[Mplayer-users] ffdivx benchmark

[Arpi]

Hi,

> 023bbe58aba3b32c7b5c45549a34b5a8  sample.light.it.up.avi
> VIDEO:  [DIV3]  640x352  24bpp  23.98 fps  823.5 kbps (100.5 kbyte/s)
> divxds:
> BENCHMARKs: V:  15.393s VO:   0.006s A:   0.000s Sys:   0.274s =   15.674s
> BENCHMARK%: V: 98.2086% VO:  0.0411% A:  0.0000% Sys:  1.7503% = 100.0000%
> divx:
> BENCHMARKs: V:  14.819s VO:   0.008s A:   0.000s Sys:   0.269s =   15.095s
> BENCHMARK%: V: 98.1672% VO:  0.0498% A:  0.0000% Sys:  1.7831% = 100.0000%
> ffdivx:
> BENCHMARKs: V:  15.202s VO:   0.012s A:   0.000s Sys:   0.300s =   15.514s
> BENCHMARK%: V: 97.9917% VO:  0.0749% A:  0.0000% Sys:  1.9334% = 100.0000%

I've recompiled libavcodec (ffmpeg stuff) with Intel's compiler, and
got better results:

BENCHMARKs: V:  14.446s VO:   0.009s A:   0.000s Sys:   0.293s =   14.749s
BENCHMARK%: V: 97.9504% VO:  0.0639% A:  0.0000% Sys:  1.9857% = 100.0000%

It detected some code parts and vectorized those, so this way it used
a little MMX. I think it can be much better optimized by hand.

(I'm waiting for answer from ffmpeg author about code syncronization and
CVS access)


A'rpi / Astral & ESP-team

--
mailto:arpi at thot.banki.hu
http://esp-team.scene.hu

_______________________________________________
Mplayer-users mailing list
Mplayer-users at lists.sourceforge.net
http://lists.sourceforge.net/lists/listinfo/mplayer-users