diff options
| author | Ronald S. Bultje <rsbultje@gmail.com> | 2011-02-19 19:07:29 -0500 |
|---|---|---|
| committer | Ronald S. Bultje <rsbultje@gmail.com> | 2011-02-20 08:31:14 -0500 |
| commit | 70aa916e4630bcec14439a2d703074b6d4c890a8 (patch) | |
| tree | 99274329bc0120c0baf1e6fcc234861c761e5567 | |
| parent | 87f1355f9b4fc11414d0e6a91404203c2745f89f (diff) | |
| download | ffmpeg-70aa916e4630bcec14439a2d703074b6d4c890a8.tar.gz | |
VC1: don't use vc1_put_block() in vc1_decode_i_blocks_adv().
Advanced profile never uses "range reduction", so vc1_put_block() quite
literally just calls put_pixels_clamped() from vc1_decode_i_blocks_adv().
By inlining the function, we can prevent calling IDCT8x8 if
CODEC_FLAG_GRAY is set, and we don't have to scale the coeffs in the
[0,256] range, but can instead use put_signed_pixels_clamped().
| -rw-r--r-- | libavcodec/vc1dec.c | 14 |
1 files changed, 11 insertions, 3 deletions
diff --git a/libavcodec/vc1dec.c b/libavcodec/vc1dec.c index 7eb9576c69..aff851487b 100644 --- a/libavcodec/vc1dec.c +++ b/libavcodec/vc1dec.c @@ -2704,7 +2704,7 @@ static void vc1_decode_i_blocks(VC1Context *v) */ static void vc1_decode_i_blocks_adv(VC1Context *v) { - int k, j; + int k; MpegEncContext *s = &v->s; int cbp, val; uint8_t *coded_val; @@ -2747,7 +2747,14 @@ static void vc1_decode_i_blocks_adv(VC1Context *v) s->mb_x = 0; ff_init_block_index(s); for(;s->mb_x < s->mb_width; s->mb_x++) { + uint8_t *dst[6]; ff_update_block_index(s); + dst[0] = s->dest[0]; + dst[1] = dst[0] + 8; + dst[2] = s->dest[0] + s->linesize * 8; + dst[3] = dst[2] + 8; + dst[4] = s->dest[1]; + dst[5] = s->dest[2]; s->dsp.clear_blocks(s->block[0]); mb_pos = s->mb_x + s->mb_y * s->mb_stride; s->current_picture.mb_type[mb_pos] = MB_TYPE_INTRA; @@ -2791,11 +2798,12 @@ static void vc1_decode_i_blocks_adv(VC1Context *v) vc1_decode_i_block_adv(v, s->block[k], k, val, (k<4)? v->codingset : v->codingset2, mquant); + if (k > 3 && (s->flags & CODEC_FLAG_GRAY)) continue; v->vc1dsp.vc1_inv_trans_8x8(s->block[k]); - for(j = 0; j < 64; j++) s->block[k][j] += 128; + s->dsp.put_signed_pixels_clamped(s->block[k], dst[k], + k & 4 ? s->uvlinesize : s->linesize); } - vc1_put_block(v, s->block); if(overlap) { if(s->mb_x) { v->vc1dsp.vc1_h_overlap(s->dest[0], s->linesize); |