25% faster tham mmx on core2, 35% if you discount fullpel, 4% overall decoding.

Originally committed as revision 11871 to svn://svn.ffmpeg.org/ffmpeg/trunk