GPU-Powered Particle-in-Cell Community Frameworks for Laser-Plasma Interaction

In the context of laser-particle acceleration, the electro-magnetic particle-in-cell codes PIConGPU and WarpX are presented. Novel developments and workflows that enable high-resolution, fast turn-around computations on manycore-powered, leadership-scale supercomputers are essential to make optimal use of upcoming Exascale machines. Both codes' software libraries and abstractions are build on top of a generalized, single-source programming model (Alpaka) or parallel-for/-reduce based kernels. While PIConGPU is designed on top of modular, single-purpose libraries, WarpX's core routines are constructed on top of a more monolithic dependency, AMReX, which is a widely used adaptive-mesh refinement framework.

Both particle-in-cell codes share the same challenges for handling PByte-scale data workflows on pre-Exascale machines. A common, open data format for particle and mesh data (openPMD) avoids duplicating I/O efforts and allows to reuse scalable data workflows with common libraries. In production runs, close bindings to scripting languages and the Jupyter platform can provide efficient control of simulations, with the goal of fast turn-arounds and good applicability to experiments. We present standardization efforts and prototypes of both communities with emphasis on reproducibility and flexibility.