Mercurial > libavcodec.hg
comparison cavs.c @ 3472:0e0cffe06fb3 libavcodec
avoid double copying of left border in luma prediction
| author | stefang |
|---|---|
| date | Fri, 14 Jul 2006 19:31:58 +0000 |
| parents | c2db6e76b269 |
| children | 419409926166 |
comparison
equal
deleted
inserted
replaced
| 3471:c2db6e76b269 | 3472:0e0cffe06fb3 |
|---|---|
| 80 int cbp; | 80 int cbp; |
| 81 | 81 |
| 82 /** intra prediction is done with un-deblocked samples | 82 /** intra prediction is done with un-deblocked samples |
| 83 they are saved here before deblocking the MB */ | 83 they are saved here before deblocking the MB */ |
| 84 uint8_t *top_border_y, *top_border_u, *top_border_v; | 84 uint8_t *top_border_y, *top_border_u, *top_border_v; |
| 85 uint8_t left_border_y[16], left_border_u[10], left_border_v[10]; | 85 uint8_t left_border_y[26], left_border_u[10], left_border_v[10]; |
| 86 uint8_t intern_border_y[26]; | |
| 86 uint8_t topleft_border_y, topleft_border_u, topleft_border_v; | 87 uint8_t topleft_border_y, topleft_border_u, topleft_border_v; |
| 87 | 88 |
| 88 void (*intra_pred_l[8])(uint8_t *d,uint8_t *top,uint8_t *left,int stride); | 89 void (*intra_pred_l[8])(uint8_t *d,uint8_t *top,uint8_t *left,int stride); |
| 89 void (*intra_pred_c[7])(uint8_t *d,uint8_t *top,uint8_t *left,int stride); | 90 void (*intra_pred_c[7])(uint8_t *d,uint8_t *top,uint8_t *left,int stride); |
| 90 uint8_t *col_type_base; | 91 uint8_t *col_type_base; |
| 149 h->topleft_border_v = h->top_border_v[h->mbx*10+8]; | 150 h->topleft_border_v = h->top_border_v[h->mbx*10+8]; |
| 150 memcpy(&h->top_border_y[h->mbx*16], h->cy + 15* h->l_stride,16); | 151 memcpy(&h->top_border_y[h->mbx*16], h->cy + 15* h->l_stride,16); |
| 151 memcpy(&h->top_border_u[h->mbx*10+1], h->cu + 7* h->c_stride,8); | 152 memcpy(&h->top_border_u[h->mbx*10+1], h->cu + 7* h->c_stride,8); |
| 152 memcpy(&h->top_border_v[h->mbx*10+1], h->cv + 7* h->c_stride,8); | 153 memcpy(&h->top_border_v[h->mbx*10+1], h->cv + 7* h->c_stride,8); |
| 153 for(i=0;i<8;i++) { | 154 for(i=0;i<8;i++) { |
| 154 h->left_border_y[i*2+0] = *(h->cy + 15 + (i*2+0)*h->l_stride); | 155 h->left_border_y[i*2+1] = *(h->cy + 15 + (i*2+0)*h->l_stride); |
| 155 h->left_border_y[i*2+1] = *(h->cy + 15 + (i*2+1)*h->l_stride); | 156 h->left_border_y[i*2+2] = *(h->cy + 15 + (i*2+1)*h->l_stride); |
| 156 h->left_border_u[i+1] = *(h->cu + 7 + i*h->c_stride); | 157 h->left_border_u[i+1] = *(h->cu + 7 + i*h->c_stride); |
| 157 h->left_border_v[i+1] = *(h->cv + 7 + i*h->c_stride); | 158 h->left_border_v[i+1] = *(h->cv + 7 + i*h->c_stride); |
| 158 } | 159 } |
| 159 if(!h->loop_filter_disable) { | 160 if(!h->loop_filter_disable) { |
| 160 /* determine bs */ | 161 /* determine bs */ |
| 209 * spatial intra prediction | 210 * spatial intra prediction |
| 210 * | 211 * |
| 211 ****************************************************************************/ | 212 ****************************************************************************/ |
| 212 | 213 |
| 213 static inline void load_intra_pred_luma(AVSContext *h, uint8_t *top, | 214 static inline void load_intra_pred_luma(AVSContext *h, uint8_t *top, |
| 214 uint8_t *left, int block) { | 215 uint8_t **left, int block) { |
| 215 int i; | 216 int i; |
| 216 | 217 |
| 217 switch(block) { | 218 switch(block) { |
| 218 case 0: | 219 case 0: |
| 219 memcpy(&left[1],h->left_border_y,16); | 220 *left = h->left_border_y; |
| 220 left[0] = left[1]; | 221 h->left_border_y[0] = h->left_border_y[1]; |
| 221 left[17] = left[16]; | 222 memset(&h->left_border_y[17],h->left_border_y[16],9); |
| 222 memcpy(&top[1],&h->top_border_y[h->mbx*16],16); | 223 memcpy(&top[1],&h->top_border_y[h->mbx*16],16); |
| 223 top[17] = top[16]; | 224 top[17] = top[16]; |
| 224 top[0] = top[1]; | 225 top[0] = top[1]; |
| 225 if((h->flags & A_AVAIL) && (h->flags & B_AVAIL)) | 226 if((h->flags & A_AVAIL) && (h->flags & B_AVAIL)) |
| 226 left[0] = top[0] = h->topleft_border_y; | 227 h->left_border_y[0] = top[0] = h->topleft_border_y; |
| 227 break; | 228 break; |
| 228 case 1: | 229 case 1: |
| 230 *left = h->intern_border_y; | |
| 229 for(i=0;i<8;i++) | 231 for(i=0;i<8;i++) |
| 230 left[i+1] = *(h->cy + 7 + i*h->l_stride); | 232 h->intern_border_y[i+1] = *(h->cy + 7 + i*h->l_stride); |
| 231 memset(&left[9],left[8],9); | 233 memset(&h->intern_border_y[9],h->intern_border_y[8],9); |
| 232 left[0] = left[1]; | 234 h->intern_border_y[0] = h->intern_border_y[1]; |
| 233 memcpy(&top[1],&h->top_border_y[h->mbx*16+8],8); | 235 memcpy(&top[1],&h->top_border_y[h->mbx*16+8],8); |
| 234 if(h->flags & C_AVAIL) | 236 if(h->flags & C_AVAIL) |
| 235 memcpy(&top[9],&h->top_border_y[(h->mbx + 1)*16],8); | 237 memcpy(&top[9],&h->top_border_y[(h->mbx + 1)*16],8); |
| 236 else | 238 else |
| 237 memset(&top[9],top[8],9); | 239 memset(&top[9],top[8],9); |
| 238 top[17] = top[16]; | 240 top[17] = top[16]; |
| 239 top[0] = top[1]; | 241 top[0] = top[1]; |
| 240 if(h->flags & B_AVAIL) | 242 if(h->flags & B_AVAIL) |
| 241 left[0] = top[0] = h->top_border_y[h->mbx*16+7]; | 243 h->intern_border_y[0] = top[0] = h->top_border_y[h->mbx*16+7]; |
| 242 break; | 244 break; |
| 243 case 2: | 245 case 2: |
| 244 memcpy(&left[1],&h->left_border_y[8],8); | 246 *left = &h->left_border_y[8]; |
| 245 memset(&left[9],left[8],9); | |
| 246 memcpy(&top[1],h->cy + 7*h->l_stride,16); | 247 memcpy(&top[1],h->cy + 7*h->l_stride,16); |
| 247 top[17] = top[16]; | 248 top[17] = top[16]; |
| 248 left[0] = h->left_border_y[7]; | |
| 249 top[0] = top[1]; | 249 top[0] = top[1]; |
| 250 if(h->flags & A_AVAIL) | 250 if(h->flags & A_AVAIL) |
| 251 top[0] = left[0]; | 251 top[0] = h->left_border_y[8]; |
| 252 break; | 252 break; |
| 253 case 3: | 253 case 3: |
| 254 for(i=0;i<9;i++) | 254 *left = &h->intern_border_y[8]; |
| 255 left[i] = *(h->cy + 7 + (i+7)*h->l_stride); | 255 for(i=0;i<8;i++) |
| 256 memset(&left[9],left[8],9); | 256 h->intern_border_y[i+9] = *(h->cy + 7 + (i+8)*h->l_stride); |
| 257 memset(&h->intern_border_y[17],h->intern_border_y[16],9); | |
| 257 memcpy(&top[0],h->cy + 7 + 7*h->l_stride,9); | 258 memcpy(&top[0],h->cy + 7 + 7*h->l_stride,9); |
| 258 memset(&top[9],top[8],9); | 259 memset(&top[9],top[8],9); |
| 259 break; | 260 break; |
| 260 } | 261 } |
| 261 } | 262 } |
| 814 | 815 |
| 815 static int decode_mb_i(AVSContext *h, int cbp_code) { | 816 static int decode_mb_i(AVSContext *h, int cbp_code) { |
| 816 GetBitContext *gb = &h->s.gb; | 817 GetBitContext *gb = &h->s.gb; |
| 817 int block, pred_mode_uv; | 818 int block, pred_mode_uv; |
| 818 uint8_t top[18]; | 819 uint8_t top[18]; |
| 819 uint8_t left[18]; | 820 uint8_t *left = NULL; |
| 820 uint8_t *d; | 821 uint8_t *d; |
| 821 | 822 |
| 822 init_mb(h); | 823 init_mb(h); |
| 823 | 824 |
| 824 /* get intra prediction modes from stream */ | 825 /* get intra prediction modes from stream */ |
| 873 h->qp = (h->qp + get_se_golomb(gb)) & 63; //qp_delta | 874 h->qp = (h->qp + get_se_golomb(gb)) & 63; //qp_delta |
| 874 | 875 |
| 875 /* luma intra prediction interleaved with residual decode/transform/add */ | 876 /* luma intra prediction interleaved with residual decode/transform/add */ |
| 876 for(block=0;block<4;block++) { | 877 for(block=0;block<4;block++) { |
| 877 d = h->cy + h->luma_scan[block]; | 878 d = h->cy + h->luma_scan[block]; |
| 878 load_intra_pred_luma(h, top, left, block); | 879 load_intra_pred_luma(h, top, &left, block); |
| 879 h->intra_pred_l[h->pred_mode_Y[scan3x3[block]]] | 880 h->intra_pred_l[h->pred_mode_Y[scan3x3[block]]] |
| 880 (d, top, left, h->l_stride); | 881 (d, top, left, h->l_stride); |
| 881 if(h->cbp & (1<<block)) | 882 if(h->cbp & (1<<block)) |
| 882 decode_residual_block(h,gb,intra_2dvlc,1,h->qp,d,h->l_stride); | 883 decode_residual_block(h,gb,intra_2dvlc,1,h->qp,d,h->l_stride); |
| 883 } | 884 } |