On Fri, 28 Nov 2003 19:26:29 -0800, Ron Kuris wrote:
Mike A. Harris wrote:
>Athlon reorders instructions internally anyway, the net result
>generally being that it makes no difference what CPU instruction
>scheduling is chosen, as the chip does it's own thing internally.
>I'd be kind of surprised if compiling with all those options
>yields any real world performance gains anyway. Placebo effect
The gentoo folks would definitely have you believe otherwise. A friend
of mine swears that rebuilding everything using the -march=athlon-xp
results in a much more responsive X subsystem.
The funny thing about the "Gentoo folks" is, their perceived performance
improvement is not reproducible by all Gentoo folks. After wasting many
hours on recompiling stuff with modified compiler options and a largely
different set of build dependencies, not everyone finds the system to be
"much more responsive". Those, that do, should attempt at proving
performance increases with benchmark studies.