>
> Before the current Xing engine, there was a thing (awfull) called tompg.exe
> from xing for mpeg audio encoding. I can send it to you if you're
> interested, but it's a win32 file.
> The interesting thing is that there is an option in it to specify to pass or
> not in the psy model, or to let the encoder choose.
> The other thing I'm quite sure is that they don't choose what huffmann table
> to use, they only use a fixed one.
>
>
Do you understand the huffmann encoding? I dont know anything about
that part of the code. Would using a fixed table give a significant
speedup? We could try and add it to the -f option. Another related
thing is to just use all scalefactors=0 for -f. This would mean just
one pass through outer_loop, makeing the code much faster. With no
psy-model, there is really no good way to tell which combination of
scalefactors is better than any other combination.
Mark
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
