Recently I am trying to perform convex hull calculations suggested by referee.
I met a problem using the computer cluster with queueing system: the CPU load begin to increase with the increase of new generated structures (0,1,2,3…). i.e., many structures (e.g. from the 10 structure or earlier) simultaneously run in a node so that the %CPU decrease fromm 99% to e.g. 50% (the correct procedure should be one by one ? 100% cpu ). Attached below are the script and ezvasp.rc files.
Thanks very much.
#!/bin/bash
#BSUB -J vasp-job
#BSUB -q snode
#BSUB -n 32
#BSUB -R "span[ptile=32]"
#BSUB -e %J.err
#BSUB -o %J.out
#Environment
. /public/home/users/app/compiler/intel-2017.8/compilers_and_libraries_2017.8.262/linux/bin/compilervars.sh intel64
CURDIR=PWD
#Generate nodelist
rm -f {CURDIR}/nodelist >& /dev/null
for i in echo $LSB_HOSTS
do
echo i >> {CURDIR}/nodelist
done
NP=cat ${CURDIR}/nodelist |wc -l
maps -d &
pollmach -s runstruct_vasp mpirun -np NP -machinefile {CURDIR}/nodelist ${HOME}/bin/vxy
.ezvasp.rc file
#!/bin/csh
#enter name of vasp executable here
set VASPCMD="vxy"
#enter the directories containing the pseudopotentials here
set POTLDA="/usr/local/lib/pot_lda"
set POTGGA="/usr/local/lib/pot_gga"
set POTPAWLDA="/usr/local/lib/paw_lda"
set POTPAWGGA="/usr/local/lib/paw_gga"
set POTPAWPBE="/public/home/users/nfpc001/potential"
~