[PATCH] rs6000: Fix incorrect fusion constraint [PR102991]

David Edelsohn dje.gcc@gmail.com
Thu Nov 4 01:59:08 GMT 2021

On Wed, Nov 3, 2021 at 9:46 PM Xionghu Luo <luoxhu@linux.ibm.com> wrote:
> On 2021/11/3 23:13, David Edelsohn wrote:
> > Did you manually change fusion.md or did you regenerate it after
> > fixing genfusion.pl?
> >
> > If you regenerated it, the ChangeLog entry should be "Regenerated" and
> > the "Fix incorrect clobber constraint." should refer to the
> > genfusion.pl change.
> >
> > I want to ensure that genfusion.pl generates the correct constraint
> > the next time it is used.
> >
> Aaron mentioned he disabled the auto generation here[1], but before
> than that, Segher suggested to enable it in stage1.
> [1] https://gcc.gnu.org/pipermail/gcc-patches/2021-February/564652.html
> [2] https://gcc.gnu.org/pipermail/gcc-patches/2021-January/564244.html
> Thus re-enable it with the followed v2 patch(Confirmed the fusion.md is
> exactly same with v1 patch.)
> [PATCH v2] rs6000: Fix incorrect fusion constraint [PR102991]
> gcc/ChangeLog:
>         * config/rs6000/fusion.md: Regenerate.
>         * config/rs6000/genfusion.pl: Fix incorrect clobber constraint.
>         * config/rs6000/t-rs6000: Uncomment regeneration of fusion.md.

I believe that there is some confusion about my request. I am not
requesting that the patch enable genfusion.pl . The Makefile fragment
rule to generate fusion.md is disabled for a reason and normally
should not be enabled.  But fusion.md should be generated by
genfusion.pl when there is a change, and any changes should be made in
genfusion.pl. In other words, change genfusion.pl, temporarily enable
the Makefile fragment rule, generate fusion.md, disable genfusion.pl.
My request was an effort to ensure that genfusion.pl correctly
regenerates the new, corrected fusion.md file.  I don't want a manual
change to fusion.md that differs from the automatically generated
file. Only the updated fusion.md and genfusion.pl should be checked

Has Aaron reviewed and confirmed the change to genfusion.pl?

Thanks, David

More information about the Gcc-patches mailing list