Encapsulating multiple communication-cost metrics in partitioning sparse rectangular matrices for parallel matrix-vector multiplies

Simple item page
buir.contributor.authorAykanat, Cevdet
dc.citation.epage1859en_US
dc.citation.issueNumber6en_US
dc.citation.spage1837en_US
dc.citation.volumeNumber25en_US
dc.contributor.authorUçar, B.en_US
dc.contributor.authorAykanat, Cevdeten_US
dc.date.accessioned2016-02-08T10:24:43Z
dc.date.available2016-02-08T10:24:43Zen_US
dc.date.issued2004en_US
dc.departmentDepartment of Computer Engineeringen_US
dc.description.abstractThis paper addresses the problem of one-dimensional partitioning of structurally unsymmetric square and rectangular sparse matrices for parallel matrix-vector and matrix-transpose-vector multiplies. The objective is to minimize the communication cost while maintaining the balance on computational loads of processors. Most of the existing partitioning models consider only the total message volume hoping that minimizing this communication-cost metric is likely to reduce other metrics. However, the total message latency (start-up time) may be more important than the total message volume. Furthermore, the maximum message volume and latency handled by a single processor are also important metrics. We propose a two-phase approach that encapsulates all these four communication-cost metrics. The objective in the first phase is to minimize the total message volume while maintaining the computational-load balance. The objective in the second phase is to encapsulate the remaining three communication-cost metrics. We propose communication-hypergraph and partitioning models for the second phase. We then present several methods for partitioning communication hypergraphs. Experiments on a wide range of test matrices show that the proposed approach yields very effective partitioning results. A parallel implementation on a PC cluster verifies that the theoretical improvements shown by partitioning results hold in practice.en_US
dc.identifier.doi10.1137/S1064827502410463en_US
dc.identifier.issn1064-8275
dc.identifier.issn1095-7197
dc.identifier.urihttp://hdl.handle.net/11693/24144en_US
dc.language.isoEnglishen_US
dc.publisherSIAMen_US
dc.relation.isversionofhttp://dx.doi.org/10.1137/S1064827502410463en_US
dc.source.titleSIAM Journal on Scientific Computingen_US
dc.subjectCommunication hypergraphen_US
dc.subjectHypergraph partitioningen_US
dc.subjectIterative methoden_US
dc.subjectMatrix partitioningen_US
dc.subjectMatrix-vector multiplyen_US
dc.subjectMessage latencyen_US
dc.subjectMessage volumeen_US
dc.subjectParallel computingen_US
dc.subjectRectangular matrixen_US
dc.subjectStructurally unsymmetric matrixen_US
dc.subjectCommunication hypergraphen_US
dc.subjectHypergraph partitioningen_US
dc.subjectMatrix partitioningen_US
dc.subjectMatrix-vector multiplyen_US
dc.subjectMessage latencyen_US
dc.subjectMessage volumeen_US
dc.subjectRectangular matrixen_US
dc.subjectStructurally unsymmetric matrixen_US
dc.subjectComputational methodsen_US
dc.subjectCostsen_US
dc.subjectGraph theoryen_US
dc.subjectIterative methodsen_US
dc.subjectMetric systemen_US
dc.subjectParallel processing systemsen_US
dc.subjectVectorsen_US
dc.subjectMatrix algebraen_US
dc.titleEncapsulating multiple communication-cost metrics in partitioning sparse rectangular matrices for parallel matrix-vector multipliesen_US
dc.typeArticleen_US

Files

Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Encapsulating multiple communication-cost metrics in partitioning sparse rectangular matrices for parallel matrix-vector multiplies.pdf
Size:
262.52 KB
Format:
Adobe Portable Document Format
Description:
Full printable version
Download

Collections

Department of Computer Engineering