Target normal sheath acceleration and laser wakefield acceleration particle-in-cell simulations performance on CPU & GPU architectures for high-power laser systems

Abstract

We present an integrated study on the scalability and performance of particle-in-cell (PIC) code simulations on CPU and GPU architectures of high parallelization focused on target normal sheath acceleration (TNSA) and laser wakefield acceleration (LWFA) experiments. The developed models follow the experimental specifications of the high-power lasers systems hosted at the infrastructures of the Institute of Plasma Physics & Lasers of the Hellenic Mediterranean University in Crete, Greece and the Extreme Light Infrastructure - Nuclear Physics in Romania. The simulations are implemented on the High-Performance Computer for Advanced Research Information System of the Greek National Infrastructures for Research and Technology. Two representative experiments for TNSA and for LWFA are initially simulated by 2D models with the minimum computational resource demands and the results are used as a reference for the scalability and performance investigation. We further extend our study to 3D models aiming to reproduce the physics involved in both laser-plasma particle accelerators. A detailed analysis of the simulation results, accompanied with the computational demands, scalability and performance of the CPU and the GPU architectures, is provided. Our research findings highlight the key features and parameters of the physical and numerical models which drive simulations to converge to reliable results by means of physics, computational and runtime demands and shed light on their influence on the efficiency and performance of the PIC simulations.