Skip to content

Latest commit

 

History

History
 
 

vasp

Folders and files

NameName
Last commit message
Last commit date

parent directory

..
Performance Study 1
Performance Study 1
 
 
Performance Study 2
Performance Study 2
 
 
README.md
README.md
 
 

README.md

  1. Running VASP
  2. VASP Documentation

Running VASP

VASP6 is available on Eagle, Swift and Vermilion. The code below demonstrates how to load different versions of VASP on Swift and Eagle and provides information on accompanying modules to load.

For information about accessing and running VASP on Vermilion, see the VASP section at the bottom of this documentation.

On Eagle

Load VASP with Intel MPI:

ml vasp

script to run VASP on Eagle with Intel MPI

Load VASP with Open MPI:

source /nopt/nrel/apps/210830a/myenv.2108301742
ml vasp/6.1.1-l2mkbb2

script to run VASP on Eagle with Open MPI

Load the GPU build of VASP:

module use /nopt/nrel/apps/220511a/modules/lmod/linux-centos7-x86_64/gcc/12.1.0
ml fftw nvhpc
export LD_LIBRARY_PATH=/nopt/nrel/apps/220511a/install/opt/spack/linux-centos7-skylake_avx512/gcc-12.1.0/nvhpc-22.3-c4qk6fly5hls3mjimoxg6vyuy5cc3vti/Linux_x86_64/22.3/compilers/extras/qd/lib:$LD_LIBRARY_PATH
export LD_LIBRARY_PATH=/nopt/nrel/apps/220511a/install/opt/spack/linux-centos7-skylake_avx512/gcc-12.1.0/nvhpc-22.3-c4qk6fly5hls3mjimoxg6vyuy5cc3vti/Linux_x86_64/22.3/compilers/extras/qd/lib:$LD_LIBRARY_PATH
export PATH=/projects/hpcapps/tkaiser2/vasp/6.3.1/nvhpc_acc:$PATH

script to run VASP on Eagle on GPU nodes

On Swift

Load VASP with Intel MPI:

ml vaspintel 
ml slurm/21-08-1-1-o2xw5ti 
ml gcc/9.4.0-v7mri5d 
ml intel-oneapi-compilers/2021.3.0-piz2usr 
ml intel-oneapi-mpi/2021.3.0-hcp2lkf 
ml intel-oneapi-mkl/2021.3.0-giz47h4

script to run VASP on Swift with Intel MPI

Load VASP with Open MPI:

ml vasp 
ml slurm/21-08-1-1-o2xw5ti 
ml openmpi/4.1.1-6vr2flz

script to run VASP on Swift with Open MPI

VASP Documentation

This repo contains the results of two separate VASP performance studies. The first, Performance Study 1, studies VASP performance on Eagle using the input files provided in the directory. The second, Performance Study 2, studies VASP performance on Eagle and Swift using benchmarks from the ESIF benchmarking suite, which can be found here or in the benchmarks folder in the Performance Harness 2 directory. Each study evaluates performance differently, as described below, and provides recommendations for running VASP most efficiently in the README files. The READMEs in each directory contain the following information.

Performance Study 1 (VASP6 on Eagle):

  • Recommendations for setting LREAL
  • Recommendations for setting cpu pinning
  • Recommendations for setting NPAR
  • Recommendations for setting NSIM
  • Instructions for using the OpenMP version of VASP
  • Instructions for running multiple VASP jobs on the same nodes (and scripts to do so)
  • Runtime comparison using VASP5

Performance Study 2 (VASP6 on Eagle and Swift):

  • Information on how runtime scales with nodecount
  • Recommendations for chosing the most efficient value of cpus/node
  • Recommendations for running VASP on Eagle's GPU nodes (and scripts to do so)
  • Recommendations for chosing Intel MPI or Open MPI (and scripts for running with both MPIs)
  • Recommendations for setting KPAR
  • Recommendations for setting cpu pinning
  • Information on k-points scaling
  • Instructions for running multiple VASP jobs on the same nodes on Swift (and scripts to do so)