Performance Optimization of an Exascale Implicit Kinetic Plasma Simulation on El Capitan
We present performance scaling and optimization of iPIC3D - an exascale-class, GPU-enabled implicit particle-in-cell code for planetary-scale magnetosphere modeling and plasma simulation - on El Capitan. Our strong and weak scaling studies demonstrate near-linear scaling up to 8,000 nodes (32,000 APUs) with a parallel efficiency of nearly 100%. We optimize iPIC3D to leverage AMD’s MI300A APUs, the Merced Lustre filesystem, and Rabbit nodes for high-bandwidth I/O. Optimizations reduce memory usage by 97% and runtime by 74%, enabling simulations that are 1.8 times larger than before. Rabbit further improve checkpointing bandwidth by 2 times, ensuring scalable fault- tolerant simulations on exascale architectures.
Authors
- Dewi Yokelson (ORCID: https://orcid.org/0000-0003-1453-5906)
- Stephanie Brink (ORCID: https://orcid.org/0000-0002-1458-8453)
- Olga Pearce (ORCID: https://orcid.org/0000-0002-1904-9627)
- Gian Luca Delzanno (ORCID: https://orcid.org/0000-0002-7030-2683)
- Ivy Peng (ORCID: https://orcid.org/0000-0003-4158-3583)
- Bronis R. de Supinski (ORCID: https://orcid.org/0000-0002-0339-1006)
- Michela Taufer (ORCID: https://orcid.org/0000-0002-0031-6377)
- Hariharan Devarajan (ORCID: https://orcid.org/0000-0001-5625-3494)
- Andong Hu (ORCID: https://orcid.org/0000-0002-6929-2158)
- Tom Scogland (ORCID: https://orcid.org/0000-0001-7234-5743)
- Stefano Markidis (ORCID: https://orcid.org/0000-0003-0639-0639)
- Ian Lumsden (ORCID: https://orcid.org/0000-0003-0009-5487)
- Luca Pennati (ORCID: https://orcid.org/0009-0009-4901-1716)
Institutions
- Lawrence Livermore National Laboratory (US)
- Advanced Micro Devices (Canada) (CA)
- Los Alamos National Laboratory (US)
- University of Tennessee at Knoxville (US)
- KTH Royal Institute of Technology (SE)
Publication Details
- Published
- 2025-09-15
- DOI
- https://doi.org/10.1109/escience65000.2025.00050
- Primary Topic
- Plasma Diagnostics and Applications
- Type
- article
- Field-Weighted Citation Impact
- 0.00