Re: libmoto PowerPC-optimized math library?

[Michel Dänzer <daenzerm@student.ethz.ch>]

Michel Dänzer wrote:

> > > > Of note is the fact that there is very little difference between
> > > > the benchmark results with no extra math library and with the
> > > > libmoto library except for the Fourier transform test, where
> > > > there's a pretty dramatic difference (improvement of about 3X with
> > > > libmoto over no special library).
> > >
> > > This rings a bell here. "MP3 encoding" :)
> >
> > My bell says images and video (and also mp3, sure =). Although I don't
> > know which algorithm is used in most of these formats. Fourier or Huffman.
> 
> MP3 encoding certainly uses Fourier, the question is how important it is...
> 
> I've tried building lame 3.83 with libmoto, and it seems to make no real
> difference :(

I've played a bit more with 3.86, removed a few optimization options which I
consider mostly important on i386 and -ffast-math (may this prevent libmoto
functions from being called?), and now the play/CPU ratio is higher than 1 for
VBR encoding with 128 bit/s minimum :)

I don't mention that on my Athlon with gogo (3Dnow), the ratio is around 8 :-/


Michel


-- 
Earthling Michel Dänzer (MrCooper)  \  CS student and free software enthusiast
Debian GNU/Linux (powerpc,i386) user \   member of XFree86 and The DRI Project