Performance of asynchronous optimized Schwarz with one-sided communication

Ichitaro Yamazaki; Edmond Chow; Aurelien Bouteiller; Jack Dongarra

doi:10.1016/j.parco.2019.05.004

Performance of asynchronous optimized Schwarz with one-sided communication

Ichitaro Yamazaki, Edmond Chow, Aurelien Bouteiller, Jack Dongarra

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

20 Scopus citations

Abstract

In asynchronous iterative methods on distributed-memory computers, processes update their local solutions using data from other processes without an implicit or explicit global synchronization that corresponds to advancing the global iteration counter. In this work, we test the asynchronous optimized Schwarz domain-decomposition iterative method using various one-sided (remote direct memory access) communication schemes with passive target completion. The results show that when one-sided communication is well-supported, the asynchronous version of optimized Schwarz can outperform the synchronous version even for perfectly balanced partitionings of the problem on a supercomputer with uniform nodes.

Original language	English
Pages (from-to)	66-81
Number of pages	16
Journal	Parallel Computing
Volume	86
DOIs	https://doi.org/10.1016/j.parco.2019.05.004
State	Published - Aug 2019
Externally published	Yes

Funding

This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research , Applied Mathematics program under Award Numbers # DE-SC0016513 and # DE-SC-0016564 . This research used resources of the National Energy Research Scientific Computing Center (NERSC), a U.S. Department of Energy Office of Science User Facility operated under Contract No. DE-AC02-05CH11231.

Funders	Funder number
U.S. Department of Energy Office of Science
U.S. Department of Energy
Office of Science
Advanced Scientific Computing Research	DE-SC-0016564, DE-SC0016513

Access to Document

10.1016/j.parco.2019.05.004

Cite this

@article{2bcdb5e7735249ee820342eb52cbe594,

title = "Performance of asynchronous optimized Schwarz with one-sided communication",

abstract = "In asynchronous iterative methods on distributed-memory computers, processes update their local solutions using data from other processes without an implicit or explicit global synchronization that corresponds to advancing the global iteration counter. In this work, we test the asynchronous optimized Schwarz domain-decomposition iterative method using various one-sided (remote direct memory access) communication schemes with passive target completion. The results show that when one-sided communication is well-supported, the asynchronous version of optimized Schwarz can outperform the synchronous version even for perfectly balanced partitionings of the problem on a supercomputer with uniform nodes.",

author = "Ichitaro Yamazaki and Edmond Chow and Aurelien Bouteiller and Jack Dongarra",

note = "Publisher Copyright: {\textcopyright} 2019",

year = "2019",

month = aug,

doi = "10.1016/j.parco.2019.05.004",

language = "English",

volume = "86",

pages = "66--81",

journal = "Parallel Computing",

issn = "0167-8191",

publisher = "Elsevier",

}

TY - JOUR

T1 - Performance of asynchronous optimized Schwarz with one-sided communication

AU - Yamazaki, Ichitaro

AU - Chow, Edmond

AU - Bouteiller, Aurelien

AU - Dongarra, Jack

PY - 2019/8

Y1 - 2019/8

N2 - In asynchronous iterative methods on distributed-memory computers, processes update their local solutions using data from other processes without an implicit or explicit global synchronization that corresponds to advancing the global iteration counter. In this work, we test the asynchronous optimized Schwarz domain-decomposition iterative method using various one-sided (remote direct memory access) communication schemes with passive target completion. The results show that when one-sided communication is well-supported, the asynchronous version of optimized Schwarz can outperform the synchronous version even for perfectly balanced partitionings of the problem on a supercomputer with uniform nodes.

AB - In asynchronous iterative methods on distributed-memory computers, processes update their local solutions using data from other processes without an implicit or explicit global synchronization that corresponds to advancing the global iteration counter. In this work, we test the asynchronous optimized Schwarz domain-decomposition iterative method using various one-sided (remote direct memory access) communication schemes with passive target completion. The results show that when one-sided communication is well-supported, the asynchronous version of optimized Schwarz can outperform the synchronous version even for perfectly balanced partitionings of the problem on a supercomputer with uniform nodes.

UR - http://www.scopus.com/inward/record.url?scp=85066284826&partnerID=8YFLogxK

U2 - 10.1016/j.parco.2019.05.004

DO - 10.1016/j.parco.2019.05.004

M3 - Article

AN - SCOPUS:85066284826

SN - 0167-8191

VL - 86

SP - 66

EP - 81

JO - Parallel Computing

JF - Parallel Computing

ER -

Performance of asynchronous optimized Schwarz with one-sided communication

Abstract

Funding

Access to Document

Other files and links

Fingerprint

Cite this