This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH][GRAPHITE] Remove support for ISL 0.14
On 02/11/2017 08:24 AM, Richard Biener wrote:
> On February 11, 2017 12:38:32 AM GMT+01:00, Jakub Jelinek <jakub@redhat.com> wrote:
>> On Fri, Feb 10, 2017 at 04:34:30PM -0700, Jeff Law wrote:
>>>> 2017-02-10 Richard Biener <rguenther@suse.de>
>>>>
>>>> config/
>>>> * isl.m4: Remove support for ISL 0.14.
>>>>
>>>> * configure: Re-generate.
>>>>
>>>> gcc/
>>>> * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
>>>> Remove.
>>>> * configure: Re-generate.
>>>> * config.in: Likewise.
>>>> * graphite-dependences.c: Simplify as if
>>>> HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
>>>> * graphite-isl-ast-to-gimple.c: Likewise.
>>>> * graphite-optimize-isl.c: Likewise.
>>>> * graphite-poly.c: Likewise.
>>>> * graphite-sese-to-poly.c: Likewise.
>>>> * graphite.h: Likewise.
>>>> * toplev.c: Include isl/version.h and use isl_version () for
>>>> printing the ISL version.
>>>> * doc/install.texi: Update ISL requirement.
>>> My concern here would be that distributions may still be shipping
>> isl-0.14.
>>> Fedora 25 (for example) still uses iso-0.14.
>>
>> isl isn't a mandatory requirement for building gcc, worst case you just
>> don't have graphite support. Fedora 26 already has isl-0.16.1.
>
> Yeah. Note that even with ISL 0.15 there are known bugs. I just hope that w/o the legacy code maintainance would be easier.
And there are at least 2 bugs (PR79471 and PR69675) which ICE with ISL 0.16 and are fixed (well no transformation is done)
in 0.18. But maybe it just hide the ICE instead of not introducing PHI nodes usage before a definition.
Martin
>
> Richard.
>
>> Jakub
>