I’m trying to accelerate the OpenMC depletion calculation on University’s high-performance computing cluster BlueBEAR. The staff from BlueBEAR group suggest using OpenMPI.
Hi @YimingZ. Depletion works by loading the OpenMC shared library directly from the Python process, so unlike openmc.run, which creates subprocesses (and can call mpiexec), you would need to call mpiexec on the Python script itself. Also note that with depletion there is no need to call openmc.run – when you have a depletion integrator, it handles calling the transport solver within OpenMC.
To work properly, you need to have the mpi4py package installed, and it needs to be built against the same MPI implementation (OpenMPI, in your case) that was used to build OpenMC.