-
This site is deprecated and will be decommissioned shortly. For current information regarding HPC visit our new site: hpc.njit.edu
Lammps
In progress. don't rely on this yet for GPU - CPU should be OK.
LAMMPS is a classical molecular dynamics code, and an acronym for Large-scale Atomic/Molecular Massively Parallel Simulator[1]
LAMMPS was compiled using Intel 2013 compilers and the CUDA 6.5 libraries.
The following instructions are for running the LAMMPS in parallel on kong using cpu and gpu.
Create a directory for this run:
<pre=code>mkdir lammps_test cd lammps_test</pre>
Get an example lammps input file: <pre=code>cp /afs/cad/linux/lammps/30Jul16/lammps-30Jul16/bench/in.lj .</pre>
Edit the submit script as directed: <pre=code>
- !/bin/bash
- EXAMPLE LAMMPS PARALLEL SCRIPT FOR SGE
- lammps_submit.sh
- To use, change all of the following to real values:
- JOB_NAME, QUEUE, UCID, NUMBER_OF_CPUS
- Please see https://wiki.hpc.arcs.njit.edu/ for additional information.
- Modified by Basement Supercomputing 1/2/2006 DJE
- Modified for NJIT - AM 18Apr06
- Updated for modules, QRSH_PORT, additional documentation. 31-Aug-2015 /KJW
- Updated for LAMMPS 28-Sep-2016/GW
-
- Change JOB_NAME to a one-word alphanumeric description. This will be used for
- part of output file name and in status listings.
- $ -N JOB_NAME
- Change QUEUE to a valid queue. On kong this would be one of "short",
- "medium", "long, or gpu." See https://wiki.hpc.arcs.njit.edu/index.php/KongQueues
- To run on GPU you mus use the gpu queue
- or https://wiki.hpc.arcs.njit.edu/index.php/SthenoQueues for more info
- $ -q QUEUE
- Send email when jobs starts and again when it finishes or if it aborts.
- $ -M UCID@njit.edu
- $ -m abe
- Use current working directory where "qsub" was invoked.
- $ -cwd
- Join stdout and stderr
- $ -j y
- Set your number of processors here by changing NUMBER_OF_CPUS to an integer
- value.
- LAMMPS uses the intel mpich2 MPI
- Do not select more than 16 CPUS if using the GPU node
- $ -pe mpich2 NUMBER_OF_CPUS
- Uncomment to request one gpu (max two)
- Note that uncomment means to remove the first "#" and leave "#$."
- $ -l gpu=1
- Run job through bash shell
- $ -S /bin/bash
- Load the correct lammps module.
- Uncomment the one to use.
- module load lammps/cpu/30Jul16
- module load lammps/gpu/30Jul16
- The following are for reporting only and not really needed to run the job;
- their results will show up in your output file.
echo "Got $NSLOTS processors." echo "cwd=$(pwd)"
- For CPU uncomment the following.
- mpirun -np $NSLOTS lmp_intel_cpu_intelmpi -v x 16 -v y 16 -v z 16 -v t 100 < in.lj
- For GPU uncomment the following.
- mpirun -np $NSLOTS lmp_intel_cpu_intelmpi -sf gpu -v x 32 -v y 32 -v z 64 -v t 100 < in.lj
</pre>
Submit the script: <pre=code>qsub lammps_submit.sh</pre>