[x264-devel] commit: Backport various speed tweak ideas from ffmpeg ( Jason Garrett-Glaser )
git version control
git at videolan.org
Mon Feb 15 10:20:25 CET 2010
x264 | branch: master | Jason Garrett-Glaser <darkshikari at gmail.com> | Fri Feb 12 21:15:12 2010 -0800| [41db4e8355e2390c430071c6f5ed9a2c8851e0a6] | committer: Jason Garrett-Glaser
Backport various speed tweak ideas from ffmpeg
Add mv0 early termination to spatial direct calculation
Up to twice as fast direct mv calculation on near-motionless video.
Branchless CAVLC level code adjustment based on trailing ones.
A few clocks faster.
Check tc value before clipping in C version of deblock functions.
Much faster, but nobody uses those anyways.
Thanks to Michael Niedermayer for the ideas.
> http://git.videolan.org/gitweb.cgi/x264.git/?a=commit;h=41db4e8355e2390c430071c6f5ed9a2c8851e0a6
---
common/frame.c | 6 ++++--
common/macroblock.c | 3 +++
encoder/cavlc.c | 7 +++----
3 files changed, 10 insertions(+), 6 deletions(-)
diff --git a/common/frame.c b/common/frame.c
index 40cc78f..d89f5ab 100644
--- a/common/frame.c
+++ b/common/frame.c
@@ -472,12 +472,14 @@ static inline void deblock_luma_c( uint8_t *pix, int xstride, int ystride, int a
int delta;
if( abs( p2 - p0 ) < beta )
{
- pix[-2*xstride] = p1 + x264_clip3( (( p2 + ((p0 + q0 + 1) >> 1)) >> 1) - p1, -tc0[i], tc0[i] );
+ if( tc0[i] )
+ pix[-2*xstride] = p1 + x264_clip3( (( p2 + ((p0 + q0 + 1) >> 1)) >> 1) - p1, -tc0[i], tc0[i] );
tc++;
}
if( abs( q2 - q0 ) < beta )
{
- pix[ 1*xstride] = q1 + x264_clip3( (( q2 + ((p0 + q0 + 1) >> 1)) >> 1) - q1, -tc0[i], tc0[i] );
+ if( tc0[i] )
+ pix[ 1*xstride] = q1 + x264_clip3( (( q2 + ((p0 + q0 + 1) >> 1)) >> 1) - q1, -tc0[i], tc0[i] );
tc++;
}
diff --git a/common/macroblock.c b/common/macroblock.c
index cc9a3fd..278659c 100644
--- a/common/macroblock.c
+++ b/common/macroblock.c
@@ -272,6 +272,9 @@ static int x264_mb_predict_mv_direct16x16_spatial( x264_t *h )
x264_macroblock_cache_mv_ptr( h, 0, 0, 4, 4, 0, mv[0] );
x264_macroblock_cache_mv_ptr( h, 0, 0, 4, 4, 1, mv[1] );
+ if( !M64( mv ) )
+ return 1;
+
if( h->param.i_threads > 1
&& ( mv[0][1] > h->mb.mv_max_spel[1]
|| mv[1][1] > h->mb.mv_max_spel[1] ) )
diff --git a/encoder/cavlc.c b/encoder/cavlc.c
index c65c9bd..85d2dde 100644
--- a/encoder/cavlc.c
+++ b/encoder/cavlc.c
@@ -147,10 +147,9 @@ static int block_residual_write_cavlc( x264_t *h, int i_ctxBlockCat, int16_t *l,
if( i_trailing < i_total )
{
- int16_t val = runlevel.level[i_trailing];
- int16_t val_original = runlevel.level[i_trailing]+LEVEL_TABLE_SIZE/2;
- if( i_trailing < 3 )
- val -= (val>>15)|1; /* as runlevel.level[i] can't be 1 for the first one if i_trailing < 3 */
+ int val = runlevel.level[i_trailing];
+ int val_original = runlevel.level[i_trailing]+LEVEL_TABLE_SIZE/2;
+ val -= ((val>>31)|1) & -(i_trailing < 3); /* as runlevel.level[i] can't be 1 for the first one if i_trailing < 3 */
val += LEVEL_TABLE_SIZE/2;
if( (unsigned)val_original < LEVEL_TABLE_SIZE )
More information about the x264-devel
mailing list