Adaptive fast multipole methods on the GPU
Division of Electricity, Department of Engineering Sciences, Uppsala University, SE-751 21, Uppsala, Sweden
The Journal of Supercomputing; arXiv:1205.4611v1 [cs.DC] (21 May 2012)
@article{Goude2012Adaptive,
author={Goude, Anders and Engblom, Stefan},
title={Adaptive fast multipole methods on the GPU},
journal={The Journal of Supercomputing},
keywords={Adaptive fast multipole method; CUDA; Graphics processing units; Tesla C2075},
year={2012},
issn={0920-8542},
doi={10.1007/s11227-012-0836-0},
url={http://dx.doi.org/10.1007/s11227-012-0836-0},
publisher={Springer US},
pages={1-22},
language={English}
}
We present a highly general implementation of fast multipole methods on graphics processing units (GPUs). Our two-dimensional double precision code features an asymmetric type of adaptive space discretization leading to a particularly elegant and flexible implementation. All steps of the multipole algorithm are efficiently performed on the GPU, including the initial phase which assembles the topological information of the input data. Through careful timing experiments we investigate the effects of the various peculiarities of the GPU architecture.
May 23, 2012 by hgpu
