An experimental study of global and local search algorithms in empirical performance tuning

Prasanna Balaprakash; Stefan M. Wild; Paul D. Hovland

doi:10.1007/978-3-642-38718-0_26

An experimental study of global and local search algorithms in empirical performance tuning

Prasanna Balaprakash, Stefan M. Wild, Paul D. Hovland

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

14 Scopus citations

Abstract

The increasing complexity, heterogeneity, and rapid evolution of modern computer architectures present obstacles for achieving high performance of scientific codes on different machines. Empirical performance tuning is a viable approach to obtain high-performing code variants based on their measured performance on the target machine. In previous work, we formulated the search for the best code variant as a numerical optimization problem. Two classes of algorithms are available to tackle this problem: global and local algorithms. We present an experimental study of some global and local search algorithms on a number of problems from the recently introduced SPAPT test suite. We show that local search algorithms are particularly attractive, where finding high-preforming code variants in a short computation time is crucial.

Original language	English
Title of host publication	High Performance Computing for Computational Science, VECPAR 2012 - 10th International Conference, Revised Selected Papers
Pages	261-269
Number of pages	9
DOIs	https://doi.org/10.1007/978-3-642-38718-0_26
State	Published - 2013
Externally published	Yes
Event	10th International Conference on High Performance Computing for Computational Science, VECPAR 2012 - Kobe, Japan Duration: Jul 17 2012 → Jul 20 2012

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	7851 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	10th International Conference on High Performance Computing for Computational Science, VECPAR 2012
Country/Territory	Japan
City	Kobe
Period	07/17/12 → 07/20/12

Funding

This paper has been created by UChicago Argonne, LLC, Operator of Argonne National Laboratory (“Argonne”). Argonne, a U.S. Department of Energy Office of Science laboratory, is operated under Contract No. DE-AC02-06CH11357.

Keywords

automatic performance tuning
black-box optimization
search

Access to Document

10.1007/978-3-642-38718-0_26

Cite this

Balaprakash, P., Wild, S. M., & Hovland, P. D. (2013). An experimental study of global and local search algorithms in empirical performance tuning. In High Performance Computing for Computational Science, VECPAR 2012 - 10th International Conference, Revised Selected Papers (pp. 261-269). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7851 LNCS). https://doi.org/10.1007/978-3-642-38718-0_26

Balaprakash, Prasanna ; Wild, Stefan M. ; Hovland, Paul D. / An experimental study of global and local search algorithms in empirical performance tuning. High Performance Computing for Computational Science, VECPAR 2012 - 10th International Conference, Revised Selected Papers. 2013. pp. 261-269 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{b16107266add417184eb5d3be7467a9f,

title = "An experimental study of global and local search algorithms in empirical performance tuning",

abstract = "The increasing complexity, heterogeneity, and rapid evolution of modern computer architectures present obstacles for achieving high performance of scientific codes on different machines. Empirical performance tuning is a viable approach to obtain high-performing code variants based on their measured performance on the target machine. In previous work, we formulated the search for the best code variant as a numerical optimization problem. Two classes of algorithms are available to tackle this problem: global and local algorithms. We present an experimental study of some global and local search algorithms on a number of problems from the recently introduced SPAPT test suite. We show that local search algorithms are particularly attractive, where finding high-preforming code variants in a short computation time is crucial.",

keywords = "automatic performance tuning, black-box optimization, search",

author = "Prasanna Balaprakash and Wild, {Stefan M.} and Hovland, {Paul D.}",

year = "2013",

doi = "10.1007/978-3-642-38718-0_26",

language = "English",

isbn = "9783642387173",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

pages = "261--269",

booktitle = "High Performance Computing for Computational Science, VECPAR 2012 - 10th International Conference, Revised Selected Papers",

note = "10th International Conference on High Performance Computing for Computational Science, VECPAR 2012 ; Conference date: 17-07-2012 Through 20-07-2012",

}

Balaprakash, P, Wild, SM & Hovland, PD 2013, An experimental study of global and local search algorithms in empirical performance tuning. in High Performance Computing for Computational Science, VECPAR 2012 - 10th International Conference, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 7851 LNCS, pp. 261-269, 10th International Conference on High Performance Computing for Computational Science, VECPAR 2012, Kobe, Japan, 07/17/12. https://doi.org/10.1007/978-3-642-38718-0_26

An experimental study of global and local search algorithms in empirical performance tuning. / Balaprakash, Prasanna; Wild, Stefan M.; Hovland, Paul D.
High Performance Computing for Computational Science, VECPAR 2012 - 10th International Conference, Revised Selected Papers. 2013. p. 261-269 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7851 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - An experimental study of global and local search algorithms in empirical performance tuning

AU - Balaprakash, Prasanna

AU - Wild, Stefan M.

AU - Hovland, Paul D.

PY - 2013

Y1 - 2013

N2 - The increasing complexity, heterogeneity, and rapid evolution of modern computer architectures present obstacles for achieving high performance of scientific codes on different machines. Empirical performance tuning is a viable approach to obtain high-performing code variants based on their measured performance on the target machine. In previous work, we formulated the search for the best code variant as a numerical optimization problem. Two classes of algorithms are available to tackle this problem: global and local algorithms. We present an experimental study of some global and local search algorithms on a number of problems from the recently introduced SPAPT test suite. We show that local search algorithms are particularly attractive, where finding high-preforming code variants in a short computation time is crucial.

AB - The increasing complexity, heterogeneity, and rapid evolution of modern computer architectures present obstacles for achieving high performance of scientific codes on different machines. Empirical performance tuning is a viable approach to obtain high-performing code variants based on their measured performance on the target machine. In previous work, we formulated the search for the best code variant as a numerical optimization problem. Two classes of algorithms are available to tackle this problem: global and local algorithms. We present an experimental study of some global and local search algorithms on a number of problems from the recently introduced SPAPT test suite. We show that local search algorithms are particularly attractive, where finding high-preforming code variants in a short computation time is crucial.

KW - automatic performance tuning

KW - black-box optimization

KW - search

UR - http://www.scopus.com/inward/record.url?scp=84883288790&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-38718-0_26

DO - 10.1007/978-3-642-38718-0_26

M3 - Conference contribution

AN - SCOPUS:84883288790

SN - 9783642387173

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 261

EP - 269

BT - High Performance Computing for Computational Science, VECPAR 2012 - 10th International Conference, Revised Selected Papers

T2 - 10th International Conference on High Performance Computing for Computational Science, VECPAR 2012

Y2 - 17 July 2012 through 20 July 2012

ER -

Balaprakash P, Wild SM, Hovland PD. An experimental study of global and local search algorithms in empirical performance tuning. In High Performance Computing for Computational Science, VECPAR 2012 - 10th International Conference, Revised Selected Papers. 2013. p. 261-269. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-642-38718-0_26

An experimental study of global and local search algorithms in empirical performance tuning

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this