SIESTA for HPC
SIESTA can take advantage of high-performance computing (HPC) infrastructure to perform large-scale simulations efficiently. We provide here a few resources to help you install and run SIESTA on HPC systems.
Build tools
The build tools repository contains CMake, easybuild and spack recipes. Even if your own HPC system is not considered here, you can easily adapt the scripts to your own needs.
You can find the repository at: https://gitlab.com/siesta-project/ecosystem/build-tools
Once you found your desired platform, you can compile SIESTA by using CMake recipes in the following manner:
cmake -S . -B _build -C CMAKE_RECIPE.cmake [more options]
Please refer to the Installing SIESTA__<how-to-build-siesta>__ section for more information on how to build SIESTA from source.
Recommended EuroHCP modules
The modules will depend on the HPC system. Be aware that even for a given supercomputer, the modules may change over time. We include here the modules used for the most recent SIESTA benchmarks.
Guide for setting up your Fugaku account and compiling SIESTA
Fugaku Account Setup
Guide for configuring your Fugaku account, understanding the file system structure, and preparing your environment for compiling scientific software.
SIESTA Compilation on Fugaku
SIESTA can be compiled on Fugaku using two different compiler toolchains. Choose the approach that best suits your needs:
GNU compilers with Fujitsu MPI: Uses GCC/GFortran compilers with Fujitsu’s MPI implementation.
Native Fujitsu compilers with MPI wrappers: Uses Fujitsu’s fcc/frt compilers with their native MPI wrappers (
mpifcc/mpifrt).
Benchmarking
SIESTA benchmarks have been performed under the MaX project umbrella. To reproduce the benchmarks, you can use the JUBE interface found at: https://gitlab.com/max-centre/JUBE4MaX
The results for past benchmarks are available at: https://gitlab.com/max-centre/benchmarks-max3