Mercurial > libavcodec.hg
comparison h264.c @ 2707:360024d31dab libavcodec
H.264 deblocking optimizations (mmx for chroma_bS4 case, convert existing cases to 8-bit math)
| author | lorenm |
|---|---|
| date | Wed, 18 May 2005 09:17:22 +0000 |
| parents | 7d86ebbccecf |
| children | 16258de8ae06 |
comparison
equal
deleted
inserted
replaced
| 2706:a23c44f7ed60 | 2707:360024d31dab |
|---|---|
| 5663 const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 ); | 5663 const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 ); |
| 5664 const int alpha = alpha_table[index_a]; | 5664 const int alpha = alpha_table[index_a]; |
| 5665 const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )]; | 5665 const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )]; |
| 5666 | 5666 |
| 5667 if( bS[0] < 4 ) { | 5667 if( bS[0] < 4 ) { |
| 5668 int tc[4]; | 5668 int8_t tc[4]; |
| 5669 for(i=0; i<4; i++) | 5669 for(i=0; i<4; i++) |
| 5670 tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] : -1; | 5670 tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] : -1; |
| 5671 h->s.dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc); | 5671 h->s.dsp.h264_h_loop_filter_luma(pix, stride, alpha, beta, tc); |
| 5672 } else { | 5672 } else { |
| 5673 /* 16px edge length, because bS=4 is triggered by being at | 5673 /* 16px edge length, because bS=4 is triggered by being at |
| 5724 const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 ); | 5724 const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 ); |
| 5725 const int alpha = alpha_table[index_a]; | 5725 const int alpha = alpha_table[index_a]; |
| 5726 const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )]; | 5726 const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )]; |
| 5727 | 5727 |
| 5728 if( bS[0] < 4 ) { | 5728 if( bS[0] < 4 ) { |
| 5729 int tc[4]; | 5729 int8_t tc[4]; |
| 5730 for(i=0; i<4; i++) | 5730 for(i=0; i<4; i++) |
| 5731 tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] + 1 : 0; | 5731 tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] + 1 : 0; |
| 5732 h->s.dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc); | 5732 h->s.dsp.h264_h_loop_filter_chroma(pix, stride, alpha, beta, tc); |
| 5733 } else { | 5733 } else { |
| 5734 /* 8px edge length, see filter_mb_edgev */ | 5734 h->s.dsp.h264_h_loop_filter_chroma_intra(pix, stride, alpha, beta); |
| 5735 for( d = 0; d < 8; d++ ){ | |
| 5736 const int p0 = pix[-1]; | |
| 5737 const int p1 = pix[-2]; | |
| 5738 const int q0 = pix[0]; | |
| 5739 const int q1 = pix[1]; | |
| 5740 | |
| 5741 if( ABS( p0 - q0 ) < alpha && | |
| 5742 ABS( p1 - p0 ) < beta && | |
| 5743 ABS( q1 - q0 ) < beta ) { | |
| 5744 | |
| 5745 pix[-1] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */ | |
| 5746 pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */ | |
| 5747 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); | |
| 5748 } | |
| 5749 pix += stride; | |
| 5750 } | |
| 5751 } | 5735 } |
| 5752 } | 5736 } |
| 5753 | 5737 |
| 5754 static void filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, int bS[8], int qp[2] ) { | 5738 static void filter_mb_mbaff_edgev( H264Context *h, uint8_t *pix, int stride, int bS[8], int qp[2] ) { |
| 5755 int i; | 5739 int i; |
| 5911 const int alpha = alpha_table[index_a]; | 5895 const int alpha = alpha_table[index_a]; |
| 5912 const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )]; | 5896 const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )]; |
| 5913 const int pix_next = stride; | 5897 const int pix_next = stride; |
| 5914 | 5898 |
| 5915 if( bS[0] < 4 ) { | 5899 if( bS[0] < 4 ) { |
| 5916 int tc[4]; | 5900 int8_t tc[4]; |
| 5917 for(i=0; i<4; i++) | 5901 for(i=0; i<4; i++) |
| 5918 tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] : -1; | 5902 tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] : -1; |
| 5919 h->s.dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc); | 5903 h->s.dsp.h264_v_loop_filter_luma(pix, stride, alpha, beta, tc); |
| 5920 } else { | 5904 } else { |
| 5921 /* 16px edge length, see filter_mb_edgev */ | 5905 /* 16px edge length, see filter_mb_edgev */ |
| 5968 static void filter_mb_edgech( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) { | 5952 static void filter_mb_edgech( H264Context *h, uint8_t *pix, int stride, int bS[4], int qp ) { |
| 5969 int i, d; | 5953 int i, d; |
| 5970 const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 ); | 5954 const int index_a = clip( qp + h->slice_alpha_c0_offset, 0, 51 ); |
| 5971 const int alpha = alpha_table[index_a]; | 5955 const int alpha = alpha_table[index_a]; |
| 5972 const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )]; | 5956 const int beta = beta_table[clip( qp + h->slice_beta_offset, 0, 51 )]; |
| 5973 const int pix_next = stride; | |
| 5974 | 5957 |
| 5975 if( bS[0] < 4 ) { | 5958 if( bS[0] < 4 ) { |
| 5976 int tc[4]; | 5959 int8_t tc[4]; |
| 5977 for(i=0; i<4; i++) | 5960 for(i=0; i<4; i++) |
| 5978 tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] + 1 : 0; | 5961 tc[i] = bS[i] ? tc0_table[index_a][bS[i] - 1] + 1 : 0; |
| 5979 h->s.dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc); | 5962 h->s.dsp.h264_v_loop_filter_chroma(pix, stride, alpha, beta, tc); |
| 5980 } else { | 5963 } else { |
| 5981 /* 8px edge length, see filter_mb_edgev */ | 5964 h->s.dsp.h264_v_loop_filter_chroma_intra(pix, stride, alpha, beta); |
| 5982 for( d = 0; d < 8; d++ ) { | |
| 5983 const int p0 = pix[-1*pix_next]; | |
| 5984 const int p1 = pix[-2*pix_next]; | |
| 5985 const int q0 = pix[0]; | |
| 5986 const int q1 = pix[1*pix_next]; | |
| 5987 | |
| 5988 if( ABS( p0 - q0 ) < alpha && | |
| 5989 ABS( p1 - p0 ) < beta && | |
| 5990 ABS( q1 - q0 ) < beta ) { | |
| 5991 | |
| 5992 pix[-pix_next] = ( 2*p1 + p0 + q1 + 2 ) >> 2; /* p0' */ | |
| 5993 pix[0] = ( 2*q1 + q0 + p1 + 2 ) >> 2; /* q0' */ | |
| 5994 tprintf("filter_mb_edgech 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], pix[-3*pix_next], p1, p0, q0, q1, pix[2*pix_next], pix[-2*pix_next], pix[-pix_next], pix[0], pix[pix_next]); | |
| 5995 } | |
| 5996 pix++; | |
| 5997 } | |
| 5998 } | 5965 } |
| 5999 } | 5966 } |
| 6000 | 5967 |
| 6001 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) { | 5968 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) { |
| 6002 MpegEncContext * const s = &h->s; | 5969 MpegEncContext * const s = &h->s; |