Dan wrote:
> > >
> > > BTW, on my Celeron 400, the MSVC 6 compiled version with "/Ox" is
about 6%
> > > faster than the version I compiled with Cygwin B20 (egcs-2.91.57) and
> > >
"-O9 -fomit-frame-pointer -march=pentium -mcpu=pentium -ffast-math -funroll-
> > > loops".
>
> have you tried -march=pentiumpro? The celeron has the p6 core so
pentiumpro is
> more appropriate than pentium. Also, -march=<chiptype>
implies -mcpu=<chiptype>
> so you don't need to specify -mcpu
Thanks for the tip. With -march=pentiumpro, the speed gap is closed to
about 1%. BTW, I don't notice any difference in speed with the /G6 (PPro
optimization) flag to MSVC compared to the default "blended" optimization.
Also, there's one more bug in LAME 3.30 under MSVC. Line 5 of
mpglib/mpg123.h should be:
#ifndef _MSC_VER
not:
#ifndef WIN32
-Jake
--
MP3 ENCODER mailing list ( http://geek.rcc.se/mp3encoder/ )
