GPAW Installations on Tetralith & Sigma

The GPAW installations are generally maintained by Weine Olovsson (weiol@nsc.liu.se).

"GPAW is a density-functional theory (DFT) Python code based on the projector-augmented wave (PAW) method and the atomic simulation environment (ASE)."

How to run

A sample batch script for running the latest installed GPAW version 20.1.0 in parallel:

#!/bin/bash
#SBATCH -n 32
#SBATCH -t 2:00:00
#SBATCH -J jobname
#SBATCH -A SNIC-xxx-yyy

export OMP_NUM_THREADS=1

export GPAW_SETUP_PATH=/software/sse/manual/GPAW-setups/0.9.20000

module load Python/3.6.7-env-nsc1-gcc-2018a-eb 
module load GPAW/20.1.0-vdwxc-nsc2-gcc-2018a-eb 
gpaw -P 32 python input.py

(Note that you should edit the jobname and the account number before submitting.)

In this case note that mpprun isn't used for parallel calculations, instead the number of ranks should be set explicitly after -P. So, for example, running on 2 nodes -N 2 or -n 64, the line would read:

gpaw -P 64 python input.py

For this installation libvdwxc 0.4.0 is included (also see last section for such installations). For ASE nsc2 includes version 3.18.0 (however 3.19.0 is recommended), whilensc1includes3.19.1` which might be incompatible.

GPAW finds the path to the potentials by setting GPAW_SETUP_PATH as shown above.

For running older versions of GPAW in parallel with mpprun, a sample batch script:

#!/bin/bash
#SBATCH -n 32
#SBATCH -t 2:00:00
#SBATCH -J jobname
#SBATCH -A SNIC-xxx-yyy

export OMP_NUM_THREADS=1

export GPAW_SETUP_PATH=/software/sse/manual/GPAW-setups/0.9.20000

module load ASE/3.17.0-nsc1 Python/3.6.7-env-nsc1-gcc-2018a-eb 
module load GPAW/1.5.2-nsc1-gcc-7.3.0-bare-eb 
mpprun $GPAW_EXE input.py

(Note that you should edit the jobname and the account number before submitting.)

In order to use GPAW, it needs to find ASE in its path. The recommended version for GPAW 1.5.2is ASE 3.17.0, therefore ASE is also loaded in the line, together with the Python version used for building GPAW:

module load ASE/3.17.0-nsc1 Python/3.6.7-env-nsc1-gcc-2018a-eb
module load GPAW/1.5.2-nsc1-gcc-7.3.0-bare-eb  

GPAW finds the path to the potentials by setting GPAW_SETUP_PATH as shown above.

For other versions of GPAW, you can load e.g.:

module load ASE/3.17.0-nsc1 Python/3.6.7-env-nsc1-gcc-2018a-eb
module load GPAW/1.5.1-nsc1-gcc-7.3.0-bare-eb

module load ASE/3.15.0-nsc1 Python/3.6.7-env-nsc1-gcc-2018a-eb
module load GPAW/1.3.0-nsc1-gcc-7.3.0-bare-eb

OBS: in order to run the serial version of the modules gcc-7.3.0-bare-eb you also need to add the following lines in your job script before gpaw input.py:

export LD_PRELOAD=/software/sse/easybuild/prefix/software/imkl/2018.3.222-iimpi-2018b/mkl/lib/intel64/libmkl_avx512.so:/software/sse/easybuild/prefix/software/imkl/2018.3.222-iimpi-2018b/mkl/lib/intel64/libmkl_core.so:/software/sse/easybuild/prefix/software/imkl/2018.3.222-iimpi-2018b/mkl/lib/intel64/libmkl_sequential.so

Installation with libvdwxc

You can find GPAW compiled with libvdwxc as a separate module, the most recent installation was already described at the top of the previous section. The modules for previous versions follow the scheme using mpprun:

module load ASE/3.18.0-nsc1 Python/3.6.7-env-nsc1-gcc-2018a-eb
module load GPAW/19.8.1-vdwxc-nsc1-gcc-2018a-eb

another version:

module load ASE/3.17.0-nsc1 Python/3.6.7-env-nsc1-gcc-2018a-eb
module load GPAW/1.5.2-vdwxc-nsc1-gcc-2018a-eb

all for libvdwxc 0.4.0. Note that these modules were compiled using OpenBLAS and ScaLAPACK and not intel MKL, which was the case of the other modules gcc-7.3.0-bare-eb. There might be a difference in performance, which needs to be checked by user.

Testing and settings

Output from GPAW tests, as well as configuration files (including installation comments), can typically be found in the respective installation paths, e.g. in:

/software/sse/manual/GPAW/20.1.0/vdwxc/nsc1

User Area

User support

Guides, documentation and FAQ.

Getting access

Applying for projects and login accounts.

System status

Everything OK!

No reported problems

Self-service

SUPR
NSC Express