VASP
VASP is an ab initio DFT program for computing electronic structures of up to few hundreds of atoms.
This page briefly describes how to use VASP on mahti.csc.fi. Usage on puhti.csc.fi is very similar. That said, VASP is a program the usage of which requires some experience. It is advised that new VASP users start out with a supervisor or an experienced colleague.
Available
See available VASP versions in with command
License prior to version 6.5.1
The usage of VASP requires a license, which must be acquired directly from the developers of the software.
VASP versions prior to 6.5.1 use unix goups 'vasp' and 'vasp6' to control access to pre-installed executables. After acquiring the license, or after your email address has been added to an existing license, please send an email to CSC Service Desk, including your username at CSC, and the email address you have registered for the VASP license in the VASP Portal.
License for version 6.5.1
VASP version 6.5.1 is available in mahti.csc.fi.
VASP versions from 6.5.1 onwards require license file '~/.vasp/vasp_license'. The license file is downloaded from the VASP portal with commands
The request_license_key.sh script will ask user's VASP Portal username and password.
This license model is currently in beta stage. The license information in the beta stage license
server is not necessarily in sync with the data in VASP Portal. Until the
new license schema is rolled out properly, the owner of the license can download the vasp
source file vasp.6.5.1.tgz from the portal to /projappl/<project name> directory in
mahti.csc.fi, and build the executables and run the tests with
Usage
Precompiled VASP executables and pseudopotentials are available through the module environment. Use the command
to see more detailed information.
An example batch job script for a small test
#!/bin/bash
#SBATCH --time=00:15:00
#SBATCH --partition=test
#SBATCH --ntasks=4
#SBATCH --mem-per-cpu=1GB
#SBATCH --account=<project>
module load vasp
srun vasp_std
For more options and details, see how to create batch job scripts for Puhti and Mahti.
VASP tutorials in JupyterLab
VASP tutorials can also be
followed using JupyterLab from the
Mahti web interface. Open the Jupyter app,
and from Settings -> Python, select Custom module and type in
py4vasp. When submitting jobs from the JupyterLab terminal window to
compute nodes, first load module vasp, and then use a command similar to
instead of the mpirun ... command shown in the tutorial.
Performance optimization
First, the performance of VASP depends crucially on the parameters in the INCAR file that control how the different k-points, bands and FFT coefficients are distributed among the MPI tasks, and that the correct version (std/gam/ncl) of the executable is used.
Second, the provided prebuilt executables are built as "vanilla" as
possible and provide a reasonable baseline. The performance
optimization for large experiments should be done on a per case basis.
The commands that created the prebuilt executables are in
$VASPDIR/README.sh, and can be used as a starting point
for building more optimized and/or otherwise modified executables.