I have been starting to use LAMMPS for modeling however, when I try and use MPI it will just run it for as many times as I specify cores for MPI instead of dividing the task. I have also tried using OpenMP but this is just as slow if not slower as just using one core. Any insight would be greatly appreciated!
This is typically due to using one version of MPI to compile LAMMPS and second (different) version of MPI to run LAMMPS. The MPI version has to be consistent for both compile and run.
Thank you, how does one see what version of MPI was used to compile LAMMPS to ensure it is consistent?
A lot of information about your LAMMPS executable is output, if you run it with the ‘-h’ or ‘-help’ flag. The MPI version info is part of that.
Whether you get acceleration from OpenMP depends on whether you are using styles that are supported by the OPENMP packages, and that that package has been installed.
There are discussions in the LAMMPS manual about how to optimize for speed.