[MPlayer-dev-eng] Make ffmpeg MPEG1/2 decoder the default?

Ivan Kalvachev ikalvachev at gmail.com
Mon Feb 18 21:27:36 CET 2008


On Feb 18, 2008 7:51 PM, Reimar Döffinger
<Reimar.Doeffinger at stud.uni-karlsruhe.de> wrote:
> On Mon, Feb 18, 2008 at 07:47:54PM +0200, Uoti Urpala wrote:
> > On Mon, 2008-02-18 at 18:12 +0100, Reimar Döffinger wrote:
> > > ffmpeg12:
> > > BENCHMARKs: VC:   9.223s VO:   9.958s A:   0.000s Sys:   0.819s = 20.000s
> > > user    0m10.165s
> > > BENCHMARKs: VC:  13.447s VO:  10.156s A:   0.000s Sys:   1.168s = 24.771s
> > > user    0m14.749s
> > > BENCHMARKs: VC:  14.440s VO:  10.213s A:   0.000s Sys:   1.306s = 25.958s
> > > user    0m15.613s
> >
> > That looks very suspicious/unreliable as a benchmark - the fastest value
> > is a single outlier.
>
> Probably forgot to disable speedstep. Feel free to do your own
> benchmarks, I'm not going to repeat these.
> On the plus side, the Athlon 800 benchmarks can not have this problem...
Don't forget to put -quiet too, the statusbar update adds a lot of
quantum noise.

Here are few benchmarks on my Athlon 1GHz TBird:

<.vob -vc ffmpeg12 -vo null
BENCHMARKs: VC:  20.149s VO:   0.017s A:   0.000s Sys:   1.015s =   21.181s
BENCHMARKs: VC:  19.401s VO:   0.014s A:   0.000s Sys:   0.957s =   20.371s

<.vob -vc mpeg12 -vo null
BENCHMARKs: VC:  22.742s VO:   0.024s A:   0.000s Sys:   0.973s =   23.740s
BENCHMARKs: VC:  21.766s VO:   0.023s A:   0.000s Sys:   0.938s =   22.728s

<.vob -vc ffmpeg12 -vo xv
BENCHMARKs: VC:  26.850s VO:  12.718s A:   0.000s Sys:   1.148s =   40.715s
BENCHMARKs: VC:  23.201s VO:  12.645s A:   0.000s Sys:   1.067s =   36.914s
BENCHMARKs: VC:  23.468s VO:  12.687s A:   0.000s Sys:   1.032s =   37.187s

<.vob -vc mpeg12 -vo xv
BENCHMARKs: VC:  23.303s VO:  18.777s A:   0.000s Sys:   1.049s =   43.130s
BENCHMARKs: VC:  23.102s VO:  18.869s A:   0.000s Sys:   1.110s =   43.081s

===

<.mpg -vo null -vc ffmpeg12
BENCHMARKs: VC:   4.810s VO:   0.017s A:   0.000s Sys:   0.549s =    5.376s
BENCHMARKs: VC:   5.107s VO:   0.013s A:   0.000s Sys:   0.640s =    5.761s

<.mpg -vo null -vc mpeg12
BENCHMARKs: VC:   5.551s VO:   0.020s A:   0.000s Sys:   0.560s =    6.131s
BENCHMARKs: VC:   5.853s VO:   0.026s A:   0.000s Sys:   0.597s =    6.477s

<.mpg -vo xv -vc ffmpeg12
BENCHMARKs: VC:   5.660s VO:   3.816s A:   0.000s Sys:   0.733s =   10.209s
BENCHMARKs: VC:   5.958s VO:   3.842s A:   0.000s Sys:   0.794s =   10.594s

<.mpg -vo xv -vc mpeg12
BENCHMARKs: VC:   5.626s VO:   4.925s A:   0.000s Sys:   0.831s =   11.382s
BENCHMARKs: VC:   5.723s VO:   4.890s A:   0.000s Sys:   0.813s =   11.427s

I must say I'm surprised ffmpeg wins in all cases. Maybe because my
build is with fbranch-probabilities (I heard somebody saying ffh264
gets noticeable speedup with it).



More information about the MPlayer-dev-eng mailing list