https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81205
Bug ID: 81205
Summary: Hybrid MPI and OpenMp: Blocking code in loops
Product: gcc
Version: 4.4.7
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: fortran
Assignee: unassigned at gcc dot gnu.org
Reporter: pasha.313 at hotmail dot com
Target Milestone: ---
I am developing a CFD code which was paralleled via MPI. I am adding OpenMp to
make it a hybrid parallel code.
I am using Gfortan (GCC version 4.4.7) right now.
The first thing that I noticed is the code blocks when a subroutine ,which has
a loop inside itself with index variable such as I,J for example, is called
from a loop which has the same index variable(counter I,J) of that subroutine.
When I disable OMP command of just those subroutine everything is fine and the
code works well.
Do you have any remedy for it?
I tried to use some update version such as GCC 4.7.2 and even 5.2.1, but the
condition exacerbated as the code block in early stage in comparison with
previous situation even if disabling those OMP commands.
I was wondering if you could kindly offer a solution to fix this problem.
Best regards
