Software: MPI Troubleshooting - nthu-ioa/cluster GitHub Wiki

:warning: This page is still being written. Please get in touch with the admins if you have any questions.

The answers to these questions are helpful for understanding problems related to running parallel codes using MPI or MPI+OpenMP.

  • Have you read about and understood the difference between multithreading and MPI?

  • Are you using mpi4py? Did you build it yourself or install it in your conda environment?

  • Have you tried setting #SBATCH --export=NONE together with srun --export=ALL (see here)?