I have a fortran program which uses MPICH library. At some point of the program rank 0 calls an external program which itself uses OPENMP while other ranks wait in an MPI_BARRIER. My issue is that waiting processes stay at 100% CPU usage. I’d like to be able to use resources from waiting processes to speed-up my OPENMP program running on rank 0.

I have a fortran program which uses MPICH library. At some point of the program rank 0 calls an external program which itself uses OPENMP while other ranks wait in an MPI_BARRIER. My issue is that waiting processes stay at 100% CPU usage. I’d like to be able to use resources from waiting processes to speed-up my OPENMP program running on rank 0.

I have a fortran program which uses MPICH library. At some point of the program rank 0 calls an external program which itself uses OPENMP while other ranks wait in an MPI_BARRIER. My issue is that waiting processes stay at 100% CPU usage. I’d like to be able to use resources from waiting processes to speed-up my OPENMP program running on rank 0.