#define DELAYED_PIC_REF 4
static VLC coeff_token_vlc[4];
+static VLC_TYPE coeff_token_vlc_tables[520+332+280+256][2];
+static const int coeff_token_vlc_tables_size[4]={520,332,280,256};
+
static VLC chroma_dc_coeff_token_vlc;
+static VLC_TYPE chroma_dc_coeff_token_vlc_table[256][2];
+static const int chroma_dc_coeff_token_vlc_table_size = 256;
static VLC total_zeros_vlc[15];
+static VLC_TYPE total_zeros_vlc_tables[15][512][2];
+static const int total_zeros_vlc_tables_size = 512;
+
static VLC chroma_dc_total_zeros_vlc[3];
+static VLC_TYPE chroma_dc_total_zeros_vlc_tables[3][8][2];
+static const int chroma_dc_total_zeros_vlc_tables_size = 8;
static VLC run_vlc[6];
+static VLC_TYPE run_vlc_tables[6][8][2];
+static const int run_vlc_tables_size = 8;
+
static VLC run7_vlc;
+static VLC_TYPE run7_vlc_table[96][2];
+static const int run7_vlc_table_size = 96;
static void svq3_luma_dc_dequant_idct_c(DCTELEM *block, int qp);
static void svq3_add_idct_c(uint8_t *dst, DCTELEM *block, int stride, int qp, int dc);
static void filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize);
static void filter_mb_fast( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr, unsigned int linesize, unsigned int uvlinesize);
-static void remove_long_at_index(H264Context *h, int i);
+static Picture * remove_long(H264Context *h, int i, int ref_mask);
static av_always_inline uint32_t pack16to32(int a, int b){
#ifdef WORDS_BIGENDIAN
#endif
}
-const uint8_t ff_rem6[52]={
+static const uint8_t rem6[52]={
0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3, 4, 5, 0, 1, 2, 3,
};
-const uint8_t ff_div6[52]={
+static const uint8_t div6[52]={
0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 5, 5, 5, 5, 5, 5, 6, 6, 6, 6, 6, 6, 7, 7, 7, 7, 7, 7, 8, 8, 8, 8,
};
+static const int left_block_options[4][8]={
+ {0,1,2,3,7,10,8,11},
+ {2,2,3,3,8,11,8,11},
+ {0,0,1,1,7,10,7,10},
+ {0,2,0,2,7,10,7,10}
+};
static void fill_caches(H264Context *h, int mb_type, int for_deblock){
MpegEncContext * const s = &h->s;
const int mb_xy= h->mb_xy;
int topleft_xy, top_xy, topright_xy, left_xy[2];
int topleft_type, top_type, topright_type, left_type[2];
- int left_block[8];
+ int * left_block;
int topleft_partition= -1;
int i;
topleft_xy = top_xy - 1;
topright_xy= top_xy + 1;
left_xy[1] = left_xy[0] = mb_xy-1;
- left_block[0]= 0;
- left_block[1]= 1;
- left_block[2]= 2;
- left_block[3]= 3;
- left_block[4]= 7;
- left_block[5]= 10;
- left_block[6]= 8;
- left_block[7]= 11;
+ left_block = left_block_options[0];
if(FRAME_MBAFF){
const int pair_xy = s->mb_x + (s->mb_y & ~1)*s->mb_stride;
const int top_pair_xy = pair_xy - s->mb_stride;
left_xy[1] = left_xy[0] = pair_xy - 1;
if (curr_mb_frame_flag) {
if (bottom) {
- left_block[0]= 2;
- left_block[1]= 2;
- left_block[2]= 3;
- left_block[3]= 3;
- left_block[4]= 8;
- left_block[5]= 11;
- left_block[6]= 8;
- left_block[7]= 11;
+ left_block = left_block_options[1];
} else {
- left_block[0]= 0;
- left_block[1]= 0;
- left_block[2]= 1;
- left_block[3]= 1;
- left_block[4]= 7;
- left_block[5]= 10;
- left_block[6]= 7;
- left_block[7]= 10;
+ left_block= left_block_options[2];
}
} else {
left_xy[1] += s->mb_stride;
- //left_block[0]= 0;
- left_block[1]= 2;
- left_block[2]= 0;
- left_block[3]= 2;
- //left_block[4]= 7;
- left_block[5]= 10;
- left_block[6]= 7;
- left_block[7]= 10;
+ left_block = left_block_options[3];
}
}
}
if(for_deblock){
topleft_type = 0;
topright_type = 0;
- top_type = h->slice_table[top_xy ] < 255 ? s->current_picture.mb_type[top_xy] : 0;
- left_type[0] = h->slice_table[left_xy[0] ] < 255 ? s->current_picture.mb_type[left_xy[0]] : 0;
- left_type[1] = h->slice_table[left_xy[1] ] < 255 ? s->current_picture.mb_type[left_xy[1]] : 0;
+ top_type = h->slice_table[top_xy ] < 0xFFFF ? s->current_picture.mb_type[top_xy] : 0;
+ left_type[0] = h->slice_table[left_xy[0] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[0]] : 0;
+ left_type[1] = h->slice_table[left_xy[1] ] < 0xFFFF ? s->current_picture.mb_type[left_xy[1]] : 0;
- if(FRAME_MBAFF && !IS_INTRA(mb_type)){
+ if(MB_MBAFF && !IS_INTRA(mb_type)){
int list;
- int v = *(uint16_t*)&h->non_zero_count[mb_xy][14];
- for(i=0; i<16; i++)
- h->non_zero_count_cache[scan8[i]] = (v>>i)&1;
for(list=0; list<h->list_count; list++){
+ //These values where changed for ease of performing MC, we need to change them back
+ //FIXME maybe we can make MC and loop filter use the same values or prevent
+ //the MC code from changing ref_cache and rather use a temporary array.
if(USES_LIST(mb_type,list)){
- uint32_t *src = (uint32_t*)s->current_picture.motion_val[list][h->mb2b_xy[mb_xy]];
- uint32_t *dst = (uint32_t*)h->mv_cache[list][scan8[0]];
int8_t *ref = &s->current_picture.ref_index[list][h->mb2b8_xy[mb_xy]];
- for(i=0; i<4; i++, dst+=8, src+=h->b_stride){
- dst[0] = src[0];
- dst[1] = src[1];
- dst[2] = src[2];
- dst[3] = src[3];
- }
*(uint32_t*)&h->ref_cache[list][scan8[ 0]] =
- *(uint32_t*)&h->ref_cache[list][scan8[ 2]] = pack16to32(ref[0],ref[1])*0x0101;
+ *(uint32_t*)&h->ref_cache[list][scan8[ 2]] = (pack16to32(ref[0],ref[1])&0x00FF00FF)*0x0101;
ref += h->b8_stride;
*(uint32_t*)&h->ref_cache[list][scan8[ 8]] =
- *(uint32_t*)&h->ref_cache[list][scan8[10]] = pack16to32(ref[0],ref[1])*0x0101;
- }else{
- fill_rectangle(&h-> mv_cache[list][scan8[ 0]], 4, 4, 8, 0, 4);
- fill_rectangle(&h->ref_cache[list][scan8[ 0]], 4, 4, 8, (uint8_t)LIST_NOT_USED, 1);
+ *(uint32_t*)&h->ref_cache[list][scan8[10]] = (pack16to32(ref[0],ref[1])&0x00FF00FF)*0x0101;
}
}
}
topright_type= h->slice_table[topright_xy] == h->slice_num ? s->current_picture.mb_type[topright_xy]: 0;
left_type[0] = h->slice_table[left_xy[0] ] == h->slice_num ? s->current_picture.mb_type[left_xy[0]] : 0;
left_type[1] = h->slice_table[left_xy[1] ] == h->slice_num ? s->current_picture.mb_type[left_xy[1]] : 0;
- }
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(!IS_INTRA(top_type) && (top_type==0 || h->pps.constrained_intra_pred)){
+ if(!(top_type & type_mask)){
h->topleft_samples_available= 0xB3FF;
h->top_samples_available= 0x33FF;
h->topright_samples_available= 0x26EA;
}
- for(i=0; i<2; i++){
- if(!IS_INTRA(left_type[i]) && (left_type[i]==0 || h->pps.constrained_intra_pred)){
+ 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 = h->slice_table[left_xy[0] + s->mb_stride ] == h->slice_num
+ ? s->current_picture.mb_type[left_xy[0] + s->mb_stride] : 0;
+ 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(!IS_INTRA(topleft_type) && (topleft_type==0 || h->pps.constrained_intra_pred))
+ if(!(topleft_type & type_mask))
h->topleft_samples_available&= 0x7FFF;
- if(!IS_INTRA(topright_type) && (topright_type==0 || h->pps.constrained_intra_pred))
+ if(!(topright_type & type_mask))
h->topright_samples_available&= 0xFBFF;
if(IS_INTRA4x4(mb_type)){
h->intra4x4_pred_mode_cache[7+8*0]= h->intra4x4_pred_mode[top_xy][3];
}else{
int pred;
- if(!top_type || (IS_INTER(top_type) && h->pps.constrained_intra_pred))
+ if(!(top_type & type_mask))
pred= -1;
else{
pred= 2;
h->intra4x4_pred_mode_cache[3+8*2 + 2*8*i]= h->intra4x4_pred_mode[left_xy[i]][left_block[1+2*i]];
}else{
int pred;
- if(!left_type[i] || (IS_INTER(left_type[i]) && h->pps.constrained_intra_pred))
+ if(!(left_type[i] & type_mask))
pred= -1;
else{
pred= 2;
}
}
}
+ }
/*
}
}
- if((for_deblock || (IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred)) && !FRAME_MBAFF)
+ if(for_deblock || ((IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred) && !FRAME_MBAFF))
continue;
if(USES_LIST(topleft_type, list)){
}
}
- if(!(h->left_samples_available&0x8000)){
+ if((h->left_samples_available&0x8888)!=0x8888){
+ static const int mask[4]={0x8000,0x2000,0x80,0x20};
for(i=0; i<4; i++){
+ if(!(h->left_samples_available&mask[i])){
int status= left[ h->intra4x4_pred_mode_cache[scan8[0] + 8*i] ];
if(status<0){
av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra4x4 mode %d at %d %d\n", status, s->mb_x, s->mb_y);
} else if(status){
h->intra4x4_pred_mode_cache[scan8[0] + 8*i]= status;
}
+ }
}
}
}
}
- if(!(h->left_samples_available&0x8000)){
+ if((h->left_samples_available&0x8080) != 0x8080){
mode= left[ mode ];
+ if(h->left_samples_available&0x8080){ //mad cow disease mode, aka MBAFF + constrained_intra_pred
+ mode= ALZHEIMER_DC_L0T_PRED8x8 + (!(h->left_samples_available&0x8000)) + 2*(mode == DC_128_PRED8x8);
+ }
if(mode<0){
av_log(h->s.avctx, AV_LOG_ERROR, "left block unavailable for requested intra mode at %d %d\n", s->mb_x, s->mb_y);
return -1;
h->non_zero_count[mb_xy][12]=h->non_zero_count_cache[1+8*5];
h->non_zero_count[mb_xy][11]=h->non_zero_count_cache[2+8*5];
h->non_zero_count[mb_xy][10]=h->non_zero_count_cache[2+8*4];
-
- if(FRAME_MBAFF){
- // store all luma nnzs, for deblocking
- int v = 0, i;
- for(i=0; i<16; i++)
- v += (!!h->non_zero_count_cache[scan8[i]]) << i;
- *(uint16_t*)&h->non_zero_count[mb_xy][14] = v;
- }
}
/**
return;
}
+static int get_scale_factor(H264Context * const h, int poc, int poc1, int i){
+ int poc0 = h->ref_list[0][i].poc;
+ int td = av_clip(poc1 - poc0, -128, 127);
+ if(td == 0 || h->ref_list[0][i].long_ref){
+ return 256;
+ }else{
+ int tb = av_clip(poc - poc0, -128, 127);
+ int tx = (16384 + (FFABS(td) >> 1)) / td;
+ return av_clip((tb*tx + 32) >> 6, -1024, 1023);
+ }
+}
+
static inline void direct_dist_scale_factor(H264Context * const h){
- const int poc = h->s.current_picture_ptr->poc;
+ MpegEncContext * const s = &h->s;
+ const int poc = h->s.current_picture_ptr->field_poc[ s->picture_structure == PICT_BOTTOM_FIELD ];
const int poc1 = h->ref_list[1][0].poc;
- int i;
+ int i, field;
+ for(field=0; field<2; field++){
+ const int poc = h->s.current_picture_ptr->field_poc[field];
+ const int poc1 = h->ref_list[1][0].field_poc[field];
+ for(i=0; i < 2*h->ref_count[0]; i++)
+ h->dist_scale_factor_field[field][i^field] = get_scale_factor(h, poc, poc1, i+16);
+ }
+
for(i=0; i<h->ref_count[0]; i++){
- int poc0 = h->ref_list[0][i].poc;
- int td = av_clip(poc1 - poc0, -128, 127);
- if(td == 0 /* FIXME || pic0 is a long-term ref */){
- h->dist_scale_factor[i] = 256;
- }else{
- int tb = av_clip(poc - poc0, -128, 127);
- int tx = (16384 + (FFABS(td) >> 1)) / td;
- h->dist_scale_factor[i] = av_clip((tb*tx + 32) >> 6, -1024, 1023);
- }
+ h->dist_scale_factor[i] = get_scale_factor(h, poc, poc1, i);
}
- if(FRAME_MBAFF){
- for(i=0; i<h->ref_count[0]; i++){
- h->dist_scale_factor_field[2*i] =
- h->dist_scale_factor_field[2*i+1] = h->dist_scale_factor[i];
+}
+
+static void fill_colmap(H264Context *h, int map[2][16+32], int list, int field, int colfield, int mbafi){
+ MpegEncContext * const s = &h->s;
+ Picture * const ref1 = &h->ref_list[1][0];
+ int j, old_ref, rfield;
+ int start= mbafi ? 16 : 0;
+ int end = mbafi ? 16+2*h->ref_count[list] : h->ref_count[list];
+ int interl= mbafi || s->picture_structure != PICT_FRAME;
+
+ /* bogus; fills in for missing frames */
+ memset(map[list], 0, sizeof(map[list]));
+
+ for(rfield=0; rfield<2; rfield++){
+ for(old_ref=0; old_ref<ref1->ref_count[colfield][list]; old_ref++){
+ int poc = ref1->ref_poc[colfield][list][old_ref];
+
+ if (!interl)
+ poc |= 3;
+ else if( interl && (poc&3) == 3) //FIXME store all MBAFF references so this isnt needed
+ poc= (poc&~3) + rfield + 1;
+
+ for(j=start; j<end; j++){
+ if(4*h->ref_list[list][j].frame_num + (h->ref_list[list][j].reference&3) == poc){
+ int cur_ref= mbafi ? (j-16)^field : j;
+ map[list][2*old_ref + (rfield^field) + 16] = cur_ref;
+ if(rfield == field)
+ map[list][old_ref] = cur_ref;
+ break;
+ }
+ }
}
}
}
+
static inline void direct_ref_list_init(H264Context * const h){
MpegEncContext * const s = &h->s;
Picture * const ref1 = &h->ref_list[1][0];
Picture * const cur = s->current_picture_ptr;
- int list, i, j;
- if(cur->pict_type == FF_I_TYPE)
- cur->ref_count[0] = 0;
- if(cur->pict_type != FF_B_TYPE)
- cur->ref_count[1] = 0;
+ int list, j, field;
+ int sidx= (s->picture_structure&1)^1;
+ int ref1sidx= (ref1->reference&1)^1;
+
for(list=0; list<2; list++){
- cur->ref_count[list] = h->ref_count[list];
+ cur->ref_count[sidx][list] = h->ref_count[list];
for(j=0; j<h->ref_count[list]; j++)
- cur->ref_poc[list][j] = h->ref_list[list][j].poc;
+ cur->ref_poc[sidx][list][j] = 4*h->ref_list[list][j].frame_num + (h->ref_list[list][j].reference&3);
}
+
+ if(s->picture_structure == PICT_FRAME){
+ memcpy(cur->ref_count[1], cur->ref_count[0], sizeof(cur->ref_count[0]));
+ memcpy(cur->ref_poc [1], cur->ref_poc [0], sizeof(cur->ref_poc [0]));
+ }
+
+ cur->mbaff= FRAME_MBAFF;
+
if(cur->pict_type != FF_B_TYPE || h->direct_spatial_mv_pred)
return;
+
for(list=0; list<2; list++){
- for(i=0; i<ref1->ref_count[list]; i++){
- const int poc = ref1->ref_poc[list][i];
- h->map_col_to_list0[list][i] = 0; /* bogus; fills in for missing frames */
- for(j=0; j<h->ref_count[list]; j++)
- if(h->ref_list[list][j].poc == poc){
- h->map_col_to_list0[list][i] = j;
- break;
- }
- }
- }
- if(FRAME_MBAFF){
- for(list=0; list<2; list++){
- for(i=0; i<ref1->ref_count[list]; i++){
- j = h->map_col_to_list0[list][i];
- h->map_col_to_list0_field[list][2*i] = 2*j;
- h->map_col_to_list0_field[list][2*i+1] = 2*j+1;
- }
- }
+ fill_colmap(h, h->map_col_to_list0, list, sidx, ref1sidx, 0);
+ for(field=0; field<2; field++)
+ fill_colmap(h, h->map_col_to_list0_field[field], list, field, field, 1);
}
}
static inline void pred_direct_motion(H264Context * const h, int *mb_type){
MpegEncContext * const s = &h->s;
- const int mb_xy = h->mb_xy;
- const int b8_xy = 2*s->mb_x + 2*s->mb_y*h->b8_stride;
- const int b4_xy = 4*s->mb_x + 4*s->mb_y*h->b_stride;
- const int mb_type_col = h->ref_list[1][0].mb_type[mb_xy];
- const int16_t (*l1mv0)[2] = (const int16_t (*)[2]) &h->ref_list[1][0].motion_val[0][b4_xy];
- const int16_t (*l1mv1)[2] = (const int16_t (*)[2]) &h->ref_list[1][0].motion_val[1][b4_xy];
- const int8_t *l1ref0 = &h->ref_list[1][0].ref_index[0][b8_xy];
- const int8_t *l1ref1 = &h->ref_list[1][0].ref_index[1][b8_xy];
+ int b8_stride = h->b8_stride;
+ int b4_stride = h->b_stride;
+ int mb_xy = h->mb_xy;
+ 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;
int i8, i4;
#define MB_TYPE_16x16_OR_INTRA (MB_TYPE_16x16|MB_TYPE_INTRA4x4|MB_TYPE_INTRA16x16|MB_TYPE_INTRA_PCM)
- if(IS_8X8(mb_type_col) && !h->sps.direct_8x8_inference_flag){
- /* FIXME save sub mb types from previous frames (or derive from MVs)
- * so we know exactly what block size to use */
- sub_mb_type = MB_TYPE_8x8|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_4x4 */
- *mb_type = MB_TYPE_8x8|MB_TYPE_L0L1;
- }else if(!is_b8x8 && (mb_type_col & MB_TYPE_16x16_OR_INTRA)){
- sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
- *mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_16x16 */
- }else{
- sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
- *mb_type = MB_TYPE_8x8|MB_TYPE_L0L1;
+
+ 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
+ int cur_poc = s->current_picture_ptr->poc;
+ int *col_poc = h->ref_list[1]->field_poc;
+ int col_parity = FFABS(col_poc[0] - cur_poc) >= FFABS(col_poc[1] - cur_poc);
+ mb_xy= s->mb_x + ((s->mb_y&~1) + col_parity)*s->mb_stride;
+ b8_stride = 0;
+ }else if(!(s->picture_structure & h->ref_list[1][0].reference) && !h->ref_list[1][0].mbaff){// FL -> FL & differ parity
+ int fieldoff= 2*(h->ref_list[1][0].reference)-3;
+ mb_xy += s->mb_stride*fieldoff;
+ }
+ goto single_col;
+ }else{ // AFL/AFR/FR/FL -> AFR/FR
+ if(IS_INTERLACED(*mb_type)){ // AFL /FL -> AFR/FR
+ 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;
+ b4_stride *= 6;
+ //FIXME IS_8X8(mb_type_col[0]) && !h->sps.direct_8x8_inference_flag
+ if( (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)
+ && (mb_type_col[1] & MB_TYPE_16x16_OR_INTRA)
+ && !is_b8x8){
+ sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+ *mb_type |= MB_TYPE_16x8 |MB_TYPE_L0L1|MB_TYPE_DIRECT2; /* B_16x8 */
+ }else{
+ sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+ *mb_type |= MB_TYPE_8x8|MB_TYPE_L0L1;
+ }
+ }else{ // AFR/FR -> AFR/FR
+single_col:
+ mb_type_col[0] =
+ mb_type_col[1] = h->ref_list[1][0].mb_type[mb_xy];
+ if(IS_8X8(mb_type_col[0]) && !h->sps.direct_8x8_inference_flag){
+ /* FIXME save sub mb types from previous frames (or derive from MVs)
+ * so we know exactly what block size to use */
+ sub_mb_type = MB_TYPE_8x8|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_4x4 */
+ *mb_type |= MB_TYPE_8x8|MB_TYPE_L0L1;
+ }else if(!is_b8x8 && (mb_type_col[0] & MB_TYPE_16x16_OR_INTRA)){
+ sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+ *mb_type |= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_16x16 */
+ }else{
+ sub_mb_type = MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2; /* B_SUB_8x8 */
+ *mb_type |= MB_TYPE_8x8|MB_TYPE_L0L1;
+ }
+ }
}
- if(!is_b8x8)
- *mb_type |= MB_TYPE_DIRECT2;
- if(MB_FIELD)
- *mb_type |= MB_TYPE_INTERLACED;
- tprintf(s->avctx, "mb_type = %08x, sub_mb_type = %08x, is_b8x8 = %d, mb_type_col = %08x\n", *mb_type, sub_mb_type, is_b8x8, mb_type_col);
+ 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]];
+ if(!b8_stride){
+ if(s->mb_y&1){
+ l1ref0 += h->b8_stride;
+ l1ref1 += h->b8_stride;
+ l1mv0 += 2*b4_stride;
+ l1mv1 += 2*b4_stride;
+ }
+ }
if(h->direct_spatial_mv_pred){
int ref[2];
int refa = h->ref_cache[list][scan8[0] - 1];
int refb = h->ref_cache[list][scan8[0] - 8];
int refc = h->ref_cache[list][scan8[0] - 8 + 4];
- if(refc == -2)
+ if(refc == PART_NOT_AVAILABLE)
refc = h->ref_cache[list][scan8[0] - 8 - 1];
ref[list] = FFMIN3((unsigned)refa, (unsigned)refb, (unsigned)refc);
if(ref[list] < 0)
sub_mb_type &= ~MB_TYPE_L0;
}
- if(IS_INTERLACED(*mb_type) != IS_INTERLACED(mb_type_col)){
- int pair_xy = s->mb_x + (s->mb_y&~1)*s->mb_stride;
- int mb_types_col[2];
- int b8_stride = h->b8_stride;
- int b4_stride = h->b_stride;
-
- *mb_type = (*mb_type & ~MB_TYPE_16x16) | MB_TYPE_8x8;
-
- if(IS_INTERLACED(*mb_type)){
- mb_types_col[0] = h->ref_list[1][0].mb_type[pair_xy];
- mb_types_col[1] = h->ref_list[1][0].mb_type[pair_xy+s->mb_stride];
- if(s->mb_y&1){
- l1ref0 -= 2*b8_stride;
- l1ref1 -= 2*b8_stride;
- l1mv0 -= 4*b4_stride;
- l1mv1 -= 4*b4_stride;
- }
- b8_stride *= 3;
- b4_stride *= 6;
- }else{
- int cur_poc = s->current_picture_ptr->poc;
- int *col_poc = h->ref_list[1]->field_poc;
- int col_parity = FFABS(col_poc[0] - cur_poc) >= FFABS(col_poc[1] - cur_poc);
- int dy = 2*col_parity - (s->mb_y&1);
- mb_types_col[0] =
- mb_types_col[1] = h->ref_list[1][0].mb_type[pair_xy + col_parity*s->mb_stride];
- l1ref0 += dy*b8_stride;
- l1ref1 += dy*b8_stride;
- l1mv0 += 2*dy*b4_stride;
- l1mv1 += 2*dy*b4_stride;
- b8_stride = 0;
- }
-
+ if(IS_INTERLACED(*mb_type) != IS_INTERLACED(mb_type_col[0])){
for(i8=0; i8<4; i8++){
int x8 = i8&1;
int y8 = i8>>1;
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);
- if(!IS_INTRA(mb_types_col[y8])
+ if(!IS_INTRA(mb_type_col[y8])
&& ( (l1ref0[xy8] == 0 && FFABS(l1mv0[xy4][0]) <= 1 && FFABS(l1mv0[xy4][1]) <= 1)
|| (l1ref0[xy8] < 0 && l1ref1[xy8] == 0 && FFABS(l1mv1[xy4][0]) <= 1 && FFABS(l1mv1[xy4][1]) <= 1))){
if(ref[0] > 0)
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);
- if(!IS_INTRA(mb_type_col)
+ if(!IS_INTRA(mb_type_col[0])
&& ( (l1ref0[0] == 0 && FFABS(l1mv0[0][0]) <= 1 && FFABS(l1mv0[0][1]) <= 1)
|| (l1ref0[0] < 0 && l1ref1[0] == 0 && FFABS(l1mv1[0][0]) <= 1 && FFABS(l1mv1[0][1]) <= 1
&& (h->x264_build>33 || !h->x264_build)))){
fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, (uint8_t)ref[1], 1);
/* col_zero_flag */
- if(!IS_INTRA(mb_type_col) && ( l1ref0[x8 + y8*h->b8_stride] == 0
- || (l1ref0[x8 + y8*h->b8_stride] < 0 && l1ref1[x8 + y8*h->b8_stride] == 0
+ if(!IS_INTRA(mb_type_col[0]) && ( l1ref0[x8 + y8*b8_stride] == 0
+ || (l1ref0[x8 + y8*b8_stride] < 0 && l1ref1[x8 + y8*b8_stride] == 0
&& (h->x264_build>33 || !h->x264_build)))){
- const int16_t (*l1mv)[2]= l1ref0[x8 + y8*h->b8_stride] == 0 ? l1mv0 : l1mv1;
+ const int16_t (*l1mv)[2]= l1ref0[x8 + y8*b8_stride] == 0 ? l1mv0 : l1mv1;
if(IS_SUB_8X8(sub_mb_type)){
- const int16_t *mv_col = l1mv[x8*3 + y8*3*h->b_stride];
+ const int16_t *mv_col = l1mv[x8*3 + y8*3*b4_stride];
if(FFABS(mv_col[0]) <= 1 && FFABS(mv_col[1]) <= 1){
if(ref[0] == 0)
fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
}
}else
for(i4=0; i4<4; i4++){
- const int16_t *mv_col = l1mv[x8*2 + (i4&1) + (y8*2 + (i4>>1))*h->b_stride];
+ 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;
}else{ /* direct temporal mv pred */
const int *map_col_to_list0[2] = {h->map_col_to_list0[0], h->map_col_to_list0[1]};
const int *dist_scale_factor = h->dist_scale_factor;
+ int ref_offset= 0;
- if(FRAME_MBAFF){
- if(IS_INTERLACED(*mb_type)){
- map_col_to_list0[0] = h->map_col_to_list0_field[0];
- map_col_to_list0[1] = h->map_col_to_list0_field[1];
- dist_scale_factor = h->dist_scale_factor_field;
- }
- if(IS_INTERLACED(*mb_type) != IS_INTERLACED(mb_type_col)){
- /* FIXME assumes direct_8x8_inference == 1 */
- const int pair_xy = s->mb_x + (s->mb_y&~1)*s->mb_stride;
- int mb_types_col[2];
- int y_shift;
-
- *mb_type = MB_TYPE_8x8|MB_TYPE_L0L1
- | (is_b8x8 ? 0 : MB_TYPE_DIRECT2)
- | (*mb_type & MB_TYPE_INTERLACED);
- sub_mb_type = MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2|MB_TYPE_16x16;
-
- if(IS_INTERLACED(*mb_type)){
- /* frame to field scaling */
- mb_types_col[0] = h->ref_list[1][0].mb_type[pair_xy];
- mb_types_col[1] = h->ref_list[1][0].mb_type[pair_xy+s->mb_stride];
- if(s->mb_y&1){
- l1ref0 -= 2*h->b8_stride;
- l1ref1 -= 2*h->b8_stride;
- l1mv0 -= 4*h->b_stride;
- l1mv1 -= 4*h->b_stride;
- }
- y_shift = 0;
+ if(FRAME_MBAFF && IS_INTERLACED(*mb_type)){
+ map_col_to_list0[0] = h->map_col_to_list0_field[s->mb_y&1][0];
+ map_col_to_list0[1] = h->map_col_to_list0_field[s->mb_y&1][1];
+ dist_scale_factor =h->dist_scale_factor_field[s->mb_y&1];
+ }
+ if(h->ref_list[1][0].mbaff && IS_INTERLACED(mb_type_col[0]))
+ ref_offset += 16;
- if( (mb_types_col[0] & MB_TYPE_16x16_OR_INTRA)
- && (mb_types_col[1] & MB_TYPE_16x16_OR_INTRA)
- && !is_b8x8)
- *mb_type |= MB_TYPE_16x8;
- else
- *mb_type |= MB_TYPE_8x8;
- }else{
- /* field to frame scaling */
- /* col_mb_y = (mb_y&~1) + (topAbsDiffPOC < bottomAbsDiffPOC ? 0 : 1)
- * but in MBAFF, top and bottom POC are equal */
- int dy = (s->mb_y&1) ? 1 : 2;
- mb_types_col[0] =
- mb_types_col[1] = h->ref_list[1][0].mb_type[pair_xy+s->mb_stride];
- l1ref0 += dy*h->b8_stride;
- l1ref1 += dy*h->b8_stride;
- l1mv0 += 2*dy*h->b_stride;
- l1mv1 += 2*dy*h->b_stride;
- y_shift = 2;
-
- if((mb_types_col[0] & (MB_TYPE_16x16_OR_INTRA|MB_TYPE_16x8))
- && !is_b8x8)
- *mb_type |= MB_TYPE_16x16;
- else
- *mb_type |= MB_TYPE_8x8;
- }
+ if(IS_INTERLACED(*mb_type) != IS_INTERLACED(mb_type_col[0])){
+ /* FIXME assumes direct_8x8_inference == 1 */
+ int y_shift = 2*!IS_INTERLACED(*mb_type);
- for(i8=0; i8<4; i8++){
- const int x8 = i8&1;
- const int y8 = i8>>1;
- int ref0, scale;
- const int16_t (*l1mv)[2]= l1mv0;
-
- if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
- continue;
- h->sub_mb_type[i8] = sub_mb_type;
-
- fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
- if(IS_INTRA(mb_types_col[y8])){
- fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, 0, 1);
- fill_rectangle(&h-> mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
- fill_rectangle(&h-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
- continue;
- }
+ for(i8=0; i8<4; i8++){
+ const int x8 = i8&1;
+ const int y8 = i8>>1;
+ int ref0, scale;
+ const int16_t (*l1mv)[2]= l1mv0;
- ref0 = l1ref0[x8 + (y8*2>>y_shift)*h->b8_stride];
- if(ref0 >= 0)
- ref0 = map_col_to_list0[0][ref0*2>>y_shift];
- else{
- ref0 = map_col_to_list0[1][l1ref1[x8 + (y8*2>>y_shift)*h->b8_stride]*2>>y_shift];
- l1mv= l1mv1;
- }
- scale = dist_scale_factor[ref0];
- fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
+ if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
+ continue;
+ h->sub_mb_type[i8] = sub_mb_type;
- {
- const int16_t *mv_col = l1mv[x8*3 + (y8*6>>y_shift)*h->b_stride];
- int my_col = (mv_col[1]<<y_shift)/2;
- int mx = (scale * mv_col[0] + 128) >> 8;
- int my = (scale * my_col + 128) >> 8;
- fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mx,my), 4);
- fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mx-mv_col[0],my-my_col), 4);
- }
+ fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
+ if(IS_INTRA(mb_type_col[y8])){
+ fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, 0, 1);
+ fill_rectangle(&h-> mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
+ fill_rectangle(&h-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
+ continue;
+ }
+
+ ref0 = l1ref0[x8 + y8*b8_stride];
+ 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];
+ l1mv= l1mv1;
+ }
+ scale = dist_scale_factor[ref0];
+ fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
+
+ {
+ const int16_t *mv_col = l1mv[x8*3 + y8*b4_stride];
+ int my_col = (mv_col[1]<<y_shift)/2;
+ int mx = (scale * mv_col[0] + 128) >> 8;
+ int my = (scale * my_col + 128) >> 8;
+ fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mx,my), 4);
+ fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mx-mv_col[0],my-my_col), 4);
}
- return;
}
+ return;
}
/* one-to-one mv scaling */
int ref, mv0, mv1;
fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, 0, 1);
- if(IS_INTRA(mb_type_col)){
+ if(IS_INTRA(mb_type_col[0])){
ref=mv0=mv1=0;
}else{
- const int ref0 = l1ref0[0] >= 0 ? map_col_to_list0[0][l1ref0[0]]
- : map_col_to_list0[1][l1ref1[0]];
+ const int ref0 = l1ref0[0] >= 0 ? map_col_to_list0[0][l1ref0[0] + ref_offset]
+ : map_col_to_list0[1][l1ref1[0] + ref_offset];
const int scale = dist_scale_factor[ref0];
const int16_t *mv_col = l1ref0[0] >= 0 ? l1mv0[0] : l1mv1[0];
int mv_l0[2];
continue;
h->sub_mb_type[i8] = sub_mb_type;
fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
- if(IS_INTRA(mb_type_col)){
+ if(IS_INTRA(mb_type_col[0])){
fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, 0, 1);
fill_rectangle(&h-> mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
fill_rectangle(&h-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
continue;
}
- ref0 = l1ref0[x8 + y8*h->b8_stride];
+ ref0 = l1ref0[x8 + y8*b8_stride] + ref_offset;
if(ref0 >= 0)
ref0 = map_col_to_list0[0][ref0];
else{
- ref0 = map_col_to_list0[1][l1ref1[x8 + y8*h->b8_stride]];
+ ref0 = map_col_to_list0[1][l1ref1[x8 + y8*b8_stride] + ref_offset];
l1mv= l1mv1;
}
scale = dist_scale_factor[ref0];
fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
if(IS_SUB_8X8(sub_mb_type)){
- const int16_t *mv_col = l1mv[x8*3 + y8*3*h->b_stride];
+ const int16_t *mv_col = l1mv[x8*3 + y8*3*b4_stride];
int mx = (scale * mv_col[0] + 128) >> 8;
int my = (scale * mv_col[1] + 128) >> 8;
fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mx,my), 4);
fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mx-mv_col[0],my-mv_col[1]), 4);
}else
for(i4=0; i4<4; i4++){
- const int16_t *mv_col = l1mv[x8*2 + (i4&1) + (y8*2 + (i4>>1))*h->b_stride];
+ const int16_t *mv_col = l1mv[x8*2 + (i4&1) + (y8*2 + (i4>>1))*b4_stride];
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;
return h->pps.chroma_qp_table[t][qscale];
}
-//FIXME need to check that this does not overflow signed 32 bit for low qp, I am not sure, it's very close
-//FIXME check that gcc inlines this (and optimizes intra & separate_dc stuff away)
-static inline int quantize_c(DCTELEM *block, uint8_t *scantable, int qscale, int intra, int separate_dc){
- int i;
- const int * const quant_table= quant_coeff[qscale];
- const int bias= intra ? (1<<QUANT_SHIFT)/3 : (1<<QUANT_SHIFT)/6;
- const unsigned int threshold1= (1<<QUANT_SHIFT) - bias - 1;
- const unsigned int threshold2= (threshold1<<1);
- int last_non_zero;
-
- if(separate_dc){
- if(qscale<=18){
- //avoid overflows
- const int dc_bias= intra ? (1<<(QUANT_SHIFT-2))/3 : (1<<(QUANT_SHIFT-2))/6;
- const unsigned int dc_threshold1= (1<<(QUANT_SHIFT-2)) - dc_bias - 1;
- const unsigned int dc_threshold2= (dc_threshold1<<1);
-
- int level= block[0]*quant_coeff[qscale+18][0];
- if(((unsigned)(level+dc_threshold1))>dc_threshold2){
- if(level>0){
- level= (dc_bias + level)>>(QUANT_SHIFT-2);
- block[0]= level;
- }else{
- level= (dc_bias - level)>>(QUANT_SHIFT-2);
- block[0]= -level;
- }
-// last_non_zero = i;
- }else{
- block[0]=0;
- }
- }else{
- const int dc_bias= intra ? (1<<(QUANT_SHIFT+1))/3 : (1<<(QUANT_SHIFT+1))/6;
- const unsigned int dc_threshold1= (1<<(QUANT_SHIFT+1)) - dc_bias - 1;
- const unsigned int dc_threshold2= (dc_threshold1<<1);
-
- int level= block[0]*quant_table[0];
- if(((unsigned)(level+dc_threshold1))>dc_threshold2){
- if(level>0){
- level= (dc_bias + level)>>(QUANT_SHIFT+1);
- block[0]= level;
- }else{
- level= (dc_bias - level)>>(QUANT_SHIFT+1);
- block[0]= -level;
- }
-// last_non_zero = i;
- }else{
- block[0]=0;
- }
- }
- last_non_zero= 0;
- i=1;
- }else{
- last_non_zero= -1;
- i=0;
- }
-
- for(; i<16; i++){
- const int j= scantable[i];
- int level= block[j]*quant_table[j];
-
-// if( bias+level >= (1<<(QMAT_SHIFT - 3))
-// || bias-level >= (1<<(QMAT_SHIFT - 3))){
- if(((unsigned)(level+threshold1))>threshold2){
- if(level>0){
- level= (bias + level)>>QUANT_SHIFT;
- block[j]= level;
- }else{
- level= (bias - level)>>QUANT_SHIFT;
- block[j]= -level;
- }
- last_non_zero = i;
- }else{
- block[j]=0;
- }
- }
-
- return last_non_zero;
-}
-
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,
if (!done) {
int i;
+ int offset;
done = 1;
+ chroma_dc_coeff_token_vlc.table = chroma_dc_coeff_token_vlc_table;
+ chroma_dc_coeff_token_vlc.table_allocated = chroma_dc_coeff_token_vlc_table_size;
init_vlc(&chroma_dc_coeff_token_vlc, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 4*5,
&chroma_dc_coeff_token_len [0], 1, 1,
- &chroma_dc_coeff_token_bits[0], 1, 1, 1);
+ &chroma_dc_coeff_token_bits[0], 1, 1,
+ INIT_VLC_USE_NEW_STATIC);
+ offset = 0;
for(i=0; i<4; i++){
+ coeff_token_vlc[i].table = coeff_token_vlc_tables+offset;
+ coeff_token_vlc[i].table_allocated = coeff_token_vlc_tables_size[i];
init_vlc(&coeff_token_vlc[i], COEFF_TOKEN_VLC_BITS, 4*17,
&coeff_token_len [i][0], 1, 1,
- &coeff_token_bits[i][0], 1, 1, 1);
+ &coeff_token_bits[i][0], 1, 1,
+ INIT_VLC_USE_NEW_STATIC);
+ offset += coeff_token_vlc_tables_size[i];
}
+ /*
+ * This is a one time safety check to make sure that
+ * the packed static coeff_token_vlc table sizes
+ * were initialized correctly.
+ */
+ assert(offset == FF_ARRAY_ELEMS(coeff_token_vlc_tables));
for(i=0; i<3; i++){
- init_vlc(&chroma_dc_total_zeros_vlc[i], CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 4,
+ chroma_dc_total_zeros_vlc[i].table = chroma_dc_total_zeros_vlc_tables[i];
+ chroma_dc_total_zeros_vlc[i].table_allocated = chroma_dc_total_zeros_vlc_tables_size;
+ init_vlc(&chroma_dc_total_zeros_vlc[i],
+ CHROMA_DC_TOTAL_ZEROS_VLC_BITS, 4,
&chroma_dc_total_zeros_len [i][0], 1, 1,
- &chroma_dc_total_zeros_bits[i][0], 1, 1, 1);
+ &chroma_dc_total_zeros_bits[i][0], 1, 1,
+ INIT_VLC_USE_NEW_STATIC);
}
for(i=0; i<15; i++){
- init_vlc(&total_zeros_vlc[i], TOTAL_ZEROS_VLC_BITS, 16,
+ total_zeros_vlc[i].table = total_zeros_vlc_tables[i];
+ total_zeros_vlc[i].table_allocated = total_zeros_vlc_tables_size;
+ init_vlc(&total_zeros_vlc[i],
+ TOTAL_ZEROS_VLC_BITS, 16,
&total_zeros_len [i][0], 1, 1,
- &total_zeros_bits[i][0], 1, 1, 1);
+ &total_zeros_bits[i][0], 1, 1,
+ INIT_VLC_USE_NEW_STATIC);
}
for(i=0; i<6; i++){
- init_vlc(&run_vlc[i], RUN_VLC_BITS, 7,
+ run_vlc[i].table = run_vlc_tables[i];
+ run_vlc[i].table_allocated = run_vlc_tables_size;
+ init_vlc(&run_vlc[i],
+ RUN_VLC_BITS, 7,
&run_len [i][0], 1, 1,
- &run_bits[i][0], 1, 1, 1);
+ &run_bits[i][0], 1, 1,
+ INIT_VLC_USE_NEW_STATIC);
}
+ run7_vlc.table = run7_vlc_table,
+ run7_vlc.table_allocated = run7_vlc_table_size;
init_vlc(&run7_vlc, RUN7_VLC_BITS, 16,
&run_len [6][0], 1, 1,
- &run_bits[6][0], 1, 1, 1);
+ &run_bits[6][0], 1, 1,
+ INIT_VLC_USE_NEW_STATIC);
}
}
av_freep(&h->mb2b_xy);
av_freep(&h->mb2b8_xy);
- for(i = 0; i < MAX_SPS_COUNT; i++)
- av_freep(h->sps_buffers + i);
-
- for(i = 0; i < MAX_PPS_COUNT; i++)
- av_freep(h->pps_buffers + i);
-
for(i = 0; i < h->s.avctx->thread_count; i++) {
hx = h->thread_context[i];
if(!hx) continue;
}
for(q=0; q<52; q++){
- int shift = ff_div6[q];
- int idx = ff_rem6[q];
+ int shift = div6[q];
+ int idx = rem6[q];
for(x=0; x<64; x++)
h->dequant8_coeff[i][q][transpose ? (x>>3)|((x&7)<<3) : x] =
((uint32_t)dequant8_coeff_init[idx][ dequant8_coeff_init_scan[((x>>1)&12) | (x&3)] ] *
continue;
for(q=0; q<52; q++){
- int shift = ff_div6[q] + 2;
- int idx = ff_rem6[q];
+ int shift = div6[q] + 2;
+ int idx = rem6[q];
for(x=0; x<16; x++)
h->dequant4_coeff[i][q][transpose ? (x>>2)|((x<<2)&0xF) : x] =
((uint32_t)dequant4_coeff_init[idx][(x&1) + ((x>>2)&1)] *
CHECKED_ALLOCZ(h->intra4x4_pred_mode, big_mb_num * 8 * sizeof(uint8_t))
CHECKED_ALLOCZ(h->non_zero_count , big_mb_num * 16 * sizeof(uint8_t))
- CHECKED_ALLOCZ(h->slice_table_base , (big_mb_num+s->mb_stride) * sizeof(uint8_t))
+ CHECKED_ALLOCZ(h->slice_table_base , (big_mb_num+s->mb_stride) * sizeof(*h->slice_table_base))
CHECKED_ALLOCZ(h->cbp_table, big_mb_num * sizeof(uint16_t))
CHECKED_ALLOCZ(h->chroma_pred_mode_table, big_mb_num * sizeof(uint8_t))
CHECKED_ALLOCZ(h->mvd_table[1], 32*big_mb_num * sizeof(uint16_t));
CHECKED_ALLOCZ(h->direct_table, 32*big_mb_num * sizeof(uint8_t));
- memset(h->slice_table_base, -1, (big_mb_num+s->mb_stride) * sizeof(uint8_t));
+ memset(h->slice_table_base, -1, (big_mb_num+s->mb_stride) * sizeof(*h->slice_table_base));
h->slice_table= h->slice_table_base + s->mb_stride*2 + 1;
CHECKED_ALLOCZ(h->mb2b_xy , big_mb_num * sizeof(uint32_t));
}
h->thread_context[0] = h;
+ h->outputed_poc = INT_MIN;
+ h->prev_poc_msb= 1<<16;
return 0;
}
/* some macroblocks will be accessed before they're available */
if(FRAME_MBAFF || s->avctx->thread_count > 1)
- memset(h->slice_table, -1, (s->mb_height*s->mb_stride-1) * sizeof(uint8_t));
+ 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;
static 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;
int i;
+ int step = 1;
+ int offset = 1;
+ int uvoffset= 1;
+ int top_idx = 1;
+ int skiplast= 0;
src_y -= linesize;
src_cb -= uvlinesize;
src_cr -= uvlinesize;
+ if(!simple && FRAME_MBAFF){
+ if(s->mb_y&1){
+ offset = MB_MBAFF ? 1 : 17;
+ uvoffset= MB_MBAFF ? 1 : 9;
+ if(!MB_MBAFF){
+ *(uint64_t*)(h->top_borders[0][s->mb_x]+ 0)= *(uint64_t*)(src_y + 15*linesize);
+ *(uint64_t*)(h->top_borders[0][s->mb_x]+ 8)= *(uint64_t*)(src_y +8+15*linesize);
+ if(simple || !ENABLE_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
+ *(uint64_t*)(h->top_borders[0][s->mb_x]+16)= *(uint64_t*)(src_cb+7*uvlinesize);
+ *(uint64_t*)(h->top_borders[0][s->mb_x]+24)= *(uint64_t*)(src_cr+7*uvlinesize);
+ }
+ }
+ }else{
+ if(!MB_MBAFF){
+ h->left_border[0]= h->top_borders[0][s->mb_x][15];
+ if(simple || !ENABLE_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
+ h->left_border[34 ]= h->top_borders[0][s->mb_x][16+7 ];
+ h->left_border[34+18]= h->top_borders[0][s->mb_x][16+8+7];
+ }
+ skiplast= 1;
+ }
+ offset =
+ uvoffset=
+ top_idx = MB_MBAFF ? 0 : 1;
+ }
+ step= MB_MBAFF ? 2 : 1;
+ }
+
// There are two lines saved, the line above the the top macroblock of a pair,
// and the line above the bottom macroblock
- h->left_border[0]= h->top_borders[0][s->mb_x][15];
- for(i=1; i<17; i++){
- h->left_border[i]= src_y[15+i* linesize];
+ h->left_border[offset]= h->top_borders[top_idx][s->mb_x][15];
+ for(i=1; i<17 - skiplast; i++){
+ h->left_border[offset+i*step]= src_y[15+i* linesize];
}
- *(uint64_t*)(h->top_borders[0][s->mb_x]+0)= *(uint64_t*)(src_y + 16*linesize);
- *(uint64_t*)(h->top_borders[0][s->mb_x]+8)= *(uint64_t*)(src_y +8+16*linesize);
+ *(uint64_t*)(h->top_borders[top_idx][s->mb_x]+0)= *(uint64_t*)(src_y + 16*linesize);
+ *(uint64_t*)(h->top_borders[top_idx][s->mb_x]+8)= *(uint64_t*)(src_y +8+16*linesize);
if(simple || !ENABLE_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
- h->left_border[17 ]= h->top_borders[0][s->mb_x][16+7];
- h->left_border[17+9]= h->top_borders[0][s->mb_x][24+7];
- for(i=1; i<9; i++){
- h->left_border[i+17 ]= src_cb[7+i*uvlinesize];
- h->left_border[i+17+9]= src_cr[7+i*uvlinesize];
+ h->left_border[uvoffset+34 ]= h->top_borders[top_idx][s->mb_x][16+7];
+ h->left_border[uvoffset+34+18]= h->top_borders[top_idx][s->mb_x][24+7];
+ for(i=1; i<9 - skiplast; i++){
+ h->left_border[uvoffset+34 +i*step]= src_cb[7+i*uvlinesize];
+ h->left_border[uvoffset+34+18+i*step]= src_cr[7+i*uvlinesize];
}
- *(uint64_t*)(h->top_borders[0][s->mb_x]+16)= *(uint64_t*)(src_cb+8*uvlinesize);
- *(uint64_t*)(h->top_borders[0][s->mb_x]+24)= *(uint64_t*)(src_cr+8*uvlinesize);
+ *(uint64_t*)(h->top_borders[top_idx][s->mb_x]+16)= *(uint64_t*)(src_cb+8*uvlinesize);
+ *(uint64_t*)(h->top_borders[top_idx][s->mb_x]+24)= *(uint64_t*)(src_cr+8*uvlinesize);
}
}
int deblock_left;
int deblock_top;
int mb_xy;
+ int step = 1;
+ int offset = 1;
+ int uvoffset= 1;
+ int top_idx = 1;
+
+ if(!simple && FRAME_MBAFF){
+ if(s->mb_y&1){
+ offset = MB_MBAFF ? 1 : 17;
+ uvoffset= MB_MBAFF ? 1 : 9;
+ }else{
+ offset =
+ uvoffset=
+ top_idx = MB_MBAFF ? 0 : 1;
+ }
+ step= MB_MBAFF ? 2 : 1;
+ }
if(h->deblocking_filter == 2) {
mb_xy = h->mb_xy;
deblock_top = h->slice_table[mb_xy] == h->slice_table[h->top_mb_xy];
} else {
deblock_left = (s->mb_x > 0);
- deblock_top = (s->mb_y > 0);
+ deblock_top = (s->mb_y > !!MB_FIELD);
}
src_y -= linesize + 1;
b= t;
if(deblock_left){
- for(i = !deblock_top; i<17; i++){
- XCHG(h->left_border[i ], src_y [i* linesize], temp8, xchg);
+ for(i = !deblock_top; i<16; i++){
+ XCHG(h->left_border[offset+i*step], src_y [i* linesize], temp8, xchg);
}
+ XCHG(h->left_border[offset+i*step], src_y [i* linesize], temp8, 1);
}
if(deblock_top){
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+0), *(uint64_t*)(src_y +1), temp64, xchg);
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+8), *(uint64_t*)(src_y +9), temp64, 1);
+ XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x]+0), *(uint64_t*)(src_y +1), temp64, xchg);
+ XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x]+8), *(uint64_t*)(src_y +9), temp64, 1);
if(s->mb_x+1 < s->mb_width){
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x+1]), *(uint64_t*)(src_y +17), temp64, 1);
+ XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x+1]), *(uint64_t*)(src_y +17), temp64, 1);
}
}
if(simple || !ENABLE_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
if(deblock_left){
- for(i = !deblock_top; i<9; i++){
- XCHG(h->left_border[i+17 ], src_cb[i*uvlinesize], temp8, xchg);
- XCHG(h->left_border[i+17+9], src_cr[i*uvlinesize], temp8, xchg);
- }
- }
- if(deblock_top){
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+16), *(uint64_t*)(src_cb+1), temp64, 1);
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+24), *(uint64_t*)(src_cr+1), temp64, 1);
- }
- }
-}
-
-static inline void backup_pair_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize){
- MpegEncContext * const s = &h->s;
- int i;
-
- src_y -= 2 * linesize;
- src_cb -= 2 * uvlinesize;
- src_cr -= 2 * uvlinesize;
-
- // There are two lines saved, the line above the the top macroblock of a pair,
- // and the line above the bottom macroblock
- h->left_border[0]= h->top_borders[0][s->mb_x][15];
- h->left_border[1]= h->top_borders[1][s->mb_x][15];
- for(i=2; i<34; i++){
- h->left_border[i]= src_y[15+i* linesize];
- }
-
- *(uint64_t*)(h->top_borders[0][s->mb_x]+0)= *(uint64_t*)(src_y + 32*linesize);
- *(uint64_t*)(h->top_borders[0][s->mb_x]+8)= *(uint64_t*)(src_y +8+32*linesize);
- *(uint64_t*)(h->top_borders[1][s->mb_x]+0)= *(uint64_t*)(src_y + 33*linesize);
- *(uint64_t*)(h->top_borders[1][s->mb_x]+8)= *(uint64_t*)(src_y +8+33*linesize);
-
- if(!ENABLE_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
- h->left_border[34 ]= h->top_borders[0][s->mb_x][16+7];
- h->left_border[34+ 1]= h->top_borders[1][s->mb_x][16+7];
- h->left_border[34+18 ]= h->top_borders[0][s->mb_x][24+7];
- h->left_border[34+18+1]= h->top_borders[1][s->mb_x][24+7];
- for(i=2; i<18; i++){
- h->left_border[i+34 ]= src_cb[7+i*uvlinesize];
- h->left_border[i+34+18]= src_cr[7+i*uvlinesize];
- }
- *(uint64_t*)(h->top_borders[0][s->mb_x]+16)= *(uint64_t*)(src_cb+16*uvlinesize);
- *(uint64_t*)(h->top_borders[0][s->mb_x]+24)= *(uint64_t*)(src_cr+16*uvlinesize);
- *(uint64_t*)(h->top_borders[1][s->mb_x]+16)= *(uint64_t*)(src_cb+17*uvlinesize);
- *(uint64_t*)(h->top_borders[1][s->mb_x]+24)= *(uint64_t*)(src_cr+17*uvlinesize);
- }
-}
-
-static inline void xchg_pair_border(H264Context *h, uint8_t *src_y, uint8_t *src_cb, uint8_t *src_cr, int linesize, int uvlinesize, int xchg){
- MpegEncContext * const s = &h->s;
- int temp8, i;
- uint64_t temp64;
- int deblock_left = (s->mb_x > 0);
- int deblock_top = (s->mb_y > 1);
-
- tprintf(s->avctx, "xchg_pair_border: src_y:%p src_cb:%p src_cr:%p ls:%d uvls:%d\n", src_y, src_cb, src_cr, linesize, uvlinesize);
-
- src_y -= 2 * linesize + 1;
- src_cb -= 2 * uvlinesize + 1;
- src_cr -= 2 * uvlinesize + 1;
-
-#define XCHG(a,b,t,xchg)\
-t= a;\
-if(xchg)\
- a= b;\
-b= t;
-
- if(deblock_left){
- for(i = (!deblock_top)<<1; i<34; i++){
- XCHG(h->left_border[i ], src_y [i* linesize], temp8, xchg);
- }
- }
-
- if(deblock_top){
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+0), *(uint64_t*)(src_y +1), temp64, xchg);
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+8), *(uint64_t*)(src_y +9), temp64, 1);
- XCHG(*(uint64_t*)(h->top_borders[1][s->mb_x]+0), *(uint64_t*)(src_y +1 +linesize), temp64, xchg);
- XCHG(*(uint64_t*)(h->top_borders[1][s->mb_x]+8), *(uint64_t*)(src_y +9 +linesize), temp64, 1);
- if(s->mb_x+1 < s->mb_width){
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x+1]), *(uint64_t*)(src_y +17), temp64, 1);
- XCHG(*(uint64_t*)(h->top_borders[1][s->mb_x+1]), *(uint64_t*)(src_y +17 +linesize), temp64, 1);
- }
- }
-
- if(!ENABLE_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
- if(deblock_left){
- for(i = (!deblock_top) << 1; i<18; i++){
- XCHG(h->left_border[i+34 ], src_cb[i*uvlinesize], temp8, xchg);
- XCHG(h->left_border[i+34+18], src_cr[i*uvlinesize], temp8, xchg);
+ for(i = !deblock_top; i<8; i++){
+ XCHG(h->left_border[uvoffset+34 +i*step], src_cb[i*uvlinesize], temp8, xchg);
+ XCHG(h->left_border[uvoffset+34+18+i*step], src_cr[i*uvlinesize], temp8, xchg);
}
+ XCHG(h->left_border[uvoffset+34 +i*step], src_cb[i*uvlinesize], temp8, 1);
+ XCHG(h->left_border[uvoffset+34+18+i*step], src_cr[i*uvlinesize], temp8, 1);
}
if(deblock_top){
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+16), *(uint64_t*)(src_cb+1), temp64, 1);
- XCHG(*(uint64_t*)(h->top_borders[0][s->mb_x]+24), *(uint64_t*)(src_cr+1), temp64, 1);
- XCHG(*(uint64_t*)(h->top_borders[1][s->mb_x]+16), *(uint64_t*)(src_cb+1 +uvlinesize), temp64, 1);
- XCHG(*(uint64_t*)(h->top_borders[1][s->mb_x]+24), *(uint64_t*)(src_cr+1 +uvlinesize), temp64, 1);
+ XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x]+16), *(uint64_t*)(src_cb+1), temp64, 1);
+ XCHG(*(uint64_t*)(h->top_borders[top_idx][s->mb_x]+24), *(uint64_t*)(src_cr+1), temp64, 1);
}
}
}
int linesize, uvlinesize /*dct_offset*/;
int i;
int *block_offset = &h->block_offset[0];
- const unsigned int bottom = mb_y & 1;
const int transform_bypass = (s->qscale == 0 && h->sps.transform_bypass), is_h264 = (simple || s->codec_id == CODEC_ID_H264);
void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride);
void (*idct_dc_add)(uint8_t *dst, DCTELEM *block, int stride);
fill_rectangle(ref, 4, 4, 8, (16+*ref)^(s->mb_y&1), 1);
}else{
for(i=0; i<16; i+=4){
- //FIXME can refs be smaller than 8x8 when !direct_8x8_inference ?
int ref = h->ref_cache[list][scan8[i]];
if(ref >= 0)
fill_rectangle(&h->ref_cache[list][scan8[i]], 2, 2, 8, (16+ref)^(s->mb_y&1), 1);
idct_add = s->dsp.h264_idct_add;
}
- if(!simple && FRAME_MBAFF && h->deblocking_filter && IS_INTRA(mb_type)
- && (!bottom || !IS_INTRA(s->current_picture.mb_type[mb_xy-s->mb_stride]))){
- int mbt_y = mb_y&~1;
- uint8_t *top_y = s->current_picture.data[0] + (mbt_y * 16* s->linesize ) + mb_x * 16;
- uint8_t *top_cb = s->current_picture.data[1] + (mbt_y * 8 * s->uvlinesize) + mb_x * 8;
- uint8_t *top_cr = s->current_picture.data[2] + (mbt_y * 8 * s->uvlinesize) + mb_x * 8;
- xchg_pair_border(h, top_y, top_cb, top_cr, s->linesize, s->uvlinesize, 1);
- }
-
if (!simple && IS_INTRA_PCM(mb_type)) {
- unsigned int x, y;
-
- // The pixels are stored in h->mb array in the same order as levels,
- // copy them in output in the correct order.
- for(i=0; i<16; i++) {
- for (y=0; y<4; y++) {
- for (x=0; x<4; x++) {
- *(dest_y + block_offset[i] + y*linesize + x) = h->mb[i*16+y*4+x];
- }
- }
- }
- for(i=16; i<16+4; i++) {
- for (y=0; y<4; y++) {
- for (x=0; x<4; x++) {
- *(dest_cb + block_offset[i] + y*uvlinesize + x) = h->mb[i*16+y*4+x];
- }
- }
+ for (i=0; i<16; i++) {
+ memcpy(dest_y + i* linesize, h->mb + i*8, 16);
}
- for(i=20; i<20+4; i++) {
- for (y=0; y<4; y++) {
- for (x=0; x<4; x++) {
- *(dest_cr + block_offset[i] + y*uvlinesize + x) = h->mb[i*16+y*4+x];
- }
- }
+ 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);
}
} else {
if(IS_INTRA(mb_type)){
- if(h->deblocking_filter && (simple || !FRAME_MBAFF))
+ if(h->deblocking_filter)
xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 1, simple);
if(simple || !ENABLE_GRAY || !(s->flags&CODEC_FLAG_GRAY)){
}else
svq3_luma_dc_dequant_idct_c(h->mb, s->qscale);
}
- if(h->deblocking_filter && (simple || !FRAME_MBAFF))
+ if(h->deblocking_filter)
xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0, simple);
}else if(is_h264){
hl_motion(h, dest_y, dest_cb, dest_cr,
}
}
if(h->deblocking_filter) {
+ backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, simple);
+ fill_caches(h, mb_type, 1); //FIXME don't fill stuff which isn't used by filter_mb
+ 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]);
if (!simple && FRAME_MBAFF) {
- //FIXME try deblocking one mb at a time?
- // the reduction in load/storing mvs and such might outweigh the extra backup/xchg_border
- const int mb_y = s->mb_y - 1;
- uint8_t *pair_dest_y, *pair_dest_cb, *pair_dest_cr;
- const int mb_xy= mb_x + mb_y*s->mb_stride;
- const int mb_type_top = s->current_picture.mb_type[mb_xy];
- const int mb_type_bottom= s->current_picture.mb_type[mb_xy+s->mb_stride];
- if (!bottom) return;
- pair_dest_y = s->current_picture.data[0] + (mb_y * 16* s->linesize ) + mb_x * 16;
- pair_dest_cb = s->current_picture.data[1] + (mb_y * 8 * s->uvlinesize) + mb_x * 8;
- pair_dest_cr = s->current_picture.data[2] + (mb_y * 8 * s->uvlinesize) + mb_x * 8;
-
- if(IS_INTRA(mb_type_top | mb_type_bottom))
- xchg_pair_border(h, pair_dest_y, pair_dest_cb, pair_dest_cr, s->linesize, s->uvlinesize, 0);
-
- backup_pair_border(h, pair_dest_y, pair_dest_cb, pair_dest_cr, s->linesize, s->uvlinesize);
- // deblock a pair
- // top
- s->mb_y--; h->mb_xy -= s->mb_stride;
- tprintf(h->s.avctx, "call mbaff filter_mb mb_x:%d mb_y:%d pair_dest_y = %p, dest_y = %p\n", mb_x, mb_y, pair_dest_y, dest_y);
- fill_caches(h, mb_type_top, 1); //FIXME don't fill stuff which isn't used by filter_mb
- 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]);
- filter_mb(h, mb_x, mb_y, pair_dest_y, pair_dest_cb, pair_dest_cr, linesize, uvlinesize);
- // bottom
- s->mb_y++; h->mb_xy += s->mb_stride;
- tprintf(h->s.avctx, "call mbaff filter_mb\n");
- fill_caches(h, mb_type_bottom, 1); //FIXME don't fill stuff which isn't used by filter_mb
- h->chroma_qp[0] = get_chroma_qp(h, 0, s->current_picture.qscale_table[mb_xy+s->mb_stride]);
- h->chroma_qp[1] = get_chroma_qp(h, 1, s->current_picture.qscale_table[mb_xy+s->mb_stride]);
- filter_mb(h, mb_x, mb_y+1, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
+ filter_mb (h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
} else {
- tprintf(h->s.avctx, "call filter_mb\n");
- backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, simple);
- fill_caches(h, mb_type, 1); //FIXME don't fill stuff which isn't used by filter_mb
- 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]);
filter_mb_fast(h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
}
}
pic->reference = parity;
pic->linesize[i] *= 2;
}
+ pic->poc= pic->field_poc[parity == PICT_BOTTOM_FIELD];
}
static int split_field_copy(Picture *dest, Picture *src,
if (match) {
*dest = *src;
- pic_as_field(dest, parity);
- dest->pic_id *= 2;
- dest->pic_id += id_add;
+ if(parity != PICT_FRAME){
+ pic_as_field(dest, parity);
+ dest->pic_id *= 2;
+ dest->pic_id += id_add;
+ }
}
return match;
}
-/**
- * Split one reference list into field parts, interleaving by parity
- * as per H.264 spec section 8.2.4.2.5. Output fields have their data pointers
- * set to look at the actual start of data for that field.
- *
- * @param dest output list
- * @param dest_len maximum number of fields to put in dest
- * @param src the source reference list containing fields and/or field pairs
- * (aka short_ref/long_ref, or
- * refFrameListXShortTerm/refFrameListLongTerm in spec-speak)
- * @param src_len number of Picture's in source (pairs and unmatched fields)
- * @param parity the parity of the picture being decoded/needing
- * these ref pics (PICT_{TOP,BOTTOM}_FIELD)
- * @return number of fields placed in dest
- */
-static int split_field_half_ref_list(Picture *dest, int dest_len,
- Picture *src, int src_len, int parity){
- int same_parity = 1;
- int same_i = 0;
- int opp_i = 0;
- int out_i;
- int field_output;
-
- for (out_i = 0; out_i < dest_len; out_i += field_output) {
- if (same_parity && same_i < src_len) {
- field_output = split_field_copy(dest + out_i, src + same_i,
- parity, 1);
- same_parity = !field_output;
- same_i++;
-
- } else if (opp_i < src_len) {
- field_output = split_field_copy(dest + out_i, src + opp_i,
- PICT_FRAME - parity, 0);
- same_parity = field_output;
- opp_i++;
+static int build_def_list(Picture *def, Picture **in, int len, int is_long, int sel){
+ int i[2]={0};
+ int index=0;
- } else {
- break;
+ while(i[0]<len || i[1]<len){
+ while(i[0]<len && !(in[ i[0] ] && (in[ i[0] ]->reference & sel)))
+ i[0]++;
+ while(i[1]<len && !(in[ i[1] ] && (in[ i[1] ]->reference & (sel^3))))
+ i[1]++;
+ if(i[0] < len){
+ in[ i[0] ]->pic_id= is_long ? i[0] : in[ i[0] ]->frame_num;
+ split_field_copy(&def[index++], in[ i[0]++ ], sel , 1);
+ }
+ if(i[1] < len){
+ in[ i[1] ]->pic_id= is_long ? i[1] : in[ i[1] ]->frame_num;
+ split_field_copy(&def[index++], in[ i[1]++ ], sel^3, 0);
}
}
- return out_i;
+ return index;
}
-/**
- * Split the reference frame list into a reference field list.
- * This implements H.264 spec 8.2.4.2.5 for a combined input list.
- * The input list contains both reference field pairs and
- * unmatched reference fields; it is ordered as spec describes
- * RefPicListX for frames in 8.2.4.2.1 and 8.2.4.2.3, except that
- * unmatched field pairs are also present. Conceptually this is equivalent
- * to concatenation of refFrameListXShortTerm with refFrameListLongTerm.
- *
- * @param dest output reference list where ordered fields are to be placed
- * @param dest_len max number of fields to place at dest
- * @param src source reference list, as described above
- * @param src_len number of pictures (pairs and unmatched fields) in src
- * @param parity parity of field being currently decoded
- * (one of PICT_{TOP,BOTTOM}_FIELD)
- * @param long_i index into src array that holds first long reference picture,
- * or src_len if no long refs present.
- */
-static int split_field_ref_list(Picture *dest, int dest_len,
- Picture *src, int src_len,
- int parity, int long_i){
+static int add_sorted(Picture **sorted, Picture **src, int len, int limit, int dir){
+ int i, best_poc;
+ int out_i= 0;
- int i = split_field_half_ref_list(dest, dest_len, src, long_i, parity);
- dest += i;
- dest_len -= i;
+ for(;;){
+ best_poc= dir ? INT_MIN : INT_MAX;
- i += split_field_half_ref_list(dest, dest_len, src + long_i,
- src_len - long_i, parity);
- return i;
+ for(i=0; i<len; i++){
+ const int poc= src[i]->poc;
+ if(((poc > limit) ^ dir) && ((poc < best_poc) ^ dir)){
+ best_poc= poc;
+ sorted[out_i]= src[i];
+ }
+ }
+ if(best_poc == (dir ? INT_MIN : INT_MAX))
+ break;
+ limit= sorted[out_i++]->poc - dir;
+ }
+ return out_i;
}
/**
*/
static int fill_default_ref_list(H264Context *h){
MpegEncContext * const s = &h->s;
- int i;
- int smallest_poc_greater_than_current = -1;
- int structure_sel;
- Picture sorted_short_ref[32];
- Picture field_entry_list[2][32];
- Picture *frame_list[2];
-
- if (FIELD_PICTURE) {
- structure_sel = PICT_FRAME;
- frame_list[0] = field_entry_list[0];
- frame_list[1] = field_entry_list[1];
- } else {
- structure_sel = 0;
- frame_list[0] = h->default_ref_list[0];
- frame_list[1] = h->default_ref_list[1];
- }
+ int i, len;
if(h->slice_type_nos==FF_B_TYPE){
- int list;
- int len[2];
- int short_len[2];
- int out_i;
- int limit= INT_MIN;
-
- /* sort frame according to POC in B slice */
- for(out_i=0; out_i<h->short_ref_count; out_i++){
- int best_i=INT_MIN;
- int best_poc=INT_MAX;
-
- for(i=0; i<h->short_ref_count; i++){
- const int poc= h->short_ref[i]->poc;
- if(poc > limit && poc < best_poc){
- best_poc= poc;
- best_i= i;
- }
- }
-
- assert(best_i != INT_MIN);
-
- limit= best_poc;
- sorted_short_ref[out_i]= *h->short_ref[best_i];
- tprintf(h->s.avctx, "sorted poc: %d->%d poc:%d fn:%d\n", best_i, out_i, sorted_short_ref[out_i].poc, sorted_short_ref[out_i].frame_num);
- if (-1 == smallest_poc_greater_than_current) {
- if (h->short_ref[best_i]->poc >= s->current_picture_ptr->poc) {
- smallest_poc_greater_than_current = out_i;
- }
- }
- }
+ Picture *sorted[32];
+ int cur_poc, list;
+ int lens[2];
- tprintf(h->s.avctx, "current poc: %d, smallest_poc_greater_than_current: %d\n", s->current_picture_ptr->poc, smallest_poc_greater_than_current);
-
- // find the largest POC
- for(list=0; list<2; list++){
- int index = 0;
- int j= -99;
- int step= list ? -1 : 1;
-
- for(i=0; i<h->short_ref_count && index < h->ref_count[list]; i++, j+=step) {
- int sel;
- while(j<0 || j>= h->short_ref_count){
- if(j != -99 && step == (list ? -1 : 1))
- return -1;
- step = -step;
- j= smallest_poc_greater_than_current + (step>>1);
- }
- sel = sorted_short_ref[j].reference | structure_sel;
- if(sel != PICT_FRAME) continue;
- frame_list[list][index ]= sorted_short_ref[j];
- frame_list[list][index++].pic_id= sorted_short_ref[j].frame_num;
- }
- short_len[list] = index;
+ if(FIELD_PICTURE)
+ cur_poc= s->current_picture_ptr->field_poc[ s->picture_structure == PICT_BOTTOM_FIELD ];
+ else
+ cur_poc= s->current_picture_ptr->poc;
- for(i = 0; i < 16 && index < h->ref_count[ list ]; i++){
- int sel;
- if(h->long_ref[i] == NULL) continue;
- sel = h->long_ref[i]->reference | structure_sel;
- if(sel != PICT_FRAME) continue;
+ for(list= 0; list<2; list++){
+ len= add_sorted(sorted , h->short_ref, h->short_ref_count, cur_poc, 1^list);
+ len+=add_sorted(sorted+len, h->short_ref, h->short_ref_count, cur_poc, 0^list);
+ assert(len<=32);
+ len= build_def_list(h->default_ref_list[list] , sorted , len, 0, s->picture_structure);
+ len+=build_def_list(h->default_ref_list[list]+len, h->long_ref, 16 , 1, s->picture_structure);
+ assert(len<=32);
- frame_list[ list ][index ]= *h->long_ref[i];
- frame_list[ list ][index++].pic_id= i;
- }
- len[list] = index;
+ if(len < h->ref_count[list])
+ memset(&h->default_ref_list[list][len], 0, sizeof(Picture)*(h->ref_count[list] - len));
+ lens[list]= len;
}
- for(list=0; list<2; list++){
- if (FIELD_PICTURE)
- len[list] = split_field_ref_list(h->default_ref_list[list],
- h->ref_count[list],
- frame_list[list],
- len[list],
- s->picture_structure,
- short_len[list]);
-
- // swap the two first elements of L1 when L0 and L1 are identical
- if(list && len[0] > 1 && len[0] == len[1])
- for(i=0; h->default_ref_list[0][i].data[0] == h->default_ref_list[1][i].data[0]; i++)
- if(i == len[0]){
- FFSWAP(Picture, h->default_ref_list[1][0], h->default_ref_list[1][1]);
- break;
- }
-
- if(len[list] < h->ref_count[ list ])
- memset(&h->default_ref_list[list][len[list]], 0, sizeof(Picture)*(h->ref_count[ list ] - len[list]));
+ if(lens[0] == lens[1] && lens[1] > 1){
+ for(i=0; h->default_ref_list[0][i].data[0] == h->default_ref_list[1][i].data[0] && i<lens[0]; i++);
+ if(i == lens[0])
+ FFSWAP(Picture, h->default_ref_list[1][0], h->default_ref_list[1][1]);
}
-
-
}else{
- int index=0;
- int short_len;
- for(i=0; i<h->short_ref_count; i++){
- int sel;
- sel = h->short_ref[i]->reference | structure_sel;
- if(sel != PICT_FRAME) continue;
- frame_list[0][index ]= *h->short_ref[i];
- frame_list[0][index++].pic_id= h->short_ref[i]->frame_num;
- }
- short_len = index;
- for(i = 0; i < 16; i++){
- int sel;
- if(h->long_ref[i] == NULL) continue;
- sel = h->long_ref[i]->reference | structure_sel;
- if(sel != PICT_FRAME) continue;
- frame_list[0][index ]= *h->long_ref[i];
- frame_list[0][index++].pic_id= i;
- }
-
- if (FIELD_PICTURE)
- index = split_field_ref_list(h->default_ref_list[0],
- h->ref_count[0], frame_list[0],
- index, s->picture_structure,
- short_len);
-
- if(index < h->ref_count[0])
- memset(&h->default_ref_list[0][index], 0, sizeof(Picture)*(h->ref_count[0] - index));
+ len = build_def_list(h->default_ref_list[0] , h->short_ref, h->short_ref_count, 0, s->picture_structure);
+ len+= build_def_list(h->default_ref_list[0]+len, h-> long_ref, 16 , 1, s->picture_structure);
+ assert(len <= 32);
+ if(len < h->ref_count[0])
+ memset(&h->default_ref_list[0][len], 0, sizeof(Picture)*(h->ref_count[0] - len));
}
#ifdef TRACE
for (i=0; i<h->ref_count[0]; i++) {
print_short_term(h);
print_long_term(h);
- if(h->slice_type_nos==FF_I_TYPE) return 0; //FIXME move before function
for(list=0; list<h->list_count; list++){
memcpy(h->ref_list[list], h->default_ref_list[list], sizeof(Picture)*h->ref_count[list]);
ref = h->short_ref[i];
assert(ref->reference);
assert(!ref->long_ref);
- if(ref->data[0] != NULL &&
+ if(
ref->frame_num == frame_num &&
- (ref->reference & pic_structure) &&
- ref->long_ref == 0) // ignore non-existing pictures by testing data[0] pointer
+ (ref->reference & pic_structure)
+ )
break;
}
if(i>=0)
}
for(list=0; list<h->list_count; list++){
for(index= 0; index < h->ref_count[list]; index++){
- if(!h->ref_list[list][index].data[0])
- h->ref_list[list][index]= s->current_picture;
+ if(!h->ref_list[list][index].data[0]){
+ av_log(h->s.avctx, AV_LOG_ERROR, "Missing reference picture\n");
+ h->ref_list[list][index]= s->current_picture; //FIXME this is not a sensible solution
+ }
}
}
- if(h->slice_type_nos==FF_B_TYPE && !h->direct_spatial_mv_pred)
- direct_dist_scale_factor(h);
- direct_ref_list_init(h);
return 0;
}
for(j=0; j<3; j++)
field[0].linesize[j] <<= 1;
field[0].reference = PICT_TOP_FIELD;
+ field[0].poc= field[0].field_poc[0];
field[1] = field[0];
for(j=0; j<3; j++)
field[1].data[j] += frame->linesize[j];
field[1].reference = PICT_BOTTOM_FIELD;
+ field[1].poc= field[1].field_poc[1];
h->luma_weight[list][16+2*i] = h->luma_weight[list][16+2*i+1] = h->luma_weight[list][i];
h->luma_offset[list][16+2*i] = h->luma_offset[list][16+2*i+1] = h->luma_offset[list][i];
h->luma_offset[list][i]= 0;
}
- chroma_weight_flag= get_bits1(&s->gb);
- if(chroma_weight_flag){
- int j;
- for(j=0; j<2; j++){
- h->chroma_weight[list][i][j]= get_se_golomb(&s->gb);
- h->chroma_offset[list][i][j]= get_se_golomb(&s->gb);
- if( h->chroma_weight[list][i][j] != chroma_def
- || h->chroma_offset[list][i][j] != 0)
- h->use_weight_chroma= 1;
- }
- }else{
- int j;
- for(j=0; j<2; j++){
- h->chroma_weight[list][i][j]= chroma_def;
- h->chroma_offset[list][i][j]= 0;
+ if(CHROMA){
+ chroma_weight_flag= get_bits1(&s->gb);
+ if(chroma_weight_flag){
+ int j;
+ for(j=0; j<2; j++){
+ h->chroma_weight[list][i][j]= get_se_golomb(&s->gb);
+ h->chroma_offset[list][i][j]= get_se_golomb(&s->gb);
+ if( h->chroma_weight[list][i][j] != chroma_def
+ || h->chroma_offset[list][i][j] != 0)
+ h->use_weight_chroma= 1;
+ }
+ }else{
+ int j;
+ for(j=0; j<2; j++){
+ h->chroma_weight[list][i][j]= chroma_def;
+ h->chroma_offset[list][i][j]= 0;
+ }
}
}
}
int i;
for(i=0; i<16; i++){
- if (h->long_ref[i] != NULL) {
- unreference_pic(h, h->long_ref[i], 0);
- remove_long_at_index(h, i);
- }
+ remove_long(h, i, 0);
}
assert(h->long_ref_count==0);
*
* @return the removed picture or NULL if an error occurs
*/
-static Picture * remove_short(H264Context *h, int frame_num){
+static Picture * remove_short(H264Context *h, int frame_num, int ref_mask){
MpegEncContext * const s = &h->s;
Picture *pic;
int i;
av_log(h->s.avctx, AV_LOG_DEBUG, "remove short %d count %d\n", frame_num, h->short_ref_count);
pic = find_short(h, frame_num, &i);
- if (pic)
+ if (pic){
+ if(unreference_pic(h, pic, ref_mask))
remove_short_at_index(h, i);
+ }
return pic;
}
/**
* Remove a picture from the long term reference list by its index in
- * that list. This does no checking on the provided index; it is assumed
- * to be valid. The removed entry is set to NULL. Other entries are unaffected.
- * @param i index into h->long_ref of picture to remove.
- */
-static void remove_long_at_index(H264Context *h, int i){
- assert(h->long_ref[i]->long_ref == 1);
- h->long_ref[i]->long_ref= 0;
- h->long_ref[i]= NULL;
- h->long_ref_count--;
-}
-
-/**
- *
+ * that list.
* @return the removed picture or NULL if an error occurs
*/
-static Picture * remove_long(H264Context *h, int i){
+static Picture * remove_long(H264Context *h, int i, int ref_mask){
Picture *pic;
pic= h->long_ref[i];
- if (pic)
- remove_long_at_index(h, i);
+ if (pic){
+ if(unreference_pic(h, pic, ref_mask)){
+ assert(h->long_ref[i]->long_ref == 1);
+ h->long_ref[i]->long_ref= 0;
+ h->long_ref[i]= NULL;
+ h->long_ref_count--;
+ }
+ }
return pic;
}
if(s->avctx->debug&FF_DEBUG_MMCO)
av_log(h->s.avctx, AV_LOG_DEBUG, "mmco:%d %d %d\n", h->mmco[i].opcode, h->mmco[i].short_pic_num, h->mmco[i].long_arg);
+ if( mmco[i].opcode == MMCO_SHORT2UNUSED
+ || mmco[i].opcode == MMCO_SHORT2LONG){
+ frame_num = pic_num_extract(h, mmco[i].short_pic_num, &structure);
+ pic = find_short(h, frame_num, &j);
+ if(!pic){
+ if(mmco[i].opcode != MMCO_SHORT2LONG || !h->long_ref[mmco[i].long_arg]
+ || h->long_ref[mmco[i].long_arg]->frame_num != frame_num)
+ av_log(h->s.avctx, AV_LOG_ERROR, "mmco: unref short failure\n");
+ continue;
+ }
+ }
+
switch(mmco[i].opcode){
case MMCO_SHORT2UNUSED:
if(s->avctx->debug&FF_DEBUG_MMCO)
av_log(h->s.avctx, AV_LOG_DEBUG, "mmco: unref short %d count %d\n", h->mmco[i].short_pic_num, h->short_ref_count);
- frame_num = pic_num_extract(h, mmco[i].short_pic_num, &structure);
- pic = find_short(h, frame_num, &j);
- if (pic) {
- if (unreference_pic(h, pic, structure ^ PICT_FRAME))
- remove_short_at_index(h, j);
- } else if(s->avctx->debug&FF_DEBUG_MMCO)
- av_log(h->s.avctx, AV_LOG_DEBUG, "mmco: unref short failure\n");
+ remove_short(h, frame_num, structure ^ PICT_FRAME);
break;
case MMCO_SHORT2LONG:
- if (FIELD_PICTURE && mmco[i].long_arg < h->long_ref_count &&
- h->long_ref[mmco[i].long_arg]->frame_num ==
- mmco[i].short_pic_num / 2) {
- /* do nothing, we've already moved this field pair. */
- } else {
- int frame_num = mmco[i].short_pic_num >> FIELD_PICTURE;
+ if (h->long_ref[mmco[i].long_arg] != pic)
+ remove_long(h, mmco[i].long_arg, 0);
- pic= remove_long(h, mmco[i].long_arg);
- if(pic) unreference_pic(h, pic, 0);
-
- h->long_ref[ mmco[i].long_arg ]= remove_short(h, frame_num);
+ remove_short_at_index(h, j);
+ h->long_ref[ mmco[i].long_arg ]= pic;
if (h->long_ref[ mmco[i].long_arg ]){
h->long_ref[ mmco[i].long_arg ]->long_ref=1;
h->long_ref_count++;
}
- }
break;
case MMCO_LONG2UNUSED:
j = pic_num_extract(h, mmco[i].long_arg, &structure);
pic = h->long_ref[j];
if (pic) {
- if (unreference_pic(h, pic, structure ^ PICT_FRAME))
- remove_long_at_index(h, j);
+ remove_long(h, j, structure ^ PICT_FRAME);
} else if(s->avctx->debug&FF_DEBUG_MMCO)
av_log(h->s.avctx, AV_LOG_DEBUG, "mmco: unref long failure\n");
break;
// Comment below left from previous code as it is an interresting note.
/* First field in pair is in short term list or
* at a different long term index.
- * This is not allowed; see 7.4.3, notes 2 and 3.
+ * This is not allowed; see 7.4.3.3, notes 2 and 3.
* Report the problem and keep the pair where it is,
* and mark this field valid.
*/
if (h->long_ref[mmco[i].long_arg] != s->current_picture_ptr) {
- pic= remove_long(h, mmco[i].long_arg);
- if(pic) unreference_pic(h, pic, 0);
+ remove_long(h, mmco[i].long_arg, 0);
h->long_ref[ mmco[i].long_arg ]= s->current_picture_ptr;
h->long_ref[ mmco[i].long_arg ]->long_ref=1;
assert(mmco[i].long_arg <= 16);
// just remove the long term which index is greater than new max
for(j = mmco[i].long_arg; j<16; j++){
- pic = remove_long(h, j);
- if (pic) unreference_pic(h, pic, 0);
+ remove_long(h, j, 0);
}
break;
case MMCO_RESET:
while(h->short_ref_count){
- pic= remove_short(h, h->short_ref[0]->frame_num);
- if(pic) unreference_pic(h, pic, 0);
+ remove_short(h, h->short_ref[0]->frame_num, 0);
}
for(j = 0; j < 16; j++) {
- pic= remove_long(h, j);
- if(pic) unreference_pic(h, pic, 0);
+ remove_long(h, j, 0);
}
s->current_picture_ptr->poc=
s->current_picture_ptr->field_poc[0]=
* which is already referenced. If short referenced, it
* should be first entry in short_ref. If not, it must exist
* in long_ref; trying to put it on the short list here is an
- * error in the encoded bit stream (ref: 7.4.3, NOTE 2 and 3).
+ * error in the encoded bit stream (ref: 7.4.3.3, NOTE 2 and 3).
*/
if (h->short_ref_count && h->short_ref[0] == s->current_picture_ptr) {
/* Just mark the second field valid */
"in complementary field pair "
"(first field is long term)\n");
} else {
- pic= remove_short(h, s->current_picture_ptr->frame_num);
+ pic= remove_short(h, s->current_picture_ptr->frame_num, 0);
if(pic){
- unreference_pic(h, pic, 0);
av_log(h->s.avctx, AV_LOG_ERROR, "illegal short term buffer state detected\n");
}
break;
assert(i < 16);
- pic = h->long_ref[i];
- remove_long_at_index(h, i);
+ remove_long(h, i, 0);
} else {
pic = h->short_ref[h->short_ref_count - 1];
- remove_short_at_index(h, h->short_ref_count - 1);
+ remove_short(h, pic->frame_num, 0);
}
- unreference_pic(h, pic, 0);
}
print_short_term(h);
unsigned int first_mb_in_slice;
unsigned int pps_id;
int num_ref_idx_active_override_flag;
- static const uint8_t slice_type_map[5]= {FF_P_TYPE, FF_B_TYPE, FF_I_TYPE, FF_SP_TYPE, FF_SI_TYPE};
unsigned int slice_type, tmp, i, j;
int default_ref_list_done = 0;
int last_pic_structure;
}else
h->slice_type_fixed=0;
- slice_type= slice_type_map[ slice_type ];
+ slice_type= golomb_to_pict_type[ slice_type ];
if (slice_type == FF_I_TYPE
|| (h0->current_slice != 0 && slice_type == h0->last_slice_type) ) {
default_ref_list_done = 1;
if(h != h0)
return -1; // width / height changed during parallelized decoding
free_tables(h);
+ flush_dpb(s->avctx);
MPV_common_end(s);
}
if (!s->context_initialized) {
h->mb_aff_frame = h->sps.mb_aff;
}
}
+ h->mb_field_decoding_flag= s->picture_structure != PICT_FRAME;
if(h0->current_slice == 0){
while(h->frame_num != h->prev_frame_num &&
fill_default_ref_list(h);
}
- if(decode_ref_pic_list_reordering(h) < 0)
+ if(h->slice_type_nos!=FF_I_TYPE && decode_ref_pic_list_reordering(h) < 0)
return -1;
+ if(h->slice_type_nos!=FF_I_TYPE){
+ s->last_picture_ptr= &h->ref_list[0][0];
+ ff_copy_picture(&s->last_picture, s->last_picture_ptr);
+ }
+ if(h->slice_type_nos==FF_B_TYPE){
+ s->next_picture_ptr= &h->ref_list[1][0];
+ ff_copy_picture(&s->next_picture, s->next_picture_ptr);
+ }
+
if( (h->pps.weighted_pred && h->slice_type_nos == FF_P_TYPE )
|| (h->pps.weighted_bipred_idc==1 && h->slice_type_nos== FF_B_TYPE ) )
pred_weight_table(h);
if(FRAME_MBAFF)
fill_mbaff_ref_list(h);
+ if(h->slice_type_nos==FF_B_TYPE && !h->direct_spatial_mv_pred)
+ direct_dist_scale_factor(h);
+ direct_ref_list_init(h);
+
if( h->slice_type_nos != FF_I_TYPE && h->pps.cabac ){
tmp = get_ue_golomb(&s->gb);
if(tmp > 2){
h0->last_slice_type = slice_type;
h->slice_num = ++h0->current_slice;
+ if(h->slice_num >= MAX_SLICES){
+ av_log(s->avctx, AV_LOG_ERROR, "Too many slices, increase MAX_SLICES and recompile\n");
+ }
for(j=0; j<2; j++){
- int *ref2frm= h->ref2frm[h->slice_num&15][j];
+ int *ref2frm= h->ref2frm[h->slice_num&(MAX_SLICES-1)][j];
ref2frm[0]=
ref2frm[1]= -1;
- for(i=0; i<48; i++)
+ for(i=0; i<16; i++)
ref2frm[i+2]= 4*h->ref_list[j][i].frame_num
+(h->ref_list[j][i].reference&3);
+ ref2frm[18+0]=
+ ref2frm[18+1]= -1;
+ for(i=16; i<48; i++)
+ ref2frm[i+4]= 4*h->ref_list[j][i].frame_num
+ +(h->ref_list[j][i].reference&3);
}
h->emu_edge_width= (s->flags&CODEC_FLAG_EMU_EDGE) ? 0 : 16;
h->emu_edge_height= (FRAME_MBAFF || FIELD_PICTURE) ? 0 : h->emu_edge_width;
if(s->avctx->debug&FF_DEBUG_PICT_INFO){
- av_log(h->s.avctx, AV_LOG_DEBUG, "slice:%d %s mb:%d %c pps:%u frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d:%d:%d weight:%d%s %s\n",
+ av_log(h->s.avctx, AV_LOG_DEBUG, "slice:%d %s mb:%d %c%s%s pps:%u frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d:%d:%d weight:%d%s %s\n",
h->slice_num,
(s->picture_structure==PICT_FRAME ? "F" : s->picture_structure==PICT_TOP_FIELD ? "T" : "B"),
first_mb_in_slice,
- av_get_pict_type_char(h->slice_type),
+ av_get_pict_type_char(h->slice_type), h->slice_type_fixed ? " fix" : "", h->nal_unit_type == NAL_IDR_SLICE ? " IDR" : "",
pps_id, h->frame_num,
s->current_picture_ptr->field_poc[0], s->current_picture_ptr->field_poc[1],
h->ref_count[0], h->ref_count[1],
if( h->slice_type_nos == FF_B_TYPE )
{
// just for fill_caches. pred_direct_motion will set the real mb_type
- mb_type|= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2|MB_TYPE_SKIP;
+ mb_type|= MB_TYPE_P0L0|MB_TYPE_P0L1|MB_TYPE_DIRECT2|MB_TYPE_SKIP;
fill_caches(h, mb_type, 0); //FIXME check what is needed and what not ...
pred_direct_motion(h, &mb_type);
if(FRAME_MBAFF){
if( (s->mb_y&1) == 0 )
h->mb_mbaff = h->mb_field_decoding_flag = get_bits1(&s->gb);
- }else
- h->mb_field_decoding_flag= (s->picture_structure!=PICT_FRAME);
+ }
h->prev_mb_skipped= 0;
h->slice_table[ mb_xy ]= h->slice_num;
if(IS_INTRA_PCM(mb_type)){
- unsigned int x, y;
+ unsigned int x;
// We assume these blocks are very rare so we do not optimize it.
align_get_bits(&s->gb);
// The pixels are stored in the same order as levels in h->mb array.
- for(y=0; y<16; y++){
- const int index= 4*(y&3) + 32*((y>>2)&1) + 128*(y>>3);
- for(x=0; x<16; x++){
- tprintf(s->avctx, "LUMA ICPM LEVEL (%3d)\n", show_bits(&s->gb, 8));
- h->mb[index + (x&3) + 16*((x>>2)&1) + 64*(x>>3)]= get_bits(&s->gb, 8);
- }
- }
- for(y=0; y<8; y++){
- const int index= 256 + 4*(y&3) + 32*(y>>2);
- for(x=0; x<8; x++){
- tprintf(s->avctx, "CHROMA U ICPM LEVEL (%3d)\n", show_bits(&s->gb, 8));
- h->mb[index + (x&3) + 16*(x>>2)]= get_bits(&s->gb, 8);
- }
- }
- for(y=0; y<8; y++){
- const int index= 256 + 64 + 4*(y&3) + 32*(y>>2);
- for(x=0; x<8; x++){
- tprintf(s->avctx, "CHROMA V ICPM LEVEL (%3d)\n", show_bits(&s->gb, 8));
- h->mb[index + (x&3) + 16*(x>>2)]= get_bits(&s->gb, 8);
- }
+ for(x=0; x < (CHROMA ? 384 : 256); x++){
+ ((uint8_t*)h->mb)[x]= get_bits(&s->gb, 8);
}
// In deblocking, the quantizer is 0
//mb_pred
if(IS_INTRA(mb_type)){
- int pred_mode;
+ int pred_mode;
// init_top_left_availability(h);
- if(IS_INTRA4x4(mb_type)){
- int i;
- int di = 1;
- if(dct8x8_allowed && get_bits1(&s->gb)){
- mb_type |= MB_TYPE_8x8DCT;
- di = 4;
- }
+ if(IS_INTRA4x4(mb_type)){
+ int i;
+ int di = 1;
+ if(dct8x8_allowed && get_bits1(&s->gb)){
+ mb_type |= MB_TYPE_8x8DCT;
+ di = 4;
+ }
// fill_intra4x4_pred_table(h);
- for(i=0; i<16; i+=di){
- int mode= pred_intra_mode(h, i);
-
- if(!get_bits1(&s->gb)){
- const int rem_mode= get_bits(&s->gb, 3);
- mode = rem_mode + (rem_mode >= mode);
- }
+ for(i=0; i<16; i+=di){
+ int mode= pred_intra_mode(h, i);
- if(di==4)
- fill_rectangle( &h->intra4x4_pred_mode_cache[ scan8[i] ], 2, 2, 8, mode, 1 );
- else
- h->intra4x4_pred_mode_cache[ scan8[i] ] = mode;
+ if(!get_bits1(&s->gb)){
+ const int rem_mode= get_bits(&s->gb, 3);
+ mode = rem_mode + (rem_mode >= mode);
}
- write_back_intra_pred_mode(h);
- if( check_intra4x4_pred_mode(h) < 0)
- return -1;
- }else{
- h->intra16x16_pred_mode= check_intra_pred_mode(h, h->intra16x16_pred_mode);
- if(h->intra16x16_pred_mode < 0)
- return -1;
- }
+ if(di==4)
+ fill_rectangle( &h->intra4x4_pred_mode_cache[ scan8[i] ], 2, 2, 8, mode, 1 );
+ else
+ h->intra4x4_pred_mode_cache[ scan8[i] ] = mode;
+ }
+ write_back_intra_pred_mode(h);
+ if( check_intra4x4_pred_mode(h) < 0)
+ return -1;
+ }else{
+ h->intra16x16_pred_mode= check_intra_pred_mode(h, h->intra16x16_pred_mode);
+ if(h->intra16x16_pred_mode < 0)
+ return -1;
+ }
+ if(CHROMA){
pred_mode= check_intra_pred_mode(h, get_ue_golomb(&s->gb));
if(pred_mode < 0)
return -1;
h->chroma_pred_mode= pred_mode;
+ }
}else if(partition_count==4){
int i, j, sub_partition_count[4], list, ref[2][4];
return -1;
}
- if(IS_INTRA4x4(mb_type))
- cbp= golomb_to_intra4x4_cbp[cbp];
- else
- cbp= golomb_to_inter_cbp[cbp];
+ if(CHROMA){
+ if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp[cbp];
+ else cbp= golomb_to_inter_cbp [cbp];
+ }else{
+ if(IS_INTRA4x4(mb_type)) cbp= golomb_to_intra4x4_cbp_gray[cbp];
+ else cbp= golomb_to_inter_cbp_gray[cbp];
+ }
}
h->cbp = cbp;
}
}
- while( coeff_count-- ) {
+ do {
uint8_t *ctx = coeff_abs_level1_ctx[node_ctx] + abs_level_m1_ctx_base;
- int j= scantable[index[coeff_count]];
+ int j= scantable[index[--coeff_count]];
if( get_cabac( CC, ctx ) == 0 ) {
node_ctx = coeff_abs_level_transition[0][node_ctx];
block[j] = (get_cabac_bypass_sign( CC, -coeff_abs ) * qmul[j] + 32) >> 6;
}
}
- }
+ } while( coeff_count );
#ifdef CABAC_ON_STACK
h->cabac.range = cc.range ;
h->cabac.low = cc.low ;
if( (s->mb_y&1) == 0 )
h->mb_mbaff =
h->mb_field_decoding_flag = decode_cabac_field_decoding_flag(h);
- }else
- h->mb_field_decoding_flag= (s->picture_structure!=PICT_FRAME);
+ }
h->prev_mb_skipped = 0;
if(IS_INTRA_PCM(mb_type)) {
const uint8_t *ptr;
- unsigned int x, y;
// We assume these blocks are very rare so we do not optimize it.
// FIXME The two following lines get the bitstream position in the cabac
}
// The pixels are stored in the same order as levels in h->mb array.
- for(y=0; y<16; y++){
- const int index= 4*(y&3) + 32*((y>>2)&1) + 128*(y>>3);
- for(x=0; x<16; x++){
- tprintf(s->avctx, "LUMA ICPM LEVEL (%3d)\n", *ptr);
- h->mb[index + (x&3) + 16*((x>>2)&1) + 64*(x>>3)]= *ptr++;
- }
- }
- for(y=0; y<8; y++){
- const int index= 256 + 4*(y&3) + 32*(y>>2);
- for(x=0; x<8; x++){
- tprintf(s->avctx, "CHROMA U ICPM LEVEL (%3d)\n", *ptr);
- h->mb[index + (x&3) + 16*(x>>2)]= *ptr++;
- }
- }
- for(y=0; y<8; y++){
- const int index= 256 + 64 + 4*(y&3) + 32*(y>>2);
- for(x=0; x<8; x++){
- tprintf(s->avctx, "CHROMA V ICPM LEVEL (%3d)\n", *ptr);
- h->mb[index + (x&3) + 16*(x>>2)]= *ptr++;
- }
+ memcpy(h->mb, ptr, 256); ptr+=256;
+ if(CHROMA){
+ memcpy(h->mb+128, ptr, 128); ptr+=128;
}
ff_init_cabac_decoder(&h->cabac, ptr, h->cabac.bytestream_end - ptr);
h->intra16x16_pred_mode= check_intra_pred_mode( h, h->intra16x16_pred_mode );
if( h->intra16x16_pred_mode < 0 ) return -1;
}
- h->chroma_pred_mode_table[mb_xy] =
- pred_mode = decode_cabac_mb_chroma_pre_mode( h );
+ if(CHROMA){
+ h->chroma_pred_mode_table[mb_xy] =
+ pred_mode = decode_cabac_mb_chroma_pre_mode( h );
- pred_mode= check_intra_pred_mode( h, pred_mode );
- if( pred_mode < 0 ) return -1;
- h->chroma_pred_mode= pred_mode;
+ pred_mode= check_intra_pred_mode( h, pred_mode );
+ if( pred_mode < 0 ) return -1;
+ h->chroma_pred_mode= pred_mode;
+ }
} else if( partition_count == 4 ) {
int i, j, sub_partition_count[4], list, ref[2][4];
if( !IS_INTRA16x16( mb_type ) ) {
cbp = decode_cabac_mb_cbp_luma( h );
- cbp |= decode_cabac_mb_cbp_chroma( h ) << 4;
+ if(CHROMA)
+ cbp |= decode_cabac_mb_cbp_chroma( h ) << 4;
}
h->cbp_table[mb_xy] = h->cbp = cbp;
if (FRAME_MBAFF
// left mb is in picture
- && h->slice_table[mb_xy-1] != 255
+ && h->slice_table[mb_xy-1] != 0xFFFF
// and current and left pair do not have the same interlaced type
&& (IS_INTERLACED(mb_type) != IS_INTERLACED(s->current_picture.mb_type[mb_xy-1]))
// and left mb is in the same slice if deblocking_filter == 2
if( IS_INTRA( s->current_picture.mb_type[mbn_xy] ) )
bS[i] = 4;
else if( h->non_zero_count_cache[12+8*(i>>1)] != 0 ||
- /* FIXME: with 8x8dct + cavlc, should check cbp instead of nnz */
- h->non_zero_count[mbn_xy][MB_FIELD ? i&3 : (i>>2)+(mb_y&1)*2] )
+ ((!h->pps.cabac && IS_8x8DCT(s->current_picture.mb_type[mbn_xy])) ?
+ (h->cbp_table[mbn_xy] & ((MB_FIELD ? (i&2) : (mb_y&1)) ? 8 : 2))
+ :
+ h->non_zero_count[mbn_xy][MB_FIELD ? i&3 : (i>>2)+(mb_y&1)*2]))
bS[i] = 2;
else
bS[i] = 1;
int edge;
const int mbm_xy = dir == 0 ? mb_xy -1 : h->top_mb_xy;
const int mbm_type = s->current_picture.mb_type[mbm_xy];
- int (*ref2frm) [48+2] = h->ref2frm[ h->slice_num &15 ];
- int (*ref2frmm)[48+2] = h->ref2frm[ h->slice_table[mbm_xy]&15 ];
- int start = h->slice_table[mbm_xy] == 255 ? 1 : 0;
+ int (*ref2frm) [64] = h->ref2frm[ h->slice_num &(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
+ int (*ref2frmm)[64] = h->ref2frm[ h->slice_table[mbm_xy]&(MAX_SLICES-1) ][0] + (MB_MBAFF ? 20 : 2);
+ int start = h->slice_table[mbm_xy] == 0xFFFF ? 1 : 0;
const int edges = (mb_type & (MB_TYPE_16x16|MB_TYPE_SKIP))
== (MB_TYPE_16x16|MB_TYPE_SKIP) ? 1 : 4;
/* mbn_xy: neighbor macroblock */
const int mbn_xy = edge > 0 ? mb_xy : mbm_xy;
const int mbn_type = s->current_picture.mb_type[mbn_xy];
- int (*ref2frmn)[48+2] = edge > 0 ? ref2frm : ref2frmm;
+ int (*ref2frmn)[64] = edge > 0 ? ref2frm : ref2frmm;
int16_t bS[4];
int qp;
int v = 0;
for( l = 0; !v && l < 1 + (h->slice_type_nos == FF_B_TYPE); l++ ) {
- v |= ref2frm[l][h->ref_cache[l][b_idx]+2] != ref2frmn[l][h->ref_cache[l][bn_idx]+2] ||
+ v |= ref2frm[l][h->ref_cache[l][b_idx]] != ref2frmn[l][h->ref_cache[l][bn_idx]] ||
FFABS( h->mv_cache[l][b_idx][0] - h->mv_cache[l][bn_idx][0] ) >= 4 ||
FFABS( h->mv_cache[l][b_idx][1] - h->mv_cache[l][bn_idx][1] ) >= mvy_limit;
}
v=0;
for( l = 0; !v && l < 2; l++ ) {
int ln= 1-l;
- v |= ref2frm[l][h->ref_cache[l][b_idx]+2] != ref2frmn[ln][h->ref_cache[ln][bn_idx]+2] ||
+ v |= ref2frm[l][h->ref_cache[l][b_idx]] != ref2frmn[ln][h->ref_cache[ln][bn_idx]] ||
FFABS( h->mv_cache[l][b_idx][0] - h->mv_cache[ln][bn_idx][0] ) >= 4 ||
FFABS( h->mv_cache[l][b_idx][1] - h->mv_cache[ln][bn_idx][1] ) >= mvy_limit;
}
{
bS[i] = 0;
for( l = 0; l < 1 + (h->slice_type_nos == FF_B_TYPE); l++ ) {
- if( ref2frm[l][h->ref_cache[l][b_idx]+2] != ref2frmn[l][h->ref_cache[l][bn_idx]+2] ||
+ if( ref2frm[l][h->ref_cache[l][b_idx]] != ref2frmn[l][h->ref_cache[l][bn_idx]] ||
FFABS( h->mv_cache[l][b_idx][0] - h->mv_cache[l][bn_idx][0] ) >= 4 ||
FFABS( h->mv_cache[l][b_idx][1] - h->mv_cache[l][bn_idx][1] ) >= mvy_limit ) {
bS[i] = 1;
bS[i] = 0;
for( l = 0; l < 2; l++ ) {
int ln= 1-l;
- if( ref2frm[l][h->ref_cache[l][b_idx]+2] != ref2frmn[ln][h->ref_cache[ln][bn_idx]+2] ||
+ if( ref2frm[l][h->ref_cache[l][b_idx]] != ref2frmn[ln][h->ref_cache[ln][bn_idx]] ||
FFABS( h->mv_cache[l][b_idx][0] - h->mv_cache[ln][bn_idx][0] ) >= 4 ||
FFABS( h->mv_cache[l][b_idx][1] - h->mv_cache[ln][bn_idx][1] ) >= mvy_limit ) {
bS[i] = 1;
}
}
-static int decode_slice(struct AVCodecContext *avctx, H264Context *h){
+static int decode_slice(struct AVCodecContext *avctx, void *arg){
+ H264Context *h = *(void**)arg;
MpegEncContext * const s = &h->s;
const int part_mask= s->partitioned_frame ? (AC_END|AC_ERROR) : 0x7F;
return -1; //not reached
}
+static int decode_picture_timing(H264Context *h){
+ MpegEncContext * const s = &h->s;
+ if(h->sps.nal_hrd_parameters_present_flag || h->sps.vcl_hrd_parameters_present_flag){
+ skip_bits(&s->gb, h->sps.cpb_removal_delay_length); /* cpb_removal_delay */
+ skip_bits(&s->gb, h->sps.dpb_output_delay_length); /* dpb_output_delay */
+ }
+ if(h->sps.pic_struct_present_flag){
+ unsigned int i, num_clock_ts;
+ h->sei_pic_struct = get_bits(&s->gb, 4);
+
+ if (h->sei_pic_struct > SEI_PIC_STRUCT_FRAME_TRIPLING)
+ return -1;
+
+ num_clock_ts = sei_num_clock_ts_table[h->sei_pic_struct];
+
+ for (i = 0 ; i < num_clock_ts ; i++){
+ if(get_bits(&s->gb, 1)){ /* clock_timestamp_flag */
+ unsigned int full_timestamp_flag;
+ skip_bits(&s->gb, 2); /* ct_type */
+ skip_bits(&s->gb, 1); /* nuit_field_based_flag */
+ skip_bits(&s->gb, 5); /* counting_type */
+ full_timestamp_flag = get_bits(&s->gb, 1);
+ skip_bits(&s->gb, 1); /* discontinuity_flag */
+ skip_bits(&s->gb, 1); /* cnt_dropped_flag */
+ skip_bits(&s->gb, 8); /* n_frames */
+ if(full_timestamp_flag){
+ skip_bits(&s->gb, 6); /* seconds_value 0..59 */
+ skip_bits(&s->gb, 6); /* minutes_value 0..59 */
+ skip_bits(&s->gb, 5); /* hours_value 0..23 */
+ }else{
+ if(get_bits(&s->gb, 1)){ /* seconds_flag */
+ skip_bits(&s->gb, 6); /* seconds_value range 0..59 */
+ if(get_bits(&s->gb, 1)){ /* minutes_flag */
+ skip_bits(&s->gb, 6); /* minutes_value 0..59 */
+ if(get_bits(&s->gb, 1)) /* hours_flag */
+ skip_bits(&s->gb, 5); /* hours_value 0..23 */
+ }
+ }
+ }
+ if(h->sps.time_offset_length > 0)
+ skip_bits(&s->gb, h->sps.time_offset_length); /* time_offset */
+ }
+ }
+ }
+ return 0;
+}
+
static int decode_unregistered_user_data(H264Context *h, int size){
MpegEncContext * const s = &h->s;
uint8_t user_data[16+256];
}while(get_bits(&s->gb, 8) == 255);
switch(type){
+ case 1: // Picture timing SEI
+ if(decode_picture_timing(h) < 0)
+ return -1;
+ break;
case 5:
if(decode_unregistered_user_data(h, size) < 0)
return -1;
get_bits1(&s->gb); /* cbr_flag */
}
get_bits(&s->gb, 5); /* initial_cpb_removal_delay_length_minus1 */
- get_bits(&s->gb, 5); /* cpb_removal_delay_length_minus1 */
- get_bits(&s->gb, 5); /* dpb_output_delay_length_minus1 */
- get_bits(&s->gb, 5); /* time_offset_length */
+ sps->cpb_removal_delay_length = get_bits(&s->gb, 5) + 1;
+ sps->dpb_output_delay_length = get_bits(&s->gb, 5) + 1;
+ sps->time_offset_length = get_bits(&s->gb, 5);
}
static inline int decode_vui_parameters(H264Context *h, SPS *sps){
MpegEncContext * const s = &h->s;
int aspect_ratio_info_present_flag;
unsigned int aspect_ratio_idc;
- int nal_hrd_parameters_present_flag, vcl_hrd_parameters_present_flag;
aspect_ratio_info_present_flag= get_bits1(&s->gb);
if( aspect_ratio_idc == EXTENDED_SAR ) {
sps->sar.num= get_bits(&s->gb, 16);
sps->sar.den= get_bits(&s->gb, 16);
- }else if(aspect_ratio_idc < sizeof(pixel_aspect)/sizeof(*pixel_aspect)){
+ }else if(aspect_ratio_idc < FF_ARRAY_ELEMS(pixel_aspect)){
sps->sar= pixel_aspect[aspect_ratio_idc];
}else{
av_log(h->s.avctx, AV_LOG_ERROR, "illegal aspect ratio\n");
sps->fixed_frame_rate_flag = get_bits1(&s->gb);
}
- nal_hrd_parameters_present_flag = get_bits1(&s->gb);
- if(nal_hrd_parameters_present_flag)
+ sps->nal_hrd_parameters_present_flag = get_bits1(&s->gb);
+ if(sps->nal_hrd_parameters_present_flag)
decode_hrd_parameters(h, sps);
- vcl_hrd_parameters_present_flag = get_bits1(&s->gb);
- if(vcl_hrd_parameters_present_flag)
+ sps->vcl_hrd_parameters_present_flag = get_bits1(&s->gb);
+ if(sps->vcl_hrd_parameters_present_flag)
decode_hrd_parameters(h, sps);
- if(nal_hrd_parameters_present_flag || vcl_hrd_parameters_present_flag)
+ if(sps->nal_hrd_parameters_present_flag || sps->vcl_hrd_parameters_present_flag)
get_bits1(&s->gb); /* low_delay_hrd_flag */
- get_bits1(&s->gb); /* pic_struct_present_flag */
+ sps->pic_struct_present_flag = get_bits1(&s->gb);
sps->bitstream_restriction_flag = get_bits1(&s->gb);
if(sps->bitstream_restriction_flag){
- unsigned int num_reorder_frames;
get_bits1(&s->gb); /* motion_vectors_over_pic_boundaries_flag */
get_ue_golomb(&s->gb); /* max_bytes_per_pic_denom */
get_ue_golomb(&s->gb); /* max_bits_per_mb_denom */
get_ue_golomb(&s->gb); /* log2_max_mv_length_horizontal */
get_ue_golomb(&s->gb); /* log2_max_mv_length_vertical */
- num_reorder_frames= get_ue_golomb(&s->gb);
+ sps->num_reorder_frames= get_ue_golomb(&s->gb);
get_ue_golomb(&s->gb); /*max_dec_frame_buffering*/
- if(num_reorder_frames > 16 /*max_dec_frame_buffering || max_dec_frame_buffering > 16*/){
- av_log(h->s.avctx, AV_LOG_ERROR, "illegal num_reorder_frames %d\n", num_reorder_frames);
+ if(sps->num_reorder_frames > 16U /*max_dec_frame_buffering || max_dec_frame_buffering > 16*/){
+ av_log(h->s.avctx, AV_LOG_ERROR, "illegal num_reorder_frames %d\n", sps->num_reorder_frames);
return -1;
}
-
- sps->num_reorder_frames= num_reorder_frames;
}
return 0;
decode_scaling_list(h,scaling_matrix8[0],64,default_scaling8[0],fallback[2]); // Intra, Y
decode_scaling_list(h,scaling_matrix8[1],64,default_scaling8[1],fallback[3]); // Inter, Y
}
- } else if(fallback_sps) {
- memcpy(scaling_matrix4, sps->scaling_matrix4, 6*16*sizeof(uint8_t));
- memcpy(scaling_matrix8, sps->scaling_matrix8, 2*64*sizeof(uint8_t));
- }
-}
-
-/**
- * Returns and optionally allocates SPS / PPS structures in the supplied array 'vec'
- */
-static void *
-alloc_parameter_set(H264Context *h, void **vec, const unsigned int id, const unsigned int max,
- const size_t size, const char *name)
-{
- if(id>=max) {
- av_log(h->s.avctx, AV_LOG_ERROR, "%s_id (%d) out of range\n", name, id);
- return NULL;
}
-
- if(!vec[id]) {
- vec[id] = av_mallocz(size);
- if(vec[id] == NULL)
- av_log(h->s.avctx, AV_LOG_ERROR, "cannot allocate memory for %s\n", name);
- }
- return vec[id];
}
static inline int decode_seq_parameter_set(H264Context *h){
MpegEncContext * const s = &h->s;
int profile_idc, level_idc;
- unsigned int sps_id, tmp, mb_width, mb_height;
+ unsigned int sps_id;
int i;
SPS *sps;
level_idc= get_bits(&s->gb, 8);
sps_id= get_ue_golomb(&s->gb);
- sps = alloc_parameter_set(h, (void **)h->sps_buffers, sps_id, MAX_SPS_COUNT, sizeof(SPS), "sps");
+ if(sps_id >= MAX_SPS_COUNT) {
+ av_log(h->s.avctx, AV_LOG_ERROR, "sps_id (%d) out of range\n", sps_id);
+ return -1;
+ }
+ sps= av_mallocz(sizeof(SPS));
if(sps == NULL)
return -1;
sps->profile_idc= profile_idc;
sps->level_idc= level_idc;
+ memset(sps->scaling_matrix4, 16, sizeof(sps->scaling_matrix4));
+ memset(sps->scaling_matrix8, 16, sizeof(sps->scaling_matrix8));
+ sps->scaling_matrix_present = 0;
+
if(sps->profile_idc >= 100){ //high profile
- if(get_ue_golomb(&s->gb) == 3) //chroma_format_idc
+ sps->chroma_format_idc= get_ue_golomb(&s->gb);
+ if(sps->chroma_format_idc == 3)
get_bits1(&s->gb); //residual_color_transform_flag
get_ue_golomb(&s->gb); //bit_depth_luma_minus8
get_ue_golomb(&s->gb); //bit_depth_chroma_minus8
sps->transform_bypass = get_bits1(&s->gb);
decode_scaling_matrices(h, sps, NULL, 1, sps->scaling_matrix4, sps->scaling_matrix8);
- }else
- sps->scaling_matrix_present = 0;
+ }else{
+ sps->chroma_format_idc= 1;
+ }
sps->log2_max_frame_num= get_ue_golomb(&s->gb) + 4;
sps->poc_type= get_ue_golomb(&s->gb);
sps->delta_pic_order_always_zero_flag= get_bits1(&s->gb);
sps->offset_for_non_ref_pic= get_se_golomb(&s->gb);
sps->offset_for_top_to_bottom_field= get_se_golomb(&s->gb);
- tmp= get_ue_golomb(&s->gb);
+ sps->poc_cycle_length = get_ue_golomb(&s->gb);
- if(tmp >= sizeof(sps->offset_for_ref_frame) / sizeof(sps->offset_for_ref_frame[0])){
- av_log(h->s.avctx, AV_LOG_ERROR, "poc_cycle_length overflow %u\n", tmp);
- return -1;
+ if((unsigned)sps->poc_cycle_length >= FF_ARRAY_ELEMS(sps->offset_for_ref_frame)){
+ av_log(h->s.avctx, AV_LOG_ERROR, "poc_cycle_length overflow %u\n", sps->poc_cycle_length);
+ goto fail;
}
- sps->poc_cycle_length= tmp;
for(i=0; i<sps->poc_cycle_length; i++)
sps->offset_for_ref_frame[i]= get_se_golomb(&s->gb);
}else if(sps->poc_type != 2){
av_log(h->s.avctx, AV_LOG_ERROR, "illegal POC type %d\n", sps->poc_type);
- return -1;
+ goto fail;
}
- tmp= get_ue_golomb(&s->gb);
- if(tmp > MAX_PICTURE_COUNT-2 || tmp >= 32){
+ sps->ref_frame_count= get_ue_golomb(&s->gb);
+ if(sps->ref_frame_count > MAX_PICTURE_COUNT-2 || sps->ref_frame_count >= 32U){
av_log(h->s.avctx, AV_LOG_ERROR, "too many reference frames\n");
- return -1;
+ goto fail;
}
- sps->ref_frame_count= tmp;
sps->gaps_in_frame_num_allowed_flag= get_bits1(&s->gb);
- mb_width= get_ue_golomb(&s->gb) + 1;
- mb_height= get_ue_golomb(&s->gb) + 1;
- if(mb_width >= INT_MAX/16 || mb_height >= INT_MAX/16 ||
- avcodec_check_dimensions(NULL, 16*mb_width, 16*mb_height)){
+ sps->mb_width = get_ue_golomb(&s->gb) + 1;
+ sps->mb_height= get_ue_golomb(&s->gb) + 1;
+ if((unsigned)sps->mb_width >= INT_MAX/16 || (unsigned)sps->mb_height >= INT_MAX/16 ||
+ avcodec_check_dimensions(NULL, 16*sps->mb_width, 16*sps->mb_height)){
av_log(h->s.avctx, AV_LOG_ERROR, "mb_width/height overflow\n");
- return -1;
+ goto fail;
}
- sps->mb_width = mb_width;
- sps->mb_height= mb_height;
sps->frame_mbs_only_flag= get_bits1(&s->gb);
if(!sps->frame_mbs_only_flag)
if(sps->mb_aff)
av_log(h->s.avctx, AV_LOG_ERROR, "MBAFF support not included; enable it at compile-time.\n");
#endif
- if(!sps->direct_8x8_inference_flag && sps->mb_aff)
- av_log(h->s.avctx, AV_LOG_ERROR, "MBAFF + !direct_8x8_inference is not implemented\n");
-
sps->crop= get_bits1(&s->gb);
if(sps->crop){
sps->crop_left = get_ue_golomb(&s->gb);
if(sps->crop_left || sps->crop_top){
av_log(h->s.avctx, AV_LOG_ERROR, "insane cropping not completely supported, this could look slightly wrong ...\n");
}
- if(sps->crop_right >= 8 || sps->crop_bottom >= (8>> !h->sps.frame_mbs_only_flag)){
+ if(sps->crop_right >= 8 || sps->crop_bottom >= (8>> !sps->frame_mbs_only_flag)){
av_log(h->s.avctx, AV_LOG_ERROR, "brainfart cropping not supported, this could look slightly wrong ...\n");
}
}else{
decode_vui_parameters(h, sps);
if(s->avctx->debug&FF_DEBUG_PICT_INFO){
- av_log(h->s.avctx, AV_LOG_DEBUG, "sps:%u profile:%d/%d poc:%d ref:%d %dx%d %s %s crop:%d/%d/%d/%d %s\n",
+ av_log(h->s.avctx, AV_LOG_DEBUG, "sps:%u profile:%d/%d poc:%d ref:%d %dx%d %s %s crop:%d/%d/%d/%d %s %s\n",
sps_id, sps->profile_idc, sps->level_idc,
sps->poc_type,
sps->ref_frame_count,
sps->direct_8x8_inference_flag ? "8B8" : "",
sps->crop_left, sps->crop_right,
sps->crop_top, sps->crop_bottom,
- sps->vui_parameters_present_flag ? "VUI" : ""
+ sps->vui_parameters_present_flag ? "VUI" : "",
+ ((const char*[]){"Gray","420","422","444"})[sps->chroma_format_idc]
);
}
+ av_free(h->sps_buffers[sps_id]);
+ h->sps_buffers[sps_id]= sps;
return 0;
+fail:
+ av_free(sps);
+ return -1;
}
static void
static inline int decode_picture_parameter_set(H264Context *h, int bit_length){
MpegEncContext * const s = &h->s;
- unsigned int tmp, pps_id= get_ue_golomb(&s->gb);
+ unsigned int pps_id= get_ue_golomb(&s->gb);
PPS *pps;
- pps = alloc_parameter_set(h, (void **)h->pps_buffers, pps_id, MAX_PPS_COUNT, sizeof(PPS), "pps");
- if(pps == NULL)
+ if(pps_id >= MAX_PPS_COUNT) {
+ av_log(h->s.avctx, AV_LOG_ERROR, "pps_id (%d) out of range\n", pps_id);
return -1;
+ }
- tmp= get_ue_golomb(&s->gb);
- if(tmp>=MAX_SPS_COUNT || h->sps_buffers[tmp] == NULL){
- av_log(h->s.avctx, AV_LOG_ERROR, "sps_id out of range\n");
+ pps= av_mallocz(sizeof(PPS));
+ if(pps == NULL)
return -1;
+ pps->sps_id= get_ue_golomb(&s->gb);
+ if((unsigned)pps->sps_id>=MAX_SPS_COUNT || h->sps_buffers[pps->sps_id] == NULL){
+ av_log(h->s.avctx, AV_LOG_ERROR, "sps_id out of range\n");
+ goto fail;
}
- pps->sps_id= tmp;
pps->cabac= get_bits1(&s->gb);
pps->pic_order_present= get_bits1(&s->gb);
pps->ref_count[1]= get_ue_golomb(&s->gb) + 1;
if(pps->ref_count[0]-1 > 32-1 || pps->ref_count[1]-1 > 32-1){
av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow (pps)\n");
- pps->ref_count[0]= pps->ref_count[1]= 1;
- return -1;
+ goto fail;
}
pps->weighted_pred= get_bits1(&s->gb);
pps->transform_8x8_mode= 0;
h->dequant_coeff_pps= -1; //contents of sps/pps can change even if id doesn't, so reinit
- memset(pps->scaling_matrix4, 16, 6*16*sizeof(uint8_t));
- memset(pps->scaling_matrix8, 16, 2*64*sizeof(uint8_t));
+ memcpy(pps->scaling_matrix4, h->sps_buffers[pps->sps_id]->scaling_matrix4, sizeof(pps->scaling_matrix4));
+ memcpy(pps->scaling_matrix8, h->sps_buffers[pps->sps_id]->scaling_matrix8, sizeof(pps->scaling_matrix8));
if(get_bits_count(&s->gb) < bit_length){
pps->transform_8x8_mode= get_bits1(&s->gb);
);
}
+ av_free(h->pps_buffers[pps_id]);
+ h->pps_buffers[pps_id]= pps;
return 0;
+fail:
+ av_free(pps);
+ return -1;
}
/**
int i;
if(context_count == 1) {
- decode_slice(avctx, h);
+ decode_slice(avctx, &h);
} else {
for(i = 1; i < context_count; i++) {
hx = h->thread_context[i];
- hx->s.error_resilience = avctx->error_resilience;
+ hx->s.error_recognition = avctx->error_recognition;
hx->s.error_count = 0;
}
avctx->execute(avctx, (void *)decode_slice,
- (void **)h->thread_context, NULL, context_count);
+ (void **)h->thread_context, NULL, context_count, sizeof(void*));
/* pull back stuff from slices to master context */
hx = h->thread_context[context_count - 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]->poc; i++)
+ for(i=1; h->delayed_pic[i] && (h->delayed_pic[i]->poc && !h->delayed_pic[i]->key_frame); i++)
if(h->delayed_pic[i]->poc < out->poc){
out = h->delayed_pic[i];
out_idx = i;
h->got_avcC = 1;
}
- if(avctx->frame_number==0 && !h->is_avc && s->avctx->extradata_size){
+ if(!h->got_avcC && !h->is_avc && s->avctx->extradata_size){
if(decode_nal_units(h, s->avctx->extradata, s->avctx->extradata_size) < 0)
return -1;
+ h->got_avcC = 1;
}
buf_index=decode_nal_units(h, buf, buf_size);
*data_size = 0;
} else {
- cur->interlaced_frame = FIELD_OR_MBAFF_PICTURE;
- /* Derive top_field_first from field pocs. */
- cur->top_field_first = cur->field_poc[0] < cur->field_poc[1];
+ cur->repeat_pict = 0;
+
+ /* Signal interlacing information externally. */
+ /* Prioritize picture timing SEI information over used decoding process if it exists. */
+ if(h->sps.pic_struct_present_flag){
+ switch (h->sei_pic_struct)
+ {
+ case SEI_PIC_STRUCT_FRAME:
+ cur->interlaced_frame = 0;
+ break;
+ case SEI_PIC_STRUCT_TOP_FIELD:
+ case SEI_PIC_STRUCT_BOTTOM_FIELD:
+ case SEI_PIC_STRUCT_TOP_BOTTOM:
+ case SEI_PIC_STRUCT_BOTTOM_TOP:
+ cur->interlaced_frame = 1;
+ 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->interlaced_frame = FIELD_OR_MBAFF_PICTURE;
+ break;
+ case SEI_PIC_STRUCT_FRAME_DOUBLING:
+ // Force progressive here, as doubling interlaced frame is a bad idea.
+ cur->interlaced_frame = 0;
+ cur->repeat_pict = 2;
+ break;
+ case SEI_PIC_STRUCT_FRAME_TRIPLING:
+ cur->interlaced_frame = 0;
+ cur->repeat_pict = 4;
+ break;
+ }
+ }else{
+ /* Derive interlacing flag from used decoding process. */
+ cur->interlaced_frame = FIELD_OR_MBAFF_PICTURE;
+ }
+
+ 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];
+ }else{
+ if(cur->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;
+ else
+ cur->top_field_first = 0;
+ }else{
+ /* Most likely progressive */
+ cur->top_field_first = 0;
+ }
+ }
//FIXME do something with unavailable reference frames
out = h->delayed_pic[0];
out_idx = 0;
- for(i=1; h->delayed_pic[i] && h->delayed_pic[i]->poc; i++)
+ for(i=1; h->delayed_pic[i] && (h->delayed_pic[i]->poc && !h->delayed_pic[i]->key_frame); i++)
if(h->delayed_pic[i]->poc < out->poc){
out = h->delayed_pic[i];
out_idx = i;
}
- cross_idr = !h->delayed_pic[0]->poc || !!h->delayed_pic[i];
+ cross_idr = !h->delayed_pic[0]->poc || !!h->delayed_pic[i] || h->delayed_pic[0]->key_frame;
out_of_order = !cross_idr && out->poc < h->outputed_poc;
{
H264Context *h = avctx->priv_data;
MpegEncContext *s = &h->s;
+ int i;
av_freep(&h->rbsp_buffer[0]);
av_freep(&h->rbsp_buffer[1]);
free_tables(h); //FIXME cleanup init stuff perhaps
+
+ for(i = 0; i < MAX_SPS_COUNT; i++)
+ av_freep(h->sps_buffers + i);
+
+ for(i = 0; i < MAX_PPS_COUNT; i++)
+ av_freep(h->pps_buffers + i);
+
MPV_common_end(s);
// memset(h, 0, sizeof(H264Context));
projects / ffmpeg / blobdiff