Supercomputer Use

Supercomputer Use

Digital Research Alliance of Canada Modules

module load StdEnv/2023
module load gcc/12.3
module load netcdf-fortran
module load tbb

Submission Scripts

We have some example submission scripts located at https://git.cs.usask.ca/numerical_simulations_lab/actors/Summa-Actors-Utils/-/tree/main/sbatch?ref_type=heads

Distributed Use

When using Summa-Actors in a mode of operation that requires all nodes in the job to be started at the same time, the following script details need to be specified. Because using actors requires an initialization step of client-server, to connect nodes together, the job script needs to specify how to start the server and the clients individually. This is straightforward when using an array job, however, when one requires that the resources be acquired at the same time then the following can be used:

Example Batch Script for starting a job on two different nodes as part of the same sbatch submission.

#!/bin/bash
#SBATCH --nodes=2
#SBATCH --ntasks-per-node=1 
#SBATCH --cpus-per-task=32
#SBATCH --mem=0
#SBATCH --time=00:2:00
#SBATCH --job-name=job-name
#SBATCH --output=test-%A-%n.out
#SBATCH --account=


output_dir=test-%A-%J.out

srun -N1 --output=$output_dir --exclusive greet_and_show_hostname.sh &
srun -N1 --output=$output_dir --exclusive greet_and_show_hostname.sh &

wait

In the above example, we ask for 2 nodes, and then we specify how to start a job on each node. In this simple example we start the same job and run the same script. However, this will be updated for how to start Summa-Actors once the development is completed.

Anvil-Modules

module load gcc/11.2.0 module load openmpi module load netcdf-fortran