A Push-Relabel-Based Maximum Cardinality Bipartite Matching Algorithm on GPUs
Dept. of Biomedical Informatics, The Ohio State University
hal-00923464, (2 January 2014)
@inproceedings{deveci:hal-00923464,
hal_id={hal-00923464},
url={http://hal.inria.fr/hal-00923464},
title={A Push-Relabel-Based Maximum Cardinality Bipartite Matching Algorithm on GPUs},
author={Deveci, Mehmet and Kaya, Kamer and U{c c}ar, Bora and Catalyurek, Umit, V.},
language={Anglais},
affiliation={Department of Biomedical Informatics, Laboratoire de l’Informatique du Parall{‘e}lisme – LIP , ROMA – ENS Lyon / CNRS / Inria Grenoble Rh{^o}ne-Alpes},
booktitle={42nd International Conference on Parallel Processing},
publisher={IEEE Computer Society},
pages={21 – 29},
address={Lyon, France},
audience={internationale},
doi={10.1109/ICPP.2013.11},
year={2013},
pdf={http://hal.inria.fr/hal-00923464/PDF/matchingGPU.pdf}
}
We design, develop, and evaluate an atomic- and lock-free GPU implementation of the push-relabel algorithm in the context of finding maximum cardinality matchings in bipartite graphs. The problem has applications on computer science, scientific computing, bioinformatics, and other areas. Although the GPU parallelization of the push-relabel technique has been investigated in the context of flow algorithms, to the best of our knowledge, ours is the first study which focuses on the maximum cardinality matching. We compare the proposed algorithms with serial, multicore, and manycore bipartite graph matching implementations from the literature on a large set of real-life problems. Our experiments show that the proposed push-relabel-based GPU algorithm is faster than the existing parallel and sequential implementations.
January 5, 2014 by hgpu
