LAMMPS

Description

LAMMPS , Large-scale Atomic/Molecular Massively Parallel Simulator, is a molecular dynamic simulator that models an ensemble of particles in liquid, solid, or gaseous state. It is a open source software, written in C++ and developed at Sandia laboratory. It can be used to model atomic, polymeric, biological, metallic, granular, and coarse-grained systems using various force fields and boundary conditions.

Environment Modules

Run module spider LAMMPS to find out what environment modules are available for this application. LAMMPS at HiperGator is installed in three flavors:

GNU compiler with openmpi
Intel compiler with openmpi
NVIDIA NGC container

System Variables

HPC_LAMMPS_DIR - installation directory
HPC_LAMMPS_BIN - executable directory
HPC_LAMMPS_LIB - library directory

Additional Information

The GNU version of LAMMPS supports external packages PLUMED,EXTRA-MOLECULE,EXTRA-COMPUTE,MISC,ML-QUIP,INTEL,KOKKOS. The Intel version of LAMMPS only includes default packages.

To execute lammps binaries set up the appropriate environment variables and execute them with srun using the pmix level corresponding to the openmpi version used to build the binary. E.g. the command will look similar to

srun --mpi=${HPC_PMIX} $LAMMPS -sf gpu -pk gpu 2 -var x 2 -var y 7 -var z 7 < in.$job

For standard or user-customized installations the fastest way is to use cmake.

Expand this section to view instructions for standard installation.

Standard installation:

cd <install-dir>
tar xzf lammps-<version>.tar.gz
cd lammps-<version>

ml purge
ml cmake/<version> intel/<version> openmpi/<version>
ml list

mkdir build
cd build
cmake -D CMAKE_INSTALL_PREFIX=<target_lammps_dir> ../cmake
make -j8
make install

Using presets for customization - e.g. for a very rich set of packages:

cmake -D CMAKE_INSTALL_PREFIX=<target_lammps_dir> -C ../cmake/presets/all_on.cmake -C ../cmake/presets/nolib.cmake ../cmake

NGC containers downloaded and installed from NVIDIA GPU Cloud - Usage:

ml purge
ml ngc-lammps
lmp <your_parameters>
OR (for a MPI run):
mpirun -n <#_of_processes> lmp <your_parameters>

Job Script Examples

Note: This section may require cleanup to meet UFRC standards. It's either outdated, has factual errors, too terse, too verbose, or inappropriate for UFRC public wiki.

Expand this section to view sample serial script.

#!/bin/bash
#SBATCH --job-name=<JOBNAME>
#SBATCH --mail-user=<EMAIL>
#SBATCH --mail-type=FAIL,END
#SBATCH --output <my_job-%j.out>
#SBATCH --error <my_job-%j.err>
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --mem-per-cpu=2G
#SBATCH --time=01:00:00
#SBATCH --account=<GROUP>
#SBATCH --array=<BEGIN-END>
 
module load intel/2016.0.109 lammps
 
LAMMPS=lmp_ufhpc
INPUT=<input_file>
 
mpiexec $LAMMPS < $INPUT > log.out 2>&1

Expand this section to view sample parallel script.

#!/bin/bash
#SBATCH --job-name=<JOBNAME>
#SBATCH --mail-user=<EMAIL>
#SBATCH --mail-type=FAIL,END
#SBATCH --output <my_job-%j.out>
#SBATCH --error <my_job-%j.err>
#SBATCH --nodes=1
#SBATCH --ntasks=<number of tasks>
#SBATCH --mem-per-cpu=2G
#SBATCH --time=01:00:00
#SBATCH --account=<GROUP>
#SBATCH --array=<BEGIN-END>
 
cd $PBS_O_WORKDIR
 
module load intel/2016.0.109 openmpi/1.10.2 lammps
 
LAMMPS=lmp_ufhpc
INPUT=<input_file>
 
mpiexec $LAMMPS < $INPUT > log.out 2>&1