A framework for enabling openMP autotuning

Vinu Sreenivasan; Rajath Javali; Mary Hall; Prasanna Balaprakash; Thomas R.W. Scogland; Bronis R. de Supinski

doi:10.1007/978-3-030-28596-8_4

A framework for enabling openMP autotuning

Vinu Sreenivasan, Rajath Javali, Mary Hall, Prasanna Balaprakash, Thomas R.W. Scogland, Bronis R. de Supinski

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

13 Scopus citations

Abstract

This paper describes a lightweight framework that enables autotuning of OpenMP pragmas to ease performance tuning of OpenMP codes across platforms. This paper describes a prototype of the framework and demonstrates its use in identifying best-performing parallel loop schedules and number of threads for five codes from the PolyBench benchmark suite. This process is facilitated by a tool for taking a compact search-space description of pragmas to apply to the loop nest and chooses the best solution using model-based search. This tool offers the potential to achieve performance portability of OpenMP across platforms without burdening the programmer with exploring this search space manually. Performance results show that the tool identifies different selections for schedule and thread count applied to parallel loops across benchmarks, data set sizes and architectures. Performance gain over the baseline with default settings of up to 1.17×, but slowdowns of 0.5× show the importance of preserving default settings. More importantly, this experiment sets the stage for more elaborate experiments to map new OpenMP features such as GPU offloading and the new loop pragma.

Original language	English
Title of host publication	OpenMP
Subtitle of host publication	Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings
Editors	Xing Fan, Oliver Sinnen, Nasser Giacaman, Bronis R. de Supinski
Publisher	Springer Verlag
Pages	50-60
Number of pages	11
ISBN (Print)	9783030285951
DOIs	https://doi.org/10.1007/978-3-030-28596-8_4
State	Published - 2019
Externally published	Yes
Event	15th International Workshop on OpenMP, IWOMP 2019 - Auckland, New Zealand Duration: Sep 11 2019 → Sep 13 2019

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11718 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	15th International Workshop on OpenMP, IWOMP 2019
Country/Territory	New Zealand
City	Auckland
Period	09/11/19 → 09/13/19

Funding

This research was supported in part by the Exascale Computing Project (17-SC-20SC), a joint project of the U.S. Department of Energy’s Office of Science and National Nuclear Security Administration, and by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research, Scientific Discovery through Advanced Computing (SciDAC) program under the RAPIDS Subcontract Award Number 4000159989.

Keywords

Autotuning
Loop scheduling
Performance portability

Access to Document

10.1007/978-3-030-28596-8_4

Cite this

Sreenivasan, V., Javali, R., Hall, M., Balaprakash, P., Scogland, T. R. W., & de Supinski, B. R. (2019). A framework for enabling openMP autotuning. In X. Fan, O. Sinnen, N. Giacaman, & B. R. de Supinski (Eds.), OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings (pp. 50-60). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11718 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-28596-8_4

Sreenivasan, Vinu ; Javali, Rajath ; Hall, Mary et al. / A framework for enabling openMP autotuning. OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings. editor / Xing Fan ; Oliver Sinnen ; Nasser Giacaman ; Bronis R. de Supinski. Springer Verlag, 2019. pp. 50-60 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{f0248659566041c98942b251fb8e0f17,

title = "A framework for enabling openMP autotuning",

abstract = "This paper describes a lightweight framework that enables autotuning of OpenMP pragmas to ease performance tuning of OpenMP codes across platforms. This paper describes a prototype of the framework and demonstrates its use in identifying best-performing parallel loop schedules and number of threads for five codes from the PolyBench benchmark suite. This process is facilitated by a tool for taking a compact search-space description of pragmas to apply to the loop nest and chooses the best solution using model-based search. This tool offers the potential to achieve performance portability of OpenMP across platforms without burdening the programmer with exploring this search space manually. Performance results show that the tool identifies different selections for schedule and thread count applied to parallel loops across benchmarks, data set sizes and architectures. Performance gain over the baseline with default settings of up to 1.17×, but slowdowns of 0.5× show the importance of preserving default settings. More importantly, this experiment sets the stage for more elaborate experiments to map new OpenMP features such as GPU offloading and the new loop pragma.",

keywords = "Autotuning, Loop scheduling, Performance portability",

author = "Vinu Sreenivasan and Rajath Javali and Mary Hall and Prasanna Balaprakash and Scogland, {Thomas R.W.} and {de Supinski}, {Bronis R.}",

note = "Publisher Copyright: {\textcopyright} Springer Nature Switzerland AG 2019.; 15th International Workshop on OpenMP, IWOMP 2019 ; Conference date: 11-09-2019 Through 13-09-2019",

year = "2019",

doi = "10.1007/978-3-030-28596-8_4",

language = "English",

isbn = "9783030285951",

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

publisher = "Springer Verlag",

pages = "50--60",

editor = "Xing Fan and Oliver Sinnen and Nasser Giacaman and {de Supinski}, {Bronis R.}",

booktitle = "OpenMP",

}

Sreenivasan, V, Javali, R, Hall, M, Balaprakash, P, Scogland, TRW & de Supinski, BR 2019, A framework for enabling openMP autotuning. in X Fan, O Sinnen, N Giacaman & BR de Supinski (eds), OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11718 LNCS, Springer Verlag, pp. 50-60, 15th International Workshop on OpenMP, IWOMP 2019, Auckland, New Zealand, 09/11/19. https://doi.org/10.1007/978-3-030-28596-8_4

A framework for enabling openMP autotuning. / Sreenivasan, Vinu; Javali, Rajath; Hall, Mary et al.
OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings. ed. / Xing Fan; Oliver Sinnen; Nasser Giacaman; Bronis R. de Supinski. Springer Verlag, 2019. p. 50-60 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11718 LNCS).

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

TY - GEN

T1 - A framework for enabling openMP autotuning

AU - Sreenivasan, Vinu

AU - Javali, Rajath

AU - Hall, Mary

AU - Balaprakash, Prasanna

AU - Scogland, Thomas R.W.

AU - de Supinski, Bronis R.

N1 - Publisher Copyright: © Springer Nature Switzerland AG 2019.

PY - 2019

Y1 - 2019

N2 - This paper describes a lightweight framework that enables autotuning of OpenMP pragmas to ease performance tuning of OpenMP codes across platforms. This paper describes a prototype of the framework and demonstrates its use in identifying best-performing parallel loop schedules and number of threads for five codes from the PolyBench benchmark suite. This process is facilitated by a tool for taking a compact search-space description of pragmas to apply to the loop nest and chooses the best solution using model-based search. This tool offers the potential to achieve performance portability of OpenMP across platforms without burdening the programmer with exploring this search space manually. Performance results show that the tool identifies different selections for schedule and thread count applied to parallel loops across benchmarks, data set sizes and architectures. Performance gain over the baseline with default settings of up to 1.17×, but slowdowns of 0.5× show the importance of preserving default settings. More importantly, this experiment sets the stage for more elaborate experiments to map new OpenMP features such as GPU offloading and the new loop pragma.

AB - This paper describes a lightweight framework that enables autotuning of OpenMP pragmas to ease performance tuning of OpenMP codes across platforms. This paper describes a prototype of the framework and demonstrates its use in identifying best-performing parallel loop schedules and number of threads for five codes from the PolyBench benchmark suite. This process is facilitated by a tool for taking a compact search-space description of pragmas to apply to the loop nest and chooses the best solution using model-based search. This tool offers the potential to achieve performance portability of OpenMP across platforms without burdening the programmer with exploring this search space manually. Performance results show that the tool identifies different selections for schedule and thread count applied to parallel loops across benchmarks, data set sizes and architectures. Performance gain over the baseline with default settings of up to 1.17×, but slowdowns of 0.5× show the importance of preserving default settings. More importantly, this experiment sets the stage for more elaborate experiments to map new OpenMP features such as GPU offloading and the new loop pragma.

KW - Autotuning

KW - Loop scheduling

KW - Performance portability

UR - http://www.scopus.com/inward/record.url?scp=85072853249&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-28596-8_4

DO - 10.1007/978-3-030-28596-8_4

M3 - Conference contribution

AN - SCOPUS:85072853249

SN - 9783030285951

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

SP - 50

EP - 60

BT - OpenMP

A2 - Fan, Xing

A2 - Sinnen, Oliver

A2 - Giacaman, Nasser

A2 - de Supinski, Bronis R.

PB - Springer Verlag

T2 - 15th International Workshop on OpenMP, IWOMP 2019

Y2 - 11 September 2019 through 13 September 2019

ER -

Sreenivasan V, Javali R, Hall M, Balaprakash P, Scogland TRW, de Supinski BR. A framework for enabling openMP autotuning. In Fan X, Sinnen O, Giacaman N, de Supinski BR, editors, OpenMP: Conquering the Full Hardware Spectrum - 15th International Workshop on OpenMP, IWOMP 2019, Proceedings. Springer Verlag. 2019. p. 50-60. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-28596-8_4

A framework for enabling openMP autotuning

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this