Dynamic scheduling for dependence loops on heterogeneous clusters

Papakonstantinou, G; Riakiotakis, I; Andronikos, T; Ciorba, FM; Chronopoulos, AT

dc.contributor.author	Papakonstantinou, G	en
dc.contributor.author	Riakiotakis, I	en
dc.contributor.author	Andronikos, T	en
dc.contributor.author	Ciorba, FM	en
dc.contributor.author	Chronopoulos, AT	en
dc.date.accessioned	2014-03-01T01:55:23Z
dc.date.available	2014-03-01T01:55:23Z
dc.date.issued	2006	en
dc.identifier.issn	10615369	en
dc.identifier.uri	https://dspace.lib.ntua.gr/xmlui/handle/123456789/27712
dc.relation.uri	http://www.scopus.com/inward/record.url?eid=2-s2.0-34249328493&partnerID=40&md5=6b51e8740a69bd4f4bbb0fe1504667aa	en
dc.subject	Dependence loops	en
dc.subject	Dynamic algorithms	en
dc.subject	Heterogeneous distributed systems	en
dc.subject	Loop scheduling	en
dc.subject	Pipelined execution	en
dc.title	Dynamic scheduling for dependence loops on heterogeneous clusters	en
heal.type	journalArticle	en
heal.publicationDate	2006	en
heal.abstract	Distributed computing systems are a viable and less expensive alternative to parallel computers. However, concurrent programming methods in distributed systems have not been studied as extensively as for parallel computers. In the past, a variety of dynamic scheduling schemes suitable for loops with independent iterations on heterogeneous computer clusters have been obtained and studied. However, no study of dynamic schemes for loops with iteration dependencies has been reported so far. In this work we study the problem of scheduling loops with iteration dependencies for heterogeneous clusters. The iteration dependencies incur an extra degree of difficulty and make the development of such schemes quite a challenge. We extend three well known dynamic schemes (CSS, TSS and DTSS) by introducing synchronization points at certain intervals so that processors compute in pipelined fashion. Our scheme is called dynamic multi-phase scheduling (DMPS) and we apply it to loops with iteration dependencies. We implemented our new scheme on a network of heterogeneous computers and studied its performance. Through extensive testing on four real-life applications we show that the proposed method is efficient for parallelizing nested loops with dependencies on heterogeneous systems. Results show that DTSS gives a speedup of 3-5 out of the ideal 7, in all cases, when applied to dependence loops. © Dynamic Publishers, Inc.	en
heal.journalName	Neural, Parallel and Scientific Computations	en
dc.identifier.volume	14	en
dc.identifier.issue	4	en
dc.identifier.spage	359	en
dc.identifier.epage	384	en