += nalsize;
}
// Now store right nal length size, that will be use to parse all other nals
- h->nal_length_size = (avctx->extradata[4] & 0x03) + 1;
+ h->nal_length_size = (buf[4] & 0x03) + 1;
} else {
h->is_avc = 0;
- if(decode_nal_units(h, avctx->extradata, avctx->extradata_size) < 0)
+ if(decode_nal_units(h, buf, size) < 0)
return -1;
}
return 0;
}
if(avctx->extradata_size > 0 && avctx->extradata &&
- ff_h264_decode_extradata(h))
+ ff_h264_decode_extradata(h, avctx->extradata, avctx->extradata_size))
return -1;
if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames < h->sps.num_reorder_frames){
copy_picture_range(h->delayed_pic, h1->delayed_pic, MAX_DELAYED_PIC_COUNT+2, s, s1);
h->last_slice_type = h1->last_slice_type;
+ h->sync = h1->sync;
if(!s->current_picture_ptr) return 0;
pics = 0;
while(h->delayed_pic[pics]) pics++;
- assert(pics <= MAX_DELAYED_PIC_COUNT);
+ av_assert0(pics <= MAX_DELAYED_PIC_COUNT);
h->delayed_pic[pics++] = cur;
if (cur->f.reference == 0)
av_log(s->avctx, AV_LOG_DEBUG, "no picture\n");
}
+ if (h->next_output_pic && h->next_output_pic->sync) {
+ h->sync |= 2;
+ }
+
if (setup_finished)
ff_thread_finish_setup(s->avctx);
}
-static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int chroma444, int simple){
+static av_always_inline void backup_mb_border(H264Context *h, uint8_t *src_y,
+ uint8_t *src_cb, uint8_t *src_cr,
+ int linesize, int uvlinesize, int simple)
+{
MpegEncContext * const s = &h->s;
uint8_t *top_border;
int top_idx = 1;
const int pixel_shift = h->pixel_shift;
+ int chroma444 = CHROMA444;
+ int chroma422 = CHROMA422;
src_y -= linesize;
src_cb -= uvlinesize;
AV_COPY128(top_border+16, src_cb + 15*uvlinesize);
AV_COPY128(top_border+32, src_cr + 15*uvlinesize);
}
+ } else if(chroma422){
+ if (pixel_shift) {
+ AV_COPY128(top_border+32, src_cb + 15*uvlinesize);
+ AV_COPY128(top_border+48, src_cr + 15*uvlinesize);
+ } else {
+ AV_COPY64(top_border+16, src_cb + 15*uvlinesize);
+ AV_COPY64(top_border+24, src_cr + 15*uvlinesize);
+ }
} else {
if (pixel_shift) {
AV_COPY128(top_border+32, src_cb+7*uvlinesize);
AV_COPY128(top_border+16, src_cb + 16*linesize);
AV_COPY128(top_border+32, src_cr + 16*linesize);
}
+ } else if(chroma422) {
+ if (pixel_shift) {
+ AV_COPY128(top_border+32, src_cb+16*uvlinesize);
+ AV_COPY128(top_border+48, src_cr+16*uvlinesize);
+ } else {
+ AV_COPY64(top_border+16, src_cb+16*uvlinesize);
+ AV_COPY64(top_border+24, src_cr+16*uvlinesize);
+ }
} else {
if (pixel_shift) {
AV_COPY128(top_border+32, src_cb+8*uvlinesize);
tr_high= ((uint16_t*)ptr)[3 - linesize/2]*0x0001000100010001ULL;
topright= (uint8_t*) &tr_high;
} else {
- tr= ptr[3 - linesize]*0x01010101;
+ tr= ptr[3 - linesize]*0x01010101u;
topright= (uint8_t*) &tr;
}
}else
static const uint8_t dc_mapping[16] = { 0*16, 1*16, 4*16, 5*16, 2*16, 3*16, 6*16, 7*16,
8*16, 9*16,12*16,13*16,10*16,11*16,14*16,15*16};
for(i = 0; i < 16; i++)
- dctcoef_set(h->mb+p*256, pixel_shift, dc_mapping[i], dctcoef_get(h->mb_luma_dc[p], pixel_shift, i));
+ dctcoef_set(h->mb+(p*256 << pixel_shift), pixel_shift, dc_mapping[i], dctcoef_get(h->mb_luma_dc[p], pixel_shift, i));
}
}
}else
}
}
-static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, int pixel_shift){
+static av_always_inline void hl_decode_mb_internal(H264Context *h, int simple, int pixel_shift)
+{
MpegEncContext * const s = &h->s;
const int mb_x= s->mb_x;
const int mb_y= s->mb_y;
/* is_h264 should always be true if SVQ3 is disabled. */
const int is_h264 = !CONFIG_SVQ3_DECODER || simple || s->codec_id == CODEC_ID_H264;
void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride);
+ const int block_h = 16 >> s->chroma_y_shift;
+ const int chroma422 = CHROMA422;
dest_y = s->current_picture.f.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16;
- dest_cb = s->current_picture.f.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * 8;
- dest_cr = s->current_picture.f.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * 8;
+ dest_cb = s->current_picture.f.data[1] + (mb_x << pixel_shift)*8 + mb_y * s->uvlinesize * block_h;
+ dest_cr = s->current_picture.f.data[2] + (mb_x << pixel_shift)*8 + mb_y * s->uvlinesize * block_h;
s->dsp.prefetch(dest_y + (s->mb_x&3)*4*s->linesize + (64 << pixel_shift), s->linesize, 4);
s->dsp.prefetch(dest_cb + (s->mb_x&7)*s->uvlinesize + (64 << pixel_shift), dest_cr - dest_cb, 2);
block_offset = &h->block_offset[48];
if(mb_y&1){ //FIXME move out of this function?
dest_y -= s->linesize*15;
- dest_cb-= s->uvlinesize*7;
- dest_cr-= s->uvlinesize*7;
+ dest_cb-= s->uvlinesize * (block_h - 1);
+ dest_cr-= s->uvlinesize * (block_h - 1);
}
if(FRAME_MBAFF) {
int list;
}
if (!simple && IS_INTRA_PCM(mb_type)) {
+ const int bit_depth = h->sps.bit_depth_luma;
if (pixel_shift) {
- const int bit_depth = h->sps.bit_depth_luma;
int j;
GetBitContext gb;
init_get_bits(&gb, (uint8_t*)h->mb, 384*bit_depth);
}
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
if (!h->sps.chroma_format_idc) {
- for (i = 0; i < 8; i++) {
+ for (i = 0; i < block_h; i++) {
uint16_t *tmp_cb = (uint16_t*)(dest_cb + i*uvlinesize);
- for (j = 0; j < 8; j++) {
- tmp_cb[j] = 1 << (bit_depth - 1);
- }
- }
- for (i = 0; i < 8; i++) {
uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize);
for (j = 0; j < 8; j++) {
- tmp_cr[j] = 1 << (bit_depth - 1);
+ tmp_cb[j] = tmp_cr[j] = 1 << (bit_depth - 1);
}
}
} else {
- for (i = 0; i < 8; i++) {
+ for (i = 0; i < block_h; i++) {
uint16_t *tmp_cb = (uint16_t*)(dest_cb + i*uvlinesize);
for (j = 0; j < 8; j++)
tmp_cb[j] = get_bits(&gb, bit_depth);
}
- for (i = 0; i < 8; i++) {
+ for (i = 0; i < block_h; i++) {
uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize);
for (j = 0; j < 8; j++)
tmp_cr[j] = get_bits(&gb, bit_depth);
}
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
if (!h->sps.chroma_format_idc) {
- for (i = 0; i < 8; i++) {
- memset(dest_cb + i*uvlinesize, 128, 8);
- memset(dest_cr + i*uvlinesize, 128, 8);
+ for (i=0; i<8; i++) {
+ memset(dest_cb + i*uvlinesize, 1 << (bit_depth - 1), 8);
+ memset(dest_cr + i*uvlinesize, 1 << (bit_depth - 1), 8);
}
} else {
- for (i = 0; i < 8; i++) {
+ for (i=0; i<block_h; i++) {
memcpy(dest_cb + i*uvlinesize, h->mb + 128 + i*4, 8);
memcpy(dest_cr + i*uvlinesize, h->mb + 160 + i*4, 8);
}
if(h->deblocking_filter)
xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0, 0, simple, pixel_shift);
}else if(is_h264){
- hl_motion(h, dest_y, dest_cb, dest_cr,
- s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab,
- s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab,
- h->h264dsp.weight_h264_pixels_tab,
- h->h264dsp.biweight_h264_pixels_tab, pixel_shift, 0);
+ if (chroma422) {
+ hl_motion_422(h, dest_y, dest_cb, dest_cr,
+ s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab,
+ s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab,
+ h->h264dsp.weight_h264_pixels_tab,
+ h->h264dsp.biweight_h264_pixels_tab,
+ pixel_shift);
+ } else {
+ hl_motion_420(h, dest_y, dest_cb, dest_cr,
+ s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab,
+ s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab,
+ h->h264dsp.weight_h264_pixels_tab,
+ h->h264dsp.biweight_h264_pixels_tab,
+ pixel_shift);
+ }
}
hl_decode_mb_idct_luma(h, mb_type, is_h264, simple, transform_bypass, pixel_shift, block_offset, linesize, dest_y, 0);
if(h->non_zero_count_cache[ scan8[i] ] || dctcoef_get(h->mb, pixel_shift, i*16))
idct_add (dest[j-1] + block_offset[i], h->mb + (i*16 << pixel_shift), uvlinesize);
}
+ if (chroma422) {
+ for(i=j*16+4; i<j*16+8; i++){
+ if(h->non_zero_count_cache[ scan8[i+4] ] || dctcoef_get(h->mb, pixel_shift, i*16))
+ idct_add (dest[j-1] + block_offset[i+4], h->mb + (i*16 << pixel_shift), uvlinesize);
+ }
+ }
}
}
}else{
if(is_h264){
+ int qp[2];
+ if (chroma422) {
+ qp[0] = h->chroma_qp[0] + 3;
+ qp[1] = h->chroma_qp[1] + 3;
+ } else {
+ qp[0] = h->chroma_qp[0];
+ qp[1] = h->chroma_qp[1];
+ }
if(h->non_zero_count_cache[ scan8[CHROMA_DC_BLOCK_INDEX+0] ])
- h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*1 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]);
+ h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*1 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][qp[0]][0]);
if(h->non_zero_count_cache[ scan8[CHROMA_DC_BLOCK_INDEX+1] ])
- h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*2 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]);
+ h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + (16*16*2 << pixel_shift), h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][qp[1]][0]);
h->h264dsp.h264_idct_add8(dest, block_offset,
h->mb, uvlinesize,
h->non_zero_count_cache);
- }else{
+ }
+#if CONFIG_SVQ3_DECODER
+ else{
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16*16*1, h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]);
h->h264dsp.h264_chroma_dc_dequant_idct(h->mb + 16*16*2, h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]);
for(j=1; j<3; j++){
}
}
}
+#endif
}
}
}
s->me.qpel_put, s->dsp.put_h264_chroma_pixels_tab,
s->me.qpel_avg, s->dsp.avg_h264_chroma_pixels_tab,
h->h264dsp.weight_h264_pixels_tab,
- h->h264dsp.biweight_h264_pixels_tab, pixel_shift, 1);
+ h->h264dsp.biweight_h264_pixels_tab, pixel_shift, 3);
}
for (p = 0; p < plane_count; p++)
}
if(field < 0){
- cur_poc = s->current_picture_ptr->poc;
+ if (s->picture_structure == PICT_FRAME) {
+ cur_poc = s->current_picture_ptr->poc;
+ } else {
+ cur_poc = s->current_picture_ptr->field_poc[s->picture_structure - 1];
+ }
if( h->ref_count[0] == 1 && h->ref_count[1] == 1 && !FRAME_MBAFF
&& h->ref_list[0][0].poc + h->ref_list[1][0].poc == 2*cur_poc){
h->use_weight= 0;
static void flush_dpb(AVCodecContext *avctx){
H264Context *h= avctx->priv_data;
int i;
- for(i=0; i<MAX_DELAYED_PIC_COUNT; i++) {
+ for(i=0; i<=MAX_DELAYED_PIC_COUNT; i++) {
if(h->delayed_pic[i])
h->delayed_pic[i]->f.reference = 0;
h->delayed_pic[i]= NULL;
h->s.first_field= 0;
ff_h264_reset_sei(h);
ff_mpeg_flush(avctx);
+ h->recovery_frame= -1;
+ h->sync= 0;
}
static int init_poc(H264Context *h){
s->me.qpel_avg= s->dsp.avg_h264_qpel_pixels_tab;
}
- first_mb_in_slice= get_ue_golomb(&s->gb);
+ first_mb_in_slice= get_ue_golomb_long(&s->gb);
if(first_mb_in_slice == 0){ //FIXME better field boundary detection
if(h0->current_slice && FIELD_PICTURE){
h->b_stride= s->mb_width*4;
+ s->chroma_y_shift = h->sps.chroma_format_idc <= 1; // 400 uses yuv420p
+
s->width = 16*s->mb_width - (2>>CHROMA444)*FFMIN(h->sps.crop_right, (8<<CHROMA444)-1);
if(h->sps.frame_mbs_only_flag)
- s->height= 16*s->mb_height - (2>>CHROMA444)*FFMIN(h->sps.crop_bottom, (8<<CHROMA444)-1);
+ s->height= 16*s->mb_height - (1<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1);
else
- s->height= 16*s->mb_height - (4>>CHROMA444)*FFMIN(h->sps.crop_bottom, (8<<CHROMA444)-1);
+ s->height= 16*s->mb_height - (2<<s->chroma_y_shift)*FFMIN(h->sps.crop_bottom, (16>>s->chroma_y_shift)-1);
if (s->context_initialized
- && ( s->width != s->avctx->width || s->height != s->avctx->height
+ && ( s->width != s->avctx->coded_width || s->height != s->avctx->coded_height
+ || s->avctx->bits_per_raw_sample != h->sps.bit_depth_luma
+ || h->cur_chroma_format_idc != h->sps.chroma_format_idc
|| av_cmp_q(h->sps.sar, s->avctx->sample_aspect_ratio))) {
if(h != h0) {
- av_log_missing_feature(s->avctx, "Width/height changing with threads is", 0);
+ av_log_missing_feature(s->avctx, "Width/height/bit depth/chroma idc changing with threads is", 0);
return -1; // width / height changed during parallelized decoding
}
free_tables(h, 0);
flush_dpb(s->avctx);
MPV_common_end(s);
+ h->list_count = 0;
}
if (!s->context_initialized) {
if (h != h0) {
s->avctx->sample_aspect_ratio= h->sps.sar;
av_assert0(s->avctx->sample_aspect_ratio.den);
+ if (s->avctx->bits_per_raw_sample != h->sps.bit_depth_luma ||
+ h->cur_chroma_format_idc != h->sps.chroma_format_idc) {
+ if (h->sps.bit_depth_luma >= 8 && h->sps.bit_depth_luma <= 10 &&
+ (h->sps.bit_depth_luma != 9 || !CHROMA422)) {
+ s->avctx->bits_per_raw_sample = h->sps.bit_depth_luma;
+ h->cur_chroma_format_idc = h->sps.chroma_format_idc;
+ h->pixel_shift = h->sps.bit_depth_luma > 8;
+
+ ff_h264dsp_init(&h->h264dsp, h->sps.bit_depth_luma, h->sps.chroma_format_idc);
+ ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma, h->sps.chroma_format_idc);
+ s->dsp.dct_bits = h->sps.bit_depth_luma > 8 ? 32 : 16;
+ dsputil_init(&s->dsp, s->avctx);
+ } else {
+ av_log(s->avctx, AV_LOG_ERROR, "Unsupported bit depth: %d chroma_idc: %d\n",
+ h->sps.bit_depth_luma, h->sps.chroma_format_idc);
+ return -1;
+ }
+ }
+
if(h->sps.video_signal_type_present_flag){
- s->avctx->color_range = h->sps.full_range ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG;
+ s->avctx->color_range = h->sps.full_range>0 ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG;
if(h->sps.colour_description_present_flag){
s->avctx->color_primaries = h->sps.color_primaries;
s->avctx->color_trc = h->sps.color_trc;
switch (h->sps.bit_depth_luma) {
case 9 :
- s->avctx->pix_fmt = CHROMA444 ? PIX_FMT_YUV444P9 : PIX_FMT_YUV420P9;
+ if (CHROMA444)
+ s->avctx->pix_fmt = PIX_FMT_YUV444P9;
+ else if (CHROMA422)
+ s->avctx->pix_fmt = PIX_FMT_YUV422P9;
+ else
+ s->avctx->pix_fmt = PIX_FMT_YUV420P9;
break;
case 10 :
- s->avctx->pix_fmt = CHROMA444 ? PIX_FMT_YUV444P10 : PIX_FMT_YUV420P10;
+ if (CHROMA444)
+ s->avctx->pix_fmt = PIX_FMT_YUV444P10;
+ else if (CHROMA422)
+ s->avctx->pix_fmt = PIX_FMT_YUV422P10;
+ else
+ s->avctx->pix_fmt = PIX_FMT_YUV420P10;
break;
default:
if (CHROMA444){
s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ444P : PIX_FMT_YUV444P;
+ if (s->avctx->colorspace == AVCOL_SPC_RGB) {
+ s->avctx->pix_fmt = PIX_FMT_GBR24P;
+ av_log(h->s.avctx, AV_LOG_DEBUG, "Detected GBR colorspace.\n");
+ } else if (s->avctx->colorspace == AVCOL_SPC_YCGCO) {
+ av_log(h->s.avctx, AV_LOG_WARNING, "Detected unsupported YCgCo colorspace.\n");
+ }
+ } else if (CHROMA422) {
+ s->avctx->pix_fmt = s->avctx->color_range == AVCOL_RANGE_JPEG ? PIX_FMT_YUVJ422P : PIX_FMT_YUV422P;
}else{
s->avctx->pix_fmt = s->avctx->get_format(s->avctx,
s->avctx->codec->pix_fmts ?
if(h->sps.frame_mbs_only_flag){
s->picture_structure= PICT_FRAME;
}else{
+ if(!h->sps.direct_8x8_inference_flag && slice_type == AV_PICTURE_TYPE_B){
+ av_log(h->s.avctx, AV_LOG_ERROR, "This stream was generated by a broken encoder, invalid 8x8 inference\n");
+ return -1;
+ }
if(get_bits1(&s->gb)) { //field_pic_flag
s->picture_structure= PICT_TOP_FIELD + get_bits1(&s->gb); //bottom_field_flag
} else {
ff_thread_report_progress((AVFrame*)s->current_picture_ptr, INT_MAX, 1);
ff_generate_sliding_window_mmcos(h);
if (ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index) < 0 &&
- s->avctx->error_recognition >= FF_ER_EXPLODE)
+ (s->avctx->err_recognition & AV_EF_EXPLODE))
return AVERROR_INVALIDDATA;
/* Error concealment: if a ref is missing, copy the previous ref in its place.
* FIXME: avoiding a memcpy would be nice, but ref handling makes many assumptions
h->ref_count[1]= h->pps.ref_count[1];
if(h->slice_type_nos != AV_PICTURE_TYPE_I){
+ unsigned max= (16<<(s->picture_structure != PICT_FRAME))-1;
if(h->slice_type_nos == AV_PICTURE_TYPE_B){
h->direct_spatial_mv_pred= get_bits1(&s->gb);
}
if(h->slice_type_nos==AV_PICTURE_TYPE_B)
h->ref_count[1]= get_ue_golomb(&s->gb) + 1;
- if(h->ref_count[0]-1 > 32-1 || h->ref_count[1]-1 > 32-1){
- av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow\n");
- h->ref_count[0]= h->ref_count[1]= 1;
- return -1;
- }
+ }
+ if(h->ref_count[0]-1 > max || h->ref_count[1]-1 > max){
+ av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow\n");
+ h->ref_count[0]= h->ref_count[1]= 1;
+ return -1;
}
if(h->slice_type_nos == AV_PICTURE_TYPE_B)
h->list_count= 2;
else
h->list_count= 1;
}else
- h->list_count= 0;
+ h->ref_count[1]= h->ref_count[0]= h->list_count= 0;
if(!default_ref_list_done){
ff_h264_fill_default_ref_list(h);
}
- if(h->slice_type_nos!=AV_PICTURE_TYPE_I && ff_h264_decode_ref_pic_list_reordering(h) < 0)
+ if(h->slice_type_nos!=AV_PICTURE_TYPE_I && ff_h264_decode_ref_pic_list_reordering(h) < 0) {
+ h->ref_count[1]= h->ref_count[0]= 0;
return -1;
+ }
if(h->slice_type_nos!=AV_PICTURE_TYPE_I){
s->last_picture_ptr= &h->ref_list[0][0];
}
if(h->nal_ref_idc && ff_h264_decode_ref_pic_marking(h0, &s->gb) < 0 &&
- s->avctx->error_recognition >= FF_ER_EXPLODE)
+ (s->avctx->err_recognition & AV_EF_EXPLODE))
return AVERROR_INVALIDDATA;
if(FRAME_MBAFF){
h0->last_slice_type = slice_type;
h->slice_num = ++h0->current_slice;
- if(h->slice_num >= MAX_SLICES){
- av_log(s->avctx, AV_LOG_ERROR, "Too many slices, increase MAX_SLICES and recompile\n");
+
+ if(h->slice_num)
+ h0->slice_row[(h->slice_num-1)&(MAX_SLICES-1)]= s->resync_mb_y;
+ if ( h0->slice_row[h->slice_num&(MAX_SLICES-1)] + 3 >= s->resync_mb_y
+ && h0->slice_row[h->slice_num&(MAX_SLICES-1)] <= s->resync_mb_y
+ && h->slice_num >= MAX_SLICES) {
+ //in case of ASO this check needs to be updated depending on how we decide to assign slice numbers in this case
+ av_log(s->avctx, AV_LOG_WARNING, "Possibly too many slices (%d >= %d), increase MAX_SLICES and recompile if there are artifacts\n", h->slice_num, MAX_SLICES);
}
for(j=0; j<2; j++){
const int end_mb_y= s->mb_y + FRAME_MBAFF;
const int old_slice_type= h->slice_type;
const int pixel_shift = h->pixel_shift;
+ const int block_h = 16 >> s->chroma_y_shift;
if(h->deblocking_filter) {
for(mb_x= start_x; mb_x<end_x; mb_x++){
s->mb_x= mb_x;
s->mb_y= mb_y;
dest_y = s->current_picture.f.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16;
- dest_cb = s->current_picture.f.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444);
- dest_cr = s->current_picture.f.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444);
+ dest_cb = s->current_picture.f.data[1] + (mb_x << pixel_shift) * (8 << CHROMA444) + mb_y * s->uvlinesize * block_h;
+ dest_cr = s->current_picture.f.data[2] + (mb_x << pixel_shift) * (8 << CHROMA444) + mb_y * s->uvlinesize * block_h;
//FIXME simplify above
if (MB_FIELD) {
uvlinesize = h->mb_uvlinesize = s->uvlinesize * 2;
if(mb_y&1){ //FIXME move out of this function?
dest_y -= s->linesize*15;
- dest_cb-= s->uvlinesize*((8 << CHROMA444)-1);
- dest_cr-= s->uvlinesize*((8 << CHROMA444)-1);
+ dest_cb-= s->uvlinesize * (block_h - 1);
+ dest_cr-= s->uvlinesize * (block_h - 1);
}
} else {
linesize = h->mb_linesize = s->linesize;
uvlinesize = h->mb_uvlinesize = s->uvlinesize;
}
- backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, CHROMA444, 0);
+ backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0);
if(fill_filter_caches(h, mb_type))
continue;
h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mb_xy]);
if(s->mb_y >= s->mb_height){
tprintf(s->avctx, "slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);
- if(get_bits_count(&s->gb) == s->gb.size_in_bits ) {
+ if( get_bits_count(&s->gb) == s->gb.size_in_bits
+ || get_bits_count(&s->gb) < s->gb.size_in_bits && s->avctx->error_recognition < FF_ER_AGGRESSIVE) {
ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
return 0;
hx = h->thread_context[i];
hx->s.error_recognition = avctx->error_recognition;
hx->s.error_count = 0;
+ hx->x264_build= h->x264_build;
}
avctx->execute(avctx, (void *)decode_slice,
s->workaround_bugs |= FF_BUG_TRUNCATED;
if(!(s->workaround_bugs & FF_BUG_TRUNCATED)){
- while(ptr[dst_length - 1] == 0 && dst_length > 0)
+ while(dst_length > 0 && ptr[dst_length - 1] == 0)
dst_length--;
}
bit_length= !dst_length ? 0 : (8*dst_length - ff_h264_decode_rbsp_trailing(h, ptr + dst_length - 1));
if(s->avctx->debug&FF_DEBUG_STARTCODE){
- av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d at %d/%d length %d\n", hx->nal_unit_type, buf_index, buf_size, dst_length);
+ av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d/%d at %d/%d length %d\n", hx->nal_unit_type, hx->nal_ref_idc, buf_index, buf_size, dst_length);
}
if (h->is_avc && (nalsize != consumed) && nalsize){
switch (hx->nal_unit_type) {
case NAL_SPS:
case NAL_PPS:
+ nals_needed = nal_index;
+ break;
case NAL_IDR_SLICE:
case NAL_SLICE:
- nals_needed = nal_index;
+ init_get_bits(&hx->s.gb, ptr, bit_length);
+ if (!get_ue_golomb(&hx->s.gb))
+ nals_needed = nal_index;
}
continue;
}
if((err = decode_slice_header(hx, h)))
break;
+ if ( h->sei_recovery_frame_cnt >= 0
+ && ((h->recovery_frame - h->frame_num) & ((1 << h->sps.log2_max_frame_num)-1)) > h->sei_recovery_frame_cnt) {
+ h->recovery_frame = (h->frame_num + h->sei_recovery_frame_cnt) %
+ (1 << h->sps.log2_max_frame_num);
+ }
+
s->current_picture_ptr->f.key_frame |=
- (hx->nal_unit_type == NAL_IDR_SLICE) ||
- (h->sei_recovery_frame_cnt >= 0);
+ (hx->nal_unit_type == NAL_IDR_SLICE);
+
+ if (h->recovery_frame == h->frame_num) {
+ h->sync |= 1;
+ h->recovery_frame = -1;
+ }
+
+ h->sync |= !!s->current_picture_ptr->f.key_frame;
+ h->sync |= 3*!!(s->flags2 & CODEC_FLAG2_SHOW_ALL);
+ s->current_picture_ptr->sync = h->sync;
if (h->current_slice == 1) {
if(!(s->flags2 & CODEC_FLAG2_CHUNKS)) {
break;
case NAL_SPS:
init_get_bits(&s->gb, ptr, bit_length);
- ff_h264_decode_seq_parameter_set(h);
+ if(ff_h264_decode_seq_parameter_set(h) < 0 && (h->is_avc ? (nalsize != consumed) && nalsize : 1)){
+ av_log(h->s.avctx, AV_LOG_DEBUG, "SPS decoding failure, trying alternative mode\n");
+ if(h->is_avc) av_assert0(next_avc - buf_index + consumed == nalsize);
+ init_get_bits(&s->gb, &buf[buf_index + 1 - consumed], 8*(next_avc - buf_index + consumed));
+ ff_h264_decode_seq_parameter_set(h);
+ }
if (s->flags& CODEC_FLAG_LOW_DELAY ||
(h->sps.bitstream_restriction_flag && !h->sps.num_reorder_frames))
if(avctx->has_b_frames < 2)
avctx->has_b_frames= !s->low_delay;
-
- if (avctx->bits_per_raw_sample != h->sps.bit_depth_luma) {
- if (h->sps.bit_depth_luma >= 8 && h->sps.bit_depth_luma <= 10) {
- avctx->bits_per_raw_sample = h->sps.bit_depth_luma;
- h->pixel_shift = h->sps.bit_depth_luma > 8;
-
- ff_h264dsp_init(&h->h264dsp, h->sps.bit_depth_luma);
- ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma);
- s->dsp.dct_bits = h->sps.bit_depth_luma > 8 ? 32 : 16;
- dsputil_init(&s->dsp, s->avctx);
- } else {
- av_log(avctx, AV_LOG_DEBUG, "Unsupported bit depth: %d\n", h->sps.bit_depth_luma);
- return -1;
- }
- }
break;
case NAL_PPS:
init_get_bits(&s->gb, ptr, bit_length);
return 0;
}
+ if(h->is_avc && buf_size >= 9 && AV_RB32(buf)==0x0164001F && buf[5] && buf[8]==0x67)
+ return ff_h264_decode_extradata(h, buf, buf_size);
buf_index=decode_nal_units(h, buf, buf_size);
if(buf_index < 0)
field_end(h, 0);
- if (!h->next_output_pic) {
- /* Wait for second field. */
- *data_size = 0;
-
- } else {
- *data_size = sizeof(AVFrame);
- *pict = *(AVFrame*)h->next_output_pic;
+ *data_size = 0; /* Wait for second field. */
+ if (h->next_output_pic && h->next_output_pic->sync) {
+ if(h->sync>1 || h->next_output_pic->f.pict_type != AV_PICTURE_TYPE_B){
+ *data_size = sizeof(AVFrame);
+ *pict = *(AVFrame*)h->next_output_pic;
+ }
}
}
#undef random
#define COUNT 8000
#define SIZE (COUNT*40)
+extern AVCodec ff_h264_decoder;
int main(void){
int i;
uint8_t temp[SIZE];
DSPContext dsp;
AVCodecContext avctx;
+ avcodec_get_context_defaults3(&avctx, &ff_h264_decoder);
+
dsputil_init(&dsp, &avctx);
init_put_bits(&pb, temp, SIZE);
{ FF_PROFILE_UNKNOWN },
};
+static const AVOption h264_options[] = {
+ {"is_avc", "is avc", offsetof(H264Context, is_avc), FF_OPT_TYPE_INT, {.dbl = 0}, 0, 1, 0},
+ {"nal_length_size", "nal_length_size", offsetof(H264Context, nal_length_size), FF_OPT_TYPE_INT, {.dbl = 0}, 0, 4, 0},
+ {NULL}
+};
+
+static const AVClass h264_class = {
+ "H264 Decoder",
+ av_default_item_name,
+ h264_options,
+ LIBAVUTIL_VERSION_INT,
+};
+
+static const AVClass h264_vdpau_class = {
+ "H264 VDPAU Decoder",
+ av_default_item_name,
+ h264_options,
+ LIBAVUTIL_VERSION_INT,
+};
+
AVCodec ff_h264_decoder = {
.name = "h264",
.type = AVMEDIA_TYPE_VIDEO,
.init_thread_copy = ONLY_IF_THREADS_ENABLED(decode_init_thread_copy),
.update_thread_context = ONLY_IF_THREADS_ENABLED(decode_update_thread_context),
.profiles = NULL_IF_CONFIG_SMALL(profiles),
+ .priv_class = &h264_class,
};
#if CONFIG_H264_VDPAU_DECODER
.long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10 (VDPAU acceleration)"),
.pix_fmts = (const enum PixelFormat[]){PIX_FMT_VDPAU_H264, PIX_FMT_NONE},
.profiles = NULL_IF_CONFIG_SMALL(profiles),
+ .priv_class = &h264_vdpau_class,
};
#endif