Simulating Laser Particle Acceleration on a Cluster of GPU Nodes
Simulating Laser Particle Acceleration on a Cluster of GPU Nodes
Bussmann, M.; Burau, H.; Widera, R.; Hönig, W.; Schmitt, F.; Juckeland, G.; Debus, A.; Kluge, T.; Schramm, U.; Cowan, T. E.; Sauerbrey, R.
We present the particle-in-cell code PIConGPU. This fully-relativistic plasma code utilizes the computational power of graphic processing units.
We show that it scales well even on a cluster of GPU-enhanced nodes, by interleaving the communication between the GPU and CPU and the inter-node communication with the local computation steps.
Keywords: GPU; particle-in-cell; latency; communication algorithm; laser; plasma; acceleration; simulation; graphic cards; pic
-
Invited lecture (Conferences)
The State of Affairs of GPU Computing in HPC, 09.02.2010, Dresden, Deutschland
Permalink: https://www.hzdr.de/publications/Publ-13821