Environment modules

The cluster uses module to manage compilers, MPI, CUDA, scientific applications, and bioinformatics tools. Load the required version to avoid conflicts. If a command is missing, first check available modules.

Find software

module av

Filter by name:

module av cuda
module av gcc
module av mpi
module av gromacs

Module directories include:

/hpcfs/fpublic/app/modulefiles/Compilers
/hpcfs/fpublic/app/modulefiles/Libraries
/hpcfs/fpublic/app/modulefiles/Science

Category	Examples
Compilers	gcc, nvhpc, aocc-compiler
Runtime libraries	cuda, glibc, node
MPI	mpi/intel, mpi/openmpi, mpi/mpich
Scientific computing	gromacs, vasp, gaussian, orca, cp2k
Bioinformatics and structural tools	ncbi-blast, rosetta, relion, PyMOL

Load a module

module load mpi/intel/oneapi/2023.1

Commands are added to the current shell environment. Check their locations:

which icc
which ifort
which mpirun

Example output:

/hpcfs/fpublic/app/intel/oneapi/2023.1/compiler/2023.1.0/linux/bin/intel64/icc
/hpcfs/fpublic/app/intel/oneapi/2023.1/compiler/2023.1.0/linux/bin/intel64/ifort
/hpcfs/fpublic/app/intel/oneapi/2023.1/mpi/2021.9.0/bin/mpirun

Inspect loaded modules

module list

Example:

Currently Loaded Modulefiles:
 1) mpi/intel/oneapi/2023.1

Change versions

Unload the old version before loading another:

module unload mpi/intel/oneapi/2023.1
module load mpi/intel/2020u4
module list

Or clear the environment and reload what you need:

module purge
module load mpi/intel/2020u4

Avoid incompatible environments

Do not mix incompatible MPI, CUDA, or compiler installations in the same shell. Check module list before submitting.

Use modules in Slurm scripts

Load dependencies explicitly inside batch scripts so the compute-node environment is reproducible.

Adapt this original example

The source example below uses qcpu, which is not listed in the platform partition guide. Replace it with an authorized CPU partition. For four MPI processes, request four tasks (--ntasks=4); --cpus-per-task=4 alone describes four CPUs for each task, not four MPI tasks. See the OpenMPI guide for a complete hybrid example.

#!/bin/bash
#SBATCH --job-name=test-module
#SBATCH --partition=qcpu
#SBATCH --cpus-per-task=4
#SBATCH --time=01:00:00
#SBATCH --output=%x_%j.out

module purge
module load mpi/intel/oneapi/2023.1

module list
which mpirun

mpirun -np 4 ./your_program

Completion and reference

Press Tab after part of a module name to complete it or list versions.

Command	Action
module av	List available modules
module av keyword	Filter modules
module load name	Load a module
module unload name	Unload a module
module list	Show loaded modules
module purge	Unload all modules

If a command remains unavailable, check module list and which, then verify the module name/version. For library conflicts, purge and load only required dependencies. Consult the application guide or administrator when unsure, and test with a small input before production use.