summaryrefslogtreecommitdiff
path: root/libavcodec/svq3.c
diff options
context:
space:
mode:
authorRonald S. Bultje <rsbultje@gmail.com>2013-02-18 21:03:02 -0800
committerMichael Niedermayer <michaelni@gmx.at>2013-02-19 16:25:50 +0100
commit1acd7d594c15aa491729c837ad3519d3469e620a (patch)
tree4a515aa31c7483d8a2fdfa1d4399c9a9d2b29cae /libavcodec/svq3.c
parenta1f1ca96b42698002920467c0aab9e636893088a (diff)
h264: integrate clear_blocks calls with IDCT.
The non-intra-pcm branch in hl_decode_mb (simple, 8bpp) goes from 700 to 672 cycles, and the complete loop of decode_mb_cabac and hl_decode_mb (in the decode_slice loop) goes from 1759 to 1733 cycles on the clip tested (cathedral), i.e. almost 30 cycles per mb faster. Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Diffstat (limited to 'libavcodec/svq3.c')
-rw-r--r--libavcodec/svq3.c4
1 files changed, 2 insertions, 2 deletions
diff --git a/libavcodec/svq3.c b/libavcodec/svq3.c
index 7817766e8c..8f910b588f 100644
--- a/libavcodec/svq3.c
+++ b/libavcodec/svq3.c
@@ -219,6 +219,8 @@ void ff_svq3_add_idct_c(uint8_t *dst, int16_t *block,
dst[i + stride * 2] = av_clip_uint8(dst[i + stride * 2] + ((z1 - z2) * qmul + rr >> 20));
dst[i + stride * 3] = av_clip_uint8(dst[i + stride * 3] + ((z0 - z3) * qmul + rr >> 20));
}
+
+ memset(block, 0, 16 * sizeof(int16_t));
}
static inline int svq3_decode_block(GetBitContext *gb, int16_t *block,
@@ -669,8 +671,6 @@ static int svq3_decode_mb(SVQ3Context *s, unsigned int mb_type)
}
if (!IS_SKIP(mb_type) || h->pict_type == AV_PICTURE_TYPE_B) {
memset(h->non_zero_count_cache + 8, 0, 14 * 8 * sizeof(uint8_t));
- h->dsp.clear_blocks(h->mb + 0);
- h->dsp.clear_blocks(h->mb + 384);
}
if (!IS_INTRA16x16(mb_type) &&