Skip to main navigation Skip to search Skip to main content

SHCOLL - A standalone implementation of OpenSHMEM-style collectives API

  • Srđan Milaković
  • , Zoran Budimlić
  • , Howard Pritchard
  • , Anthony Curtis
  • , Barbara Chapman
  • , Vivek Sarkar
  • Rice University
  • Los Alamos National Laboratory
  • Stony Brook University
  • Georgia Institute of Technology

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

1 Scopus citations

Abstract

The performance of collective operations has a large impact on overall performance in many HPC applications. Implementing multiple algorithms and selecting optimal one depending on message size and the number of processes involved in the operation is essential to achieve good performance. In this paper, we will present SHCOLL, a collective routines library that was developed on top of OpenSHMEM API point to point operations: puts, gets, atomic memory update, and memory synchronization routines. The library is designed to serve as a plug-in to OpenSHMEM implementations and will be used by the OSSS OpenSHMEM reference implementation to support OpenSHMEM collective operations. In this paper, we describe the algorithms that have been incorporated in the implementation of each OpenSHMEM API collective routine and evaluate them on a Cray XC30 system. For long messages, SHCOLL shows an improvement by up to a factor of 12 compared to the vendor’s implementation. We also discuss future development of the library, as well as how it will be incorporated into the OSSS OpenSHMEM reference implementation.

Original languageEnglish
Title of host publicationOpenSHMEM and Related Technologies. OpenSHMEM in the Era of Extreme Heterogeneity - 5th Workshop, OpenSHMEM 2018, Revised Selected Papers
EditorsSwaroop Pophale, Neena Imam, Ferrol Aderholdt, Manjunath Gorentla Venkata
PublisherSpringer Verlag
Pages90-106
Number of pages17
ISBN (Print)9783030049171
DOIs
StatePublished - 2019
Event5th Workshop on OpenSHMEM and Related Technologies, 2018 - Baltimore, United States
Duration: Aug 21 2018Aug 23 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11283 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference5th Workshop on OpenSHMEM and Related Technologies, 2018
Country/TerritoryUnited States
CityBaltimore
Period08/21/1808/23/18

Fingerprint

Dive into the research topics of 'SHCOLL - A standalone implementation of OpenSHMEM-style collectives API'. Together they form a unique fingerprint.

Cite this