Performance Portability of Heterogeneous Components of SeisSol
This discussion explores the recent results of integrating the oneAPI programming model into SeisSol, a software package for simulating seismic waves and earthquake dynamics. See a demonstration comparing various SeisSol-specific benchmarks compiled and run using oneAPI, hipSYCL, and CUDA*. The presentation ends with performance of the whole application obtained with two NVIDIA* GeForce RTX* 3090 turbo GPUs using oneAPI compiled with a CUDA back end.
Ravil Dorozhinskii is a PhD candidate at the Technical University of Munich. His current research topic is heterogeneous computing in computational seismology as part of the ChEESE project. He obtained his first degree in engineering at Bauman Moscow State Technical University with a focus on numerical methods and simulations. After a few years of working in the industry, he joined the computational science and engineering master’s program at the Technical University of Munich and focused on high-performance computing (HPC). He is working on generating GPU code for SeisSol. His current research interest is in compiler design and construction. As a part of his PhD study, he teaches HPC algorithms and applications as well as 3D game physics.
Product and Performance Information
Performance varies by use, configuration and other factors. Learn more at www.Intel.com/PerformanceIndex.