Examples¶

Objectives

Be able to find some typical software modules that are installed
Learn how to load those software modules (and any prerequisites)
Be able to find out if Python packages are installed as their own modules
Learn how to find bundles of R packages installed as modules

In the previous section we looked at toolchains, which are often used to install or build your own software. In this chapter we will look at some examples of how to find and load some of the typical software modules that are installed at many of the Swedish HPC centres.

Loading modules works the same whether the modules are toolchains or standalone packages, with and without prerequisites. The procedure is usually some variation on the following:

Use ml spider <package> to see whether a module is installed, and if so, view the available versions. (At facilities that do not hide packages depending on prerequisites, module avail <package> may be preferred.)
Use ml spider <package>/<version> to view prerequisites for a specific version of the module, if any.
Load the modules with ml <prerequisite>/<version> <package>/<version>.

Most of the examples below use outputs from Cosmos, but the workflows will be similar at other institutions except where otherwise noted.

Important

Both R and Python packages often include many so-called extensions (dependent packages that are usually called modules when not dealing with LMOD modules) that are installed, but cannot be found with ml spider or ml avail.

In these cases, if you have loaded at least the prerequisites of a standalone package containing the extension, you can use ml show <package> on the standalone package to view the Lua module file, which usually has a section on included extensions near the top.

For example, NumPy, SciPy, and Pandas, among other packages, are all are included in SciPy-bundle. If you have at least loaded a GCC version, you can use ml show SciPy-bundle to view all of the included extensions (Python modules) in the compatible SciPy-bundle.

Tip

grep does not work directly on the outputs of module commands like ml show <module>. To search for an extension in a very long Lua module file, copy the full path of the .lua file from the ml show output, and use less /path/to/module.lua | grep <extension>. If there is no output, the extension is not present.

Python-based packages¶

It varies between centres how many packages are installed with the base Python packages, how many are installed as separate modules, what prerequsites are required (most centers apart from PDC are the same), and the available version numbers.

Note

HPC2N: Little is installed with the Python module, but instead most of the common Python packages are available as extra modules (SciPy-bundle, Jupyter, mpi4py, matplotlib, tensorflow, PyTorch, Python-bundle-PyPi, …)
LUNARC: Little is installed with the Python module, but instead most of the common Python packages are available as extra modules (SciPy-bundle, Jupyter, mpi4py, matplotlib, tensorflow, PyTorch, Python-bundle-PyPi, …). Anaconda3 bundles more into one module (SciPy, Pandas, Matplotlib, etc), but does not incorporate other modules as well unless they are installed in a custom environment.
C3SE: Little is installed with the Python module, but instead most of the common Python packages are available as extra modules (SciPy-bundle, matplotlib, mpi4py, PyTorch, Python-bundle-PyPi, Jupyter, Horovod)
NSC: Little is installed with the Python module, but instead most of the common Python packages are available as extra modules (SciPy-bundle, matplotlib, mpi4py, PyTorch, Python-bundle-PyPi, Jupyter, …). Most programs on Tetralith also have an extra prerequisite, buildtool-easybuild/4.8.0-hpc<version> that must be loaded before anything else.
PDC: most modules included in SciPy-bundle (NumPy, SciPy, Pandas, etc.) are in cray-python modules, and these are compatible with a couple of the installed versions of matplotlib. Python modules that do not include the cray- prefix have very little installed in them and are not compatible with most other Python-adjacent modules; these are typically intended as bases for users to build their own environments. Most programs on Dardel also have an extra prerequisite, PDC/XX.XX or PDCOLD/XX.XX that must be loaded before anything else.

Example 1: Matplotlib¶

Matplotlib prerequisites vary significantly across HPC centers: some require none, some need one, some need two, and in some cases Matplotlib is only an extension of another module (more info on how to find Matplotlib at different HPC centers here). A good starting point is to view the output of ml spider matplotlib (or ml avail matplotlib on NSC), pick an arbitrary version, and view ml spider matplotlib/<version>.

All of the following code blocks in this example are taken from Cosmos.

$ ml spider matplotlib

----------------------------------------------------------------------------
  matplotlib:
----------------------------------------------------------------------------
    Description:
      matplotlib is a python 2D plotting library which produces publication
      quality figures in a variety of hardcopy formats and interactive
      environments across platforms. matplotlib can be used in python
      scripts, the python and ipython shell, web application servers, and
      six graphical user interface toolkits.

     Versions:
        matplotlib/2.2.5-Python-2.7.18
        matplotlib/3.3.3
        matplotlib/3.4.2
        matplotlib/3.4.3
        matplotlib/3.5.2
        matplotlib/3.7.0
        matplotlib/3.7.2
        matplotlib/3.8.2
        matplotlib/3.9.2

----------------------------------------------------------------------------

If you try the above command at your local HPC center and get a “not found” error, that probably means Matplotlib is an extension of another module (e.g. on Rackham, it is part of the base Python module).

Let us look at matplotlib/3.8.2, for example:

$ ml spider matplotlib/3.8.2

---------------------------------------------------------------------------------
  matplotlib: matplotlib/3.8.2
---------------------------------------------------------------------------------
    Description:
      matplotlib is a python 2D plotting library which produces publication
      quality figures in a variety of hardcopy formats and interactive
      environments across platforms. matplotlib can be used in python scripts,
      the python and ipython shell, web application servers, and six graphical
      user interface toolkits.


    You will need to load all module(s) on any one of the lines below before the "mat
plotlib/3.8.2" module is available to load.

      GCC/13.2.0

So, at Cosmos, only GCC must be loaded before Matplotlib. However, for all practical purposes, Matplotlib is unusable without the tools needed to read in or create the data arrays, so NumPy and/or Pandas are necessary. At most facilities, that means SciPy-bundle is also required.

Note that ml show matplotlib/<version> does not show which Python version is associated with that version of Matplotlib.

If you are constrained by an exact GCC version, you can typically omit the version numbers for everything else unless, e.g. you’re using a very old GCC version that makes both Python 2 and Python 3 modules available. For instance, if you only care that you use GCC/12.3.0, then you can do the following:

$ ml GCC/12.3.0 Python matplotlib SciPy-bundle

However, if you want to move code developed on a personal laptop to the cluster, you will typically only be constrained to a Python version (use python --version to check), or more practically, a range of versions.

Say you built a Matplotlib-based script using Python 3.11.8 on your own laptop. Glob patterns do not work to filter ml spider or ml avail, so it is necessary to view the full list with ml spider Python (ml spider cray-python on Dardel). Here is the output on Cosmos:

$ ml spider Python
---------------------------------------------------------------------------------
  Python:
---------------------------------------------------------------------------------
    Description:
      Python is a programming language that lets you work more quickly and integrate
your systems more effectively.

     Versions:
        Python/2.7.18-bare
        Python/2.7.18
        Python/3.8.6
        Python/3.9.5-bare
        Python/3.9.5
        Python/3.9.6-bare
        Python/3.9.6
        Python/3.10.4-bare
        Python/3.10.4
        Python/3.10.8-bare
        Python/3.10.8
        Python/3.11.3
        Python/3.11.5
        Python/3.12.3
     Other possible modules matches:
        Biopython  GitPython  IPython  Python-bundle  Python-bundle-PyPI
        bx-python  flatbuffers-python  graphviz-python  meson-python  ...

The closest result is Python/3.11.5 (though probably anything from 3.10.x to 3.12.x would work). Let us check what that requires:

$ ml spider Python/3.11.5
---------------------------------------------------------------------------------
  Python: Python/3.11.5
---------------------------------------------------------------------------------
    Description:
      Python is a programming language that lets you work more quickly and
      integrate your systems more effectively.


    You will need to load all module(s) on any one of the lines below before the
    "Python/3.11.5" module is available to load.

      GCCcore/13.2.0

    Help:

      Description
      ===========
      Python is a programming language that lets you work more quickly and integrate 
your systems more effectively.


      More information
      ================
       - Homepage: https://python.org/


      Included extensions
      ===================
      flit_core-3.9.0, packaging-23.2, pip-23.2.1, setuptools-68.2.2, setuptools-
      scm-8.0.4, tomli-2.0.1, typing_extensions-4.8.0, wheel-0.41.2

On this cluster, the base Python module requires GCCcore, but we already saw that Matplotlib requires GCC (whch GCCcore is part of). In fact, nearly every other Python-based module requires GCC, so you may as well use GCC every time.

Each version of Matplotlib and SciPy-bundle should only be associated with one Python version, so now we can load them all at once like this:

ml GCC/13.2.0 Python matplotlib SciPy-bundle

And if we check what was loaded, we get this:

$ ml

Currently Loaded Modules:
  1) SoftwareTree/Milan         (S)  26) expat/2.5.0
  2) GCCcore/13.2.0                  27) util-linux/2.39
  3) zlib/1.2.13                     28) fontconfig/2.14.2
  4) binutils/2.40                   29) xorg-macros/1.20.0
  5) GCC/13.2.0                      30) libpciaccess/0.17
  6) bzip2/1.0.8                     31) X11/20231019
  7) ncurses/6.4                     32) Tk/8.6.13
  8) libreadline/8.2                 33) Tkinter/3.11.5
  9) Tcl/8.6.13                      34) NASM/2.16.01
 10) SQLite/3.43.1                   35) libjpeg-turbo/3.0.1
 11) XZ/5.4.4                        36) jbigkit/2.1
 12) libffi/3.4.4                    37) gzip/1.13
 13) OpenSSL/1.1                     38) lz4/1.9.4
 14) Python/3.11.5                   39) zstd/1.5.5
 15) OpenBLAS/0.3.24                 40) libdeflate/1.19
 16) FlexiBLAS/3.3.1                 41) LibTIFF/4.6.0
 17) FFTW/3.3.10                     42) giflib/5.2.1
 18) cffi/1.15.1                     43) libwebp/1.3.2
 19) cryptography/41.0.5             44) OpenJPEG/2.5.0
 20) virtualenv/20.24.6              45) LittleCMS/2.15
 21) Python-bundle-PyPI/2023.10      46) Pillow/10.2.0
 22) pybind11/2.11.1                 47) Qhull/2020.2
 23) libpng/1.6.40                   48) matplotlib/3.8.2
 24) Brotli/1.1.0                    49) SciPy-bundle/2023.11
 25) freetype/2.13.2

  Where:
   S:  Module is Sticky, requires --force to unload or purge

If you are comfortable editing code in a basic text editor and running at the command-line, the above is all you need. For more information on choosing and loading IDEs to work with Matplotlib graphics interactively, we refer readers to this documentation from the Python for HPC course.

Example 2: MPI4Py¶

Here is an example of what ml spider MPI4Py would output (note that ml spider and ml avail are only case-sensitive when the module version number is included):

$ ml spider MPI4Py

---------------------------------------------------------------------------------
  mpi4py:
---------------------------------------------------------------------------------
    Description:
      MPI for Python (mpi4py) provides bindings of the Message Passing Interface
      (MPI) standard for the Python programming language, allowing any Python
      program to exploit multiple processors.

     Versions:
        mpi4py/3.1.4
        mpi4py/3.1.5
        mpi4py/4.0.1

---------------------------------------------------------------------------------
  For detailed information about a specific "mpi4py" package (including how to load t
he modules) use the module's full name.
  Note that names that have a trailing (E) are extensions provided by other modules.
  For example:

     $ module spider mpi4py/4.0.1
---------------------------------------------------------------------------------

Note

UPPMAX does not currently have mpi4py installed, per their documentation page on parallel programming in Python.

Here again, you will typically only know which Python version you need and/or which GCC version to load, not which version of mpi4py you need. However, since the previous example showed how to work from a known Python version, this example will show a workflow based on working backwards from the latest version of a module.

In the above output from Cosmos, the latest version is mpi4py/4.0.1, so here is how to view its dependencies:

$ ml spider mpi4py/4.0.1

---------------------------------------------------------------------------------
  mpi4py: mpi4py/4.0.1
---------------------------------------------------------------------------------
    Description:
      MPI for Python (mpi4py) provides bindings of the Message Passing Interface
      (MPI) standard for the Python programming language, allowing any Python
      program to exploit multiple processors.


    You will need to load all module(s) on any one of the lines below before the "mpi
4py/4.0.1" module is available to load.

      GCC/13.3.0  OpenMPI/5.0.3

    Help:

      Description
      ===========
      MPI for Python (mpi4py) provides bindings of the Message Passing Interface (MPI
) standard for
       the Python programming language, allowing any Python program to exploit multip
le processors.


      More information
      ================
       - Homepage: https://github.com/mpi4py/mpi4py


      Included extensions
      ===================
      mpi4py-4.0.1

Now we can just copy and paste the dependencies, and then load the main module, like so:

$ ml GCC/13.3.0  OpenMPI/5.0.3
$ ml mpi4py/4.0.1

Or you could combine them into one line, and omit the mpi4py version since the prerequisites make only 1 version available, as in ml GCC/13.3.0 OpenMPI/5.0.3 mpi4py.

Now, if we check what was loaded, we see the following:

$ ml

Currently Loaded Modules:
  1) SoftwareTree/Milan  (S)  10) hwloc/2.10.0      19) bzip2/1.0.8
  2) GCCcore/13.3.0           11) OpenSSL/3         20) ncurses/6.5
  3) zlib/1.3.1               12) libevent/2.1.12   21) libreadline/8.2
  4) binutils/2.42            13) UCX/1.16.0        22) Tcl/8.6.14
  5) GCC/13.3.0               14) libfabric/1.21.0  23) SQLite/3.45.3
  6) numactl/2.0.18           15) PMIx/5.0.2        24) libffi/3.4.5
  7) XZ/5.4.5                 16) PRRTE/3.0.5       25) Python/3.12.3
  8) libxml2/2.12.7           17) UCC/1.3.0         26) mpi4py/4.0.1
  9) libpciaccess/0.18.1      18) OpenMPI/5.0.3

  Where:
   S:  Module is Sticky, requires --force to unload or purge

As you can see, on the cluster where this code was run, Python was loaded automatically.

However, note that no BLAS or LAPACK libraries are loaded, which might be unexpected given the types of problems MPI is typically used for. These libraries are available through either a foss toolchain or a SciPy-bundle, and in practice, you will almost always load one or both of those when using mpi4py.

Example 3: Check if a Python extension is loaded¶

Extensions can be hard to find without knowing what includes them, but it is easy to check if modules that are already loaded added the extension silently. If you cannot find a package you want with ml avail, ml spider, or ml show <module>, you should also check pip list and grep for the package after loading the rest of your modules.

For example, psutil is part of Python-bundle-PyPI, which is silently loaded with any SciPy-bundle. Here is the easiest way to find psutil:

$ pip list | grep psutil
psutil                            5.9.5

[notice] A new release of pip is available: 23.1.2 -> 25.1.1
[notice] To update, run: pip install --upgrade pip

The pip list | grep approach is also helpful if you want to see the version of a package without having to open a Python interpreter.

Tip

The same list (and grep) approach works for Anaconda3. The only difference is that you use conda list instead of pip list. The Anaconda3 module file does not list the included extensions, so conda list | grep <package> is also the only way to see if a package is included without starting up a Python command line interface.

R-based packages¶

At most HPC centers, the base R module usually contains relatively few extensions. Most of the popular packages are in additional bundles like R-bundle-CRAN and R-bundle-Bioconductor. Most HPC centers have prerequisites for R, but at a few, like Alvis, R can be loaded directly. Always check the prerequisites with ml spider or ml avail.

Note

HPC2N: Little is installed with the basic R module, but most common packages are available as extensions of R-bundle-CRAN, R-bundle-CRAN-extra, or R-bundle-Bioconductor. RStudio is a separate module and only runs on the login nodes via Thinlinc, so it should be used sparingly.
LUNARC: Little is installed with the basic R module, but most common packages are available as extensions of R-bundle-CRAN or R-bundle-Bioconductor. RStudio is also a separate module, and is available as an On-Demand application that automatically loads R and various bundles at start-up.
UPPMAX (Rackham) and C3SE (Alvis): R can be loaded directly, but has few installed packages. More common modules are available as extensions with the R_packages module. RStudio is also a separate module.
NSC (Tetralith): R can be loaded directly, but contains few installed packages, and there are no bundles to provide more. Users are typically expected to install their own extension libraries. RStudio is included in the base R module, however.
PDC (Dardel): Like most programs on Dardel, R also has the prerequisite PDC/XX.XX or PDCOLD/XX.XX, but the compiler and MPI library are chosen for you. There are about 250 packages available in the basic R module, and there are no additional bundles to provide more packages. Users are typically expected to install their own extension libraries.

Important

Most facilities have only built a handful of R releases, so many of the dependent modules are adapted for multiple versions. The version of such any dependent package should always be specified to ensure reproducibility. If the version number is omitted, the latest will be loaded by default, and that version may change without warning.

Example 1: Bioconductor¶

Many R-packages conveniently specify the version of R they are compatible with in the module name. One example of this is Bioconductor.

$ ml spider bioconductor

---------------------------------------------------------------------------------
  R-bundle-Bioconductor:
---------------------------------------------------------------------------------
    Description:
      Bioconductor provides tools for the analysis and coprehension of
      high-throughput genomic data.

     Versions:
        R-bundle-Bioconductor/3.15-R-4.2.1
        R-bundle-Bioconductor/3.18-R-4.3.2
        R-bundle-Bioconductor/3.18-R-4.4.1
        R-bundle-Bioconductor/3.19-R-4.4.1

---------------------------------------------------------------------------------
  For detailed information about a specific "R-bundle-Bioconductor" package (includin
g how to load the modules) use the module's full name.
  Note that names that have a trailing (E) are extensions provided by other modules.
  For example:

     $ module spider R-bundle-Bioconductor/3.19-R-4.4.1
---------------------------------------------------------------------------------

Notice that in this case, there are 2 versions of the Bioconductor bundle associated with R/4.4.1, and that there are 2 versions of R associated with R-bundle-Bioconductor/3.18. Do not rely on the prerequisites to set which version of R-bundle-Bioconductor gets loaded.

To check the prerequisites with ml spider, the specific version number must be included anyway, for all software.

$ ml spider R-bundle-Bioconductor/3.18-R-4.4.1

---------------------------------------------------------------------------------
  R-bundle-Bioconductor: R-bundle-Bioconductor/3.18-R-4.4.1
---------------------------------------------------------------------------------
    Description:
      Bioconductor provides tools for the analysis and coprehension of
      high-throughput genomic data.


    You will need to load all module(s) on any one of the lines below before the "R-b
undle-Bioconductor/3.18-R-4.4.1" module is available to load.

      GCC/12.3.0  OpenMPI/4.1.5

    Help:

      Description
      ===========
      Bioconductor provides tools for the analysis and coprehension
       of high-throughput genomic data.


      More information
      ================
       - Homepage: https://bioconductor.org


      Included extensions
      ===================
      affxparser-1.74.0, affy-1.80.0, affycoretools-1.74.0, affyio-1.72.0,
      AgiMicroRna-2.52.0, agricolae-1.3-7, ALDEx2-1.34.0, ALL-1.44.0, ANCOMBC-2.4.0,
      annaffy-1.74.0, annotate-1.80.0, AnnotationDbi-1.64.1,
      AnnotationFilter-1.26.0, AnnotationForge-1.44.0, AnnotationHub-3.10.0,
      anytime-0.3.9, aroma.affymetrix-3.2.1, aroma.apd-0.7.0, aroma.core-3.3.0,
      aroma.light-3.32.0, ash-1.0-15, ATACseqQC-1.26.0, AUCell-1.24.0,
      aws.s3-0.3.21, aws.signature-0.6.0, babelgene-22.9, ballgown-2.34.0,
      basilisk-1.14.2, basilisk.utils-1.14.1, batchelor-1.18.1, baySeq-2.36.0,
      beachmat-2.18.0, BH-1.84.0-0, Biobase-2.62.0, BiocBaseUtils-1.4.0, ...

The list of extensions is too long to copy here, but some popular extensions included in this module are: DeSeq2, GenomeInfoDb, MStats, Seurat, Rsamtools, and more.

The above prerequisites and the main package can be loaded either one at a time or all at once with,

$ ml GCC/12.3.0  OpenMPI/4.1.5  R-bundle-Bioconductor/3.18-R-4.4.1

In this case, R-bundle-Bioconductor loads the version of R that it is based on automatically (along with about 130 other modules!). That is not the case for all R-bundles at all HPC centers, so pay attention to the prerequisites.

Matlab¶

At most HPC centers, Matlab can be loaded directly, but a couple of centers require a basic software tree as a prerequisite. On Dardel (PDC), all Matlab versions, like nearly every other module, have a module called something like PDC/xx.xx or PDCOLD/xx.xx as a prerequisite. The example further down in this section shows how finding and determining the prerequisites for Matlab works at another such facility.

Capitalization and other naming conventions also vary between HPC centers; for more information, refer to this section of the R, Matlab, and Julia for HPC course.

All Add-Ons and Toolboxes should be available through the Matlab GUI.

Important

The Matlab GUI is prone to hogging resources if not launched carefully, which makes it risky to run on a login node. In general, the GUI should only be run via either Desktop On Demand or after booking interactive allocations on compute nodes with salloc or interactive. For more particulars on running Matlab, see the relevant page of the R, Matlab, and Julia for HPC course materials.

Example: Matlab on Tetralith (NSC)¶

At most centers, where modules are hidden if prerequisites are not loaded, it is better to use ml spider to see what versions are available before accounting for preconditions. At centers where all modules are searchable without loading prerequisites, it may be better to use ml avail to avoid listing modules that only exist as extensions or aliases of other modules, as in the case of Tetralith at NSC:

$ ml avail matlab

--------------------- /software/sse2/tetralith_el9/modules ---------------------
   MATLAB/recommendation (D)    MATLAB/2023b-bdist
   MATLAB/2023a-bdist           MATLAB/2024a-hpc1-bdist

  Where:
   D:  Default Module

Once you have chosen a specific version, use ml spider to check if there are prerequisites, like so:

$ ml spider MATLAB/2024a-hpc1-bdist

The full output was too verbose to reprint in full here, but the one important line reads:

    This module can be loaded directly: module load MATLAB/2024a-hpc1-bdist

The command after the colon (:) can be copied, pasted, and entered directly into the bash prompt to load the module, or you can type the short version as follows:

$ ml MATLAB/2024a-hpc1-bdist

Specialized Applications¶

For most specialized packages (Amber, GROMACS, Nextflow, VASP, etc), unless there is reason to believe it is included in a larger package or you include a spurious non-alphanumeric character, ml spider will tell you whether it is installed or not. If the full name of a module includes CUDA, then the relevant CUDA version will typically be loaded automatically, without the need to choose a CUDA-containing toolchain.

Important

Some specialized modules (e.g., Abaqus, Gaussian, and VASP) are license-restricted. If you run ml spider on a specific version of licensed software, the description may (as with VASP) or may not (as with Gaussian) specify that a license is required. It is encumbant on users to determine the licensing requirements of specialized software packages.

Example: OpenFOAM¶

As usual, we start by checking the versions available with ml spider OpenFOAM (or ml avail OpenFOAM at a facility like NSC where all modules are visible regardless of the presence of prerequisites).

An example output might look like this (from Cosmos):

$ ml spider OpenFOAM

--------------------------------------------------------------------------------------
  OpenFOAM:
--------------------------------------------------------------------------------------
    Description:
      OpenFOAM is a free, open source CFD software package. OpenFOAM has an extensive
      range of features to solve anything from complex fluid flows involving chemical
      reactions, turbulence and heat transfer, to solid dynamics and
      electromagnetics.

     Versions:
        OpenFOAM/v2112
        OpenFOAM/v2206
        OpenFOAM/v2306
        OpenFOAM/v2406
        OpenFOAM/7-20200508
        OpenFOAM/9
        OpenFOAM/10
        OpenFOAM/11

--------------------------------------------------------------------------------------
  For detailed information about a specific "OpenFOAM" package (including how to load the 
modules) use the module's full name.
  Note that names that have a trailing (E) are extensions provided by other modules.
  For example:

     $ module spider OpenFOAM/11
--------------------------------------------------------------------------------------

Let us look at a recent version:

$ ml spider OpenFOAM/11
--------------------------------------------------------------------------------------
  OpenFOAM: OpenFOAM/11
--------------------------------------------------------------------------------------
    Description:
      OpenFOAM is a free, open source CFD software package. OpenFOAM has an extensive
      range of features to solve anything from complex fluid flows involving chemical
      reactions, turbulence and heat transfer, to solid dynamics and
      electromagnetics.


    You will need to load all module(s) on any one of the lines below before the "OpenFOAM
/11" module is available to load.

      GCC/11.3.0  OpenMPI/4.1.4

    Help:
      Description
      ===========
      OpenFOAM is a free, open source CFD software package.
       OpenFOAM has an extensive range of features to solve
       anything from complex fluid flows involving chemical
       reactions, turbulence and heat transfer, to solid dynamics
       and electromagnetics.


      More information
      ================
       - Homepage: https://www.openfoam.org/

On this system, GCC and OpenMPI must be loaded first, but this is not true for every system. Indeed, some versions on some systems (e.g. at NSC) load and use the compilers, MPI libraries, and mathematics libraries from Intel toolchains.

Now we can load everything all at once like so:

$ ml GCC/11.3.0  OpenMPI/4.1.4 OpenFOAM/11

or load each one at a time. The above command loads almost 90 modules, including several Python packages and visualization libraries, all of which can be viewed by entering ml.