Optimize split MC, so we don't always do 4x4 blocks of 4x4pixels each, but
authorRonald S. Bultje <rsbultje@gmail.com>
Mon, 28 Jun 2010 13:50:55 +0000 (13:50 +0000)
committerRonald S. Bultje <rsbultje@gmail.com>
Mon, 28 Jun 2010 13:50:55 +0000 (13:50 +0000)
commit7c4dcf81658103b9506adcbf848bd23efe3b0b4e
treef04ecb6acdd8e925caaebc856f50c42c3aeeabb3
parent4332bfbff865c41d27d09eaf73409eac544e1350
Optimize split MC, so we don't always do 4x4 blocks of 4x4pixels each, but
we apply them as 16x8/8x16/8x8 subblocks where possible. Since this allows
us to use width=8/16 instead of width=4 MC functions, we can now take more
advantage of SSE2/SSSE3 optimizations, leading to a total speedup for splitMV
filter of about 10%.

Originally committed as revision 23853 to svn://svn.ffmpeg.org/ffmpeg/trunk
libavcodec/vp8.c
libavcodec/vp8data.h