"Replace implicit use of fast_memcpy via macro by explicit use to allow
for future optimization."
This is not yet done for ffmpeg when compiled within MPlayer.

Originally committed as revision 9225 to svn://svn.ffmpeg.org/ffmpeg/trunk