Mercurial > libavcodec.hg
comparison snow.c @ 2223:b26474e72d6d libavcodec
use h264 MC code if possible
50% faster predict_plane() if mmx2/3dnow is available
0.1% bitrate increase
| author | michael |
|---|---|
| date | Sun, 12 Sep 2004 21:32:36 +0000 |
| parents | 3543987dccad |
| children | 11d54cb7ac4e |
comparison
equal
deleted
inserted
replaced
| 2222:ef568cc0972c | 2223:b26474e72d6d |
|---|---|
| 1986 tmp += stride; | 1986 tmp += stride; |
| 1987 } | 1987 } |
| 1988 STOP_TIMER("mc_block") | 1988 STOP_TIMER("mc_block") |
| 1989 } | 1989 } |
| 1990 | 1990 |
| 1991 #define mcb(dx,dy,b_w)\ | |
| 1992 static void mc_block ## dx ## dy(uint8_t *dst, uint8_t *src, int stride){\ | |
| 1993 uint8_t tmp[stride*(b_w+5)];\ | |
| 1994 mc_block(dst, src-2-2*stride, tmp, stride, b_w, b_w, dx, dy);\ | |
| 1995 } | |
| 1996 | |
| 1997 mcb( 0, 0,16) | |
| 1998 mcb( 4, 0,16) | |
| 1999 mcb( 8, 0,16) | |
| 2000 mcb(12, 0,16) | |
| 2001 mcb( 0, 4,16) | |
| 2002 mcb( 4, 4,16) | |
| 2003 mcb( 8, 4,16) | |
| 2004 mcb(12, 4,16) | |
| 2005 mcb( 0, 8,16) | |
| 2006 mcb( 4, 8,16) | |
| 2007 mcb( 8, 8,16) | |
| 2008 mcb(12, 8,16) | |
| 2009 mcb( 0,12,16) | |
| 2010 mcb( 4,12,16) | |
| 2011 mcb( 8,12,16) | |
| 2012 mcb(12,12,16) | |
| 2013 | |
| 2014 #define mca(dx,dy,b_w)\ | 1991 #define mca(dx,dy,b_w)\ |
| 2015 static void mc_block_hpel ## dx ## dy(uint8_t *dst, uint8_t *src, int stride, int h){\ | 1992 static void mc_block_hpel ## dx ## dy(uint8_t *dst, uint8_t *src, int stride, int h){\ |
| 2016 uint8_t tmp[stride*(b_w+5)];\ | 1993 uint8_t tmp[stride*(b_w+5)];\ |
| 2017 assert(h==b_w);\ | 1994 assert(h==b_w);\ |
| 2018 mc_block(dst, src-2-2*stride, tmp, stride, b_w, b_w, dx, dy);\ | 1995 mc_block(dst, src-2-2*stride, tmp, stride, b_w, b_w, dx, dy);\ |
| 2034 } | 2011 } |
| 2035 }else{ | 2012 }else{ |
| 2036 const int scale= plane_index ? s->mv_scale : 2*s->mv_scale; | 2013 const int scale= plane_index ? s->mv_scale : 2*s->mv_scale; |
| 2037 int mx= block->mx*scale; | 2014 int mx= block->mx*scale; |
| 2038 int my= block->my*scale; | 2015 int my= block->my*scale; |
| 2016 const int dx= mx&15; | |
| 2017 const int dy= my&15; | |
| 2039 sx += (mx>>4) - 2; | 2018 sx += (mx>>4) - 2; |
| 2040 sy += (my>>4) - 2; | 2019 sy += (my>>4) - 2; |
| 2041 src += sx + sy*stride; | 2020 src += sx + sy*stride; |
| 2042 if( (unsigned)sx >= w - b_w - 4 | 2021 if( (unsigned)sx >= w - b_w - 4 |
| 2043 || (unsigned)sy >= h - b_h - 4){ | 2022 || (unsigned)sy >= h - b_h - 4){ |
| 2044 ff_emulated_edge_mc(tmp + MB_SIZE, src, stride, b_w+5, b_h+5, sx, sy, w, h); | 2023 ff_emulated_edge_mc(tmp + MB_SIZE, src, stride, b_w+5, b_h+5, sx, sy, w, h); |
| 2045 src= tmp + MB_SIZE; | 2024 src= tmp + MB_SIZE; |
| 2046 } | 2025 } |
| 2047 mc_block(dst, src, tmp, stride, b_w, b_h, mx&15, my&15); | 2026 if((dx&3) || (dy&3) || b_w!=b_h || (b_w!=4 && b_w!=8 && b_w!=16)) |
| 2027 mc_block(dst, src, tmp, stride, b_w, b_h, dx, dy); | |
| 2028 else | |
| 2029 s->dsp.put_h264_qpel_pixels_tab[2-(b_w>>3)][dy+(dx>>2)](dst,src + 2 + 2*stride,stride); | |
| 2048 } | 2030 } |
| 2049 } | 2031 } |
| 2050 | 2032 |
| 2051 static always_inline int same_block(BlockNode *a, BlockNode *b){ | 2033 static always_inline int same_block(BlockNode *a, BlockNode *b){ |
| 2052 return !((a->mx - b->mx) | (a->my - b->my) | a->type | b->type); | 2034 return !((a->mx - b->mx) | (a->my - b->my) | a->type | b->type); |
| 2462 dsputil_init(&s->dsp, avctx); | 2444 dsputil_init(&s->dsp, avctx); |
| 2463 | 2445 |
| 2464 #define mcf(dx,dy)\ | 2446 #define mcf(dx,dy)\ |
| 2465 s->dsp.put_qpel_pixels_tab [0][dy+dx/4]=\ | 2447 s->dsp.put_qpel_pixels_tab [0][dy+dx/4]=\ |
| 2466 s->dsp.put_no_rnd_qpel_pixels_tab[0][dy+dx/4]=\ | 2448 s->dsp.put_no_rnd_qpel_pixels_tab[0][dy+dx/4]=\ |
| 2467 mc_block ## dx ## dy; | 2449 s->dsp.put_h264_qpel_pixels_tab[0][dy+dx/4]; |
| 2468 | 2450 |
| 2469 mcf( 0, 0) | 2451 mcf( 0, 0) |
| 2470 mcf( 4, 0) | 2452 mcf( 4, 0) |
| 2471 mcf( 8, 0) | 2453 mcf( 8, 0) |
| 2472 mcf(12, 0) | 2454 mcf(12, 0) |