Improving the performance of multithreaded sparse matrix-vector multiplication using index and value compression

Kourtis, K; Goumas, G; Koziris, N

dc.contributor.author	Kourtis, K	en
dc.contributor.author	Goumas, G	en
dc.contributor.author	Koziris, N	en
dc.date.accessioned	2014-03-01T02:45:30Z
dc.date.available	2014-03-01T02:45:30Z
dc.date.issued	2008	en
dc.identifier.issn	01903918	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/32283
dc.subject	Coarse Grained	en
dc.subject	Memory Bandwidth	en
dc.subject	Shared Memory	en
dc.subject	Sparse Matrix	en
dc.subject	Structured Data	en
dc.subject.other	Coarse grains	en
dc.subject.other	Compression schemes	en
dc.subject.other	Delta encoding	en
dc.subject.other	Memory bandwidth requirements	en
dc.subject.other	Multithreaded	en
dc.subject.other	Numerical values	en
dc.subject.other	Shared memories	en
dc.subject.other	Sparse matrixes	en
dc.subject.other	Structural datums	en
dc.subject.other	Vector multiplications	en
dc.subject.other	Data compression	en
dc.subject.other	Encoding (symbols)	en
dc.subject.other	Online searching	en
dc.subject.other	Vectors	en
dc.subject.other	Data storage equipment	en
dc.title	Improving the performance of multithreaded sparse matrix-vector multiplication using index and value compression	en
heal.type	conferenceItem	en
heal.identifier.primary	10.1109/ICPP.2008.62	en
heal.identifier.secondary	http://dx.doi.org/10.1109/ICPP.2008.62	en
heal.identifier.secondary	4625888	en
heal.publicationDate	2008	en
heal.abstract	The Sparse Matrix-Vector Multiplication kernel exhibits limited potential for taking advantage of modern shared memory architectures due to its large memory bandwidth requirements. To decrease memory contention and improve the performance of the kernel we propose two compression schemes. The first, called CSR-DU, targets the reduction of the matrix structural data by applying coarse grain delta encoding for the column indices. The second scheme, called CSR-VI, targets the reduction of the numerical values using indirect indexing and can only be applied to matrices which contain a small number of unique values. Evaluation of both methods on a rich matrix set showed that they can significantly improve the performance of the multithreaded version of the kernel and achieve good scalability for large matrices. © 2008 IEEE.	en
heal.journalName	Proceedings of the International Conference on Parallel Processing	en
dc.identifier.doi	10.1109/ICPP.2008.62	en
dc.identifier.spage	511	en
dc.identifier.epage	519	en