Accelerated Many-Core GPU Computing for Physics and Astrophysics on Three Continents

Graphical processing units (GPUs) have become widely used to accelerate a broad range of applications, including computational physics and astrophysics, image/video processing, engineering simulations, and quantum chemistry. This chapter presents results obtained from GPU clusters with previous generations of GPU accelerators, which have no or very limited double-precision support. It provides an astrophysical N-body application for star clusters and galactic nuclei, which is currently the well-tested and heavily used application. The chapter explains exemplary implementations of parallel codes using many GPUs as accelerators, so combining message passing parallelization with many-core parallelization. It discusses their benchmarks using up to 512 Fermi Tesla GPUs in parallel on the Mole-8.5 hardware of the Institute of Process Engineering of the Chinese Academy of Sciences (IPE/CAS) in Beijing, and the Laohu Tesla C1070 cluster of the National Astronomical Observatories of CAS in Beijing and smaller clusters in Germany and in the United States.

galactic nuclei; graphics processing units; star clusters