IOP Chester 2019

Lectures

DFT, PW, and PAW: "VASP: The basics. DFT, plane waves, PAW, ...".

Hybrid functionals: "VASP: Hybrid functionals".

Beyond DFT: RPA: "VASP: beyond DFT. The Random-Phase-Approximation".

Performance: "VASP: running on HPC resources".

Allocating Computing Nodes and Environment Setup

An interactive shell should be allocated after login. The following command allocates an interactive node with 8 CPUs for 4 hours

qsub -IVl select=1:ncpus=8,walltime=4:0:0,place=scatter:excl -A y15

After successful allocation, one has to setup the environment as follows.

To have access to the vasp binaries, the corresponding module has to be loaded into the environment. Furthermore, the job scripts found in the tutorial tar files (job.sh, doall.sh, etc) work only if the environment variables "vasp_std, vasp_gam, vasp_ncl" are defined. Enter following commands in the terminal window after login, to setup up the environment.

module load vasp
export vasp_std="mpirun -ppn 8 -np 8 /lustre/home/y07/vasp5/5.4.4-intel17-mpt214/bin/vasp_std"
export vasp_gam="mpirun -ppn 8 -np 8 /lustre/home/y07/vasp5/5.4.4-intel17-mpt214/bin/vasp_gam"
export vasp_ncl="mpirun -ppn 8 -np 8 /lustre/home/y07/vasp5/5.4.4-intel17-mpt214/bin/vasp_ncl"

Submitting jobs

Alternative to an interactive shell, one may submit jobs to the cluster as follows

qsub vasp.job

where the jobfile "vasp.job" reads

#!/bin/bash --login

#PBS -N VASP-Test
#PBS -l select=1:ncpus=36
#PBS -l place=scatter:excl
#PBS -l walltime=00:30:00
#PBS -A y15 

cd $PBS_O_WORKDIR 
module load vasp
mpiexec_mpt -ppn 36 -n 36 vasp_std | tee vasp.out