Frames per Second is the new CPU hours - Particle-in-Cell Simulations on large GPU clusters

GPUs are one implementation of the many-core architecture that could be an important building block of future Exascale HPC-systems.
I present PIConGPU, an electromagnetic particle-in-cell code for laser-plasma interaction, discussing the important building blocks and techniques implemented in PIConGPU to leverage the power of large-scale GPU clusters.
PIConGPU hints at what to expect in the near-term future of massively-parallel particle-in-cell simulations in terms of parameter surveys, new physics modules and large-scale simulations.