Using Software Modules on REPACSS¶
REPACSS offers a wide selection of system-provided research and instructional software using the Spack package manager and Lmod environment module system. This guide shows you how to find, load, and manage software modules, and provides an up-to-date summary of available software environments.
This covers system-provided HPC applications, compilers, and system software - no installation required, just load what you need. For user-installed data science applications, see Getting Started with MiniForge next.
Available Software Modules¶
Software modules are grouped by repository path and hardware partition. You can load any module on any partition, but only GPU-enabled packages can utilize GPU resources when running on the GPU (h100) partition. Some modules (such as MPI libraries) are available in both partitions, but with different capabilities.
zen4 Partition¶
/opt/apps/nfs/modules/zen4/rocky9.4/Core¶
| Module | Version | Description |
|---|---|---|
| boost | 1.86.0 | C++ libraries for tasks and structures |
| bwa | 0.7.17 | Burrows-Wheeler Aligner for sequence alignment |
| eigen | 3.4.0 | C++ template library for linear algebra |
| ffmpeg | 6.1.1 | Audio and video processing tools |
| gdal | 3.10.0 | Geospatial Data Abstraction Library |
| geos | 3.13.0 | Geometry Engine - Open Source |
| gcc | 14.2.0 | GNU Compiler Collection (C, C++, Fortran) |
| gsl | 2.8 | GNU Scientific Library |
| gnuplot | 6.0.0 | Portable command-line driven graphing utility |
| htop | 3.3.0 | Interactive process viewer |
| imagemagick | 7.1.1-39 | Image manipulation tools |
| intel-oneapi-compilers | 2023.0.0 | Intel OneAPI Compilers (C++, Fortran) |
| metis | 5.1.0 | Serial graph partitioning and fill-reducing matrix ordering |
| mpich | 4.1.2 | MPI implementation (default) |
| ninja | 1.12.1 | Small build system with a focus on speed |
| openblas | 0.3.28 | Optimized BLAS library |
| openmpi | 5.0.4 | High-performance MPI (default) |
| proj | 9.4.1 | Cartographic projections and coordinate transformations |
| pmix | 5.0.3 | Standardized API for launching and managing parallel applications |
| python | 3.10.10 | Python programming language |
| python | 3.12.5 | Python programming language (default) |
| r | 4.4.1 | Statistical computing and graphics |
| samtools | 1.19.2 | Tools for manipulating next-generation sequencing data |
| sparsehash | 2.0.4 | Memory-efficient hash map implementation |
| swig | 4.1.0-fortran | Simplified Wrapper and Interface Generator for Fortran |
| tmux | 3.4 | Terminal multiplexer |
/opt/apps/nfs/modules/zen4/rocky9.4/mpich/4.1.2-a5xh3ge/Core¶
Note
These modules are only available after loading the mpich module (default in zen4). Run:
module load mpich
module show mpich
/opt/apps/nfs/modules/zen4/rocky9.4/mpich/4.1.2-a5xh3ge/Core.
| Module | Version | Description |
|---|---|---|
| fftw | 3.3.10 | Fast Fourier Transform library used in signal and image processing |
| hdf5 | 1.14.5 | Hierarchical Data Format library for large/complex data |
| lammps | 20240829.1 | Molecular dynamics simulator |
| netcdf-c | 4.9.2 | Data format for array-oriented scientific data |
| openfoam | 2312 | Open source computational fluid dynamics toolbox |
| p3dfft3 | 3.0.0 | Parallel 3D FFT library |
| quantum-espresso | 7.4 | Electronic-structure and materials modeling suite |
| scotch | 7.0.4 | Graph partitioning, static mapping, and clustering library |
| wrf | 4.6.1 | Weather Research and Forecasting Model |
| wps | 4.5 | WRF Preprocessing System |
/opt/apps/nfs/modules/zen4/rocky9.4/oneapi/2023.0.0¶
Note
These modules are only available after loading the intel-oneapi-compilers module. Run:
module load intel-oneapi-compilers
| Module | Version | Description |
|---|---|---|
| intel-oneapi-mkl | 2024.2.2 | Intel Math Kernel Library |
| intel-oneapi-mpi | 2021.4.0 | Intel MPI Library |
| quantum-espresso | 7.4 | Electronic-structure and materials modeling suite |
/opt/apps/nfs/modules/zen4/rocky9.4/intel-oneapi-mpi/2021.4.0-5ksoj4d/oneapi/2023.0.0¶
Note
These modules are only available after loading the intel-oneapi-mpi module. Run:
module load intel-oneapi-mpi
| Module | Version | Description |
|---|---|---|
| fftw | 3.3.10 | Fast Fourier Transform library used in signal and image processing |
| hdf5 | 1.14.5 | Hierarchical Data Format library for large/complex data |
h100 (GPU Partition)¶
/opt/apps/nfs/modules/h100/rocky9.4/Core¶
These modules are GPU-enabled. To access them, prepend the h100 path to your MODULEPATH (see MPI section below):
| Module | Version | Description |
|---|---|---|
| cuda | 12.6.2 | NVIDIA CUDA Toolkit |
| cudnn | 9.2.0.82-12 | NVIDIA CUDA Deep Neural Network library |
| mpich | 4.1.2 | MPI implementation (GPU-enabled) |
| nccl | 2.22.3-1 | NVIDIA Collective Communications Library |
| openmpi | 5.0.4 | High-performance implementation of MPI (GPU-enabled) |
MPI Implementations and Usage Guidance¶
Overview¶
Three MPI implementations are available: OpenMPI, MPICH, and Intel MPI. OpenMPI and MPICH are provided for both CPU (zen4) and GPU (h100) partitions, while Intel MPI is only available in the CPU (zen4) partition. The h100 versions are CUDA-enabled for GPU workloads.
- zen4: CPU-only MPI modules (default: mpich, openmpi) + Intel MPI (after loading intel-oneapi-compilers)
- h100: GPU-enabled MPI modules (mpich, openmpi)
Using MPI-Dependent Packages¶
Some software packages in the zen4 partition require specific MPI modules to be loaded first. Two sets of MPI-dependent packages are available:
MPICH-Dependent Packages¶
Software packages like fftw, hdf5, lammps, openfoam, quantum-espresso, wrf, etc. (see the MPICH table above) require the mpich module to be loaded first:
module load mpich
To make sure you are loading the default zen4 mpich, not the h100 mpich, verify with:
module show mpich
The output should reference /opt/apps/nfs/modules/zen4/rocky9.4/mpich/4.1.2-a5xh3ge/Core.
Intel MPI-Dependent Packages¶
Software packages like fftw and hdf5 (see the Intel MPI table above) require the Intel MPI environment to be loaded first:
module load intel-oneapi-compilers
module load intel-oneapi-mpi
Running MPI Applications¶
When running MPI applications, use the appropriate launcher for your MPI implementation:
-
For OpenMPI and MPICH: Use
srun(SLURM's native launcher)srun -n 4 ./your_mpi_program -
For Intel MPI: Use
mpirun(Intel MPI's launcher)mpirun -np 4 ./your_mpi_program
This ensures optimal performance and compatibility with each MPI implementation.
Accessing GPU-Enabled MPI Modules¶
To use CUDA-enabled mpich or openmpi in the GPU (h100) partition:
1. Prepend the h100 module path to your MODULEPATH:
module use /opt/apps/nfs/modules/h100/rocky9.4/Core
mpich or openmpi).
Switching Between MPI Implementations¶
To switch from one MPI implementation to another, use:
module swap mpich openmpi
How to Use Modules¶
List Available Modules¶
Show all modules you can load:
module avail
Show only default modules:
module --default avail
# or
ml -d av
See module categories and counts:
module overview
# or
ml ov
Search for Modules¶
Find all available software and extensions:
module spider
Search for a specific module or keyword:
module spider python
Get details for a specific version:
module spider python/3.11.0
Loading and Unloading Modules¶
Load a module to use its software:
module load gcc/14.2.0
module load openmpi/4.1.6
You can load multiple modules at once:
module load gcc/14.2.0 cuda/12.0
Specifying exact versions is recommended for reproducibility.
To remove a specific software module from the current environment:
module unload gcc/14.2.0
To remove all loaded modules and restore a clean environment:
module purge
Viewing Module Details and Loaded Modules¶
See what a module changes in your environment:
module show hdf5
List all modules currently loaded:
module list
This helps with troubleshooting and documenting your environment.
Best Practices for Job Scripts¶
- Start job scripts with
module purgefor a clean environment. - Load only the modules you need.
- Record loaded modules for reproducibility.
More Information and Support¶
For a full list of available software, see the tables above.
For Python-specific help, see the Installing Packages guide.
Need help? Contact REPACSS support:
repacss.support@ttu.edu