>
> > Could I have a quick indication of how much faster quantize_xrpow() is over a
> > normal quantize(). I've got a few ideas for doing some more stuff to
> > quantize(), but if quantize_xrpow is now a better option I'll put my efforts
> > into that. (just had a quick peek and there's some asm in there. Mmmm.. :)
> > (for starters, the count_bits change in quantize could be put into
> > quantize_xrpow.
> >
>
> On a sample run on a pentium II 266 (compiled using gcc), quantize() takes
> 12 sec, while quantize_xrpow() takes 2.56 sec. I'll add some comments to
> the function to explain the assembly. The quick explanation is that this
> way the repeated setting of the FPU control word to specify the rounding mode
> is avoided, resulting in a huge performance gain for this function.
> --
Dont do any more work on quantize()!
I've actually already removed it from what will be the next release.
(I'm in the process of removing all the ISO code that is not used by
lame/gpsycho, which includes quantize() now that quantize_xrpow is
the default)
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
