Re: [PATCH] vect-pragma-target-2.C: Compile with -march=x86-64

On Tue, Aug 18, 2026 at 11:33 AM Liu, Hongtao <[email protected]> wrote:
>
>
>
> > -----Original Message-----
> > From: Liu, Hongtao
> > Sent: Tuesday, August 18, 2026 9:43 AM
> > To: H.J. Lu <[email protected]>; GCC Patches <[email protected]>;
> > Uros Bizjak <[email protected]>
> > Subject: RE: [PATCH] vect-pragma-target-2.C: Compile with -march=x86-64
> >
> > > -----Original Message-----
> > > From: H.J. Lu <[email protected]>
> > > Sent: Tuesday, August 18, 2026 9:12 AM
> > > To: GCC Patches <[email protected]>; Uros Bizjak
> > > <[email protected]>; Liu, Hongtao <[email protected]>
> > > Subject: [PATCH] vect-pragma-target-2.C: Compile with -march=x86-64
> > >
> > > Compile vect-pragma-target-2.C with -march=x86-64 so that x86-64-v2
> > > and above features aren't enabled on x86-64-v4 machines when GCC is
> > > configured with
> > >
> > > --with-arch=native --with-cpu=native
> >
> > There are quite a few test cases that fail under --with-arch=native. As I
> > understand it, the primary intent of using --with-arch=native is to serve as an
> > indicator for detecting genuine regressions, rather than to actively investigate
> > and fix those individual failures.
>
> Just to clarify, I am not opposed to the patch itself,  but rather discussing the strategy around regression testing with --with-arch=native.

Configure GCC with --with-arch=native --with-cpu=native provides additional
coverage.  In the case of this patch series:

https://patchwork.sourceware.org/project/gcc/list/?series=65022

the original patch set passed the GCC regression test, but failed to bootstrap
when configured with --with-arch=native --with-cpu=native

-- 
H.J.