CÉCI High-Performance Computing#

The CÉCI (Consortium des Équipements de Calcul Intensif in French) is the High Performance Computing infrastructure of UCLouvain, ULB, ULiège, UMons, and UNamur and is supported by the F.R.S-FNRS and the Walloon Region. It gives access to several HPC clusters composed of: thousands of CPU cores, large amounts of shared and distributed memory and high-performance storage systems.

Using HPC cluster is recommended in the following situations:

simulations running longer than a few hours;
two-dimensional or three-dimensional numerical models;
parametric studies involving multiple simulation runs;
parallel computations (MPI, OpenMP, …);
post-processing of large datasets.

In the context of Watlab, most hydrodynamic or sediment transport simulations fall into at least one of those categories.

Accessing the CÉCI and installing required tools#

To access CÉCI clusters, you must be officially affiliated with a university member of the CÉCI consortium, be holder of an official university email address, and be endorsed by a supervisor with a permanent academic or scientific position. If this is not the case, unfortunately you will not be able to use the CÉCI clusters, but informations below may be useful if you plan to use another HPC cluster.

Accessing Lemaitre4 CÉCI cluster requires:

a valid user account;
a properly configured SSH connection;
an appropriate client depending on the operating system (e.g. MobaXterm on Windows).

These steps are fully described and regularly updated in the official CÉCI documentation. To avoid duplication and ensure consistency, we refer users directly to the official pages:

Once these steps have been completed, you should be able to log in to Lemaitre 4 cluster in MobaXterm (on Windows).

Logo

Compile hydroflow on Lemaitre4 cluster#

Once access to the CÉCI is proprtly configured, you will copy-paste the source code and compile with the following commands :

cd hydroflow
cmake -S . -B _build/
cmake --build _build

It will build several executables : hydroflow and hydroflow_parallel for hydrodynamic simulations and meshchecker for mesh verification/correction (and executables for testing purposes).

Usefull executable files to run a simulation are located in the folder hydroflow/src/watlab/bin.

Run a simulation on Lemaitre4 cluster#

After generating the input files with the model.export.data() function, copy and paste the input and output folders onto the cluster and run the following command sbatch bash_file.sh. The bash_file.sh should contain the following commands and be placed in the same folder as input and output folders:

#!/bin/bash
#Submission script for Lemaitre4
#SBATCH --job-name=Exemple_Hotton
#SBATCH --time=0-00:05:00 # days-hh:mm:ss
#
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=14
#SBATCH --mem-per-cpu=20 # megabytes
#SBATCH --partition=batch
#
#SBATCH --mail-user=martin.petitjean@uclouvain.be
#SBATCH --mail-type=ALL
#
#SBATCH --comment=Exemple_Hotton
#
#SBATCH --output=Exemple_Hotton_out

module purge
module load foss
export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK

ml releases/2021b parallel

export results_dir="output/"
mkdir -p ${results_dir}

../hydroflow/src/watlab/bin/hydroflow-parallel-linux -f "input/data.json"