EPOCpp

Published: Dec 19, 2023 by Liam Pattinson

EPOCH is a particle-in-cell (PIC) code used to simulate laser-plasma interactions, with research applications including inertial confinement fusion, laser wakefield acceleration, and quantum electrodynamics. However, EPOCH is written in Fortran, which will make it difficult to optimise for next-generation HPC architectures which will rely heavily on GPU offloading. As a result, there is an effort underway to reimplement EPOCH in C++ using performance-portable frameworks.

Our work on this project was aimed at implementing a CMake build system, testing/documentation frameworks, and continuous integration workflows. EPOCpp is planned to be usable either as a standalone executable or via Python bindings, so the CMake system had to be quite flexible. We chose to use scikit-build-core to enable building and installing the project as a Python library, which greatly simplified the interface between the compiled parts of the project defined in CMakeLists.txt and the Python parts in pyproject.toml.

Tests for the C++ portion of the code were implemented using GoogleTest, while the Python tests were implemented using pytest. The documentation was implemented using Doxygen in C++ and Sphinx in Python, and the two were bridged using the Sphinx extension Breathe to provide a single consistent set of docs across the two languages.

Finally, a series of GitHub workflows were implemented to automate building, testing, linting, and formatting. Linting was performed using ruff in Python and clang-tidy in C++, while formatting was performed using black in Python and clang-format in C++.

The most difficult aspect of this project was learning to use such a large range of tools and ensuring all of them could function across the language barrier. It was also challenging to write the CMakeLists.txt files in such a way that users and developers could have sufficient control over the build process while also permitting multiple build strategies.