Fast Sparse Matrix-Vector Multiplication on GPUs: Implications for Graph Mining
Department of Computer Science and Engineering, Ohio State University, Columbus, OH 43210
arXiv:1103.2405v1 [cs.DB] (12 Mar 2011), VLDB2011; Proceedings of the VLDB Endowment (PVLDB), Vol. 4, No. 4, pp. 231-242 (2011)
@article{2011arXiv1103.2405Y,
author={Yang}, X. and {Parthasarathy}, S. and {Sadayappan}, P.},
title={“{Fast Sparse Matrix-Vector Multiplication on GPUs: Implications for Graph Mining}”},
journal={ArXiv e-prints},
archivePrefix={“arXiv”},
eprint={1103.2405},
primaryClass={“cs.DB”},
keywords={Computer Science – Databases},
year={2011},
month={mar},
adsurl={http://adsabs.harvard.edu/abs/2011arXiv1103.2405Y},
adsnote={Provided by the SAO/NASA Astrophysics Data System}
}
Scaling up the sparse matrix-vector multiplication kernel on modern Graphics Processing Units (GPU) has been at the heart of numerous studies in both academia and industry. In this article we present a novel non-parametric, self-tunable, approach to data representation for computing this kernel, particularly targeting sparse matrices representing power-law graphs. Using real data, we show how our representation scheme, coupled with a novel tiling algorithm, can yield significant benefits over the current state of the art GPU efforts on a number of core data mining algorithms such as PageRank, HITS and Random Walk with Restart.
March 15, 2011 by hgpu
