Exascale multiphase flow solver - 2025 Gordon Bell Prize Finalist

MFC conducted the largest known public CFD simulation at 200 trillion grid points on 43,000+ AMD APUs on El Capitan and 33,000+ AMD GPUs on Frontier. It is written in ~40K lines of Fortran with Fypp metaprogramming.

Get started

• Repository: MFlowCode/MFC (pinned below)
• Documentation: mflowcode.github.io
• Quick start: Getting Started guide or open a Codespace
• Videos: YouTube channel - simulation visualizations and demos

Capabilities

• Exascale GPU performance - Ideal weak scaling on El Capitan (MI300A), Frontier (MI250X), and Alps (GH200). Near compute-roofline. Compile-time case optimization for up to 10x speedup.
• Multi-phase/physics - 4, 5, and 6-equation models, phase change, surface tension, bubble dynamics, MHD, relativistic MHD, hyper/hypoelasticity, chemistry, and Euler-Lagrange particle tracking.
• Portable - NVIDIA and AMD GPUs, CPUs, laptops to exascale. Docker, Codespaces, Homebrew, and 16+ HPC system templates.

Community

MFC nucleated from Tim Colonius's group at Caltech. The Bryngelson Group at Georgia Tech leads development, with close collaborators at Brown (Rodriguez), WPI (Gnanaskandan), and others.

MFC is a SPEChpc benchmark candidate, part of the JSC JUPITER Early Access Program, and used OLCF Frontier and LLNL El Capitan early access systems.

Questions? Join the MFC Slack or contact Spencer Bryngelson.