GPU-Acceleration of RHEA Flow Solver with MPI+OpenACC

In computational fluid dynamics, large-scale simulations of turbulence rely heavily on ‌a flow solver’s speed and adaptability, where functions such as inviscid flux and viscous flux can become bottlenecks to fast computations. Project RHEA, Reproducible Hybrid-architecture flow solver Engineered for Academia, is an open-source solution that enables groundbreaking engineering applications by successfully simulating turbulent flows with complex physics.

RHEA can be adaptable for many cases, including different thermodynamic models, transport coefficient models, and Riemann solver schemes. The flow solver targets hybrid supercomputing architectures based on ‌state-of-the-art parallel and scalable MPI and is accelerated by OpenACC.

In his presentation at the Open Accelerated Computing Summit, Ahmed Abdellatif discusses GPU porting, data management, solver performance, and scalability on different multi-node/GPU systems for simulations of turbulent flows. Learn how the team achieved a 6x speedup of the solver, with individual functions running up to 27 times faster. Watch now.