Installing LAMMPS

Every LAMMPS release released after January 23th, 2025 contains the functionalities needed for descriptor calculation with MALA. If you are working on a high perfomance computing cluster, check if you have a LAMMPS version installed that may already be used with MALA. Note that in order to use GPU acceleration, you need access to a GPU version (i.e., Kokkos enabled version) of LAMMPS. In the following, a brief overview of how to build LAMMPS is provided. The MALA team recommends to build LAMMPS with cmake. For a full overview of how to build LAMMPS, please refer to the official instructions.

Prerequisites

Make sure you have cmake and suitable compilers (e.g. gcc) available. LAMMPS can be build with GPU support (recommended for production settings!), in which case CUDA is required to be installed on your machine.

Build LAMMPS

  • Checkout LAMMPS at https://github.com/lammps/lammps

  • Inside the LAMMPS folder create a build folder (named, e.g., build)

  • In the build folder, configure your cmake build: cmake ../cmake -D OPTION1 -D OPTION2 ...; Options for a typical LAMMPS build for usage with MALA:

    • BUILD_SHARED_LIBS=yes: Necessary to link the LAMMPS library to Python

    • PKG_ML-SNAP=yes: Enables the calculation of SNAP descriptors with LAMMPS

    • BUILD_MPI=yes: Enables MPI aware calculations; set this option if plan to use MPI and have an MPI aware compiler loaded

    • PKG_KOKKOS=yes: Enables the Kokkos package which is needed to calculate bispectrum descriptors on a GPU. Without this option, bispectrum descriptor calculation will solely be performed on CPU. If you want to use a GPU via Kokkos (recommended for production settings!), you will have to set further options:

      • Kokkos_ENABLE_CUDA=yes: Tells Kokkos to use CUDA

      • Kokkos_ARCH_HOSTARCH=???: Your CPU architecture (see Kokkos instructions)

      • Kokkos_ARCH_GPUARCH=???: Your GPU architecture (see see Kokkos instructions)

      • CMAKE_CXX_COMPILER=???: Path to the nvcc_wrapper executable shipped with the LAMMPS code, should be at /your/path/to/lammps/lib/kokkos/bin/nvcc_wrapper

      For example, this configures the LAMMPS cmake build with Kokkos support for an Intel Haswell CPU and an Nvidia Volta GPU, with MPI support:

      cmake ../cmake -D PKG_KOKKOS=yes -D BUILD_MPI=yes -D PKG_ML-SNAP=yes -D Kokkos_ENABLE_CUDA=yes -D Kokkos_ARCH_HSW=yes -D Kokkos_ARCH_VOLTA70=yes -D CMAKE_CXX_COMPILER=/path/to/lammps/lib/kokkos/bin/nvcc_wrapper -D BUILD_SHARED_LIBS=yes

Note

When using a GPU by setting parameters.use_gpu = True, you need to have a GPU version of LAMMPS installed. See Predictions on GPUs for details.

  • Build the library and executable with cmake --build . (Add --parallel=8 for a faster build)

Installing the Python extension

  • After successfully having built LAMMPS, Change into the python folder of the LAMMPS code

  • python3 install.py -p lammps -l ../<build_folder>/liblammps.so -v ../src/version.h, where <build_folder> is whatever folder you performed the cmake build in

  • Note: The Python installation process may give an shutil.SameFileError after successful installation; this is LAMMPS related and can be ignored here.