Communication-aware supernode shape

Goumas, G; Drosinos, N; Koziris, N

dc.contributor.author	Goumas, G	en
dc.contributor.author	Drosinos, N	en
dc.contributor.author	Koziris, N	en
dc.date.accessioned	2014-03-01T01:30:00Z
dc.date.available	2014-03-01T01:30:00Z
dc.date.issued	2009	en
dc.identifier.issn	1045-9219	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/19443
dc.subject	Loop tiling	en
dc.subject	MPI	en
dc.subject	Process grid	en
dc.subject	Scheduling	en
dc.subject	Supernode transformation	en
dc.subject	Tile shape	en
dc.subject.classification	Computer Science, Theory & Methods	en
dc.subject.classification	Engineering, Electrical & Electronic	en
dc.subject.other	Cartesians	en
dc.subject.other	Loop tiling	en
dc.subject.other	MPI	en
dc.subject.other	Parallel execution time	en
dc.subject.other	Parallel executions	en
dc.subject.other	Process communication	en
dc.subject.other	Process grid	en
dc.subject.other	Selection problems	en
dc.subject.other	Supernode transformation	en
dc.subject.other	Cache memory	en
dc.subject.other	Communication	en
dc.subject.other	Optimization	en
dc.subject.other	Parallel algorithms	en
dc.title	Communication-aware supernode shape	en
heal.type	journalArticle	en
heal.identifier.primary	10.1109/TPDS.2008.114	en
heal.identifier.secondary	http://dx.doi.org/10.1109/TPDS.2008.114	en
heal.language	English	en
heal.publicationDate	2009	en
heal.abstract	In this paper we revisit the supernode-shape selection problem, that has been widely discussed in bibliography. In general, the selection of the supernode transformation greatly affects the parallel execution time of the transformed algorithm. Since the minimization of the overall parallel execution time via an appropriate supernode transformation is very difficult to accomplish, researchers have focused on scheduling-aware supernode transformations that maximize parallelism during the execution. In this paper we argue that the communication volume of the transformed algorithm is an important criterion, and its minimization should be given high priority. For this reason we define the metric of the per process communication volume and propose a method to miminize this metric by selecting a communication-aware supernode shape. Our approach is equivalent to defining a proper Cartesian process grid with MPI_Cart_Create, which means that it can be incorporated in applications in a straightforward manner. Our experimental results illustrate that by selecting the tile shape with the proposed method, the total parallel execution time is significantly reduced due to the minimization of the communication volume, despite the fact that a few more parallel execution steps are required. © 2009 IEEE.	en
heal.publisher	IEEE COMPUTER SOC	en
heal.journalName	IEEE Transactions on Parallel and Distributed Systems	en
dc.identifier.doi	10.1109/TPDS.2008.114	en
dc.identifier.isi	ISI:000263762400006	en
dc.identifier.volume	20	en
dc.identifier.issue	4	en
dc.identifier.spage	498	en
dc.identifier.epage	511	en