Best Optimization Flags for Athlon XP

[grün]

I am trying to identify the most useful optimization flags in the GCC/MingW
environment for systems using the Athlon-XP processor.  I am compiling
versions of the Firefox browser, along with some other software,
specifically for this cpu and want to make it run as fast as possible.
Right now I am using the following optimization flags:

-march=athlon-xp
-mfpmath=387
-ftracer
-fforce-addr
-O2
-s
-pipe

These work very well, but the end products are not as fast as version built
from within MS VC++ ( .NET Studio Pro ).  Note I am not using SSE
instruction set as I have found that it actually SLOWS the software I
compile.

What other optimization flags, if any, do you recommend for Athlon-XPs?