Re: Building for K6-2 CPU: i586 or i686?

[Jean Francois Martinez <jfm2 at club-internet dot fr>]

Michael Meissner a écrit :

> On Tue, Jun 19, 2001 at 04:23:21PM -0500, steve.snyder@philips.com wrote:
> > Hello.
> >
> > Which compiler option generates better code for the AMD K6-2 CPU,
> > -march=i586 or -march=i686?
>
> IIRC, you don't want -march=i686, since the K6 doesn't have the conditional
> move instructions that the i686 processors do.
>
> > Please don't rub salt into my wounds by suggesting "-march=k6" as the way
> > to go.  I want the code to actually *run* after it's compiled.  (See
> > PR2676 in GNATS ("Illegal loop instruction when using -O2 -march=k6" as
> > well as post http://gcc.gnu.org/ml/gcc/2001-06/msg01058.html.)
> >

-mcpu=i686 beats (or beated in gcc 2.x) -march=i586 hands down according to
the benchmarks I ran.
 It does not generate
non 386 instructions but on the other hand it schedules for a processor (P6
family) whose
architecture is closer to K6s than Pentium's architecture


                                                    JFM



>
> > Thank you.
>
> --
> Michael Meissner, Red Hat, Inc.  (GCC group)
> PMB 198, 174 Littleton Road #3, Westford, Massachusetts 01886, USA
> Work:     meissner@redhat.com           phone: +1 978-486-9304
> Non-work: meissner@spectacle-pond.org   fax:   +1 978-692-4482

--
                        JF Martinez

Independence Linux (http://independence.seul.org)

Because Linux should be for everyone.