NAMD AMD GPUs (Kebnekaise)

Although some parallel software could experience some performance overhead by using containers instead of the compiled versions, sometimes this is still a good alternative. One case is NAMD3 a software for Molecular Dynamics when it needs to run on AMD GPUs.

Setup

AMD offers containers for this NAMD version that can run straightforwardly on AMD GPUs, for instance, MI100s. On the machine you are targeting:

apptainer pull docker://amdih/namd3:3.0a9

apptainer shell namd3_3.0a9.sif

This software requires that you accept the terms of a license. The license together with examples and benchmarks can be found at here. One can use the following batch script to run this software.

#!/bin/bash
#SBATCH -A Project_ID
#SBATCH -t 20:17:00
#SBATCH -c 14
#SBATCH -o output_%j.out          # output file
#SBATCH -e error_%j.err           # error messages
#SBATCH --gpus-per-node=mi100:1   # using MI100 GPU cards

ml purge > /dev/null 2>&1
apptainer run /your-project-folder/NAMD3_ROCM/namd3_3.0a9.sif namd3 apoa1_hip.namd +p1 +setcpuaffinity  +devices 0 > output.log

In the output.log file you can see that NAMD3 is using the allocated resources:

Charm++> Running on 1 hosts (1 sockets x 14 cores x 1 PUs = 14-way SMP)
Charm++> cpu topology info is gathered in 0.002 seconds.
Pe 0 physical rank 0 binding to CUDA device 0 on b-cn1612.hpc2n.umu.se: ''  Mem: 32752MB  Rev: 9.0  PCI: 0:27:0
Info: NAMD 3.0alpha9 for Linux-x86_64-multicore-HIP
Info: 
Info: Please visit http://www.ks.uiuc.edu/Research/namd/
Info: for updates, documentation, and support information.
Info: 
Info: Please cite Phillips et al., J. Chem. Phys. 153:044130 (2020) doi:10.1063/5.0014475
Info: in all publications reporting results obtained with NAMD.