On 9 December 2025, we wrapped up our webinar programme for the year with a session focused on a timely challenge across high-performance computing: how to prepare long-standing Fortran simulation codes for today’s GPU-based systems. In this final webinar of 2025, our own Spencer Starr shared insights from the Numerics Research Group at the Institute of Aerodynamics and Gas Dynamics through the Lense of their experience with GALÆXI, a CFD solver developed to run efficiently on both traditional CPUs and modern GPUs.
The webinar started with a clear, beginner-friendly introduction to GPU computing, explaining how GPUs work and why they differ from CPUs in the way they handle data and parallel work. From there, the talk addressed a common dilemma faced by many research teams: while Fortran remains widely used in scientific computing, support for it in GPU programming tools is often limited. The presentation explored one practical solution—combining existing Fortran code with CUDA and HIP in C++—and discussed the questions developers need to consider before choosing this path.
To keep the discussion grounded, the speaker used GALÆXI as a real-world example, walking through how an existing CFD code was adapted for GPUs, how researchers can minimise intrusive code changes, and how results can be carefully verified to ensure scientific correctness across different hardware platforms. Whether you are actively porting a code or simply trying to understand what’s involved, the webinar offers a valuable overview of the process, challenges, and design choices.
The full recording and slides are now available below for anyone who missed the live session or would like a closer look. This webinar brings CEEC’s 2025 webinar programme to a close, but we’ll be back in the new year with more talks focused on practical, hands-on topics for the HPC and CFD community. Stay tuned for upcoming CEEC webinars in 2026!