Mainz 2022 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
P: Fachverband Plasmaphysik
P 19: Poster II
P 19.3: Poster
Thursday, March 31, 2022, 16:00–17:30, P
GPU Offloading of the Gyrokinetic Turbulence Code GENE-X — •Jordy Trilaksono1, Dominik Michels1, Andreas Stegmeir1, and Frank Jenko1,2 — 1Max Planck Institute for Plasma Physics, Boltzmannstraße 2, 85748 Garching, Germany — 2University of Texas at Austin, Austin, TX 78712, USA
Turbulence in magnetic confinement fusion devices is a non-linear phenomenon which involves multi-scale and multi-physics modeling. Simulating turbulence requires a large number of computing resources exploited in parallel which is provided by modern supercomputers. The recently developed gyrokinetic turbulence code GENE-X [1] extends the typical coverage of gyrokinetic turbulence simulations from the core to the edge and scrape-off layer of magnetic confinement fusion devices. Currently, GENE-X uses a heterogenous parallelization featuring OpenMP for intranode and MPI for internode parallelism respectively. To enable simulations of the edge and scrape-off layer of reactor relevant fusion devices, like ITER, the scalability of GENE-X needs to be improved. Therefore, we present progress towards GPU offloading in GENE-X in this work. This includes improving the current offloading approach of GENE-X by implementing a separate C++ layer to the code using modern Fortran’s C interoperability and CUDA.
[1] D. Michels, et. al., Comput. Phys. Commun. 264, 107986 (2021)