// Error resilience puts the current picture in the ref list.
// Don't try to wait on these as it will cause a deadlock.
// Fields can wait on each other, though.
- if(ref->thread_opaque != s->current_picture.thread_opaque ||
- (ref->reference&3) != s->picture_structure) {
+ if (ref->f.thread_opaque != s->current_picture.f.thread_opaque ||
+ (ref->f.reference & 3) != s->picture_structure) {
my = get_lowest_part_list_y(h, ref, n, height, y_offset, 0);
if (refs[0][ref_n] < 0) nrefs[0] += 1;
refs[0][ref_n] = FFMAX(refs[0][ref_n], my);
int ref_n = h->ref_cache[1][ scan8[n] ];
Picture *ref= &h->ref_list[1][ref_n];
- if(ref->thread_opaque != s->current_picture.thread_opaque ||
- (ref->reference&3) != s->picture_structure) {
+ if (ref->f.thread_opaque != s->current_picture.f.thread_opaque ||
+ (ref->f.reference & 3) != s->picture_structure) {
my = get_lowest_part_list_y(h, ref, n, height, y_offset, 1);
if (refs[1][ref_n] < 0) nrefs[1] += 1;
refs[1][ref_n] = FFMAX(refs[1][ref_n], my);
static void await_references(H264Context *h){
MpegEncContext * const s = &h->s;
const int mb_xy= h->mb_xy;
- const int mb_type= s->current_picture.mb_type[mb_xy];
+ const int mb_type = s->current_picture.f.mb_type[mb_xy];
int refs[2][48];
int nrefs[2] = {0};
int ref, list;
int row = refs[list][ref];
if(row >= 0){
Picture *ref_pic = &h->ref_list[list][ref];
- int ref_field = ref_pic->reference - 1;
+ int ref_field = ref_pic->f.reference - 1;
int ref_field_picture = ref_pic->field_picture;
int pic_height = 16*s->mb_height >> ref_field_picture;
}
#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 luma_xy= (mx&3) + ((my&3)<<2);
int offset = ((mx>>2) << pixel_shift) + (my>>2)*h->mb_linesize;
- uint8_t * src_y = pic->data[0] + offset;
+ uint8_t * src_y = pic->f.data[0] + offset;
uint8_t * src_cb, * src_cr;
int extra_width= h->emu_edge_width;
int extra_height= h->emu_edge_height;
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){
- src_cb = pic->data[1] + offset;
+ 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,
16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, pic_width, pic_height);
qpix_op[luma_xy](dest_cb + delta, src_cb + delta, h->mb_linesize);
}
- src_cr = pic->data[2] + offset;
+ src_cr = pic->f.data[2] + offset;
if(emu){
s->dsp.emulated_edge_mc(s->edge_emu_buffer, src_cr - (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);
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->reference - 1));
+ my += 2 * ((s->mb_y & 1) - (pic->f.reference - 1));
emu |= (my>>3) < 0 || (my>>3) + 8 >= (pic_height>>1);
}
- src_cb= pic->data[1] + ((mx>>3) << pixel_shift) + (my>>3)*h->mb_uvlinesize;
- src_cr= pic->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;
if(refn >= 0){
const int mx= (h->mv_cache[list][scan8[0]][0]>>2) + 16*s->mb_x + 8;
const int my= (h->mv_cache[list][scan8[0]][1]>>2) + 16*s->mb_y;
- uint8_t **src= h->ref_list[list][refn].data;
+ 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{
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.mb_type[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);
+ ff_h264dsp_init(&h->h264dsp, 8, 1);
+ ff_h264_pred_init(&h->hpc, s->codec_id, 8, 1);
h->dequant_coeff_pps= -1;
s->unrestricted_mv=1;
- s->decode=1; //FIXME
dsputil_init(&s->dsp, s->avctx); // needed so that idct permutation is known early
p += 6;
for (i = 0; i < cnt; i++) {
nalsize = AV_RB16(p) + 2;
+ if (p - avctx->extradata + nalsize > avctx->extradata_size)
+ 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 (p - avctx->extradata + nalsize > avctx->extradata_size)
+ 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;
av_cold int ff_h264_decode_init(AVCodecContext *avctx){
H264Context *h= avctx->priv_data;
MpegEncContext * const s = &h->s;
+ int i;
MPV_decode_defaults(s);
h->thread_context[0] = h;
h->outputed_poc = h->next_outputed_poc = INT_MIN;
+ for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++)
+ h->last_pocs[i] = INT_MIN;
h->prev_poc_msb= 1<<16;
h->x264_build = -1;
ff_h264_reset_sei(h);
memcpy(&h->s + 1, &h1->s + 1, sizeof(H264Context) - sizeof(MpegEncContext)); //copy all fields after MpegEnc
memset(h->sps_buffers, 0, sizeof(h->sps_buffers));
memset(h->pps_buffers, 0, sizeof(h->pps_buffers));
- ff_h264_alloc_tables(h);
+ if (ff_h264_alloc_tables(h) < 0) {
+ av_log(dst, AV_LOG_ERROR, "Could not allocate memory for h264\n");
+ return AVERROR(ENOMEM);
+ }
context_init(h);
for(i=0; i<2; i++){
if(!s->current_picture_ptr) return 0;
if(!s->dropable) {
- ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
+ err = ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
h->prev_poc_msb = h->poc_msb;
h->prev_poc_lsb = h->poc_lsb;
}
h->prev_frame_num = h->frame_num;
h->outputed_poc = h->next_outputed_poc;
- return 0;
+ return err;
}
int ff_h264_frame_start(H264Context *h){
* Zero here; IDR markings per slice in frame or fields are ORed in later.
* See decode_nal_units().
*/
- s->current_picture_ptr->key_frame= 0;
+ s->current_picture_ptr->f.key_frame = 0;
s->current_picture_ptr->mmco_reset= 0;
assert(s->linesize && s->uvlinesize);
/* some macroblocks can be accessed before they're available in case of lost slices, mbaff or threading*/
memset(h->slice_table, -1, (s->mb_height*s->mb_stride-1) * sizeof(*h->slice_table));
-// s->decode= (s->flags&CODEC_FLAG_PSNR) || !s->encoding || s->current_picture.reference /*|| h->contains_intra*/ || 1;
+// s->decode = (s->flags & CODEC_FLAG_PSNR) || !s->encoding || s->current_picture.f.reference /*|| h->contains_intra*/ || 1;
// 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
// get released even with set reference, besides SVQ3 and others do not
// mark frames as reference later "naturally".
if(s->codec_id != CODEC_ID_SVQ3)
- s->current_picture_ptr->reference= 0;
+ s->current_picture_ptr->f.reference = 0;
s->current_picture_ptr->field_poc[0]=
s->current_picture_ptr->field_poc[1]= INT_MAX;
Picture *cur = s->current_picture_ptr;
int i, pics, out_of_order, out_idx;
- s->current_picture_ptr->qscale_type= FF_QSCALE_TYPE_H264;
- s->current_picture_ptr->pict_type= s->pict_type;
+ s->current_picture_ptr->f.qscale_type = FF_QSCALE_TYPE_H264;
+ s->current_picture_ptr->f.pict_type = s->pict_type;
if (h->next_output_pic) return;
return;
}
- cur->interlaced_frame = 0;
- cur->repeat_pict = 0;
+ cur->f.interlaced_frame = 0;
+ cur->f.repeat_pict = 0;
/* Signal interlacing information externally. */
/* Prioritize picture timing SEI information over used decoding process if it exists. */
break;
case SEI_PIC_STRUCT_TOP_FIELD:
case SEI_PIC_STRUCT_BOTTOM_FIELD:
- cur->interlaced_frame = 1;
+ cur->f.interlaced_frame = 1;
break;
case SEI_PIC_STRUCT_TOP_BOTTOM:
case SEI_PIC_STRUCT_BOTTOM_TOP:
if (FIELD_OR_MBAFF_PICTURE)
- cur->interlaced_frame = 1;
+ cur->f.interlaced_frame = 1;
else
// try to flag soft telecine progressive
- cur->interlaced_frame = h->prev_interlaced_frame;
+ cur->f.interlaced_frame = h->prev_interlaced_frame;
break;
case SEI_PIC_STRUCT_TOP_BOTTOM_TOP:
case SEI_PIC_STRUCT_BOTTOM_TOP_BOTTOM:
// Signal the possibility of telecined film externally (pic_struct 5,6)
// From these hints, let the applications decide if they apply deinterlacing.
- cur->repeat_pict = 1;
+ cur->f.repeat_pict = 1;
break;
case SEI_PIC_STRUCT_FRAME_DOUBLING:
// Force progressive here, as doubling interlaced frame is a bad idea.
- cur->repeat_pict = 2;
+ cur->f.repeat_pict = 2;
break;
case SEI_PIC_STRUCT_FRAME_TRIPLING:
- cur->repeat_pict = 4;
+ cur->f.repeat_pict = 4;
break;
}
if ((h->sei_ct_type & 3) && h->sei_pic_struct <= SEI_PIC_STRUCT_BOTTOM_TOP)
- cur->interlaced_frame = (h->sei_ct_type & (1<<1)) != 0;
+ cur->f.interlaced_frame = (h->sei_ct_type & (1 << 1)) != 0;
}else{
/* Derive interlacing flag from used decoding process. */
- cur->interlaced_frame = FIELD_OR_MBAFF_PICTURE;
+ cur->f.interlaced_frame = FIELD_OR_MBAFF_PICTURE;
}
- h->prev_interlaced_frame = cur->interlaced_frame;
+ h->prev_interlaced_frame = cur->f.interlaced_frame;
if (cur->field_poc[0] != cur->field_poc[1]){
/* Derive top_field_first from field pocs. */
- cur->top_field_first = cur->field_poc[0] < cur->field_poc[1];
+ cur->f.top_field_first = cur->field_poc[0] < cur->field_poc[1];
}else{
- if(cur->interlaced_frame || h->sps.pic_struct_present_flag){
+ if (cur->f.interlaced_frame || h->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_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM
|| h->sei_pic_struct == SEI_PIC_STRUCT_TOP_BOTTOM_TOP)
- cur->top_field_first = 1;
+ cur->f.top_field_first = 1;
else
- cur->top_field_first = 0;
+ cur->f.top_field_first = 0;
}else{
/* Most likely progressive */
- cur->top_field_first = 0;
+ cur->f.top_field_first = 0;
}
}
assert(pics <= MAX_DELAYED_PIC_COUNT);
h->delayed_pic[pics++] = cur;
- if(cur->reference == 0)
- cur->reference = DELAYED_PIC_REF;
+ if (cur->f.reference == 0)
+ cur->f.reference = DELAYED_PIC_REF;
out = h->delayed_pic[0];
out_idx = 0;
- for(i=1; h->delayed_pic[i] && !h->delayed_pic[i]->key_frame && !h->delayed_pic[i]->mmco_reset; i++)
+ for (i = 1; h->delayed_pic[i] && !h->delayed_pic[i]->f.key_frame && !h->delayed_pic[i]->mmco_reset; i++)
if(h->delayed_pic[i]->poc < out->poc){
out = h->delayed_pic[i];
out_idx = i;
}
- if(s->avctx->has_b_frames == 0 && (h->delayed_pic[0]->key_frame || h->delayed_pic[0]->mmco_reset))
+ if (s->avctx->has_b_frames == 0 && (h->delayed_pic[0]->f.key_frame || h->delayed_pic[0]->mmco_reset))
h->next_outputed_poc= INT_MIN;
out_of_order = out->poc < h->next_outputed_poc;
if(h->sps.bitstream_restriction_flag && s->avctx->has_b_frames >= h->sps.num_reorder_frames)
{ }
- else if((out_of_order && pics-1 == s->avctx->has_b_frames && s->avctx->has_b_frames < MAX_DELAYED_PIC_COUNT)
- || (s->low_delay &&
- ((h->next_outputed_poc != INT_MIN && out->poc > h->next_outputed_poc + 2)
- || cur->pict_type == AV_PICTURE_TYPE_B)))
- {
+ else if (out_of_order && pics-1 == s->avctx->has_b_frames &&
+ s->avctx->has_b_frames < MAX_DELAYED_PIC_COUNT) {
+ int cnt = 0, invalid = 0;
+ for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++) {
+ cnt += out->poc < h->last_pocs[i];
+ invalid += h->last_pocs[i] == INT_MIN;
+ }
+ if (invalid + cnt < MAX_DELAYED_PIC_COUNT) {
+ s->avctx->has_b_frames = FFMAX(s->avctx->has_b_frames, cnt);
+ } else if (cnt) {
+ for (i = 0; i < MAX_DELAYED_PIC_COUNT; i++)
+ h->last_pocs[i] = INT_MIN;
+ }
+ s->low_delay = 0;
+ } else if (s->low_delay &&
+ ((h->next_outputed_poc != INT_MIN && out->poc > h->next_outputed_poc + 2) ||
+ cur->f.pict_type == AV_PICTURE_TYPE_B)) {
s->low_delay = 0;
s->avctx->has_b_frames++;
}
if(out_of_order || pics > s->avctx->has_b_frames){
- out->reference &= ~DELAYED_PIC_REF;
+ out->f.reference &= ~DELAYED_PIC_REF;
out->owner2 = s; // for frame threading, the owner must be the second field's thread
// or else the first thread can release the picture and reuse it unsafely
for(i=out_idx; h->delayed_pic[i]; i++)
h->delayed_pic[i] = h->delayed_pic[i+1];
}
+ memmove(h->last_pocs, &h->last_pocs[1], sizeof(*h->last_pocs) * (MAX_DELAYED_PIC_COUNT - 1));
+ h->last_pocs[MAX_DELAYED_PIC_COUNT - 1] = out->poc;
if(!out_of_order && pics > s->avctx->has_b_frames){
h->next_output_pic = out;
- if(out_idx==0 && h->delayed_pic[0] && (h->delayed_pic[0]->key_frame || h->delayed_pic[0]->mmco_reset)) {
+ if (out_idx == 0 && h->delayed_pic[0] && (h->delayed_pic[0]->f.key_frame || h->delayed_pic[0]->mmco_reset)) {
h->next_outputed_poc = INT_MIN;
} else
h->next_outputed_poc = out->poc;
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;
const int mb_xy= h->mb_xy;
- const int mb_type= s->current_picture.mb_type[mb_xy];
+ const int mb_type = s->current_picture.f.mb_type[mb_xy];
uint8_t *dest_y, *dest_cb, *dest_cr;
int linesize, uvlinesize /*dct_offset*/;
int i, j;
/* 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.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16;
- dest_cb = s->current_picture.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * 8;
- dest_cr = s->current_picture.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * 8;
+ 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)*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;
tmp_y[j] = get_bits(&gb, bit_depth);
}
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
- for (i = 0; i < 8; 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++) {
- 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 (!h->sps.chroma_format_idc) {
+ 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 < block_h; i++) {
+ uint16_t *tmp_cr = (uint16_t*)(dest_cr + i*uvlinesize);
+ for (j = 0; j < 8; j++) {
+ tmp_cr[j] = 1 << (bit_depth - 1);
+ }
+ }
+ } else {
+ 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 < 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);
+ }
}
}
} else {
memcpy(dest_y + i* linesize, h->mb + i*8, 16);
}
if(simple || !CONFIG_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
- for (i=0; i<8; 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->sps.chroma_format_idc) {
+ for (i = 0; i < block_h; i++) {
+ memset(dest_cb + i*uvlinesize, 128, 8);
+ memset(dest_cr + i*uvlinesize, 128, 8);
+ }
+ } else {
+ 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);
const int mb_x= s->mb_x;
const int mb_y= s->mb_y;
const int mb_xy= h->mb_xy;
- const int mb_type= s->current_picture.mb_type[mb_xy];
+ const int mb_type = s->current_picture.f.mb_type[mb_xy];
uint8_t *dest[3];
int linesize;
int i, j, p;
for (p = 0; p < plane_count; p++)
{
- dest[p] = s->current_picture.data[p] + ((mb_x << pixel_shift) + mb_y * s->linesize) * 16;
+ dest[p] = s->current_picture.f.data[p] + ((mb_x << pixel_shift) + mb_y * s->linesize) * 16;
s->dsp.prefetch(dest[p] + (s->mb_x&3)*4*s->linesize + (64 << pixel_shift), s->linesize, 4);
}
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++)
void ff_h264_hl_decode_mb(H264Context *h){
MpegEncContext * const s = &h->s;
const int mb_xy= h->mb_xy;
- const int mb_type= s->current_picture.mb_type[mb_xy];
+ const int mb_type = s->current_picture.f.mb_type[mb_xy];
int is_complex = CONFIG_SMALL || h->is_complex || IS_INTRA_PCM(mb_type) || s->qscale == 0;
if (CHROMA444) {
}
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;
for(ref0=ref_start; ref0 < ref_count0; ref0++){
int poc0 = h->ref_list[0][ref0].poc;
for(ref1=ref_start; ref1 < ref_count1; ref1++){
- int poc1 = h->ref_list[1][ref1].poc;
- int td = av_clip(poc1 - poc0, -128, 127);
- int w= 32;
- if(td){
- int tb = av_clip(cur_poc - poc0, -128, 127);
- int tx = (16384 + (FFABS(td) >> 1)) / td;
- int dist_scale_factor = (tb*tx + 32) >> 8;
- if(dist_scale_factor >= -64 && dist_scale_factor <= 128)
- w = 64 - dist_scale_factor;
+ int w = 32;
+ if (!h->ref_list[0][ref0].long_ref && !h->ref_list[1][ref1].long_ref) {
+ int poc1 = h->ref_list[1][ref1].poc;
+ int td = av_clip(poc1 - poc0, -128, 127);
+ if(td){
+ int tb = av_clip(cur_poc - poc0, -128, 127);
+ int tx = (16384 + (FFABS(td) >> 1)) / td;
+ int dist_scale_factor = (tb*tx + 32) >> 8;
+ if(dist_scale_factor >= -64 && dist_scale_factor <= 128)
+ w = 64 - dist_scale_factor;
+ }
}
if(field<0){
h->implicit_weight[ref0][ref1][0]=
int i;
for(i=0; i<MAX_DELAYED_PIC_COUNT; i++) {
if(h->delayed_pic[i])
- h->delayed_pic[i]->reference= 0;
+ h->delayed_pic[i]->f.reference = 0;
h->delayed_pic[i]= NULL;
}
h->outputed_poc=h->next_outputed_poc= INT_MIN;
h->prev_interlaced_frame = 1;
idr(h);
if(h->s.current_picture_ptr)
- h->s.current_picture_ptr->reference= 0;
+ h->s.current_picture_ptr->f.reference = 0;
h->s.first_field= 0;
ff_h264_reset_sei(h);
ff_mpeg_flush(avctx);
}
}
-static void field_end(H264Context *h, int in_setup){
+static int field_end(H264Context *h, int in_setup){
MpegEncContext * const s = &h->s;
AVCodecContext * const avctx= s->avctx;
+ int err = 0;
s->mb_y= 0;
if (!in_setup && !s->dropable)
if(in_setup || !(avctx->active_thread_type&FF_THREAD_FRAME)){
if(!s->dropable) {
- ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
+ err = ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
h->prev_poc_msb= h->poc_msb;
h->prev_poc_lsb= h->poc_lsb;
}
MPV_frame_end(s);
h->current_slice=0;
+
+ return err;
}
/**
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->avctx->sample_aspect_ratio= h->sps.sar;
av_assert0(s->avctx->sample_aspect_ratio.den);
- h->s.avctx->coded_width = 16*s->mb_width;
- h->s.avctx->coded_height = 16*s->mb_height;
-
if(h->sps.video_signal_type_present_flag){
s->avctx->color_range = h->sps.full_range ? AVCOL_RANGE_JPEG : AVCOL_RANGE_MPEG;
if(h->sps.colour_description_present_flag){
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;
+ } 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 ?
h->prev_interlaced_frame = 1;
init_scan_tables(h);
- ff_h264_alloc_tables(h);
+ if (ff_h264_alloc_tables(h) < 0) {
+ av_log(h->s.avctx, AV_LOG_ERROR, "Could not allocate memory for h264\n");
+ return AVERROR(ENOMEM);
+ }
if (!HAVE_THREADS || !(s->avctx->active_thread_type&FF_THREAD_SLICE)) {
if (context_init(h) < 0) {
ff_thread_report_progress((AVFrame*)s->current_picture_ptr, INT_MAX, 0);
ff_thread_report_progress((AVFrame*)s->current_picture_ptr, INT_MAX, 1);
ff_generate_sliding_window_mmcos(h);
- ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index);
+ if (ff_h264_execute_ref_pic_marking(h, h->mmco, h->mmco_index) < 0 &&
+ (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
* about there being no actual duplicates.
* be fixed. */
if (h->short_ref_count) {
if (prev) {
- av_image_copy(h->short_ref[0]->data, h->short_ref[0]->linesize,
- (const uint8_t**)prev->data, prev->linesize,
+ av_image_copy(h->short_ref[0]->f.data, h->short_ref[0]->f.linesize,
+ (const uint8_t**)prev->f.data, prev->f.linesize,
s->avctx->pix_fmt, s->mb_width*16, s->mb_height*16);
h->short_ref[0]->poc = prev->poc+2;
}
/* See if we have a decoded first field looking for a pair... */
if (s0->first_field) {
assert(s0->current_picture_ptr);
- assert(s0->current_picture_ptr->data[0]);
- assert(s0->current_picture_ptr->reference != DELAYED_PIC_REF);
+ assert(s0->current_picture_ptr->f.data[0]);
+ assert(s0->current_picture_ptr->f.reference != DELAYED_PIC_REF);
/* figure out if we have a complementary field pair */
if (!FIELD_PICTURE || s->picture_structure == last_pic_structure) {
} else {
if (h->nal_ref_idc &&
- s0->current_picture_ptr->reference &&
+ s0->current_picture_ptr->f.reference &&
s0->current_picture_ptr->frame_num != h->frame_num) {
/*
* This and previous field were reference, but had
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);
+ if(h->nal_ref_idc && ff_h264_decode_ref_pic_marking(h0, &s->gb) < 0 &&
+ (s->avctx->err_recognition & AV_EF_EXPLODE))
+ return AVERROR_INVALIDDATA;
if(FRAME_MBAFF){
ff_h264_fill_mbaff_ref_list(h);
int *ref2frm= h->ref2frm[h->slice_num&(MAX_SLICES-1)][j];
for(i=0; i<16; i++){
id_list[i]= 60;
- if(h->ref_list[j][i].data[0]){
+ if (h->ref_list[j][i].f.data[0]) {
int k;
- uint8_t *base= h->ref_list[j][i].base[0];
+ uint8_t *base = h->ref_list[j][i].f.base[0];
for(k=0; k<h->short_ref_count; k++)
- if(h->short_ref[k]->base[0] == base){
+ if (h->short_ref[k]->f.base[0] == base) {
id_list[i]= k;
break;
}
for(k=0; k<h->long_ref_count; k++)
- if(h->long_ref[k] && h->long_ref[k]->base[0] == base){
+ if (h->long_ref[k] && h->long_ref[k]->f.base[0] == base) {
id_list[i]= h->short_ref_count + k;
break;
}
ref2frm[1]= -1;
for(i=0; i<16; i++)
ref2frm[i+2]= 4*id_list[i]
- +(h->ref_list[j][i].reference&3);
+ + (h->ref_list[j][i].f.reference & 3);
ref2frm[18+0]=
ref2frm[18+1]= -1;
for(i=16; i<48; i++)
ref2frm[i+4]= 4*id_list[(i-16)>>1]
- +(h->ref_list[j][i].reference&3);
+ + (h->ref_list[j][i].f.reference & 3);
}
//FIXME: fix draw_edges+PAFF+frame threads
const int b_xy= h->mb2b_xy[top_xy] + 3*b_stride;
const int b8_xy= 4*top_xy + 2;
int (*ref2frm)[64] = h->ref2frm[ h->slice_table[top_xy]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
- AV_COPY128(mv_dst - 1*8, s->current_picture.motion_val[list][b_xy + 0]);
+ AV_COPY128(mv_dst - 1*8, s->current_picture.f.motion_val[list][b_xy + 0]);
ref_cache[0 - 1*8]=
- ref_cache[1 - 1*8]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 0]];
+ ref_cache[1 - 1*8]= ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 0]];
ref_cache[2 - 1*8]=
- ref_cache[3 - 1*8]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 1]];
+ ref_cache[3 - 1*8]= ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 1]];
}else{
AV_ZERO128(mv_dst - 1*8);
AV_WN32A(&ref_cache[0 - 1*8], ((LIST_NOT_USED)&0xFF)*0x01010101u);
const int b_xy= h->mb2b_xy[left_xy[LTOP]] + 3;
const int b8_xy= 4*left_xy[LTOP] + 1;
int (*ref2frm)[64] = h->ref2frm[ h->slice_table[left_xy[LTOP]]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
- AV_COPY32(mv_dst - 1 + 0, s->current_picture.motion_val[list][b_xy + b_stride*0]);
- AV_COPY32(mv_dst - 1 + 8, s->current_picture.motion_val[list][b_xy + b_stride*1]);
- AV_COPY32(mv_dst - 1 +16, s->current_picture.motion_val[list][b_xy + b_stride*2]);
- AV_COPY32(mv_dst - 1 +24, s->current_picture.motion_val[list][b_xy + b_stride*3]);
+ AV_COPY32(mv_dst - 1 + 0, s->current_picture.f.motion_val[list][b_xy + b_stride*0]);
+ AV_COPY32(mv_dst - 1 + 8, s->current_picture.f.motion_val[list][b_xy + b_stride*1]);
+ AV_COPY32(mv_dst - 1 + 16, s->current_picture.f.motion_val[list][b_xy + b_stride*2]);
+ AV_COPY32(mv_dst - 1 + 24, s->current_picture.f.motion_val[list][b_xy + b_stride*3]);
ref_cache[-1 + 0]=
- ref_cache[-1 + 8]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 2*0]];
+ ref_cache[-1 + 8]= ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 2*0]];
ref_cache[-1 + 16]=
- ref_cache[-1 + 24]= ref2frm[list][s->current_picture.ref_index[list][b8_xy + 2*1]];
+ ref_cache[-1 + 24]= ref2frm[list][s->current_picture.f.ref_index[list][b8_xy + 2*1]];
}else{
AV_ZERO32(mv_dst - 1 + 0);
AV_ZERO32(mv_dst - 1 + 8);
}
{
- int8_t *ref = &s->current_picture.ref_index[list][4*mb_xy];
+ int8_t *ref = &s->current_picture.f.ref_index[list][4*mb_xy];
int (*ref2frm)[64] = h->ref2frm[ h->slice_num&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 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;
}
{
- int16_t (*mv_src)[2] = &s->current_picture.motion_val[list][4*s->mb_x + 4*s->mb_y*b_stride];
+ int16_t (*mv_src)[2] = &s->current_picture.f.motion_val[list][4*s->mb_x + 4*s->mb_y*b_stride];
AV_COPY128(mv_dst + 8*0, mv_src + 0*b_stride);
AV_COPY128(mv_dst + 8*1, mv_src + 1*b_stride);
AV_COPY128(mv_dst + 8*2, mv_src + 2*b_stride);
left_xy[LBOT] = left_xy[LTOP] = mb_xy-1;
if(FRAME_MBAFF){
- const int left_mb_field_flag = IS_INTERLACED(s->current_picture.mb_type[mb_xy-1]);
+ const int left_mb_field_flag = IS_INTERLACED(s->current_picture.f.mb_type[mb_xy - 1]);
const int curr_mb_field_flag = IS_INTERLACED(mb_type);
if(s->mb_y&1){
if (left_mb_field_flag != curr_mb_field_flag) {
}
}else{
if(curr_mb_field_flag){
- top_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy ]>>7)&1)-1);
+ top_xy += s->mb_stride & (((s->current_picture.f.mb_type[top_xy] >> 7) & 1) - 1);
}
if (left_mb_field_flag != curr_mb_field_flag) {
left_xy[LBOT] += s->mb_stride;
//for sufficiently low qp, filtering wouldn't do anything
//this is a conservative estimate: could also check beta_offset and more accurate chroma_qp
int qp_thresh = h->qp_thresh; //FIXME strictly we should store qp_thresh for each mb of a slice
- int qp = s->current_picture.qscale_table[mb_xy];
+ int qp = s->current_picture.f.qscale_table[mb_xy];
if(qp <= qp_thresh
- && (left_xy[LTOP]<0 || ((qp + s->current_picture.qscale_table[left_xy[LTOP]] + 1)>>1) <= qp_thresh)
- && (top_xy <0 || ((qp + s->current_picture.qscale_table[top_xy ] + 1)>>1) <= qp_thresh)){
+ && (left_xy[LTOP] < 0 || ((qp + s->current_picture.f.qscale_table[left_xy[LTOP]] + 1) >> 1) <= qp_thresh)
+ && (top_xy < 0 || ((qp + s->current_picture.f.qscale_table[top_xy ] + 1) >> 1) <= qp_thresh)) {
if(!FRAME_MBAFF)
return 1;
- if( (left_xy[LTOP]< 0 || ((qp + s->current_picture.qscale_table[left_xy[LBOT] ] + 1)>>1) <= qp_thresh)
- && (top_xy < s->mb_stride || ((qp + s->current_picture.qscale_table[top_xy -s->mb_stride] + 1)>>1) <= qp_thresh))
+ if ((left_xy[LTOP] < 0 || ((qp + s->current_picture.f.qscale_table[left_xy[LBOT] ] + 1) >> 1) <= qp_thresh) &&
+ (top_xy < s->mb_stride || ((qp + s->current_picture.f.qscale_table[top_xy - s->mb_stride] + 1) >> 1) <= qp_thresh))
return 1;
}
}
- top_type = s->current_picture.mb_type[top_xy];
- left_type[LTOP] = s->current_picture.mb_type[left_xy[LTOP]];
- left_type[LBOT] = s->current_picture.mb_type[left_xy[LBOT]];
+ top_type = s->current_picture.f.mb_type[top_xy];
+ left_type[LTOP] = s->current_picture.f.mb_type[left_xy[LTOP]];
+ left_type[LBOT] = s->current_picture.f.mb_type[left_xy[LBOT]];
if(h->deblocking_filter == 2){
if(h->slice_table[top_xy ] != h->slice_num) top_type= 0;
if(h->slice_table[left_xy[LBOT]] != h->slice_num) left_type[LTOP]= left_type[LBOT]= 0;
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++){
int mb_xy, mb_type;
mb_xy = h->mb_xy = mb_x + mb_y*s->mb_stride;
h->slice_num= h->slice_table[mb_xy];
- mb_type= s->current_picture.mb_type[mb_xy];
+ mb_type = s->current_picture.f.mb_type[mb_xy];
h->list_count= h->list_counts[mb_xy];
if(FRAME_MBAFF)
s->mb_x= mb_x;
s->mb_y= mb_y;
- dest_y = s->current_picture.data[0] + ((mb_x << pixel_shift) + mb_y * s->linesize ) * 16;
- dest_cb = s->current_picture.data[1] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444);
- dest_cr = s->current_picture.data[2] + ((mb_x << pixel_shift) + mb_y * s->uvlinesize) * (8 << CHROMA444);
+ 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) * (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.qscale_table[mb_xy]);
- h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.qscale_table[mb_xy]);
+ h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.f.qscale_table[mb_xy]);
+ h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.f.qscale_table[mb_xy]);
if (FRAME_MBAFF) {
ff_h264_filter_mb (h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
MpegEncContext * const s = &h->s;
const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
int mb_type = (h->slice_table[mb_xy-1] == h->slice_num)
- ? s->current_picture.mb_type[mb_xy-1]
+ ? s->current_picture.f.mb_type[mb_xy - 1]
: (h->slice_table[mb_xy-s->mb_stride] == h->slice_num)
- ? s->current_picture.mb_type[mb_xy-s->mb_stride]
+ ? s->current_picture.f.mb_type[mb_xy - s->mb_stride]
: 0;
h->mb_mbaff = h->mb_field_decoding_flag = IS_INTERLACED(mb_type) ? 1 : 0;
}
}
}
}
-
-#if 0
- for(;s->mb_y < s->mb_height; s->mb_y++){
- for(;s->mb_x < s->mb_width; s->mb_x++){
- int ret= decode_mb(h);
-
- ff_h264_hl_decode_mb(h);
-
- if(ret<0){
- av_log(s->avctx, AV_LOG_ERROR, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
-
- return -1;
- }
-
- if(++s->mb_x >= s->mb_width){
- s->mb_x=0;
- if(++s->mb_y >= s->mb_height){
- if(get_bits_count(s->gb) == s->gb.size_in_bits){
- 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;
- }else{
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
-
- return -1;
- }
- }
- }
-
- if(get_bits_count(s->?gb) >= s->gb?.size_in_bits){
- if(get_bits_count(s->gb) == s->gb.size_in_bits){
- 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;
- }else{
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
-
- return -1;
- }
- }
- }
- s->mb_x=0;
- ff_draw_horiz_band(s, 16*s->mb_y, 16);
- }
-#endif
}
/**
* @param h h264 master context
* @param context_count number of contexts to execute
*/
-static void execute_decode_slices(H264Context *h, int context_count){
+static int execute_decode_slices(H264Context *h, int context_count){
MpegEncContext * const s = &h->s;
AVCodecContext * const avctx= s->avctx;
H264Context *hx;
int i;
- if (s->avctx->hwaccel)
- return;
- if(s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
- return;
+ if (s->avctx->hwaccel || s->avctx->codec->capabilities&CODEC_CAP_HWACCEL_VDPAU)
+ return 0;
if(context_count == 1) {
- decode_slice(avctx, &h);
+ return decode_slice(avctx, &h);
} else {
for(i = 1; i < context_count; i++) {
hx = h->thread_context[i];
for(i = 1; i < context_count; i++)
h->s.error_count += h->thread_context[i]->s.error_count;
}
+
+ return 0;
}
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;
- s->current_picture_ptr->key_frame |=
+ s->current_picture_ptr->f.key_frame |=
(hx->nal_unit_type == NAL_IDR_SLICE) ||
(h->sei_recovery_frame_cnt >= 0);
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 (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) {
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);
- ff_h264_pred_init(&h->hpc, s->codec_id, h->sps.bit_depth_luma);
+ 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(avctx, AV_LOG_DEBUG, "Unsupported bit depth: %d\n", h->sps.bit_depth_luma);
+ av_log(avctx, AV_LOG_ERROR, "Unsupported bit depth: %d\n", h->sps.bit_depth_luma);
return -1;
}
}
//FIXME factorize this with the output code below
out = h->delayed_pic[0];
out_idx = 0;
- for(i=1; h->delayed_pic[i] && !h->delayed_pic[i]->key_frame && !h->delayed_pic[i]->mmco_reset; i++)
+ for (i = 1; h->delayed_pic[i] && !h->delayed_pic[i]->f.key_frame && !h->delayed_pic[i]->mmco_reset; i++)
if(h->delayed_pic[i]->poc < out->poc){
out = h->delayed_pic[i];
out_idx = i;
STOP_TIMER("get_se_golomb");
}
-#if 0
- printf("testing 4x4 (I)DCT\n");
-
- DCTELEM block[16];
- uint8_t src[16], ref[16];
- uint64_t error= 0, max_error=0;
-
- for(i=0; i<COUNT; i++){
- int j;
-// printf("%d %d %d\n", r1, r2, (r2-r1)*16);
- for(j=0; j<16; j++){
- ref[j]= random()%255;
- src[j]= random()%255;
- }
-
- h264_diff_dct_c(block, src, ref, 4);
-
- //normalize
- for(j=0; j<16; j++){
-// printf("%d ", block[j]);
- block[j]= block[j]*4;
- if(j&1) block[j]= (block[j]*4 + 2)/5;
- if(j&4) block[j]= (block[j]*4 + 2)/5;
- }
-// printf("\n");
-
- h->h264dsp.h264_idct_add(ref, block, 4);
-/* for(j=0; j<16; j++){
- printf("%d ", ref[j]);
- }
- printf("\n");*/
-
- for(j=0; j<16; j++){
- int diff= FFABS(src[j] - ref[j]);
-
- error+= diff*diff;
- max_error= FFMAX(max_error, diff);
- }
- }
- printf("error=%f max_error=%d\n", ((float)error)/COUNT/16, (int)max_error );
- printf("testing quantizer\n");
- for(qp=0; qp<52; qp++){
- for(i=0; i<16; i++)
- src1_block[i]= src2_block[i]= random()%255;
-
- }
- printf("Testing NAL layer\n");
-
- uint8_t bitstream[COUNT];
- uint8_t nal[COUNT*2];
- H264Context h;
- memset(&h, 0, sizeof(H264Context));
-
- for(i=0; i<COUNT; i++){
- int zeros= i;
- int nal_length;
- int consumed;
- int out_length;
- uint8_t *out;
- int j;
-
- for(j=0; j<COUNT; j++){
- bitstream[j]= (random() % 255) + 1;
- }
-
- for(j=0; j<zeros; j++){
- int pos= random() % COUNT;
- while(bitstream[pos] == 0){
- pos++;
- pos %= COUNT;
- }
- bitstream[pos]=0;
- }
-
- START_TIMER
-
- nal_length= encode_nal(&h, nal, bitstream, COUNT, COUNT*2);
- if(nal_length<0){
- printf("encoding failed\n");
- return -1;
- }
-
- out= ff_h264_decode_nal(&h, nal, &out_length, &consumed, nal_length);
-
- STOP_TIMER("NAL")
-
- if(out_length != COUNT){
- printf("incorrect length %d %d\n", out_length, COUNT);
- return -1;
- }
-
- if(consumed != nal_length){
- printf("incorrect consumed length %d %d\n", nal_length, consumed);
- return -1;
- }
-
- if(memcmp(bitstream, out, COUNT)){
- printf("mismatch\n");
- return -1;
- }
- }
-#endif
-
printf("Testing RBSP\n");
};
AVCodec ff_h264_decoder = {
- "h264",
- AVMEDIA_TYPE_VIDEO,
- CODEC_ID_H264,
- sizeof(H264Context),
- ff_h264_decode_init,
- NULL,
- ff_h264_decode_end,
- decode_frame,
- /*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | CODEC_CAP_DELAY |
- CODEC_CAP_SLICE_THREADS | CODEC_CAP_FRAME_THREADS,
+ .name = "h264",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .id = CODEC_ID_H264,
+ .priv_data_size = sizeof(H264Context),
+ .init = ff_h264_decode_init,
+ .close = ff_h264_decode_end,
+ .decode = decode_frame,
+ .capabilities = /*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | CODEC_CAP_DELAY |
+ CODEC_CAP_SLICE_THREADS | CODEC_CAP_FRAME_THREADS,
.flush= flush_dpb,
.long_name = NULL_IF_CONFIG_SMALL("H.264 / AVC / MPEG-4 AVC / MPEG-4 part 10"),
.init_thread_copy = ONLY_IF_THREADS_ENABLED(decode_init_thread_copy),
#if CONFIG_H264_VDPAU_DECODER
AVCodec ff_h264_vdpau_decoder = {
- "h264_vdpau",
- AVMEDIA_TYPE_VIDEO,
- CODEC_ID_H264,
- sizeof(H264Context),
- ff_h264_decode_init,
- NULL,
- ff_h264_decode_end,
- decode_frame,
- CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,
+ .name = "h264_vdpau",
+ .type = AVMEDIA_TYPE_VIDEO,
+ .id = CODEC_ID_H264,
+ .priv_data_size = sizeof(H264Context),
+ .init = ff_h264_decode_init,
+ .close = ff_h264_decode_end,
+ .decode = decode_frame,
+ .capabilities = CODEC_CAP_DR1 | CODEC_CAP_DELAY | CODEC_CAP_HWACCEL_VDPAU,
.flush= flush_dpb,
.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},
projects / ffmpeg / blobdiff