OpenMP catastrophic error Internal compiler error

[pat97269]

Hi all,here i report the catastrophic error i've got with a simple code :[fortran]module tt contains function test(p0) integer,dimension(:),allocatable :: p0 integer :: test,np[/fortran][fortran] np=10 !$omp parallel do shared(np) default(private) do i=1,np print *,'i',p0 end do !$omp end parallel do test=1 end function test end module tt program mp use tt implicit none integer :: i,np np=10 print *,test([1]) end program[/fortran] As soon as the openmp deal with an allocatable array which is an argument of the function , i get internal error.Thanks

[mecej4]

The variable np is undefined in the module.

[pat97269]

And so ? Did you compile the code ?

a undefined variable never stop the compiler and throw an catastrophic error, it's an execution time problem.

I fix the undefined variable and still have the problem, the compiler stop without compiling the code. (I have test that with v12.1.0 and the last v12.1.2)

[Udit_P_Intel]

I can reproduce this reported problem with just the following code. Note that I have removed NP here:

subroutine test(p0)
integer,dimension(:),allocatable :: p0

!$omp parallel do default(private)
do i=1,2
print *,p0
end do
!$omp end parallel do

end

-bash-3.2$ ifort -c a1234.f90 -openmp
a1234.f90: catastrophic error: **Internal compiler error: segmentation violation signal raised** Please report this error along with the circumstances in which it occurred in a Software Problem Report. Note: File and line given may not beexplicit cause of this error.

-----------
I noticed that if we remove DEFAULT(PRIVATE) then there is no problem. Maybe the other OpenMP experts can comment more on this issue. Or if there is a bug, then the support team can start a tracker.

Best!
-Udit

[TimP]

On the basis that any internal error is a bug, even if provoked by invalid code, this does appear to demonstrate a compiler bug. The compiler ought to point to invalid source code before quitting. It may be a tough question of policy, as the compiler does little to diagnose OpenMP related problems unless you engage -diag-enable openmp and have Inspector installed. Unless I have missed something, the compile with that option engaged dies without results.

[Anonymous66]

Hi,

I have escalated this issue to the developers. The issue number is DPD200177818. I will post any updated I receive on this issue to this thread.

Regards,
Annalee

[pat97269]

The same issue appears when you try to work with output variable of a function. So it is both in and out 1.

Thanks

[Anonymous66]

A fix has been found for this issue. We are currently planning to include it in the next major release which isscheduled for later this year.

[Anonymous66]

This is issue has been fixed in Intel® Fortran Composer XE for Linux* 2013 which is now available at the Intel® Registration Center. Regards, Annalee Intel Developer Support * Other names and brands may be claimed as the property of others.