[Bug middle-end/64017] New: Support ISL 0.14.0 (to fix ICE with gfortran.dg/graphite/pr42393.f90)

[Bug middle-end/64017] New: Support ISL 0.14.0 (to fix ICE with gfortran.dg/graphite/pr42393.f90)

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64017

            Bug ID: 64017
           Summary: Support ISL 0.14.0 (to fix ICE with
                    gfortran.dg/graphite/pr42393.f90)
           Product: gcc
           Version: 5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: middle-end
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
                CC: grosser at gcc dot gnu.org, romangareev at gcc dot gnu.org
        Depends on: 62289

GCC 5 should (optionally) support ISL 0.14.0; currently, only 0.12.2 is
supported. For compatibility with GCC 4.8/4.9, supporting ISL 0.12.x would be
nice.

ISL 0.14.0 is required to fix the ICE with gfortran.dg/graphite/pr42393.f90
(see PR62289).

Cf. https://gcc.gnu.org/ml/gcc-patches/2014-11/msg00906.html


BTW: Currently, there is no version check for ISL in GCC's configure script.
BTW2: There are in gcc/graphite* still references to CLooG in the comments,
which are obsolete.

[Bug middle-end/64017] Support ISL 0.14.0 (to fix ICE with gfortran.dg/graphite/pr42393.f90)

[howarth at bromo dot med.uc.edu]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64017

howarth at bromo dot med.uc.edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |howarth at bromo dot med.uc.edu

--- Comment #1 from howarth at bromo dot med.uc.edu ---
It appears that Polly in llvm is only coded for isl libraries containing the
newer isl_schedule_constraints_compute_schedule() call without fallback support
for the older isl_union_set_compute_schedule() call. If we want to support isl
0.12.2 as well as 0.14, we could add a test for the presence of
isl_union_set_compute_schedule() to config/isl,m4 and use that to set define in
auto-host.h that could be leveraged in gcc/graphite-optimize-isl.c to determine
which code path to use.

[Bug middle-end/64017] Support ISL 0.14.0 (to fix ICE with gfortran.dg/graphite/pr42393.f90)

[rguenth at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64017

--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
That sounds like sth sensible.  Note that I'd put that test into gcc/configure
as there you can perform link tests even to an in-tree compiled ISL.

[Bug middle-end/64017] Support ISL 0.14.0 (to fix ICE with gfortran.dg/graphite/pr42393.f90)

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64017

Tobias Burnus <burnus at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|---                         |FIXED

--- Comment #4 from Tobias Burnus <burnus at gcc dot gnu.org> ---
FIXED

[Bug middle-end/64017] Support ISL 0.14.0 (to fix ICE with gfortran.dg/graphite/pr42393.f90)

[burnus at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64017

--- Comment #3 from Tobias Burnus <burnus at gcc dot gnu.org> ---
Author: burnus
Date: Mon Dec  1 19:07:37 2014
New Revision: 218247

URL: https://gcc.gnu.org/viewcvs?rev=218247&root=gcc&view=rev
Log:
2014-12-01  Tobias Burnus  <burnus@net-b.de>
            Jack Howarth  <howarth@bromo.med.uc.edu>

        PR middle-end/64017
        * configure.ac
        * (ac_has_isl_schedule_constraints_compute_schedule):
        New check.
        * doc/install.texi (ISL): Permit ISL 0.14.
        * graphite-optimize-isl.c (getScheduleForBandList,
        * optimize_isl):
        Conditionally use ISL 0.13+ functions.
        * graphite-interchange.c: Make 'extern "C"' conditional.
        * graphite-isl-ast-to-gimple.c: Ditto.
        * graphite-poly.c: Ditto.
        * graphite-sese-to-poly.c: Ditto.
        * config.in: Regenerate.
        * gcc/configure: Regenerate.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/config.in
    trunk/gcc/configure
    trunk/gcc/configure.ac
    trunk/gcc/doc/install.texi
    trunk/gcc/graphite-interchange.c
    trunk/gcc/graphite-isl-ast-to-gimple.c
    trunk/gcc/graphite-optimize-isl.c
    trunk/gcc/graphite-poly.c
    trunk/gcc/graphite-sese-to-poly.c

[Bug middle-end/64017] Support ISL 0.14.0 (to fix ICE with gfortran.dg/graphite/pr42393.f90)

[howarth at bromo dot med.uc.edu]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64017

--- Comment #5 from howarth at bromo dot med.uc.edu ---
Do we still need to be passing -DCLOOG_INT_GMP from config/isl.m4 now that the
cloog support is fully deprecated in gcc 5.0?

[Bug middle-end/64017] Support ISL 0.14.0 (to fix ICE with gfortran.dg/graphite/pr42393.f90)

[howarth at bromo dot med.uc.edu]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64017

--- Comment #6 from howarth at bromo dot med.uc.edu ---
AFAIKS, the define of CLOOG_INT_GMP is only used in the cloog headers...

https://groups.google.com/forum/#!msg/cloog-development/IFqcHdgGpM4/UOAXxd00MXIJ

so it 

islinc="-DCLOOG_INT_GMP ${islinc}"

should be removed now from config/isl.m4 and the top-level configure
regenerated.