Accelerating Scientific Applications on Heterogeneous Systems with HybridOMP

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

High Performance Computing relies on accelerators (such as GPGPUs) to achieve fast execution of scientific applications. Traditionally these accelerators have been programmed with specialized languages, such as CUDA or OpenCL. In recent years, OpenMP emerged as a promising alternative for supporting accelerators, providing advantages such as maintaining a single code base for the host and different accelerator types and providing a simple way to extend support for accelerators to existing code. Efficiently using this support requires solving several challenges, related to performance, work partitioning, and concurrent execution on multiple device types. In this paper, we discuss these challenges and introduce a library, HybridOMP, that addresses several of them, thus enabling the effective use of OpenMP for accelerators. We apply HybridOMP to a scientific application, PlasCom2, that has not previously been able to use accelerators. Experiments on three architectures show that HybridOMP results in performance gains of up to 10x compared to CPU-only execution. Concurrent execution on the host and GPU resulted in additional gains of up to 10% compared to running on the GPU only.

Original languageEnglish (US)
Title of host publicationHigh Performance Computing for Computational Science – VECPAR 2018 - 13th International Conference, Revised Selected Papers
EditorsVeronica Gil-Costa, Hermes Senger, Osni Marques, Rogerio Garcia, Tatiana Pinheiro de Brito, Rogério Iope, Silvio Stanzani
PublisherSpringer-Verlag
Pages174-187
Number of pages14
ISBN (Print)9783030159955
DOIs
StatePublished - Jan 1 2019
Event13th International Conference on High Performance Computing in Computational Science, VECPAR 2018 - São Pedro, Brazil
Duration: Sep 17 2018Sep 19 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11333 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference13th International Conference on High Performance Computing in Computational Science, VECPAR 2018
CountryBrazil
CitySão Pedro
Period9/17/189/19/18

Fingerprint

Heterogeneous Systems
Accelerator
Particle accelerators
OpenMP
Concurrent
GPGPU
Program processors
Partitioning
High Performance
Computing
Alternatives
Experiment

Keywords

  • Accelerators
  • GPGPU
  • Heterogeneous computing
  • OpenMP

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Diener, M., Bodony, D. J., & Kale, L. V. (2019). Accelerating Scientific Applications on Heterogeneous Systems with HybridOMP. In V. Gil-Costa, H. Senger, O. Marques, R. Garcia, T. P. de Brito, R. Iope, & S. Stanzani (Eds.), High Performance Computing for Computational Science – VECPAR 2018 - 13th International Conference, Revised Selected Papers (pp. 174-187). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11333 LNCS). Springer-Verlag. https://doi.org/10.1007/978-3-030-15996-2_13

Accelerating Scientific Applications on Heterogeneous Systems with HybridOMP. / Diener, Matthias; Bodony, Daniel J; Kale, Laxmikant V.

High Performance Computing for Computational Science – VECPAR 2018 - 13th International Conference, Revised Selected Papers. ed. / Veronica Gil-Costa; Hermes Senger; Osni Marques; Rogerio Garcia; Tatiana Pinheiro de Brito; Rogério Iope; Silvio Stanzani. Springer-Verlag, 2019. p. 174-187 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11333 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Diener, M, Bodony, DJ & Kale, LV 2019, Accelerating Scientific Applications on Heterogeneous Systems with HybridOMP. in V Gil-Costa, H Senger, O Marques, R Garcia, TP de Brito, R Iope & S Stanzani (eds), High Performance Computing for Computational Science – VECPAR 2018 - 13th International Conference, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11333 LNCS, Springer-Verlag, pp. 174-187, 13th International Conference on High Performance Computing in Computational Science, VECPAR 2018, São Pedro, Brazil, 9/17/18. https://doi.org/10.1007/978-3-030-15996-2_13
Diener M, Bodony DJ, Kale LV. Accelerating Scientific Applications on Heterogeneous Systems with HybridOMP. In Gil-Costa V, Senger H, Marques O, Garcia R, de Brito TP, Iope R, Stanzani S, editors, High Performance Computing for Computational Science – VECPAR 2018 - 13th International Conference, Revised Selected Papers. Springer-Verlag. 2019. p. 174-187. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-15996-2_13
Diener, Matthias ; Bodony, Daniel J ; Kale, Laxmikant V. / Accelerating Scientific Applications on Heterogeneous Systems with HybridOMP. High Performance Computing for Computational Science – VECPAR 2018 - 13th International Conference, Revised Selected Papers. editor / Veronica Gil-Costa ; Hermes Senger ; Osni Marques ; Rogerio Garcia ; Tatiana Pinheiro de Brito ; Rogério Iope ; Silvio Stanzani. Springer-Verlag, 2019. pp. 174-187 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{3b3f3aa3bf8945bf80afd69a331ffa8d,
title = "Accelerating Scientific Applications on Heterogeneous Systems with HybridOMP",
abstract = "High Performance Computing relies on accelerators (such as GPGPUs) to achieve fast execution of scientific applications. Traditionally these accelerators have been programmed with specialized languages, such as CUDA or OpenCL. In recent years, OpenMP emerged as a promising alternative for supporting accelerators, providing advantages such as maintaining a single code base for the host and different accelerator types and providing a simple way to extend support for accelerators to existing code. Efficiently using this support requires solving several challenges, related to performance, work partitioning, and concurrent execution on multiple device types. In this paper, we discuss these challenges and introduce a library, HybridOMP, that addresses several of them, thus enabling the effective use of OpenMP for accelerators. We apply HybridOMP to a scientific application, PlasCom2, that has not previously been able to use accelerators. Experiments on three architectures show that HybridOMP results in performance gains of up{\^A} to 10x compared to CPU-only execution. Concurrent execution on the host and GPU resulted in additional gains of up{\^A} to 10{\%} compared to running on the GPU only.",
keywords = "Accelerators, GPGPU, Heterogeneous computing, OpenMP",
author = "Matthias Diener and Bodony, {Daniel J} and Kale, {Laxmikant V}",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-3-030-15996-2_13",
language = "English (US)",
isbn = "9783030159955",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer-Verlag",
pages = "174--187",
editor = "Veronica Gil-Costa and Hermes Senger and Osni Marques and Rogerio Garcia and {de Brito}, {Tatiana Pinheiro} and Rog{\'e}rio Iope and Silvio Stanzani",
booktitle = "High Performance Computing for Computational Science – VECPAR 2018 - 13th International Conference, Revised Selected Papers",

}

TY - GEN

T1 - Accelerating Scientific Applications on Heterogeneous Systems with HybridOMP

AU - Diener, Matthias

AU - Bodony, Daniel J

AU - Kale, Laxmikant V

PY - 2019/1/1

Y1 - 2019/1/1

N2 - High Performance Computing relies on accelerators (such as GPGPUs) to achieve fast execution of scientific applications. Traditionally these accelerators have been programmed with specialized languages, such as CUDA or OpenCL. In recent years, OpenMP emerged as a promising alternative for supporting accelerators, providing advantages such as maintaining a single code base for the host and different accelerator types and providing a simple way to extend support for accelerators to existing code. Efficiently using this support requires solving several challenges, related to performance, work partitioning, and concurrent execution on multiple device types. In this paper, we discuss these challenges and introduce a library, HybridOMP, that addresses several of them, thus enabling the effective use of OpenMP for accelerators. We apply HybridOMP to a scientific application, PlasCom2, that has not previously been able to use accelerators. Experiments on three architectures show that HybridOMP results in performance gains of up to 10x compared to CPU-only execution. Concurrent execution on the host and GPU resulted in additional gains of up to 10% compared to running on the GPU only.

AB - High Performance Computing relies on accelerators (such as GPGPUs) to achieve fast execution of scientific applications. Traditionally these accelerators have been programmed with specialized languages, such as CUDA or OpenCL. In recent years, OpenMP emerged as a promising alternative for supporting accelerators, providing advantages such as maintaining a single code base for the host and different accelerator types and providing a simple way to extend support for accelerators to existing code. Efficiently using this support requires solving several challenges, related to performance, work partitioning, and concurrent execution on multiple device types. In this paper, we discuss these challenges and introduce a library, HybridOMP, that addresses several of them, thus enabling the effective use of OpenMP for accelerators. We apply HybridOMP to a scientific application, PlasCom2, that has not previously been able to use accelerators. Experiments on three architectures show that HybridOMP results in performance gains of up to 10x compared to CPU-only execution. Concurrent execution on the host and GPU resulted in additional gains of up to 10% compared to running on the GPU only.

KW - Accelerators

KW - GPGPU

KW - Heterogeneous computing

KW - OpenMP

UR - http://www.scopus.com/inward/record.url?scp=85064640353&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85064640353&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-15996-2_13

DO - 10.1007/978-3-030-15996-2_13

M3 - Conference contribution

SN - 9783030159955

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

SP - 174

EP - 187

BT - High Performance Computing for Computational Science – VECPAR 2018 - 13th International Conference, Revised Selected Papers

A2 - Gil-Costa, Veronica

A2 - Senger, Hermes

A2 - Marques, Osni

A2 - Garcia, Rogerio

A2 - de Brito, Tatiana Pinheiro

A2 - Iope, Rogério

A2 - Stanzani, Silvio

PB - Springer-Verlag

ER -