Elastic tasks: Unifying task parallelism and SPMD parallelism with an adaptive runtime

Alina Sbîrlea, Kunal Agrawal, Vivek Sarkar

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

7 Scopus citations

Abstract

In this paper, we introduce elastic tasks, a new high-level parallel programming primitive that can be used to unify task parallelism and SPMD parallelism in a common adaptive scheduling framework. Elastic tasks are internally parallel tasks and can run on a single worker or expand to take over multiple workers. An elastic task can be an ordinary task or an SPMD region that must be executed by one or more workers simultaneously, in a tightly coupled manner. This paper demonstrates the following benefits of elastic tasks: (1) they offer theoretical guarantees: in a work-stealing environment computations complete in expected time O(W/P + S + E lg P), where E =# of elastic tasks, W =work, S =span, P =# cores. (2) they offer performance benefits in practice by co-scheduling tightly coupled parallel/ SPMD subcomputations within a single elastic task, and (3) they can adapt at runtime to the state of the application and work-load of the machine. We also introduce ElastiJ — a runtime system that includes worksharing and work-stealing scheduling algorithms to support computations with regular and elastic tasks. This scheduler dynamically decides the allocation for each elastic task in a non-centralized manner, and provides close to asymptotically optimal running times for computations with elastic tasks.

Original languageEnglish
Title of host publicationEuro-Par 2015
Subtitle of host publicationParallel Processing - 21st International Conference on Parallel and Distributed Computing, Proceedings
EditorsJesper Larsson Traff, Sascha Hunold, Francesco Versaci
PublisherSpringer Verlag
Pages491-503
Number of pages13
ISBN (Print)9783662480953
DOIs
StatePublished - 2015
Event21st International Conference on Parallel and Distributed Computing, Euro-Par 2015 - Vienna, Austria
Duration: Aug 24 2015Aug 28 2015

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9233
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference21st International Conference on Parallel and Distributed Computing, Euro-Par 2015
Country/TerritoryAustria
CityVienna
Period08/24/1508/28/15

Fingerprint

Dive into the research topics of 'Elastic tasks: Unifying task parallelism and SPMD parallelism with an adaptive runtime'. Together they form a unique fingerprint.

Cite this