-static inline void pred_pskip_motion(H264Context * const h, int * const mx, int * const my);
-
-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[2];
- 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[1] = left_xy[0] = mb_xy-1;
- h->left_block = left_block_options[0];
- if(FRAME_MBAFF){
- const int left_mb_field_flag = IS_INTERLACED(s->current_picture.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[1] = left_xy[0] = mb_xy - s->mb_stride - 1;
- if (curr_mb_field_flag) {
- left_xy[1] += 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.mb_type[top_xy - 1]>>7)&1)-1);
- topright_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy + 1]>>7)&1)-1);
- top_xy += s->mb_stride & (((s->current_picture.mb_type[top_xy ]>>7)&1)-1);
- }
- if (left_mb_field_flag != curr_mb_field_flag) {
- if (curr_mb_field_flag) {
- left_xy[1] += 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[0] = left_xy[0];
- h->left_mb_xy[1] = left_xy[1];
- //FIXME do we need all in the context?
-
- h->topleft_type = s->current_picture.mb_type[topleft_xy] ;
- h->top_type = s->current_picture.mb_type[top_xy] ;
- h->topright_type= s->current_picture.mb_type[topright_xy];
- h->left_type[0] = s->current_picture.mb_type[left_xy[0]] ;
- h->left_type[1] = s->current_picture.mb_type[left_xy[1]] ;
-
- 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[0] ] != h->slice_num) h->left_type[0] = h->left_type[1] = 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[0] ] != h->slice_num) h->left_type[0] = h->left_type[1] = 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[2];
- int topleft_type, top_type, topright_type, left_type[2];
- const uint8_t * left_block= h->left_block;
- int i;
-
- topleft_xy = h->topleft_mb_xy ;
- top_xy = h->top_mb_xy ;
- topright_xy = h->topright_mb_xy;
- left_xy[0] = h->left_mb_xy[0] ;
- left_xy[1] = h->left_mb_xy[1] ;
- topleft_type = h->topleft_type ;
- top_type = h->top_type ;
- topright_type= h->topright_type ;
- left_type[0] = h->left_type[0] ;
- left_type[1] = h->left_type[1] ;
-
- 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[0])){
- if(IS_INTERLACED(mb_type)){
- if(!(left_type[0] & type_mask)){
- h->topleft_samples_available&= 0xDFFF;
- h->left_samples_available&= 0x5FFF;
- }
- if(!(left_type[1] & type_mask)){
- h->topleft_samples_available&= 0xFF5F;
- h->left_samples_available&= 0xFF5F;
- }
- }else{
- int left_typei = s->current_picture.mb_type[left_xy[0] + s->mb_stride];
-
- assert(left_xy[0] == left_xy[1]);
- if(!((left_typei & type_mask) && (left_type[0] & type_mask))){
- h->topleft_samples_available&= 0xDF5F;
- h->left_samples_available&= 0x5F5F;
- }
- }
- }else{
- if(!(left_type[0] & 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[i])){
- int8_t *mode= h->intra4x4_pred_mode + h->mb2br_xy[left_xy[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[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)
- if(top_type){
- AV_COPY32(&h->non_zero_count_cache[4+8* 0], &h->non_zero_count[top_xy][4*3]);
- if(CHROMA444){
- AV_COPY32(&h->non_zero_count_cache[4+8* 5], &h->non_zero_count[top_xy][4* 7]);
- AV_COPY32(&h->non_zero_count_cache[4+8*10], &h->non_zero_count[top_xy][4*11]);
- }else{
- AV_COPY32(&h->non_zero_count_cache[4+8* 5], &h->non_zero_count[top_xy][4* 5]);
- AV_COPY32(&h->non_zero_count_cache[4+8*10], &h->non_zero_count[top_xy][4* 9]);
- }
- }else{
- uint32_t top_empty = CABAC && !IS_INTRA(mb_type) ? 0 : 0x40404040;
- AV_WN32A(&h->non_zero_count_cache[4+8* 0], top_empty);
- AV_WN32A(&h->non_zero_count_cache[4+8* 5], top_empty);
- AV_WN32A(&h->non_zero_count_cache[4+8*10], top_empty);
- }
-
- for (i=0; i<2; i++) {
- if(left_type[i]){
- h->non_zero_count_cache[3+8* 1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]];
- h->non_zero_count_cache[3+8* 2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]];
- if(CHROMA444){
- h->non_zero_count_cache[3+8* 6 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]+4*4];
- h->non_zero_count_cache[3+8* 7 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]+4*4];
- h->non_zero_count_cache[3+8*11 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+0+2*i]+8*4];
- h->non_zero_count_cache[3+8*12 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[8+1+2*i]+8*4];
- }else{
- h->non_zero_count_cache[3+8* 6 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+4+2*i]];
- h->non_zero_count_cache[3+8*11 + 8*i]= h->non_zero_count[left_xy[i]][left_block[8+5+2*i]];
- }
- }else{
- h->non_zero_count_cache[3+8* 1 + 2*8*i]=
- h->non_zero_count_cache[3+8* 2 + 2*8*i]=
- h->non_zero_count_cache[3+8* 6 + 2*8*i]=
- h->non_zero_count_cache[3+8* 7 + 2*8*i]=
- h->non_zero_count_cache[3+8*11 + 2*8*i]=
- h->non_zero_count_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[0]) {
- h->left_cbp = (h->cbp_table[left_xy[0]] & 0x7F0)
- | ((h->cbp_table[left_xy[0]]>>(left_block[0]&(~1)))&2)
- | (((h->cbp_table[left_xy[1]]>>(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;
- for(list=0; list<h->list_count; list++){
- if(!USES_LIST(mb_type, list)){
- /*if(!h->mv_cache_clean[list]){
- memset(h->mv_cache [list], 0, 8*5*2*sizeof(int16_t)); //FIXME clean only input? clean at all?
- memset(h->ref_cache[list], PART_NOT_AVAILABLE, 8*5*sizeof(int8_t));
- h->mv_cache_clean[list]= 1;
- }*/
- continue;
- }
- assert(!(IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred));
-
- h->mv_cache_clean[list]= 0;
-
- if(USES_LIST(top_type, list)){
- const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
- AV_COPY128(h->mv_cache[list][scan8[0] + 0 - 1*8], s->current_picture.motion_val[list][b_xy + 0]);
- h->ref_cache[list][scan8[0] + 0 - 1*8]=
- h->ref_cache[list][scan8[0] + 1 - 1*8]= s->current_picture.ref_index[list][4*top_xy + 2];
- h->ref_cache[list][scan8[0] + 2 - 1*8]=
- h->ref_cache[list][scan8[0] + 3 - 1*8]= s->current_picture.ref_index[list][4*top_xy + 3];
- }else{
- AV_ZERO128(h->mv_cache[list][scan8[0] + 0 - 1*8]);
- AV_WN32A(&h->ref_cache[list][scan8[0] + 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 = scan8[0] - 1 + i*2*8;
- if(USES_LIST(left_type[i], list)){
- const int b_xy= h->mb2b_xy[left_xy[i]] + 3;
- const int b8_xy= 4*left_xy[i] + 1;
- AV_COPY32(h->mv_cache[list][cache_idx ], s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[0+i*2]]);
- AV_COPY32(h->mv_cache[list][cache_idx+8], s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[1+i*2]]);
- h->ref_cache[list][cache_idx ]= s->current_picture.ref_index[list][b8_xy + (left_block[0+i*2]&~1)];
- h->ref_cache[list][cache_idx+8]= s->current_picture.ref_index[list][b8_xy + (left_block[1+i*2]&~1)];
- }else{
- AV_ZERO32(h->mv_cache [list][cache_idx ]);
- AV_ZERO32(h->mv_cache [list][cache_idx+8]);
- h->ref_cache[list][cache_idx ]=
- h->ref_cache[list][cache_idx+8]= (left_type[i]) ? LIST_NOT_USED : PART_NOT_AVAILABLE;
- }
- }
- }else{
- if(USES_LIST(left_type[0], list)){
- const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
- const int b8_xy= 4*left_xy[0] + 1;
- AV_COPY32(h->mv_cache[list][scan8[0] - 1], s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[0]]);
- h->ref_cache[list][scan8[0] - 1]= s->current_picture.ref_index[list][b8_xy + (left_block[0]&~1)];
- }else{
- AV_ZERO32(h->mv_cache [list][scan8[0] - 1]);
- h->ref_cache[list][scan8[0] - 1]= left_type[0] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
- }
- }
-
- if(USES_LIST(topright_type, list)){
- const int b_xy= h->mb2b_xy[topright_xy] + 3*h->b_stride;
- AV_COPY32(h->mv_cache[list][scan8[0] + 4 - 1*8], s->current_picture.motion_val[list][b_xy]);
- h->ref_cache[list][scan8[0] + 4 - 1*8]= s->current_picture.ref_index[list][4*topright_xy + 2];
- }else{
- AV_ZERO32(h->mv_cache [list][scan8[0] + 4 - 1*8]);
- h->ref_cache[list][scan8[0] + 4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
- }
- if(h->ref_cache[list][scan8[0] + 4 - 1*8] < 0){
- if(USES_LIST(topleft_type, list)){
- const int b_xy = h->mb2b_xy [topleft_xy] + 3 + h->b_stride + (h->topleft_partition & 2*h->b_stride);
- const int b8_xy= 4*topleft_xy + 1 + (h->topleft_partition & 2);
- AV_COPY32(h->mv_cache[list][scan8[0] - 1 - 1*8], s->current_picture.motion_val[list][b_xy]);
- h->ref_cache[list][scan8[0] - 1 - 1*8]= s->current_picture.ref_index[list][b8_xy];
- }else{
- AV_ZERO32(h->mv_cache[list][scan8[0] - 1 - 1*8]);
- h->ref_cache[list][scan8[0] - 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))) {
- h->ref_cache[list][scan8[4 ]] =
- h->ref_cache[list][scan8[12]] = PART_NOT_AVAILABLE;
- AV_ZERO32(h->mv_cache [list][scan8[4 ]]);
- AV_ZERO32(h->mv_cache [list][scan8[12]]);
-
- if( CABAC ) {
- /* XXX beurk, Load mvd */
- if(USES_LIST(top_type, list)){
- const int b_xy= h->mb2br_xy[top_xy];
- AV_COPY64(h->mvd_cache[list][scan8[0] + 0 - 1*8], h->mvd_table[list][b_xy + 0]);
- }else{
- AV_ZERO64(h->mvd_cache[list][scan8[0] + 0 - 1*8]);
- }
- if(USES_LIST(left_type[0], list)){
- const int b_xy= h->mb2br_xy[left_xy[0]] + 6;
- AV_COPY16(h->mvd_cache[list][scan8[0] - 1 + 0*8], h->mvd_table[list][b_xy - left_block[0]]);
- AV_COPY16(h->mvd_cache[list][scan8[0] - 1 + 1*8], h->mvd_table[list][b_xy - left_block[1]]);
- }else{
- AV_ZERO16(h->mvd_cache [list][scan8[0] - 1 + 0*8]);
- AV_ZERO16(h->mvd_cache [list][scan8[0] - 1 + 1*8]);
- }
- if(USES_LIST(left_type[1], list)){
- const int b_xy= h->mb2br_xy[left_xy[1]] + 6;
- AV_COPY16(h->mvd_cache[list][scan8[0] - 1 + 2*8], h->mvd_table[list][b_xy - left_block[2]]);
- AV_COPY16(h->mvd_cache[list][scan8[0] - 1 + 3*8], h->mvd_table[list][b_xy - left_block[3]]);
- }else{
- AV_ZERO16(h->mvd_cache [list][scan8[0] - 1 + 2*8]);
- AV_ZERO16(h->mvd_cache [list][scan8[0] - 1 + 3*8]);
- }
- AV_ZERO16(h->mvd_cache [list][scan8[4 ]]);
- AV_ZERO16(h->mvd_cache [list][scan8[12]]);
- if(h->slice_type_nos == AV_PICTURE_TYPE_B){
- fill_rectangle(&h->direct_cache[scan8[0]], 4, 4, 8, MB_TYPE_16x16>>1, 1);
-
- if(IS_DIRECT(top_type)){
- AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101u*(MB_TYPE_DIRECT2>>1));
- }else if(IS_8X8(top_type)){
- int b8_xy = 4*top_xy;
- h->direct_cache[scan8[0] + 0 - 1*8]= h->direct_table[b8_xy + 2];
- h->direct_cache[scan8[0] + 2 - 1*8]= h->direct_table[b8_xy + 3];
- }else{
- AV_WN32A(&h->direct_cache[scan8[0] - 1*8], 0x01010101*(MB_TYPE_16x16>>1));
- }
-
- if(IS_DIRECT(left_type[0]))
- h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_DIRECT2>>1;
- else if(IS_8X8(left_type[0]))
- h->direct_cache[scan8[0] - 1 + 0*8]= h->direct_table[4*left_xy[0] + 1 + (left_block[0]&~1)];
- else
- h->direct_cache[scan8[0] - 1 + 0*8]= MB_TYPE_16x16>>1;
-
- if(IS_DIRECT(left_type[1]))
- h->direct_cache[scan8[0] - 1 + 2*8]= MB_TYPE_DIRECT2>>1;
- else if(IS_8X8(left_type[1]))
- h->direct_cache[scan8[0] - 1 + 2*8]= h->direct_table[4*left_xy[1] + 1 + (left_block[2]&~1)];
- else
- h->direct_cache[scan8[0] - 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[0])\
- MAP_F2F(scan8[0] - 1 + 1*8, left_type[0])\
- MAP_F2F(scan8[0] - 1 + 2*8, left_type[1])\
- MAP_F2F(scan8[0] - 1 + 3*8, left_type[1])
- 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[0]);
-}
-