Message Passing Interface (MPI) is a standardized and portable message-passing system designed for parallel computing architectures. MPI is commonly used to develop parallel applications in High-Performance Computing (HPC) environments.

The POLARIS cluster supports OpenMPI and uses Slurm for job scheduling.

1. Load the OpenMPI Module

Before compiling or running MPI programs, load the OpenMPI module:

module avail
module load mpi/openmpi-x86_64

Verify the module is loaded:

mpicc --version

2. Compile an MPI Program

Use mpicc (for C), mpicxx (for C++), or mpifort (for Fortran) to compile your MPI program.

Example (C program):

mpicc -o hello_mpi hello_mpi.c

3. Running MPI Jobs with Slurm (Recommended)

We recommend using srun within a Slurm job script rather than running mpirun directly on the login node.

Example Slurm Job Script (MPI)

#!/bin/bash
#SBATCH --job-name=mpi_job
#SBATCH --nodes=2
#SBATCH --ntasks-per-node=32
#SBATCH --time=02:00:00
#SBATCH --partition=cpucluster

module load mpi/openmpi-x86_64

srun ./hello_mpi

Submit the job using:

sbatch your_job_script.sh

4. Running MPI Interactively (Optional)

If you need an interactive session:

salloc --nodes=1 --ntasks=8 --time=01:00:00
module load mpi/openmpi-x86_64
srun ./hello_mpi

5. Checking Job Status

To monitor your job:

squeue -u your_username

To view completed jobs:

sacct -u your_username

To cancel a job:

scancel jobID

Additional Resources

• OpenMPI Documentation
• MPI Tutorial