A scalable implementation of the Particle-in-Cell Algorithm on GPU Clusters

We present PIConGPU, an implementation of the particle-in-cell algorithm for relativistic laser-plasma interactions. PIConGPU is currently used to simulate laser-wakefield acceleration of electrons as a model for strongly anisotropic many-particle systems. We discuss caveats when porting the particle-in-cell algorithm to GPUs and lessons learned from extending the code from a single-GPU version to a multiple-GPU version.