-
Reimar Döffinger authored
This allows for some simplifications like removing some outer loops and gives much better performance for thread_count > number of idle CPUs. Originally committed as revision 20211 to svn://svn.ffmpeg.org/ffmpeg/trunk
2a1294b9