Hi all:
I’d like to run openmc.depletion on a multi-node computer to speed up calculations. I’m attempting to use 4 nodes with a total of 512 cores, but it seems that the computation speed isn’t significantly faster than running on a single node (only about 20% faster). I have limited knowledge of OPENMP and MPI. Can anyone help me identify if there’s an issue with my configuration?
Hello, normally to take advantage of parallel computing performance you must ensure that the number of particles is high, i.e. there is a balance between computation time and communication time between nodes. I recommend you to have a look at some speedup concepts. In the past I did a project on parallel computing performance, in the link I give you you can see that as you increase the number of particles the performance increases with respect to serial computing. MPI efficiency.
Therefore, try lowering the number of processors and look at the performance with respect to the serial computation by setting a number of particles. Or increase the number of particles a lot and compare with the serial computation.
1 Like
Thank you for your input! I’ll review your suggestions. Much appreciated!
1 Like