[Bug fortran/68829] New: [4.7/4.8/4.9/5.3/6.0 Regression] Segfaults with -Ofast

[Joost.VandeVondele at mat dot ethz.ch]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68829

            Bug ID: 68829
           Summary: [4.7/4.8/4.9/5.3/6.0 Regression] Segfaults with -Ofast
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: Joost.VandeVondele at mat dot ethz.ch
  Target Milestone: ---

Currently, switching from -O3 to -Ofast might result in segfaulting Fortran
programs:

> cat test.f90
MODULE foo
CONTAINS
  SUBROUTINE mysum(a)
    INTEGER :: a(:)
    WRITE(6,*) SUM(a)
  END SUBROUTINE
END MODULE foo

USE foo
INTEGER, ALLOCATABLE :: a(:)
INTEGER, PARAMETER :: N=2**26 ! 256Mb array
ALLOCATE(a(N)) ; a=1
CALL mysum(a*a)
END

> gfortran -O3 test.f90 && ./a.out
    67108864

> gfortran -Ofast test.f90 && ./a.out
Segmentation fault

The reason for this is that -Ofast enables -fstack-arrays. This options puts
temporary arrays off unlimited size on the stack. While there is precedent for
this (ifort), I think this is a poor choice. I think -fstack-arrays should only
do this for sufficiently small arrays.

This is a regression from gcc 4.6 (presumably because -Ofast didn't imply
-fstack-arrays).