[CLUE-Tech] Athlon = i686 in kernel 2.4.7 Makefile

[Jed S. Baer]

On Sun, 24 Mar 2002 14:55:30 -0700
"Timothy C. Klein" <teece at silverklein.net> wrote:

> * Jed S. Baer (thag at frii.com) wrote:
> > > > I think this is the last strangeness from the RH7.2 upgrade.
> > > > Searching through the kernel mailing list archive, I see
> > > > speculation on why make would write -march=i686 instead of
> > > > march=athlon, 
> 
> Isn't the 2.96 that comes with Redhat the 3.0 prerelease?  If so, it
> would support athlon as an arch.  And I would definetly say use i686
> rather than K6 if you can't use Athlon.  However, I looked through
> kernel makefiles, and it *does* support arch Athlon if your compiler
> supports it, it would seem.  At least in 2.4.18.  It does a test to see
> of -march=athlon is accepted, and if not, it uses -march=686.

As I mentioned before, the test you mention isn't in my makefile. This is
the 2.4.7-10 kernel. It explicitly sets march=i686 when CONFIG-MK7 is
defined. The test you mention must be a later kernel release.

> I usually compile with gcc 2.95 on Debian Sid, which is the default.
> However, if I edit the top-level kernel Makefile to use gcc-3.0, I have
> support for the -march=athlon flag, and the kernel Makefiles are using
> it automatically.  The kernel folks say that gcc 3.0 is dangerous,
> though.  It is currently compiling now, I guess I will find out if
> compiles all the way to the end.

I'll be interested to here how it comes out. I was hoping there was some
"known" article on why RH (or whomever) didn't use the athlon capability
in a distribution where the c compiler supports it, as, AFAICT, 2.96 does.
Or maybe RH didn't tweak the makefile, even though they included an
athlon-capable compiler.

jed
-- 
"Those who expect to reap the blessings of freedom must, like men,
 undergo the fatigue of supporting it."
 - Thomas Paine