This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] Fix fallout of PR37491 fix
On Tue, 23 Sep 2008, Ira Rosen wrote:
>
>
> gcc-patches-owner@gcc.gnu.org wrote on 22/09/2008 16:30:31:
>
> >
> > This fixes the fallout from the fix for PR37491. The vectorizer
> > disregards restrict qualification from the original data-ref pointer
> > if building the vectorized data pointer. The following patch fixes
> > that and makes sure to initialize and copy DECL_POINTER_ALIAS_SET
> > (otherwise the alias machinery will break down, as using alias sets
> > for restrict vs. non-restrict pointers isn't really ok).
> >
> > It also moves the alias sanity checks to where they actually belong,
> > so we can test the new reference vs. the original reference. But
> > this means the check is now an assert as we cannot safely abort
> > vectorization there.
> >
> > Bootstrapped on x86_64-unknown-linux-gnu, tested with -m64 and -m32.
> > Ira is testing ppc currently.
>
> Bootstrapped with vectorization enabled and tested on ppc-linux.>
Applied as r140781.
Richard.