Job performance is lower when scheduled through pbs

[adarsh]

Thank you @tpan1039 , could you please try the below script and check whether it makes any difference.

#PBS -q gpu
#PBS -l walltime=168:00:00
#PBS -l select=1:ncpus=24:ngpus=1:mem=48gb:mpiprocs=24:host=g6:ompthreads=1
cd $PBS_O_WORKDIR
module purge
module load lammps/2021-Sep-29-gpu-cuda-11.0-gcc-9.2
mpirun -np 24 lmp -sf gpu -pk gpu 0 -in in.NiPd > lmp_out_pbs.dat

@tpan1039

I used openmpi-4.0.6 and compiled it from the source code. I don’t know what is openPBS TM, so I think I didn’t use that.

Compile OpenMPI with PBSpro 14.1.10

Please refer to an earlier post in this thread: Compile OpenMPI with PBSpro 14.1.10 - #5 by mkaro You’ll need to customize the script so that the --libs parameter returns “/opt/pbs/lib/libpbs.a -lpthread -lcrypto” The autotools macros used to detect PBS are not very sophisticated and rely on this script being present. If you look for files in the OpenMPI tree that end with check_tm.m4 you’ll see how they rely on the script being present. The only other option I see would be to update the macr…