Skip to content

F
ffmpeg.wasm-core
Commit 80621249 authored by Diego Biurrun's avatar Diego Biurrun
Browse files
Options

h264: move fill_decode_neighbors()/fill_decode_caches() to h264_mvpred.h 

This fixes a bunch of unused function warnings.
parent ac4a85f4
Show whitespace changes
Inline Side-by-side
Showing with 418 additions and 418 deletions
libavcodec/h264.h
View file @ 80621249
......@@ -770,424 +770,6 @@ static av_always_inline int get_chroma_qp(H264Context *h, int t, int qscale){
return h->pps.chroma_qp_table[t][qscale];
}
static void fill_decode_neighbors(H264Context *h, int mb_type){
MpegEncContext * const s = &h->s;
const int mb_xy= h->mb_xy;
int topleft_xy, top_xy, topright_xy, left_xy[LEFT_MBS];
static const uint8_t left_block_options[4][32]={
{0,1,2,3,7,10,8,11,3+0*4, 3+1*4, 3+2*4, 3+3*4, 1+4*4, 1+8*4, 1+5*4, 1+9*4},
{2,2,3,3,8,11,8,11,3+2*4, 3+2*4, 3+3*4, 3+3*4, 1+5*4, 1+9*4, 1+5*4, 1+9*4},
{0,0,1,1,7,10,7,10,3+0*4, 3+0*4, 3+1*4, 3+1*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4},
{0,2,0,2,7,10,7,10,3+0*4, 3+2*4, 3+0*4, 3+2*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4}
};
h->topleft_partition= -1;
top_xy = mb_xy - (s->mb_stride << MB_FIELD);
/* Wow, what a mess, why didn't they simplify the interlacing & intra
* stuff, I can't imagine that these complex rules are worth it. */
topleft_xy = top_xy - 1;
topright_xy= top_xy + 1;
left_xy[LBOT] = left_xy[LTOP] = mb_xy-1;
h->left_block = left_block_options[0];
if(FRAME_MBAFF){
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) {
left_xy[LBOT] = left_xy[LTOP] = mb_xy - s->mb_stride - 1;
if (curr_mb_field_flag) {
left_xy[LBOT] += s->mb_stride;
h->left_block = left_block_options[3];
} else {
topleft_xy += s->mb_stride;
// take top left mv from the middle of the mb, as opposed to all other modes which use the bottom right partition
h->topleft_partition = 0;
h->left_block = left_block_options[1];
}
}
}else{
if(curr_mb_field_flag){
topleft_xy += s->mb_stride & (((s->current_picture.f.mb_type[top_xy - 1] >> 7) & 1) - 1);
topright_xy += s->mb_stride & (((s->current_picture.f.mb_type[top_xy + 1] >> 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) {
if (curr_mb_field_flag) {
left_xy[LBOT] += s->mb_stride;
h->left_block = left_block_options[3];
} else {
h->left_block = left_block_options[2];
}
}
}
}
h->topleft_mb_xy = topleft_xy;
h->top_mb_xy = top_xy;
h->topright_mb_xy= topright_xy;
h->left_mb_xy[LTOP] = left_xy[LTOP];
h->left_mb_xy[LBOT] = left_xy[LBOT];
//FIXME do we need all in the context?
h->topleft_type = s->current_picture.f.mb_type[topleft_xy];
h->top_type = s->current_picture.f.mb_type[top_xy];
h->topright_type = s->current_picture.f.mb_type[topright_xy];
h->left_type[LTOP] = s->current_picture.f.mb_type[left_xy[LTOP]];
h->left_type[LBOT] = s->current_picture.f.mb_type[left_xy[LBOT]];
if(FMO){
if(h->slice_table[topleft_xy ] != h->slice_num) h->topleft_type = 0;
if(h->slice_table[top_xy ] != h->slice_num) h->top_type = 0;
if(h->slice_table[left_xy[LTOP] ] != h->slice_num) h->left_type[LTOP] = h->left_type[LBOT] = 0;
}else{
if(h->slice_table[topleft_xy ] != h->slice_num){
h->topleft_type = 0;
if(h->slice_table[top_xy ] != h->slice_num) h->top_type = 0;
if(h->slice_table[left_xy[LTOP] ] != h->slice_num) h->left_type[LTOP] = h->left_type[LBOT] = 0;
}
}
if(h->slice_table[topright_xy] != h->slice_num) h->topright_type= 0;
}
static void fill_decode_caches(H264Context *h, int mb_type){
MpegEncContext * const s = &h->s;
int topleft_xy, top_xy, topright_xy, left_xy[LEFT_MBS];
int topleft_type, top_type, topright_type, left_type[LEFT_MBS];
const uint8_t * left_block= h->left_block;
int i;
uint8_t *nnz;
uint8_t *nnz_cache;
topleft_xy = h->topleft_mb_xy;
top_xy = h->top_mb_xy;
topright_xy = h->topright_mb_xy;
left_xy[LTOP] = h->left_mb_xy[LTOP];
left_xy[LBOT] = h->left_mb_xy[LBOT];
topleft_type = h->topleft_type;
top_type = h->top_type;
topright_type = h->topright_type;
left_type[LTOP]= h->left_type[LTOP];
left_type[LBOT]= h->left_type[LBOT];
if(!IS_SKIP(mb_type)){
if(IS_INTRA(mb_type)){
int type_mask= h->pps.constrained_intra_pred ? IS_INTRA(-1) : -1;
h->topleft_samples_available=
h->top_samples_available=
h->left_samples_available= 0xFFFF;
h->topright_samples_available= 0xEEEA;
if(!(top_type & type_mask)){
h->topleft_samples_available= 0xB3FF;
h->top_samples_available= 0x33FF;
h->topright_samples_available= 0x26EA;
}
if(IS_INTERLACED(mb_type) != IS_INTERLACED(left_type[LTOP])){
if(IS_INTERLACED(mb_type)){
if(!(left_type[LTOP] & type_mask)){
h->topleft_samples_available&= 0xDFFF;
h->left_samples_available&= 0x5FFF;
}
if(!(left_type[LBOT] & type_mask)){
h->topleft_samples_available&= 0xFF5F;
h->left_samples_available&= 0xFF5F;
}
}else{
int left_typei = s->current_picture.f.mb_type[left_xy[LTOP] + s->mb_stride];
assert(left_xy[LTOP] == left_xy[LBOT]);
if(!((left_typei & type_mask) && (left_type[LTOP] & type_mask))){
h->topleft_samples_available&= 0xDF5F;
h->left_samples_available&= 0x5F5F;
}
}
}else{
if(!(left_type[LTOP] & type_mask)){
h->topleft_samples_available&= 0xDF5F;
h->left_samples_available&= 0x5F5F;
}
}
if(!(topleft_type & type_mask))
h->topleft_samples_available&= 0x7FFF;
if(!(topright_type & type_mask))
h->topright_samples_available&= 0xFBFF;
if(IS_INTRA4x4(mb_type)){
if(IS_INTRA4x4(top_type)){
AV_COPY32(h->intra4x4_pred_mode_cache+4+8*0, h->intra4x4_pred_mode + h->mb2br_xy[top_xy]);
}else{
h->intra4x4_pred_mode_cache[4+8*0]=
h->intra4x4_pred_mode_cache[5+8*0]=
h->intra4x4_pred_mode_cache[6+8*0]=
h->intra4x4_pred_mode_cache[7+8*0]= 2 - 3*!(top_type & type_mask);
}
for(i=0; i<2; i++){
if(IS_INTRA4x4(left_type[LEFT(i)])){
int8_t *mode= h->intra4x4_pred_mode + h->mb2br_xy[left_xy[LEFT(i)]];
h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]= mode[6-left_block[0+2*i]];
h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= mode[6-left_block[1+2*i]];
}else{
h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]=
h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= 2 - 3*!(left_type[LEFT(i)] & type_mask);
}
}
}
}
/*
0 . T T. T T T T
1 L . .L . . . .
2 L . .L . . . .
3 . T TL . . . .
4 L . .L . . . .
5 L . .. . . . .
*/
//FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec)
nnz_cache = h->non_zero_count_cache;
if(top_type){
nnz = h->non_zero_count[top_xy];
AV_COPY32(&nnz_cache[4+8* 0], &nnz[4*3]);
if(CHROMA444){
AV_COPY32(&nnz_cache[4+8* 5], &nnz[4* 7]);
AV_COPY32(&nnz_cache[4+8*10], &nnz[4*11]);
}else{
AV_COPY32(&nnz_cache[4+8* 5], &nnz[4* 5]);
AV_COPY32(&nnz_cache[4+8*10], &nnz[4* 9]);
}
}else{
uint32_t top_empty = CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040;
AV_WN32A(&nnz_cache[4+8* 0], top_empty);
AV_WN32A(&nnz_cache[4+8* 5], top_empty);
AV_WN32A(&nnz_cache[4+8*10], top_empty);
}
for (i=0; i<2; i++) {
if(left_type[LEFT(i)]){
nnz = h->non_zero_count[left_xy[LEFT(i)]];
nnz_cache[3+8* 1 + 2*8*i]= nnz[left_block[8+0+2*i]];
nnz_cache[3+8* 2 + 2*8*i]= nnz[left_block[8+1+2*i]];
if(CHROMA444){
nnz_cache[3+8* 6 + 2*8*i]= nnz[left_block[8+0+2*i]+4*4];
nnz_cache[3+8* 7 + 2*8*i]= nnz[left_block[8+1+2*i]+4*4];
nnz_cache[3+8*11 + 2*8*i]= nnz[left_block[8+0+2*i]+8*4];
nnz_cache[3+8*12 + 2*8*i]= nnz[left_block[8+1+2*i]+8*4];
}else{
nnz_cache[3+8* 6 + 8*i]= nnz[left_block[8+4+2*i]];
nnz_cache[3+8*11 + 8*i]= nnz[left_block[8+5+2*i]];
}
}else{
nnz_cache[3+8* 1 + 2*8*i]=
nnz_cache[3+8* 2 + 2*8*i]=
nnz_cache[3+8* 6 + 2*8*i]=
nnz_cache[3+8* 7 + 2*8*i]=
nnz_cache[3+8*11 + 2*8*i]=
nnz_cache[3+8*12 + 2*8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64;
}
}
if( CABAC ) {
// top_cbp
if(top_type) {
h->top_cbp = h->cbp_table[top_xy];
} else {
h->top_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F;
}
// left_cbp
if (left_type[LTOP]) {
h->left_cbp = (h->cbp_table[left_xy[LTOP]] & 0x7F0)
| ((h->cbp_table[left_xy[LTOP]]>>(left_block[0]&(~1)))&2)
| (((h->cbp_table[left_xy[LBOT]]>>(left_block[2]&(~1)))&2) << 2);
} else {
h->left_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F;
}
}
}
if(IS_INTER(mb_type) || (IS_DIRECT(mb_type) && h->direct_spatial_mv_pred)){
int list;
int b_stride = h->b_stride;
for(list=0; list<h->list_count; list++){
int8_t *ref_cache = &h->ref_cache[list][scan8[0]];
int8_t *ref = s->current_picture.f.ref_index[list];
int16_t (*mv_cache)[2] = &h->mv_cache[list][scan8[0]];
int16_t (*mv)[2] = s->current_picture.f.motion_val[list];
if(!USES_LIST(mb_type, list)){
continue;
}
assert(!(IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred));
if(USES_LIST(top_type, list)){
const int b_xy= h->mb2b_xy[top_xy] + 3*b_stride;
AV_COPY128(mv_cache[0 - 1*8], mv[b_xy + 0]);
ref_cache[0 - 1*8]=
ref_cache[1 - 1*8]= ref[4*top_xy + 2];
ref_cache[2 - 1*8]=
ref_cache[3 - 1*8]= ref[4*top_xy + 3];
}else{
AV_ZERO128(mv_cache[0 - 1*8]);
AV_WN32A(&ref_cache[0 - 1*8], ((top_type ? LIST_NOT_USED : PART_NOT_AVAILABLE)&0xFF)*0x01010101);
}
if(mb_type & (MB_TYPE_16x8|MB_TYPE_8x8)){
for(i=0; i<2; i++){
int cache_idx = -1 + i*2*8;
if(USES_LIST(left_type[LEFT(i)], list)){
const int b_xy= h->mb2b_xy[left_xy[LEFT(i)]] + 3;
const int b8_xy= 4*left_xy[LEFT(i)] + 1;
AV_COPY32(mv_cache[cache_idx ], mv[b_xy + b_stride*left_block[0+i*2]]);
AV_COPY32(mv_cache[cache_idx+8], mv[b_xy + b_stride*left_block[1+i*2]]);
ref_cache[cache_idx ]= ref[b8_xy + (left_block[0+i*2]&~1)];
ref_cache[cache_idx+8]= ref[b8_xy + (left_block[1+i*2]&~1)];
}else{
AV_ZERO32(mv_cache[cache_idx ]);
AV_ZERO32(mv_cache[cache_idx+8]);
ref_cache[cache_idx ]=
ref_cache[cache_idx+8]= (left_type[LEFT(i)]) ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
}
}else{
if(USES_LIST(left_type[LTOP], list)){
const int b_xy= h->mb2b_xy[left_xy[LTOP]] + 3;
const int b8_xy= 4*left_xy[LTOP] + 1;
AV_COPY32(mv_cache[-1], mv[b_xy + b_stride*left_block[0]]);
ref_cache[-1]= ref[b8_xy + (left_block[0]&~1)];
}else{
AV_ZERO32(mv_cache[-1]);
ref_cache[-1]= left_type[LTOP] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
}
if(USES_LIST(topright_type, list)){
const int b_xy= h->mb2b_xy[topright_xy] + 3*b_stride;
AV_COPY32(mv_cache[4 - 1*8], mv[b_xy]);
ref_cache[4 - 1*8]= ref[4*topright_xy + 2];
}else{
AV_ZERO32(mv_cache[4 - 1*8]);
ref_cache[4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
if(ref_cache[4 - 1*8] < 0){
if(USES_LIST(topleft_type, list)){
const int b_xy = h->mb2b_xy[topleft_xy] + 3 + b_stride + (h->topleft_partition & 2*b_stride);
const int b8_xy= 4*topleft_xy + 1 + (h->topleft_partition & 2);
AV_COPY32(mv_cache[-1 - 1*8], mv[b_xy]);
ref_cache[-1 - 1*8]= ref[b8_xy];
}else{
AV_ZERO32(mv_cache[-1 - 1*8]);
ref_cache[-1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
}
if((mb_type&(MB_TYPE_SKIP|MB_TYPE_DIRECT2)) && !FRAME_MBAFF)
continue;
if(!(mb_type&(MB_TYPE_SKIP|MB_TYPE_DIRECT2))){
uint8_t (*mvd_cache)[2] = &h->mvd_cache[list][scan8[0]];
uint8_t (*mvd)[2] = h->mvd_table[list];
ref_cache[2+8*0] =
ref_cache[2+8*2] = PART_NOT_AVAILABLE;
AV_ZERO32(mv_cache[2+8*0]);
AV_ZERO32(mv_cache[2+8*2]);
if( CABAC ) {
if(USES_LIST(top_type, list)){
const int b_xy= h->mb2br_xy[top_xy];
AV_COPY64(mvd_cache[0 - 1*8], mvd[b_xy + 0]);
}else{
AV_ZERO64(mvd_cache[0 - 1*8]);
}
if(USES_LIST(left_type[LTOP], list)){
const int b_xy= h->mb2br_xy[left_xy[LTOP]] + 6;
AV_COPY16(mvd_cache[-1 + 0*8], mvd[b_xy - left_block[0]]);
AV_COPY16(mvd_cache[-1 + 1*8], mvd[b_xy - left_block[1]]);
}else{
AV_ZERO16(mvd_cache[-1 + 0*8]);
AV_ZERO16(mvd_cache[-1 + 1*8]);
}
if(USES_LIST(left_type[LBOT], list)){
const int b_xy= h->mb2br_xy[left_xy[LBOT]] + 6;
AV_COPY16(mvd_cache[-1 + 2*8], mvd[b_xy - left_block[2]]);
AV_COPY16(mvd_cache[-1 + 3*8], mvd[b_xy - left_block[3]]);
}else{
AV_ZERO16(mvd_cache[-1 + 2*8]);
AV_ZERO16(mvd_cache[-1 + 3*8]);
}
AV_ZERO16(mvd_cache[2+8*0]);
AV_ZERO16(mvd_cache[2+8*2]);
if(h->slice_type_nos == AV_PICTURE_TYPE_B){
uint8_t *direct_cache = &h->direct_cache[scan8[0]];
uint8_t *direct_table = h->direct_table;
fill_rectangle(direct_cache, 4, 4, 8, MB_TYPE_16x16>>1, 1);
if(IS_DIRECT(top_type)){
AV_WN32A(&direct_cache[-1*8], 0x01010101u*(MB_TYPE_DIRECT2>>1));
}else if(IS_8X8(top_type)){
int b8_xy = 4*top_xy;
direct_cache[0 - 1*8]= direct_table[b8_xy + 2];
direct_cache[2 - 1*8]= direct_table[b8_xy + 3];
}else{
AV_WN32A(&direct_cache[-1*8], 0x01010101*(MB_TYPE_16x16>>1));
}
if(IS_DIRECT(left_type[LTOP]))
direct_cache[-1 + 0*8]= MB_TYPE_DIRECT2>>1;
else if(IS_8X8(left_type[LTOP]))
direct_cache[-1 + 0*8]= direct_table[4*left_xy[LTOP] + 1 + (left_block[0]&~1)];
else
direct_cache[-1 + 0*8]= MB_TYPE_16x16>>1;
if(IS_DIRECT(left_type[LBOT]))
direct_cache[-1 + 2*8]= MB_TYPE_DIRECT2>>1;
else if(IS_8X8(left_type[LBOT]))
direct_cache[-1 + 2*8]= direct_table[4*left_xy[LBOT] + 1 + (left_block[2]&~1)];
else
direct_cache[-1 + 2*8]= MB_TYPE_16x16>>1;
}
}
}
if(FRAME_MBAFF){
#define MAP_MVS\
MAP_F2F(scan8[0] - 1 - 1*8, topleft_type)\
MAP_F2F(scan8[0] + 0 - 1*8, top_type)\
MAP_F2F(scan8[0] + 1 - 1*8, top_type)\
MAP_F2F(scan8[0] + 2 - 1*8, top_type)\
MAP_F2F(scan8[0] + 3 - 1*8, top_type)\
MAP_F2F(scan8[0] + 4 - 1*8, topright_type)\
MAP_F2F(scan8[0] - 1 + 0*8, left_type[LTOP])\
MAP_F2F(scan8[0] - 1 + 1*8, left_type[LTOP])\
MAP_F2F(scan8[0] - 1 + 2*8, left_type[LBOT])\
MAP_F2F(scan8[0] - 1 + 3*8, left_type[LBOT])
if(MB_FIELD){
#define MAP_F2F(idx, mb_type)\
if(!IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
h->ref_cache[list][idx] <<= 1;\
h->mv_cache[list][idx][1] /= 2;\
h->mvd_cache[list][idx][1] >>=1;\
}
MAP_MVS
#undef MAP_F2F
}else{
#define MAP_F2F(idx, mb_type)\
if(IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
h->ref_cache[list][idx] >>= 1;\
h->mv_cache[list][idx][1] <<= 1;\
h->mvd_cache[list][idx][1] <<= 1;\
}
MAP_MVS
#undef MAP_F2F
}
}
}
}
h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[LTOP]);
}
/**
* gets the predicted intra4x4 prediction mode.
*/
......
libavcodec/h264_mvpred.h
View file @ 80621249
......@@ -327,6 +327,424 @@ zeromv:
return;
}
static void fill_decode_neighbors(H264Context *h, int mb_type){
MpegEncContext * const s = &h->s;
const int mb_xy= h->mb_xy;
int topleft_xy, top_xy, topright_xy, left_xy[LEFT_MBS];
static const uint8_t left_block_options[4][32]={
{0,1,2,3,7,10,8,11,3+0*4, 3+1*4, 3+2*4, 3+3*4, 1+4*4, 1+8*4, 1+5*4, 1+9*4},
{2,2,3,3,8,11,8,11,3+2*4, 3+2*4, 3+3*4, 3+3*4, 1+5*4, 1+9*4, 1+5*4, 1+9*4},
{0,0,1,1,7,10,7,10,3+0*4, 3+0*4, 3+1*4, 3+1*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4},
{0,2,0,2,7,10,7,10,3+0*4, 3+2*4, 3+0*4, 3+2*4, 1+4*4, 1+8*4, 1+4*4, 1+8*4}
};
h->topleft_partition= -1;
top_xy = mb_xy - (s->mb_stride << MB_FIELD);
/* Wow, what a mess, why didn't they simplify the interlacing & intra
* stuff, I can't imagine that these complex rules are worth it. */
topleft_xy = top_xy - 1;
topright_xy= top_xy + 1;
left_xy[LBOT] = left_xy[LTOP] = mb_xy-1;
h->left_block = left_block_options[0];
if(FRAME_MBAFF){
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) {
left_xy[LBOT] = left_xy[LTOP] = mb_xy - s->mb_stride - 1;
if (curr_mb_field_flag) {
left_xy[LBOT] += s->mb_stride;
h->left_block = left_block_options[3];
} else {
topleft_xy += s->mb_stride;
// take top left mv from the middle of the mb, as opposed to all other modes which use the bottom right partition
h->topleft_partition = 0;
h->left_block = left_block_options[1];
}
}
}else{
if(curr_mb_field_flag){
topleft_xy += s->mb_stride & (((s->current_picture.f.mb_type[top_xy - 1] >> 7) & 1) - 1);
topright_xy += s->mb_stride & (((s->current_picture.f.mb_type[top_xy + 1] >> 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) {
if (curr_mb_field_flag) {
left_xy[LBOT] += s->mb_stride;
h->left_block = left_block_options[3];
} else {
h->left_block = left_block_options[2];
}
}
}
}
h->topleft_mb_xy = topleft_xy;
h->top_mb_xy = top_xy;
h->topright_mb_xy= topright_xy;
h->left_mb_xy[LTOP] = left_xy[LTOP];
h->left_mb_xy[LBOT] = left_xy[LBOT];
//FIXME do we need all in the context?
h->topleft_type = s->current_picture.f.mb_type[topleft_xy];
h->top_type = s->current_picture.f.mb_type[top_xy];
h->topright_type = s->current_picture.f.mb_type[topright_xy];
h->left_type[LTOP] = s->current_picture.f.mb_type[left_xy[LTOP]];
h->left_type[LBOT] = s->current_picture.f.mb_type[left_xy[LBOT]];
if(FMO){
if(h->slice_table[topleft_xy ] != h->slice_num) h->topleft_type = 0;
if(h->slice_table[top_xy ] != h->slice_num) h->top_type = 0;
if(h->slice_table[left_xy[LTOP] ] != h->slice_num) h->left_type[LTOP] = h->left_type[LBOT] = 0;
}else{
if(h->slice_table[topleft_xy ] != h->slice_num){
h->topleft_type = 0;
if(h->slice_table[top_xy ] != h->slice_num) h->top_type = 0;
if(h->slice_table[left_xy[LTOP] ] != h->slice_num) h->left_type[LTOP] = h->left_type[LBOT] = 0;
}
}
if(h->slice_table[topright_xy] != h->slice_num) h->topright_type= 0;
}
static void fill_decode_caches(H264Context *h, int mb_type){
MpegEncContext * const s = &h->s;
int topleft_xy, top_xy, topright_xy, left_xy[LEFT_MBS];
int topleft_type, top_type, topright_type, left_type[LEFT_MBS];
const uint8_t * left_block= h->left_block;
int i;
uint8_t *nnz;
uint8_t *nnz_cache;
topleft_xy = h->topleft_mb_xy;
top_xy = h->top_mb_xy;
topright_xy = h->topright_mb_xy;
left_xy[LTOP] = h->left_mb_xy[LTOP];
left_xy[LBOT] = h->left_mb_xy[LBOT];
topleft_type = h->topleft_type;
top_type = h->top_type;
topright_type = h->topright_type;
left_type[LTOP]= h->left_type[LTOP];
left_type[LBOT]= h->left_type[LBOT];
if(!IS_SKIP(mb_type)){
if(IS_INTRA(mb_type)){
int type_mask= h->pps.constrained_intra_pred ? IS_INTRA(-1) : -1;
h->topleft_samples_available=
h->top_samples_available=
h->left_samples_available= 0xFFFF;
h->topright_samples_available= 0xEEEA;
if(!(top_type & type_mask)){
h->topleft_samples_available= 0xB3FF;
h->top_samples_available= 0x33FF;
h->topright_samples_available= 0x26EA;
}
if(IS_INTERLACED(mb_type) != IS_INTERLACED(left_type[LTOP])){
if(IS_INTERLACED(mb_type)){
if(!(left_type[LTOP] & type_mask)){
h->topleft_samples_available&= 0xDFFF;
h->left_samples_available&= 0x5FFF;
}
if(!(left_type[LBOT] & type_mask)){
h->topleft_samples_available&= 0xFF5F;
h->left_samples_available&= 0xFF5F;
}
}else{
int left_typei = s->current_picture.f.mb_type[left_xy[LTOP] + s->mb_stride];
assert(left_xy[LTOP] == left_xy[LBOT]);
if(!((left_typei & type_mask) && (left_type[LTOP] & type_mask))){
h->topleft_samples_available&= 0xDF5F;
h->left_samples_available&= 0x5F5F;
}
}
}else{
if(!(left_type[LTOP] & type_mask)){
h->topleft_samples_available&= 0xDF5F;
h->left_samples_available&= 0x5F5F;
}
}
if(!(topleft_type & type_mask))
h->topleft_samples_available&= 0x7FFF;
if(!(topright_type & type_mask))
h->topright_samples_available&= 0xFBFF;
if(IS_INTRA4x4(mb_type)){
if(IS_INTRA4x4(top_type)){
AV_COPY32(h->intra4x4_pred_mode_cache+4+8*0, h->intra4x4_pred_mode + h->mb2br_xy[top_xy]);
}else{
h->intra4x4_pred_mode_cache[4+8*0]=
h->intra4x4_pred_mode_cache[5+8*0]=
h->intra4x4_pred_mode_cache[6+8*0]=
h->intra4x4_pred_mode_cache[7+8*0]= 2 - 3*!(top_type & type_mask);
}
for(i=0; i<2; i++){
if(IS_INTRA4x4(left_type[LEFT(i)])){
int8_t *mode= h->intra4x4_pred_mode + h->mb2br_xy[left_xy[LEFT(i)]];
h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]= mode[6-left_block[0+2*i]];
h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= mode[6-left_block[1+2*i]];
}else{
h->intra4x4_pred_mode_cache[3+8*1 + 2*8*i]=
h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= 2 - 3*!(left_type[LEFT(i)] & type_mask);
}
}
}
}
/*
0 . T T. T T T T
1 L . .L . . . .
2 L . .L . . . .
3 . T TL . . . .
4 L . .L . . . .
5 L . .. . . . .
*/
//FIXME constraint_intra_pred & partitioning & nnz (let us hope this is just a typo in the spec)
nnz_cache = h->non_zero_count_cache;
if(top_type){
nnz = h->non_zero_count[top_xy];
AV_COPY32(&nnz_cache[4+8* 0], &nnz[4*3]);
if(CHROMA444){
AV_COPY32(&nnz_cache[4+8* 5], &nnz[4* 7]);
AV_COPY32(&nnz_cache[4+8*10], &nnz[4*11]);
}else{
AV_COPY32(&nnz_cache[4+8* 5], &nnz[4* 5]);
AV_COPY32(&nnz_cache[4+8*10], &nnz[4* 9]);
}
}else{
uint32_t top_empty = CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040;
AV_WN32A(&nnz_cache[4+8* 0], top_empty);
AV_WN32A(&nnz_cache[4+8* 5], top_empty);
AV_WN32A(&nnz_cache[4+8*10], top_empty);
}
for (i=0; i<2; i++) {
if(left_type[LEFT(i)]){
nnz = h->non_zero_count[left_xy[LEFT(i)]];
nnz_cache[3+8* 1 + 2*8*i]= nnz[left_block[8+0+2*i]];
nnz_cache[3+8* 2 + 2*8*i]= nnz[left_block[8+1+2*i]];
if(CHROMA444){
nnz_cache[3+8* 6 + 2*8*i]= nnz[left_block[8+0+2*i]+4*4];
nnz_cache[3+8* 7 + 2*8*i]= nnz[left_block[8+1+2*i]+4*4];
nnz_cache[3+8*11 + 2*8*i]= nnz[left_block[8+0+2*i]+8*4];
nnz_cache[3+8*12 + 2*8*i]= nnz[left_block[8+1+2*i]+8*4];
}else{
nnz_cache[3+8* 6 + 8*i]= nnz[left_block[8+4+2*i]];
nnz_cache[3+8*11 + 8*i]= nnz[left_block[8+5+2*i]];
}
}else{
nnz_cache[3+8* 1 + 2*8*i]=
nnz_cache[3+8* 2 + 2*8*i]=
nnz_cache[3+8* 6 + 2*8*i]=
nnz_cache[3+8* 7 + 2*8*i]=
nnz_cache[3+8*11 + 2*8*i]=
nnz_cache[3+8*12 + 2*8*i]= CABAC && !IS_INTRA(mb_type) ? 0 : 64;
}
}
if( CABAC ) {
// top_cbp
if(top_type) {
h->top_cbp = h->cbp_table[top_xy];
} else {
h->top_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F;
}
// left_cbp
if (left_type[LTOP]) {
h->left_cbp = (h->cbp_table[left_xy[LTOP]] & 0x7F0)
| ((h->cbp_table[left_xy[LTOP]]>>(left_block[0]&(~1)))&2)
| (((h->cbp_table[left_xy[LBOT]]>>(left_block[2]&(~1)))&2) << 2);
} else {
h->left_cbp = IS_INTRA(mb_type) ? 0x7CF : 0x00F;
}
}
}
if(IS_INTER(mb_type) || (IS_DIRECT(mb_type) && h->direct_spatial_mv_pred)){
int list;
int b_stride = h->b_stride;
for(list=0; list<h->list_count; list++){
int8_t *ref_cache = &h->ref_cache[list][scan8[0]];
int8_t *ref = s->current_picture.f.ref_index[list];
int16_t (*mv_cache)[2] = &h->mv_cache[list][scan8[0]];
int16_t (*mv)[2] = s->current_picture.f.motion_val[list];
if(!USES_LIST(mb_type, list)){
continue;
}
assert(!(IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred));
if(USES_LIST(top_type, list)){
const int b_xy= h->mb2b_xy[top_xy] + 3*b_stride;
AV_COPY128(mv_cache[0 - 1*8], mv[b_xy + 0]);
ref_cache[0 - 1*8]=
ref_cache[1 - 1*8]= ref[4*top_xy + 2];
ref_cache[2 - 1*8]=
ref_cache[3 - 1*8]= ref[4*top_xy + 3];
}else{
AV_ZERO128(mv_cache[0 - 1*8]);
AV_WN32A(&ref_cache[0 - 1*8], ((top_type ? LIST_NOT_USED : PART_NOT_AVAILABLE)&0xFF)*0x01010101);
}
if(mb_type & (MB_TYPE_16x8|MB_TYPE_8x8)){
for(i=0; i<2; i++){
int cache_idx = -1 + i*2*8;
if(USES_LIST(left_type[LEFT(i)], list)){
const int b_xy= h->mb2b_xy[left_xy[LEFT(i)]] + 3;
const int b8_xy= 4*left_xy[LEFT(i)] + 1;
AV_COPY32(mv_cache[cache_idx ], mv[b_xy + b_stride*left_block[0+i*2]]);
AV_COPY32(mv_cache[cache_idx+8], mv[b_xy + b_stride*left_block[1+i*2]]);
ref_cache[cache_idx ]= ref[b8_xy + (left_block[0+i*2]&~1)];
ref_cache[cache_idx+8]= ref[b8_xy + (left_block[1+i*2]&~1)];
}else{
AV_ZERO32(mv_cache[cache_idx ]);
AV_ZERO32(mv_cache[cache_idx+8]);
ref_cache[cache_idx ]=
ref_cache[cache_idx+8]= (left_type[LEFT(i)]) ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
}
}else{
if(USES_LIST(left_type[LTOP], list)){
const int b_xy= h->mb2b_xy[left_xy[LTOP]] + 3;
const int b8_xy= 4*left_xy[LTOP] + 1;
AV_COPY32(mv_cache[-1], mv[b_xy + b_stride*left_block[0]]);
ref_cache[-1]= ref[b8_xy + (left_block[0]&~1)];
}else{
AV_ZERO32(mv_cache[-1]);
ref_cache[-1]= left_type[LTOP] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
}
if(USES_LIST(topright_type, list)){
const int b_xy= h->mb2b_xy[topright_xy] + 3*b_stride;
AV_COPY32(mv_cache[4 - 1*8], mv[b_xy]);
ref_cache[4 - 1*8]= ref[4*topright_xy + 2];
}else{
AV_ZERO32(mv_cache[4 - 1*8]);
ref_cache[4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
if(ref_cache[4 - 1*8] < 0){
if(USES_LIST(topleft_type, list)){
const int b_xy = h->mb2b_xy[topleft_xy] + 3 + b_stride + (h->topleft_partition & 2*b_stride);
const int b8_xy= 4*topleft_xy + 1 + (h->topleft_partition & 2);
AV_COPY32(mv_cache[-1 - 1*8], mv[b_xy]);
ref_cache[-1 - 1*8]= ref[b8_xy];
}else{
AV_ZERO32(mv_cache[-1 - 1*8]);
ref_cache[-1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
}
if((mb_type&(MB_TYPE_SKIP|MB_TYPE_DIRECT2)) && !FRAME_MBAFF)
continue;
if(!(mb_type&(MB_TYPE_SKIP|MB_TYPE_DIRECT2))){
uint8_t (*mvd_cache)[2] = &h->mvd_cache[list][scan8[0]];
uint8_t (*mvd)[2] = h->mvd_table[list];
ref_cache[2+8*0] =
ref_cache[2+8*2] = PART_NOT_AVAILABLE;
AV_ZERO32(mv_cache[2+8*0]);
AV_ZERO32(mv_cache[2+8*2]);
if( CABAC ) {
if(USES_LIST(top_type, list)){
const int b_xy= h->mb2br_xy[top_xy];
AV_COPY64(mvd_cache[0 - 1*8], mvd[b_xy + 0]);
}else{
AV_ZERO64(mvd_cache[0 - 1*8]);
}
if(USES_LIST(left_type[LTOP], list)){
const int b_xy= h->mb2br_xy[left_xy[LTOP]] + 6;
AV_COPY16(mvd_cache[-1 + 0*8], mvd[b_xy - left_block[0]]);
AV_COPY16(mvd_cache[-1 + 1*8], mvd[b_xy - left_block[1]]);
}else{
AV_ZERO16(mvd_cache[-1 + 0*8]);
AV_ZERO16(mvd_cache[-1 + 1*8]);
}
if(USES_LIST(left_type[LBOT], list)){
const int b_xy= h->mb2br_xy[left_xy[LBOT]] + 6;
AV_COPY16(mvd_cache[-1 + 2*8], mvd[b_xy - left_block[2]]);
AV_COPY16(mvd_cache[-1 + 3*8], mvd[b_xy - left_block[3]]);
}else{
AV_ZERO16(mvd_cache[-1 + 2*8]);
AV_ZERO16(mvd_cache[-1 + 3*8]);
}
AV_ZERO16(mvd_cache[2+8*0]);
AV_ZERO16(mvd_cache[2+8*2]);
if(h->slice_type_nos == AV_PICTURE_TYPE_B){
uint8_t *direct_cache = &h->direct_cache[scan8[0]];
uint8_t *direct_table = h->direct_table;
fill_rectangle(direct_cache, 4, 4, 8, MB_TYPE_16x16>>1, 1);
if(IS_DIRECT(top_type)){
AV_WN32A(&direct_cache[-1*8], 0x01010101u*(MB_TYPE_DIRECT2>>1));
}else if(IS_8X8(top_type)){
int b8_xy = 4*top_xy;
direct_cache[0 - 1*8]= direct_table[b8_xy + 2];
direct_cache[2 - 1*8]= direct_table[b8_xy + 3];
}else{
AV_WN32A(&direct_cache[-1*8], 0x01010101*(MB_TYPE_16x16>>1));
}
if(IS_DIRECT(left_type[LTOP]))
direct_cache[-1 + 0*8]= MB_TYPE_DIRECT2>>1;
else if(IS_8X8(left_type[LTOP]))
direct_cache[-1 + 0*8]= direct_table[4*left_xy[LTOP] + 1 + (left_block[0]&~1)];
else
direct_cache[-1 + 0*8]= MB_TYPE_16x16>>1;
if(IS_DIRECT(left_type[LBOT]))
direct_cache[-1 + 2*8]= MB_TYPE_DIRECT2>>1;
else if(IS_8X8(left_type[LBOT]))
direct_cache[-1 + 2*8]= direct_table[4*left_xy[LBOT] + 1 + (left_block[2]&~1)];
else
direct_cache[-1 + 2*8]= MB_TYPE_16x16>>1;
}
}
}
if(FRAME_MBAFF){
#define MAP_MVS\
MAP_F2F(scan8[0] - 1 - 1*8, topleft_type)\
MAP_F2F(scan8[0] + 0 - 1*8, top_type)\
MAP_F2F(scan8[0] + 1 - 1*8, top_type)\
MAP_F2F(scan8[0] + 2 - 1*8, top_type)\
MAP_F2F(scan8[0] + 3 - 1*8, top_type)\
MAP_F2F(scan8[0] + 4 - 1*8, topright_type)\
MAP_F2F(scan8[0] - 1 + 0*8, left_type[LTOP])\
MAP_F2F(scan8[0] - 1 + 1*8, left_type[LTOP])\
MAP_F2F(scan8[0] - 1 + 2*8, left_type[LBOT])\
MAP_F2F(scan8[0] - 1 + 3*8, left_type[LBOT])
if(MB_FIELD){
#define MAP_F2F(idx, mb_type)\
if(!IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
h->ref_cache[list][idx] <<= 1;\
h->mv_cache[list][idx][1] /= 2;\
h->mvd_cache[list][idx][1] >>=1;\
}
MAP_MVS
#undef MAP_F2F
}else{
#define MAP_F2F(idx, mb_type)\
if(IS_INTERLACED(mb_type) && h->ref_cache[list][idx] >= 0){\
h->ref_cache[list][idx] >>= 1;\
h->mv_cache[list][idx][1] <<= 1;\
h->mvd_cache[list][idx][1] <<= 1;\
}
MAP_MVS
#undef MAP_F2F
}
}
}
}
h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[LTOP]);
}
/**
* decodes a P_SKIP or B_SKIP macroblock
*/
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment