[Bug fortran/122910] New: [OpenMP] 'firstprivate' on 'target' mishandled – should privatize pointer not (also) the pointee

[burnus at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122910

            Bug ID: 122910
           Summary: [OpenMP] 'firstprivate' on 'target' mishandled –
                    should privatize pointer not (also) the pointee
           Product: gcc
           Version: 16.0
            Status: UNCONFIRMED
          Keywords: openmp, wrong-code
          Severity: normal
          Priority: P3
         Component: fortran
          Assignee: unassigned at gcc dot gnu.org
          Reporter: burnus at gcc dot gnu.org
  Target Milestone: ---

Testcase:

https://github.com/OpenMP-Validation-and-Verification/OpenMP_VV/blob/master/tests/5.1/target/test_target_is_accessible.F90

For
    ! Assume the device can access host memory
    ! If you don't have one, try OMP_OFFLOAD_TARGET=disabled (hostfallback)
    INTEGER, POINTER :: fptr(:)
    integer, parameter :: N = 5
    ALLOCATE(fptr(N))

    fptr = 0
    !$omp target firstprivate(fptr)
    DO i=1, N
      fptr(i) = 5*i
    END DO
    !$omp end target
    print *, fptr
    end

GCC privatizes the pointee, i.e. it prints 0 instead of 5*i.


OpenMP - here TR14 - requires (at least meanwhile):

"7.3.4 firstprivate Clause
"If an original list item has the POINTER attribute, the new list
 items receive the same association status as the original list item,
 as if by pointer assignment."