Gromacs Job Scripts

From UFRC
Jump to navigation Jump to search
Note
Use 'module spider gromacs' to find available gromacs versions. The module loads below may be outdated.

back to the Gromacs page

Parallel (MPI)

#!/bin/bash
#SBATCH --job-name=gromacs
#SBATCH --mail-user=YOUR_MAIL_ADDRESS_HERE
#SBATCH --mail-type=FAIL,BEGIN,END
#SBATCH --output=gmx-%j.out
#SBATCH --ntasks=2
#SBATCH --cpus-per-task=4
#SBATCH --ntasks-per-socket=1
#SBATCH --distribution=cyclic:block
#SBATCH --time=24:00:00
#SBATCH --mem-per-cpu=1gb

module purge
ml gcc/8.2.0 openmpi/4.0.1 gromacs/2019.2

export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
srun --mpi=pmix_v3 gmx mdrun -ntomp ${SLURM_CPUS_PER_TASK} -s topol.tpr

GPU Acceleration (MPI + multiple GPUs)

#!/bin/bash
#SBATCH --job-name=multi-gpu
#SBATCH --mail-user=YOUR_MAIL_ADDRESS_HERE
#SBATCH --mail-type=FAIL,BEGIN,END
#SBATCH --output=gromacs_%j.log
#SBATCH --nodes=1
#SBATCH --ntasks=2
#SBATCH --tasks-per-node=2
#SBATCH --cpus-per-task=7
#SBATCH --ntasks-per-socket=1
#SBATCH --distribution=cyclic:block
#SBATCH --time=2:00:00
#SBATCH --mem-per-cpu=1gb
#SBATCH --partition=hpg2-gpu
#SBATCH --gres=gpu:tesla:2

module load gcc/5.2.0  openmpi/1.10.2 gromacs/2016.3-CUDA

export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
srun --mpi=pmi2 --accel-bind=g --ntasks=$SLURM_NTASKS gmx mdrun -v