Benchmarking GPU and CPU codes for Heisenberg spin glass overrelaxation
Istituto Applicazioni Calcolo, CNR, Viale Manzoni, 30 – 00185 Rome, Italy
Computer Physics Communications (03 March 2011)
@article{Bernaschi2011,
title={“Benchmarking GPU and CPU codes for Heisenberg spin glass overrelaxation”},
journal={“Computer Physics Communications”},
volume={“InPress},
number={“”},
pages={“-“},
year={“2011”},
note={“”},
issn={“0010-4655”},
doi={“DOI:10.1016/j.cpc.2011.02.012”},
url={“http://www.sciencedirect.com/science/article/B6TJ5-529CNR2-3/2/d7cae37f6747ec711090bf3fa0cd8c20”},
author={“M.Bernaschi and G.Parisi and L.Parisi”},
keywords={“GPU”}
}
We present a set of possible implementations for Graphics Processing Units (GPU) of the Overrelaxation technique applied to the 3D Heisenberg spin glass model. The results show that a carefully tuned code can achieve more than 100 GFlops/sec. of sustained performance and update a single spin in about 0.6 nanoseconds. A multi-hit technique that exploits the GPU shared memory further reduces this time. Such results are compared with those obtained by means of a highly-tuned vector-parallel code on latest generation multi-core CPUs.
March 9, 2011 by hgpu
