We are pleased to share our publication "Towards a Platform-Portable Linear Algebra Backend for OpenFOAM", which has been published in the journal Meccanica. The work is authored by Gregor Olenik, Marcel Koch, Ziad Boutanios, and Hartwig Anzt.

This paper presents the development of the OpenFOAM Ginkgo Layer (OGL), a plugin designed to leverage the Ginkgo library for GPU-accelerated linear algebra computations in OpenFOAM. The study emphasizes platform portability and evaluates performance across different hardware systems, including Intel, AMD, and NVIDIA GPUs.

Notable contributions include methods to reduce computational overhead and improve efficiency through device-persistent data structures and optimized stopping criteria. The results highlight the potential for enhanced computational efficiency in large-scale CFD simulations.

To read the full paper, visit Springer Link . For more details on the implementation, the project's repository can be found at GitHub . We extend our congratulations to the authors for their contributions!

An assortment of OpenFOAM cases designed for HPC testing is incorporated in the EXASIM benchmark suite. They are setup to be used with our dedicated framework for large parametric OpenFOAM studies based on signac. You can find the code for the OpenFOAM Benchmark Runner (OBR) on Github.

The Windsor body micro-benchmark is designed to enable weak scaling analysis, so we can identify bottlenecks in the code and push towards exascale computing. Identifying optimisation potential based on industrially relevant cases will help us develop innovative technology for efficient simulations in real world design processes. The EXASIM team is looking forward to seeing how our hard work will have a positive impact for industry.