Optimizing Sparse Matrix-Matrix Multiplication for the GPU

Dalton, Steven; Bell, Nathan; Olson, Luke

Permalink

https://hdl.handle.net/2142/42667 Copy

Description

Title

Optimizing Sparse Matrix-Matrix Multiplication for the GPU

Author(s)

• Dalton, Steven
• Bell, Nathan
• Olson, Luke

Issue Date

2013-03-26

Keyword(s)

• parallel
• sparse
• gpu
• matrix-matrix

Abstract

Sparse matrix-matrix multiplication (SpMM) is a key operation in numerous ar- eas from information to the physical sciences. Implementing SpMM 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 SpMM operation into three, highly-parallel phases: expansion, sorting, and compres- sion, and introduce a set of complementary bandwidth-saving performance optimiza- tions. Our implementation is fully general and our optimizations lead to substantial efficiencies for a SpMM product.

Type of Resource

text

Language

en

Permalink

http://hdl.handle.net/2142/42667

Owning Collections