Add -m{arch,tune}=local for x86/x86-64

Richard Guenther richard.guenther@gmail.com
Wed Mar 15 09:41:00 GMT 2006


On 14 Mar 2006 21:30:29 -0800, Ian Lance Taylor <ian@airs.com> wrote:
> "H. J. Lu" <hjl@lucon.org> writes:
>
> > On Tue, Mar 14, 2006 at 08:23:42PM -0800, Ian Lance Taylor wrote:
> > > Michael Matz <matz@suse.de> writes:
> > >
> > > > I convinced myself somehow that gcc.c will not include the target but the
> > > > host header, which would avoid this problem.  On reflection that's not
> > > > true.  Hmm.  Any advice?  I could simply #ifdef CROSS_COMPILE the define
> > > > and the function definition.
> > >
> > > It seems to me that -march=local is only meaningful for a native
> > > compiler.  So it seems appropriate to check CROSS_COMPILE, which will
> > > be defined for a non-native compiler.
> > >
> >
> > Isn't -march=native better than -march=local?
>
> Yes, I agree.

No, I don't agree.  With "native" I can see no connection to the local
CPU (native CPU??)
in the local machine (native machine??).  -march=local fits it best -
see other compilers
for precedence (like gcc4ss ... *runs*).

Richard.



More information about the Gcc-patches mailing list