Implementing a GPU-Portable Field Line Tracing Application with OpenMP Offload

Diego Jiménez; Javier Herrera-Mora; Markus Rampp; Erwin Laure; Esteban Meneses

doi:10.1007/978-3-031-23821-5_3

Implementing a GPU-Portable Field Line Tracing Application with OpenMP Offload

Diego Jiménez, Javier Herrera-Mora, Markus Rampp, Erwin Laure, Esteban Meneses

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

1 Scopus citations

Abstract

Accelerated computing is becoming more diverse as new vendors and architectures come into play. Although platform-specific programming models promise ease of development and better control over performance, they still restrict the portability of scientific applications. As the OpenMP offloading specification becomes adopted by more compilers, this programming model stands out as a vendor-neutral portable approach to heterogeneous programming. In this study, we port a plasma physics oriented field line tracing code from a CPU-based MPI+OpenMP approach to a GPU accelerated version, using OpenMP’s offloading capabilities. We analyze GPU performance across different vendors with respect to the original CPU version and test both prescriptive and descriptive approaches to accelerator programming. A maximum acceleration over the CPU implementation was achieved using OpenMP’s high-level offloading directives. In addition, we demonstrate portability across three different vendor GPUs with no code modifications.

Original language	English
Title of host publication	High Performance Computing - 9th Latin American Conference, CARLA 2022, Revised Selected Papers
Editors	Philippe Navaux, Carlos J. Barrios H., Carla Osthoff, Ginés Guerrero
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	31-46
Number of pages	16
ISBN (Print)	9783031238208
DOIs	https://doi.org/10.1007/978-3-031-23821-5_3
State	Published - 2022
Event	9th Latin American High Performance Computing Conference, CARLA 2022 - Porto Alegre, Brazil Duration: 26 Sep 2022 → 30 Sep 2022

Publication series

Name	Communications in Computer and Information Science
Volume	1660 CCIS
ISSN (Print)	1865-0929
ISSN (Electronic)	1865-0937

Conference

Conference	9th Latin American High Performance Computing Conference, CARLA 2022
Country/Territory	Brazil
City	Porto Alegre
Period	26/09/22 → 30/09/22

Keywords

Computational plasma physics
High performance computing
OpenMP GPU offload

Access to Document

10.1007/978-3-031-23821-5_3

Cite this

Jiménez, D., Herrera-Mora, J., Rampp, M., Laure, E., & Meneses, E. (2022). Implementing a GPU-Portable Field Line Tracing Application with OpenMP Offload. In P. Navaux, C. J. Barrios H., C. Osthoff, & G. Guerrero (Eds.), High Performance Computing - 9th Latin American Conference, CARLA 2022, Revised Selected Papers (pp. 31-46). (Communications in Computer and Information Science; Vol. 1660 CCIS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-23821-5_3

Jiménez, Diego ; Herrera-Mora, Javier ; Rampp, Markus et al. / Implementing a GPU-Portable Field Line Tracing Application with OpenMP Offload. High Performance Computing - 9th Latin American Conference, CARLA 2022, Revised Selected Papers. editor / Philippe Navaux ; Carlos J. Barrios H. ; Carla Osthoff ; Ginés Guerrero. Springer Science and Business Media Deutschland GmbH, 2022. pp. 31-46 (Communications in Computer and Information Science).

@inproceedings{1052df8894684879a8743dc39b93cc57,

title = "Implementing a GPU-Portable Field Line Tracing Application with OpenMP Offload",

abstract = "Accelerated computing is becoming more diverse as new vendors and architectures come into play. Although platform-specific programming models promise ease of development and better control over performance, they still restrict the portability of scientific applications. As the OpenMP offloading specification becomes adopted by more compilers, this programming model stands out as a vendor-neutral portable approach to heterogeneous programming. In this study, we port a plasma physics oriented field line tracing code from a CPU-based MPI+OpenMP approach to a GPU accelerated version, using OpenMP{\textquoteright}s offloading capabilities. We analyze GPU performance across different vendors with respect to the original CPU version and test both prescriptive and descriptive approaches to accelerator programming. A maximum acceleration over the CPU implementation was achieved using OpenMP{\textquoteright}s high-level offloading directives. In addition, we demonstrate portability across three different vendor GPUs with no code modifications.",

keywords = "Computational plasma physics, High performance computing, OpenMP GPU offload",

author = "Diego Jim{\'e}nez and Javier Herrera-Mora and Markus Rampp and Erwin Laure and Esteban Meneses",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 9th Latin American High Performance Computing Conference, CARLA 2022 ; Conference date: 26-09-2022 Through 30-09-2022",

year = "2022",

doi = "10.1007/978-3-031-23821-5_3",

language = "Ingl{\'e}s",

isbn = "9783031238208",

series = "Communications in Computer and Information Science",

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "31--46",

editor = "Philippe Navaux and {Barrios H.}, {Carlos J.} and Carla Osthoff and Gin{\'e}s Guerrero",

booktitle = "High Performance Computing - 9th Latin American Conference, CARLA 2022, Revised Selected Papers",

}

Jiménez, D, Herrera-Mora, J, Rampp, M, Laure, E & Meneses, E 2022, Implementing a GPU-Portable Field Line Tracing Application with OpenMP Offload. in P Navaux, CJ Barrios H., C Osthoff & G Guerrero (eds), High Performance Computing - 9th Latin American Conference, CARLA 2022, Revised Selected Papers. Communications in Computer and Information Science, vol. 1660 CCIS, Springer Science and Business Media Deutschland GmbH, pp. 31-46, 9th Latin American High Performance Computing Conference, CARLA 2022, Porto Alegre, Brazil, 26/09/22. https://doi.org/10.1007/978-3-031-23821-5_3

Implementing a GPU-Portable Field Line Tracing Application with OpenMP Offload. / Jiménez, Diego; Herrera-Mora, Javier; Rampp, Markus et al.
High Performance Computing - 9th Latin American Conference, CARLA 2022, Revised Selected Papers. ed. / Philippe Navaux; Carlos J. Barrios H.; Carla Osthoff; Ginés Guerrero. Springer Science and Business Media Deutschland GmbH, 2022. p. 31-46 (Communications in Computer and Information Science; Vol. 1660 CCIS).

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

TY - GEN

T1 - Implementing a GPU-Portable Field Line Tracing Application with OpenMP Offload

AU - Jiménez, Diego

AU - Herrera-Mora, Javier

AU - Rampp, Markus

AU - Laure, Erwin

AU - Meneses, Esteban

PY - 2022

Y1 - 2022

N2 - Accelerated computing is becoming more diverse as new vendors and architectures come into play. Although platform-specific programming models promise ease of development and better control over performance, they still restrict the portability of scientific applications. As the OpenMP offloading specification becomes adopted by more compilers, this programming model stands out as a vendor-neutral portable approach to heterogeneous programming. In this study, we port a plasma physics oriented field line tracing code from a CPU-based MPI+OpenMP approach to a GPU accelerated version, using OpenMP’s offloading capabilities. We analyze GPU performance across different vendors with respect to the original CPU version and test both prescriptive and descriptive approaches to accelerator programming. A maximum acceleration over the CPU implementation was achieved using OpenMP’s high-level offloading directives. In addition, we demonstrate portability across three different vendor GPUs with no code modifications.

AB - Accelerated computing is becoming more diverse as new vendors and architectures come into play. Although platform-specific programming models promise ease of development and better control over performance, they still restrict the portability of scientific applications. As the OpenMP offloading specification becomes adopted by more compilers, this programming model stands out as a vendor-neutral portable approach to heterogeneous programming. In this study, we port a plasma physics oriented field line tracing code from a CPU-based MPI+OpenMP approach to a GPU accelerated version, using OpenMP’s offloading capabilities. We analyze GPU performance across different vendors with respect to the original CPU version and test both prescriptive and descriptive approaches to accelerator programming. A maximum acceleration over the CPU implementation was achieved using OpenMP’s high-level offloading directives. In addition, we demonstrate portability across three different vendor GPUs with no code modifications.

KW - Computational plasma physics

KW - High performance computing

KW - OpenMP GPU offload

UR - http://www.scopus.com/inward/record.url?scp=85146697077&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-23821-5_3

DO - 10.1007/978-3-031-23821-5_3

M3 - Contribución a la conferencia

AN - SCOPUS:85146697077

SN - 9783031238208

T3 - Communications in Computer and Information Science

SP - 31

EP - 46

BT - High Performance Computing - 9th Latin American Conference, CARLA 2022, Revised Selected Papers

A2 - Navaux, Philippe

A2 - Barrios H., Carlos J.

A2 - Osthoff, Carla

A2 - Guerrero, Ginés

PB - Springer Science and Business Media Deutschland GmbH

T2 - 9th Latin American High Performance Computing Conference, CARLA 2022

Y2 - 26 September 2022 through 30 September 2022

ER -

Jiménez D, Herrera-Mora J, Rampp M, Laure E, Meneses E. Implementing a GPU-Portable Field Line Tracing Application with OpenMP Offload. In Navaux P, Barrios H. CJ, Osthoff C, Guerrero G, editors, High Performance Computing - 9th Latin American Conference, CARLA 2022, Revised Selected Papers. Springer Science and Business Media Deutschland GmbH. 2022. p. 31-46. (Communications in Computer and Information Science). doi: 10.1007/978-3-031-23821-5_3