Mercurial > libavcodec.hg
comparison rv34.c @ 10117:d0f2f4e4436c libavcodec
Adjust r->avail_cache[] indexes to avoid alignment issues when zeroing block
inside it.
| author | kostya |
|---|---|
| date | Sun, 30 Aug 2009 06:16:33 +0000 |
| parents | 305536ce781f |
| children | 4dafbd9c8918 |
comparison
equal
deleted
inserted
replaced
| 10116:6cade2cdd63b | 10117:d0f2f4e4436c |
|---|---|
| 438 | 438 |
| 439 /** macroblock partition height in 8x8 blocks */ | 439 /** macroblock partition height in 8x8 blocks */ |
| 440 static const uint8_t part_sizes_h[RV34_MB_TYPES] = { 2, 2, 2, 1, 2, 2, 2, 2, 1, 2, 2, 2 }; | 440 static const uint8_t part_sizes_h[RV34_MB_TYPES] = { 2, 2, 2, 1, 2, 2, 2, 2, 1, 2, 2, 2 }; |
| 441 | 441 |
| 442 /** availability index for subblocks */ | 442 /** availability index for subblocks */ |
| 443 static const uint8_t avail_indexes[4] = { 5, 6, 9, 10 }; | 443 static const uint8_t avail_indexes[4] = { 6, 7, 10, 11 }; |
| 444 | 444 |
| 445 /** | 445 /** |
| 446 * motion vector prediction | 446 * motion vector prediction |
| 447 * | 447 * |
| 448 * Motion prediction performed for the block by using median prediction of | 448 * Motion prediction performed for the block by using median prediction of |
| 551 int type = cur_pic->mb_type[mb_pos]; | 551 int type = cur_pic->mb_type[mb_pos]; |
| 552 | 552 |
| 553 memset(A, 0, sizeof(A)); | 553 memset(A, 0, sizeof(A)); |
| 554 memset(B, 0, sizeof(B)); | 554 memset(B, 0, sizeof(B)); |
| 555 memset(C, 0, sizeof(C)); | 555 memset(C, 0, sizeof(C)); |
| 556 if((r->avail_cache[5-1] & type) & mask){ | 556 if((r->avail_cache[6-1] & type) & mask){ |
| 557 A[0] = cur_pic->motion_val[dir][mv_pos - 1][0]; | 557 A[0] = cur_pic->motion_val[dir][mv_pos - 1][0]; |
| 558 A[1] = cur_pic->motion_val[dir][mv_pos - 1][1]; | 558 A[1] = cur_pic->motion_val[dir][mv_pos - 1][1]; |
| 559 has_A = 1; | 559 has_A = 1; |
| 560 } | 560 } |
| 561 if((r->avail_cache[5-4] & type) & mask){ | 561 if((r->avail_cache[6-4] & type) & mask){ |
| 562 B[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride][0]; | 562 B[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride][0]; |
| 563 B[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride][1]; | 563 B[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride][1]; |
| 564 has_B = 1; | 564 has_B = 1; |
| 565 } | 565 } |
| 566 if(r->avail_cache[5-4] && (r->avail_cache[5-2] & type) & mask){ | 566 if(r->avail_cache[6-4] && (r->avail_cache[6-2] & type) & mask){ |
| 567 C[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride + 2][0]; | 567 C[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride + 2][0]; |
| 568 C[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride + 2][1]; | 568 C[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride + 2][1]; |
| 569 has_C = 1; | 569 has_C = 1; |
| 570 }else if((s->mb_x+1) == s->mb_width && (r->avail_cache[5-5] & type) & mask){ | 570 }else if((s->mb_x+1) == s->mb_width && (r->avail_cache[6-5] & type) & mask){ |
| 571 C[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride - 1][0]; | 571 C[0] = cur_pic->motion_val[dir][mv_pos - s->b8_stride - 1][0]; |
| 572 C[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride - 1][1]; | 572 C[1] = cur_pic->motion_val[dir][mv_pos - s->b8_stride - 1][1]; |
| 573 has_C = 1; | 573 has_C = 1; |
| 574 } | 574 } |
| 575 | 575 |
| 956 int itype; | 956 int itype; |
| 957 int avail[6*8] = {0}; | 957 int avail[6*8] = {0}; |
| 958 int idx; | 958 int idx; |
| 959 | 959 |
| 960 // Set neighbour information. | 960 // Set neighbour information. |
| 961 if(r->avail_cache[0]) | 961 if(r->avail_cache[1]) |
| 962 avail[0] = 1; | 962 avail[0] = 1; |
| 963 if(r->avail_cache[1]) | 963 if(r->avail_cache[2]) |
| 964 avail[1] = avail[2] = 1; | 964 avail[1] = avail[2] = 1; |
| 965 if(r->avail_cache[2]) | 965 if(r->avail_cache[3]) |
| 966 avail[3] = avail[4] = 1; | 966 avail[3] = avail[4] = 1; |
| 967 if(r->avail_cache[3]) | 967 if(r->avail_cache[4]) |
| 968 avail[5] = 1; | 968 avail[5] = 1; |
| 969 if(r->avail_cache[4]) | 969 if(r->avail_cache[5]) |
| 970 avail[8] = avail[16] = 1; | 970 avail[8] = avail[16] = 1; |
| 971 if(r->avail_cache[8]) | 971 if(r->avail_cache[9]) |
| 972 avail[24] = avail[32] = 1; | 972 avail[24] = avail[32] = 1; |
| 973 | 973 |
| 974 Y = s->dest[0]; | 974 Y = s->dest[0]; |
| 975 U = s->dest[1]; | 975 U = s->dest[1]; |
| 976 V = s->dest[2]; | 976 V = s->dest[2]; |
| 985 } | 985 } |
| 986 Y += s->linesize * 4 - 4*4; | 986 Y += s->linesize * 4 - 4*4; |
| 987 intra_types += r->intra_types_stride; | 987 intra_types += r->intra_types_stride; |
| 988 } | 988 } |
| 989 intra_types -= r->intra_types_stride * 4; | 989 intra_types -= r->intra_types_stride * 4; |
| 990 fill_rectangle(r->avail_cache + 5, 2, 2, 4, 0, 4); | 990 fill_rectangle(r->avail_cache + 6, 2, 2, 4, 0, 4); |
| 991 for(j = 0; j < 2; j++){ | 991 for(j = 0; j < 2; j++){ |
| 992 idx = 5 + j*4; | 992 idx = 6 + j*4; |
| 993 for(i = 0; i < 2; i++, cbp >>= 1, idx++){ | 993 for(i = 0; i < 2; i++, cbp >>= 1, idx++){ |
| 994 rv34_pred_4x4_block(r, U + i*4 + j*4*s->uvlinesize, s->uvlinesize, ittrans[intra_types[i*2+j*2*r->intra_types_stride]], r->avail_cache[idx-4], r->avail_cache[idx-1], !i && !j, r->avail_cache[idx-3]); | 994 rv34_pred_4x4_block(r, U + i*4 + j*4*s->uvlinesize, s->uvlinesize, ittrans[intra_types[i*2+j*2*r->intra_types_stride]], r->avail_cache[idx-4], r->avail_cache[idx-1], !i && !j, r->avail_cache[idx-3]); |
| 995 rv34_pred_4x4_block(r, V + i*4 + j*4*s->uvlinesize, s->uvlinesize, ittrans[intra_types[i*2+j*2*r->intra_types_stride]], r->avail_cache[idx-4], r->avail_cache[idx-1], !i && !j, r->avail_cache[idx-3]); | 995 rv34_pred_4x4_block(r, V + i*4 + j*4*s->uvlinesize, s->uvlinesize, ittrans[intra_types[i*2+j*2*r->intra_types_stride]], r->avail_cache[idx-4], r->avail_cache[idx-1], !i && !j, r->avail_cache[idx-3]); |
| 996 r->avail_cache[idx] = 1; | 996 r->avail_cache[idx] = 1; |
| 997 if(cbp & 0x01) | 997 if(cbp & 0x01) |
| 1000 rv34_add_4x4_block(V + i*4 + j*4*s->uvlinesize, s->uvlinesize, s->block[5], i*4+j*32); | 1000 rv34_add_4x4_block(V + i*4 + j*4*s->uvlinesize, s->uvlinesize, s->block[5], i*4+j*32); |
| 1001 } | 1001 } |
| 1002 } | 1002 } |
| 1003 }else{ | 1003 }else{ |
| 1004 itype = ittrans16[intra_types[0]]; | 1004 itype = ittrans16[intra_types[0]]; |
| 1005 itype = adjust_pred16(itype, r->avail_cache[5-4], r->avail_cache[5-1]); | 1005 itype = adjust_pred16(itype, r->avail_cache[6-4], r->avail_cache[6-1]); |
| 1006 r->h.pred16x16[itype](Y, s->linesize); | 1006 r->h.pred16x16[itype](Y, s->linesize); |
| 1007 dsp->add_pixels_clamped(s->block[0], Y, s->linesize); | 1007 dsp->add_pixels_clamped(s->block[0], Y, s->linesize); |
| 1008 dsp->add_pixels_clamped(s->block[1], Y + 8, s->linesize); | 1008 dsp->add_pixels_clamped(s->block[1], Y + 8, s->linesize); |
| 1009 Y += s->linesize * 8; | 1009 Y += s->linesize * 8; |
| 1010 dsp->add_pixels_clamped(s->block[2], Y, s->linesize); | 1010 dsp->add_pixels_clamped(s->block[2], Y, s->linesize); |
| 1011 dsp->add_pixels_clamped(s->block[3], Y + 8, s->linesize); | 1011 dsp->add_pixels_clamped(s->block[3], Y + 8, s->linesize); |
| 1012 | 1012 |
| 1013 itype = ittrans16[intra_types[0]]; | 1013 itype = ittrans16[intra_types[0]]; |
| 1014 if(itype == PLANE_PRED8x8) itype = DC_PRED8x8; | 1014 if(itype == PLANE_PRED8x8) itype = DC_PRED8x8; |
| 1015 itype = adjust_pred16(itype, r->avail_cache[5-4], r->avail_cache[5-1]); | 1015 itype = adjust_pred16(itype, r->avail_cache[6-4], r->avail_cache[6-1]); |
| 1016 r->h.pred8x8[itype](U, s->uvlinesize); | 1016 r->h.pred8x8[itype](U, s->uvlinesize); |
| 1017 dsp->add_pixels_clamped(s->block[4], U, s->uvlinesize); | 1017 dsp->add_pixels_clamped(s->block[4], U, s->uvlinesize); |
| 1018 r->h.pred8x8[itype](V, s->uvlinesize); | 1018 r->h.pred8x8[itype](V, s->uvlinesize); |
| 1019 dsp->add_pixels_clamped(s->block[5], V, s->uvlinesize); | 1019 dsp->add_pixels_clamped(s->block[5], V, s->uvlinesize); |
| 1020 } | 1020 } |
| 1171 int dist; | 1171 int dist; |
| 1172 int mb_pos = s->mb_x + s->mb_y * s->mb_stride; | 1172 int mb_pos = s->mb_x + s->mb_y * s->mb_stride; |
| 1173 | 1173 |
| 1174 // Calculate which neighbours are available. Maybe it's worth optimizing too. | 1174 // Calculate which neighbours are available. Maybe it's worth optimizing too. |
| 1175 memset(r->avail_cache, 0, sizeof(r->avail_cache)); | 1175 memset(r->avail_cache, 0, sizeof(r->avail_cache)); |
| 1176 fill_rectangle(r->avail_cache + 5, 2, 2, 4, 1, 4); | 1176 fill_rectangle(r->avail_cache + 6, 2, 2, 4, 1, 4); |
| 1177 dist = (s->mb_x - s->resync_mb_x) + (s->mb_y - s->resync_mb_y) * s->mb_width; | 1177 dist = (s->mb_x - s->resync_mb_x) + (s->mb_y - s->resync_mb_y) * s->mb_width; |
| 1178 if(s->mb_x && dist) | 1178 if(s->mb_x && dist) |
| 1179 r->avail_cache[4] = | 1179 r->avail_cache[5] = |
| 1180 r->avail_cache[8] = s->current_picture_ptr->mb_type[mb_pos - 1]; | 1180 r->avail_cache[9] = s->current_picture_ptr->mb_type[mb_pos - 1]; |
| 1181 if(dist >= s->mb_width) | 1181 if(dist >= s->mb_width) |
| 1182 r->avail_cache[1] = | 1182 r->avail_cache[2] = |
| 1183 r->avail_cache[2] = s->current_picture_ptr->mb_type[mb_pos - s->mb_stride]; | 1183 r->avail_cache[3] = s->current_picture_ptr->mb_type[mb_pos - s->mb_stride]; |
| 1184 if(((s->mb_x+1) < s->mb_width) && dist >= s->mb_width - 1) | 1184 if(((s->mb_x+1) < s->mb_width) && dist >= s->mb_width - 1) |
| 1185 r->avail_cache[3] = s->current_picture_ptr->mb_type[mb_pos - s->mb_stride + 1]; | 1185 r->avail_cache[4] = s->current_picture_ptr->mb_type[mb_pos - s->mb_stride + 1]; |
| 1186 if(s->mb_x && dist > s->mb_width) | 1186 if(s->mb_x && dist > s->mb_width) |
| 1187 r->avail_cache[0] = s->current_picture_ptr->mb_type[mb_pos - s->mb_stride - 1]; | 1187 r->avail_cache[1] = s->current_picture_ptr->mb_type[mb_pos - s->mb_stride - 1]; |
| 1188 | 1188 |
| 1189 s->qscale = r->si.quant; | 1189 s->qscale = r->si.quant; |
| 1190 cbp = cbp2 = rv34_decode_mb_header(r, intra_types); | 1190 cbp = cbp2 = rv34_decode_mb_header(r, intra_types); |
| 1191 r->cbp_luma [mb_pos] = cbp; | 1191 r->cbp_luma [mb_pos] = cbp; |
| 1192 r->cbp_chroma[mb_pos] = cbp >> 16; | 1192 r->cbp_chroma[mb_pos] = cbp >> 16; |