Mercurial > libavcodec.hg
comparison h264.c @ 12492:58a960d6e34c libavcodec
Rename h264_idct_sse2.asm to h264_idct.asm; move inline IDCT asm from
h264dsp_mmx.c to h264_idct.asm (as yasm code). Because the loops are now
coded in asm instead of C, this is (depending on the function) up to 50%
faster for cases where gcc didn't do a great job at looping.
Since h264_idct_add8() is now faster than the manual loop setup in h264.c,
in-asm idct calling can now be enabled for chroma as well (see r16207). For
MMX, this is 5% faster. For SSE2 (which isn't done for chroma if h264.c does
the looping), this makes it up to 50% faster. Speed gain overall is ~0.5-1.0%.
author | rbultje |
---|---|
date | Tue, 14 Sep 2010 13:36:26 +0000 |
parents | f7bedc1ce1bc |
children | 7c0dbd8eb53a |
comparison
equal
deleted
inserted
replaced
12491:990f8a5fc8af | 12492:58a960d6e34c |
---|---|
1316 } | 1316 } |
1317 }else{ | 1317 }else{ |
1318 chroma_dc_dequant_idct_c(h->mb + 16*16, h->chroma_qp[0], h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]); | 1318 chroma_dc_dequant_idct_c(h->mb + 16*16, h->chroma_qp[0], h->dequant4_coeff[IS_INTRA(mb_type) ? 1:4][h->chroma_qp[0]][0]); |
1319 chroma_dc_dequant_idct_c(h->mb + 16*16+4*16, h->chroma_qp[1], h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]); | 1319 chroma_dc_dequant_idct_c(h->mb + 16*16+4*16, h->chroma_qp[1], h->dequant4_coeff[IS_INTRA(mb_type) ? 2:5][h->chroma_qp[1]][0]); |
1320 if(is_h264){ | 1320 if(is_h264){ |
1321 idct_add = h->h264dsp.h264_idct_add; | 1321 h->h264dsp.h264_idct_add8(dest, block_offset, |
1322 idct_dc_add = h->h264dsp.h264_idct_dc_add; | 1322 h->mb, uvlinesize, |
1323 for(i=16; i<16+8; i++){ | 1323 h->non_zero_count_cache); |
1324 if(h->non_zero_count_cache[ scan8[i] ]) | |
1325 idct_add (dest[(i&4)>>2] + block_offset[i], h->mb + i*16, uvlinesize); | |
1326 else if(h->mb[i*16]) | |
1327 idct_dc_add(dest[(i&4)>>2] + block_offset[i], h->mb + i*16, uvlinesize); | |
1328 } | |
1329 }else{ | 1324 }else{ |
1330 for(i=16; i<16+8; i++){ | 1325 for(i=16; i<16+8; i++){ |
1331 if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ | 1326 if(h->non_zero_count_cache[ scan8[i] ] || h->mb[i*16]){ |
1332 uint8_t * const ptr= dest[(i&4)>>2] + block_offset[i]; | 1327 uint8_t * const ptr= dest[(i&4)>>2] + block_offset[i]; |
1333 ff_svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, ff_h264_chroma_qp[s->qscale + 12] - 12, 2); | 1328 ff_svq3_add_idct_c(ptr, h->mb + i*16, uvlinesize, ff_h264_chroma_qp[s->qscale + 12] - 12, 2); |