LAMMPS on t2

LAMMPS stands for Large-scale Atomic/Molecular Massively Parallel Simulator. It allows scientists to model, analyze, and predict the physical movements and behaviors of atoms and molecules.

How to load LAMMPS

To load LAMMPS into your computing environment, type in

module load lammps

This loads LAMMPS 2025.07.22, and its dependencies, into your computing environment.

In order to run calculations with this software, request resources on the compute nodes on t2 by using the following job submission script (set of instructions):

#!/bin/bash

#BSUB -J jobName                 # name of your job
#BSUB -e %J.err                  # file name for std. error output
#BSUB -o %J.out                  # file name for std. out output
#BSUB -q normal                  # request run on normal queue
#BSUB -n 20                      # request 1 core
#BSUB -W 2:00                    # request 2 hours of runtime
#BSUB -P XYZ                     # your projectID
#BSUB -gpu "num=1"               # request a single NVIDIA V100 GPU

module load lammps

## Launch 20 tasks that use 1 GPU
mpirun lmp -sf gpu -pk gpu 1 -in in.lj

## If launching multiple distributed. Adjust bsub parameters accordingly

If Kokkos support is desired, then one needs to load LAMMPS 2024.08.29 instead

#BSUB -J jobName                 # name of your job
#BSUB -e %J.err                  # file name for std. error output
#BSUB -o %J.out                  # file name for std. out output
#BSUB -q normal                  # request run on normal queue
#BSUB -n 20                      # request 1 core
#BSUB -W 2:00                    # request 2 hours of runtime
#BSUB -P XYZ                     # your projectID
#BSUB -gpu "num=2"               # request two NVIDIA V100 GPU

module load lammps/20240829.4-gcc-13.4.0-rlyq2s7

export OMP_PROC_BIND=spread
export OMP_PLACES=threads

## Launch 20 tasks
mpirun lmp -in in.lj -k on g 2 -sf kk -pk kokkos ..........