Mercurial > libavcodec.hg
comparison h264_cabac.c @ 10906:1b5fba731e24 libavcodec
Rearchitecturing the stiched up goose part 1
Run loop filter per row instead of per MB, this also should make it
much easier to switch to per frame filtering and also doing so in a
seperate thread in the future if some volunteer wants to try.
Overall decoding speedup of 1.7% (single thread on pentium dual / cathedral sample)
This change also allows some optimizations to be tried that would not have
been possible before.
author | michael |
---|---|
date | Sun, 17 Jan 2010 20:35:55 +0000 |
parents | 2aafcafbe1f0 |
children | 2d82b73b12ef |
comparison
equal
deleted
inserted
replaced
10905:06d20a468d1e | 10906:1b5fba731e24 |
---|---|
1390 h->cbp_table[mb_xy] = 0x1ef; | 1390 h->cbp_table[mb_xy] = 0x1ef; |
1391 h->chroma_pred_mode_table[mb_xy] = 0; | 1391 h->chroma_pred_mode_table[mb_xy] = 0; |
1392 // In deblocking, the quantizer is 0 | 1392 // In deblocking, the quantizer is 0 |
1393 s->current_picture.qscale_table[mb_xy]= 0; | 1393 s->current_picture.qscale_table[mb_xy]= 0; |
1394 // All coeffs are present | 1394 // All coeffs are present |
1395 memset(h->non_zero_count[mb_xy], 16, 16); | 1395 memset(h->non_zero_count[mb_xy], 16, 32); |
1396 s->current_picture.mb_type[mb_xy]= mb_type; | 1396 s->current_picture.mb_type[mb_xy]= mb_type; |
1397 h->last_qscale_diff = 0; | 1397 h->last_qscale_diff = 0; |
1398 return 0; | 1398 return 0; |
1399 } | 1399 } |
1400 | 1400 |