* H.26L/H.264/AVC/JVT/14496-10/... direct mb/block decoding
* Copyright (c) 2003 Michael Niedermayer <michaelni@gmx.at>
*
- * This file is part of FFmpeg.
+ * This file is part of Libav.
*
- * FFmpeg is free software; you can redistribute it and/or
+ * Libav 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.
*
- * FFmpeg is distributed in the hope that it will be useful,
+ * Libav 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 FFmpeg; if not, write to the Free Software
+ * License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
- * @file libavcodec/h264_direct.c
+ * @file
* H.264 / AVC / MPEG4 part10 direct mb/block decoding.
* @author Michael Niedermayer <michaelni@gmx.at>
*/
#include "mpegvideo.h"
#include "h264.h"
#include "rectangle.h"
+#include "thread.h"
//#undef NDEBUG
#include <assert.h>
poc= (poc&~3) + rfield + 1;
for(j=start; j<end; j++){
- if(4*h->ref_list[0][j].frame_num + (h->ref_list[0][j].reference&3) == poc){
+ if (4 * h->ref_list[0][j].frame_num + (h->ref_list[0][j].f.reference & 3) == poc) {
int cur_ref= mbafi ? (j-16)^field : j;
map[list][2*old_ref + (rfield^field) + 16] = cur_ref;
if(rfield == field || !interl)
Picture * const cur = s->current_picture_ptr;
int list, j, field;
int sidx= (s->picture_structure&1)^1;
- int ref1sidx= (ref1->reference&1)^1;
+ int ref1sidx = (ref1->f.reference&1)^1;
for(list=0; list<2; list++){
cur->ref_count[sidx][list] = h->ref_count[list];
for(j=0; j<h->ref_count[list]; j++)
- cur->ref_poc[sidx][list][j] = 4*h->ref_list[list][j].frame_num + (h->ref_list[list][j].reference&3);
+ cur->ref_poc[sidx][list][j] = 4 * h->ref_list[list][j].frame_num + (h->ref_list[list][j].f.reference & 3);
}
if(s->picture_structure == PICT_FRAME){
int *col_poc = h->ref_list[1]->field_poc;
h->col_parity= (FFABS(col_poc[0] - cur_poc) >= FFABS(col_poc[1] - cur_poc));
ref1sidx=sidx= h->col_parity;
- }else if(!(s->picture_structure & h->ref_list[1][0].reference) && !h->ref_list[1][0].mbaff){ // FL -> FL & differ parity
- h->col_fieldoff= s->mb_stride*(2*(h->ref_list[1][0].reference) - 3);
+ } else if (!(s->picture_structure & h->ref_list[1][0].f.reference) && !h->ref_list[1][0].mbaff) { // FL -> FL & differ parity
+ h->col_fieldoff = 2 * h->ref_list[1][0].f.reference - 3;
}
- if(cur->pict_type != FF_B_TYPE || h->direct_spatial_mv_pred)
+ if (cur->f.pict_type != AV_PICTURE_TYPE_B || h->direct_spatial_mv_pred)
return;
for(list=0; list<2; list++){
}
}
+static void await_reference_mb_row(H264Context * const h, Picture *ref, int mb_y)
+{
+ int ref_field = ref->f.reference - 1;
+ int ref_field_picture = ref->field_picture;
+ int ref_height = 16*h->s.mb_height >> ref_field_picture;
+
+ if(!HAVE_THREADS || !(h->s.avctx->active_thread_type&FF_THREAD_FRAME))
+ return;
+
+ //FIXME it can be safe to access mb stuff
+ //even if pixels aren't deblocked yet
+
+ ff_thread_await_progress(&ref->f,
+ FFMIN(16 * mb_y >> ref_field_picture, ref_height - 1),
+ ref_field_picture && ref_field);
+}
+
static void pred_spatial_direct_motion(H264Context * const h, int *mb_type){
MpegEncContext * const s = &h->s;
- int b8_stride = h->b8_stride;
+ int b8_stride = 2;
int b4_stride = h->b_stride;
- int mb_xy = h->mb_xy;
+ int mb_xy = h->mb_xy, mb_y = s->mb_y;
int mb_type_col[2];
const int16_t (*l1mv0)[2], (*l1mv1)[2];
const int8_t *l1ref0, *l1ref1;
const int is_b8x8 = IS_8X8(*mb_type);
- unsigned int sub_mb_type= MB_TYPE_L0L1;;
+ unsigned int sub_mb_type= MB_TYPE_L0L1;
int i8, i4;
int ref[2];
int mv[2];
int list;
- assert(h->ref_list[1][0].reference&3);
+ assert(h->ref_list[1][0].f.reference & 3);
+
+ await_reference_mb_row(h, &h->ref_list[1][0], s->mb_y + !!IS_INTERLACED(*mb_type));
#define MB_TYPE_16x16_OR_INTRA (MB_TYPE_16x16|MB_TYPE_INTRA4x4|MB_TYPE_INTRA16x16|MB_TYPE_INTRA_PCM)
- *mb_type |= MB_TYPE_L0L1;
/* ref = min(neighbors) */
for(list=0; list<2; list++){
int match_count= (left_ref==ref[list]) + (top_ref==ref[list]) + (refc==ref[list]);
if(match_count > 1){ //most common
- mv[list]= (mid_pred(A[0], B[0], C[0])&0xFFFF)
- +(mid_pred(A[1], B[1], C[1])<<16);
+ mv[list]= pack16to32(mid_pred(A[0], B[0], C[0]),
+ mid_pred(A[1], B[1], C[1]) );
}else {
assert(match_count==1);
if(left_ref==ref[list]){
- mv[list]= *(uint32_t*)A;
+ mv[list]= AV_RN32A(A);
}else if(top_ref==ref[list]){
- mv[list]= *(uint32_t*)B;
+ mv[list]= AV_RN32A(B);
}else{
- mv[list]= *(uint32_t*)C;
+ mv[list]= AV_RN32A(C);
}
}
}else{
sub_mb_type |= MB_TYPE_L0L1;
}
- if(!is_b8x8 && (mv[0]|mv[1]) == 0){
+ if(!(is_b8x8|mv[0]|mv[1])){
fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, (uint8_t)ref[0], 1);
fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, (uint8_t)ref[1], 1);
fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, 0, 4);
return;
}
- if(IS_INTERLACED(h->ref_list[1][0].mb_type[mb_xy])){ // AFL/AFR/FR/FL -> AFL/FL
- if(!IS_INTERLACED(*mb_type)){ // AFR/FR -> AFL/FL
+ if (IS_INTERLACED(h->ref_list[1][0].f.mb_type[mb_xy])) { // AFL/AFR/FR/FL -> AFL/FL
+ if (!IS_INTERLACED(*mb_type)) { // AFR/FR -> AFL/FL
+ mb_y = (s->mb_y&~1) + h->col_parity;
mb_xy= s->mb_x + ((s->mb_y&~1) + h->col_parity)*s->mb_stride;
b8_stride = 0;
}else{
- mb_xy += h->col_fieldoff; // non zero for FL -> FL & differ parity
+ mb_y += h->col_fieldoff;
+ mb_xy += s->mb_stride*h->col_fieldoff; // non zero for FL -> FL & differ parity
}
goto single_col;
}else{ // AFL/AFR/FR/FL -> AFR/FR
if(IS_INTERLACED(*mb_type)){ // AFL /FL -> AFR/FR
+ mb_y = s->mb_y&~1;
mb_xy= s->mb_x + (s->mb_y&~1)*s->mb_stride;
- mb_type_col[0] = h->ref_list[1][0].mb_type[mb_xy];
- mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy + s->mb_stride];
- b8_stride *= 3;
+ mb_type_col[0] = h->ref_list[1][0].f.mb_type[mb_xy];
+ mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy + s->mb_stride];
+ b8_stride = 2+4*s->mb_stride;
b4_stride *= 6;
+ if (IS_INTERLACED(mb_type_col[0]) != IS_INTERLACED(mb_type_col[1])) {
+ mb_type_col[0] &= ~MB_TYPE_INTERLACED;
+ mb_type_col[1] &= ~MB_TYPE_INTERLACED;
+ }
sub_mb_type |= MB_TYPE_16x16|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
if( (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)
}else{ // AFR/FR -> AFR/FR
single_col:
mb_type_col[0] =
- mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy];
+ mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy];
sub_mb_type |= MB_TYPE_16x16|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
if(!is_b8x8 && (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)){
}
}
- l1mv0 = &h->ref_list[1][0].motion_val[0][h->mb2b_xy [mb_xy]];
- l1mv1 = &h->ref_list[1][0].motion_val[1][h->mb2b_xy [mb_xy]];
- l1ref0 = &h->ref_list[1][0].ref_index [0][h->mb2b8_xy[mb_xy]];
- l1ref1 = &h->ref_list[1][0].ref_index [1][h->mb2b8_xy[mb_xy]];
+ await_reference_mb_row(h, &h->ref_list[1][0], mb_y);
+
+ l1mv0 = &h->ref_list[1][0].f.motion_val[0][h->mb2b_xy [mb_xy]];
+ l1mv1 = &h->ref_list[1][0].f.motion_val[1][h->mb2b_xy [mb_xy]];
+ l1ref0 = &h->ref_list[1][0].f.ref_index [0][4 * mb_xy];
+ l1ref1 = &h->ref_list[1][0].f.ref_index [1][4 * mb_xy];
if(!b8_stride){
if(s->mb_y&1){
- l1ref0 += h->b8_stride;
- l1ref1 += h->b8_stride;
+ l1ref0 += 2;
+ l1ref1 += 2;
l1mv0 += 2*b4_stride;
l1mv1 += 2*b4_stride;
}
fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[0], 1);
fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[1], 1);
+ assert(b8_stride==2);
/* col_zero_flag */
- if(!IS_INTRA(mb_type_col[0]) && !h->ref_list[1][0].long_ref && ( l1ref0[x8 + y8*b8_stride] == 0
- || (l1ref0[x8 + y8*b8_stride] < 0 && l1ref1[x8 + y8*b8_stride] == 0
+ if(!IS_INTRA(mb_type_col[0]) && !h->ref_list[1][0].long_ref && ( l1ref0[i8] == 0
+ || (l1ref0[i8] < 0 && l1ref1[i8] == 0
&& h->x264_build>33U))){
- const int16_t (*l1mv)[2]= l1ref0[x8 + y8*b8_stride] == 0 ? l1mv0 : l1mv1;
+ const int16_t (*l1mv)[2]= l1ref0[i8] == 0 ? l1mv0 : l1mv1;
if(IS_SUB_8X8(sub_mb_type)){
const int16_t *mv_col = l1mv[x8*3 + y8*3*b4_stride];
if(FFABS(mv_col[0]) <= 1 && FFABS(mv_col[1]) <= 1){
const int16_t *mv_col = l1mv[x8*2 + (i4&1) + (y8*2 + (i4>>1))*b4_stride];
if(FFABS(mv_col[0]) <= 1 && FFABS(mv_col[1]) <= 1){
if(ref[0] == 0)
- *(uint32_t*)h->mv_cache[0][scan8[i8*4+i4]] = 0;
+ AV_ZERO32(h->mv_cache[0][scan8[i8*4+i4]]);
if(ref[1] == 0)
- *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] = 0;
+ AV_ZERO32(h->mv_cache[1][scan8[i8*4+i4]]);
m++;
}
}
static void pred_temp_direct_motion(H264Context * const h, int *mb_type){
MpegEncContext * const s = &h->s;
- int b8_stride = h->b8_stride;
+ int b8_stride = 2;
int b4_stride = h->b_stride;
- int mb_xy = h->mb_xy;
+ int mb_xy = h->mb_xy, mb_y = s->mb_y;
int mb_type_col[2];
const int16_t (*l1mv0)[2], (*l1mv1)[2];
const int8_t *l1ref0, *l1ref1;
unsigned int sub_mb_type;
int i8, i4;
- assert(h->ref_list[1][0].reference&3);
+ assert(h->ref_list[1][0].f.reference & 3);
- if(IS_INTERLACED(h->ref_list[1][0].mb_type[mb_xy])){ // AFL/AFR/FR/FL -> AFL/FL
- if(!IS_INTERLACED(*mb_type)){ // AFR/FR -> AFL/FL
+ await_reference_mb_row(h, &h->ref_list[1][0], s->mb_y + !!IS_INTERLACED(*mb_type));
+
+ if (IS_INTERLACED(h->ref_list[1][0].f.mb_type[mb_xy])) { // AFL/AFR/FR/FL -> AFL/FL
+ if (!IS_INTERLACED(*mb_type)) { // AFR/FR -> AFL/FL
+ mb_y = (s->mb_y&~1) + h->col_parity;
mb_xy= s->mb_x + ((s->mb_y&~1) + h->col_parity)*s->mb_stride;
b8_stride = 0;
}else{
- mb_xy += h->col_fieldoff; // non zero for FL -> FL & differ parity
+ mb_y += h->col_fieldoff;
+ mb_xy += s->mb_stride*h->col_fieldoff; // non zero for FL -> FL & differ parity
}
goto single_col;
}else{ // AFL/AFR/FR/FL -> AFR/FR
if(IS_INTERLACED(*mb_type)){ // AFL /FL -> AFR/FR
+ mb_y = s->mb_y&~1;
mb_xy= s->mb_x + (s->mb_y&~1)*s->mb_stride;
- mb_type_col[0] = h->ref_list[1][0].mb_type[mb_xy];
- mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy + s->mb_stride];
- b8_stride *= 3;
+ mb_type_col[0] = h->ref_list[1][0].f.mb_type[mb_xy];
+ mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy + s->mb_stride];
+ b8_stride = 2+4*s->mb_stride;
b4_stride *= 6;
+ if (IS_INTERLACED(mb_type_col[0]) != IS_INTERLACED(mb_type_col[1])) {
+ mb_type_col[0] &= ~MB_TYPE_INTERLACED;
+ mb_type_col[1] &= ~MB_TYPE_INTERLACED;
+ }
sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
}else{ // AFR/FR -> AFR/FR
single_col:
mb_type_col[0] =
- mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy];
+ mb_type_col[1] = h->ref_list[1][0].f.mb_type[mb_xy];
sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
if(!is_b8x8 && (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)){
}
}
- l1mv0 = &h->ref_list[1][0].motion_val[0][h->mb2b_xy [mb_xy]];
- l1mv1 = &h->ref_list[1][0].motion_val[1][h->mb2b_xy [mb_xy]];
- l1ref0 = &h->ref_list[1][0].ref_index [0][h->mb2b8_xy[mb_xy]];
- l1ref1 = &h->ref_list[1][0].ref_index [1][h->mb2b8_xy[mb_xy]];
+ await_reference_mb_row(h, &h->ref_list[1][0], mb_y);
+
+ l1mv0 = &h->ref_list[1][0].f.motion_val[0][h->mb2b_xy [mb_xy]];
+ l1mv1 = &h->ref_list[1][0].f.motion_val[1][h->mb2b_xy [mb_xy]];
+ l1ref0 = &h->ref_list[1][0].f.ref_index [0][4 * mb_xy];
+ l1ref1 = &h->ref_list[1][0].f.ref_index [1][4 * mb_xy];
if(!b8_stride){
if(s->mb_y&1){
- l1ref0 += h->b8_stride;
- l1ref1 += h->b8_stride;
+ l1ref0 += 2;
+ l1ref1 += 2;
l1mv0 += 2*b4_stride;
l1mv1 += 2*b4_stride;
}
continue;
}
- ref0 = l1ref0[x8 + y8*b8_stride];
+ assert(b8_stride == 2);
+ ref0 = l1ref0[i8];
if(ref0 >= 0)
ref0 = map_col_to_list0[0][ref0 + ref_offset];
else{
- ref0 = map_col_to_list0[1][l1ref1[x8 + y8*b8_stride] + ref_offset];
+ ref0 = map_col_to_list0[1][l1ref1[i8] + ref_offset];
l1mv= l1mv1;
}
scale = dist_scale_factor[ref0];
int16_t *mv_l0 = h->mv_cache[0][scan8[i8*4+i4]];
mv_l0[0] = (scale * mv_col[0] + 128) >> 8;
mv_l0[1] = (scale * mv_col[1] + 128) >> 8;
- *(uint32_t*)h->mv_cache[1][scan8[i8*4+i4]] =
- pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]);
+ AV_WN32A(h->mv_cache[1][scan8[i8*4+i4]],
+ pack16to32(mv_l0[0]-mv_col[0],mv_l0[1]-mv_col[1]));
}
}
}
projects / ffmpeg / blobdiff