Communication-Avoiding Parallel Sparse-Dense Matrix-Matrix Multiplication

Multiplication of a sparse matrix with a dense matrix is a building block of an increasing number of applications in many areas such as machine learning and graph algorithms. However, most previous work on parallel matrix multiplication considered only both dense or both sparse matrix operands. This...

Full description

Saved in:
Bibliographic Details
Published in:Proceedings - IEEE International Parallel and Distributed Processing Symposium pp. 842 - 853
Main Authors: Koanantakool, Penporn, Azad, Ariful, Buluc, Aydin, Morozov, Dmitriy, Sang-Yun Oh, Oliker, Leonid, Yelick, Katherine
Format: Conference Proceeding
Language:English
Published: IEEE 01.05.2016
Subjects:
Algorithm design and analysis
communication-avoiding algorithms
Covariance matrices
linear algebra
Machine learning algorithms
Parallel algorithms
Partitioning algorithms
Sparse matrices
sparse-dense matrix-matrix multiplication
Three-dimensional displays
ISSN:1530-2075
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Multiplication of a sparse matrix with a dense matrix is a building block of an increasing number of applications in many areas such as machine learning and graph algorithms. However, most previous work on parallel matrix multiplication considered only both dense or both sparse matrix operands. This paper analyzes the communication lower bounds and compares the communication costs of various classic parallel algorithms in the context of sparse-dense matrix-matrix multiplication. We also present new communication-avoiding algorithms based on a 1D decomposition, called 1.5D, which - while suboptimal in dense-dense and sparse-sparse cases - outperform the 2D and 3D variants both theoretically and in practice for sparse-dense multiplication. Our analysis separates one-time costs from per iteration costs in an iterative machine learning context. Experiments demonstrate speedups up to 100x over a baseline 3D SUMMA implementation and show parallel scaling over 10 thousand cores.
ISSN:1530-2075
DOI:10.1109/IPDPS.2016.117