Erlangen 2026 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

P: Fachverband Plasmaphysik

P 16: Poster Session Plasma Physics

P 16.1: Poster

Donnerstag, 19. März 2026, 13:45–15:45, Redoutensaal

GPU Acceleration and Portability of the TRIMEG Code for Global Gyrokinetic Plasma Simulations using OpenMP — •Giorgio Daneri, Zhixin Lu, and Matthias Hoelzl — Max Planck Institute for Plasma Physics, Boltzmannstr. 2, 85748 Garching b. M., Germany

The TRIangular MEsh-based Gyrokinetic (TRIMEG) code solves the gyrokinetic equations using a particle-in-cell (PIC) scheme to simulate electromagnetic instabilities throughout the full tokamak volume.

In this work, we focus on the acceleration and portability of the TRIMEG code on multiple GPU architectures. The OpenMP framework is chosen as the methodological approach for GPU offloading in Fortran on NVIDIA and AMD platforms. The particle pushing procedure, as well as particle-to-grid operations have been adapted for GPU execution. The kernels were analyzed with the available profiling tools to gather metrics about resource occupancy, throughput, and memory usage. Their performance was evaluated by carrying out GPU grid size exploration, as well as multi-node scalability studies. In addition, the efficiency of hybrid MPI-OpenMP offloading parallelization was assessed.

The Ion Temperature Gradient (ITG) mode was simulated using the GPU-accelerated version for the economical Cyclone case and the TCV case, and its correctness was verified by comparing the physics results in terms of the energy growth rate and the two-dimensional mode structures.

Keywords: gyrokinetics; particle-in-cell; GPU acceleration; openMP offloading; performance portability