Efficient sparse-matrix multi-vector product on GPUs
Open Access
- 11 June 2018
- conference paper
- Published by Association for Computing Machinery (ACM)
- p. 66-79
Abstract
Sparse Matrix-Vector (SpMV) and Sparse Matrix-Multivector (SpMM) products are key kernels for computational science and data science. While GPUs offer significantly higher peak performance and memory bandwidth than multicore CPUs, achieving high performance on sparse computations on GPUs is very challenging. A tremendous amount of recent research has focused on various GPU implementations of the SpMV kernel. But the multi-vector SpMM kernel has received much less attention. In this paper, we present an in-depth analysis to contrast SpMV and SpMM, and develop a new sparse-matrix representation and computation approach suited to achieving high data-movement efficiency and effective GPU parallelization of SpMM. Experimental evaluation using the entire SuiteSparse matrix suite demonstrates significant performance improvement over existing SpMM implementations from vendor libraries.Keywords
Funding Information
- Defense Sciences Office, DARPA (D16PC00183)
- National Science Foundation (1404995,1513120,1629548,1645599,1747447)
This publication has 27 references indexed in Scilit:
- FastSpMM: An Efficient Library for Sparse Matrix Matrix Product on GPUsThe Computer Journal, 2013
- A reflection on the implicitly restarted Arnoldi method for computing eigenvalues near a vertical lineLinear Algebra and its Applications, 2012
- Strategies for spectrum slicing based on restarted Lanczos methodsNumerical Algorithms, 2012
- CSXACM SIGPLAN Notices, 2011
- Restarting the Nonsymmetric Lanczos Algorithm for Eigenvalues and Linear Equations Including Multiple Right-Hand SidesSIAM Journal on Scientific Computing, 2011
- A thick-restarted block Arnoldi algorithm with modified Ritz vectors for large eigenproblemsComputers & Mathematics with Applications, 2010
- The WEKA data mining softwareACM SIGKDD Explorations Newsletter, 2009
- RooflineCommunications of the ACM, 2009
- On Improving Linear Solver Performance: A Block Variant of GMRESSIAM Journal on Scientific Computing, 2006
- Toward the Optimal Preconditioned Eigensolver: Locally Optimal Block Preconditioned Conjugate Gradient MethodSIAM Journal on Scientific Computing, 2001