Misleading error message about profiling region in mixed MPI-OpenMP mode
If there is a mismatch between the number of core available (e.g. requested by the job) and what is required, i.e.:
ncores = 10* nsplit_i * nsplit_j * omp_level_size
One then gets error message of the likes of:
38: Impossible to reset profiling : a profiling region is still open 38: Thread, depth 0 0 38: Thread, depth 1 3 38: level, name 1 timeloop 38: level, name 2 MPI 38: level, name 3 MPI_omp_barrier 38: Abort ! 38: Impossible to reset profiling : a profiling region is still open 38: Abort(-1) on node 38 (rank 38 in comm 0): application called MPI_Abort(MPI_COMM_WORLD, -1) - process 38
Quite misleading...
Ideally we should add some extra checking at the beginning of the run that the number of cores for the job is adequate.