gpu_ocean_saga_test.job

#!/bin/bash
# Job name:
#SBATCH --job-name=gpu_ocean_test
#
# Project:
#SBATCH --account=nn9550k
#
# Wall clock limit:
#SBATCH --time=00:30:00
#
# Ask for 1 GPU (max is 2)
# Note: The environment variable CUDA_VISIBLE_DEVICES will show which GPU 
# device(s) to use. It will have values '0', '1' or '0,1' corresponding to 
# /dev/nvidia0, /dev/nvidia1 or both, respectively.
#SBATCH --partition=accel --gres=gpu:1
#
# Max memory usage per task (core) - increasing this will cost more core hours:
#SBATCH --mem-per-cpu=4G
#
# Number of tasks (for performance benchmarking, use --nodes switch):
##SBATCH --nodes=16 --ntasks-per-node=2 # gives exclusive access to all GPU-nodes
#SBATCH --nodes=4 --ntasks-per-node=1

## Set up job environment: (this is done automatically behind the scenes)
## (make sure to comment '#' or remove the following line 'source ...')
# source /cluster/bin/jobsetup

module restore system   # instead of 'module purge' rather set module environment to the system default
module load Python/3.6.6-fosscuda-2018b #nb: 'Versions' is mandatory! There are no default versions of modules as on Abel!

# It is also recommended to to list loaded modules, for easier debugging:
module list  

export PATH=$HOME/.local/bin:$PATH

set -o errexit # exit on errors
set -o nounset # Treat unset variables as errors (added for more easily discovering issues in your batch script)

## Copy input files to the work directory:
cp -r gpu_ocean/ $SCRATCH

## Make sure the results are copied back to the submit directory (see Work Directory below):
# chkfile MyResultFile
# chkfile is replaced by 'savefile' on Saga
savefile "$SCRATCH/gpu_ocean/prototypes/scripts/mpi_run_benchmark.log"
savefile "$SCRATCH/gpu_ocean/prototypes/scripts/netcdf*/*.nc"

## Do some work:
cd $SCRATCH/gpu_ocean/prototypes/scripts/
srun python mpi_run_benchmark.py --nx 256 --ny 256 --iterations 200 --simulator CDKLM > mpi_run_benchmark.log