The truth about GCC and Athlons

[Kevin]

I'm finding nothing but confusing and outdated information about
optimizing CGG for Athlon. I have GCC 2.95.3 on my Mandrake 7.2 host
system, and 2.95.2 and 2.95.2.1 patched with pgcc-2.95.2.1 on the LFS I'm
building. I'm finding all kinds of conflicting information about what
works best for Athlons, and I know that in the end all I can do is
benchmark and find out for myself. So if anyone can recommend some good
benchmarking software, I'd be much obliged.

One thing I haven't found is any independent review of the various GCC
patches and optimizations. AthlonPGCC and AthlonGCC seem to be dead in
the water; PGCC seems to be fairly up to date, but not many people seem
to be using it, and they all disagree about what the optimal switches
are. The "other" PGCC, from Portland Group, claims to be the industry
performance leader, and it has an "athlon" -mcpu/-march type option built
in, but Portland Group seems only to be interested in selling
workstations pre-configured with their proprietary software. Bummer.

Can anyone shed some light on this subject?

Domo,
Krum