Iterative-method performance evaluation for multiple vectors associated with a large-scale sparse matrix

Seigo Imamura, Kenji Ono, Mitsuo Yokokawa

Research output: Contribution to journalArticle

1 Citation (Scopus)

Abstract

Ensemble computing, which is an instance of capacity computing, is an effective computing scenario for exascale parallel supercomputers. In ensemble computing, there are multiple linear systems associated with a common coefficient matrix. We improve the performance of iterative solvers for multiple vectors by solving them at the same time, that is, by solving for the product of the matrices. We implemented several iterative methods and compared their performance. The maximum performance on Sparc VIIIfx was 7.6 times higher than that of a naïve implementation. Finally, to deal with the different convergence processes of linear systems, we introduced a control method to eliminate the calculation of already converged vectors.

Original languageEnglish
Pages (from-to)395-401
Number of pages7
JournalInternational Journal of Computational Fluid Dynamics
Volume30
Issue number6
DOIs
Publication statusPublished - Jul 2 2016

Fingerprint

Iterative methods
Linear systems
linear systems
evaluation
Supercomputers
supercomputers
coefficients
products
matrices

All Science Journal Classification (ASJC) codes

  • Computational Mechanics
  • Aerospace Engineering
  • Condensed Matter Physics
  • Energy Engineering and Power Technology
  • Mechanics of Materials
  • Mechanical Engineering

Cite this

Iterative-method performance evaluation for multiple vectors associated with a large-scale sparse matrix. / Imamura, Seigo; Ono, Kenji; Yokokawa, Mitsuo.

In: International Journal of Computational Fluid Dynamics, Vol. 30, No. 6, 02.07.2016, p. 395-401.

Research output: Contribution to journalArticle

@article{3737656fad224952bd9c0a760d54d49c,
title = "Iterative-method performance evaluation for multiple vectors associated with a large-scale sparse matrix",
abstract = "Ensemble computing, which is an instance of capacity computing, is an effective computing scenario for exascale parallel supercomputers. In ensemble computing, there are multiple linear systems associated with a common coefficient matrix. We improve the performance of iterative solvers for multiple vectors by solving them at the same time, that is, by solving for the product of the matrices. We implemented several iterative methods and compared their performance. The maximum performance on Sparc VIIIfx was 7.6 times higher than that of a na{\"i}ve implementation. Finally, to deal with the different convergence processes of linear systems, we introduced a control method to eliminate the calculation of already converged vectors.",
author = "Seigo Imamura and Kenji Ono and Mitsuo Yokokawa",
year = "2016",
month = "7",
day = "2",
doi = "10.1080/10618562.2016.1234046",
language = "English",
volume = "30",
pages = "395--401",
journal = "International Journal of Computational Fluid Dynamics",
issn = "1061-8562",
publisher = "Taylor and Francis Ltd.",
number = "6",

}

TY - JOUR

T1 - Iterative-method performance evaluation for multiple vectors associated with a large-scale sparse matrix

AU - Imamura, Seigo

AU - Ono, Kenji

AU - Yokokawa, Mitsuo

PY - 2016/7/2

Y1 - 2016/7/2

N2 - Ensemble computing, which is an instance of capacity computing, is an effective computing scenario for exascale parallel supercomputers. In ensemble computing, there are multiple linear systems associated with a common coefficient matrix. We improve the performance of iterative solvers for multiple vectors by solving them at the same time, that is, by solving for the product of the matrices. We implemented several iterative methods and compared their performance. The maximum performance on Sparc VIIIfx was 7.6 times higher than that of a naïve implementation. Finally, to deal with the different convergence processes of linear systems, we introduced a control method to eliminate the calculation of already converged vectors.

AB - Ensemble computing, which is an instance of capacity computing, is an effective computing scenario for exascale parallel supercomputers. In ensemble computing, there are multiple linear systems associated with a common coefficient matrix. We improve the performance of iterative solvers for multiple vectors by solving them at the same time, that is, by solving for the product of the matrices. We implemented several iterative methods and compared their performance. The maximum performance on Sparc VIIIfx was 7.6 times higher than that of a naïve implementation. Finally, to deal with the different convergence processes of linear systems, we introduced a control method to eliminate the calculation of already converged vectors.

UR - http://www.scopus.com/inward/record.url?scp=84991254641&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84991254641&partnerID=8YFLogxK

U2 - 10.1080/10618562.2016.1234046

DO - 10.1080/10618562.2016.1234046

M3 - Article

AN - SCOPUS:84991254641

VL - 30

SP - 395

EP - 401

JO - International Journal of Computational Fluid Dynamics

JF - International Journal of Computational Fluid Dynamics

SN - 1061-8562

IS - 6

ER -