[Bug fortran/65819] New: overzealous checking in gfc_check_dependency for identical=true

[Bug fortran/65819] New: overzealous checking in gfc_check_dependency for identical=true

[tkoenig at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65819

            Bug ID: 65819
           Summary: overzealous checking in gfc_check_dependency for
                    identical=true
           Product: gcc
           Version: 6.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: tkoenig at gcc dot gnu.org

Just ran across this while trying to cook up a test case for matrix inlining.

In the patch proposed for PR 37131, I have

  if (gfc_check_dependency (expr1, matrix_a, true)
      || gfc_check_dependency (expr1, matrix_b, true))
    return 0;

In the test case, I want to do

  f(3,1:eight:2,:) = matmul(f(1,1:3,2:3), f(2,2:3,:))

where clearly the lhs does not depend on the rhs.  However, specifying true to
gfc_dep_dependency leads to

      if (identical)
    return 1;

before any checking is done.  I'm not sure if there is a real-world test
case yet, but there will be one soon, so I'll keep this PR as a reminder.

[Bug fortran/65819] overzealous checking in gfc_check_dependency for identical=true

[tkoenig at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65819

Thomas Koenig <tkoenig at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Blocks|                            |37131
           Severity|normal                      |enhancement

[Bug fortran/65819] overzealous checking in gfc_check_dependency for identical=true

[dominiq at lps dot ens.fr]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65819

Dominique d'Humieres <dominiq at lps dot ens.fr> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |WAITING
   Last reconfirmed|                            |2015-09-08
     Ever confirmed|0                           |1

--- Comment #1 from Dominique d'Humieres <dominiq at lps dot ens.fr> ---
Test case?

[Bug fortran/65819] overzealous checking in gfc_check_dependency for identical=true

[tkoenig at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65819

Thomas Koenig <tkoenig at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |NEW

--- Comment #2 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
The LHS and the arrays on the RHS cannot possibly overlap,
because of the difference in the first index.

ig25@linux-fd1f:~/Krempel/Dep-f> cat foo.f90
program main
  implicit none
  real, dimension(3,3,3) :: f
  integer :: three
  call random_number(f)
  three = 3
  f(1, 1:three, :) = matmul(f(2,1:3,2:3), f(3,2:3,:))
  print *,f
end program main
ig25@linux-fd1f:~/Krempel/Dep-f> gfortran -O -Warray-temporaries foo.f90
foo.f90:7:21:

   f(1, 1:three, :) = matmul(f(2,1:3,2:3), f(3,2:3,:))
                     1
Warning: Creating array temporary at (1) [-Warray-temporaries]

[Bug fortran/65819] overzealous checking in gfc_check_dependency for identical=true

[sandra at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=65819

sandra at gcc dot gnu.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |FIXED
                 CC|                            |sandra at gcc dot gnu.org
             Status|NEW                         |RESOLVED

--- Comment #8 from sandra at gcc dot gnu.org ---
This issue seems to have been fixed by the above patch.