Software environment
A rich user software environment is available on MeluXina and can be used through environment modules. It comprises:
-
tools for HPC, HPDA and AI application development
- compilers & programming languages
- common use libraries for math & data
- tools for performance engineering
-
scientific computing applications, libraries and frameworks
- general-purpose HPC, HPDA or AI
- or domain-specific
The MeluXina User Software Environment (MUSE)
The EasyBuild system is used on MeluXina to deploy software stacks, with applications and libraries provided in several revisions: compiled with different compilers, MPI suites, accelerated and non-accelerated versions. EasyBuild automatically generated software modules that enables the use of the different application revisions, and dependencies on particular support libraries.
The naming scheme used for the modules follows the application/version schema, e.g. NAMD/2.14-foss-2022a-ucx-CUDA-11.7.0
which shows that the NAMD application is available in version 2.14 built with the foss toolchain in
the 2022a release, and with CUDA support.
For a deeper overview on versioning and toolchains see the EasyBuild documentation specific to Common toolchains and Available toolchains.
We release a new software stack at least once a year. The software stacks can be either: Production (current), Maintained (previous, not updated with new tools/versions), Retired (not maintained and should not be used any more) or Staging (preproduction envioronment for testing purposes).
MUSE release | Based on EasyBuild toolchain(s) | Status | Path |
---|---|---|---|
2022.1 | 2022a | Next Production and default(!) from 2023-01-15 | /apps/USE/easybuild/release/2022.1/modules/all/ |
2021.3 | 2021a | Current Production and default until 2023-01-14, maintained afterwards | /apps/USE/easybuild/release/2021.3/modules/all |
2021.5 | 2021a | Production, optimized version of 2021.3 stack and tracking it | /apps/USE/easybuild/release/2021.5/modules/all/ |
2022.1 | 2022a | Staging | /apps/USE/easybuild/staging/2022.1/modules/all/ |
Using different MUSE releases
To load modules from a software release other than the default Production stack, you can:
- Use the dedicated software environment modules in the 'env/' branch, e.g.
module load env/staging/2022.1
- Or directly configure LMod to use the appropriate modules path, e.g.:
module use /apps/USE/easybuild/staging/2022.1/modules/all/
The following table lists highlighted HPC applications, libraries and support tools available in the latest release of the MeluXina User Software Environment.
Software | Available versions | Has GPU accelerated version? |
---|---|---|
AMD-uProf | 3.6.449 | |
AOCC | 3.2.0-GCCcore-11.3.0 | |
Advisor | 2022.1.0 | |
Arm-Forge | 22.0.4-GCC-11.3.0 | |
Autoconf | 2.69-GCCcore-11.3.0, 2.71-GCCcore-11.3.0 | |
aws-cli | 2.7.1 | |
BLIS | 0.9.0-GCC-11.3.0 | |
Bazel | 5.1.1-GCCcore-11.3.0 | |
Blender | 3.2.2-linux-x64 | |
Boost | 1.79.0-GCC-11.3.0 | |
CGAL | 4.14.3-gompi-2022a, 5.2-gompi-2022a | |
CMake | 3.23.1-GCCcore-11.3.0 | |
CUDA | 11.7.0 | |
Clang | 13.0.1-GCCcore-11.3.0-CUDA-11.7.0, 13.0.1-GCCcore-11.3.0 | Yes |
cuDNN | 8.4.1.50-CUDA-11.7.0 | Yes |
Dakota | 6.16.0-foss-2022a | |
Doxygen | 1.9.4-GCCcore-11.3.0 | |
DualSPHysics | 5.0.175-foss-2022a, 5.0.175-GCC-11.3.0-CUDA-11.7.0, 5.0.175-GCC-11.3.0 | Yes |
[Eigen]Eigen | 3.4.0-GCCcore-11.3.0 | |
FFTW | 3.3.10-GCC-11.3.0 | |
FFTW.MPI | 3.3.10-gompi-2022a | |
FlexiBLAS | 3.2.0-GCC-11.3.0 | |
foss | 2022a | |
GCC | 11.3.0 | |
GCCcore | 11.3.0 | |
GDAL | 3.5.1-foss-2022a | |
GDB | 11.1-GCCcore-11.3.0 | |
GROMACS | 2022.3-foss-2022a-CUDA-11.7.0, 2022.3-foss-2022a | Yes |
gnuplot | 5.4.2-GCCcore-11.3.0, 5.4.4-GCCcore-11.3.0 | |
Go | 1.19.1 | |
gompi | 2022a-prte, 2022a | |
gperf | 3.1-GCCcore-11.3.0 | |
gperftools | 2.10-GCCcore-11.3.0 | |
h5py | 3.7.0-foss-2022a-CUDA-11.7.0, 3.7.0-foss-2022a | Yes |
HDF5 | 1.12.2-gompi-2022a-serial, 1.12.2-gompi-2022a | |
HPCX | 2.12-GCCcore-11.3.0 | |
Horovod | 0.26.0-foss-2022a-CUDA-11.7.0-PyTorch-1.12.0 | Yes |
hwloc | 2.7.1-GCCcore-11.3.0 | |
iOR | 3.3.0-gompi-2022a | |
ifpgasdk | 20.4 | |
iimpi | 2022a | |
imkl | 2022.1.0 | |
imkl-FFTW | 2022.1.0-iimpi-2022a | |
impi | 2021.6.0-intel-compilers-2022.1.0 | |
intel | 2022a | |
intel-compilers | 2022.1.0 | |
itac | 2021.6.0 | |
Java | 11.0.16 | |
Julia | 1.8.2-linux-x86_64 | |
JupyterHub | 2.3.1-GCCcore-11.3.0 | |
JupyterLab | 3.2.8-GCCcore-11.3.0 | |
Keras | 2.9.0-foss-2022a | |
Kokkos | 3.6.01-GCC-11.3.0 | |
LAMMPS | 23Jun2022-foss-2022a-kokkos-CUDA-11.7.0, 23Jun2022-foss-2022a-kokkos | Yes |
MCR | R2022a | |
METIS | 5.1.0-GCCcore-11.3.0 | |
mpifileutils | 0.11.1-foss-2022a | |
NAMD | 2.14-foss-2022a-mpi, 2.14-foss-2022a-ucx-CUDA-11.7.0 | Yes |
NCCL | 2.12.12-GCCcore-11.3.0-CUDA-11.7.0 | Yes |
NVHPC | 22.7-CUDA-11.7.0, 22.7 | Yes |
NVSHMEM | 2.6.0-gompi-2022a-CUDA-11.7.0 | Yes |
netCDF | 4.9.0-gompi-2022a | |
Ninja | 1.10.2-GCCcore-11.3.0 | |
numactl | 2.0.14-GCCcore-11.3.0 | |
nvompic | 2022a | |
OSU-Micro-Benchmarks | 6.0-gompi-2022a-CUDA-11.7.0, 6.0-gompi-2022a-prte, 6.0-gompi-2022a, 6.0-iimpi-2022a | Yes |
OpenBLAS | 0.3.20-GCC-11.3.0 | |
OpenCV | 4.6.0-foss-2022a-contrib, 4.6.0-foss-2022a-CUDA-11.7.0-contrib | Yes |
OpenFOAM | v2206-foss-2022a, 9-foss-2022a | |
OpenMPI | 4.1.4-GCC-11.3.0-prte, 4.1.4-GCC-11.3.0, 4.1.4-NVHPC-22.7-CUDA-11.7.0, 4.1.4-NVHPC-22.7 | Yes |
PAPI | 6.0.0.1-GCCcore-11.3.0 | |
PETSc | 3.17.4-foss-2022a | |
PLUMED | 2.8.0-foss-2022a | |
PMIx | 4.2.2-GCCcore-11.3.0 | |
ParaView | 5.10.1-foss-2022a-mpi | |
psmpi | 5.6.0-1-GCC-11.3.0-CUDA-11.7.0, 5.6.0-1-GCC-11.3.0 | Yes |
PyTorch | 1.12.0-foss-2022a-CUDA-11.7.0, 1.12.0-foss-2022a | Yes |
Python | 2.7.18-GCCcore-11.3.0-bare, 3.10.4-GCCcore-11.3.0-bare, 3.10.4-GCCcore-11.3.0 | |
QMCPACK | 3.14.0-foss-2022a | |
QUDA | 1.1.0-foss-2022a-CUDA-11.7.0 | Yes |
Qsimcirq | 0.14.0-foss-2022a | |
QuantumESPRESSO | 7.1-foss-2022a, 7.1-nvompic-2022a-NVHPC-22.7 | |
R | 4.2.1-foss-2022a-bare, 4.2.1-foss-2022a | |
ReFrame | 3.12.0 | |
Rust | 1.60.0-GCCcore-11.3.0 | |
s3cmd | 2.2.0 | |
SCOTCH | 7.0.1-gompi-2022a | |
ScaLAPACK | 2.2.0-gompi-2022a-fb | |
Scalasca | 2.6-gompi-2022a-CUDA-11.7.0, 2.6-gompi-2022a | Yes |
SciPy-bundle | 2022.05-foss-2022a | |
Scikit-image | 0.19.3-foss-2022a | |
Scikit-learn | 1.1.2-foss-2022a | |
Score-P | 7.1-gompi-2022a-CUDA-11.7.0, 7.1-gompi-2022a | Yes |
Singularity-CE | 3.10.2-GCCcore-11.3.0 | |
Spark | 3.3.0-foss-2022a-CUDA-11.7.0, 3.3.0-foss-2022a | Yes |
TBB | 2021.5.0-GCCcore-11.3.0 | |
TensorFlow | 2.9.1-foss-2022a | |
Torchtext | 0.12.0-foss-2022a | |
Torchvision | 0.13.1-foss-2022a-CUDA-11.7.0, 0.13.1-foss-2022a | Yes |
UCX | 1.13.1-GCCcore-11.3.0 | |
UCX-CUDA | 1.13.1-GCCcore-11.3.0-CUDA-11.7.0 | Yes |
VTune | 2022.3.0 | |
Valgrind | 3.19.0-gompi-2022a | |
Voro++ | 0.4.6-GCCcore-11.3.0 | |
520nmx | 20.4 |
Requesting new software and features
You may request the installation of new applications or features through the servicedesk.lxp.lu. Depending on a project's requirements, new software may be installed in the corresponding project directory or in the global software stack. All new tools are first deployed in the Staging area for testing.
Environment modules
The environment modules system simplifies the use of applications and supporting libraries that may come in different versions and revisions. The software modules provide a way to easily switch between e.g. multiple revisions of the same application, where one revision may provide a set of functionality not available in another revision.
The core command by which software modules can be listed, loaded (activated),
and unloaded (deactivated) is module
, followed by an appropriate command.
The modules work by setting specific environment variables needed for the
respective software program when the software module corresponding to the
application is loaded. Often, this is simply adding the program to the $PATH
variable, but software containing libraries and headers will also set $LD_LIBRARY_PATH
.
Any other variable that the software may need can be set and so the contents of the
modules can be fairly simple or complex.
There are several advantages to using software modules to set up your environment, especially on a supercomputer:
- ease of use
- ability to revert to your previous environment
- ability to easily switch your environment to try different versions of a program
- e.g. when single and double-precision versions of the program exist
- e.g. when a program has been compiled with different features that cannot coexist in a single build of the program
- e.g. when a program has been compiled with different compilers or MPI suites
Using modules
The module
command is only available on compute nodes and not on the login nodes.
The applications provided through the software modules system must be used only on compute nodes.
MeluXina uses the Lmod software modules system,
the table below summarizes the most common module
commands:
Module Command | Description |
---|---|
module avail | List/browse available modules |
module list | Show modules currently loaded (active) |
module load module_name | Load a specific module (may load additional modules as dependencies) |
module unload module_name | Unload a loaded module (does not unload modules activated as dependencies) |
module swap module_name1 module_name2 | Unload module_name1 and load module_name2 |
module purge | Unload all loaded modules |
module reset | Reset loaded modules to system defaults |
module show module_name | Display the contents of a selected module |
module spider | List all modules and the short description of each |
module spider package | Display the description and various versions available of an application |
module use path | Add path to the MODULEPATH search path |
Using environment modules
- Finding applications: module avail
To lists all available (loadable) modules and module groups. With the information of these two commands:
module avail
Output
---------------------------------
AOCC/3.0.0-GCCcore-10.2.0 GCC/10.2.0 ...
Autoconf/2.69-GCCcore-10.2.0 GCCcore/10.2.0 ...
Automake/1.16.2-GCCcore-10.2.0 GDRCopy/2.1-GCCcore-10.2.0-CUDA-11.2.2 ...
Autotools/20200321-GCCcore-10.2.0 GMP/6.2.0-GCCcore-10.2.0 ...
BLIS/0.8.1-GCCcore-10.2.0 Go/1.16.3 ...
BLIS/3.0-GCCcore-10.2.0 (D) HPCX/2.8.1 ...
Bison/3.7.1-GCCcore-10.2.0 M4/1.4.18-GCCcore-10.2.0 ...
Bison/3.7.1 M4/1.4.18 (D) ...
Bison/3.7.4 (D) NVHPC/21.3 ...
CMake/3.18.4-GCCcore-10.2.0 OSU-Micro-Benchmarks/5.7-gompic-2020b ...
CUDA/11.2.2-GCC-10.2.0 OpenBLAS/0.3.12-GCC-10.2.0 ...
CUDAcore/11.2.2 OpenMPI/4.0.5-GCC-10.2.0-with-orte ...
Check/0.15.2-GCCcore-10.2.0 OpenMPI/4.0.5-GCC-10.2.0 ...
DB/18.1.40-GCCcore-10.2.0 OpenMPI/4.0.5-gcccuda-2020b (D) ...
FFTW/3.3.8-gompi-2020b PMIx/3.2.3-GCCcore-10.2.0 ...
FFTW/3.3.8-gompic-2020b (D)
- Listing loaded application profiles: module list
To get a list of all currently loaded modules:
module list
- Loading or unloading application profiles: module load/unload
To load a specific module
module load <module_name>
Default version
In case of multiple software versions, one version will always be defined as the default version, and can be identified by its (D) mark in the module avail
output.
When loading a software module, if the version is not specified, the default is loaded (e.g. module load FFTW
will activate FFTW/3.3.8-gompic-2020b
if this is the default).
Fully specifying the software module (name+version) should always be preferred to ensure that the correct version is being activated.
Unloading an environment module will undo the changes that module made to the environment, restoring any variables set to their previous values. To unload a specific module you can use the following:
module unload <module_name>
- Unloading all profiles: module purge
module purge
- Switching profiles: module switch
To swap a specific module for another one (especially useful to switch between different versions of the same program) use the following:
module switch <old_module_name> <new_module_name>