Hi Magnus!
Magnus Holmgren schrieb am Sam, 12 Aug 2000:
> Robert Hegemann wrote:
>
> > > Btw, after a test run, I noticed that 405 bytes differ (at least
> > > according to the simple cmp program I wrote :). Is that something to
> > > worry about?
> >
> > Well, for me using the gogo routine decreases the sound quality a lot.
> > It's true, it's fast, but I'm not an assembler freak so I don't know
> > what's wrong with it.
>
> I was a bit unclear on it, but the test run was with a plain compile (no
> assembler or hacks activated; just plain C). That worries me a little
> (after all, 400 is a lot more than 30)... :)
>
> (Btw, which optimization do you mean with "the gogo routine"?)
>
> --
> Magnus Holmgren
You wrote you used the fft_FPU_FXCH code, that is a gogo routine.
I think Takehiro brought it back into LAME's code base.
So you usually have 30 bytes different by the short test?
Hm, I don't know what makes the difference to 400 bytes now.
Ciao Robert
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
