Sparse matrix-vector multiplication on GPGPU clusters: A new storage format and a scalable implementation
Erlangen Regional Computing Center, Erlangen, Germany
arXiv:1112.5588v1 [cs.DC] (23 Dec 2011)
@article{2011arXiv1112.5588K,
author={Kreutzer}, M. and {Hager}, G. and {Wellein}, G. and {Fehske}, H. and {Basermann}, A. and {Bishop}, A.~R.},
title={"{Sparse matrix-vector multiplication on GPGPU clusters: A new storage format and a scalable implementation}"},
journal={ArXiv e-prints},
archivePrefix={"arXiv"},
eprint={1112.5588},
primaryClass={"cs.DC"},
keywords={Computer Science – Distributed, Parallel, and Cluster Computing, Computer Science – Performance},
year={2011},
month={dec},
adsurl={http://adsabs.harvard.edu/abs/2011arXiv1112.5588K},
adsnote={Provided by the SAO/NASA Astrophysics Data System}
}
Sparse matrix-vector multiplication (spMVM) is the dominant operation in many sparse solvers. We investigate performance properties of spMVM with matrices of various sparsity patterns on the nVidia "Fermi" class of GPGPUs. A new "padded jagged diagonals storage" (pJDS) format is proposed which may substantially reduce the memory overhead intrinsic to the widespread ELLPACK-R scheme. In our test scenarios the pJDS format cuts the overall spMVM memory footprint on the GPGPU by up to 70%, and achieves 95% to 130% of the ELLPACK-R performance. Using a suitable performance model we identify performance bottlenecks on the node level that invalidate some types of matrix structures for efficient multi-GPGPU parallelization. For appropriate sparsity patterns we extend previous work on distributed-memory parallel spMVM to demonstrate a scalable hybrid MPI-GPGPU code, achieving efficient overlap of communication and computation.
December 26, 2011 by hgpu
