Athlon Incompatible Packages

[Warren Togami]

> On Tue, 2004-02-24 at 07:00, Ivan Gyurdiev wrote:
>> > Seems to me that this is not a bug, but an intentional decision.
>> Either
>> > way, outside of the fact that any efficiencies you got from the
>> recompile
>> > would probably never make up the number of CPU cycles required for the
>> > build,
>>
>> Why do gentoo users make such a huge deal out of this?
> Do they have any _real_ benchmark telling the diffrence ?
>
> Note that comparing say FC1 directly with an optimized Gentoo
> is not a real benchmark, its probably not the same kernel with
> the same features. Diffrent version of lots-of-things.
>
> It's also interresting to find where the diffrence is.
> e.g. Gentoo have some kernel features enabled which is what
> makes the diffrence, you cannot say optimizing everything is
> what made Evolution a bit quicker.
> Or comparing a grep that's dog slow cause the user local is UTF-8
> vs a locale of C..
>
> Also beware of the placebo effect. Tell a user that this'n'that is much
> better/faster, and he thinks so.

I remember reading some comparative benchmarks of gcc-2.95 and gcc-3.2.x
somewhere that showed relative effects of compiling stuff for different
architectures.  IIRC, back during 2.95 compiling for athlon did make a
large difference when compared to i386, but the gap narrowed significantly
with gcc-3.2.x.  So Gentoo was perhaps effective in optimization back
then, but the reasons for doing so have quickly gone away.

Note that I'm relying completely on my memory and I know nothing about the
compiler.

Warren