Speaker
Description
Laser-plasma accelerators (LPAs), offering an enormous accelerating gradient of GV/m, have garnered significant attention as a promising avenue to a variety of advanced applications from tabletop radiation sources to high-energy compact accelerators. Here, we present a collection of open-source simulation tools to further our understanding and support experiments on plasma acceleration. In particular, the 3D, GPU-portable quasi-static particle-in-cell code HIPACE++ makes abundant optimization on memory management and computing performance, demonstrating orders-of-magnitude speedups on modern scientific GPUs over CPU-only implementations. The longitudinal parallelization is done through a custom pipeline algorithm, enabling near-optimal strong scaling from 1 to 512 GPUs.
Key words: Laser-plasma accelerator, Particle-in-cell simulation, High performance computing
| Speed talk: | I am unwilling/unable to present a speed talk |
|---|
