Browse Prior Art Database

Sparse Matrix-Vector Multiplication on Processors with Small Memory Footprints

IP.com Disclosure Number: IPCOM000184518D
Original Publication Date: 2009-Jun-29
Included in the Prior Art Database: 2009-Jun-29
Document File: 5 page(s) / 32K

Publishing Venue

IBM

Abstract

We present a sparse matrix-dense vector multiplication procedure which is suitable for a subclass of sparse matrices (which we call block-sparse matrices), and which effectively harnesses the compute power of the Cell/B.E. processor. This technique could be mapped to GPUs, Larrabee, or any low-memory footprint microprocessor. .