Back

GPAW

GPAW

Description

GPAW is an efficient program package for electronic structure calculations. It is based on the density-functional theory (DFT) implemented within the projector augmented wave (PAW) method using uniform real-space grids. GPAW can be used for systems containing up to hundres of atoms and thousands of electrons.

Some features of the software:

  • total energy calculations
  • structural optimizations
  • different boundary conditions (finite, wire, slab, bulk)
  • efficient parallelization
  • excited state properties within time-dependent density-functional theory

 


Available

Version on CSC's Servers

Sisu: 0.9.0, 0.10.0, 0.11.0, 1.1.0, 1.2.0, 1.3.0, 1.4.0

Taito: 0.9.0, 0.10.0, 0.11.0, 1.1.0, 1.2.0, 1.3.0, 1.4.0


Usage

GPAW can be used from command line or with limited functionality through the SOMA2 modeling environment (GUI through web browser). The GPAW environment can be set up with the module command.

On Sisu, one loads the default version with the command

module load gpaw

and a specific version (e.g. 1.4.0) with a command such as

module load gpaw/1.4.0

On Taito, the commands are the same, but the name of the module is gpaw-env.

module load gpaw-env

module load gpaw-env/1.4.0

GPAW input files are python scripts and interactive and serial calculations are run by python interpreter as

python input.py

For parallel calculations a special gpaw-python executable is used, whose full pathname in CSC environment is given in $GPAW_PYTHON variable.

gpaw-python input.py

Below are more specific instructions for different CSC machines.

Sisu

Sample job script for parallel calculations:

#!/bin/bash
#SBATCH -J jobname
#SBATCH -e job_err
#SBATCH -o job_out
#SBATCH -N 8
#SBATCH -p small
#SBATCH -t 00:01:10
#SBATCH --no-requeue

((ncores = 24 * SLURM_NNODES))

module load gpaw/1.4.0

aprun -n $ncores gpaw-python input.py

Taito

Sample job script for parallel calculations:

#!/bin/bash
#SBATCH -J jobname
#SBATCH -e job_err
#SBATCH -o job_out
#SBATCH -N 2
#SBATCH -n 48
#SBATCH -p parallel
#SBATCH -t 00:01:10
#SBATCH --mem-per-cpu=1000

module load gpaw-env/1.4.0

srun gpaw-python input.py

# this script will print some usage statistics to the end of job_out file.
used_slurm_resources.bash 

FGI grid

You can find the instructions to run GPAW in the FGI Grid from the GPAW runtime environment page.

DataWarp on Sisu

Compute nodes on Sisu have so called DataWarps available, i.e. fast flash-based temporary storage attached directly to the nodes. These may offer improved burst-I/O buffering for I/O patterns that are not optimal for the Lustre filesystem. If your calculation involves a lot of frequent reads/writes, it is worthwhile to test if DataWarp gives you a performance boost. For more information, please see Sisu User Guide.


Discipline

Chemistry
Physics

References


Support


Manual

Manual and further information can be found from GPAW-wiki.