*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
- * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*
*/
-
+
/**
* @file h264.c
* H.264 / AVC / MPEG4 part10 codec.
* Sequence parameter set
*/
typedef struct SPS{
-
+
int profile_idc;
int level_idc;
+ int transform_bypass; ///< qpprime_y_zero_transform_bypass_flag
int log2_max_frame_num; ///< log2_max_frame_num_minus4 + 4
int poc_type; ///< pic_order_cnt_type
int log2_max_poc_lsb; ///< log2_max_pic_order_cnt_lsb_minus4
uint32_t time_scale;
int fixed_frame_rate_flag;
short offset_for_ref_frame[256]; //FIXME dyn aloc?
+ int bitstream_restriction_flag;
+ int num_reorder_frames;
+ int scaling_matrix_present;
+ uint8_t scaling_matrix4[6][16];
+ uint8_t scaling_matrix8[2][64];
}SPS;
/**
int deblocking_filter_parameters_present; ///< deblocking_filter_parameters_present_flag
int constrained_intra_pred; ///< constrained_intra_pred_flag
int redundant_pic_cnt_present; ///< redundant_pic_cnt_present_flag
+ int transform_8x8_mode; ///< transform_8x8_mode_flag
+ uint8_t scaling_matrix4[6][16];
+ uint8_t scaling_matrix8[2][64];
}PPS;
/**
MMCO_LONG2UNUSED,
MMCO_SHORT2LONG,
MMCO_SET_MAX_LONG,
- MMCO_RESET,
+ MMCO_RESET,
MMCO_LONG,
} MMCOOpcode;
*/
typedef struct H264Context{
MpegEncContext s;
- int nal_ref_idc;
+ int nal_ref_idc;
int nal_unit_type;
-#define NAL_SLICE 1
-#define NAL_DPA 2
-#define NAL_DPB 3
-#define NAL_DPC 4
-#define NAL_IDR_SLICE 5
-#define NAL_SEI 6
-#define NAL_SPS 7
-#define NAL_PPS 8
-#define NAL_PICTURE_DELIMITER 9
-#define NAL_FILTER_DATA 10
+#define NAL_SLICE 1
+#define NAL_DPA 2
+#define NAL_DPB 3
+#define NAL_DPC 4
+#define NAL_IDR_SLICE 5
+#define NAL_SEI 6
+#define NAL_SPS 7
+#define NAL_PPS 8
+#define NAL_AUD 9
+#define NAL_END_SEQUENCE 10
+#define NAL_END_STREAM 11
+#define NAL_FILLER_DATA 12
+#define NAL_SPS_EXT 13
+#define NAL_AUXILIARY_SLICE 19
uint8_t *rbsp_buffer;
int rbsp_buffer_size;
int chroma_qp; //QPc
- int prev_mb_skiped; //FIXME remove (IMHO not used)
+ int prev_mb_skipped; //FIXME remove (IMHO not used)
//prediction stuff
int chroma_pred_mode;
int intra16x16_pred_mode;
-
+
+ int top_mb_xy;
+ int left_mb_xy[2];
+
int8_t intra4x4_pred_mode_cache[5*8];
int8_t (*intra4x4_pred_mode)[8];
void (*pred4x4 [9+3])(uint8_t *src, uint8_t *topright, int stride);//FIXME move to dsp?
+ void (*pred8x8l [9+3])(uint8_t *src, int topleft, int topright, int stride);
void (*pred8x8 [4+3])(uint8_t *src, int stride);
void (*pred16x16[4+3])(uint8_t *src, int stride);
unsigned int topleft_samples_available;
unsigned int top_samples_available;
unsigned int topright_samples_available;
unsigned int left_samples_available;
- uint8_t (*top_border)[16+2*8];
- uint8_t left_border[17+2*9];
+ uint8_t (*top_borders[2])[16+2*8];
+ uint8_t left_border[2*(17+2*9)];
/**
* non zero coeff count cache.
int8_t ref_cache[2][5*8] __align8;
#define LIST_NOT_USED -1 //FIXME rename?
#define PART_NOT_AVAILABLE -2
-
+
/**
* is 1 if the specific list MV&references are set to 0,0,-2.
*/
int mv_cache_clean[2];
- int block_offset[16+8];
- int chroma_subblock_offset[16]; //FIXME remove
-
- uint16_t *mb2b_xy; //FIXME are these 4 a good idea?
- uint16_t *mb2b8_xy;
+ /**
+ * number of neighbors (top and/or left) that used 8x8 dct
+ */
+ int neighbor_transform_size;
+
+ /**
+ * block_offset[ 0..23] for frame macroblocks
+ * block_offset[24..47] for field macroblocks
+ */
+ int block_offset[2*(16+8)];
+
+ uint32_t *mb2b_xy; //FIXME are these 4 a good idea?
+ uint32_t *mb2b8_xy;
int b_stride; //FIXME use s->b4_stride
int b8_stride;
SPS sps_buffer[MAX_SPS_COUNT];
SPS sps; ///< current sps
-
+
PPS pps_buffer[MAX_PPS_COUNT];
/**
* current pps
*/
- PPS pps; //FIXME move tp Picture perhaps? (->no) do we need that?
+ PPS pps; //FIXME move to Picture perhaps? (->no) do we need that?
+
+ uint32_t dequant4_buffer[6][52][16];
+ uint32_t dequant8_buffer[2][52][64];
+ uint32_t (*dequant4_coeff[6])[16];
+ uint32_t (*dequant8_coeff[2])[64];
+ int dequant_coeff_pps; ///< reinit tables when pps changes
int slice_num;
uint8_t *slice_table_base;
uint8_t *slice_table; ///< slice_table_base + mb_stride + 1
int slice_type;
int slice_type_fixed;
-
+
//interlacing specific flags
+ int mb_aff_frame;
int mb_field_decoding_flag;
-
+
int sub_mb_type[4];
-
+
//POC stuff
int poc_lsb;
int poc_msb;
* frame_num for frames or 2*frame_num for field pics.
*/
int curr_pic_num;
-
+
/**
* max_frame_num or 2*max_frame_num for field pics.
*/
int chroma_weight[2][16][2];
int chroma_offset[2][16][2];
int implicit_weight[16][16];
-
+
//deblock
- int deblocking_filter; ///< disable_deblocking_filter_idc with 1<->0
+ int deblocking_filter; ///< disable_deblocking_filter_idc with 1<->0
int slice_alpha_c0_offset;
int slice_beta_offset;
-
+
int redundant_pic_count;
-
+
int direct_spatial_mv_pred;
int dist_scale_factor[16];
+ int map_col_to_list0[2][16];
/**
* num_ref_idx_l0/1_active_minus1 + 1
*/
int ref_count[2];// FIXME split for AFF
- Picture *short_ref[16];
- Picture *long_ref[16];
+ Picture *short_ref[32];
+ Picture *long_ref[32];
Picture default_ref_list[2][32];
Picture ref_list[2][32]; //FIXME size?
Picture field_ref_list[2][32]; //FIXME size?
Picture *delayed_pic[16]; //FIXME size?
-
+ Picture *delayed_output_pic;
+
/**
* memory management control operations buffer.
*/
MMCO mmco[MAX_MMCO_COUNT];
int mmco_index;
-
+
int long_ref_count; ///< number of actual long term references
int short_ref_count; ///< number of actual short term references
-
+
//data partitioning
GetBitContext intra_gb;
GetBitContext inter_gb;
GetBitContext *intra_gb_ptr;
GetBitContext *inter_gb_ptr;
-
+
DCTELEM mb[16*24] __align8;
/**
* Cabac
*/
CABACContext cabac;
- uint8_t cabac_state[399];
+ uint8_t cabac_state[460];
int cabac_init_idc;
/* 0x100 -> non null luma_dc, 0x80/0x40 -> non null chroma_dc (cb/cr), 0x?0 -> chroma_cbp(0,1,2), 0x0? luma_cbp */
uint8_t *direct_table;
uint8_t direct_cache[5*8];
+ uint8_t zigzag_scan[16];
+ uint8_t field_scan[16];
+ const uint8_t *zigzag_scan_q0;
+ const uint8_t *field_scan_q0;
+
+ int x264_build;
}H264Context;
static VLC coeff_token_vlc[4];
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);
+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 inline uint32_t pack16to32(int a, int b){
+static always_inline uint32_t pack16to32(int a, int b){
#ifdef WORDS_BIGENDIAN
return (b&0xFFFF) + (a<<16);
#else
* @param w width of the rectangle, should be a constant
* @param size the size of val (1 or 4), should be a constant
*/
-static inline void fill_rectangle(void *vp, int w, int h, int stride, uint32_t val, int size){ //FIXME ensure this IS inlined
+static always_inline void fill_rectangle(void *vp, int w, int h, int stride, uint32_t val, int size){
uint8_t *p= (uint8_t*)vp;
assert(size==1 || size==4);
-
+
w *= size;
stride *= size;
-
- assert((((int)vp)&(FFMIN(w, STRIDE_ALIGN)-1)) == 0);
-//FIXME check what gcc generates for 64 bit on x86 and possible write a 32 bit ver of it
+
+ assert((((long)vp)&(FFMIN(w, STRIDE_ALIGN)-1)) == 0);
+ assert((stride&(w-1))==0);
+//FIXME check what gcc generates for 64 bit on x86 and possibly write a 32 bit ver of it
if(w==2 && h==2){
*(uint16_t*)(p + 0)=
*(uint16_t*)(p + stride)= size==4 ? val : val*0x0101;
assert(0);
}
-static inline void fill_caches(H264Context *h, int mb_type, int for_deblock){
+static void fill_caches(H264Context *h, int mb_type, int for_deblock){
MpegEncContext * const s = &h->s;
const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
int topleft_xy, top_xy, topright_xy, left_xy[2];
int topleft_type, top_type, topright_type, left_type[2];
- int left_block[4];
+ int left_block[8];
int i;
- //wow what a mess, why didnt they simplify the interlacing&intra stuff, i cant imagine that these complex rules are worth it
-
- if(h->sps.mb_aff){
- //FIXME
- topleft_xy = 0; /* avoid warning */
- top_xy = 0; /* avoid warning */
- topright_xy = 0; /* avoid warning */
- }else{
- topleft_xy = mb_xy-1 - s->mb_stride;
- top_xy = mb_xy - s->mb_stride;
- topright_xy= mb_xy+1 - s->mb_stride;
- left_xy[0] = mb_xy-1;
- left_xy[1] = mb_xy-1;
- left_block[0]= 0;
- left_block[1]= 1;
- left_block[2]= 2;
- left_block[3]= 3;
+ //FIXME deblocking can skip fill_caches much of the time with multiple slices too.
+ // the actual condition is whether we're on the edge of a slice,
+ // and even then the intra and nnz parts are unnecessary.
+ if(for_deblock && h->slice_num == 1)
+ return;
+
+ //wow what a mess, why didn't they simplify the interlacing&intra stuff, i can't imagine that these complex rules are worth it
+
+ top_xy = mb_xy - s->mb_stride;
+ 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;
+ if(h->mb_aff_frame){
+ const int pair_xy = s->mb_x + (s->mb_y & ~1)*s->mb_stride;
+ const int top_pair_xy = pair_xy - s->mb_stride;
+ const int topleft_pair_xy = top_pair_xy - 1;
+ const int topright_pair_xy = top_pair_xy + 1;
+ const int topleft_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[topleft_pair_xy]);
+ const int top_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[top_pair_xy]);
+ const int topright_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[topright_pair_xy]);
+ const int left_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[pair_xy-1]);
+ const int curr_mb_frame_flag = !IS_INTERLACED(mb_type);
+ const int bottom = (s->mb_y & 1);
+ tprintf("fill_caches: curr_mb_frame_flag:%d, left_mb_frame_flag:%d, topleft_mb_frame_flag:%d, top_mb_frame_flag:%d, topright_mb_frame_flag:%d\n", curr_mb_frame_flag, left_mb_frame_flag, topleft_mb_frame_flag, top_mb_frame_flag, topright_mb_frame_flag);
+ if (bottom
+ ? !curr_mb_frame_flag // bottom macroblock
+ : (!curr_mb_frame_flag && !top_mb_frame_flag) // top macroblock
+ ) {
+ top_xy -= s->mb_stride;
+ }
+ if (bottom
+ ? !curr_mb_frame_flag // bottom macroblock
+ : (!curr_mb_frame_flag && !topleft_mb_frame_flag) // top macroblock
+ ) {
+ topleft_xy -= s->mb_stride;
+ }
+ if (bottom
+ ? !curr_mb_frame_flag // bottom macroblock
+ : (!curr_mb_frame_flag && !topright_mb_frame_flag) // top macroblock
+ ) {
+ topright_xy -= s->mb_stride;
+ }
+ if (left_mb_frame_flag != curr_mb_frame_flag) {
+ 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;
+ } 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;
+ }
+ } 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;
+ }
+ }
}
+ h->top_mb_xy = top_xy;
+ h->left_mb_xy[0] = left_xy[0];
+ h->left_mb_xy[1] = left_xy[1];
if(for_deblock){
topleft_type = h->slice_table[topleft_xy ] < 255 ? s->current_picture.mb_type[topleft_xy] : 0;
top_type = h->slice_table[top_xy ] < 255 ? s->current_picture.mb_type[top_xy] : 0;
}
if(IS_INTRA(mb_type)){
- h->topleft_samples_available=
- h->top_samples_available=
+ h->topleft_samples_available=
+ h->top_samples_available=
h->left_samples_available= 0xFFFF;
h->topright_samples_available= 0xEEEA;
h->left_samples_available&= 0x5F5F;
}
}
-
+
if(!IS_INTRA(topleft_type) && (topleft_type==0 || h->pps.constrained_intra_pred))
h->topleft_samples_available&= 0x7FFF;
-
+
if(!IS_INTRA(topright_type) && (topright_type==0 || h->pps.constrained_intra_pred))
h->topright_samples_available&= 0xFBFF;
-
+
if(IS_INTRA4x4(mb_type)){
if(IS_INTRA4x4(top_type)){
h->intra4x4_pred_mode_cache[4+8*0]= h->intra4x4_pred_mode[top_xy][4];
}
}
}
-
-
+
+
/*
-0 . T T. T T T T
-1 L . .L . . . .
-2 L . .L . . . .
-3 . T TL . . . .
-4 L . .L . . . .
-5 L . .. . . . .
+0 . T T. T T T T
+1 L . .L . . . .
+2 L . .L . . . .
+3 . T TL . . . .
+4 L . .L . . . .
+5 L . .. . . . .
*/
//FIXME constraint_intra_pred & partitioning & nnz (lets hope this is just a typo in the spec)
if(top_type){
- h->non_zero_count_cache[4+8*0]= h->non_zero_count[top_xy][0];
- h->non_zero_count_cache[5+8*0]= h->non_zero_count[top_xy][1];
- h->non_zero_count_cache[6+8*0]= h->non_zero_count[top_xy][2];
+ h->non_zero_count_cache[4+8*0]= h->non_zero_count[top_xy][4];
+ h->non_zero_count_cache[5+8*0]= h->non_zero_count[top_xy][5];
+ h->non_zero_count_cache[6+8*0]= h->non_zero_count[top_xy][6];
h->non_zero_count_cache[7+8*0]= h->non_zero_count[top_xy][3];
-
- h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][7];
+
+ h->non_zero_count_cache[1+8*0]= h->non_zero_count[top_xy][9];
h->non_zero_count_cache[2+8*0]= h->non_zero_count[top_xy][8];
-
- h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][10];
+
+ h->non_zero_count_cache[1+8*3]= h->non_zero_count[top_xy][12];
h->non_zero_count_cache[2+8*3]= h->non_zero_count[top_xy][11];
-
- h->top_cbp= h->cbp_table[top_xy];
+
}else{
- h->non_zero_count_cache[4+8*0]=
+ h->non_zero_count_cache[4+8*0]=
h->non_zero_count_cache[5+8*0]=
h->non_zero_count_cache[6+8*0]=
h->non_zero_count_cache[7+8*0]=
-
+
h->non_zero_count_cache[1+8*0]=
h->non_zero_count_cache[2+8*0]=
-
+
h->non_zero_count_cache[1+8*3]=
h->non_zero_count_cache[2+8*3]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
-
- if(IS_INTRA(mb_type)) h->top_cbp= 0x1C0;
- else h->top_cbp= 0;
- }
-
- if(left_type[0]){
- h->non_zero_count_cache[3+8*1]= h->non_zero_count[left_xy[0]][6];
- h->non_zero_count_cache[3+8*2]= h->non_zero_count[left_xy[0]][5];
- h->non_zero_count_cache[0+8*1]= h->non_zero_count[left_xy[0]][9]; //FIXME left_block
- h->non_zero_count_cache[0+8*4]= h->non_zero_count[left_xy[0]][12];
- h->left_cbp= h->cbp_table[left_xy[0]]; //FIXME interlacing
- }else{
- h->non_zero_count_cache[3+8*1]=
- h->non_zero_count_cache[3+8*2]=
- h->non_zero_count_cache[0+8*1]=
- h->non_zero_count_cache[0+8*4]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
-
- if(IS_INTRA(mb_type)) h->left_cbp= 0x1C0;//FIXME interlacing
- else h->left_cbp= 0;
- }
-
- if(left_type[1]){
- h->non_zero_count_cache[3+8*3]= h->non_zero_count[left_xy[1]][4];
- h->non_zero_count_cache[3+8*4]= h->non_zero_count[left_xy[1]][3];
- h->non_zero_count_cache[0+8*2]= h->non_zero_count[left_xy[1]][8];
- h->non_zero_count_cache[0+8*5]= h->non_zero_count[left_xy[1]][11];
- }else{
- h->non_zero_count_cache[3+8*3]=
- h->non_zero_count_cache[3+8*4]=
- h->non_zero_count_cache[0+8*2]=
- h->non_zero_count_cache[0+8*5]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
+
}
-
+
+ for (i=0; i<2; i++) {
+ if(left_type[i]){
+ h->non_zero_count_cache[3+8*1 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[0+2*i]];
+ h->non_zero_count_cache[3+8*2 + 2*8*i]= h->non_zero_count[left_xy[i]][left_block[1+2*i]];
+ h->non_zero_count_cache[0+8*1 + 8*i]= h->non_zero_count[left_xy[i]][left_block[4+2*i]];
+ h->non_zero_count_cache[0+8*4 + 8*i]= h->non_zero_count[left_xy[i]][left_block[5+2*i]];
+ }else{
+ h->non_zero_count_cache[3+8*1 + 2*8*i]=
+ h->non_zero_count_cache[3+8*2 + 2*8*i]=
+ h->non_zero_count_cache[0+8*1 + 8*i]=
+ h->non_zero_count_cache[0+8*4 + 8*i]= h->pps.cabac && !IS_INTRA(mb_type) ? 0 : 64;
+ }
+ }
+
+ if( h->pps.cabac ) {
+ // top_cbp
+ if(top_type) {
+ h->top_cbp = h->cbp_table[top_xy];
+ } else if(IS_INTRA(mb_type)) {
+ h->top_cbp = 0x1C0;
+ } else {
+ h->top_cbp = 0;
+ }
+ // left_cbp
+ if (left_type[0]) {
+ h->left_cbp = h->cbp_table[left_xy[0]] & 0x1f0;
+ } else if(IS_INTRA(mb_type)) {
+ h->left_cbp = 0x1C0;
+ } else {
+ h->left_cbp = 0;
+ }
+ if (left_type[0]) {
+ h->left_cbp |= ((h->cbp_table[left_xy[0]]>>((left_block[0]&(~1))+1))&0x1) << 1;
+ }
+ if (left_type[1]) {
+ h->left_cbp |= ((h->cbp_table[left_xy[1]]>>((left_block[2]&(~1))+1))&0x1) << 3;
+ }
+ }
+
#if 1
//FIXME direct mb can skip much of this
- if(IS_INTER(mb_type) || (IS_DIRECT(mb_type) && h->direct_spatial_mv_pred)){
+ if(IS_INTER(mb_type) || IS_DIRECT(mb_type)){
int list;
- for(list=0; list<2; list++){
- if((!IS_8X8(mb_type)) && !USES_LIST(mb_type, list) && !IS_DIRECT(mb_type)){
+ for(list=0; list<1+(h->slice_type==B_TYPE); list++){
+ if(!USES_LIST(mb_type, list) && !IS_DIRECT(mb_type) && !h->deblocking_filter){
/*if(!h->mv_cache_clean[list]){
memset(h->mv_cache [list], 0, 8*5*2*sizeof(int16_t)); //FIXME clean only input? clean at all?
memset(h->ref_cache[list], PART_NOT_AVAILABLE, 8*5*sizeof(int8_t));
continue;
}
h->mv_cache_clean[list]= 0;
-
- if(IS_INTER(topleft_type)){
- const int b_xy = h->mb2b_xy[topleft_xy] + 3 + 3*h->b_stride;
- const int b8_xy= h->mb2b8_xy[topleft_xy] + 1 + h->b8_stride;
- *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
- h->ref_cache[list][scan8[0] - 1 - 1*8]= s->current_picture.ref_index[list][b8_xy];
- }else{
- *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= 0;
- h->ref_cache[list][scan8[0] - 1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
- }
-
+
if(IS_INTER(top_type)){
const int b_xy= h->mb2b_xy[top_xy] + 3*h->b_stride;
const int b8_xy= h->mb2b8_xy[top_xy] + h->b8_stride;
h->ref_cache[list][scan8[0] + 2 - 1*8]=
h->ref_cache[list][scan8[0] + 3 - 1*8]= s->current_picture.ref_index[list][b8_xy + 1];
}else{
- *(uint32_t*)h->mv_cache [list][scan8[0] + 0 - 1*8]=
- *(uint32_t*)h->mv_cache [list][scan8[0] + 1 - 1*8]=
- *(uint32_t*)h->mv_cache [list][scan8[0] + 2 - 1*8]=
+ *(uint32_t*)h->mv_cache [list][scan8[0] + 0 - 1*8]=
+ *(uint32_t*)h->mv_cache [list][scan8[0] + 1 - 1*8]=
+ *(uint32_t*)h->mv_cache [list][scan8[0] + 2 - 1*8]=
*(uint32_t*)h->mv_cache [list][scan8[0] + 3 - 1*8]= 0;
*(uint32_t*)&h->ref_cache[list][scan8[0] + 0 - 1*8]= ((top_type ? LIST_NOT_USED : PART_NOT_AVAILABLE)&0xFF)*0x01010101;
}
- if(IS_INTER(topright_type)){
- const int b_xy= h->mb2b_xy[topright_xy] + 3*h->b_stride;
- const int b8_xy= h->mb2b8_xy[topright_xy] + h->b8_stride;
- *(uint32_t*)h->mv_cache[list][scan8[0] + 4 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
- h->ref_cache[list][scan8[0] + 4 - 1*8]= s->current_picture.ref_index[list][b8_xy];
- }else{
- *(uint32_t*)h->mv_cache [list][scan8[0] + 4 - 1*8]= 0;
- h->ref_cache[list][scan8[0] + 4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
- }
-
//FIXME unify cleanup or sth
if(IS_INTER(left_type[0])){
const int b_xy= h->mb2b_xy[left_xy[0]] + 3;
const int b8_xy= h->mb2b8_xy[left_xy[0]] + 1;
*(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 0*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[0]];
*(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[1]];
- h->ref_cache[list][scan8[0] - 1 + 0*8]=
+ h->ref_cache[list][scan8[0] - 1 + 0*8]=
h->ref_cache[list][scan8[0] - 1 + 1*8]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[0]>>1)];
}else{
*(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 0*8]=
h->ref_cache[list][scan8[0] - 1 + 0*8]=
h->ref_cache[list][scan8[0] - 1 + 1*8]= left_type[0] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
}
-
+
if(IS_INTER(left_type[1])){
const int b_xy= h->mb2b_xy[left_xy[1]] + 3;
const int b8_xy= h->mb2b8_xy[left_xy[1]] + 1;
*(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 2*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[2]];
*(uint32_t*)h->mv_cache[list][scan8[0] - 1 + 3*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy + h->b_stride*left_block[3]];
- h->ref_cache[list][scan8[0] - 1 + 2*8]=
+ h->ref_cache[list][scan8[0] - 1 + 2*8]=
h->ref_cache[list][scan8[0] - 1 + 3*8]= s->current_picture.ref_index[list][b8_xy + h->b8_stride*(left_block[2]>>1)];
}else{
*(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 2*8]=
*(uint32_t*)h->mv_cache [list][scan8[0] - 1 + 3*8]= 0;
h->ref_cache[list][scan8[0] - 1 + 2*8]=
h->ref_cache[list][scan8[0] - 1 + 3*8]= left_type[0] ? LIST_NOT_USED : PART_NOT_AVAILABLE;
+ assert((!left_type[0]) == (!left_type[1]));
}
- if(for_deblock)
+ if(for_deblock || (IS_DIRECT(mb_type) && !h->direct_spatial_mv_pred))
continue;
- h->ref_cache[list][scan8[5 ]+1] =
- h->ref_cache[list][scan8[7 ]+1] =
- h->ref_cache[list][scan8[13]+1] = //FIXME remove past 3 (init somewher else)
- h->ref_cache[list][scan8[4 ]] =
+ if(IS_INTER(topleft_type)){
+ const int b_xy = h->mb2b_xy[topleft_xy] + 3 + 3*h->b_stride;
+ const int b8_xy= h->mb2b8_xy[topleft_xy] + 1 + h->b8_stride;
+ *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
+ h->ref_cache[list][scan8[0] - 1 - 1*8]= s->current_picture.ref_index[list][b8_xy];
+ }else{
+ *(uint32_t*)h->mv_cache[list][scan8[0] - 1 - 1*8]= 0;
+ h->ref_cache[list][scan8[0] - 1 - 1*8]= topleft_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
+ }
+
+ if(IS_INTER(topright_type)){
+ const int b_xy= h->mb2b_xy[topright_xy] + 3*h->b_stride;
+ const int b8_xy= h->mb2b8_xy[topright_xy] + h->b8_stride;
+ *(uint32_t*)h->mv_cache[list][scan8[0] + 4 - 1*8]= *(uint32_t*)s->current_picture.motion_val[list][b_xy];
+ h->ref_cache[list][scan8[0] + 4 - 1*8]= s->current_picture.ref_index[list][b8_xy];
+ }else{
+ *(uint32_t*)h->mv_cache [list][scan8[0] + 4 - 1*8]= 0;
+ h->ref_cache[list][scan8[0] + 4 - 1*8]= topright_type ? LIST_NOT_USED : PART_NOT_AVAILABLE;
+ }
+
+
+ h->ref_cache[list][scan8[5 ]+1] =
+ h->ref_cache[list][scan8[7 ]+1] =
+ h->ref_cache[list][scan8[13]+1] = //FIXME remove past 3 (init somewhere else)
+ h->ref_cache[list][scan8[4 ]] =
h->ref_cache[list][scan8[12]] = PART_NOT_AVAILABLE;
*(uint32_t*)h->mv_cache [list][scan8[5 ]+1]=
*(uint32_t*)h->mv_cache [list][scan8[7 ]+1]=
- *(uint32_t*)h->mv_cache [list][scan8[13]+1]= //FIXME remove past 3 (init somewher else)
+ *(uint32_t*)h->mv_cache [list][scan8[13]+1]= //FIXME remove past 3 (init somewhere else)
*(uint32_t*)h->mv_cache [list][scan8[4 ]]=
*(uint32_t*)h->mv_cache [list][scan8[12]]= 0;
*(uint32_t*)h->mvd_cache[list][scan8[0] + 2 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 2];
*(uint32_t*)h->mvd_cache[list][scan8[0] + 3 - 1*8]= *(uint32_t*)h->mvd_table[list][b_xy + 3];
}else{
- *(uint32_t*)h->mvd_cache [list][scan8[0] + 0 - 1*8]=
- *(uint32_t*)h->mvd_cache [list][scan8[0] + 1 - 1*8]=
- *(uint32_t*)h->mvd_cache [list][scan8[0] + 2 - 1*8]=
+ *(uint32_t*)h->mvd_cache [list][scan8[0] + 0 - 1*8]=
+ *(uint32_t*)h->mvd_cache [list][scan8[0] + 1 - 1*8]=
+ *(uint32_t*)h->mvd_cache [list][scan8[0] + 2 - 1*8]=
*(uint32_t*)h->mvd_cache [list][scan8[0] + 3 - 1*8]= 0;
}
if(IS_INTER(left_type[0])){
}
*(uint32_t*)h->mvd_cache [list][scan8[5 ]+1]=
*(uint32_t*)h->mvd_cache [list][scan8[7 ]+1]=
- *(uint32_t*)h->mvd_cache [list][scan8[13]+1]= //FIXME remove past 3 (init somewher else)
+ *(uint32_t*)h->mvd_cache [list][scan8[13]+1]= //FIXME remove past 3 (init somewhere else)
*(uint32_t*)h->mvd_cache [list][scan8[4 ]]=
*(uint32_t*)h->mvd_cache [list][scan8[12]]= 0;
}else{
*(uint32_t*)&h->direct_cache[scan8[0] - 1*8]= 0;
}
-
+
//FIXME interlacing
if(IS_DIRECT(left_type[0])){
h->direct_cache[scan8[0] - 1 + 0*8]=
}
}
#endif
+
+ h->neighbor_transform_size= !!IS_8x8DCT(top_type) + !!IS_8x8DCT(left_type[0]);
}
static inline void write_back_intra_pred_mode(H264Context *h){
static const int8_t top [12]= {-1, 0,LEFT_DC_PRED,-1,-1,-1,-1,-1, 0};
static const int8_t left[12]= { 0,-1, TOP_DC_PRED, 0,-1,-1,-1, 0,-1,DC_128_PRED};
int i;
-
+
if(!(h->top_samples_available&0x8000)){
for(i=0; i<4; i++){
int status= top[ h->intra4x4_pred_mode_cache[scan8[0] + i] ];
}
}
}
-
+
if(!(h->left_samples_available&0x8000)){
for(i=0; i<4; i++){
int status= left[ h->intra4x4_pred_mode_cache[scan8[0] + 8*i] ];
MpegEncContext * const s = &h->s;
static const int8_t top [7]= {LEFT_DC_PRED8x8, 1,-1,-1};
static const int8_t left[7]= { TOP_DC_PRED8x8,-1, 2,-1,DC_128_PRED8x8};
-
+
if(mode < 0 || mode > 6) {
av_log(h->s.avctx, AV_LOG_ERROR, "out of range intra chroma pred mode at %d %d\n", s->mb_x, s->mb_y);
return -1;
}
-
+
if(!(h->top_samples_available&0x8000)){
mode= top[ mode ];
if(mode<0){
return -1;
}
}
-
+
if(!(h->left_samples_available&0x8000)){
mode= left[ mode ];
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;
- }
+ }
}
return mode;
MpegEncContext * const s = &h->s;
const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
- h->non_zero_count[mb_xy][0]= h->non_zero_count_cache[4+8*4];
- h->non_zero_count[mb_xy][1]= h->non_zero_count_cache[5+8*4];
- h->non_zero_count[mb_xy][2]= h->non_zero_count_cache[6+8*4];
+ h->non_zero_count[mb_xy][0]= h->non_zero_count_cache[7+8*1];
+ h->non_zero_count[mb_xy][1]= h->non_zero_count_cache[7+8*2];
+ h->non_zero_count[mb_xy][2]= h->non_zero_count_cache[7+8*3];
h->non_zero_count[mb_xy][3]= h->non_zero_count_cache[7+8*4];
- h->non_zero_count[mb_xy][4]= h->non_zero_count_cache[7+8*3];
- h->non_zero_count[mb_xy][5]= h->non_zero_count_cache[7+8*2];
- h->non_zero_count[mb_xy][6]= h->non_zero_count_cache[7+8*1];
-
- h->non_zero_count[mb_xy][7]= h->non_zero_count_cache[1+8*2];
+ h->non_zero_count[mb_xy][4]= h->non_zero_count_cache[4+8*4];
+ h->non_zero_count[mb_xy][5]= h->non_zero_count_cache[5+8*4];
+ h->non_zero_count[mb_xy][6]= h->non_zero_count_cache[6+8*4];
+
+ h->non_zero_count[mb_xy][9]= h->non_zero_count_cache[1+8*2];
h->non_zero_count[mb_xy][8]= h->non_zero_count_cache[2+8*2];
- h->non_zero_count[mb_xy][9]= h->non_zero_count_cache[2+8*1];
+ h->non_zero_count[mb_xy][7]= h->non_zero_count_cache[2+8*1];
- h->non_zero_count[mb_xy][10]=h->non_zero_count_cache[1+8*5];
+ 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][12]=h->non_zero_count_cache[2+8*4];
+ h->non_zero_count[mb_xy][10]=h->non_zero_count_cache[2+8*4];
}
/**
const int left= h->non_zero_count_cache[index8 - 1];
const int top = h->non_zero_count_cache[index8 - 8];
int i= left + top;
-
+
if(i<64) i= (i+1)>>1;
tprintf("pred_nnz L%X T%X n%d s%d P%X\n", left, top, n, scan8[n], i&31);
assert(part_width==1 || part_width==2 || part_width==4);
/* mv_cache
- B . . A T T T T
+ B . . A T T T T
U . . L . . , .
U . . L . . . .
U . . L . . , .
}else if(match_count==1){
if(left_ref==ref){
*mx= A[0];
- *my= A[1];
+ *my= A[1];
}else if(top_ref==ref){
*mx= B[0];
- *my= B[1];
+ *my= B[1];
}else{
*mx= C[0];
- *my= C[1];
+ *my= C[1];
}
}else{
if(top_ref == PART_NOT_AVAILABLE && diagonal_ref == PART_NOT_AVAILABLE && left_ref != PART_NOT_AVAILABLE){
*mx= A[0];
- *my= A[1];
+ *my= A[1];
}else{
*mx= mid_pred(A[0], B[0], C[0]);
*my= mid_pred(A[1], B[1], C[1]);
}
}
-
+
tprintf("pred_motion (%2d %2d %2d) (%2d %2d %2d) (%2d %2d %2d) -> (%2d %2d %2d) at %2d %2d %d list %d\n", top_ref, B[0], B[1], diagonal_ref, C[0], C[1], left_ref, A[0], A[1], ref, *mx, *my, h->s.mb_x, h->s.mb_y, n, list);
}
const int16_t * const B= h->mv_cache[list][ scan8[0] - 8 ];
tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d\n", top_ref, B[0], B[1], h->s.mb_x, h->s.mb_y, n, list);
-
+
if(top_ref == ref){
*mx= B[0];
*my= B[1];
}else{
const int left_ref= h->ref_cache[list][ scan8[8] - 1 ];
const int16_t * const A= h->mv_cache[list][ scan8[8] - 1 ];
-
+
tprintf("pred_16x8: (%2d %2d %2d) at %2d %2d %d list %d\n", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
if(left_ref == ref){
if(n==0){
const int left_ref= h->ref_cache[list][ scan8[0] - 1 ];
const int16_t * const A= h->mv_cache[list][ scan8[0] - 1 ];
-
+
tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d\n", left_ref, A[0], A[1], h->s.mb_x, h->s.mb_y, n, list);
if(left_ref == ref){
int diagonal_ref;
diagonal_ref= fetch_diagonal_mv(h, &C, scan8[4], list, 2);
-
+
tprintf("pred_8x16: (%2d %2d %2d) at %2d %2d %d list %d\n", diagonal_ref, C[0], C[1], h->s.mb_x, h->s.mb_y, n, list);
- if(diagonal_ref == ref){
+ if(diagonal_ref == ref){
*mx= C[0];
*my= C[1];
return;
if(top_ref == PART_NOT_AVAILABLE || left_ref == PART_NOT_AVAILABLE
|| (top_ref == 0 && *(uint32_t*)h->mv_cache[0][ scan8[0] - 8 ] == 0)
|| (left_ref == 0 && *(uint32_t*)h->mv_cache[0][ scan8[0] - 1 ] == 0)){
-
+
*mx = *my = 0;
return;
}
-
+
pred_motion(h, 0, 4, 0, 0, mx, my);
return;
}
}
}
+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 == I_TYPE)
+ cur->ref_count[0] = 0;
+ if(cur->pict_type != B_TYPE)
+ cur->ref_count[1] = 0;
+ for(list=0; list<2; list++){
+ cur->ref_count[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;
+ }
+ if(cur->pict_type != 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] = PART_NOT_AVAILABLE;
+ 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;
+ }
+ }
+ }
+}
static inline void pred_direct_motion(H264Context * const h, int *mb_type){
MpegEncContext * const s = &h->s;
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];
const int is_b8x8 = IS_8X8(*mb_type);
int sub_mb_type;
int i8, i4;
/* 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 = MB_TYPE_8x8|MB_TYPE_L0L1;
}else if(!is_b8x8 && (IS_16X16(mb_type_col) || IS_INTRA(mb_type_col))){
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 = MB_TYPE_8x8|MB_TYPE_L0L1;
}
if(!is_b8x8)
*mb_type |= MB_TYPE_DIRECT2;
tprintf("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);
-
+
if(h->direct_spatial_mv_pred){
int ref[2];
int mv[2][2];
}
if(IS_16X16(*mb_type)){
- fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, ref[0], 1);
- fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, ref[1], 1);
- if(!IS_INTRA(mb_type_col) && l1ref0[0] == 0 &&
- ABS(l1mv0[0][0]) <= 1 && ABS(l1mv0[0][1]) <= 1){
+ fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, (uint8_t)ref[0], 1);
+ fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, (uint8_t)ref[1], 1);
+ if(!IS_INTRA(mb_type_col)
+ && ( (l1ref0[0] == 0 && ABS(l1mv0[0][0]) <= 1 && ABS(l1mv0[0][1]) <= 1)
+ || (l1ref0[0] < 0 && l1ref1[0] == 0 && ABS(l1mv1[0][0]) <= 1 && ABS(l1mv1[0][1]) <= 1
+ && (h->x264_build>33 || !h->x264_build)))){
if(ref[0] > 0)
fill_rectangle(&h->mv_cache[0][scan8[0]], 4, 4, 8, pack16to32(mv[0][0],mv[0][1]), 4);
else
for(i8=0; i8<4; i8++){
const int x8 = i8&1;
const int y8 = i8>>1;
-
+
if(is_b8x8 && !IS_DIRECT(h->sub_mb_type[i8]))
continue;
h->sub_mb_type[i8] = sub_mb_type;
-
+
fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, pack16to32(mv[0][0],mv[0][1]), 4);
fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, pack16to32(mv[1][0],mv[1][1]), 4);
- fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref[0], 1);
- fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, ref[1], 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);
+
/* col_zero_flag */
- if(!IS_INTRA(mb_type_col) && l1ref0[x8 + y8*h->b8_stride] == 0){
+ 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
+ && (h->x264_build>33 || !h->x264_build)))){
+ const int16_t (*l1mv)[2]= l1ref0[x8 + y8*h->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];
+ if(ABS(mv_col[0]) <= 1 && ABS(mv_col[1]) <= 1){
+ if(ref[0] == 0)
+ fill_rectangle(&h->mv_cache[0][scan8[i8*4]], 2, 2, 8, 0, 4);
+ if(ref[1] == 0)
+ fill_rectangle(&h->mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
+ }
+ }else
for(i4=0; i4<4; i4++){
- const int16_t *mv_col = l1mv0[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))*h->b_stride];
if(ABS(mv_col[0]) <= 1 && ABS(mv_col[1]) <= 1){
if(ref[0] == 0)
*(uint32_t*)h->mv_cache[0][scan8[i8*4+i4]] = 0;
}
}
}else{ /* direct temporal mv pred */
- /* FIXME assumes that L1ref0 used the same ref lists as current frame */
if(IS_16X16(*mb_type)){
fill_rectangle(&h->ref_cache[1][scan8[0]], 4, 4, 8, 0, 1);
if(IS_INTRA(mb_type_col)){
fill_rectangle(&h-> mv_cache[0][scan8[0]], 4, 4, 8, 0, 4);
fill_rectangle(&h-> mv_cache[1][scan8[0]], 4, 4, 8, 0, 4);
}else{
- const int ref0 = l1ref0[0];
+ const int ref0 = l1ref0[0] >= 0 ? h->map_col_to_list0[0][l1ref0[0]]
+ : h->map_col_to_list0[1][l1ref1[0]];
const int dist_scale_factor = h->dist_scale_factor[ref0];
- const int16_t *mv_col = l1mv0[0];
+ const int16_t *mv_col = l1ref0[0] >= 0 ? l1mv0[0] : l1mv1[0];
int mv_l0[2];
mv_l0[0] = (dist_scale_factor * mv_col[0] + 128) >> 8;
mv_l0[1] = (dist_scale_factor * mv_col[1] + 128) >> 8;
const int x8 = i8&1;
const int y8 = i8>>1;
int ref0, dist_scale_factor;
-
+ 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-> mv_cache[1][scan8[i8*4]], 2, 2, 8, 0, 4);
continue;
}
-
+
ref0 = l1ref0[x8 + y8*h->b8_stride];
+ if(ref0 >= 0)
+ ref0 = h->map_col_to_list0[0][ref0];
+ else{
+ ref0 = h->map_col_to_list0[1][l1ref1[x8 + y8*h->b8_stride]];
+ l1mv= l1mv1;
+ }
dist_scale_factor = h->dist_scale_factor[ref0];
-
+
fill_rectangle(&h->ref_cache[0][scan8[i8*4]], 2, 2, 8, ref0, 1);
fill_rectangle(&h->ref_cache[1][scan8[i8*4]], 2, 2, 8, 0, 1);
+ if(IS_SUB_8X8(sub_mb_type)){
+ const int16_t *mv_col = l1mv[x8*3 + y8*3*h->b_stride];
+ int mx = (dist_scale_factor * mv_col[0] + 128) >> 8;
+ int my = (dist_scale_factor * 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 = l1mv0[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))*h->b_stride];
int16_t *mv_l0 = h->mv_cache[0][scan8[i8*4+i4]];
mv_l0[0] = (dist_scale_factor * mv_col[0] + 128) >> 8;
mv_l0[1] = (dist_scale_factor * mv_col[1] + 128) >> 8;
for(list=0; list<2; list++){
int y;
- if((!IS_8X8(mb_type)) && !USES_LIST(mb_type, list)){
- if(1){ //FIXME skip or never read if mb_type doesnt use it
+ if(!USES_LIST(mb_type, list)){
+ if(1){ //FIXME skip or never read if mb_type doesn't use it
for(y=0; y<4; y++){
*(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]=
*(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= 0;
}
}
for(y=0; y<2; y++){
- *(uint16_t*)&s->current_picture.ref_index[list][b8_xy + y*h->b8_stride]= (LIST_NOT_USED&0xFF)*0x0101;
+ s->current_picture.ref_index[list][b8_xy + 0 + y*h->b8_stride]=
+ s->current_picture.ref_index[list][b8_xy + 1 + y*h->b8_stride]= LIST_NOT_USED;
}
}
continue;
}
-
+
for(y=0; y<4; y++){
*(uint64_t*)s->current_picture.motion_val[list][b_xy + 0 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+0 + 8*y];
*(uint64_t*)s->current_picture.motion_val[list][b_xy + 2 + y*h->b_stride]= *(uint64_t*)h->mv_cache[list][scan8[0]+2 + 8*y];
s->current_picture.ref_index[list][b8_xy + 1 + y*h->b8_stride]= h->ref_cache[list][scan8[0]+2 + 16*y];
}
}
-
+
if(h->slice_type == B_TYPE && h->pps.cabac){
if(IS_8X8(mb_type)){
h->direct_table[b8_xy+1+0*h->b8_stride] = IS_DIRECT(h->sub_mb_type[1]) ? 1 : 0;
* Decodes a network abstraction layer unit.
* @param consumed is the number of bytes used as input
* @param length is the length of the array
- * @param dst_length is the number of decoded bytes FIXME here or a decode rbsp ttailing?
- * @returns decoded bytes, might be src+1 if no escapes
+ * @param dst_length is the number of decoded bytes FIXME here or a decode rbsp tailing?
+ * @returns decoded bytes, might be src+1 if no escapes
*/
static uint8_t *decode_nal(H264Context *h, uint8_t *src, int *dst_length, int *consumed, int length){
int i, si, di;
uint8_t *dst;
-// src[0]&0x80; //forbidden bit
+// src[0]&0x80; //forbidden bit
h->nal_ref_idc= src[0]>>5;
h->nal_unit_type= src[0]&0x1F;
src++; length--;
-#if 0
+#if 0
for(i=0; i<length; i++)
printf("%2X ", src[i]);
#endif
if(i>=length-1){ //no escaped 0
*dst_length= length;
*consumed= length+1; //+1 for the header
- return src;
+ return src;
}
h->rbsp_buffer= av_fast_realloc(h->rbsp_buffer, &h->rbsp_buffer_size, length);
dst= h->rbsp_buffer;
-//printf("deoding esc\n");
+//printf("decoding esc\n");
si=di=0;
- while(si<length){
+ while(si<length){
//remove escapes (very rare 1:2^22)
if(si+2<length && src[si]==0 && src[si+1]==0 && src[si+2]<=3){
if(src[si+2]==3){ //escape
static int encode_nal(H264Context *h, uint8_t *dst, uint8_t *src, int length, int dst_length){
int i, escape_count, si, di;
uint8_t *temp;
-
+
assert(length>=0);
assert(dst_length>0);
-
+
dst[0]= (h->nal_ref_idc<<5) + h->nal_unit_type;
if(length==0) return 1;
escape_count= 0;
for(i=0; i<length; i+=2){
if(src[i]) continue;
- if(i>0 && src[i-1]==0)
+ if(i>0 && src[i-1]==0)
i--;
if(i+2<length && src[i+1]==0 && src[i+2]<=3){
escape_count++;
i+=2;
}
}
-
- if(escape_count==0){
+
+ if(escape_count==0){
if(dst+1 != src)
memcpy(dst+1, src, length);
return length + 1;
}
-
+
if(length + escape_count + 1> dst_length)
return -1;
h->rbsp_buffer= av_fast_realloc(h->rbsp_buffer, &h->rbsp_buffer_size, length + escape_count);
temp= h->rbsp_buffer;
//printf("encoding esc\n");
-
+
si= 0;
di= 0;
while(si < length){
if(si+2<length && src[si]==0 && src[si+1]==0 && src[si+2]<=3){
temp[di++]= 0; si++;
temp[di++]= 0; si++;
- temp[di++]= 3;
+ temp[di++]= 3;
temp[di++]= src[si++];
}
else
temp[di++]= src[si++];
}
memcpy(dst+1, temp, length+escape_count);
-
+
assert(di == length+escape_count);
-
+
return di + 1;
}
* idct tranforms the 16 dc values and dequantize them.
* @param qp quantization parameter
*/
-static void h264_luma_dc_dequant_idct_c(DCTELEM *block, int qp){
- const int qmul= dequant_coeff[qp][0];
+static void h264_luma_dc_dequant_idct_c(DCTELEM *block, int qp, int qmul){
#define stride 16
int i;
int temp[16]; //FIXME check if this is a good idea
const int z2= temp[4*1+i] - temp[4*3+i];
const int z3= temp[4*1+i] + temp[4*3+i];
- block[stride*0 +offset]= ((z0 + z3)*qmul + 2)>>2; //FIXME think about merging this into decode_resdual
- block[stride*2 +offset]= ((z1 + z2)*qmul + 2)>>2;
- block[stride*8 +offset]= ((z1 - z2)*qmul + 2)>>2;
- block[stride*10+offset]= ((z0 - z3)*qmul + 2)>>2;
+ block[stride*0 +offset]= ((((z0 + z3)*qmul + 128 ) >> 8)); //FIXME think about merging this into decode_resdual
+ block[stride*2 +offset]= ((((z1 + z2)*qmul + 128 ) >> 8));
+ block[stride*8 +offset]= ((((z1 - z2)*qmul + 128 ) >> 8));
+ block[stride*10+offset]= ((((z0 - z3)*qmul + 128 ) >> 8));
}
}
#undef xStride
#undef stride
-static void chroma_dc_dequant_idct_c(DCTELEM *block, int qp){
- const int qmul= dequant_coeff[qp][0];
+static void chroma_dc_dequant_idct_c(DCTELEM *block, int qp, int qmul){
const int stride= 16*2;
const int xStride= 16;
int a,b,c,d,e;
b= c-d;
c= c+d;
- block[stride*0 + xStride*0]= ((a+c)*qmul + 0)>>1;
- block[stride*0 + xStride*1]= ((e+b)*qmul + 0)>>1;
- block[stride*1 + xStride*0]= ((a-c)*qmul + 0)>>1;
- block[stride*1 + xStride*1]= ((e-b)*qmul + 0)>>1;
+ block[stride*0 + xStride*0]= ((a+c)*qmul) >> 7;
+ block[stride*0 + xStride*1]= ((e+b)*qmul) >> 7;
+ block[stride*1 + xStride*0]= ((a-c)*qmul) >> 7;
+ block[stride*1 + xStride*1]= ((e-b)*qmul) >> 7;
}
#if 0
/**
* gets the chroma qp.
*/
-static inline int get_chroma_qp(H264Context *h, int qscale){
-
- return chroma_qp[clip(qscale + h->pps.chroma_qp_index_offset, 0, 51)];
+static inline int get_chroma_qp(int chroma_qp_index_offset, int qscale){
+
+ return chroma_qp[clip(qscale + chroma_qp_index_offset, 0, 51)];
}
static void h264_diff_dct_c(DCTELEM *block, uint8_t *src1, uint8_t *src2, int stride){
int i;
//FIXME try int temp instead of block
-
+
for(i=0; i<4; i++){
const int d0= src1[0 + i*stride] - src2[0 + i*stride];
const int d1= src1[1 + i*stride] - src2[1 + i*stride];
const int z3= d0 - d3;
const int z1= d1 + d2;
const int z2= d1 - d2;
-
+
block[0 + 4*i]= z0 + z1;
block[1 + 4*i]= 2*z3 + z2;
block[2 + 4*i]= z0 - z1;
block[3 + 4*i]= z3 - 2*z2;
- }
+ }
for(i=0; i<4; i++){
const int z0= block[0*4 + i] + block[3*4 + i];
const int z3= block[0*4 + i] - block[3*4 + i];
const int z1= block[1*4 + i] + block[2*4 + i];
const int z2= block[1*4 + i] - block[2*4 + i];
-
+
block[0*4 + i]= z0 + z1;
block[1*4 + i]= 2*z3 + z2;
block[2*4 + i]= z0 - z1;
}
#endif
-//FIXME need to check that this doesnt overflow signed 32 bit for low qp, iam not sure, its very close
+//FIXME need to check that this doesnt overflow signed 32 bit for low qp, i am not sure, it's very close
//FIXME check that gcc inlines this (and optimizes intra & seperate_dc stuff away)
static inline int quantize_c(DCTELEM *block, uint8_t *scantable, int qscale, int intra, int seperate_dc){
int i;
static void pred4x4_dc_c(uint8_t *src, uint8_t *topright, int stride){
const int dc= ( src[-stride] + src[1-stride] + src[2-stride] + src[3-stride]
+ src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 4) >>3;
-
- ((uint32_t*)(src+0*stride))[0]=
- ((uint32_t*)(src+1*stride))[0]=
- ((uint32_t*)(src+2*stride))[0]=
- ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
+
+ ((uint32_t*)(src+0*stride))[0]=
+ ((uint32_t*)(src+1*stride))[0]=
+ ((uint32_t*)(src+2*stride))[0]=
+ ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
}
static void pred4x4_left_dc_c(uint8_t *src, uint8_t *topright, int stride){
const int dc= ( src[-1+0*stride] + src[-1+1*stride] + src[-1+2*stride] + src[-1+3*stride] + 2) >>2;
-
- ((uint32_t*)(src+0*stride))[0]=
- ((uint32_t*)(src+1*stride))[0]=
- ((uint32_t*)(src+2*stride))[0]=
- ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
+
+ ((uint32_t*)(src+0*stride))[0]=
+ ((uint32_t*)(src+1*stride))[0]=
+ ((uint32_t*)(src+2*stride))[0]=
+ ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
}
static void pred4x4_top_dc_c(uint8_t *src, uint8_t *topright, int stride){
const int dc= ( src[-stride] + src[1-stride] + src[2-stride] + src[3-stride] + 2) >>2;
-
- ((uint32_t*)(src+0*stride))[0]=
- ((uint32_t*)(src+1*stride))[0]=
- ((uint32_t*)(src+2*stride))[0]=
- ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
+
+ ((uint32_t*)(src+0*stride))[0]=
+ ((uint32_t*)(src+1*stride))[0]=
+ ((uint32_t*)(src+2*stride))[0]=
+ ((uint32_t*)(src+3*stride))[0]= dc* 0x01010101;
}
static void pred4x4_128_dc_c(uint8_t *src, uint8_t *topright, int stride){
- ((uint32_t*)(src+0*stride))[0]=
- ((uint32_t*)(src+1*stride))[0]=
- ((uint32_t*)(src+2*stride))[0]=
+ ((uint32_t*)(src+0*stride))[0]=
+ ((uint32_t*)(src+1*stride))[0]=
+ ((uint32_t*)(src+2*stride))[0]=
((uint32_t*)(src+3*stride))[0]= 128U*0x01010101U;
}
LOAD_TOP_EDGE
LOAD_LEFT_EDGE
- src[0+3*stride]=(l3 + 2*l2 + l1 + 2)>>2;
+ src[0+3*stride]=(l3 + 2*l2 + l1 + 2)>>2;
src[0+2*stride]=
- src[1+3*stride]=(l2 + 2*l1 + l0 + 2)>>2;
+ src[1+3*stride]=(l2 + 2*l1 + l0 + 2)>>2;
src[0+1*stride]=
src[1+2*stride]=
- src[2+3*stride]=(l1 + 2*l0 + lt + 2)>>2;
+ src[2+3*stride]=(l1 + 2*l0 + lt + 2)>>2;
src[0+0*stride]=
src[1+1*stride]=
src[2+2*stride]=
- src[3+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
+ src[3+3*stride]=(l0 + 2*lt + t0 + 2)>>2;
src[1+0*stride]=
src[2+1*stride]=
src[3+2*stride]=(lt + 2*t0 + t1 + 2)>>2;
}
static void pred4x4_down_left_c(uint8_t *src, uint8_t *topright, int stride){
- LOAD_TOP_EDGE
- LOAD_TOP_RIGHT_EDGE
-// LOAD_LEFT_EDGE
+ LOAD_TOP_EDGE
+ LOAD_TOP_RIGHT_EDGE
+// LOAD_LEFT_EDGE
src[0+0*stride]=(t0 + t2 + 2*t1 + 2)>>2;
src[1+0*stride]=
static void pred4x4_vertical_right_c(uint8_t *src, uint8_t *topright, int stride){
const int lt= src[-1-1*stride];
- LOAD_TOP_EDGE
- LOAD_LEFT_EDGE
+ LOAD_TOP_EDGE
+ LOAD_LEFT_EDGE
const __attribute__((unused)) int unu= l3;
src[0+0*stride]=
}
static void pred4x4_vertical_left_c(uint8_t *src, uint8_t *topright, int stride){
- LOAD_TOP_EDGE
- LOAD_TOP_RIGHT_EDGE
+ LOAD_TOP_EDGE
+ LOAD_TOP_RIGHT_EDGE
const __attribute__((unused)) int unu= t7;
src[0+0*stride]=(t0 + t1 + 1)>>1;
}
static void pred4x4_horizontal_up_c(uint8_t *src, uint8_t *topright, int stride){
- LOAD_LEFT_EDGE
+ LOAD_LEFT_EDGE
src[0+0*stride]=(l0 + l1 + 1)>>1;
src[1+0*stride]=(l0 + 2*l1 + l2 + 2)>>2;
src[2+3*stride]=
src[3+3*stride]=l3;
}
-
+
static void pred4x4_horizontal_down_c(uint8_t *src, uint8_t *topright, int stride){
const int lt= src[-1-1*stride];
- LOAD_TOP_EDGE
- LOAD_LEFT_EDGE
+ LOAD_TOP_EDGE
+ LOAD_LEFT_EDGE
const __attribute__((unused)) int unu= t3;
src[0+0*stride]=
const uint32_t b= ((uint32_t*)(src-stride))[1];
const uint32_t c= ((uint32_t*)(src-stride))[2];
const uint32_t d= ((uint32_t*)(src-stride))[3];
-
+
for(i=0; i<16; i++){
((uint32_t*)(src+i*stride))[0]= a;
((uint32_t*)(src+i*stride))[1]= b;
for(i=0;i<16; i++){
dc+= src[-1+i*stride];
}
-
+
for(i=0;i<16; i++){
dc+= src[i-stride];
}
for(i=0;i<16; i++){
dc+= src[-1+i*stride];
}
-
+
dc= 0x01010101*((dc + 8)>>4);
for(i=0; i<16; i++){
int i;
const uint32_t a= ((uint32_t*)(src-stride))[0];
const uint32_t b= ((uint32_t*)(src-stride))[1];
-
+
for(i=0; i<8; i++){
((uint32_t*)(src+i*stride))[0]= a;
((uint32_t*)(src+i*stride))[1]= b;
static void pred8x8_128_dc_c(uint8_t *src, int stride){
int i;
- for(i=0; i<4; i++){
- ((uint32_t*)(src+i*stride))[0]=
- ((uint32_t*)(src+i*stride))[1]= 0x01010101U*128U;
- }
- for(i=4; i<8; i++){
- ((uint32_t*)(src+i*stride))[0]=
+ for(i=0; i<8; i++){
+ ((uint32_t*)(src+i*stride))[0]=
((uint32_t*)(src+i*stride))[1]= 0x01010101U*128U;
}
}
}
}
+#define SRC(x,y) src[(x)+(y)*stride]
+#define PL(y) \
+ const int l##y = (SRC(-1,y-1) + 2*SRC(-1,y) + SRC(-1,y+1) + 2) >> 2;
+#define PREDICT_8x8_LOAD_LEFT \
+ const int l0 = ((has_topleft ? SRC(-1,-1) : SRC(-1,0)) \
+ + 2*SRC(-1,0) + SRC(-1,1) + 2) >> 2; \
+ PL(1) PL(2) PL(3) PL(4) PL(5) PL(6) \
+ const int l7 attribute_unused = (SRC(-1,6) + 3*SRC(-1,7) + 2) >> 2
+
+#define PT(x) \
+ const int t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
+#define PREDICT_8x8_LOAD_TOP \
+ const int t0 = ((has_topleft ? SRC(-1,-1) : SRC(0,-1)) \
+ + 2*SRC(0,-1) + SRC(1,-1) + 2) >> 2; \
+ PT(1) PT(2) PT(3) PT(4) PT(5) PT(6) \
+ const int t7 attribute_unused = ((has_topright ? SRC(8,-1) : SRC(7,-1)) \
+ + 2*SRC(7,-1) + SRC(6,-1) + 2) >> 2
+
+#define PTR(x) \
+ t##x = (SRC(x-1,-1) + 2*SRC(x,-1) + SRC(x+1,-1) + 2) >> 2;
+#define PREDICT_8x8_LOAD_TOPRIGHT \
+ int t8, t9, t10, t11, t12, t13, t14, t15; \
+ if(has_topright) { \
+ PTR(8) PTR(9) PTR(10) PTR(11) PTR(12) PTR(13) PTR(14) \
+ t15 = (SRC(14,-1) + 3*SRC(15,-1) + 2) >> 2; \
+ } else t8=t9=t10=t11=t12=t13=t14=t15= SRC(7,-1);
+
+#define PREDICT_8x8_LOAD_TOPLEFT \
+ const int lt = (SRC(-1,0) + 2*SRC(-1,-1) + SRC(0,-1) + 2) >> 2
+
+#define PREDICT_8x8_DC(v) \
+ int y; \
+ for( y = 0; y < 8; y++ ) { \
+ ((uint32_t*)src)[0] = \
+ ((uint32_t*)src)[1] = v; \
+ src += stride; \
+ }
+
+static void pred8x8l_128_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_DC(0x80808080);
+}
+static void pred8x8l_left_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_LEFT;
+ const uint32_t dc = ((l0+l1+l2+l3+l4+l5+l6+l7+4) >> 3) * 0x01010101;
+ PREDICT_8x8_DC(dc);
+}
+static void pred8x8l_top_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_TOP;
+ const uint32_t dc = ((t0+t1+t2+t3+t4+t5+t6+t7+4) >> 3) * 0x01010101;
+ PREDICT_8x8_DC(dc);
+}
+static void pred8x8l_dc_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_LEFT;
+ PREDICT_8x8_LOAD_TOP;
+ const uint32_t dc = ((l0+l1+l2+l3+l4+l5+l6+l7
+ +t0+t1+t2+t3+t4+t5+t6+t7+8) >> 4) * 0x01010101;
+ PREDICT_8x8_DC(dc);
+}
+static void pred8x8l_horizontal_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_LEFT;
+#define ROW(y) ((uint32_t*)(src+y*stride))[0] =\
+ ((uint32_t*)(src+y*stride))[1] = 0x01010101 * l##y
+ ROW(0); ROW(1); ROW(2); ROW(3); ROW(4); ROW(5); ROW(6); ROW(7);
+#undef ROW
+}
+static void pred8x8l_vertical_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ int y;
+ PREDICT_8x8_LOAD_TOP;
+ src[0] = t0;
+ src[1] = t1;
+ src[2] = t2;
+ src[3] = t3;
+ src[4] = t4;
+ src[5] = t5;
+ src[6] = t6;
+ src[7] = t7;
+ for( y = 1; y < 8; y++ )
+ *(uint64_t*)(src+y*stride) = *(uint64_t*)src;
+}
+static void pred8x8l_down_left_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_TOP;
+ PREDICT_8x8_LOAD_TOPRIGHT;
+ SRC(0,0)= (t0 + 2*t1 + t2 + 2) >> 2;
+ SRC(0,1)=SRC(1,0)= (t1 + 2*t2 + t3 + 2) >> 2;
+ SRC(0,2)=SRC(1,1)=SRC(2,0)= (t2 + 2*t3 + t4 + 2) >> 2;
+ SRC(0,3)=SRC(1,2)=SRC(2,1)=SRC(3,0)= (t3 + 2*t4 + t5 + 2) >> 2;
+ SRC(0,4)=SRC(1,3)=SRC(2,2)=SRC(3,1)=SRC(4,0)= (t4 + 2*t5 + t6 + 2) >> 2;
+ SRC(0,5)=SRC(1,4)=SRC(2,3)=SRC(3,2)=SRC(4,1)=SRC(5,0)= (t5 + 2*t6 + t7 + 2) >> 2;
+ SRC(0,6)=SRC(1,5)=SRC(2,4)=SRC(3,3)=SRC(4,2)=SRC(5,1)=SRC(6,0)= (t6 + 2*t7 + t8 + 2) >> 2;
+ SRC(0,7)=SRC(1,6)=SRC(2,5)=SRC(3,4)=SRC(4,3)=SRC(5,2)=SRC(6,1)=SRC(7,0)= (t7 + 2*t8 + t9 + 2) >> 2;
+ SRC(1,7)=SRC(2,6)=SRC(3,5)=SRC(4,4)=SRC(5,3)=SRC(6,2)=SRC(7,1)= (t8 + 2*t9 + t10 + 2) >> 2;
+ SRC(2,7)=SRC(3,6)=SRC(4,5)=SRC(5,4)=SRC(6,3)=SRC(7,2)= (t9 + 2*t10 + t11 + 2) >> 2;
+ SRC(3,7)=SRC(4,6)=SRC(5,5)=SRC(6,4)=SRC(7,3)= (t10 + 2*t11 + t12 + 2) >> 2;
+ SRC(4,7)=SRC(5,6)=SRC(6,5)=SRC(7,4)= (t11 + 2*t12 + t13 + 2) >> 2;
+ SRC(5,7)=SRC(6,6)=SRC(7,5)= (t12 + 2*t13 + t14 + 2) >> 2;
+ SRC(6,7)=SRC(7,6)= (t13 + 2*t14 + t15 + 2) >> 2;
+ SRC(7,7)= (t14 + 3*t15 + 2) >> 2;
+}
+static void pred8x8l_down_right_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_TOP;
+ PREDICT_8x8_LOAD_LEFT;
+ PREDICT_8x8_LOAD_TOPLEFT;
+ SRC(0,7)= (l7 + 2*l6 + l5 + 2) >> 2;
+ SRC(0,6)=SRC(1,7)= (l6 + 2*l5 + l4 + 2) >> 2;
+ SRC(0,5)=SRC(1,6)=SRC(2,7)= (l5 + 2*l4 + l3 + 2) >> 2;
+ SRC(0,4)=SRC(1,5)=SRC(2,6)=SRC(3,7)= (l4 + 2*l3 + l2 + 2) >> 2;
+ SRC(0,3)=SRC(1,4)=SRC(2,5)=SRC(3,6)=SRC(4,7)= (l3 + 2*l2 + l1 + 2) >> 2;
+ SRC(0,2)=SRC(1,3)=SRC(2,4)=SRC(3,5)=SRC(4,6)=SRC(5,7)= (l2 + 2*l1 + l0 + 2) >> 2;
+ SRC(0,1)=SRC(1,2)=SRC(2,3)=SRC(3,4)=SRC(4,5)=SRC(5,6)=SRC(6,7)= (l1 + 2*l0 + lt + 2) >> 2;
+ SRC(0,0)=SRC(1,1)=SRC(2,2)=SRC(3,3)=SRC(4,4)=SRC(5,5)=SRC(6,6)=SRC(7,7)= (l0 + 2*lt + t0 + 2) >> 2;
+ SRC(1,0)=SRC(2,1)=SRC(3,2)=SRC(4,3)=SRC(5,4)=SRC(6,5)=SRC(7,6)= (lt + 2*t0 + t1 + 2) >> 2;
+ SRC(2,0)=SRC(3,1)=SRC(4,2)=SRC(5,3)=SRC(6,4)=SRC(7,5)= (t0 + 2*t1 + t2 + 2) >> 2;
+ SRC(3,0)=SRC(4,1)=SRC(5,2)=SRC(6,3)=SRC(7,4)= (t1 + 2*t2 + t3 + 2) >> 2;
+ SRC(4,0)=SRC(5,1)=SRC(6,2)=SRC(7,3)= (t2 + 2*t3 + t4 + 2) >> 2;
+ SRC(5,0)=SRC(6,1)=SRC(7,2)= (t3 + 2*t4 + t5 + 2) >> 2;
+ SRC(6,0)=SRC(7,1)= (t4 + 2*t5 + t6 + 2) >> 2;
+ SRC(7,0)= (t5 + 2*t6 + t7 + 2) >> 2;
+
+}
+static void pred8x8l_vertical_right_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_TOP;
+ PREDICT_8x8_LOAD_LEFT;
+ PREDICT_8x8_LOAD_TOPLEFT;
+ SRC(0,6)= (l5 + 2*l4 + l3 + 2) >> 2;
+ SRC(0,7)= (l6 + 2*l5 + l4 + 2) >> 2;
+ SRC(0,4)=SRC(1,6)= (l3 + 2*l2 + l1 + 2) >> 2;
+ SRC(0,5)=SRC(1,7)= (l4 + 2*l3 + l2 + 2) >> 2;
+ SRC(0,2)=SRC(1,4)=SRC(2,6)= (l1 + 2*l0 + lt + 2) >> 2;
+ SRC(0,3)=SRC(1,5)=SRC(2,7)= (l2 + 2*l1 + l0 + 2) >> 2;
+ SRC(0,1)=SRC(1,3)=SRC(2,5)=SRC(3,7)= (l0 + 2*lt + t0 + 2) >> 2;
+ SRC(0,0)=SRC(1,2)=SRC(2,4)=SRC(3,6)= (lt + t0 + 1) >> 1;
+ SRC(1,1)=SRC(2,3)=SRC(3,5)=SRC(4,7)= (lt + 2*t0 + t1 + 2) >> 2;
+ SRC(1,0)=SRC(2,2)=SRC(3,4)=SRC(4,6)= (t0 + t1 + 1) >> 1;
+ SRC(2,1)=SRC(3,3)=SRC(4,5)=SRC(5,7)= (t0 + 2*t1 + t2 + 2) >> 2;
+ SRC(2,0)=SRC(3,2)=SRC(4,4)=SRC(5,6)= (t1 + t2 + 1) >> 1;
+ SRC(3,1)=SRC(4,3)=SRC(5,5)=SRC(6,7)= (t1 + 2*t2 + t3 + 2) >> 2;
+ SRC(3,0)=SRC(4,2)=SRC(5,4)=SRC(6,6)= (t2 + t3 + 1) >> 1;
+ SRC(4,1)=SRC(5,3)=SRC(6,5)=SRC(7,7)= (t2 + 2*t3 + t4 + 2) >> 2;
+ SRC(4,0)=SRC(5,2)=SRC(6,4)=SRC(7,6)= (t3 + t4 + 1) >> 1;
+ SRC(5,1)=SRC(6,3)=SRC(7,5)= (t3 + 2*t4 + t5 + 2) >> 2;
+ SRC(5,0)=SRC(6,2)=SRC(7,4)= (t4 + t5 + 1) >> 1;
+ SRC(6,1)=SRC(7,3)= (t4 + 2*t5 + t6 + 2) >> 2;
+ SRC(6,0)=SRC(7,2)= (t5 + t6 + 1) >> 1;
+ SRC(7,1)= (t5 + 2*t6 + t7 + 2) >> 2;
+ SRC(7,0)= (t6 + t7 + 1) >> 1;
+}
+static void pred8x8l_horizontal_down_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_TOP;
+ PREDICT_8x8_LOAD_LEFT;
+ PREDICT_8x8_LOAD_TOPLEFT;
+ SRC(0,7)= (l6 + l7 + 1) >> 1;
+ SRC(1,7)= (l5 + 2*l6 + l7 + 2) >> 2;
+ SRC(0,6)=SRC(2,7)= (l5 + l6 + 1) >> 1;
+ SRC(1,6)=SRC(3,7)= (l4 + 2*l5 + l6 + 2) >> 2;
+ SRC(0,5)=SRC(2,6)=SRC(4,7)= (l4 + l5 + 1) >> 1;
+ SRC(1,5)=SRC(3,6)=SRC(5,7)= (l3 + 2*l4 + l5 + 2) >> 2;
+ SRC(0,4)=SRC(2,5)=SRC(4,6)=SRC(6,7)= (l3 + l4 + 1) >> 1;
+ SRC(1,4)=SRC(3,5)=SRC(5,6)=SRC(7,7)= (l2 + 2*l3 + l4 + 2) >> 2;
+ SRC(0,3)=SRC(2,4)=SRC(4,5)=SRC(6,6)= (l2 + l3 + 1) >> 1;
+ SRC(1,3)=SRC(3,4)=SRC(5,5)=SRC(7,6)= (l1 + 2*l2 + l3 + 2) >> 2;
+ SRC(0,2)=SRC(2,3)=SRC(4,4)=SRC(6,5)= (l1 + l2 + 1) >> 1;
+ SRC(1,2)=SRC(3,3)=SRC(5,4)=SRC(7,5)= (l0 + 2*l1 + l2 + 2) >> 2;
+ SRC(0,1)=SRC(2,2)=SRC(4,3)=SRC(6,4)= (l0 + l1 + 1) >> 1;
+ SRC(1,1)=SRC(3,2)=SRC(5,3)=SRC(7,4)= (lt + 2*l0 + l1 + 2) >> 2;
+ SRC(0,0)=SRC(2,1)=SRC(4,2)=SRC(6,3)= (lt + l0 + 1) >> 1;
+ SRC(1,0)=SRC(3,1)=SRC(5,2)=SRC(7,3)= (l0 + 2*lt + t0 + 2) >> 2;
+ SRC(2,0)=SRC(4,1)=SRC(6,2)= (t1 + 2*t0 + lt + 2) >> 2;
+ SRC(3,0)=SRC(5,1)=SRC(7,2)= (t2 + 2*t1 + t0 + 2) >> 2;
+ SRC(4,0)=SRC(6,1)= (t3 + 2*t2 + t1 + 2) >> 2;
+ SRC(5,0)=SRC(7,1)= (t4 + 2*t3 + t2 + 2) >> 2;
+ SRC(6,0)= (t5 + 2*t4 + t3 + 2) >> 2;
+ SRC(7,0)= (t6 + 2*t5 + t4 + 2) >> 2;
+}
+static void pred8x8l_vertical_left_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_TOP;
+ PREDICT_8x8_LOAD_TOPRIGHT;
+ SRC(0,0)= (t0 + t1 + 1) >> 1;
+ SRC(0,1)= (t0 + 2*t1 + t2 + 2) >> 2;
+ SRC(0,2)=SRC(1,0)= (t1 + t2 + 1) >> 1;
+ SRC(0,3)=SRC(1,1)= (t1 + 2*t2 + t3 + 2) >> 2;
+ SRC(0,4)=SRC(1,2)=SRC(2,0)= (t2 + t3 + 1) >> 1;
+ SRC(0,5)=SRC(1,3)=SRC(2,1)= (t2 + 2*t3 + t4 + 2) >> 2;
+ SRC(0,6)=SRC(1,4)=SRC(2,2)=SRC(3,0)= (t3 + t4 + 1) >> 1;
+ SRC(0,7)=SRC(1,5)=SRC(2,3)=SRC(3,1)= (t3 + 2*t4 + t5 + 2) >> 2;
+ SRC(1,6)=SRC(2,4)=SRC(3,2)=SRC(4,0)= (t4 + t5 + 1) >> 1;
+ SRC(1,7)=SRC(2,5)=SRC(3,3)=SRC(4,1)= (t4 + 2*t5 + t6 + 2) >> 2;
+ SRC(2,6)=SRC(3,4)=SRC(4,2)=SRC(5,0)= (t5 + t6 + 1) >> 1;
+ SRC(2,7)=SRC(3,5)=SRC(4,3)=SRC(5,1)= (t5 + 2*t6 + t7 + 2) >> 2;
+ SRC(3,6)=SRC(4,4)=SRC(5,2)=SRC(6,0)= (t6 + t7 + 1) >> 1;
+ SRC(3,7)=SRC(4,5)=SRC(5,3)=SRC(6,1)= (t6 + 2*t7 + t8 + 2) >> 2;
+ SRC(4,6)=SRC(5,4)=SRC(6,2)=SRC(7,0)= (t7 + t8 + 1) >> 1;
+ SRC(4,7)=SRC(5,5)=SRC(6,3)=SRC(7,1)= (t7 + 2*t8 + t9 + 2) >> 2;
+ SRC(5,6)=SRC(6,4)=SRC(7,2)= (t8 + t9 + 1) >> 1;
+ SRC(5,7)=SRC(6,5)=SRC(7,3)= (t8 + 2*t9 + t10 + 2) >> 2;
+ SRC(6,6)=SRC(7,4)= (t9 + t10 + 1) >> 1;
+ SRC(6,7)=SRC(7,5)= (t9 + 2*t10 + t11 + 2) >> 2;
+ SRC(7,6)= (t10 + t11 + 1) >> 1;
+ SRC(7,7)= (t10 + 2*t11 + t12 + 2) >> 2;
+}
+static void pred8x8l_horizontal_up_c(uint8_t *src, int has_topleft, int has_topright, int stride)
+{
+ PREDICT_8x8_LOAD_LEFT;
+ SRC(0,0)= (l0 + l1 + 1) >> 1;
+ SRC(1,0)= (l0 + 2*l1 + l2 + 2) >> 2;
+ SRC(0,1)=SRC(2,0)= (l1 + l2 + 1) >> 1;
+ SRC(1,1)=SRC(3,0)= (l1 + 2*l2 + l3 + 2) >> 2;
+ SRC(0,2)=SRC(2,1)=SRC(4,0)= (l2 + l3 + 1) >> 1;
+ SRC(1,2)=SRC(3,1)=SRC(5,0)= (l2 + 2*l3 + l4 + 2) >> 2;
+ SRC(0,3)=SRC(2,2)=SRC(4,1)=SRC(6,0)= (l3 + l4 + 1) >> 1;
+ SRC(1,3)=SRC(3,2)=SRC(5,1)=SRC(7,0)= (l3 + 2*l4 + l5 + 2) >> 2;
+ SRC(0,4)=SRC(2,3)=SRC(4,2)=SRC(6,1)= (l4 + l5 + 1) >> 1;
+ SRC(1,4)=SRC(3,3)=SRC(5,2)=SRC(7,1)= (l4 + 2*l5 + l6 + 2) >> 2;
+ SRC(0,5)=SRC(2,4)=SRC(4,3)=SRC(6,2)= (l5 + l6 + 1) >> 1;
+ SRC(1,5)=SRC(3,4)=SRC(5,3)=SRC(7,2)= (l5 + 2*l6 + l7 + 2) >> 2;
+ SRC(0,6)=SRC(2,5)=SRC(4,4)=SRC(6,3)= (l6 + l7 + 1) >> 1;
+ SRC(1,6)=SRC(3,5)=SRC(5,4)=SRC(7,3)= (l6 + 3*l7 + 2) >> 2;
+ SRC(0,7)=SRC(1,7)=SRC(2,6)=SRC(2,7)=SRC(3,6)=
+ SRC(3,7)=SRC(4,5)=SRC(4,6)=SRC(4,7)=SRC(5,5)=
+ SRC(5,6)=SRC(5,7)=SRC(6,4)=SRC(6,5)=SRC(6,6)=
+ SRC(6,7)=SRC(7,4)=SRC(7,5)=SRC(7,6)=SRC(7,7)= l7;
+}
+#undef PREDICT_8x8_LOAD_LEFT
+#undef PREDICT_8x8_LOAD_TOP
+#undef PREDICT_8x8_LOAD_TOPLEFT
+#undef PREDICT_8x8_LOAD_TOPRIGHT
+#undef PREDICT_8x8_DC
+#undef PTR
+#undef PT
+#undef PL
+#undef SRC
+
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,
int emu=0;
const int full_mx= mx>>2;
const int full_my= my>>2;
-
+ const int pic_width = 16*s->mb_width;
+ const int pic_height = 16*s->mb_height;
+
assert(pic->data[0]);
-
+
if(mx&7) extra_width -= 3;
if(my&7) extra_height -= 3;
-
- if( full_mx < 0-extra_width
- || full_my < 0-extra_height
- || full_mx + 16/*FIXME*/ > s->width + extra_width
- || full_my + 16/*FIXME*/ > s->height + extra_height){
- ff_emulated_edge_mc(s->edge_emu_buffer, src_y - 2 - 2*s->linesize, s->linesize, 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, s->width, s->height);
+
+ if( full_mx < 0-extra_width
+ || full_my < 0-extra_height
+ || full_mx + 16/*FIXME*/ > pic_width + extra_width
+ || full_my + 16/*FIXME*/ > pic_height + extra_height){
+ ff_emulated_edge_mc(s->edge_emu_buffer, src_y - 2 - 2*s->linesize, s->linesize, 16+5, 16+5/*FIXME*/, full_mx-2, full_my-2, pic_width, pic_height);
src_y= s->edge_emu_buffer + 2 + 2*s->linesize;
emu=1;
}
-
+
qpix_op[luma_xy](dest_y, src_y, s->linesize); //FIXME try variable height perhaps?
if(!square){
qpix_op[luma_xy](dest_y + delta, src_y + delta, s->linesize);
}
-
+
if(s->flags&CODEC_FLAG_GRAY) return;
-
+
if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, src_cb, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), s->width>>1, s->height>>1);
+ ff_emulated_edge_mc(s->edge_emu_buffer, src_cb, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), pic_width>>1, pic_height>>1);
src_cb= s->edge_emu_buffer;
}
chroma_op(dest_cb, src_cb, s->uvlinesize, chroma_height, mx&7, my&7);
if(emu){
- ff_emulated_edge_mc(s->edge_emu_buffer, src_cr, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), s->width>>1, s->height>>1);
+ ff_emulated_edge_mc(s->edge_emu_buffer, src_cr, s->uvlinesize, 9, 9/*FIXME*/, (mx>>3), (my>>3), pic_width>>1, pic_height>>1);
src_cr= s->edge_emu_buffer;
}
chroma_op(dest_cr, src_cr, s->uvlinesize, chroma_height, mx&7, my&7);
MpegEncContext * const s = &h->s;
qpel_mc_func *qpix_op= qpix_put;
h264_chroma_mc_func chroma_op= chroma_put;
-
+
dest_y += 2*x_offset + 2*y_offset*s-> linesize;
dest_cb += x_offset + y_offset*s->uvlinesize;
dest_cr += x_offset + y_offset*s->uvlinesize;
x_offset += 8*s->mb_x;
y_offset += 8*s->mb_y;
-
+
if(list0){
Picture *ref= &h->ref_list[0][ h->ref_cache[0][ scan8[n] ] ];
mc_dir_part(h, ref, n, square, chroma_height, delta, 0,
dest_cr += x_offset + y_offset*s->uvlinesize;
x_offset += 8*s->mb_x;
y_offset += 8*s->mb_y;
-
+
if(list0 && list1){
/* don't optimize for luma-only case, since B-frames usually
* use implicit weights => chroma too. */
if(h->use_weight == 2){
int weight0 = h->implicit_weight[refn0][refn1];
int weight1 = 64 - weight0;
- luma_weight_avg( dest_y, tmp_y, s-> linesize, 5, weight0, weight1, 0, 0);
- chroma_weight_avg(dest_cb, tmp_cb, s->uvlinesize, 5, weight0, weight1, 0, 0);
- chroma_weight_avg(dest_cr, tmp_cr, s->uvlinesize, 5, weight0, weight1, 0, 0);
+ luma_weight_avg( dest_y, tmp_y, s-> linesize, 5, weight0, weight1, 0);
+ chroma_weight_avg(dest_cb, tmp_cb, s->uvlinesize, 5, weight0, weight1, 0);
+ chroma_weight_avg(dest_cr, tmp_cr, s->uvlinesize, 5, weight0, weight1, 0);
}else{
luma_weight_avg(dest_y, tmp_y, s->linesize, h->luma_log2_weight_denom,
- h->luma_weight[0][refn0], h->luma_weight[1][refn1],
- h->luma_offset[0][refn0], h->luma_offset[1][refn1]);
+ h->luma_weight[0][refn0], h->luma_weight[1][refn1],
+ h->luma_offset[0][refn0] + h->luma_offset[1][refn1]);
chroma_weight_avg(dest_cb, tmp_cb, s->uvlinesize, h->chroma_log2_weight_denom,
- h->chroma_weight[0][refn0][0], h->chroma_weight[1][refn1][0],
- h->chroma_offset[0][refn0][0], h->chroma_offset[1][refn1][0]);
+ h->chroma_weight[0][refn0][0], h->chroma_weight[1][refn1][0],
+ h->chroma_offset[0][refn0][0] + h->chroma_offset[1][refn1][0]);
chroma_weight_avg(dest_cr, tmp_cr, s->uvlinesize, h->chroma_log2_weight_denom,
- h->chroma_weight[0][refn0][1], h->chroma_weight[1][refn1][1],
- h->chroma_offset[0][refn0][1], h->chroma_offset[1][refn1][1]);
+ h->chroma_weight[0][refn0][1], h->chroma_weight[1][refn1][1],
+ h->chroma_offset[0][refn0][1] + h->chroma_offset[1][refn1][1]);
}
}else{
int list = list1 ? 1 : 0;
int x_offset, int y_offset,
qpel_mc_func *qpix_put, h264_chroma_mc_func chroma_put,
qpel_mc_func *qpix_avg, h264_chroma_mc_func chroma_avg,
- h264_weight_func *weight_op, h264_biweight_func *weight_avg,
+ h264_weight_func *weight_op, h264_biweight_func *weight_avg,
int list0, int list1){
if((h->use_weight==2 && list0 && list1
&& (h->implicit_weight[ h->ref_cache[0][scan8[n]] ][ h->ref_cache[1][scan8[n]] ] != 32))
MpegEncContext * const s = &h->s;
const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
const int mb_type= s->current_picture.mb_type[mb_xy];
-
+
assert(IS_INTER(mb_type));
-
+
if(IS_16X16(mb_type)){
mc_part(h, 0, 1, 8, 0, dest_y, dest_cb, dest_cr, 0, 0,
qpix_put[0], chroma_put[0], qpix_avg[0], chroma_avg[0],
IS_DIR(mb_type, 1, 0), IS_DIR(mb_type, 1, 1));
}else{
int i;
-
+
assert(IS_8X8(mb_type));
for(i=0; i<4; i++){
int i;
done = 1;
- init_vlc(&chroma_dc_coeff_token_vlc, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 4*5,
+ 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);
for(i=0; i<4; i++){
- init_vlc(&coeff_token_vlc[i], COEFF_TOKEN_VLC_BITS, 4*17,
+ 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);
}
&chroma_dc_total_zeros_bits[i][0], 1, 1, 1);
}
for(i=0; i<15; i++){
- init_vlc(&total_zeros_vlc[i], TOTAL_ZEROS_VLC_BITS, 16,
+ 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);
}
for(i=0; i<6; i++){
- init_vlc(&run_vlc[i], RUN_VLC_BITS, 7,
+ init_vlc(&run_vlc[i], RUN_VLC_BITS, 7,
&run_len [i][0], 1, 1,
&run_bits[i][0], 1, 1, 1);
}
- init_vlc(&run7_vlc, RUN7_VLC_BITS, 16,
+ init_vlc(&run7_vlc, RUN7_VLC_BITS, 16,
&run_len [6][0], 1, 1,
&run_bits[6][0], 1, 1, 1);
}
h->pred4x4[TOP_DC_PRED ]= pred4x4_top_dc_c;
h->pred4x4[DC_128_PRED ]= pred4x4_128_dc_c;
+ h->pred8x8l[VERT_PRED ]= pred8x8l_vertical_c;
+ h->pred8x8l[HOR_PRED ]= pred8x8l_horizontal_c;
+ h->pred8x8l[DC_PRED ]= pred8x8l_dc_c;
+ h->pred8x8l[DIAG_DOWN_LEFT_PRED ]= pred8x8l_down_left_c;
+ h->pred8x8l[DIAG_DOWN_RIGHT_PRED]= pred8x8l_down_right_c;
+ h->pred8x8l[VERT_RIGHT_PRED ]= pred8x8l_vertical_right_c;
+ h->pred8x8l[HOR_DOWN_PRED ]= pred8x8l_horizontal_down_c;
+ h->pred8x8l[VERT_LEFT_PRED ]= pred8x8l_vertical_left_c;
+ h->pred8x8l[HOR_UP_PRED ]= pred8x8l_horizontal_up_c;
+ h->pred8x8l[LEFT_DC_PRED ]= pred8x8l_left_dc_c;
+ h->pred8x8l[TOP_DC_PRED ]= pred8x8l_top_dc_c;
+ h->pred8x8l[DC_128_PRED ]= pred8x8l_128_dc_c;
+
h->pred8x8[DC_PRED8x8 ]= pred8x8_dc_c;
h->pred8x8[VERT_PRED8x8 ]= pred8x8_vertical_c;
h->pred8x8[HOR_PRED8x8 ]= pred8x8_horizontal_c;
av_freep(&h->direct_table);
av_freep(&h->non_zero_count);
av_freep(&h->slice_table_base);
- av_freep(&h->top_border);
+ av_freep(&h->top_borders[1]);
+ av_freep(&h->top_borders[0]);
h->slice_table= NULL;
av_freep(&h->mb2b_xy);
av_freep(&h->s.obmc_scratchpad);
}
+static void init_dequant8_coeff_table(H264Context *h){
+ int i,q,x;
+ h->dequant8_coeff[0] = h->dequant8_buffer[0];
+ h->dequant8_coeff[1] = h->dequant8_buffer[1];
+
+ for(i=0; i<2; i++ ){
+ if(i && !memcmp(h->pps.scaling_matrix8[0], h->pps.scaling_matrix8[1], 64*sizeof(uint8_t))){
+ h->dequant8_coeff[1] = h->dequant8_buffer[0];
+ break;
+ }
+
+ for(q=0; q<52; q++){
+ int shift = div6[q];
+ int idx = rem6[q];
+ for(x=0; x<64; x++)
+ h->dequant8_coeff[i][q][x] = ((uint32_t)dequant8_coeff_init[idx][
+ dequant8_coeff_init_scan[((x>>1)&12) | (x&3)] ] * h->pps.scaling_matrix8[i][x]) << shift;
+ }
+ }
+}
+
+static void init_dequant4_coeff_table(H264Context *h){
+ int i,j,q,x;
+ const int transpose = (h->s.dsp.h264_idct_add != ff_h264_idct_add_c); //FIXME ugly
+ for(i=0; i<6; i++ ){
+ h->dequant4_coeff[i] = h->dequant4_buffer[i];
+ for(j=0; j<i; j++){
+ if(!memcmp(h->pps.scaling_matrix4[j], h->pps.scaling_matrix4[i], 16*sizeof(uint8_t))){
+ h->dequant4_coeff[i] = h->dequant4_buffer[j];
+ break;
+ }
+ }
+ if(j<i)
+ continue;
+
+ for(q=0; q<52; 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)] *
+ h->pps.scaling_matrix4[i][x]) << shift;
+ }
+ }
+}
+
+static void init_dequant_tables(H264Context *h){
+ int i,x;
+ init_dequant4_coeff_table(h);
+ if(h->pps.transform_8x8_mode)
+ init_dequant8_coeff_table(h);
+ if(h->sps.transform_bypass){
+ for(i=0; i<6; i++)
+ for(x=0; x<16; x++)
+ h->dequant4_coeff[i][0][x] = 1<<6;
+ if(h->pps.transform_8x8_mode)
+ for(i=0; i<2; i++)
+ for(x=0; x<64; x++)
+ h->dequant8_coeff[i][0][x] = 1<<6;
+ }
+}
+
+
/**
* allocates tables.
- * needs widzh/height
+ * needs width/height
*/
static int alloc_tables(H264Context *h){
MpegEncContext * const s = &h->s;
CHECKED_ALLOCZ(h->non_zero_count , big_mb_num * 16 * sizeof(uint8_t))
CHECKED_ALLOCZ(h->slice_table_base , big_mb_num * sizeof(uint8_t))
- CHECKED_ALLOCZ(h->top_border , s->mb_width * (16+8+8) * sizeof(uint8_t))
+ CHECKED_ALLOCZ(h->top_borders[0] , s->mb_width * (16+8+8) * sizeof(uint8_t))
+ CHECKED_ALLOCZ(h->top_borders[1] , s->mb_width * (16+8+8) * sizeof(uint8_t))
CHECKED_ALLOCZ(h->cbp_table, big_mb_num * sizeof(uint16_t))
if( h->pps.cabac ) {
memset(h->slice_table_base, -1, big_mb_num * sizeof(uint8_t));
h->slice_table= h->slice_table_base + s->mb_stride + 1;
- CHECKED_ALLOCZ(h->mb2b_xy , big_mb_num * sizeof(uint16_t));
- CHECKED_ALLOCZ(h->mb2b8_xy , big_mb_num * sizeof(uint16_t));
+ CHECKED_ALLOCZ(h->mb2b_xy , big_mb_num * sizeof(uint32_t));
+ CHECKED_ALLOCZ(h->mb2b8_xy , big_mb_num * sizeof(uint32_t));
for(y=0; y<s->mb_height; y++){
for(x=0; x<s->mb_width; x++){
const int mb_xy= x + y*s->mb_stride;
const int b_xy = 4*x + 4*y*h->b_stride;
const int b8_xy= 2*x + 2*y*h->b8_stride;
-
+
h->mb2b_xy [mb_xy]= b_xy;
h->mb2b8_xy[mb_xy]= b8_xy;
}
s->obmc_scratchpad = NULL;
+ if(!h->dequant4_coeff[0])
+ init_dequant_tables(h);
+
return 0;
fail:
free_tables(h);
s->width = s->avctx->width;
s->height = s->avctx->height;
s->codec_id= s->avctx->codec->id;
-
+
init_pred_ptrs(h);
+ h->dequant_coeff_pps= -1;
s->unrestricted_mv=1;
s->decode=1; //FIXME
+
+ memset(h->pps.scaling_matrix4, 16, 6*16*sizeof(uint8_t));
+ memset(h->pps.scaling_matrix8, 16, 2*64*sizeof(uint8_t));
}
static int decode_init(AVCodecContext *avctx){
MpegEncContext * const s = &h->s;
MPV_decode_defaults(s);
-
+
s->avctx = avctx;
common_init(h);
avctx->pix_fmt= PIX_FMT_YUV420P;
decode_init_vlc(h);
-
- if(avctx->codec_tag != 0x31637661 && avctx->codec_tag != 0x31435641) // avc1
+
+ if(avctx->extradata_size > 0 && avctx->extradata &&
+ *(char *)avctx->extradata == 1){
+ h->is_avc = 1;
+ h->got_avcC = 0;
+ } else {
h->is_avc = 0;
- else {
- if((avctx->extradata_size == 0) || (avctx->extradata == NULL)) {
- av_log(avctx, AV_LOG_ERROR, "AVC codec requires avcC data\n");
- return -1;
- }
- h->is_avc = 1;
- h->got_avcC = 0;
}
return 0;
}
-static void frame_start(H264Context *h){
+static int frame_start(H264Context *h){
MpegEncContext * const s = &h->s;
int i;
- MPV_frame_start(s, s->avctx);
+ if(MPV_frame_start(s, s->avctx) < 0)
+ return -1;
ff_er_frame_start(s);
assert(s->linesize && s->uvlinesize);
for(i=0; i<16; i++){
h->block_offset[i]= 4*((scan8[i] - scan8[0])&7) + 4*s->linesize*((scan8[i] - scan8[0])>>3);
- h->chroma_subblock_offset[i]= 2*((scan8[i] - scan8[0])&7) + 2*s->uvlinesize*((scan8[i] - scan8[0])>>3);
+ h->block_offset[24+i]= 4*((scan8[i] - scan8[0])&7) + 8*s->linesize*((scan8[i] - scan8[0])>>3);
}
for(i=0; i<4; i++){
h->block_offset[16+i]=
h->block_offset[20+i]= 4*((scan8[i] - scan8[0])&7) + 4*s->uvlinesize*((scan8[i] - scan8[0])>>3);
+ h->block_offset[24+16+i]=
+ h->block_offset[24+20+i]= 4*((scan8[i] - scan8[0])&7) + 8*s->uvlinesize*((scan8[i] - scan8[0])>>3);
}
/* can't be in alloc_tables because linesize isn't known there.
s->obmc_scratchpad = av_malloc(16*s->linesize + 2*8*s->uvlinesize);
// s->decode= (s->flags&CODEC_FLAG_PSNR) || !s->encoding || s->current_picture.reference /*|| h->contains_intra*/ || 1;
+ return 0;
}
static inline void backup_mb_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 -= linesize;
src_cb -= uvlinesize;
src_cr -= uvlinesize;
- h->left_border[0]= h->top_border[s->mb_x][15];
+ // 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];
}
-
- *(uint64_t*)(h->top_border[s->mb_x]+0)= *(uint64_t*)(src_y + 16*linesize);
- *(uint64_t*)(h->top_border[s->mb_x]+8)= *(uint64_t*)(src_y +8+16*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);
if(!(s->flags&CODEC_FLAG_GRAY)){
- h->left_border[17 ]= h->top_border[s->mb_x][16+7];
- h->left_border[17+9]= h->top_border[s->mb_x][24+7];
+ 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];
}
- *(uint64_t*)(h->top_border[s->mb_x]+16)= *(uint64_t*)(src_cb+8*uvlinesize);
- *(uint64_t*)(h->top_border[s->mb_x]+24)= *(uint64_t*)(src_cr+8*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);
}
}
}
if(deblock_top){
- XCHG(*(uint64_t*)(h->top_border[s->mb_x]+0), *(uint64_t*)(src_y +1), temp64, xchg);
- XCHG(*(uint64_t*)(h->top_border[s->mb_x]+8), *(uint64_t*)(src_y +9), temp64, 1);
+ 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);
+ 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);
+ }
}
if(!(s->flags&CODEC_FLAG_GRAY)){
}
}
if(deblock_top){
- XCHG(*(uint64_t*)(h->top_border[s->mb_x]+16), *(uint64_t*)(src_cb+1), temp64, 1);
- XCHG(*(uint64_t*)(h->top_border[s->mb_x]+24), *(uint64_t*)(src_cr+1), temp64, 1);
+ 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(!(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 > 0);
+
+ tprintf("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->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);
+ }
+ }
+ 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);
}
}
}
uint8_t *dest_y, *dest_cb, *dest_cr;
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);
+ void (*idct_add)(uint8_t *dst, DCTELEM *block, int stride);
if(!s->decode)
return;
- if(s->mb_skiped){
- }
-
dest_y = s->current_picture.data[0] + (mb_y * 16* s->linesize ) + mb_x * 16;
dest_cb = s->current_picture.data[1] + (mb_y * 8 * s->uvlinesize) + mb_x * 8;
dest_cr = s->current_picture.data[2] + (mb_y * 8 * s->uvlinesize) + mb_x * 8;
if (h->mb_field_decoding_flag) {
linesize = s->linesize * 2;
uvlinesize = s->uvlinesize * 2;
+ block_offset = &h->block_offset[24];
if(mb_y&1){ //FIXME move out of this func?
dest_y -= s->linesize*15;
- dest_cb-= s->linesize*7;
- dest_cr-= s->linesize*7;
+ dest_cb-= s->uvlinesize*7;
+ dest_cr-= s->uvlinesize*7;
}
} else {
linesize = s->linesize;
// dct_offset = s->linesize * 16;
}
+ idct_add = transform_bypass
+ ? IS_8x8DCT(mb_type) ? s->dsp.add_pixels8 : s->dsp.add_pixels4
+ : IS_8x8DCT(mb_type) ? s->dsp.h264_idct8_add : s->dsp.h264_idct_add;
+
if (IS_INTRA_PCM(mb_type)) {
unsigned int x, y;
for(i=0; i<16; i++) {
for (y=0; y<4; y++) {
for (x=0; x<4; x++) {
- *(dest_y + h->block_offset[i] + y*linesize + x) = h->mb[i*16+y*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 + h->block_offset[i] + y*uvlinesize + x) = h->mb[i*16+y*4+x];
+ *(dest_cb + block_offset[i] + y*uvlinesize + x) = h->mb[i*16+y*4+x];
}
}
}
for(i=20; i<20+4; i++) {
for (y=0; y<4; y++) {
for (x=0; x<4; x++) {
- *(dest_cr + h->block_offset[i] + y*uvlinesize + x) = h->mb[i*16+y*4+x];
+ *(dest_cr + block_offset[i] + y*uvlinesize + x) = h->mb[i*16+y*4+x];
}
}
}
- goto deblock;
- }
- if(IS_INTRA(mb_type)){
- if(h->deblocking_filter)
- xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 1);
+ } else {
+ if(IS_INTRA(mb_type)){
+ if(h->deblocking_filter) {
+ if (h->mb_aff_frame) {
+ if (!bottom)
+ xchg_pair_border(h, dest_y, dest_cb, dest_cr, s->linesize, s->uvlinesize, 1);
+ } else {
+ xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 1);
+ }
+ }
- if(!(s->flags&CODEC_FLAG_GRAY)){
- h->pred8x8[ h->chroma_pred_mode ](dest_cb, uvlinesize);
- h->pred8x8[ h->chroma_pred_mode ](dest_cr, uvlinesize);
- }
+ if(!(s->flags&CODEC_FLAG_GRAY)){
+ h->pred8x8[ h->chroma_pred_mode ](dest_cb, uvlinesize);
+ h->pred8x8[ h->chroma_pred_mode ](dest_cr, uvlinesize);
+ }
- if(IS_INTRA4x4(mb_type)){
- if(!s->encoding){
- for(i=0; i<16; i++){
- uint8_t * const ptr= dest_y + h->block_offset[i];
- uint8_t *topright;
- const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ];
- int tr;
-
- if(dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED){
- const int topright_avail= (h->topright_samples_available<<i)&0x8000;
- assert(mb_y || linesize <= h->block_offset[i]);
- if(!topright_avail){
- tr= ptr[3 - linesize]*0x01010101;
- topright= (uint8_t*) &tr;
- }else if(i==5 && h->deblocking_filter){
- tr= *(uint32_t*)h->top_border[mb_x+1];
- topright= (uint8_t*) &tr;
- }else
- topright= ptr + 4 - linesize;
+ if(IS_INTRA4x4(mb_type)){
+ if(!s->encoding){
+ if(IS_8x8DCT(mb_type)){
+ for(i=0; i<16; i+=4){
+ uint8_t * const ptr= dest_y + block_offset[i];
+ const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ];
+ h->pred8x8l[ dir ](ptr, (h->topleft_samples_available<<i)&0x8000,
+ (h->topright_samples_available<<(i+1))&0x8000, linesize);
+ if(h->non_zero_count_cache[ scan8[i] ])
+ idct_add(ptr, h->mb + i*16, linesize);
+ }
}else
- topright= NULL;
-
- h->pred4x4[ dir ](ptr, topright, linesize);
- if(h->non_zero_count_cache[ scan8[i] ]){
- if(s->codec_id == CODEC_ID_H264)
- s->dsp.h264_idct_add(ptr, h->mb + i*16, linesize);
- else
- svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, 0);
+ for(i=0; i<16; i++){
+ uint8_t * const ptr= dest_y + block_offset[i];
+ uint8_t *topright;
+ const int dir= h->intra4x4_pred_mode_cache[ scan8[i] ];
+ int tr;
+
+ if(dir == DIAG_DOWN_LEFT_PRED || dir == VERT_LEFT_PRED){
+ const int topright_avail= (h->topright_samples_available<<i)&0x8000;
+ assert(mb_y || linesize <= block_offset[i]);
+ if(!topright_avail){
+ tr= ptr[3 - linesize]*0x01010101;
+ topright= (uint8_t*) &tr;
+ }else
+ topright= ptr + 4 - linesize;
+ }else
+ topright= NULL;
+
+ h->pred4x4[ dir ](ptr, topright, linesize);
+ if(h->non_zero_count_cache[ scan8[i] ]){
+ if(s->codec_id == CODEC_ID_H264)
+ idct_add(ptr, h->mb + i*16, linesize);
+ else
+ svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, 0);
+ }
}
}
+ }else{
+ h->pred16x16[ h->intra16x16_pred_mode ](dest_y , linesize);
+ if(s->codec_id == CODEC_ID_H264){
+ if(!transform_bypass)
+ h264_luma_dc_dequant_idct_c(h->mb, s->qscale, h->dequant4_coeff[IS_INTRA(mb_type) ? 0:3][s->qscale][0]);
+ }else
+ svq3_luma_dc_dequant_idct_c(h->mb, s->qscale);
}
- }else{
- h->pred16x16[ h->intra16x16_pred_mode ](dest_y , linesize);
- if(s->codec_id == CODEC_ID_H264)
- h264_luma_dc_dequant_idct_c(h->mb, s->qscale);
- else
- svq3_luma_dc_dequant_idct_c(h->mb, s->qscale);
+ if(h->deblocking_filter) {
+ if (h->mb_aff_frame) {
+ if (bottom) {
+ uint8_t *pair_dest_y = s->current_picture.data[0] + ((mb_y-1) * 16* s->linesize ) + mb_x * 16;
+ uint8_t *pair_dest_cb = s->current_picture.data[1] + ((mb_y-1) * 8 * s->uvlinesize) + mb_x * 8;
+ uint8_t *pair_dest_cr = s->current_picture.data[2] + ((mb_y-1) * 8 * s->uvlinesize) + mb_x * 8;
+ s->mb_y--;
+ xchg_pair_border(h, pair_dest_y, pair_dest_cb, pair_dest_cr, s->linesize, s->uvlinesize, 0);
+ s->mb_y++;
+ }
+ } else {
+ xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0);
+ }
+ }
+ }else if(s->codec_id == CODEC_ID_H264){
+ hl_motion(h, dest_y, dest_cb, dest_cr,
+ s->dsp.put_h264_qpel_pixels_tab, s->dsp.put_h264_chroma_pixels_tab,
+ s->dsp.avg_h264_qpel_pixels_tab, s->dsp.avg_h264_chroma_pixels_tab,
+ s->dsp.weight_h264_pixels_tab, s->dsp.biweight_h264_pixels_tab);
}
- if(h->deblocking_filter)
- xchg_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize, 0);
- }else if(s->codec_id == CODEC_ID_H264){
- hl_motion(h, dest_y, dest_cb, dest_cr,
- s->dsp.put_h264_qpel_pixels_tab, s->dsp.put_h264_chroma_pixels_tab,
- s->dsp.avg_h264_qpel_pixels_tab, s->dsp.avg_h264_chroma_pixels_tab,
- s->dsp.weight_h264_pixels_tab, s->dsp.biweight_h264_pixels_tab);
- }
- if(!IS_INTRA4x4(mb_type)){
- if(s->codec_id == CODEC_ID_H264){
- for(i=0; i<16; i++){
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
- uint8_t * const ptr= dest_y + h->block_offset[i];
- s->dsp.h264_idct_add(ptr, h->mb + i*16, linesize);
+ if(!IS_INTRA4x4(mb_type)){
+ if(s->codec_id == CODEC_ID_H264){
+ const int di = IS_8x8DCT(mb_type) ? 4 : 1;
+ for(i=0; i<16; i+=di){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
+ uint8_t * const ptr= dest_y + block_offset[i];
+ idct_add(ptr, h->mb + i*16, linesize);
+ }
}
- }
- }else{
- for(i=0; i<16; i++){
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
- uint8_t * const ptr= dest_y + h->block_offset[i];
- svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0);
+ }else{
+ for(i=0; i<16; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ //FIXME benchmark weird rule, & below
+ uint8_t * const ptr= dest_y + block_offset[i];
+ svq3_add_idct_c(ptr, h->mb + i*16, linesize, s->qscale, IS_INTRA(mb_type) ? 1 : 0);
+ }
}
}
}
- }
- if(!(s->flags&CODEC_FLAG_GRAY)){
- chroma_dc_dequant_idct_c(h->mb + 16*16, h->chroma_qp);
- chroma_dc_dequant_idct_c(h->mb + 16*16+4*16, h->chroma_qp);
- if(s->codec_id == CODEC_ID_H264){
- for(i=16; i<16+4; i++){
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
- uint8_t * const ptr= dest_cb + h->block_offset[i];
- s->dsp.h264_idct_add(ptr, h->mb + i*16, uvlinesize);
- }
+ if(!(s->flags&CODEC_FLAG_GRAY)){
+ idct_add = transform_bypass ? s->dsp.add_pixels4 : s->dsp.h264_idct_add;
+ if(!transform_bypass){
+ chroma_dc_dequant_idct_c(h->mb + 16*16, h->chroma_qp, h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp][0]);
+ chroma_dc_dequant_idct_c(h->mb + 16*16+4*16, h->chroma_qp, h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp][0]);
}
- for(i=20; i<20+4; i++){
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
- uint8_t * const ptr= dest_cr + h->block_offset[i];
- s->dsp.h264_idct_add(ptr, h->mb + i*16, uvlinesize);
+ if(s->codec_id == CODEC_ID_H264){
+ for(i=16; i<16+4; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
+ uint8_t * const ptr= dest_cb + block_offset[i];
+ idct_add(ptr, h->mb + i*16, uvlinesize);
+ }
}
- }
- }else{
- for(i=16; i<16+4; i++){
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
- uint8_t * const ptr= dest_cb + h->block_offset[i];
- svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, chroma_qp[s->qscale + 12] - 12, 2);
+ for(i=20; i<20+4; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
+ uint8_t * const ptr= dest_cr + block_offset[i];
+ idct_add(ptr, h->mb + i*16, uvlinesize);
+ }
}
- }
- for(i=20; i<20+4; i++){
- if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
- uint8_t * const ptr= dest_cr + h->block_offset[i];
- svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, chroma_qp[s->qscale + 12] - 12, 2);
+ }else{
+ for(i=16; i<16+4; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
+ uint8_t * const ptr= dest_cb + block_offset[i];
+ svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, chroma_qp[s->qscale + 12] - 12, 2);
+ }
+ }
+ for(i=20; i<20+4; i++){
+ if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){
+ uint8_t * const ptr= dest_cr + block_offset[i];
+ svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, chroma_qp[s->qscale + 12] - 12, 2);
+ }
}
}
}
}
-deblock:
if(h->deblocking_filter) {
- backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
- fill_caches(h, mb_type, 1); //FIXME dont fill stuff which isnt used by filter_mb
- filter_mb(h, mb_x, mb_y, dest_y, dest_cb, dest_cr);
+ if (h->mb_aff_frame) {
+ 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];
+ uint8_t tmp = s->current_picture.data[1][384];
+ 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;
+
+ backup_pair_border(h, pair_dest_y, pair_dest_cb, pair_dest_cr, s->linesize, s->uvlinesize);
+ // TODO deblock a pair
+ // top
+ s->mb_y--;
+ tprintf("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
+ filter_mb(h, mb_x, mb_y, pair_dest_y, pair_dest_cb, pair_dest_cr, linesize, uvlinesize);
+ if (tmp != s->current_picture.data[1][384]) {
+ tprintf("modified pixel 8,1 (1)\n");
+ }
+ // bottom
+ s->mb_y++;
+ tprintf("call mbaff filter_mb\n");
+ fill_caches(h, mb_type_bottom, 1); //FIXME don't fill stuff which isn't used by filter_mb
+ filter_mb(h, mb_x, mb_y+1, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
+ if (tmp != s->current_picture.data[1][384]) {
+ tprintf("modified pixel 8,1 (2)\n");
+ }
+ } else {
+ tprintf("call filter_mb\n");
+ backup_mb_border(h, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
+ fill_caches(h, mb_type, 1); //FIXME don't fill stuff which isn't used by filter_mb
+ filter_mb(h, mb_x, mb_y, dest_y, dest_cb, dest_cr, linesize, uvlinesize);
+ }
}
}
MpegEncContext * const s = &h->s;
int i;
int smallest_poc_greater_than_current = -1;
- Picture sorted_short_ref[16];
-
+ Picture sorted_short_ref[32];
+
if(h->slice_type==B_TYPE){
int out_i;
- int limit= -1;
+ 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=-1;
+ int best_i=INT_MIN;
int best_poc=INT_MAX;
for(i=0; i<h->short_ref_count; i++){
best_i= i;
}
}
-
- assert(best_i != -1);
-
+
+ assert(best_i != INT_MIN);
+
limit= best_poc;
sorted_short_ref[out_i]= *h->short_ref[best_i];
tprintf("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);
for(i=0; i<h->short_ref_count && index < h->ref_count[list]; i++, j+=step) {
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);
}
h->default_ref_list[ list ][index ]= *h->long_ref[i];
h->default_ref_list[ list ][index++].pic_id= i;;
}
-
+
if(list && (smallest_poc_greater_than_current<=0 || smallest_poc_greater_than_current>=h->short_ref_count) && (1 < index)){
// swap the two first elements of L1 when
// L0 and L1 are identical
Picture temp= h->default_ref_list[1][0];
h->default_ref_list[1][0] = h->default_ref_list[1][1];
- h->default_ref_list[1][0] = temp;
+ h->default_ref_list[1][1] = temp;
}
if(index < h->ref_count[ list ])
}
}else{
int index=0;
- for(i=0; i<h->short_ref_count && index < h->ref_count[0]; i++){
+ for(i=0; i<h->short_ref_count; i++){
if(h->short_ref[i]->reference != 3) continue; //FIXME refernce field shit
h->default_ref_list[0][index ]= *h->short_ref[i];
h->default_ref_list[0][index++].pic_id= h->short_ref[i]->frame_num;
}
- for(i = 0; i < 16 && index < h->ref_count[0]; i++){
+ for(i = 0; i < 16; i++){
if(h->long_ref[i] == NULL) continue;
if(h->long_ref[i]->reference != 3) continue;
h->default_ref_list[0][index ]= *h->long_ref[i];
static int decode_ref_pic_list_reordering(H264Context *h){
MpegEncContext * const s = &h->s;
- int list;
-
+ int list, index;
+
print_short_term(h);
print_long_term(h);
- if(h->slice_type==I_TYPE || h->slice_type==SI_TYPE) return 0; //FIXME move beofre func
-
+ if(h->slice_type==I_TYPE || h->slice_type==SI_TYPE) return 0; //FIXME move before func
+
for(list=0; list<2; list++){
memcpy(h->ref_list[list], h->default_ref_list[list], sizeof(Picture)*h->ref_count[list]);
if(get_bits1(&s->gb)){
int pred= h->curr_pic_num;
- int index;
for(index=0; ; index++){
int reordering_of_pic_nums_idc= get_ue_golomb(&s->gb);
int pic_id;
int i;
-
- if(reordering_of_pic_nums_idc==3)
+ Picture *ref = NULL;
+
+ if(reordering_of_pic_nums_idc==3)
break;
-
+
if(index >= h->ref_count[list]){
av_log(h->s.avctx, AV_LOG_ERROR, "reference count overflow\n");
return -1;
}
-
+
if(reordering_of_pic_nums_idc<3){
if(reordering_of_pic_nums_idc<2){
const int abs_diff_pic_num= get_ue_golomb(&s->gb) + 1;
if(reordering_of_pic_nums_idc == 0) pred-= abs_diff_pic_num;
else pred+= abs_diff_pic_num;
pred &= h->max_pic_num - 1;
-
- for(i= h->ref_count[list]-1; i>=0; i--){
- if(h->ref_list[list][i].data[0] != NULL && h->ref_list[list][i].pic_id == pred && h->ref_list[list][i].long_ref==0) // ignore non existing pictures by testing data[0] pointer
+
+ for(i= h->short_ref_count-1; i>=0; i--){
+ ref = h->short_ref[i];
+ assert(ref->reference == 3);
+ assert(!ref->long_ref);
+ if(ref->data[0] != NULL && ref->frame_num == pred && ref->long_ref == 0) // ignore non existing pictures by testing data[0] pointer
break;
}
+ if(i>=0)
+ ref->pic_id= ref->frame_num;
}else{
pic_id= get_ue_golomb(&s->gb); //long_term_pic_idx
-
- for(i= h->ref_count[list]-1; i>=0; i--){
- if(h->ref_list[list][i].pic_id == pic_id && h->ref_list[list][i].long_ref==1) // no need to ignore non existing pictures as non existing pictures have long_ref==0
- break;
- }
+ ref = h->long_ref[pic_id];
+ ref->pic_id= pic_id;
+ assert(ref->reference == 3);
+ assert(ref->long_ref);
+ i=0;
}
if (i < 0) {
av_log(h->s.avctx, AV_LOG_ERROR, "reference picture missing during reorder\n");
memset(&h->ref_list[list][index], 0, sizeof(Picture)); //FIXME
- } else if (i != index) /* this test is not necessary, it is only an optimisation to skip double copy of Picture structure in this case */ {
- Picture tmp= h->ref_list[list][i];
- if (i < index) {
- i = h->ref_count[list];
+ } else {
+ for(i=index; i+1<h->ref_count[list]; i++){
+ if(ref->long_ref == h->ref_list[list][i].long_ref && ref->pic_id == h->ref_list[list][i].pic_id)
+ break;
}
for(; i > index; i--){
h->ref_list[list][i]= h->ref_list[list][i-1];
}
- h->ref_list[list][index]= tmp;
+ h->ref_list[list][index]= *ref;
}
}else{
av_log(h->s.avctx, AV_LOG_ERROR, "illegal reordering_of_pic_nums_idc\n");
if(h->slice_type!=B_TYPE) break;
}
-
+ for(list=0; list<2; 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->slice_type!=B_TYPE) break;
+ }
+
if(h->slice_type==B_TYPE && !h->direct_spatial_mv_pred)
direct_dist_scale_factor(h);
- return 0;
+ direct_ref_list_init(h);
+ return 0;
}
static int pred_weight_table(H264Context *h){
MpegEncContext * const s = &h->s;
int list, i;
int luma_def, chroma_def;
-
+
h->use_weight= 0;
h->use_weight_chroma= 0;
h->luma_log2_weight_denom= get_ue_golomb(&s->gb);
for(list=0; list<2; list++){
for(i=0; i<h->ref_count[list]; i++){
int luma_weight_flag, chroma_weight_flag;
-
+
luma_weight_flag= get_bits1(&s->gb);
if(luma_weight_flag){
h->luma_weight[list][i]= get_se_golomb(&s->gb);
for(ref0=0; ref0 < h->ref_count[0]; ref0++){
int poc0 = h->ref_list[0][ref0].poc;
for(ref1=0; ref1 < h->ref_count[1]; ref1++){
- int poc1 = h->ref_list[0][ref1].poc;
+ int poc1 = h->ref_list[1][ref1].poc;
int td = clip(poc1 - poc0, -128, 127);
if(td){
int tb = clip(cur_poc - poc0, -128, 127);
}
}
+static inline void unreference_pic(H264Context *h, Picture *pic){
+ int i;
+ pic->reference=0;
+ if(pic == h->delayed_output_pic)
+ pic->reference=1;
+ else{
+ for(i = 0; h->delayed_pic[i]; i++)
+ if(pic == h->delayed_pic[i]){
+ pic->reference=1;
+ break;
+ }
+ }
+}
+
/**
* instantaneous decoder refresh.
*/
static void idr(H264Context *h){
- int i,j;
-
-#define CHECK_DELAY(pic) \
- for(j = 0; h->delayed_pic[j]; j++) \
- if(pic == h->delayed_pic[j]){ \
- pic->reference=1; \
- break; \
- }
+ int i;
for(i=0; i<16; i++){
if (h->long_ref[i] != NULL) {
- h->long_ref[i]->reference=0;
- CHECK_DELAY(h->long_ref[i]);
+ unreference_pic(h, h->long_ref[i]);
h->long_ref[i]= NULL;
}
}
h->long_ref_count=0;
for(i=0; i<h->short_ref_count; i++){
- h->short_ref[i]->reference=0;
- CHECK_DELAY(h->short_ref[i]);
+ unreference_pic(h, h->short_ref[i]);
h->short_ref[i]= NULL;
}
h->short_ref_count=0;
}
-#undef CHECK_DELAY
+
+/* forget old pics after a seek */
+static void flush_dpb(AVCodecContext *avctx){
+ H264Context *h= avctx->priv_data;
+ int i;
+ for(i=0; i<16; i++)
+ h->delayed_pic[i]= NULL;
+ h->delayed_output_pic= NULL;
+ idr(h);
+ if(h->s.current_picture_ptr)
+ h->s.current_picture_ptr->reference= 0;
+}
/**
*
- * @return the removed picture or NULL if an error occures
+ * @return the removed picture or NULL if an error occurs
*/
static Picture * remove_short(H264Context *h, int frame_num){
MpegEncContext * const s = &h->s;
int i;
-
+
if(s->avctx->debug&FF_DEBUG_MMCO)
av_log(h->s.avctx, AV_LOG_DEBUG, "remove short %d count %d\n", frame_num, h->short_ref_count);
-
+
for(i=0; i<h->short_ref_count; i++){
Picture *pic= h->short_ref[i];
if(s->avctx->debug&FF_DEBUG_MMCO)
/**
*
- * @return the removed picture or NULL if an error occures
+ * @return the removed picture or NULL if an error occurs
*/
static Picture * remove_long(H264Context *h, int i){
Picture *pic;
int i, j;
int current_is_long=0;
Picture *pic;
-
+
if((s->avctx->debug&FF_DEBUG_MMCO) && mmco_count==0)
av_log(h->s.avctx, AV_LOG_DEBUG, "no mmco here\n");
-
+
for(i=0; i<mmco_count; i++){
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_frame_num, h->mmco[i].long_index);
switch(mmco[i].opcode){
case MMCO_SHORT2UNUSED:
pic= remove_short(h, mmco[i].short_frame_num);
- if(pic==NULL) return -1;
- pic->reference= 0;
+ if(pic)
+ unreference_pic(h, pic);
+ else if(s->avctx->debug&FF_DEBUG_MMCO)
+ av_log(h->s.avctx, AV_LOG_DEBUG, "mmco: remove_short() failure\n");
break;
case MMCO_SHORT2LONG:
pic= remove_long(h, mmco[i].long_index);
- if(pic) pic->reference=0;
-
+ if(pic) unreference_pic(h, pic);
+
h->long_ref[ mmco[i].long_index ]= remove_short(h, mmco[i].short_frame_num);
h->long_ref[ mmco[i].long_index ]->long_ref=1;
h->long_ref_count++;
break;
case MMCO_LONG2UNUSED:
pic= remove_long(h, mmco[i].long_index);
- if(pic==NULL) return -1;
- pic->reference= 0;
+ if(pic)
+ unreference_pic(h, pic);
+ else if(s->avctx->debug&FF_DEBUG_MMCO)
+ av_log(h->s.avctx, AV_LOG_DEBUG, "mmco: remove_long() failure\n");
break;
case MMCO_LONG:
pic= remove_long(h, mmco[i].long_index);
- if(pic) pic->reference=0;
-
+ if(pic) unreference_pic(h, pic);
+
h->long_ref[ mmco[i].long_index ]= s->current_picture_ptr;
h->long_ref[ mmco[i].long_index ]->long_ref=1;
h->long_ref_count++;
-
+
current_is_long=1;
break;
case MMCO_SET_MAX_LONG:
// just remove the long term which index is greater than new max
for(j = mmco[i].long_index; j<16; j++){
pic = remove_long(h, j);
- if (pic) pic->reference=0;
+ if (pic) unreference_pic(h, pic);
}
break;
case MMCO_RESET:
while(h->short_ref_count){
pic= remove_short(h, h->short_ref[0]->frame_num);
- pic->reference=0;
+ unreference_pic(h, pic);
}
for(j = 0; j < 16; j++) {
pic= remove_long(h, j);
- if(pic) pic->reference=0;
+ if(pic) unreference_pic(h, pic);
}
break;
default: assert(0);
}
}
-
+
if(!current_is_long){
pic= remove_short(h, s->current_picture_ptr->frame_num);
if(pic){
- pic->reference=0;
+ unreference_pic(h, pic);
av_log(h->s.avctx, AV_LOG_ERROR, "illegal short term buffer state detected\n");
}
-
+
if(h->short_ref_count)
memmove(&h->short_ref[1], &h->short_ref[0], h->short_ref_count*sizeof(Picture*));
h->short_ref[0]->long_ref=0;
h->short_ref_count++;
}
-
+
print_short_term(h);
print_long_term(h);
- return 0;
+ return 0;
}
static int decode_ref_pic_marking(H264Context *h){
MpegEncContext * const s = &h->s;
int i;
-
+
if(h->nal_unit_type == NAL_IDR_SLICE){ //FIXME fields
s->broken_link= get_bits1(&s->gb) -1;
h->mmco[0].long_index= get_bits1(&s->gb) - 1; // current_long_term_idx
else{
h->mmco[0].opcode= MMCO_LONG;
h->mmco_index= 1;
- }
+ }
}else{
if(get_bits1(&s->gb)){ // adaptive_ref_pic_marking_mode_flag
- for(i= 0; i<MAX_MMCO_COUNT; i++) {
+ for(i= 0; i<MAX_MMCO_COUNT; i++) {
MMCOOpcode opcode= get_ue_golomb(&s->gb);;
h->mmco[i].opcode= opcode;
if(opcode==MMCO_SHORT2UNUSED || opcode==MMCO_SHORT2LONG){
h->mmco[i].short_frame_num= (h->frame_num - get_ue_golomb(&s->gb) - 1) & ((1<<h->sps.log2_max_frame_num)-1); //FIXME fields
/* if(h->mmco[i].short_frame_num >= h->short_ref_count || h->short_ref[ h->mmco[i].short_frame_num ] == NULL){
- fprintf(stderr, "illegal short ref in memory management control operation %d\n", mmco);
+ av_log(s->avctx, AV_LOG_ERROR, "illegal short ref in memory management control operation %d\n", mmco);
return -1;
}*/
}
return -1;
}
}
-
+
if(opcode > MMCO_LONG){
av_log(h->s.avctx, AV_LOG_ERROR, "illegal memory management control operation %d\n", opcode);
return -1;
h->mmco_index= 0;
}
}
-
- return 0;
+
+ return 0;
}
static int init_poc(H264Context *h){
if(h->sps.poc_type==0){
const int max_poc_lsb= 1<<h->sps.log2_max_poc_lsb;
+ if(h->nal_unit_type == NAL_IDR_SLICE){
+ h->prev_poc_msb=
+ h->prev_poc_lsb= 0;
+ }
+
if (h->poc_lsb < h->prev_poc_lsb && h->prev_poc_lsb - h->poc_lsb >= max_poc_lsb/2)
h->poc_msb = h->prev_poc_msb + max_poc_lsb;
else if(h->poc_lsb > h->prev_poc_lsb && h->prev_poc_lsb - h->poc_lsb < -max_poc_lsb/2)
else
h->poc_msb = h->prev_poc_msb;
//printf("poc: %d %d\n", h->poc_msb, h->poc_lsb);
- field_poc[0] =
+ field_poc[0] =
field_poc[1] = h->poc_msb + h->poc_lsb;
- if(s->picture_structure == PICT_FRAME)
+ if(s->picture_structure == PICT_FRAME)
field_poc[1] += h->delta_poc_bottom;
}else if(h->sps.poc_type==1){
int abs_frame_num, expected_delta_per_poc_cycle, expectedpoc;
if(h->nal_ref_idc==0 && abs_frame_num > 0)
abs_frame_num--;
-
+
expected_delta_per_poc_cycle = 0;
for(i=0; i < h->sps.poc_cycle_length; i++)
expected_delta_per_poc_cycle += h->sps.offset_for_ref_frame[ i ]; //FIXME integrate during sps parse
} else
expectedpoc = 0;
- if(h->nal_ref_idc == 0)
+ if(h->nal_ref_idc == 0)
expectedpoc = expectedpoc + h->sps.offset_for_non_ref_pic;
-
+
field_poc[0] = expectedpoc + h->delta_poc[0];
field_poc[1] = field_poc[0] + h->sps.offset_for_top_to_bottom_field;
field_poc[0]= poc;
field_poc[1]= poc;
}
-
+
if(s->picture_structure != PICT_BOTTOM_FIELD)
s->current_picture_ptr->field_poc[0]= field_poc[0];
if(s->picture_structure != PICT_TOP_FIELD)
int default_ref_list_done = 0;
s->current_picture.reference= h->nal_ref_idc != 0;
+ s->dropable= h->nal_ref_idc == 0;
first_mb_in_slice= get_ue_golomb(&s->gb);
h->slice_type_fixed=1;
}else
h->slice_type_fixed=0;
-
+
slice_type= slice_type_map[ slice_type ];
if (slice_type == I_TYPE
|| (h->slice_num != 0 && slice_type == h->slice_type) ) {
}
h->slice_type= slice_type;
- s->pict_type= h->slice_type; // to make a few old func happy, its wrong though
-
+ s->pict_type= h->slice_type; // to make a few old func happy, it's wrong though
+
pps_id= get_ue_golomb(&s->gb);
if(pps_id>255){
av_log(h->s.avctx, AV_LOG_ERROR, "pps_id out of range\n");
av_log(h->s.avctx, AV_LOG_ERROR, "non existing SPS referenced\n");
return -1;
}
-
+
+ if(h->dequant_coeff_pps != pps_id){
+ h->dequant_coeff_pps = pps_id;
+ init_dequant_tables(h);
+ }
+
s->mb_width= h->sps.mb_width;
- s->mb_height= h->sps.mb_height;
-
+ s->mb_height= h->sps.mb_height * (2 - h->sps.frame_mbs_only_flag);
+
h->b_stride= s->mb_width*4 + 1;
h->b8_stride= s->mb_width*2 + 1;
- s->resync_mb_x = s->mb_x = first_mb_in_slice % s->mb_width;
- s->resync_mb_y = s->mb_y = first_mb_in_slice / s->mb_width; //FIXME AFFW
-
s->width = 16*s->mb_width - 2*(h->sps.crop_left + h->sps.crop_right );
if(h->sps.frame_mbs_only_flag)
s->height= 16*s->mb_height - 2*(h->sps.crop_top + h->sps.crop_bottom);
else
s->height= 16*s->mb_height - 4*(h->sps.crop_top + h->sps.crop_bottom); //FIXME recheck
-
- if (s->context_initialized
+
+ if (s->context_initialized
&& ( s->width != s->avctx->width || s->height != s->avctx->height)) {
free_tables(h);
MPV_common_end(s);
if (MPV_common_init(s) < 0)
return -1;
+ if(s->dsp.h264_idct_add == ff_h264_idct_add_c){ //FIXME little ugly
+ memcpy(h->zigzag_scan, zigzag_scan, 16*sizeof(uint8_t));
+ memcpy(h-> field_scan, field_scan, 16*sizeof(uint8_t));
+ }else{
+ int i;
+ for(i=0; i<16; i++){
+#define T(x) (x>>2) | ((x<<2) & 0xF)
+ h->zigzag_scan[i] = T(zigzag_scan[i]);
+ h-> field_scan[i] = T( field_scan[i]);
+ }
+ }
+ if(h->sps.transform_bypass){ //FIXME same ugly
+ h->zigzag_scan_q0 = zigzag_scan;
+ h->field_scan_q0 = field_scan;
+ }else{
+ h->zigzag_scan_q0 = h->zigzag_scan;
+ h->field_scan_q0 = h->field_scan;
+ }
+
alloc_tables(h);
s->avctx->width = s->width;
if(!s->avctx->sample_aspect_ratio.den)
s->avctx->sample_aspect_ratio.den = 1;
- if(h->sps.timing_info_present_flag && h->sps.fixed_frame_rate_flag){
- s->avctx->frame_rate = h->sps.time_scale;
- s->avctx->frame_rate_base = h->sps.num_units_in_tick;
+ if(h->sps.timing_info_present_flag){
+ s->avctx->time_base= (AVRational){h->sps.num_units_in_tick, h->sps.time_scale};
}
}
if(h->slice_num == 0){
- frame_start(h);
+ if(frame_start(h) < 0)
+ return -1;
}
s->current_picture_ptr->frame_num= //FIXME frame_num cleanup
h->frame_num= get_bits(&s->gb, h->sps.log2_max_frame_num);
+ h->mb_aff_frame = 0;
if(h->sps.frame_mbs_only_flag){
s->picture_structure= PICT_FRAME;
}else{
- if(get_bits1(&s->gb)) //field_pic_flag
+ if(get_bits1(&s->gb)) { //field_pic_flag
s->picture_structure= PICT_TOP_FIELD + get_bits1(&s->gb); //bottom_field_flag
- else
+ } else {
s->picture_structure= PICT_FRAME;
+ first_mb_in_slice <<= h->sps.mb_aff;
+ h->mb_aff_frame = h->sps.mb_aff;
+ }
+ }
+
+ s->resync_mb_x = s->mb_x = first_mb_in_slice % s->mb_width;
+ s->resync_mb_y = s->mb_y = first_mb_in_slice / s->mb_width;
+ if(s->mb_y >= s->mb_height){
+ return -1;
}
if(s->picture_structure==PICT_FRAME){
h->curr_pic_num= 2*h->frame_num;
h->max_pic_num= 1<<(h->sps.log2_max_frame_num + 1);
}
-
+
if(h->nal_unit_type == NAL_IDR_SLICE){
get_ue_golomb(&s->gb); /* idr_pic_id */
}
-
+
if(h->sps.poc_type==0){
h->poc_lsb= get_bits(&s->gb, h->sps.log2_max_poc_lsb);
-
+
if(h->pps.pic_order_present==1 && s->picture_structure==PICT_FRAME){
h->delta_poc_bottom= get_se_golomb(&s->gb);
}
}
-
+
if(h->sps.poc_type==1 && !h->sps.delta_pic_order_always_zero_flag){
h->delta_poc[0]= get_se_golomb(&s->gb);
-
+
if(h->pps.pic_order_present==1 && s->picture_structure==PICT_FRAME)
h->delta_poc[1]= get_se_golomb(&s->gb);
}
-
+
init_poc(h);
-
+
if(h->pps.redundant_pic_cnt_present){
h->redundant_pic_count= get_ue_golomb(&s->gb);
}
h->direct_spatial_mv_pred= get_bits1(&s->gb);
}
num_ref_idx_active_override_flag= get_bits1(&s->gb);
-
+
if(num_ref_idx_active_override_flag){
h->ref_count[0]= get_ue_golomb(&s->gb) + 1;
if(h->slice_type==B_TYPE)
fill_default_ref_list(h);
}
- decode_ref_pic_list_reordering(h);
+ if(decode_ref_pic_list_reordering(h) < 0)
+ return -1;
- if( (h->pps.weighted_pred && (h->slice_type == P_TYPE || h->slice_type == SP_TYPE ))
+ if( (h->pps.weighted_pred && (h->slice_type == P_TYPE || h->slice_type == SP_TYPE ))
|| (h->pps.weighted_bipred_idc==1 && h->slice_type==B_TYPE ) )
pred_weight_table(h);
else if(h->pps.weighted_bipred_idc==2 && h->slice_type==B_TYPE)
implicit_weight_table(h);
else
h->use_weight = 0;
-
+
if(s->current_picture.reference)
decode_ref_pic_marking(h);
av_log(s->avctx, AV_LOG_ERROR, "QP %d out of range\n", s->qscale);
return -1;
}
- h->chroma_qp = get_chroma_qp(h, s->qscale);
+ h->chroma_qp = get_chroma_qp(h->pps.chroma_qp_index_offset, s->qscale);
//FIXME qscale / qp ... stuff
if(h->slice_type == SP_TYPE){
get_bits1(&s->gb); /* sp_for_switch_flag */
h->slice_beta_offset = 0;
if( h->pps.deblocking_filter_parameters_present ) {
h->deblocking_filter= get_ue_golomb(&s->gb);
- if(h->deblocking_filter < 2)
+ if(h->deblocking_filter < 2)
h->deblocking_filter^= 1; // 1<->0
if( h->deblocking_filter ) {
h->slice_beta_offset = get_se_golomb(&s->gb) << 1;
}
}
+ if( s->avctx->skip_loop_filter >= AVDISCARD_ALL
+ ||(s->avctx->skip_loop_filter >= AVDISCARD_NONKEY && h->slice_type != I_TYPE)
+ ||(s->avctx->skip_loop_filter >= AVDISCARD_BIDIR && h->slice_type == B_TYPE)
+ ||(s->avctx->skip_loop_filter >= AVDISCARD_NONREF && h->nal_ref_idc == 0))
+ h->deblocking_filter= 0;
#if 0 //FMO
if( h->pps.num_slice_groups > 1 && h->pps.mb_slice_group_map_type >= 3 && h->pps.mb_slice_group_map_type <= 5)
h->slice_num++;
if(s->avctx->debug&FF_DEBUG_PICT_INFO){
- av_log(h->s.avctx, AV_LOG_DEBUG, "slice:%d mb:%d %c pps:%d frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d weight:%d%s\n",
- h->slice_num, first_mb_in_slice,
+ av_log(h->s.avctx, AV_LOG_DEBUG, "slice:%d %s mb:%d %c pps:%d frame:%d poc:%d/%d ref:%d/%d qp:%d loop:%d:%d:%d weight:%d%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),
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],
s->qscale,
- h->deblocking_filter,
+ h->deblocking_filter, h->slice_alpha_c0_offset/2, h->slice_beta_offset/2,
h->use_weight,
h->use_weight==1 && h->use_weight_chroma ? "c" : ""
);
static inline int get_level_prefix(GetBitContext *gb){
unsigned int buf;
int log;
-
+
OPEN_READER(re, gb);
UPDATE_CACHE(re, gb);
buf=GET_CACHE(re, gb);
-
+
log= 32 - av_log2(buf);
#ifdef TRACE
print_bin(buf>>(32-log), log);
return log-1;
}
+static inline int get_dct8x8_allowed(H264Context *h){
+ int i;
+ for(i=0; i<4; i++){
+ if(!IS_SUB_8X8(h->sub_mb_type[i])
+ || (!h->sps.direct_8x8_inference_flag && IS_DIRECT(h->sub_mb_type[i])))
+ return 0;
+ }
+ return 1;
+}
+
/**
* decodes a residual block.
* @param n block index
* @param max_coeff number of coefficients in the block
* @return <0 if an error occured
*/
-static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, int n, const uint8_t *scantable, int qp, int max_coeff){
+static int decode_residual(H264Context *h, GetBitContext *gb, DCTELEM *block, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff){
MpegEncContext * const s = &h->s;
- const uint16_t *qmul= dequant_coeff[qp];
static const int coeff_token_table_index[17]= {0, 0, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 3, 3, 3, 3, 3};
- int level[16], run[16];
- int suffix_length, zeros_left, coeff_num, coeff_token, total_coeff, i, trailing_ones;
+ int level[16];
+ int zeros_left, coeff_num, coeff_token, total_coeff, i, j, trailing_ones, run_before;
//FIXME put trailing_onex into the context
if(n == CHROMA_DC_BLOCK_INDEX){
coeff_token= get_vlc2(gb, chroma_dc_coeff_token_vlc.table, CHROMA_DC_COEFF_TOKEN_VLC_BITS, 1);
total_coeff= coeff_token>>2;
- }else{
+ }else{
if(n == LUMA_DC_BLOCK_INDEX){
total_coeff= pred_non_zero_count(h, 0);
coeff_token= get_vlc2(gb, coeff_token_vlc[ coeff_token_table_index[total_coeff] ].table, COEFF_TOKEN_VLC_BITS, 2);
if(total_coeff==0)
return 0;
-
+
trailing_ones= coeff_token&3;
tprintf("trailing:%d, total:%d\n", trailing_ones, total_coeff);
assert(total_coeff<=16);
-
+
for(i=0; i<trailing_ones; i++){
level[i]= 1 - 2*get_bits1(gb);
}
- suffix_length= total_coeff > 10 && trailing_ones < 3;
-
- for(; i<total_coeff; i++){
- const int prefix= get_level_prefix(gb);
+ if(i<total_coeff) {
int level_code, mask;
+ int suffix_length = total_coeff > 10 && trailing_ones < 3;
+ int prefix= get_level_prefix(gb);
+ //first coefficient has suffix_length equal to 0 or 1
if(prefix<14){ //FIXME try to build a large unified VLC table for all this
if(suffix_length)
level_code= (prefix<<suffix_length) + get_bits(gb, suffix_length); //part
level_code= prefix + get_bits(gb, 4); //part
}else if(prefix==15){
level_code= (prefix<<suffix_length) + get_bits(gb, 12); //part
- if(suffix_length==0) level_code+=15; //FIXME doesnt make (much)sense
+ if(suffix_length==0) level_code+=15; //FIXME doesn't make (much)sense
}else{
av_log(h->s.avctx, AV_LOG_ERROR, "prefix too large at %d %d\n", s->mb_x, s->mb_y);
return -1;
}
- if(i==trailing_ones && i<3) level_code+= 2; //FIXME split first iteration
+ if(trailing_ones < 3) level_code += 2;
+ suffix_length = 1;
+ if(level_code > 5)
+ suffix_length++;
mask= -(level_code&1);
level[i]= (((2+level_code)>>1) ^ mask) - mask;
-
- if(suffix_length==0) suffix_length=1; //FIXME split first iteration
-
-#if 1
- if(ABS(level[i]) > (3<<(suffix_length-1)) && suffix_length<6) suffix_length++;
-#else
- if((2+level_code)>>1) > (3<<(suffix_length-1)) && suffix_length<6) suffix_length++;
- /* ? == prefix > 2 or sth */
-#endif
- tprintf("level: %d suffix_length:%d\n", level[i], suffix_length);
+ i++;
+
+ //remaining coefficients have suffix_length > 0
+ for(;i<total_coeff;i++) {
+ static const int suffix_limit[7] = {0,5,11,23,47,95,INT_MAX };
+ prefix = get_level_prefix(gb);
+ if(prefix<15){
+ level_code = (prefix<<suffix_length) + get_bits(gb, suffix_length);
+ }else if(prefix==15){
+ level_code = (prefix<<suffix_length) + get_bits(gb, 12);
+ }else{
+ av_log(h->s.avctx, AV_LOG_ERROR, "prefix too large at %d %d\n", s->mb_x, s->mb_y);
+ return -1;
+ }
+ mask= -(level_code&1);
+ level[i]= (((2+level_code)>>1) ^ mask) - mask;
+ if(level_code > suffix_limit[suffix_length])
+ suffix_length++;
+ }
}
if(total_coeff == max_coeff)
else
zeros_left= get_vlc2(gb, total_zeros_vlc[ total_coeff-1 ].table, TOTAL_ZEROS_VLC_BITS, 1);
}
-
- for(i=0; i<total_coeff-1; i++){
- if(zeros_left <=0)
- break;
- else if(zeros_left < 7){
- run[i]= get_vlc2(gb, run_vlc[zeros_left-1].table, RUN_VLC_BITS, 1);
- }else{
- run[i]= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2);
- }
- zeros_left -= run[i];
- }
-
- if(zeros_left<0){
- av_log(h->s.avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", s->mb_x, s->mb_y);
- return -1;
- }
-
- for(; i<total_coeff-1; i++){
- run[i]= 0;
- }
- run[i]= zeros_left;
-
- coeff_num=-1;
+ coeff_num = zeros_left + total_coeff - 1;
+ j = scantable[coeff_num];
if(n > 24){
- for(i=total_coeff-1; i>=0; i--){ //FIXME merge into rundecode?
- int j;
-
- coeff_num += run[i] + 1; //FIXME add 1 earlier ?
+ block[j] = level[0];
+ for(i=1;i<total_coeff;i++) {
+ if(zeros_left <= 0)
+ run_before = 0;
+ else if(zeros_left < 7){
+ run_before= get_vlc2(gb, run_vlc[zeros_left-1].table, RUN_VLC_BITS, 1);
+ }else{
+ run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2);
+ }
+ zeros_left -= run_before;
+ coeff_num -= 1 + run_before;
j= scantable[ coeff_num ];
block[j]= level[i];
}
}else{
- for(i=total_coeff-1; i>=0; i--){ //FIXME merge into rundecode?
- int j;
-
- coeff_num += run[i] + 1; //FIXME add 1 earlier ?
+ block[j] = (level[0] * qmul[j] + 32)>>6;
+ for(i=1;i<total_coeff;i++) {
+ if(zeros_left <= 0)
+ run_before = 0;
+ else if(zeros_left < 7){
+ run_before= get_vlc2(gb, run_vlc[zeros_left-1].table, RUN_VLC_BITS, 1);
+ }else{
+ run_before= get_vlc2(gb, run7_vlc.table, RUN7_VLC_BITS, 2);
+ }
+ zeros_left -= run_before;
+ coeff_num -= 1 + run_before;
j= scantable[ coeff_num ];
- block[j]= level[i] * qmul[j];
-// printf("%d %d ", block[j], qmul[j]);
+ block[j]= (level[i] * qmul[j] + 32)>>6;
}
}
+
+ if(zeros_left<0){
+ av_log(h->s.avctx, AV_LOG_ERROR, "negative number of zero coeffs at %d %d\n", s->mb_x, s->mb_y);
+ return -1;
+ }
+
return 0;
}
static void decode_mb_skip(H264Context *h){
MpegEncContext * const s = &h->s;
const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
- int mb_type;
-
+ int mb_type=0;
+
memset(h->non_zero_count[mb_xy], 0, 16);
memset(h->non_zero_count_cache + 8, 0, 8*5); //FIXME ugly, remove pfui
+ if(h->mb_aff_frame && s->mb_skip_run==0 && (s->mb_y&1)==0){
+ h->mb_field_decoding_flag= get_bits1(&s->gb);
+ }
+ if(h->mb_field_decoding_flag)
+ mb_type|= MB_TYPE_INTERLACED;
+
if( h->slice_type == 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;
- //FIXME mbaff
+ mb_type|= MB_TYPE_16x16|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);
else
{
int mx, my;
- mb_type= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P1L0|MB_TYPE_SKIP;
+ mb_type|= MB_TYPE_16x16|MB_TYPE_P0L0|MB_TYPE_P1L0|MB_TYPE_SKIP;
- if(h->sps.mb_aff && s->mb_skip_run==0 && (s->mb_y&1)==0){
- h->mb_field_decoding_flag= get_bits1(&s->gb);
- }
- if(h->mb_field_decoding_flag)
- mb_type|= MB_TYPE_INTERLACED;
-
fill_caches(h, mb_type, 0); //FIXME check what is needed and what not ...
pred_pskip_motion(h, &mx, &my);
fill_rectangle(&h->ref_cache[0][scan8[0]], 4, 4, 8, 0, 1);
s->current_picture.mb_type[mb_xy]= mb_type|MB_TYPE_SKIP;
s->current_picture.qscale_table[mb_xy]= s->qscale;
h->slice_table[ mb_xy ]= h->slice_num;
- h->prev_mb_skiped= 1;
+ h->prev_mb_skipped= 1;
}
/**
MpegEncContext * const s = &h->s;
const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
int mb_type, partition_count, cbp;
+ int dct8x8_allowed= h->pps.transform_8x8_mode;
- s->dsp.clear_blocks(h->mb); //FIXME avoid if allready clear (move after skip handlong?
+ s->dsp.clear_blocks(h->mb); //FIXME avoid if already clear (move after skip handlong?
tprintf("pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
cbp = 0; /* avoid warning. FIXME: find a solution without slowing
if(h->slice_type != I_TYPE && h->slice_type != SI_TYPE){
if(s->mb_skip_run==-1)
s->mb_skip_run= get_ue_golomb(&s->gb);
-
+
if (s->mb_skip_run--) {
decode_mb_skip(h);
return 0;
}
}
- if(h->sps.mb_aff /* && !field pic FIXME needed? */){
- if((s->mb_y&1)==0)
+ if(h->mb_aff_frame){
+ if ( ((s->mb_y&1) == 0) || h->prev_mb_skipped)
h->mb_field_decoding_flag = get_bits1(&s->gb);
}else
- h->mb_field_decoding_flag=0; //FIXME som ed note ?!
-
- h->prev_mb_skiped= 0;
-
+ h->mb_field_decoding_flag= (s->picture_structure!=PICT_FRAME);
+
+ h->prev_mb_skipped= 0;
+
mb_type= get_ue_golomb(&s->gb);
if(h->slice_type == B_TYPE){
if(mb_type < 23){
if(h->mb_field_decoding_flag)
mb_type |= MB_TYPE_INTERLACED;
- s->current_picture.mb_type[mb_xy]= mb_type;
h->slice_table[ mb_xy ]= h->slice_num;
-
+
if(IS_INTRA_PCM(mb_type)){
unsigned int x, y;
-
+
// we assume these blocks are very rare so we dont 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);
h->mb[index + (x&3) + 16*(x>>2)]= get_bits(&s->gb, 8);
}
}
-
- // In deblocking, the quantiser is 0
+
+ // In deblocking, the quantizer is 0
s->current_picture.qscale_table[mb_xy]= 0;
- h->chroma_qp = get_chroma_qp(h, 0);
- // All coeffs are presents
+ h->chroma_qp = get_chroma_qp(h->pps.chroma_qp_index_offset, 0);
+ // All coeffs are present
memset(h->non_zero_count[mb_xy], 16, 16);
-
+
+ s->current_picture.mb_type[mb_xy]= mb_type;
return 0;
}
-
+
fill_caches(h, mb_type, 0);
//mb_pred
// 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;
+ }
// fill_intra4x4_pred_table(h);
- for(i=0; i<16; i++){
+ for(i=0; i<16; i+=di){
const int mode_coded= !get_bits1(&s->gb);
const int predicted_mode= pred_intra_mode(h, i);
int mode;
}else{
mode= predicted_mode;
}
-
- h->intra4x4_pred_mode_cache[ scan8[i] ] = mode;
+
+ 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 if(partition_count==4){
int i, j, sub_partition_count[4], list, ref[2][4];
-
+
if(h->slice_type == B_TYPE){
for(i=0; i<4; i++){
h->sub_mb_type[i]= get_ue_golomb(&s->gb);
h->sub_mb_type[i]= b_sub_mb_type_info[ h->sub_mb_type[i] ].type;
}
if( IS_DIRECT(h->sub_mb_type[0]) || IS_DIRECT(h->sub_mb_type[1])
- || IS_DIRECT(h->sub_mb_type[2]) || IS_DIRECT(h->sub_mb_type[3]))
+ || IS_DIRECT(h->sub_mb_type[2]) || IS_DIRECT(h->sub_mb_type[3])) {
pred_direct_motion(h, &mb_type);
+ h->ref_cache[0][scan8[4]] =
+ h->ref_cache[1][scan8[4]] =
+ h->ref_cache[0][scan8[12]] =
+ h->ref_cache[1][scan8[12]] = PART_NOT_AVAILABLE;
+ }
}else{
assert(h->slice_type == P_TYPE || h->slice_type == SP_TYPE); //FIXME SP correct ?
for(i=0; i<4; i++){
h->sub_mb_type[i]= p_sub_mb_type_info[ h->sub_mb_type[i] ].type;
}
}
-
+
for(list=0; list<2; list++){
- const int ref_count= IS_REF0(mb_type) ? 1 : h->ref_count[list];
+ int ref_count= IS_REF0(mb_type) ? 1 : h->ref_count[list];
if(ref_count == 0) continue;
+ if (h->mb_aff_frame && h->mb_field_decoding_flag) {
+ ref_count <<= 1;
+ }
for(i=0; i<4; i++){
if(IS_DIRECT(h->sub_mb_type[i])) continue;
if(IS_DIR(h->sub_mb_type[i], 0, list)){
}
}
}
-
+
+ if(dct8x8_allowed)
+ dct8x8_allowed = get_dct8x8_allowed(h);
+
for(list=0; list<2; list++){
const int ref_count= IS_REF0(mb_type) ? 1 : h->ref_count[list];
if(ref_count == 0) continue;
for(i=0; i<4; i++){
- if(IS_DIRECT(h->sub_mb_type[i])) continue;
+ if(IS_DIRECT(h->sub_mb_type[i])) {
+ h->ref_cache[list][ scan8[4*i] ] = h->ref_cache[list][ scan8[4*i]+1 ];
+ continue;
+ }
h->ref_cache[list][ scan8[4*i] ]=h->ref_cache[list][ scan8[4*i]+1 ]=
h->ref_cache[list][ scan8[4*i]+8 ]=h->ref_cache[list][ scan8[4*i]+9 ]= ref[list][i];
tprintf("final mv:%d %d\n", mx, my);
if(IS_SUB_8X8(sub_mb_type)){
- mv_cache[ 0 ][0]= mv_cache[ 1 ][0]=
+ mv_cache[ 0 ][0]= mv_cache[ 1 ][0]=
mv_cache[ 8 ][0]= mv_cache[ 9 ][0]= mx;
- mv_cache[ 0 ][1]= mv_cache[ 1 ][1]=
+ mv_cache[ 0 ][1]= mv_cache[ 1 ][1]=
mv_cache[ 8 ][1]= mv_cache[ 9 ][1]= my;
}else if(IS_SUB_8X4(sub_mb_type)){
mv_cache[ 0 ][0]= mv_cache[ 1 ][0]= mx;
}
}else if(IS_DIRECT(mb_type)){
pred_direct_motion(h, &mb_type);
- s->current_picture.mb_type[mb_xy]= mb_type;
+ dct8x8_allowed &= h->sps.direct_8x8_inference_flag;
}else{
int list, mx, my, i;
//FIXME we should set ref_idx_l? to 0 if we use that later ...
if(IS_DIR(mb_type, 0, list)){
const int val= get_te0_golomb(&s->gb, h->ref_count[list]);
fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, val, 1);
- }
+ }else
+ fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, (LIST_NOT_USED&0xFF), 1);
}
}
for(list=0; list<2; list++){
tprintf("final mv:%d %d\n", mx, my);
fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
- }
+ }else
+ fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, 0, 4);
}
}
else if(IS_16X8(mb_type)){
if(IS_DIR(mb_type, i, list)){
const int val= get_te0_golomb(&s->gb, h->ref_count[list]);
fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, val, 1);
- }else // needed only for mixed refs (e.g. B_L0_L1_16x8)
+ }else
fill_rectangle(&h->ref_cache[list][ scan8[0] + 16*i ], 4, 2, 8, (LIST_NOT_USED&0xFF), 1);
}
}
if(IS_DIR(mb_type, i, list)){ //FIXME optimize
const int val= get_te0_golomb(&s->gb, h->ref_count[list]);
fill_rectangle(&h->ref_cache[list][ scan8[0] + 2*i ], 2, 4, 8, val, 1);
- }else // needed only for mixed refs
+ }else
fill_rectangle(&h->ref_cache[list][ scan8[0] + 2*i ], 2, 4, 8, (LIST_NOT_USED&0xFF), 1);
}
}
}
}
}
-
+
if(IS_INTER(mb_type))
write_back_motion(h, mb_type);
-
+
if(!IS_INTRA16x16(mb_type)){
cbp= get_ue_golomb(&s->gb);
if(cbp > 47){
av_log(h->s.avctx, AV_LOG_ERROR, "cbp too large (%d) at %d %d\n", cbp, s->mb_x, s->mb_y);
return -1;
}
-
+
if(IS_INTRA4x4(mb_type))
cbp= golomb_to_intra4x4_cbp[cbp];
else
cbp= golomb_to_inter_cbp[cbp];
}
+ if(dct8x8_allowed && (cbp&15) && !IS_INTRA(mb_type)){
+ if(get_bits1(&s->gb))
+ mb_type |= MB_TYPE_8x8DCT;
+ }
+ s->current_picture.mb_type[mb_xy]= mb_type;
+
if(cbp || IS_INTRA16x16(mb_type)){
int i8x8, i4x4, chroma_idx;
int chroma_qp, dquant;
GetBitContext *gb= IS_INTRA(mb_type) ? h->intra_gb_ptr : h->inter_gb_ptr;
const uint8_t *scan, *dc_scan;
-
+
// fill_non_zero_count_cache(h);
if(IS_INTERLACED(mb_type)){
- scan= field_scan;
+ scan= s->qscale ? h->field_scan : h->field_scan_q0;
dc_scan= luma_dc_field_scan;
}else{
- scan= zigzag_scan;
+ scan= s->qscale ? h->zigzag_scan : h->zigzag_scan_q0;
dc_scan= luma_dc_zigzag_scan;
}
av_log(h->s.avctx, AV_LOG_ERROR, "dquant out of range (%d) at %d %d\n", dquant, s->mb_x, s->mb_y);
return -1;
}
-
+
s->qscale += dquant;
if(((unsigned)s->qscale) > 51){
if(s->qscale<0) s->qscale+= 52;
else s->qscale-= 52;
}
-
- h->chroma_qp= chroma_qp= get_chroma_qp(h, s->qscale);
+
+ h->chroma_qp= chroma_qp= get_chroma_qp(h->pps.chroma_qp_index_offset, s->qscale);
if(IS_INTRA16x16(mb_type)){
- if( decode_residual(h, h->intra_gb_ptr, h->mb, LUMA_DC_BLOCK_INDEX, dc_scan, s->qscale, 16) < 0){
- return -1; //FIXME continue if partotioned and other retirn -1 too
+ if( decode_residual(h, h->intra_gb_ptr, h->mb, LUMA_DC_BLOCK_INDEX, dc_scan, h->dequant4_coeff[0][s->qscale], 16) < 0){
+ return -1; //FIXME continue if partitioned and other return -1 too
}
assert((cbp&15) == 0 || (cbp&15) == 15);
for(i8x8=0; i8x8<4; i8x8++){
for(i4x4=0; i4x4<4; i4x4++){
const int index= i4x4 + 4*i8x8;
- if( decode_residual(h, h->intra_gb_ptr, h->mb + 16*index, index, scan + 1, s->qscale, 15) < 0 ){
+ if( decode_residual(h, h->intra_gb_ptr, h->mb + 16*index, index, scan + 1, h->dequant4_coeff[0][s->qscale], 15) < 0 ){
return -1;
}
}
}else{
for(i8x8=0; i8x8<4; i8x8++){
if(cbp & (1<<i8x8)){
- for(i4x4=0; i4x4<4; i4x4++){
- const int index= i4x4 + 4*i8x8;
-
- if( decode_residual(h, gb, h->mb + 16*index, index, scan, s->qscale, 16) <0 ){
- return -1;
+ if(IS_8x8DCT(mb_type)){
+ DCTELEM *buf = &h->mb[64*i8x8];
+ uint8_t *nnz;
+ for(i4x4=0; i4x4<4; i4x4++){
+ if( decode_residual(h, gb, buf, i4x4+4*i8x8, zigzag_scan8x8_cavlc+16*i4x4,
+ h->dequant8_coeff[IS_INTRA( mb_type ) ? 0:1][s->qscale], 16) <0 )
+ return -1;
+ }
+ nnz= &h->non_zero_count_cache[ scan8[4*i8x8] ];
+ nnz[0] |= nnz[1] | nnz[8] | nnz[9];
+ }else{
+ for(i4x4=0; i4x4<4; i4x4++){
+ const int index= i4x4 + 4*i8x8;
+
+ if( decode_residual(h, gb, h->mb + 16*index, index, scan, h->dequant4_coeff[IS_INTRA( mb_type ) ? 0:3][s->qscale], 16) <0 ){
+ return -1;
+ }
}
}
}else{
}
}
}
-
+
if(cbp&0x30){
for(chroma_idx=0; chroma_idx<2; chroma_idx++)
- if( decode_residual(h, gb, h->mb + 256 + 16*4*chroma_idx, CHROMA_DC_BLOCK_INDEX, chroma_dc_scan, chroma_qp, 4) < 0){
+ if( decode_residual(h, gb, h->mb + 256 + 16*4*chroma_idx, CHROMA_DC_BLOCK_INDEX, chroma_dc_scan, NULL, 4) < 0){
return -1;
}
}
for(chroma_idx=0; chroma_idx<2; chroma_idx++){
for(i4x4=0; i4x4<4; i4x4++){
const int index= 16 + 4*chroma_idx + i4x4;
- if( decode_residual(h, gb, h->mb + 16*index, index, scan + 1, chroma_qp, 15) < 0){
+ if( decode_residual(h, gb, h->mb + 16*index, index, scan + 1, h->dequant4_coeff[chroma_idx+1+(IS_INTRA( mb_type ) ? 0:3)][chroma_qp], 15) < 0){
return -1;
}
}
return 0;
}
+static int decode_cabac_field_decoding_flag(H264Context *h) {
+ MpegEncContext * const s = &h->s;
+ const int mb_x = s->mb_x;
+ const int mb_y = s->mb_y & ~1;
+ const int mba_xy = mb_x - 1 + mb_y *s->mb_stride;
+ const int mbb_xy = mb_x + (mb_y-2)*s->mb_stride;
+
+ unsigned int ctx = 0;
+
+ if( h->slice_table[mba_xy] == h->slice_num && IS_INTERLACED( s->current_picture.mb_type[mba_xy] ) ) {
+ ctx += 1;
+ }
+ if( h->slice_table[mbb_xy] == h->slice_num && IS_INTERLACED( s->current_picture.mb_type[mbb_xy] ) ) {
+ ctx += 1;
+ }
+
+ return get_cabac( &h->cabac, &h->cabac_state[70 + ctx] );
+}
+
static int decode_cabac_intra_mb_type(H264Context *h, int ctx_base, int intra_slice) {
uint8_t *state= &h->cabac_state[ctx_base];
int mb_type;
-
+
if(intra_slice){
MpegEncContext * const s = &h->s;
- const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
- const int mba_xy = mb_xy - 1;
- const int mbb_xy = mb_xy - s->mb_stride;
+ const int mba_xy = h->left_mb_xy[0];
+ const int mbb_xy = h->top_mb_xy;
int ctx=0;
if( h->slice_table[mba_xy] == h->slice_num && !IS_INTRA4x4( s->current_picture.mb_type[mba_xy] ) )
ctx++;
return decode_cabac_intra_mb_type(h, 17, 0) + 5;
}
} else if( h->slice_type == B_TYPE ) {
- const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
- const int mba_xy = mb_xy - 1;
- const int mbb_xy = mb_xy - s->mb_stride;
+ const int mba_xy = h->left_mb_xy[0];
+ const int mbb_xy = h->top_mb_xy;
int ctx = 0;
int bits;
if( get_cabac( &h->cabac, &h->cabac_state[68] ) )
return pred_mode;
- if( get_cabac( &h->cabac, &h->cabac_state[69] ) )
- mode += 1;
- if( get_cabac( &h->cabac, &h->cabac_state[69] ) )
- mode += 2;
- if( get_cabac( &h->cabac, &h->cabac_state[69] ) )
- mode += 4;
+ mode += 1 * get_cabac( &h->cabac, &h->cabac_state[69] );
+ mode += 2 * get_cabac( &h->cabac, &h->cabac_state[69] );
+ mode += 4 * get_cabac( &h->cabac, &h->cabac_state[69] );
+
if( mode >= pred_mode )
return mode + 1;
else
}
static int decode_cabac_mb_chroma_pre_mode( H264Context *h) {
- MpegEncContext * const s = &h->s;
- const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
- const int mba_xy = mb_xy - 1;
- const int mbb_xy = mb_xy - s->mb_stride;
+ const int mba_xy = h->left_mb_xy[0];
+ const int mbb_xy = h->top_mb_xy;
int ctx = 0;
static int decode_cabac_mb_cbp_luma( H264Context *h) {
MpegEncContext * const s = &h->s;
- const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
int cbp = 0;
int i8x8;
- h->cbp_table[mb_xy] = 0; /* FIXME aaahahahah beurk */
-
for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
- int mba_xy = -1;
- int mbb_xy = -1;
+ int cbp_a = -1;
+ int cbp_b = -1;
int x, y;
int ctx = 0;
y = block_idx_y[4*i8x8];
if( x > 0 )
- mba_xy = mb_xy;
- else if( s->mb_x > 0 ) {
- mba_xy = mb_xy - 1;
- if (h->slice_table[mba_xy] != h->slice_num) {
- mba_xy = -1;
- }
+ cbp_a = cbp;
+ else if( s->mb_x > 0 && (h->slice_table[h->left_mb_xy[0]] == h->slice_num)) {
+ cbp_a = h->left_cbp;
+ tprintf("cbp_a = left_cbp = %x\n", cbp_a);
}
if( y > 0 )
- mbb_xy = mb_xy;
- else if( s->mb_y > 0 ) {
- mbb_xy = mb_xy - s->mb_stride;
- if (h->slice_table[mbb_xy] != h->slice_num) {
- mbb_xy = -1;
- }
+ cbp_b = cbp;
+ else if( s->mb_y > 0 && (h->slice_table[h->top_mb_xy] == h->slice_num)) {
+ cbp_b = h->top_cbp;
+ tprintf("cbp_b = top_cbp = %x\n", cbp_b);
}
/* No need to test for skip as we put 0 for skip block */
- if( mba_xy >= 0 ) {
+ /* No need to test for IPCM as we put 1 for IPCM block */
+ if( cbp_a >= 0 ) {
int i8x8a = block_idx_xy[(x-1)&0x03][y]/4;
- if( ((h->cbp_table[mba_xy] >> i8x8a)&0x01) == 0 )
+ if( ((cbp_a >> i8x8a)&0x01) == 0 )
ctx++;
}
- if( mbb_xy >= 0 ) {
+ if( cbp_b >= 0 ) {
int i8x8b = block_idx_xy[x][(y-1)&0x03]/4;
- if( ((h->cbp_table[mbb_xy] >> i8x8b)&0x01) == 0 )
+ if( ((cbp_b >> i8x8b)&0x01) == 0 )
ctx += 2;
}
if( get_cabac( &h->cabac, &h->cabac_state[73 + ctx] ) ) {
cbp |= 1 << i8x8;
- h->cbp_table[mb_xy] = cbp; /* FIXME aaahahahah beurk */
}
}
return cbp;
else
ctx = 3;
val++;
+ if(val > 52) //prevent infinite loop
+ return INT_MIN;
}
if( val&0x01 )
return type;
}
+static inline int decode_cabac_mb_transform_size( H264Context *h ) {
+ return get_cabac( &h->cabac, &h->cabac_state[399 + h->neighbor_transform_size] );
+}
+
static int decode_cabac_mb_ref( H264Context *h, int list, int n ) {
int refa = h->ref_cache[list][scan8[n] - 1];
int refb = h->ref_cache[list][scan8[n] - 8];
return ctx + 4 * cat;
}
-static int inline decode_cabac_residual( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, int qp, int max_coeff) {
+static int decode_cabac_residual( H264Context *h, DCTELEM *block, int cat, int n, const uint8_t *scantable, const uint32_t *qmul, int max_coeff) {
const int mb_xy = h->s.mb_x + h->s.mb_y*h->s.mb_stride;
- const uint16_t *qmul= dequant_coeff[qp];
- static const int significant_coeff_flag_offset[5] = { 0, 15, 29, 44, 47 };
- static const int coeff_abs_level_m1_offset[5] = {227+ 0, 227+10, 227+20, 227+30, 227+39 };
-
- int index[16];
+ static const int significant_coeff_flag_field_offset[2] = { 105, 277 };
+ static const int last_significant_coeff_flag_field_offset[2] = { 166, 338 };
+ static const int significant_coeff_flag_offset[6] = { 0, 15, 29, 44, 47, 297 };
+ static const int last_significant_coeff_flag_offset[6] = { 0, 15, 29, 44, 47, 251 };
+ static const int coeff_abs_level_m1_offset[6] = { 227+0, 227+10, 227+20, 227+30, 227+39, 426 };
+ static const int significant_coeff_flag_offset_8x8[63] = {
+ 0, 1, 2, 3, 4, 5, 5, 4, 4, 3, 3, 4, 4, 4, 5, 5,
+ 4, 4, 4, 4, 3, 3, 6, 7, 7, 7, 8, 9,10, 9, 8, 7,
+ 7, 6,11,12,13,11, 6, 7, 8, 9,14,10, 9, 8, 6,11,
+ 12,13,11, 6, 9,14,10, 9,11,12,13,11,14,10,12
+ };
+ static const int last_coeff_flag_offset_8x8[63] = {
+ 0, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1, 1,
+ 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2, 2,
+ 3, 3, 3, 3, 3, 3, 3, 3, 4, 4, 4, 4, 4, 4, 4, 4,
+ 5, 5, 5, 5, 6, 6, 6, 6, 7, 7, 7, 7, 8, 8, 8
+ };
+
+ int index[64];
int i, last;
int coeff_count = 0;
int abslevel1 = 1;
int abslevelgt1 = 0;
+ uint8_t *significant_coeff_ctx_base;
+ uint8_t *last_coeff_ctx_base;
+ uint8_t *abs_level_m1_ctx_base;
+
/* cat: 0-> DC 16x16 n = 0
* 1-> AC 16x16 n = luma4x4idx
* 2-> Luma4x4 n = luma4x4idx
* 3-> DC Chroma n = iCbCr
* 4-> AC Chroma n = 4 * iCbCr + chroma4x4idx
+ * 5-> Luma8x8 n = 4 * luma8x8idx
*/
/* read coded block flag */
- if( get_cabac( &h->cabac, &h->cabac_state[85 + get_cabac_cbf_ctx( h, cat, n ) ] ) == 0 ) {
- if( cat == 1 || cat == 2 )
- h->non_zero_count_cache[scan8[n]] = 0;
- else if( cat == 4 )
- h->non_zero_count_cache[scan8[16+n]] = 0;
+ if( cat != 5 ) {
+ if( get_cabac( &h->cabac, &h->cabac_state[85 + get_cabac_cbf_ctx( h, cat, n ) ] ) == 0 ) {
+ if( cat == 1 || cat == 2 )
+ h->non_zero_count_cache[scan8[n]] = 0;
+ else if( cat == 4 )
+ h->non_zero_count_cache[scan8[16+n]] = 0;
- return 0;
+ return 0;
+ }
}
- for(last= 0; last < max_coeff - 1; last++) {
- if( get_cabac( &h->cabac, &h->cabac_state[105+significant_coeff_flag_offset[cat]+last] )) {
- index[coeff_count++] = last;
- if( get_cabac( &h->cabac, &h->cabac_state[166+significant_coeff_flag_offset[cat]+last] ) ) {
- last= max_coeff;
- break;
- }
- }
+ significant_coeff_ctx_base = h->cabac_state
+ + significant_coeff_flag_offset[cat]
+ + significant_coeff_flag_field_offset[h->mb_field_decoding_flag];
+ last_coeff_ctx_base = h->cabac_state
+ + last_significant_coeff_flag_offset[cat]
+ + last_significant_coeff_flag_field_offset[h->mb_field_decoding_flag];
+ abs_level_m1_ctx_base = h->cabac_state
+ + coeff_abs_level_m1_offset[cat];
+
+ if( cat == 5 ) {
+#define DECODE_SIGNIFICANCE( coefs, sig_off, last_off ) \
+ for(last= 0; last < coefs; last++) { \
+ uint8_t *sig_ctx = significant_coeff_ctx_base + sig_off; \
+ if( get_cabac( &h->cabac, sig_ctx )) { \
+ uint8_t *last_ctx = last_coeff_ctx_base + last_off; \
+ index[coeff_count++] = last; \
+ if( get_cabac( &h->cabac, last_ctx ) ) { \
+ last= max_coeff; \
+ break; \
+ } \
+ } \
+ }
+ DECODE_SIGNIFICANCE( 63, significant_coeff_flag_offset_8x8[last],
+ last_coeff_flag_offset_8x8[last] );
+ } else {
+ DECODE_SIGNIFICANCE( max_coeff - 1, last, last );
}
if( last == max_coeff -1 ) {
index[coeff_count++] = last;
h->non_zero_count_cache[scan8[n]] = coeff_count;
else if( cat == 3 )
h->cbp_table[mb_xy] |= 0x40 << n;
- else {
- assert( cat == 4 );
+ else if( cat == 4 )
h->non_zero_count_cache[scan8[16+n]] = coeff_count;
+ else {
+ assert( cat == 5 );
+ fill_rectangle(&h->non_zero_count_cache[scan8[n]], 2, 2, 8, 1, 1);
}
for( i = coeff_count - 1; i >= 0; i-- ) {
- int ctx = (abslevelgt1 != 0 ? 0 : FFMIN( 4, abslevel1 )) + coeff_abs_level_m1_offset[cat];
+ uint8_t *ctx = (abslevelgt1 != 0 ? 0 : FFMIN( 4, abslevel1 )) + abs_level_m1_ctx_base;
int j= scantable[index[i]];
- if( get_cabac( &h->cabac, &h->cabac_state[ctx] ) == 0 ) {
- if( cat == 0 || cat == 3 ) {
+ if( get_cabac( &h->cabac, ctx ) == 0 ) {
+ if( !qmul ) {
if( get_cabac_bypass( &h->cabac ) ) block[j] = -1;
else block[j] = 1;
}else{
- if( get_cabac_bypass( &h->cabac ) ) block[j] = -qmul[j];
- else block[j] = qmul[j];
+ if( get_cabac_bypass( &h->cabac ) ) block[j] = (-qmul[j] + 32) >> 6;
+ else block[j] = ( qmul[j] + 32) >> 6;
}
-
+
abslevel1++;
} else {
int coeff_abs = 2;
- ctx = 5 + FFMIN( 4, abslevelgt1 ) + coeff_abs_level_m1_offset[cat];
- while( coeff_abs < 15 && get_cabac( &h->cabac, &h->cabac_state[ctx] ) ) {
+ ctx = 5 + FFMIN( 4, abslevelgt1 ) + abs_level_m1_ctx_base;
+ while( coeff_abs < 15 && get_cabac( &h->cabac, ctx ) ) {
coeff_abs++;
}
coeff_abs += 1 << j;
j++;
}
-
+
while( j-- ) {
if( get_cabac_bypass( &h->cabac ) )
coeff_abs += 1 << j ;
}
}
- if( cat == 0 || cat == 3 ) {
+ if( !qmul ) {
if( get_cabac_bypass( &h->cabac ) ) block[j] = -coeff_abs;
else block[j] = coeff_abs;
}else{
- if( get_cabac_bypass( &h->cabac ) ) block[j] = -coeff_abs * qmul[j];
- else block[j] = coeff_abs * qmul[j];
+ if( get_cabac_bypass( &h->cabac ) ) block[j] = (-coeff_abs * qmul[j] + 32) >> 6;
+ else block[j] = ( coeff_abs * qmul[j] + 32) >> 6;
}
-
+
abslevelgt1++;
}
}
return 0;
}
+void inline compute_mb_neighboors(H264Context *h)
+{
+ MpegEncContext * const s = &h->s;
+ const int mb_xy = s->mb_x + s->mb_y*s->mb_stride;
+ h->top_mb_xy = mb_xy - s->mb_stride;
+ h->left_mb_xy[0] = mb_xy - 1;
+ if(h->mb_aff_frame){
+ const int pair_xy = s->mb_x + (s->mb_y & ~1)*s->mb_stride;
+ const int top_pair_xy = pair_xy - s->mb_stride;
+ const int top_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[top_pair_xy]);
+ const int left_mb_frame_flag = !IS_INTERLACED(s->current_picture.mb_type[pair_xy-1]);
+ const int curr_mb_frame_flag = !h->mb_field_decoding_flag;
+ const int bottom = (s->mb_y & 1);
+ if (bottom
+ ? !curr_mb_frame_flag // bottom macroblock
+ : (!curr_mb_frame_flag && !top_mb_frame_flag) // top macroblock
+ ) {
+ h->top_mb_xy -= s->mb_stride;
+ }
+ if (left_mb_frame_flag != curr_mb_frame_flag) {
+ h->left_mb_xy[0] = pair_xy - 1;
+ }
+ }
+ return;
+}
+
/**
* decodes a macroblock
* @returns 0 if ok, AC_ERROR / DC_ERROR / MV_ERROR if an error is noticed
MpegEncContext * const s = &h->s;
const int mb_xy= s->mb_x + s->mb_y*s->mb_stride;
int mb_type, partition_count, cbp = 0;
+ int dct8x8_allowed= h->pps.transform_8x8_mode;
- s->dsp.clear_blocks(h->mb); //FIXME avoid if allready clear (move after skip handlong?)
-
- if( h->sps.mb_aff ) {
- av_log( h->s.avctx, AV_LOG_ERROR, "Fields not supported with CABAC\n" );
- return -1;
- }
+ s->dsp.clear_blocks(h->mb); //FIXME avoid if already clear (move after skip handlong?)
tprintf("pic:%d mb:%d/%d\n", h->frame_num, s->mb_x, s->mb_y);
if( h->slice_type != I_TYPE && h->slice_type != SI_TYPE ) {
}
}
- h->prev_mb_skiped = 0;
+ if(h->mb_aff_frame){
+ if ( ((s->mb_y&1) == 0) || h->prev_mb_skipped)
+ 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;
+ compute_mb_neighboors(h);
if( ( mb_type = decode_cabac_mb_type( h ) ) < 0 ) {
av_log( h->s.avctx, AV_LOG_ERROR, "decode_cabac_mb_type failed\n" );
return -1;
h->intra16x16_pred_mode= i_mb_type_info[mb_type].pred_mode;
mb_type= i_mb_type_info[mb_type].type;
}
-#if 0
if(h->mb_field_decoding_flag)
mb_type |= MB_TYPE_INTERLACED;
-#endif
- s->current_picture.mb_type[mb_xy]= mb_type;
h->slice_table[ mb_xy ]= h->slice_num;
if(IS_INTRA_PCM(mb_type)) {
const uint8_t *ptr;
unsigned int x, y;
-
+
// We assume these blocks are very rare so we dont optimize it.
// FIXME The two following lines get the bitstream position in the cabac
// decode, I think it should be done by a function in cabac.h (or cabac.c).
ff_init_cabac_decoder(&h->cabac, ptr, h->cabac.bytestream_end - ptr);
- // All blocks are presents
+ // All blocks are present
h->cbp_table[mb_xy] = 0x1ef;
h->chroma_pred_mode_table[mb_xy] = 0;
- // In deblocking, the quantiser is 0
+ // In deblocking, the quantizer is 0
s->current_picture.qscale_table[mb_xy]= 0;
- h->chroma_qp = get_chroma_qp(h, 0);
- // All coeffs are presents
+ h->chroma_qp = get_chroma_qp(h->pps.chroma_qp_index_offset, 0);
+ // All coeffs are present
memset(h->non_zero_count[mb_xy], 16, 16);
+ s->current_picture.mb_type[mb_xy]= mb_type;
return 0;
}
fill_caches(h, mb_type, 0);
if( IS_INTRA( mb_type ) ) {
+ int i;
if( IS_INTRA4x4( mb_type ) ) {
- int i;
- for( i = 0; i < 16; i++ ) {
- int pred = pred_intra_mode( h, i );
- h->intra4x4_pred_mode_cache[ scan8[i] ] = decode_cabac_mb_intra4x4_pred_mode( h, pred );
+ if( dct8x8_allowed && decode_cabac_mb_transform_size( h ) ) {
+ mb_type |= MB_TYPE_8x8DCT;
+ for( i = 0; i < 16; i+=4 ) {
+ int pred = pred_intra_mode( h, i );
+ int mode = decode_cabac_mb_intra4x4_pred_mode( h, pred );
+ fill_rectangle( &h->intra4x4_pred_mode_cache[ scan8[i] ], 2, 2, 8, mode, 1 );
+ }
+ } else {
+ for( i = 0; i < 16; i++ ) {
+ int pred = pred_intra_mode( h, i );
+ h->intra4x4_pred_mode_cache[ scan8[i] ] = decode_cabac_mb_intra4x4_pred_mode( h, pred );
//av_log( s->avctx, AV_LOG_ERROR, "i4x4 pred=%d mode=%d\n", pred, h->intra4x4_pred_mode_cache[ scan8[i] ] );
+ }
}
write_back_intra_pred_mode(h);
if( check_intra4x4_pred_mode(h) < 0 ) return -1;
}
}
+ if(dct8x8_allowed)
+ dct8x8_allowed = get_dct8x8_allowed(h);
+
for(list=0; list<2; list++){
for(i=0; i<4; i++){
if(IS_DIRECT(h->sub_mb_type[i])){
}
} else if( IS_DIRECT(mb_type) ) {
pred_direct_motion(h, &mb_type);
- s->current_picture.mb_type[mb_xy]= mb_type;
fill_rectangle(h->mvd_cache[0][scan8[0]], 4, 4, 8, 0, 4);
fill_rectangle(h->mvd_cache[1][scan8[0]], 4, 4, 8, 0, 4);
+ dct8x8_allowed &= h->sps.direct_8x8_inference_flag;
} else {
int list, mx, my, i, mpx, mpy;
if(IS_16X16(mb_type)){
const int ref = h->ref_count[list] > 1 ? decode_cabac_mb_ref( h, list, 0 ) : 0;
fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, ref, 1);
}
- }
+ }else
+ fill_rectangle(&h->ref_cache[list][ scan8[0] ], 4, 4, 8, (uint8_t)LIST_NOT_USED, 1);
}
for(list=0; list<2; list++){
if(IS_DIR(mb_type, 0, list)){
fill_rectangle(h->mvd_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx-mpx,my-mpy), 4);
fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, pack16to32(mx,my), 4);
- }
+ }else
+ fill_rectangle(h->mv_cache[list][ scan8[0] ], 4, 4, 8, 0, 4);
}
}
else if(IS_16X8(mb_type)){
fill_rectangle(h->mvd_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack16to32(mx-mpx,my-mpy), 4);
fill_rectangle(h->mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, pack16to32(mx,my), 4);
- }else{ // needed only for mixed refs
+ }else{
fill_rectangle(h->mvd_cache[list][ scan8[0] + 16*i ], 4, 2, 8, 0, 4);
fill_rectangle(h-> mv_cache[list][ scan8[0] + 16*i ], 4, 2, 8, 0, 4);
}
tprintf("final mv:%d %d\n", mx, my);
fill_rectangle(h->mvd_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack16to32(mx-mpx,my-mpy), 4);
fill_rectangle(h->mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, pack16to32(mx,my), 4);
- }else{ // needed only for mixed refs
+ }else{
fill_rectangle(h->mvd_cache[list][ scan8[0] + 2*i ], 2, 4, 8, 0, 4);
fill_rectangle(h-> mv_cache[list][ scan8[0] + 2*i ], 2, 4, 8, 0, 4);
}
h->cbp_table[mb_xy] = cbp;
+ if( dct8x8_allowed && (cbp&15) && !IS_INTRA( mb_type ) ) {
+ if( decode_cabac_mb_transform_size( h ) )
+ mb_type |= MB_TYPE_8x8DCT;
+ }
+ s->current_picture.mb_type[mb_xy]= mb_type;
+
if( cbp || IS_INTRA16x16( mb_type ) ) {
const uint8_t *scan, *dc_scan;
int dqp;
if(IS_INTERLACED(mb_type)){
- scan= field_scan;
+ scan= s->qscale ? h->field_scan : h->field_scan_q0;
dc_scan= luma_dc_field_scan;
}else{
- scan= zigzag_scan;
+ scan= s->qscale ? h->zigzag_scan : h->zigzag_scan_q0;
dc_scan= luma_dc_zigzag_scan;
}
h->last_qscale_diff = dqp = decode_cabac_mb_dqp( h );
+ if( dqp == INT_MIN ){
+ av_log(h->s.avctx, AV_LOG_ERROR, "cabac decode of qscale diff failed at %d %d\n", s->mb_x, s->mb_y);
+ return -1;
+ }
s->qscale += dqp;
if(((unsigned)s->qscale) > 51){
if(s->qscale<0) s->qscale+= 52;
else s->qscale-= 52;
}
- h->chroma_qp = get_chroma_qp(h, s->qscale);
+ h->chroma_qp = get_chroma_qp(h->pps.chroma_qp_index_offset, s->qscale);
if( IS_INTRA16x16( mb_type ) ) {
int i;
//av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 DC\n" );
- if( decode_cabac_residual( h, h->mb, 0, 0, dc_scan, s->qscale, 16) < 0)
+ if( decode_cabac_residual( h, h->mb, 0, 0, dc_scan, NULL, 16) < 0)
return -1;
if( cbp&15 ) {
for( i = 0; i < 16; i++ ) {
//av_log( s->avctx, AV_LOG_ERROR, "INTRA16x16 AC:%d\n", i );
- if( decode_cabac_residual(h, h->mb + 16*i, 1, i, scan + 1, s->qscale, 15) < 0 )
+ if( decode_cabac_residual(h, h->mb + 16*i, 1, i, scan + 1, h->dequant4_coeff[0][s->qscale], 15) < 0 )
return -1;
}
} else {
int i8x8, i4x4;
for( i8x8 = 0; i8x8 < 4; i8x8++ ) {
if( cbp & (1<<i8x8) ) {
+ if( IS_8x8DCT(mb_type) ) {
+ if( decode_cabac_residual(h, h->mb + 64*i8x8, 5, 4*i8x8,
+ zigzag_scan8x8, h->dequant8_coeff[IS_INTRA( mb_type ) ? 0:1][s->qscale], 64) < 0 )
+ return -1;
+ } else
for( i4x4 = 0; i4x4 < 4; i4x4++ ) {
const int index = 4*i8x8 + i4x4;
//av_log( s->avctx, AV_LOG_ERROR, "Luma4x4: %d\n", index );
- if( decode_cabac_residual(h, h->mb + 16*index, 2, index, scan, s->qscale, 16) < 0 )
+ if( decode_cabac_residual(h, h->mb + 16*index, 2, index, scan, h->dequant4_coeff[IS_INTRA( mb_type ) ? 0:3][s->qscale], 16) < 0 )
return -1;
}
} else {
int c;
for( c = 0; c < 2; c++ ) {
//av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-DC\n",c );
- if( decode_cabac_residual(h, h->mb + 256 + 16*4*c, 3, c, chroma_dc_scan, h->chroma_qp, 4) < 0)
+ if( decode_cabac_residual(h, h->mb + 256 + 16*4*c, 3, c, chroma_dc_scan, NULL, 4) < 0)
return -1;
}
}
for( i = 0; i < 4; i++ ) {
const int index = 16 + 4 * c + i;
//av_log( s->avctx, AV_LOG_ERROR, "INTRA C%d-AC %d\n",c, index - 16 );
- if( decode_cabac_residual(h, h->mb + 16*index, 4, index - 16, scan + 1, h->chroma_qp, 15) < 0)
+ if( decode_cabac_residual(h, h->mb + 16*index, 4, index - 16, scan + 1, h->dequant4_coeff[c+1+(IS_INTRA( mb_type ) ? 0:3)][h->chroma_qp], 15) < 0)
return -1;
}
}
const int alpha = alpha_table[index_a];
const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
- for( i = 0; i < 4; i++ ) {
- if( bS[i] == 0 ) {
- pix += 4 * stride;
- continue;
- }
-
- if( bS[i] < 4 ) {
- const int tc0 = tc0_table[index_a][bS[i] - 1];
- /* 4px edge length */
- for( d = 0; d < 4; d++ ) {
- const int p0 = pix[-1];
- const int p1 = pix[-2];
- const int p2 = pix[-3];
- const int q0 = pix[0];
- const int q1 = pix[1];
- const int q2 = pix[2];
-
- if( ABS( p0 - q0 ) < alpha &&
- ABS( p1 - p0 ) < beta &&
- ABS( q1 - q0 ) < beta ) {
- int tc = tc0;
- int i_delta;
-
- if( ABS( p2 - p0 ) < beta ) {
- pix[-2] = p1 + clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
- tc++;
- }
- if( ABS( q2 - q0 ) < beta ) {
- pix[1] = q1 + clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
- tc++;
- }
-
- i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
- pix[-1] = clip_uint8( p0 + i_delta ); /* p0' */
- pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
- }
- pix += stride;
- }
- }else{
- /* 4px edge length */
- for( d = 0; d < 4; d++ ) {
+ if( bS[0] < 4 ) {
+ int8_t tc[4];
+ for(i=0; i<4; i++)
+ tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] : -1;
+ h->s.dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc);
+ } else {
+ /* 16px edge length, because bS=4 is triggered by being at
+ * the edge of an intra MB, so all 4 bS are the same */
+ for( d = 0; d < 16; d++ ) {
const int p0 = pix[-1];
const int p1 = pix[-2];
const int p2 = pix[-3];
pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
pix[ 0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
}
+ tprintf("filter_mb_edgev i:%d d:%d\n# bS:4 -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, d, p2, p1, p0, q0, q1, q2, pix[-2], pix[-1], pix[0], pix[1]);
}
pix += stride;
}
- }
}
}
static void filter_mb_edgecv( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
- int i, d;
+ int i;
const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
const int alpha = alpha_table[index_a];
const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
- for( i = 0; i < 4; i++ ) {
- if( bS[i] == 0 ) {
- pix += 2 * stride;
+ if( bS[0] < 4 ) {
+ int8_t tc[4];
+ for(i=0; i<4; i++)
+ tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] + 1 : 0;
+ h->s.dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc);
+ } else {
+ h->s.dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta);
+ }
+}
+
+static void filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, int bS[8], int qp[2] ) {
+ int i;
+ for( i = 0; i < 16; i++, pix += stride) {
+ int index_a;
+ int alpha;
+ int beta;
+
+ int qp_index;
+ int bS_index = (i >> 1);
+ if (h->mb_field_decoding_flag) {
+ bS_index &= ~1;
+ bS_index |= (i & 1);
+ }
+
+ if( bS[bS_index] == 0 ) {
continue;
}
- if( bS[i] < 4 ) {
- const int tc = tc0_table[index_a][bS[i] - 1] + 1;
- /* 2px edge length (because we use same bS than the one for luma) */
- for( d = 0; d < 2; d++ ){
- const int p0 = pix[-1];
- const int p1 = pix[-2];
- const int q0 = pix[0];
- const int q1 = pix[1];
+ qp_index = h->mb_field_decoding_flag ? (i & 1) : (i >> 3);
+ index_a = clip( qp[qp_index] + h->slice_alpha_c0_offset, 0, 51 );
+ alpha = alpha_table[index_a];
+ beta = beta_table[clip( qp[qp_index] + h->slice_beta_offset, 0, 51 )];
- if( ABS( p0 - q0 ) < alpha &&
- ABS( p1 - p0 ) < beta &&
- ABS( q1 - q0 ) < beta ) {
- const int i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
- pix[-1] = clip_uint8( p0 + i_delta ); /* p0' */
- pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
- //tprintf("filter_mb_edgecv i:%d d:%d, qp:%d, indexA:%d, alpha:%d, beta:%d, tc:%d\n# bS:%d -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, d, qp, index_a, alpha, beta, tc, bS[i], pix[-3], p1, p0, q0, q1, pix[2], p1, pix[-1], pix[0], q1);
+ if( bS[bS_index] < 4 ) {
+ const int tc0 = tc0_table[index_a][bS[bS_index] - 1];
+ /* 4px edge length */
+ const int p0 = pix[-1];
+ const int p1 = pix[-2];
+ const int p2 = pix[-3];
+ const int q0 = pix[0];
+ const int q1 = pix[1];
+ const int q2 = pix[2];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+ int tc = tc0;
+ int i_delta;
+
+ if( ABS( p2 - p0 ) < beta ) {
+ pix[-2] = p1 + clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
+ tc++;
}
- pix += stride;
+ if( ABS( q2 - q0 ) < beta ) {
+ pix[1] = q1 + clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
+ tc++;
+ }
+
+ i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+ pix[-1] = clip_uint8( p0 + i_delta ); /* p0' */
+ pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
+ tprintf("filter_mb_mbaff_edgev i:%d, qp:%d, indexA:%d, alpha:%d, beta:%d, tc:%d\n# bS:%d -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, qp[qp_index], index_a, alpha, beta, tc, bS[bS_index], pix[-3], p1, p0, q0, q1, pix[2], p1, pix[-1], pix[0], q1);
}
}else{
- /* 2px edge length (because we use same bS than the one for luma) */
- for( d = 0; d < 2; d++ ){
- const int p0 = pix[-1];
- const int p1 = pix[-2];
- const int q0 = pix[0];
- const int q1 = pix[1];
+ /* 4px edge length */
+ const int p0 = pix[-1];
+ const int p1 = pix[-2];
+ const int p2 = pix[-3];
+
+ const int q0 = pix[0];
+ const int q1 = pix[1];
+ const int q2 = pix[2];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+
+ if(ABS( p0 - q0 ) < (( alpha >> 2 ) + 2 )){
+ if( ABS( p2 - p0 ) < beta)
+ {
+ const int p3 = pix[-4];
+ /* p0', p1', p2' */
+ pix[-1] = ( p2 + 2*p1 + 2*p0 + 2*q0 + q1 + 4 ) >> 3;
+ pix[-2] = ( p2 + p1 + p0 + q0 + 2 ) >> 2;
+ pix[-3] = ( 2*p3 + 3*p2 + p1 + p0 + q0 + 4 ) >> 3;
+ } else {
+ /* p0' */
+ pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+ }
+ if( ABS( q2 - q0 ) < beta)
+ {
+ const int q3 = pix[3];
+ /* q0', q1', q2' */
+ pix[0] = ( p1 + 2*p0 + 2*q0 + 2*q1 + q2 + 4 ) >> 3;
+ pix[1] = ( p0 + q0 + q1 + q2 + 2 ) >> 2;
+ pix[2] = ( 2*q3 + 3*q2 + q1 + q0 + p0 + 4 ) >> 3;
+ } else {
+ /* q0' */
+ pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+ }
+ }else{
+ /* p0', q0' */
+ pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
+ pix[ 0] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
+ }
+ tprintf("filter_mb_mbaff_edgev i:%d, qp:%d, indexA:%d, alpha:%d, beta:%d\n# bS:4 -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x, %02x, %02x]\n", i, qp[qp_index], index_a, alpha, beta, p2, p1, p0, q0, q1, q2, pix[-3], pix[-2], pix[-1], pix[0], pix[1], pix[2]);
+ }
+ }
+ }
+}
+static void filter_mb_mbaff_edgecv( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp[2] ) {
+ int i;
+ for( i = 0; i < 8; i++, pix += stride) {
+ int index_a;
+ int alpha;
+ int beta;
- if( ABS( p0 - q0 ) < alpha &&
- ABS( p1 - p0 ) < beta &&
- ABS( q1 - q0 ) < beta ) {
+ int qp_index;
+ int bS_index = i;
- pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */
- pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */
- //tprintf("filter_mb_edgecv i:%d d:%d\n# bS:4 -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, d, pix[-3], p1, p0, q0, q1, pix[2], p1, pix[-1], pix[0], q1);
- }
- pix += stride;
+ if( bS[bS_index] == 0 ) {
+ continue;
+ }
+
+ qp_index = h->mb_field_decoding_flag ? (i & 1) : (i >> 3);
+ index_a = clip( qp[qp_index] + h->slice_alpha_c0_offset, 0, 51 );
+ alpha = alpha_table[index_a];
+ beta = beta_table[clip( qp[qp_index] + h->slice_beta_offset, 0, 51 )];
+ if( bS[bS_index] < 4 ) {
+ const int tc = tc0_table[index_a][bS[bS_index] - 1] + 1;
+ /* 2px edge length (because we use same bS than the one for luma) */
+ const int p0 = pix[-1];
+ const int p1 = pix[-2];
+ const int q0 = pix[0];
+ const int q1 = pix[1];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+ const int i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+
+ pix[-1] = clip_uint8( p0 + i_delta ); /* p0' */
+ pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
+ tprintf("filter_mb_mbaff_edgecv i:%d, qp:%d, indexA:%d, alpha:%d, beta:%d, tc:%d\n# bS:%d -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, qp[qp_index], index_a, alpha, beta, tc, bS[bS_index], pix[-3], p1, p0, q0, q1, pix[2], p1, pix[-1], pix[0], q1);
+ }
+ }else{
+ const int p0 = pix[-1];
+ const int p1 = pix[-2];
+ const int q0 = pix[0];
+ const int q1 = pix[1];
+
+ if( ABS( p0 - q0 ) < alpha &&
+ ABS( p1 - p0 ) < beta &&
+ ABS( q1 - q0 ) < beta ) {
+
+ pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */
+ pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */
+ tprintf("filter_mb_mbaff_edgecv i:%d\n# bS:4 -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x, %02x, %02x]\n", i, pix[-3], p1, p0, q0, q1, pix[2], pix[-3], pix[-2], pix[-1], pix[0], pix[1], pix[2]);
}
}
}
const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
const int pix_next = stride;
- for( i = 0; i < 4; i++ ) {
- if( bS[i] == 0 ) {
- pix += 4;
- continue;
- }
-
- if( bS[i] < 4 ) {
- const int tc0 = tc0_table[index_a][bS[i] - 1];
- /* 4px edge length */
- for( d = 0; d < 4; d++ ) {
- const int p0 = pix[-1*pix_next];
- const int p1 = pix[-2*pix_next];
- const int p2 = pix[-3*pix_next];
- const int q0 = pix[0];
- const int q1 = pix[1*pix_next];
- const int q2 = pix[2*pix_next];
-
- if( ABS( p0 - q0 ) < alpha &&
- ABS( p1 - p0 ) < beta &&
- ABS( q1 - q0 ) < beta ) {
-
- int tc = tc0;
- int i_delta;
-
- if( ABS( p2 - p0 ) < beta ) {
- pix[-2*pix_next] = p1 + clip( ( p2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( p1 << 1 ) ) >> 1, -tc0, tc0 );
- tc++;
- }
- if( ABS( q2 - q0 ) < beta ) {
- pix[pix_next] = q1 + clip( ( q2 + ( ( p0 + q0 + 1 ) >> 1 ) - ( q1 << 1 ) ) >> 1, -tc0, tc0 );
- tc++;
- }
-
- i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
- pix[-pix_next] = clip_uint8( p0 + i_delta ); /* p0' */
- pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
- }
- pix++;
- }
- }else{
- /* 4px edge length */
- for( d = 0; d < 4; d++ ) {
+ if( bS[0] < 4 ) {
+ int8_t tc[4];
+ for(i=0; i<4; i++)
+ tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] : -1;
+ h->s.dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc);
+ } else {
+ /* 16px edge length, see filter_mb_edgev */
+ for( d = 0; d < 16; d++ ) {
const int p0 = pix[-1*pix_next];
const int p1 = pix[-2*pix_next];
const int p2 = pix[-3*pix_next];
pix[-1*pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2;
pix[ 0*pix_next] = ( 2*q1 + q0 + p1 + 2 ) >> 2;
}
+ tprintf("filter_mb_edgeh i:%d d:%d, qp:%d, indexA:%d, alpha:%d, beta:%d\n# bS:%d -> [%02x, %02x, %02x, %02x, %02x, %02x] =>[%02x, %02x, %02x, %02x]\n", i, d, qp, index_a, alpha, beta, bS[i], p2, p1, p0, q0, q1, q2, pix[-2*pix_next], pix[-pix_next], pix[0], pix[pix_next]);
}
pix++;
}
- }
}
}
static void filter_mb_edgech( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) {
- int i, d;
+ int i;
const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 );
const int alpha = alpha_table[index_a];
const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )];
- const int pix_next = stride;
-
- for( i = 0; i < 4; i++ )
- {
- if( bS[i] == 0 ) {
- pix += 2;
- continue;
- }
-
- if( bS[i] < 4 ) {
- int tc = tc0_table[index_a][bS[i] - 1] + 1;
- /* 2px edge length (see deblocking_filter_edgecv) */
- for( d = 0; d < 2; d++ ) {
- const int p0 = pix[-1*pix_next];
- const int p1 = pix[-2*pix_next];
- const int q0 = pix[0];
- const int q1 = pix[1*pix_next];
- if( ABS( p0 - q0 ) < alpha &&
- ABS( p1 - p0 ) < beta &&
- ABS( q1 - q0 ) < beta ) {
+ if( bS[0] < 4 ) {
+ int8_t tc[4];
+ for(i=0; i<4; i++)
+ tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] + 1 : 0;
+ h->s.dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc);
+ } else {
+ h->s.dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta);
+ }
+}
- int i_delta = clip( (((q0 - p0 ) << 2) + (p1 - q1) + 4) >> 3, -tc, tc );
+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) {
+ MpegEncContext * const s = &h->s;
+ const int mb_xy= mb_x + mb_y*s->mb_stride;
+ int first_vertical_edge_done = 0;
+ int dir;
+ /* FIXME: A given frame may occupy more than one position in
+ * the reference list. So ref2frm should be populated with
+ * frame numbers, not indices. */
+ static const int ref2frm[18] = {-1,-1,0,1,2,3,4,5,6,7,8,9,10,11,12,13,14,15};
+
+ if (h->mb_aff_frame
+ // left mb is in picture
+ && h->slice_table[mb_xy-1] != 255
+ // and current and left pair do not have the same interlaced type
+ && (IS_INTERLACED(s->current_picture.mb_type[mb_xy]) != IS_INTERLACED(s->current_picture.mb_type[mb_xy-1]))
+ // and left mb is in the same slice if deblocking_filter == 2
+ && (h->deblocking_filter!=2 || h->slice_table[mb_xy-1] == h->slice_table[mb_xy])) {
+ /* First vertical edge is different in MBAFF frames
+ * There are 8 different bS to compute and 2 different Qp
+ */
+ int bS[8];
+ int qp[2];
+ int chroma_qp[2];
- pix[-pix_next] = clip_uint8( p0 + i_delta ); /* p0' */
- pix[0] = clip_uint8( q0 - i_delta ); /* q0' */
- }
- pix++;
- }
- }else{
- /* 2px edge length (see deblocking_filter_edgecv) */
- for( d = 0; d < 2; d++ ) {
- const int p0 = pix[-1*pix_next];
- const int p1 = pix[-2*pix_next];
- const int q0 = pix[0];
- const int q1 = pix[1*pix_next];
+ int i;
+ first_vertical_edge_done = 1;
+ for( i = 0; i < 8; i++ ) {
+ int y = i>>1;
+ int b_idx= 8 + 4 + 8*y;
+ int bn_idx= b_idx - 1;
- if( ABS( p0 - q0 ) < alpha &&
- ABS( p1 - p0 ) < beta &&
- ABS( q1 - q0 ) < beta ) {
+ int mbn_xy = h->mb_field_decoding_flag ? h->left_mb_xy[i>>2] : h->left_mb_xy[i&1];
- pix[-pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */
- pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */
+ if( IS_INTRA( s->current_picture.mb_type[mb_xy] ) ||
+ IS_INTRA( s->current_picture.mb_type[mbn_xy] ) ) {
+ bS[i] = 4;
+ } else if( h->non_zero_count_cache[b_idx] != 0 ||
+ /* FIXME: with 8x8dct + cavlc, should check cbp instead of nnz */
+ h->non_zero_count_cache[bn_idx] != 0 ) {
+ bS[i] = 2;
+ } else {
+ int l;
+ bS[i] = 0;
+ for( l = 0; l < 1 + (h->slice_type == B_TYPE); l++ ) {
+ if( ref2frm[h->ref_cache[l][b_idx]+2] != ref2frm[h->ref_cache[l][bn_idx]+2] ||
+ ABS( h->mv_cache[l][b_idx][0] - h->mv_cache[l][bn_idx][0] ) >= 4 ||
+ ABS( h->mv_cache[l][b_idx][1] - h->mv_cache[l][bn_idx][1] ) >= 4 ) {
+ bS[i] = 1;
+ break;
+ }
}
- pix++;
}
}
- }
-}
-
-static void filter_mb( H264Context *h, int mb_x, int mb_y, uint8_t *img_y, uint8_t *img_cb, uint8_t *img_cr) {
- MpegEncContext * const s = &h->s;
- const int mb_xy= mb_x + mb_y*s->mb_stride;
- int linesize, uvlinesize;
- int dir;
+ if(bS[0]+bS[1]+bS[2]+bS[3] != 0) {
+ // Do not use s->qscale as luma quantizer because it has not the same
+ // value in IPCM macroblocks.
+ qp[0] = ( s->current_picture.qscale_table[mb_xy] + s->current_picture.qscale_table[h->left_mb_xy[0]] + 1 ) >> 1;
+ chroma_qp[0] = ( get_chroma_qp( h->pps.chroma_qp_index_offset, s->current_picture.qscale_table[mb_xy] ) +
+ get_chroma_qp( h->pps.chroma_qp_index_offset, s->current_picture.qscale_table[h->left_mb_xy[0]] ) + 1 ) >> 1;
+ qp[1] = ( s->current_picture.qscale_table[mb_xy] + s->current_picture.qscale_table[h->left_mb_xy[1]] + 1 ) >> 1;
+ chroma_qp[1] = ( get_chroma_qp( h->pps.chroma_qp_index_offset, s->current_picture.qscale_table[mb_xy] ) +
+ get_chroma_qp( h->pps.chroma_qp_index_offset, s->current_picture.qscale_table[h->left_mb_xy[1]] ) + 1 ) >> 1;
- /* FIXME Implement deblocking filter for field MB */
- if( h->sps.mb_aff ) {
- return;
+ /* Filter edge */
+ tprintf("filter mb:%d/%d MBAFF, QPy:%d/%d, QPc:%d/%d ls:%d uvls:%d", mb_x, mb_y, qp[0], qp[1], chroma_qp[0], chroma_qp[1], linesize, uvlinesize);
+ { int i; for (i = 0; i < 8; i++) tprintf(" bS[%d]:%d", i, bS[i]); tprintf("\n"); }
+ filter_mb_mbaff_edgev ( h, &img_y [0], linesize, bS, qp );
+ filter_mb_mbaff_edgecv( h, &img_cb[0], uvlinesize, bS, chroma_qp );
+ filter_mb_mbaff_edgecv( h, &img_cr[0], uvlinesize, bS, chroma_qp );
+ }
}
- linesize = s->linesize;
- uvlinesize = s->uvlinesize;
-
/* dir : 0 -> vertical edge, 1 -> horizontal edge */
for( dir = 0; dir < 2; dir++ )
{
int edge;
- const int mbm_xy = dir == 0 ? mb_xy -1 : mb_xy - s->mb_stride;
+ const int mbm_xy = dir == 0 ? mb_xy -1 : h->top_mb_xy;
+ const int mb_type = s->current_picture.mb_type[mb_xy];
+ const int mbm_type = s->current_picture.mb_type[mbm_xy];
int start = h->slice_table[mbm_xy] == 255 ? 1 : 0;
+ const int edges = ((mb_type & mbm_type) & (MB_TYPE_16x16|MB_TYPE_SKIP))
+ == (MB_TYPE_16x16|MB_TYPE_SKIP) ? 1 : 4;
+ // how often to recheck mv-based bS when iterating between edges
+ const int mask_edge = (mb_type & (MB_TYPE_16x16 | (MB_TYPE_16x8 << dir))) ? 3 :
+ (mb_type & (MB_TYPE_8x16 >> dir)) ? 1 : 0;
+ // how often to recheck mv-based bS when iterating along each edge
+ const int mask_par0 = mb_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir));
+
+ if (first_vertical_edge_done) {
+ start = 1;
+ first_vertical_edge_done = 0;
+ }
+
if (h->deblocking_filter==2 && h->slice_table[mbm_xy] != h->slice_table[mb_xy])
start = 1;
/* Calculate bS */
- for( edge = start; edge < 4; edge++ ) {
- /* mbn_xy: neighbour macroblock (how that works for field ?) */
- int mbn_xy = edge > 0 ? mb_xy : mbm_xy;
+ for( edge = start; edge < edges; edge++ ) {
+ /* 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 bS[4];
int qp;
- if( IS_INTRA( s->current_picture.mb_type[mb_xy] ) ||
- IS_INTRA( s->current_picture.mb_type[mbn_xy] ) ) {
- bS[0] = bS[1] = bS[2] = bS[3] = ( edge == 0 ? 4 : 3 );
+ if( (edge&1) && IS_8x8DCT(mb_type) )
+ continue;
+
+ if (h->mb_aff_frame && (dir == 1) && (edge == 0) && ((mb_y & 1) == 0)
+ && !IS_INTERLACED(mb_type)
+ && IS_INTERLACED(mbn_type)
+ ) {
+ // This is a special case in the norm where the filtering must
+ // be done twice (one each of the field) even if we are in a
+ // frame macroblock.
+ //
+ unsigned int tmp_linesize = 2 * linesize;
+ unsigned int tmp_uvlinesize = 2 * uvlinesize;
+ int mbn_xy = mb_xy - 2 * s->mb_stride;
+ int qp, chroma_qp;
+
+ // first filtering
+ if( IS_INTRA(mb_type) ||
+ IS_INTRA(s->current_picture.mb_type[mbn_xy]) ) {
+ bS[0] = bS[1] = bS[2] = bS[3] = 3;
+ } else {
+ // TODO
+ av_log(h->s.avctx, AV_LOG_ERROR, "both non intra (TODO)\n");
+ }
+ /* Filter edge */
+ // Do not use s->qscale as luma quantizer because it has not the same
+ // value in IPCM macroblocks.
+ qp = ( s->current_picture.qscale_table[mb_xy] + s->current_picture.qscale_table[mbn_xy] + 1 ) >> 1;
+ tprintf("filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize);
+ { int i; for (i = 0; i < 4; i++) tprintf(" bS[%d]:%d", i, bS[i]); tprintf("\n"); }
+ filter_mb_edgeh( h, &img_y[0], tmp_linesize, bS, qp );
+ chroma_qp = ( h->chroma_qp +
+ get_chroma_qp( h->pps.chroma_qp_index_offset, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
+ filter_mb_edgech( h, &img_cb[0], tmp_uvlinesize, bS, chroma_qp );
+ filter_mb_edgech( h, &img_cr[0], tmp_uvlinesize, bS, chroma_qp );
+
+ // second filtering
+ mbn_xy += s->mb_stride;
+ if( IS_INTRA(mb_type) ||
+ IS_INTRA(mbn_type) ) {
+ bS[0] = bS[1] = bS[2] = bS[3] = 3;
+ } else {
+ // TODO
+ av_log(h->s.avctx, AV_LOG_ERROR, "both non intra (TODO)\n");
+ }
+ /* Filter edge */
+ // Do not use s->qscale as luma quantizer because it has not the same
+ // value in IPCM macroblocks.
+ qp = ( s->current_picture.qscale_table[mb_xy] + s->current_picture.qscale_table[mbn_xy] + 1 ) >> 1;
+ tprintf("filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, tmp_linesize, tmp_uvlinesize);
+ { int i; for (i = 0; i < 4; i++) tprintf(" bS[%d]:%d", i, bS[i]); tprintf("\n"); }
+ filter_mb_edgeh( h, &img_y[linesize], tmp_linesize, bS, qp );
+ chroma_qp = ( h->chroma_qp +
+ get_chroma_qp( h->pps.chroma_qp_index_offset, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
+ filter_mb_edgech( h, &img_cb[uvlinesize], tmp_uvlinesize, bS, chroma_qp );
+ filter_mb_edgech( h, &img_cr[uvlinesize], tmp_uvlinesize, bS, chroma_qp );
+ continue;
+ }
+ if( IS_INTRA(mb_type) ||
+ IS_INTRA(mbn_type) ) {
+ int value;
+ if (edge == 0) {
+ if ( (!IS_INTERLACED(mb_type) && !IS_INTERLACED(mbm_type))
+ || ((h->mb_aff_frame || (s->picture_structure != PICT_FRAME)) && (dir == 0))
+ ) {
+ value = 4;
+ } else {
+ value = 3;
+ }
+ } else {
+ value = 3;
+ }
+ bS[0] = bS[1] = bS[2] = bS[3] = value;
} else {
- int i;
+ int i, l;
+ int mv_done;
+
+ if( edge & mask_edge ) {
+ bS[0] = bS[1] = bS[2] = bS[3] = 0;
+ mv_done = 1;
+ }
+ else if( mask_par0 && (edge || (mbn_type & (MB_TYPE_16x16 | (MB_TYPE_8x16 >> dir)))) ) {
+ int b_idx= 8 + 4 + edge * (dir ? 8:1);
+ int bn_idx= b_idx - (dir ? 8:1);
+ int v = 0;
+ for( l = 0; !v && l < 1 + (h->slice_type == B_TYPE); l++ ) {
+ v |= ref2frm[h->ref_cache[l][b_idx]+2] != ref2frm[h->ref_cache[l][bn_idx]+2] ||
+ ABS( h->mv_cache[l][b_idx][0] - h->mv_cache[l][bn_idx][0] ) >= 4 ||
+ ABS( h->mv_cache[l][b_idx][1] - h->mv_cache[l][bn_idx][1] ) >= 4;
+ }
+ bS[0] = bS[1] = bS[2] = bS[3] = v;
+ mv_done = 1;
+ }
+ else
+ mv_done = 0;
+
for( i = 0; i < 4; i++ ) {
int x = dir == 0 ? edge : i;
int y = dir == 0 ? i : edge;
h->non_zero_count_cache[bn_idx] != 0 ) {
bS[i] = 2;
}
- else if( h->slice_type == P_TYPE ) {
- if( h->ref_cache[0][b_idx] != h->ref_cache[0][bn_idx] ||
- ABS( h->mv_cache[0][b_idx][0] - h->mv_cache[0][bn_idx][0] ) >= 4 ||
- ABS( h->mv_cache[0][b_idx][1] - h->mv_cache[0][bn_idx][1] ) >= 4 )
- bS[i] = 1;
- else
- bS[i] = 0;
- } else {
- /* FIXME Add support for B frame */
- return;
+ else if(!mv_done)
+ {
+ bS[i] = 0;
+ for( l = 0; l < 1 + (h->slice_type == B_TYPE); l++ ) {
+ if( ref2frm[h->ref_cache[l][b_idx]+2] != ref2frm[h->ref_cache[l][bn_idx]+2] ||
+ ABS( h->mv_cache[l][b_idx][0] - h->mv_cache[l][bn_idx][0] ) >= 4 ||
+ ABS( h->mv_cache[l][b_idx][1] - h->mv_cache[l][bn_idx][1] ) >= 4 ) {
+ bS[i] = 1;
+ break;
+ }
+ }
}
}
}
/* Filter edge */
- // Do not use s->qscale as luma quantiser because it has not the same
+ // Do not use s->qscale as luma quantizer because it has not the same
// value in IPCM macroblocks.
qp = ( s->current_picture.qscale_table[mb_xy] + s->current_picture.qscale_table[mbn_xy] + 1 ) >> 1;
//tprintf("filter mb:%d/%d dir:%d edge:%d, QPy:%d, QPc:%d, QPcn:%d\n", mb_x, mb_y, dir, edge, qp, h->chroma_qp, s->current_picture.qscale_table[mbn_xy]);
+ tprintf("filter mb:%d/%d dir:%d edge:%d, QPy:%d ls:%d uvls:%d", mb_x, mb_y, dir, edge, qp, linesize, uvlinesize);
+ { int i; for (i = 0; i < 4; i++) tprintf(" bS[%d]:%d", i, bS[i]); tprintf("\n"); }
if( dir == 0 ) {
filter_mb_edgev( h, &img_y[4*edge], linesize, bS, qp );
if( (edge&1) == 0 ) {
int chroma_qp = ( h->chroma_qp +
- get_chroma_qp( h, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
+ get_chroma_qp( h->pps.chroma_qp_index_offset, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
filter_mb_edgecv( h, &img_cb[2*edge], uvlinesize, bS, chroma_qp );
filter_mb_edgecv( h, &img_cr[2*edge], uvlinesize, bS, chroma_qp );
}
filter_mb_edgeh( h, &img_y[4*edge*linesize], linesize, bS, qp );
if( (edge&1) == 0 ) {
int chroma_qp = ( h->chroma_qp +
- get_chroma_qp( h, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
+ get_chroma_qp( h->pps.chroma_qp_index_offset, s->current_picture.qscale_table[mbn_xy] ) + 1 ) >> 1;
filter_mb_edgech( h, &img_cb[2*edge*uvlinesize], uvlinesize, bS, chroma_qp );
filter_mb_edgech( h, &img_cr[2*edge*uvlinesize], uvlinesize, bS, chroma_qp );
}
s->gb.buffer + get_bits_count(&s->gb)/8,
( s->gb.size_in_bits - get_bits_count(&s->gb) + 7)/8);
/* calculate pre-state */
- for( i= 0; i < 399; i++ ) {
+ for( i= 0; i < 460; i++ ) {
int pre;
if( h->slice_type == I_TYPE )
pre = clip( ((cabac_context_init_I[i][0] * s->qscale) >>4 ) + cabac_context_init_I[i][1], 1, 126 );
for(;;){
int ret = decode_mb_cabac(h);
- int eos = get_cabac_terminate( &h->cabac ); /* End of Slice flag */
+ int eos;
if(ret>=0) hl_decode_mb(h);
/* XXX: useless as decode_mb_cabac it doesn't support that ... */
- if( ret >= 0 && h->sps.mb_aff ) { //FIXME optimal? or let mb_decode decode 16x32 ?
+ if( ret >= 0 && h->mb_aff_frame ) { //FIXME optimal? or let mb_decode decode 16x32 ?
s->mb_y++;
if(ret>=0) ret = decode_mb_cabac(h);
- eos = get_cabac_terminate( &h->cabac );
- hl_decode_mb(h);
+ if(ret>=0) hl_decode_mb(h);
s->mb_y--;
}
+ eos = get_cabac_terminate( &h->cabac );
if( ret < 0 || h->cabac.bytestream > h->cabac.bytestream_end + 1) {
av_log(h->s.avctx, AV_LOG_ERROR, "error while decoding MB %d %d\n", s->mb_x, s->mb_y);
s->mb_x = 0;
ff_draw_horiz_band(s, 16*s->mb_y, 16);
++s->mb_y;
+ if(h->mb_aff_frame) {
+ ++s->mb_y;
+ }
}
if( eos || s->mb_y >= s->mb_height ) {
ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
return 0;
}
-#if 0
- /* TODO test over-reading in cabac code */
- else if( read too much in h->cabac ) {
- ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x, s->mb_y, (AC_ERROR|DC_ERROR|MV_ERROR)&part_mask);
- return -1;
- }
-#endif
}
} else {
if(ret>=0) hl_decode_mb(h);
- if(ret>=0 && h->sps.mb_aff){ //FIXME optimal? or let mb_decode decode 16x32 ?
+ if(ret>=0 && h->mb_aff_frame){ //FIXME optimal? or let mb_decode decode 16x32 ?
s->mb_y++;
ret = decode_mb_cavlc(h);
if(++s->mb_x >= s->mb_width){
s->mb_x=0;
ff_draw_horiz_band(s, 16*s->mb_y, 16);
- if(++s->mb_y >= s->mb_height){
+ ++s->mb_y;
+ if(h->mb_aff_frame) {
+ ++s->mb_y;
+ }
+ if(s->mb_y >= s->mb_height){
tprintf("slice end %d %d\n", get_bits_count(&s->gb), s->gb.size_in_bits);
if(get_bits_count(&s->gb) == s->gb.size_in_bits ) {
for(;s->mb_y < s->mb_height; s->mb_y++){
for(;s->mb_x < s->mb_width; s->mb_x++){
int ret= decode_mb(h);
-
+
hl_decode_mb(h);
if(ret<0){
return -1;
}
-
+
if(++s->mb_x >= s->mb_width){
s->mb_x=0;
if(++s->mb_y >= s->mb_height){
}
}
}
-
+
if(get_bits_count(s->?gb) >= s->gb?.size_in_bits){
if(get_bits_count(s->gb) == s->gb.size_in_bits){
ff_er_add_slice(s, s->resync_mb_x, s->resync_mb_y, s->mb_x-1, s->mb_y, (AC_END|DC_END|MV_END)&part_mask);
return -1; //not reached
}
+static int decode_unregistered_user_data(H264Context *h, int size){
+ MpegEncContext * const s = &h->s;
+ uint8_t user_data[16+256];
+ int e, build, i;
+
+ if(size<16)
+ return -1;
+
+ for(i=0; i<sizeof(user_data)-1 && i<size; i++){
+ user_data[i]= get_bits(&s->gb, 8);
+ }
+
+ user_data[i]= 0;
+ e= sscanf(user_data+16, "x264 - core %d"/*%s - H.264/MPEG-4 AVC codec - Copyleft 2005 - http://www.videolan.org/x264.html*/, &build);
+ if(e==1 && build>=0)
+ h->x264_build= build;
+
+ if(s->avctx->debug & FF_DEBUG_BUGS)
+ av_log(s->avctx, AV_LOG_DEBUG, "user data:\"%s\"\n", user_data+16);
+
+ for(; i<size; i++)
+ skip_bits(&s->gb, 8);
+
+ return 0;
+}
+
+static int decode_sei(H264Context *h){
+ MpegEncContext * const s = &h->s;
+
+ while(get_bits_count(&s->gb) + 16 < s->gb.size_in_bits){
+ int size, type;
+
+ type=0;
+ do{
+ type+= show_bits(&s->gb, 8);
+ }while(get_bits(&s->gb, 8) == 255);
+
+ size=0;
+ do{
+ size+= show_bits(&s->gb, 8);
+ }while(get_bits(&s->gb, 8) == 255);
+
+ switch(type){
+ case 5:
+ if(decode_unregistered_user_data(h, size) < 0);
+ return -1;
+ break;
+ default:
+ skip_bits(&s->gb, 8*size);
+ }
+
+ //FIXME check bits here
+ align_get_bits(&s->gb);
+ }
+
+ return 0;
+}
+
+static inline void decode_hrd_parameters(H264Context *h, SPS *sps){
+ MpegEncContext * const s = &h->s;
+ int cpb_count, i;
+ cpb_count = get_ue_golomb(&s->gb) + 1;
+ get_bits(&s->gb, 4); /* bit_rate_scale */
+ get_bits(&s->gb, 4); /* cpb_size_scale */
+ for(i=0; i<cpb_count; i++){
+ get_ue_golomb(&s->gb); /* bit_rate_value_minus1 */
+ get_ue_golomb(&s->gb); /* cpb_size_value_minus1 */
+ 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 */
+}
+
static inline int decode_vui_parameters(H264Context *h, SPS *sps){
MpegEncContext * const s = &h->s;
int aspect_ratio_info_present_flag, 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_info_present_flag ) {
aspect_ratio_idc= get_bits(&s->gb, 8);
if( aspect_ratio_idc == EXTENDED_SAR ) {
return -1;
}
}else{
- sps->sar.num=
+ sps->sar.num=
sps->sar.den= 0;
}
// s->avctx->aspect_ratio= sar_width*s->width / (float)(s->height*sar_height);
sps->fixed_frame_rate_flag = get_bits1(&s->gb);
}
-#if 0
-| nal_hrd_parameters_present_flag |0 |u(1) |
-| if( nal_hrd_parameters_present_flag = = 1) | | |
-| hrd_parameters( ) | | |
-| vcl_hrd_parameters_present_flag |0 |u(1) |
-| if( vcl_hrd_parameters_present_flag = = 1) | | |
-| hrd_parameters( ) | | |
-| if( ( nal_hrd_parameters_present_flag = = 1 | || | |
-| | | |
-|( vcl_hrd_parameters_present_flag = = 1 ) ) | | |
-| low_delay_hrd_flag |0 |u(1) |
-| bitstream_restriction_flag |0 |u(1) |
-| if( bitstream_restriction_flag ) { |0 |u(1) |
-| motion_vectors_over_pic_boundaries_flag |0 |u(1) |
-| max_bytes_per_pic_denom |0 |ue(v) |
-| max_bits_per_mb_denom |0 |ue(v) |
-| log2_max_mv_length_horizontal |0 |ue(v) |
-| log2_max_mv_length_vertical |0 |ue(v) |
-| num_reorder_frames |0 |ue(v) |
-| max_dec_frame_buffering |0 |ue(v) |
-| } | | |
-|} | | |
-#endif
+ nal_hrd_parameters_present_flag = get_bits1(&s->gb);
+ if(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)
+ decode_hrd_parameters(h, sps);
+ if(nal_hrd_parameters_present_flag || vcl_hrd_parameters_present_flag)
+ get_bits1(&s->gb); /* low_delay_hrd_flag */
+ get_bits1(&s->gb); /* pic_struct_present_flag */
+
+ sps->bitstream_restriction_flag = get_bits1(&s->gb);
+ if(sps->bitstream_restriction_flag){
+ 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 */
+ sps->num_reorder_frames = get_ue_golomb(&s->gb);
+ get_ue_golomb(&s->gb); /* max_dec_frame_buffering */
+ }
+
return 0;
}
+static void decode_scaling_list(H264Context *h, uint8_t *factors, int size,
+ const uint8_t *jvt_list, const uint8_t *fallback_list){
+ MpegEncContext * const s = &h->s;
+ int i, last = 8, next = 8;
+ const uint8_t *scan = size == 16 ? zigzag_scan : zigzag_scan8x8;
+ if(!get_bits1(&s->gb)) /* matrix not written, we use the predicted one */
+ memcpy(factors, fallback_list, size*sizeof(uint8_t));
+ else
+ for(i=0;i<size;i++){
+ if(next)
+ next = (last + get_se_golomb(&s->gb)) & 0xff;
+ if(!i && !next){ /* matrix not written, we use the preset one */
+ memcpy(factors, jvt_list, size*sizeof(uint8_t));
+ break;
+ }
+ last = factors[scan[i]] = next ? next : last;
+ }
+}
+
+static void decode_scaling_matrices(H264Context *h, SPS *sps, PPS *pps, int is_sps,
+ uint8_t (*scaling_matrix4)[16], uint8_t (*scaling_matrix8)[64]){
+ MpegEncContext * const s = &h->s;
+ int fallback_sps = !is_sps && sps->scaling_matrix_present;
+ const uint8_t *fallback[4] = {
+ fallback_sps ? sps->scaling_matrix4[0] : default_scaling4[0],
+ fallback_sps ? sps->scaling_matrix4[3] : default_scaling4[1],
+ fallback_sps ? sps->scaling_matrix8[0] : default_scaling8[0],
+ fallback_sps ? sps->scaling_matrix8[1] : default_scaling8[1]
+ };
+ if(get_bits1(&s->gb)){
+ sps->scaling_matrix_present |= is_sps;
+ decode_scaling_list(h,scaling_matrix4[0],16,default_scaling4[0],fallback[0]); // Intra, Y
+ decode_scaling_list(h,scaling_matrix4[1],16,default_scaling4[0],scaling_matrix4[0]); // Intra, Cr
+ decode_scaling_list(h,scaling_matrix4[2],16,default_scaling4[0],scaling_matrix4[1]); // Intra, Cb
+ decode_scaling_list(h,scaling_matrix4[3],16,default_scaling4[1],fallback[1]); // Inter, Y
+ decode_scaling_list(h,scaling_matrix4[4],16,default_scaling4[1],scaling_matrix4[3]); // Inter, Cr
+ decode_scaling_list(h,scaling_matrix4[5],16,default_scaling4[1],scaling_matrix4[4]); // Inter, Cb
+ if(is_sps || pps->transform_8x8_mode){
+ 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));
+ }
+}
+
static inline int decode_seq_parameter_set(H264Context *h){
MpegEncContext * const s = &h->s;
int profile_idc, level_idc;
int sps_id, i;
SPS *sps;
-
+
profile_idc= get_bits(&s->gb, 8);
get_bits1(&s->gb); //constraint_set0_flag
get_bits1(&s->gb); //constraint_set1_flag
get_bits(&s->gb, 4); // reserved
level_idc= get_bits(&s->gb, 8);
sps_id= get_ue_golomb(&s->gb);
-
+
sps= &h->sps_buffer[ sps_id ];
sps->profile_idc= profile_idc;
sps->level_idc= level_idc;
+ if(sps->profile_idc >= 100){ //high profile
+ if(get_ue_golomb(&s->gb) == 3) //chroma_format_idc
+ 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;
+
sps->log2_max_frame_num= get_ue_golomb(&s->gb) + 4;
sps->poc_type= get_ue_golomb(&s->gb);
-
+
if(sps->poc_type == 0){ //FIXME #define
sps->log2_max_poc_lsb= get_ue_golomb(&s->gb) + 4;
} else if(sps->poc_type == 1){//FIXME #define
sps->offset_for_non_ref_pic= get_se_golomb(&s->gb);
sps->offset_for_top_to_bottom_field= get_se_golomb(&s->gb);
sps->poc_cycle_length= get_ue_golomb(&s->gb);
-
+
for(i=0; i<sps->poc_cycle_length; i++)
sps->offset_for_ref_frame[i]= get_se_golomb(&s->gb);
}
sps->gaps_in_frame_num_allowed_flag= get_bits1(&s->gb);
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 ||
+ 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))
return -1;
sps->crop_top = get_ue_golomb(&s->gb);
sps->crop_bottom= get_ue_golomb(&s->gb);
if(sps->crop_left || sps->crop_top){
- av_log(h->s.avctx, AV_LOG_ERROR, "insane cropping not completly supported, this could look slightly wrong ...\n");
+ av_log(h->s.avctx, AV_LOG_ERROR, "insane cropping not completely supported, this could look slightly wrong ...\n");
}
}else{
- sps->crop_left =
- sps->crop_right =
- sps->crop_top =
+ sps->crop_left =
+ sps->crop_right =
+ sps->crop_top =
sps->crop_bottom= 0;
}
sps->vui_parameters_present_flag= get_bits1(&s->gb);
if( sps->vui_parameters_present_flag )
decode_vui_parameters(h, sps);
-
+
if(s->avctx->debug&FF_DEBUG_PICT_INFO){
- av_log(h->s.avctx, AV_LOG_DEBUG, "sps:%d 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:%d profile:%d/%d poc:%d ref:%d %dx%d %s %s crop:%d/%d/%d/%d %s\n",
sps_id, sps->profile_idc, sps->level_idc,
sps->poc_type,
sps->ref_frame_count,
sps->mb_width, sps->mb_height,
sps->frame_mbs_only_flag ? "FRM" : (sps->mb_aff ? "MB-AFF" : "PIC-AFF"),
sps->direct_8x8_inference_flag ? "8B8" : "",
- sps->crop_left, sps->crop_right,
- sps->crop_top, sps->crop_bottom,
+ sps->crop_left, sps->crop_right,
+ sps->crop_top, sps->crop_bottom,
sps->vui_parameters_present_flag ? "VUI" : ""
);
}
return 0;
}
-static inline int decode_picture_parameter_set(H264Context *h){
+static inline int decode_picture_parameter_set(H264Context *h, int bit_length){
MpegEncContext * const s = &h->s;
int pps_id= get_ue_golomb(&s->gb);
PPS *pps= &h->pps_buffer[pps_id];
-
+
pps->sps_id= get_ue_golomb(&s->gb);
pps->cabac= get_bits1(&s->gb);
pps->pic_order_present= get_bits1(&s->gb);
av_log(h->s.avctx, AV_LOG_ERROR, "reference overflow (pps)\n");
return -1;
}
-
+
pps->weighted_pred= get_bits1(&s->gb);
pps->weighted_bipred_idc= get_bits(&s->gb, 2);
pps->init_qp= get_se_golomb(&s->gb) + 26;
pps->deblocking_filter_parameters_present= get_bits1(&s->gb);
pps->constrained_intra_pred= get_bits1(&s->gb);
pps->redundant_pic_cnt_present = get_bits1(&s->gb);
-
+
+ memset(pps->scaling_matrix4, 16, 6*16*sizeof(uint8_t));
+ memset(pps->scaling_matrix8, 16, 2*64*sizeof(uint8_t));
+
+ if(get_bits_count(&s->gb) < bit_length){
+ pps->transform_8x8_mode= get_bits1(&s->gb);
+ decode_scaling_matrices(h, &h->sps_buffer[pps->sps_id], pps, 0, pps->scaling_matrix4, pps->scaling_matrix8);
+ get_se_golomb(&s->gb); //second_chroma_qp_index_offset
+ }
+
if(s->avctx->debug&FF_DEBUG_PICT_INFO){
- av_log(h->s.avctx, AV_LOG_DEBUG, "pps:%d sps:%d %s slice_groups:%d ref:%d/%d %s qp:%d/%d/%d %s %s %s\n",
+ av_log(h->s.avctx, AV_LOG_DEBUG, "pps:%d sps:%d %s slice_groups:%d ref:%d/%d %s qp:%d/%d/%d %s %s %s %s\n",
pps_id, pps->sps_id,
pps->cabac ? "CABAC" : "CAVLC",
pps->slice_group_count,
pps->init_qp, pps->init_qs, pps->chroma_qp_index_offset,
pps->deblocking_filter_parameters_present ? "LPAR" : "",
pps->constrained_intra_pred ? "CONSTR" : "",
- pps->redundant_pic_cnt_present ? "REDU" : ""
+ pps->redundant_pic_cnt_present ? "REDU" : "",
+ pps->transform_8x8_mode ? "8x8DCT" : ""
);
}
-
+
return 0;
}
// first_mb_in_slice is 0, probably the first nal of a new
// slice
tprintf("find_frame_end frame_end_found, state = %08x, pos = %d\n", state, i);
- pc->state=-1;
+ pc->state=-1;
pc->frame_start_found= 0;
return i-4;
}
}
pc->frame_start_found = 1;
}
+ if((state&0xFFFFFF1F) == 0x107 || (state&0xFFFFFF1F) == 0x108 || (state&0xFFFFFF1F) == 0x109){
+ if(pc->frame_start_found){
+ pc->state=-1;
+ pc->frame_start_found= 0;
+ return i-4;
+ }
+ }
if (i<buf_size)
state= (state<<8) | buf[i];
}
-
+
pc->state= state;
return END_NOT_FOUND;
}
static int h264_parse(AVCodecParserContext *s,
AVCodecContext *avctx,
- uint8_t **poutbuf, int *poutbuf_size,
+ uint8_t **poutbuf, int *poutbuf_size,
const uint8_t *buf, int buf_size)
{
H264Context *h = s->priv_data;
ParseContext *pc = &h->s.parse_context;
int next;
-
+
next= find_frame_end(h, buf, buf_size);
if (ff_combine_frame(pc, next, (uint8_t **)&buf, &buf_size) < 0) {
return next;
}
+static int h264_split(AVCodecContext *avctx,
+ const uint8_t *buf, int buf_size)
+{
+ int i;
+ uint32_t state = -1;
+ int has_sps= 0;
+
+ for(i=0; i<=buf_size; i++){
+ if((state&0xFFFFFF1F) == 0x107)
+ has_sps=1;
+/* if((state&0xFFFFFF1F) == 0x101 || (state&0xFFFFFF1F) == 0x102 || (state&0xFFFFFF1F) == 0x105){
+ }*/
+ if((state&0xFFFFFF00) == 0x100 && (state&0xFFFFFF1F) != 0x107 && (state&0xFFFFFF1F) != 0x108 && (state&0xFFFFFF1F) != 0x109){
+ if(has_sps){
+ while(i>4 && buf[i-5]==0) i--;
+ return i-4;
+ }
+ }
+ if (i<buf_size)
+ state= (state<<8) | buf[i];
+ }
+ return 0;
+}
+
+
static int decode_nal_units(H264Context *h, uint8_t *buf, int buf_size){
MpegEncContext * const s = &h->s;
AVCodecContext * const avctx= s->avctx;
int buf_index=0;
#if 0
int i;
- for(i=0; i<32; i++){
- printf("%X ", buf[i]);
+ for(i=0; i<50; i++){
+ av_log(NULL, AV_LOG_ERROR,"%02X ", buf[i]);
}
#endif
h->slice_num = 0;
+ s->current_picture_ptr= NULL;
for(;;){
int consumed;
int dst_length;
int bit_length;
uint8_t *ptr;
int i, nalsize = 0;
-
+
if(h->is_avc) {
if(buf_index >= buf_size) break;
nalsize = 0;
if(buf[buf_index] == 0 && buf[buf_index+1] == 0 && buf[buf_index+2] == 1)
break;
}
-
+
if(buf_index+3 >= buf_size) break;
-
+
buf_index+=3;
- }
-
+ }
+
ptr= decode_nal(h, buf + buf_index, &dst_length, &consumed, h->is_avc ? nalsize : buf_size - buf_index);
if(ptr[dst_length - 1] == 0) dst_length--;
bit_length= 8*dst_length - decode_rbsp_trailing(ptr + dst_length - 1);
if(s->avctx->debug&FF_DEBUG_STARTCODE){
av_log(h->s.avctx, AV_LOG_DEBUG, "NAL %d at %d/%d length %d\n", h->nal_unit_type, buf_index, buf_size, dst_length);
}
-
+
if (h->is_avc && (nalsize != consumed))
av_log(h->s.avctx, AV_LOG_ERROR, "AVC: Consumed only %d bytes instead of %d\n", consumed, nalsize);
buf_index += consumed;
- if( s->hurry_up == 1 && h->nal_ref_idc == 0 )
+ if( (s->hurry_up == 1 && h->nal_ref_idc == 0) //FIXME dont discard SEI id
+ ||(avctx->skip_frame >= AVDISCARD_NONREF && h->nal_ref_idc == 0))
continue;
-
+
switch(h->nal_unit_type){
case NAL_IDR_SLICE:
- idr(h); //FIXME ensure we dont loose some frames if there is reordering
+ idr(h); //FIXME ensure we don't loose some frames if there is reordering
case NAL_SLICE:
init_get_bits(&s->gb, ptr, bit_length);
h->intra_gb_ptr=
h->inter_gb_ptr= &s->gb;
s->data_partitioning = 0;
-
- if(decode_slice_header(h) < 0) return -1;
- if(h->redundant_pic_count==0 && s->hurry_up < 5 )
+
+ if(decode_slice_header(h) < 0){
+ av_log(h->s.avctx, AV_LOG_ERROR, "decode_slice_header error\n");
+ break;
+ }
+ if(h->redundant_pic_count==0 && s->hurry_up < 5
+ && (avctx->skip_frame < AVDISCARD_NONREF || h->nal_ref_idc)
+ && (avctx->skip_frame < AVDISCARD_BIDIR || h->slice_type!=B_TYPE)
+ && (avctx->skip_frame < AVDISCARD_NONKEY || h->slice_type==I_TYPE)
+ && avctx->skip_frame < AVDISCARD_ALL)
decode_slice(h);
break;
case NAL_DPA:
h->intra_gb_ptr=
h->inter_gb_ptr= NULL;
s->data_partitioning = 1;
-
- if(decode_slice_header(h) < 0) return -1;
+
+ if(decode_slice_header(h) < 0){
+ av_log(h->s.avctx, AV_LOG_ERROR, "decode_slice_header error\n");
+ }
break;
case NAL_DPB:
init_get_bits(&h->intra_gb, ptr, bit_length);
init_get_bits(&h->inter_gb, ptr, bit_length);
h->inter_gb_ptr= &h->inter_gb;
- if(h->redundant_pic_count==0 && h->intra_gb_ptr && s->data_partitioning && s->hurry_up < 5 )
+ if(h->redundant_pic_count==0 && h->intra_gb_ptr && s->data_partitioning
+ && s->hurry_up < 5
+ && (avctx->skip_frame < AVDISCARD_NONREF || h->nal_ref_idc)
+ && (avctx->skip_frame < AVDISCARD_BIDIR || h->slice_type!=B_TYPE)
+ && (avctx->skip_frame < AVDISCARD_NONKEY || h->slice_type==I_TYPE)
+ && avctx->skip_frame < AVDISCARD_ALL)
decode_slice(h);
break;
case NAL_SEI:
+ init_get_bits(&s->gb, ptr, bit_length);
+ decode_sei(h);
break;
case NAL_SPS:
init_get_bits(&s->gb, ptr, bit_length);
decode_seq_parameter_set(h);
-
+
if(s->flags& CODEC_FLAG_LOW_DELAY)
s->low_delay=1;
-
- avctx->has_b_frames= !s->low_delay;
+
+ if(avctx->has_b_frames < 2)
+ avctx->has_b_frames= !s->low_delay;
break;
case NAL_PPS:
init_get_bits(&s->gb, ptr, bit_length);
-
- decode_picture_parameter_set(h);
+
+ decode_picture_parameter_set(h, bit_length);
break;
- case NAL_PICTURE_DELIMITER:
- break;
- case NAL_FILTER_DATA:
+ case NAL_AUD:
+ case NAL_END_SEQUENCE:
+ case NAL_END_STREAM:
+ case NAL_FILLER_DATA:
+ case NAL_SPS_EXT:
+ case NAL_AUXILIARY_SLICE:
break;
- default:
- av_log(avctx, AV_LOG_ERROR, "Unknown NAL code: %d\n", h->nal_unit_type);
- }
-
- //FIXME move after where irt is set
- s->current_picture.pict_type= s->pict_type;
- s->current_picture.key_frame= s->pict_type == I_TYPE;
+ default:
+ av_log(avctx, AV_LOG_ERROR, "Unknown NAL code: %d\n", h->nal_unit_type);
+ }
}
-
+
if(!s->current_picture_ptr) return buf_index; //no frame
-
+
+ s->current_picture_ptr->qscale_type= FF_QSCALE_TYPE_H264;
+ s->current_picture_ptr->pict_type= s->pict_type;
+ s->current_picture_ptr->key_frame= s->pict_type == I_TYPE && h->nal_unit_type == NAL_IDR_SLICE;
+
h->prev_frame_num_offset= h->frame_num_offset;
h->prev_frame_num= h->frame_num;
if(s->current_picture_ptr->reference){
}
/**
- * retunrs the number of bytes consumed for building the current frame
+ * returns the number of bytes consumed for building the current frame
*/
static int get_consumed_bytes(MpegEncContext *s, int pos, int buf_size){
if(s->flags&CODEC_FLAG_TRUNCATED){
pos -= s->parse_context.last_index;
- if(pos<0) pos=0; // FIXME remove (uneeded?)
-
+ if(pos<0) pos=0; // FIXME remove (unneeded?)
+
return pos;
}else{
if(pos==0) pos=1; //avoid infinite loops (i doubt thats needed but ...)
}
}
-static int decode_frame(AVCodecContext *avctx,
+static int decode_frame(AVCodecContext *avctx,
void *data, int *data_size,
uint8_t *buf, int buf_size)
{
H264Context *h = avctx->priv_data;
MpegEncContext *s = &h->s;
- AVFrame *pict = data;
+ AVFrame *pict = data;
int buf_index;
-
+
s->flags= avctx->flags;
s->flags2= avctx->flags2;
if (buf_size == 0) {
return 0;
}
-
+
if(s->flags&CODEC_FLAG_TRUNCATED){
int next= find_frame_end(h, buf, buf_size);
-
+
if( ff_combine_frame(&s->parse_context, next, &buf, &buf_size) < 0 )
return buf_size;
//printf("next:%d buf_size:%d last_index:%d\n", next, buf_size, s->parse_context.last_index);
p += 6;
for (i = 0; i < cnt; i++) {
nalsize = BE_16(p) + 2;
- if(decode_nal_units(h, p, nalsize) != nalsize) {
+ if(decode_nal_units(h, p, nalsize) < 0) {
av_log(avctx, AV_LOG_ERROR, "Decoding sps %d from avcC failed\n", i);
return -1;
}
p += nalsize;
- }
+ }
// Decode pps from avcC
cnt = *(p++); // Number of pps
for (i = 0; i < cnt; i++) {
return -1;
}
p += nalsize;
- }
+ }
// Now store right nal length size, that will be use to parse all other nals
h->nal_length_size = ((*(((char*)(avctx->extradata))+4))&0x03)+1;
// Do not reparse avcC
}
if(!h->is_avc && s->avctx->extradata_size && s->picture_number==0){
- if(0 < decode_nal_units(h, s->avctx->extradata, s->avctx->extradata_size) )
+ if(decode_nal_units(h, s->avctx->extradata, s->avctx->extradata_size) < 0)
return -1;
}
buf_index=decode_nal_units(h, buf, buf_size);
- if(buf_index < 0)
+ if(buf_index < 0)
return -1;
- //FIXME do something with unavailable reference frames
-
-// if(ret==FRAME_SKIPED) return get_consumed_bytes(s, buf_index, buf_size);
+ //FIXME do something with unavailable reference frames
+
+// if(ret==FRAME_SKIPPED) return get_consumed_bytes(s, buf_index, buf_size);
if(!s->current_picture_ptr){
av_log(h->s.avctx, AV_LOG_DEBUG, "error, NO frame\n");
return -1;
}
{
-//#define DECODE_ORDER
Picture *out = s->current_picture_ptr;
-#ifndef DECODE_ORDER
- /* Sort B-frames into display order
- * FIXME doesn't allow for multiple delayed frames */
+#if 0 //decode order
+ *data_size = sizeof(AVFrame);
+#else
+ /* Sort B-frames into display order */
Picture *cur = s->current_picture_ptr;
- Picture *prev = h->delayed_pic[0];
+ Picture *prev = h->delayed_output_pic;
+ int out_idx = 0;
+ int pics = 0;
+ int out_of_order;
+ int cross_idr = 0;
+ int dropped_frame = 0;
+ int i;
- if(s->low_delay
- && (cur->pict_type == B_TYPE
- || (!h->sps.gaps_in_frame_num_allowed_flag
- && prev && cur->poc - prev->poc > 2))){
+ if(h->sps.bitstream_restriction_flag
+ && s->avctx->has_b_frames < h->sps.num_reorder_frames){
+ s->avctx->has_b_frames = h->sps.num_reorder_frames;
s->low_delay = 0;
- s->avctx->has_b_frames = 1;
- if(prev && prev->poc > cur->poc)
- // too late to display this frame
- cur = prev;
}
- if(s->low_delay || !prev || cur->pict_type == B_TYPE)
- out = cur;
- else
+ while(h->delayed_pic[pics]) pics++;
+ h->delayed_pic[pics++] = cur;
+ if(cur->reference == 0)
+ cur->reference = 1;
+
+ for(i=0; h->delayed_pic[i]; i++)
+ if(h->delayed_pic[i]->key_frame || h->delayed_pic[i]->poc==0)
+ cross_idr = 1;
+
+ out = h->delayed_pic[0];
+ for(i=1; h->delayed_pic[i] && !h->delayed_pic[i]->key_frame; i++)
+ if(h->delayed_pic[i]->poc < out->poc){
+ out = h->delayed_pic[i];
+ out_idx = i;
+ }
+
+ out_of_order = !cross_idr && prev && out->poc < prev->poc;
+ if(prev && pics <= s->avctx->has_b_frames)
+ out = prev;
+ else if((out_of_order && pics-1 == s->avctx->has_b_frames && pics < 15)
+ || (s->low_delay &&
+ ((!cross_idr && prev && out->poc > prev->poc + 2)
+ || cur->pict_type == B_TYPE)))
+ {
+ s->low_delay = 0;
+ s->avctx->has_b_frames++;
out = prev;
- if(s->low_delay || !prev || out == prev){
- if(prev && prev->reference == 1)
- prev->reference = 0;
- h->delayed_pic[0] = cur;
}
+ else if(out_of_order)
+ out = prev;
+
+ if(out_of_order || pics > s->avctx->has_b_frames){
+ dropped_frame = (out != h->delayed_pic[out_idx]);
+ for(i=out_idx; h->delayed_pic[i]; i++)
+ h->delayed_pic[i] = h->delayed_pic[i+1];
+ }
+
+ if(prev == out && !dropped_frame)
+ *data_size = 0;
+ else
+ *data_size = sizeof(AVFrame);
+ if(prev && prev != out && prev->reference == 1)
+ prev->reference = 0;
+ h->delayed_output_pic = out;
#endif
- *pict= *(AVFrame*)out;
+ if(out)
+ *pict= *(AVFrame*)out;
+ else
+ av_log(avctx, AV_LOG_DEBUG, "no picture\n");
}
+ assert(pict->data[0] || !*data_size);
ff_print_debug_info(s, pict);
- assert(pict->data[0]);
//printf("out %d\n", (int)pict->data[0]);
#if 0 //?
/* we substract 1 because it is added on utils.c */
avctx->frame_number = s->picture_number - 1;
#endif
-#if 0
- /* dont output the last pic after seeking */
- if(s->last_picture_ptr || s->low_delay)
- //Note this isnt a issue as a IDR pic should flush the buffers
-#endif
- *data_size = sizeof(AVFrame);
return get_consumed_bytes(s, buf_index, buf_size);
}
#if 0
// int int_temp[10000];
DSPContext dsp;
AVCodecContext avctx;
-
+
dsputil_init(&dsp, &avctx);
init_put_bits(&pb, temp, SIZE);
STOP_TIMER("set_ue_golomb");
}
flush_put_bits(&pb);
-
+
init_get_bits(&gb, temp, 8*SIZE);
for(i=0; i<COUNT; i++){
int j, s;
-
+
s= show_bits(&gb, 24);
-
+
START_TIMER
j= get_ue_golomb(&gb);
if(j != i){
}
STOP_TIMER("get_ue_golomb");
}
-
-
+
+
init_put_bits(&pb, temp, SIZE);
printf("testing signed exp golomb\n");
for(i=0; i<COUNT; i++){
STOP_TIMER("set_se_golomb");
}
flush_put_bits(&pb);
-
+
init_get_bits(&gb, temp, 8*SIZE);
for(i=0; i<COUNT; i++){
int j, s;
-
+
s= show_bits(&gb, 24);
-
+
START_TIMER
j= get_se_golomb(&gb);
if(j != i - COUNT/2){
}
printf("testing 4x4 (I)DCT\n");
-
+
DCTELEM block[16];
uint8_t src[16], ref[16];
uint64_t error= 0, max_error=0;
}
h264_diff_dct_c(block, src, ref, 4);
-
+
//normalize
for(j=0; j<16; j++){
// printf("%d ", block[j]);
if(j&4) block[j]= (block[j]*4 + 2)/5;
}
// printf("\n");
-
+
s->dsp.h264_idct_add(ref, block, 4);
/* for(j=0; j<16; j++){
printf("%d ", ref[j]);
}
printf("\n");*/
-
+
for(j=0; j<16; j++){
int diff= ABS(src[j] - ref[j]);
-
+
error+= diff*diff;
max_error= FFMAX(max_error, diff);
}
for(qp=0; qp<52; qp++){
for(i=0; i<16; i++)
src1_block[i]= src2_block[i]= random()%255;
-
+
}
#endif
printf("Testing NAL layer\n");
-
+
uint8_t bitstream[COUNT];
uint8_t nal[COUNT*2];
H264Context h;
memset(&h, 0, sizeof(H264Context));
-
+
for(i=0; i<COUNT; i++){
int zeros= i;
int nal_length;
int out_length;
uint8_t *out;
int j;
-
+
for(j=0; j<COUNT; j++){
bitstream[j]= (random() % 255) + 1;
}
-
+
for(j=0; j<zeros; j++){
int pos= random() % COUNT;
while(bitstream[pos] == 0){
}
bitstream[pos]=0;
}
-
+
START_TIMER
-
+
nal_length= encode_nal(&h, nal, bitstream, COUNT, COUNT*2);
if(nal_length<0){
printf("encoding failed\n");
return -1;
}
-
+
out= decode_nal(&h, nal, &out_length, &consumed, nal_length);
STOP_TIMER("NAL")
-
+
if(out_length != COUNT){
printf("incorrect length %d %d\n", out_length, COUNT);
return -1;
}
-
+
if(consumed != nal_length){
printf("incorrect consumed length %d %d\n", nal_length, consumed);
return -1;
}
-
+
if(memcmp(bitstream, out, COUNT)){
printf("missmatch\n");
return -1;
}
}
-
+
printf("Testing RBSP\n");
-
-
+
+
return 0;
}
#endif
{
H264Context *h = avctx->priv_data;
MpegEncContext *s = &h->s;
-
+
+ av_freep(&h->rbsp_buffer);
free_tables(h); //FIXME cleanup init stuff perhaps
MPV_common_end(s);
// memset(h, 0, sizeof(H264Context));
-
+
return 0;
}
decode_end,
decode_frame,
/*CODEC_CAP_DRAW_HORIZ_BAND |*/ CODEC_CAP_DR1 | CODEC_CAP_TRUNCATED | CODEC_CAP_DELAY,
+ .flush= flush_dpb,
};
AVCodecParser h264_parser = {
NULL,
h264_parse,
ff_parse_close,
+ h264_split,
};
#include "svq3.c"