Composing low-overhead scheduling strategies for improving performance of scientific applications

Vivek Kale, William D. Gropp

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Many different sources of overheads impact the efficiency of a scheduling strategy applied to a parallel loop within a scientific application. In prior work, we handled these overheads using multiple loop scheduling strategies, with each scheduling strategy focusing on mitigating a subset of the overheads. However, mitigating the impact of one source of overhead can lead to an increase in the impact of another source of overhead, and vice versa. In this work, we show that in order to improve efficiency of loop scheduling strategies, one must adapt the loop scheduling strategies so as to handle all overheads simultaneously. To show this, we describe a composition of our existing loop scheduling strategies, and experiment with the composed scheduling strategy on standard benchmarks and application codes. Applying the composed scheduling strategy to three MPI+OpenMP scientific codes run on a cluster of SMPs improves performance an average of 31% over standard OpenMP static scheduling.

Original languageEnglish (US)
Title of host publicationOpenMP
Subtitle of host publicationHeterogenous Execution and Data Movements - 11th International Workshop on OpenMP, IWOMP 2015, Proceedings
EditorsChristian Terboven, Pablo Reble, Matthias S. Müller, Barbara M. Chapman, Bronis R. de Supinski
PublisherSpringer
Pages18-29
Number of pages12
ISBN (Print)9783319245942
DOIs
StatePublished - 2015
Event11th International Workshop on OpenMP, IWOMP 2015 - Aachen, Germany
Duration: Oct 1 2015Oct 2 2015

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9342
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other11th International Workshop on OpenMP, IWOMP 2015
Country/TerritoryGermany
CityAachen
Period10/1/1510/2/15

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Fingerprint

Dive into the research topics of 'Composing low-overhead scheduling strategies for improving performance of scientific applications'. Together they form a unique fingerprint.

Cite this