VP5/6/8: ~7% faster arithmetic decoding
authorJason Garrett-Glaser <darkshikari@gmail.com>
Thu, 12 Aug 2010 01:11:32 +0000 (01:11 +0000)
committerJason Garrett-Glaser <darkshikari@gmail.com>
Thu, 12 Aug 2010 01:11:32 +0000 (01:11 +0000)
commit05c04cdf54fee9332d337380fd4cd8502bdac2be
treea34738cea5f0b5286f5f31c996c8f634361cc31c
parentd2064fd42b2dd1cf1c44e5c4fc4b8aaba6698637
VP5/6/8: ~7% faster arithmetic decoding
Grab from the bitstream in 16-bit chunks instead of 8-bit chunks.
TODO: grab in 32-bit chunks on 64-bit systems.

Originally committed as revision 24783 to svn://svn.ffmpeg.org/ffmpeg/trunk
libavcodec/vp56.h
libavcodec/vp56rac.c
libavcodec/x86/vp56_arith.h