[FFmpeg-cvslog] r12286 - trunk/libavcodec/ac3dec.c

[Justin Ruggles]

Ivan Kalvachev wrote:
> You know, multiply is usually faster than division, so if these sums
> are constant, you can probably calculate reciprocals and multiply in
> the loop.
> {sr0=1/s0;...; v0*=sr0};

Thanks. It definitely helps. :)

Updated 5.1-to-stereo benchmark:
liba52  1.756
ffac3   2.420

-Justin