Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Optimizing Sparse Matrix—Matrix Multiplication for the GPU

Optimizing Sparse Matrix—Matrix Multiplication for the GPU Optimizing Sparse Matrix--Matrix Multiplication for the GPU STEVEN DALTON and LUKE OLSON, University of Illinois at Urbana­Champaign NATHAN BELL, Google Sparse matrix­matrix multiplication (SpGEMM) is a key operation in numerous areas from information to the physical sciences. Implementing SpGEMM efficiently on throughput-oriented processors, such as the graphics processing unit (GPU), requires the programmer to expose substantial fine-grained parallelism while conserving the limited off-chip memory bandwidth. Balancing these concerns, we decompose the SpGEMM operation into three highly parallel phases: expansion, sorting, and contraction, and introduce a set of complementary bandwidth-saving performance optimizations. Our implementation is fully general and our optimization strategy adaptively processes the SpGEMM workload row-wise to substantially improve performance by decreasing the work complexity and utilizing the memory hierarchy more effectively. Categories and Subject Descriptors: G.4 [Mathematical Software]: Algorithm Design and Analysis General Terms: Algorithms, Performance Additional Key Words and Phrases: Parallel, sparse, GPU, matrix­matrix ACM Reference Format: Steven Dalton, Luke Olson, and Nathan Bell. 2015. Optimizing sparse matrix­matrix multiplication for the GPU. ACM Trans. Math. Softw. 41, 4, Article 25 (October 2015), 20 pages. DOI: http://dx.doi.org/10.1145/2699470 1. INTRODUCTION Operations on sparse data structures abound in all areas of information and physical science. In particular, the sparse http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png ACM Transactions on Mathematical Software (TOMS) Association for Computing Machinery

Optimizing Sparse Matrix—Matrix Multiplication for the GPU

Download PDF
20 pages

Loading next page...
 
/lp/association-for-computing-machinery/optimizing-sparse-matrix-matrix-multiplication-for-the-gpu-6ayk1u1hd0

References

References for this paper are not available at this time. We will be adding them shortly, thank you for your patience.

Publisher
Association for Computing Machinery
Copyright
Copyright © 2015 by ACM Inc.
ISSN
0098-3500
DOI
10.1145/2699470
Publisher site
See Article on Publisher Site

Abstract

Optimizing Sparse Matrix--Matrix Multiplication for the GPU STEVEN DALTON and LUKE OLSON, University of Illinois at Urbana­Champaign NATHAN BELL, Google Sparse matrix­matrix multiplication (SpGEMM) is a key operation in numerous areas from information to the physical sciences. Implementing SpGEMM efficiently on throughput-oriented processors, such as the graphics processing unit (GPU), requires the programmer to expose substantial fine-grained parallelism while conserving the limited off-chip memory bandwidth. Balancing these concerns, we decompose the SpGEMM operation into three highly parallel phases: expansion, sorting, and contraction, and introduce a set of complementary bandwidth-saving performance optimizations. Our implementation is fully general and our optimization strategy adaptively processes the SpGEMM workload row-wise to substantially improve performance by decreasing the work complexity and utilizing the memory hierarchy more effectively. Categories and Subject Descriptors: G.4 [Mathematical Software]: Algorithm Design and Analysis General Terms: Algorithms, Performance Additional Key Words and Phrases: Parallel, sparse, GPU, matrix­matrix ACM Reference Format: Steven Dalton, Luke Olson, and Nathan Bell. 2015. Optimizing sparse matrix­matrix multiplication for the GPU. ACM Trans. Math. Softw. 41, 4, Article 25 (October 2015), 20 pages. DOI: http://dx.doi.org/10.1145/2699470 1. INTRODUCTION Operations on sparse data structures abound in all areas of information and physical science. In particular, the sparse

Journal

ACM Transactions on Mathematical Software (TOMS) – Association for Computing Machinery

Published: Oct 26, 2015

There are no references for this article.