> Ben Reser <[EMAIL PROTECTED]> writes:
>
> > > surely this should be something that is done after installation.
If you are
> > > going to say 'so people who have i686 machines will have
faster/better
> > > machines' then it should also optimise for athlons as well!!!
yes but what is important is the instruction set of the CPU not the
manufacturer. Athlons (If I'm correct) are basically i686 so it doesn't
pay to waste time doing the same thing twice. (Like 300 windows drivers
for all the tulip based nic cards vs one on Unix.) However i586 vs i686
is a rather large difference. Enough that it not only can make a
performance difference it can make a difference in the ability to boot.
(btw RH's i386 compiled kernels won't run on a real i386 ... only i486
go figure.) So there is a difference there. Paging, Memeory handling
(DDR vs PC1xx vs PC66) and more. The rules that apply to applications
(Abi-Word, Evolution etc.) Don't always apply to system level programs.
and really don't apply to kernels.
James
> >
> > I'm sure if someone submitted a patch to do so it'd be accepted in this
> > particular case. There are only a couple libraries in the i686
> > directory (the i586 ones are just in /lib). They are libraries that
>
> Nope, the problem is more of the performance you gain with that,
> compared to the diskspace it uses. And gentoo or mandrake users
> compiling all for athlon target seem to report that it doesn't
> make a significant difference.
>
> --
> Guillaume Cottenceau - http://people.mandrakesoft.com/~gc/
>
>
James S. Sparenberg
Director New Product Development
Open Country Inc.
(408) 248-8530 office
(408) 248-8506
