Eliminating contention bottlenecks in multithreaded MPI

Hoang Vu Dang; Marc Snir; William Gropp

doi:10.1016/j.parco.2017.08.003

Eliminating contention bottlenecks in multithreaded MPI

Research output: Contribution to journal › Article › peer-review

Abstract

We explore in this paper the advantages that accrue from avoiding the use of wildcards in MPI. We show that, with this change, one can efficiently support millions of concurrently communicating light-weight threads using send-receive communication. This is achieved through a streamlined implementation of MPI and a tight coupling between the communication runtime and the thread scheduler.

Original language	English (US)
Pages (from-to)	1-23
Number of pages	23
Journal	Parallel Computing
Volume	69
DOIs	https://doi.org/10.1016/j.parco.2017.08.003
State	Published - Nov 2017

Keywords

Communication
MPI
Networking
Scheduling
User-level threads

ASJC Scopus subject areas

Software
Theoretical Computer Science
Hardware and Architecture
Computer Networks and Communications
Computer Graphics and Computer-Aided Design
Artificial Intelligence

Online availability

10.1016/j.parco.2017.08.003

Library availability

Discover UIUC Full Text

Cite this

@article{69047a45a06d4592a1810ac06da518b9,

title = "Eliminating contention bottlenecks in multithreaded MPI",

abstract = "We explore in this paper the advantages that accrue from avoiding the use of wildcards in MPI. We show that, with this change, one can efficiently support millions of concurrently communicating light-weight threads using send-receive communication. This is achieved through a streamlined implementation of MPI and a tight coupling between the communication runtime and the thread scheduler.",

keywords = "Communication, MPI, Networking, Scheduling, User-level threads",

author = "Dang, {Hoang Vu} and Marc Snir and William Gropp",

note = "Funding Information: This material is based upon work supported by the U.S. Department of Energy, Office of Science, under contract number DE-AC02-06CH11357 and by a subcontract from Sandia National Laboratories. It was performed while the second author was at the MCS Division of the Argonne National Laboratory. We thank Ron Brightwell, Pavan Balaji, Nikoli Dryden and Alex Brooks for their help with this work. Publisher Copyright: {\textcopyright} 2017 Elsevier B.V.",

year = "2017",

month = nov,

doi = "10.1016/j.parco.2017.08.003",

language = "English (US)",

volume = "69",

pages = "1--23",

journal = "Parallel Computing",

issn = "0167-8191",

publisher = "Elsevier",

}

TY - JOUR

T1 - Eliminating contention bottlenecks in multithreaded MPI

AU - Dang, Hoang Vu

AU - Snir, Marc

AU - Gropp, William

N1 - Funding Information: This material is based upon work supported by the U.S. Department of Energy, Office of Science, under contract number DE-AC02-06CH11357 and by a subcontract from Sandia National Laboratories. It was performed while the second author was at the MCS Division of the Argonne National Laboratory. We thank Ron Brightwell, Pavan Balaji, Nikoli Dryden and Alex Brooks for their help with this work. Publisher Copyright: © 2017 Elsevier B.V.

PY - 2017/11

Y1 - 2017/11

N2 - We explore in this paper the advantages that accrue from avoiding the use of wildcards in MPI. We show that, with this change, one can efficiently support millions of concurrently communicating light-weight threads using send-receive communication. This is achieved through a streamlined implementation of MPI and a tight coupling between the communication runtime and the thread scheduler.

AB - We explore in this paper the advantages that accrue from avoiding the use of wildcards in MPI. We show that, with this change, one can efficiently support millions of concurrently communicating light-weight threads using send-receive communication. This is achieved through a streamlined implementation of MPI and a tight coupling between the communication runtime and the thread scheduler.

KW - Communication

KW - MPI

KW - Networking

KW - Scheduling

KW - User-level threads

UR - http://www.scopus.com/inward/record.url?scp=85028022500&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85028022500&partnerID=8YFLogxK

U2 - 10.1016/j.parco.2017.08.003

DO - 10.1016/j.parco.2017.08.003

M3 - Article

AN - SCOPUS:85028022500

SN - 0167-8191

VL - 69

SP - 1

EP - 23

JO - Parallel Computing

JF - Parallel Computing

ER -

Eliminating contention bottlenecks in multithreaded MPI

Abstract

Keywords

ASJC Scopus subject areas

Online availability

Library availability

Related links

Fingerprint

Cite this