diff options
| author | Jason Garrett-Glaser <darkshikari@gmail.com> | 2009-06-16 09:00:55 +0000 |
|---|---|---|
| committer | Jason Garrett-Glaser <darkshikari@gmail.com> | 2009-06-16 09:00:55 +0000 |
| commit | 4f717c69ed25a701f8b6613ca00e5e632a6382a6 (patch) | |
| tree | 0c82c716bd1f4f88d7645499692f3e213f4ffe68 /libavcodec/vc1dec.c | |
| parent | 41faa87886e6fc54f159da6940b9edbfcd194714 (diff) | |
| download | ffmpeg-4f717c69ed25a701f8b6613ca00e5e632a6382a6.tar.gz | |
idct_dc for VC-1/WMV3 decoder; ~11% faster decoding overall.
Includes mmx2 asm for the various functions.
Note that the actual idct still does not have an x86 SIMD implemtation.
For wmv3 files using regular idct, the decoder just falls back to simple_idct,
since simple_idct_dc doesn't exist (yet).
Originally committed as revision 19204 to svn://svn.ffmpeg.org/ffmpeg/trunk
Diffstat (limited to 'libavcodec/vc1dec.c')
| -rw-r--r-- | libavcodec/vc1dec.c | 23 |
1 files changed, 18 insertions, 5 deletions
diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c index 5d4dd0633b..6172e0c047 100644 --- a/libavcodec/vc1dec.c +++ b/libavcodec/vc1dec.c @@ -2028,8 +2028,12 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan block[idx] += (block[idx] < 0) ? -mquant : mquant; } if(!skip_block){ - s->dsp.vc1_inv_trans_8x8(block); - s->dsp.add_pixels_clamped(block, dst, linesize); + if(i==1) + s->dsp.vc1_inv_trans_8x8_dc(dst, linesize, block); + else{ + s->dsp.vc1_inv_trans_8x8(block); + s->dsp.add_pixels_clamped(block, dst, linesize); + } if(apply_filter && cbp_top & 0xC) s->dsp.vc1_v_loop_filter8(dst, linesize, v->pq); if(apply_filter && cbp_left & 0xA) @@ -2053,7 +2057,10 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan block[idx + off] += (block[idx + off] < 0) ? -mquant : mquant; } if(!(subblkpat & (1 << (3 - j))) && !skip_block){ - s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off); + if(i==1) + s->dsp.vc1_inv_trans_4x4_dc(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off); + else + s->dsp.vc1_inv_trans_4x4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, block + off); if(apply_filter && (j&2 ? pat & (1<<(j-2)) : (cbp_top & (1 << (j + 2))))) s->dsp.vc1_v_loop_filter4(dst + (j&1)*4 + (j&2)*2*linesize, linesize, v->pq); if(apply_filter && (j&1 ? pat & (1<<(j-1)) : (cbp_left & (1 << (j + 1))))) @@ -2078,7 +2085,10 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan block[idx] += (block[idx] < 0) ? -mquant : mquant; } if(!(subblkpat & (1 << (1 - j))) && !skip_block){ - s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off); + if(i==1) + s->dsp.vc1_inv_trans_8x4_dc(dst + j*4*linesize, linesize, block + off); + else + s->dsp.vc1_inv_trans_8x4(dst + j*4*linesize, linesize, block + off); if(apply_filter && j ? pat & 0x3 : (cbp_top & 0xC)) s->dsp.vc1_v_loop_filter8(dst + j*4*linesize, linesize, v->pq); if(apply_filter && cbp_left & (2 << j)) @@ -2103,7 +2113,10 @@ static int vc1_decode_p_block(VC1Context *v, DCTELEM block[64], int n, int mquan block[idx] += (block[idx] < 0) ? -mquant : mquant; } if(!(subblkpat & (1 << (1 - j))) && !skip_block){ - s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off); + if(i==1) + s->dsp.vc1_inv_trans_4x8_dc(dst + j*4, linesize, block + off); + else + s->dsp.vc1_inv_trans_4x8(dst + j*4, linesize, block + off); if(apply_filter && cbp_top & (2 << j)) s->dsp.vc1_v_loop_filter4(dst + j*4, linesize, v->pq); if(apply_filter && j ? pat & 0x5 : (cbp_left & 0xA)) |