Exploiting compression opportunities to improve spmxv performance on shared memory systems

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	2014-03-01T01:33:28Z
dc.date.available	2014-03-01T01:33:28Z
dc.date.issued	2010	en
dc.identifier.issn	1544-3566	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/20440
dc.subject	Data compression	en
dc.subject	Memory bandwidth	en
dc.subject	Shared memory systems	en
dc.subject	Sparse matrix	en
dc.subject.classification	Computer Science, Hardware & Architecture	en
dc.subject.classification	Computer Science, Theory & Methods	en
dc.subject.other	Coarse-grained	en
dc.subject.other	Compression scheme	en
dc.subject.other	Experimental evaluation	en
dc.subject.other	ITS data	en
dc.subject.other	matrix	en
dc.subject.other	Memory bandwidths	en
dc.subject.other	Memory contentions	en
dc.subject.other	Multithreaded	en
dc.subject.other	Shared memory system	en
dc.subject.other	Sparse matrices	en
dc.subject.other	Sparse matrix-vector multiplication	en
dc.subject.other	State-of-the-art approach	en
dc.subject.other	Structural data	en
dc.subject.other	Bandwidth compression	en
dc.subject.other	Matrix algebra	en
dc.subject.other	Data compression	en
dc.title	Exploiting compression opportunities to improve spmxv performance on shared memory systems	en
heal.type	journalArticle	en
heal.identifier.primary	10.1145/1880037.1880041	en
heal.identifier.secondary	http://dx.doi.org/10.1145/1880037.1880041	en
heal.identifier.secondary	16	en
heal.language	English	en
heal.publicationDate	2010	en
heal.abstract	The Sparse Matrix-Vector Multiplication (SpMxV) kernel exhibits poor scaling on shared memory systems, due to the streaming nature of its data access pattern. To decrease memory contention and improve kernel performance we propose two compression schemes: CSR-DU, that targets the reduction of the matrix structural data by applying coarse-grained delta-encoding, and CSR-VI, that targets the reduction of the values using indirect indexing, applicable to matrices with a small number of unique values. Thorough experimental evaluation of the proposed methods and their combination, on two modern shared memory systems, demonstrated that they can significantly improve multithreaded SpMxV performance upon standard and state-of-the-art approaches. © 2010 ACM.	en
heal.publisher	ASSOC COMPUTING MACHINERY	en
heal.journalName	Transactions on Architecture and Code Optimization	en
dc.identifier.doi	10.1145/1880037.1880041	en
dc.identifier.isi	ISI:000285826500004	en
dc.identifier.volume	7	en
dc.identifier.issue	3	en