[MPlayer-dev-eng] Re: [PATCH] sync libfaad2 CVS 5-1-04 large

[D Richard Felker III]

On Mon, May 03, 2004 at 04:14:50AM +0000, adland wrote:
> Some status update
> 
> Built your patch as well
> 
> I have done some quick testing of the new CVS sync patch and it breaks
> playback of some samples which still played with your patch.
> (both failing samples were encoded using a nero encoder
> one audio only - playback corrupt, 
> other audio/video - audio playback silent)
> 
> the faac encoded audio only mp4 sample I have plays on both versions.
> 
> MPlayer/samples/MPEG-4/video.mp4 and video2.mp4 played back on both versions.
> 
> Question: Is this example valid if I wanted to benchmark them to compare
> performance
> 
> Machine info:
> gcc version 3.3.3 AMD 
> 
> vendor_id       : AuthenticAMD
> cpu family      : 6
> model           : 6
> model name      : AMD Athlon(tm)

Athlon has very fast fpu, so using fixed point won't help much, and
might even hurt. (Did you leave -DFIXED_POINT in the Makefile?) The
performance is bound entirely by faad's horrible thrashing of the
cache...

> stepping        : 2
> cpu MHz         : 1149.928
> cache size      : 256 KB
> 
> test command:
> ./mplayer -benchmark -vo null samples/ct_faac.mp4
> 
> CVS patch version
> BENCHMARKs: VC:   0.000s VO:   0.000s A:   0.401s Sys:  19.384s =   19.784s
> BENCHMARK%: VC:  0.0000% VO:  0.0000% A:  2.0254% Sys: 97.9746% = 100.0000%
> 
> your patch version
> BENCHMARKs: VC:   0.000s VO:   0.000s A:   0.456s Sys:  19.329s =   19.785s
> BENCHMARK%: VC:  0.0000% VO:  0.0000% A:  2.3068% Sys: 97.6932% = 100.0000%
> 
> they were very close with this example.
> thanks

Do you have any "SBR" AAC files? I think this is what the windows kids
call "HE-AAC" but I could be mistaken. This seems to be the type where
performance is abysmal.

Rich