Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution

Hamidreza Khaleghzadeh; Muhammad  Fahad; Ravi Reddy Manumachu; Alexey Lastovetsky

doi:10.1007/978-3-030-48340-1_25

Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution

Hamidreza Khaleghzadeh, Muhammad Fahad, Ravi Reddy Manumachu, Alexey Lastovetsky

School of Computing

University College Dublin

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

Abstract

Energy is one of the most important objectives for optimization on modern heterogeneous high performance computing (HPC) platforms. The tight integration of multicore CPUs with accelerators in these platforms present several challenges to optimization of multithreaded data-parallel applications for dynamic energy.

In this work, we formulate the optimization problem of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution. We propose a solution method to solve the problem. It consists of a data-partitioning algorithm that employs load imbalancing technique to determine the workload distribution minimizing the dynamic energy consumption of the parallel execution of an application. The inputs to the algorithm are discrete dynamic energy profiles of individual computing devices.

We experimentally analyse the proposed algorithm using two multithreaded data-parallel applications, matrix multiplication and 2D fast Fourier transform. The load-imbalanced solutions provided by the algorithm achieve significant dynamic energy reductions (on the average 130% and 44%) compared to the load-balanced ones for the applications.

Original language	English
Title of host publication	Euro-Par 2019: Parallel Processing Workshops
Subtitle of host publication	Euro-Par 2019 International Workshops, Göttingen, Germany, August 26–30, 2019, Revised Selected Papers
Editors	Ulrich Schwardmann, Christian Boehme, Dora B. Heras, Valeria Cardellini, Emmanuel Jeannot, Antonio Salis, Claudio Schifanella, Ravi Reddy Manumachu, Dieter Schwamborn, Laura Ricci, Oh Sangyoon, Thomas Gruber, Laura Antonelli, Stephen L. Scott
Publisher	Springer
Pages	320-332
ISBN (Electronic)	9783030483401
ISBN (Print)	9783030483395
DOIs	https://doi.org/10.1007/978-3-030-48340-1_25
Publication status	Published - 29 May 2020
Event	Euro-Par 2019 International Workshops - Göttingen, Germany Duration: 26 Aug 2019 → 30 Aug 2019

Publication series

Name	Lecture Notes in Computer Science
Publisher	Springer
Volume	11997
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Workshop

Workshop	Euro-Par 2019 International Workshops
Country/Territory	Germany
City	Göttingen
Period	26/08/19 → 30/08/19

Keywords

High performance computing
Heterogeneous platforms
Energy of computation
Multicore CPU
GPU
Xeon Phi

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

Access to Document

10.1007/978-3-030-48340-1_25

Cite this

Khaleghzadeh, H., Fahad, M., Manumachu, R. R., & Lastovetsky, A. (2020). Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution. In U. Schwardmann, C. Boehme, D. B. Heras, V. Cardellini, E. Jeannot, A. Salis, C. Schifanella, R. R. Manumachu, D. Schwamborn, L. Ricci, O. Sangyoon, T. Gruber, L. Antonelli, & S. L. Scott (Eds.), Euro-Par 2019: Parallel Processing Workshops: Euro-Par 2019 International Workshops, Göttingen, Germany, August 26–30, 2019, Revised Selected Papers (pp. 320-332). (Lecture Notes in Computer Science; Vol. 11997). Springer. https://doi.org/10.1007/978-3-030-48340-1_25

Khaleghzadeh, Hamidreza ; Fahad, Muhammad ; Manumachu, Ravi Reddy et al. / Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution. Euro-Par 2019: Parallel Processing Workshops: Euro-Par 2019 International Workshops, Göttingen, Germany, August 26–30, 2019, Revised Selected Papers. editor / Ulrich Schwardmann ; Christian Boehme ; Dora B. Heras ; Valeria Cardellini ; Emmanuel Jeannot ; Antonio Salis ; Claudio Schifanella ; Ravi Reddy Manumachu ; Dieter Schwamborn ; Laura Ricci ; Oh Sangyoon ; Thomas Gruber ; Laura Antonelli ; Stephen L. Scott. Springer, 2020. pp. 320-332 (Lecture Notes in Computer Science).

@inproceedings{c566dd050d1d4960a7c45c1159298269,

title = "Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution",

abstract = "Energy is one of the most important objectives for optimization on modern heterogeneous high performance computing (HPC) platforms. The tight integration of multicore CPUs with accelerators in these platforms present several challenges to optimization of multithreaded data-parallel applications for dynamic energy.In this work, we formulate the optimization problem of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution. We propose a solution method to solve the problem. It consists of a data-partitioning algorithm that employs load imbalancing technique to determine the workload distribution minimizing the dynamic energy consumption of the parallel execution of an application. The inputs to the algorithm are discrete dynamic energy profiles of individual computing devices.We experimentally analyse the proposed algorithm using two multithreaded data-parallel applications, matrix multiplication and 2D fast Fourier transform. The load-imbalanced solutions provided by the algorithm achieve significant dynamic energy reductions (on the average 130% and 44%) compared to the load-balanced ones for the applications.",

keywords = "High performance computing, Heterogeneous platforms, Energy of computation, Multicore CPU, GPU, Xeon Phi",

author = "Hamidreza Khaleghzadeh and Muhammad Fahad and Manumachu, {Ravi Reddy} and Alexey Lastovetsky",

year = "2020",

month = may,

day = "29",

doi = "10.1007/978-3-030-48340-1_25",

language = "English",

isbn = "9783030483395",

series = "Lecture Notes in Computer Science",

publisher = "Springer",

pages = "320--332",

editor = "Ulrich Schwardmann and Christian Boehme and Heras, {Dora B.} and Valeria Cardellini and Emmanuel Jeannot and Antonio Salis and Claudio Schifanella and Manumachu, {Ravi Reddy} and Dieter Schwamborn and Laura Ricci and Oh Sangyoon and Thomas Gruber and Laura Antonelli and Scott, {Stephen L.}",

booktitle = "Euro-Par 2019: Parallel Processing Workshops",

note = "Euro-Par 2019 International Workshops ; Conference date: 26-08-2019 Through 30-08-2019",

}

Khaleghzadeh, H, Fahad, M, Manumachu, RR & Lastovetsky, A 2020, Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution. in U Schwardmann, C Boehme, DB Heras, V Cardellini, E Jeannot, A Salis, C Schifanella, RR Manumachu, D Schwamborn, L Ricci, O Sangyoon, T Gruber, L Antonelli & SL Scott (eds), Euro-Par 2019: Parallel Processing Workshops: Euro-Par 2019 International Workshops, Göttingen, Germany, August 26–30, 2019, Revised Selected Papers. Lecture Notes in Computer Science, vol. 11997, Springer, pp. 320-332, Euro-Par 2019 International Workshops, Göttingen, Germany, 26/08/19. https://doi.org/10.1007/978-3-030-48340-1_25

Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution. / Khaleghzadeh, Hamidreza; Fahad, Muhammad ; Manumachu, Ravi Reddy et al.
Euro-Par 2019: Parallel Processing Workshops: Euro-Par 2019 International Workshops, Göttingen, Germany, August 26–30, 2019, Revised Selected Papers. ed. / Ulrich Schwardmann; Christian Boehme; Dora B. Heras; Valeria Cardellini; Emmanuel Jeannot; Antonio Salis; Claudio Schifanella; Ravi Reddy Manumachu; Dieter Schwamborn; Laura Ricci; Oh Sangyoon; Thomas Gruber; Laura Antonelli; Stephen L. Scott. Springer, 2020. p. 320-332 (Lecture Notes in Computer Science; Vol. 11997).

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

TY - GEN

T1 - Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution

AU - Khaleghzadeh, Hamidreza

AU - Fahad, Muhammad

AU - Manumachu, Ravi Reddy

AU - Lastovetsky, Alexey

PY - 2020/5/29

Y1 - 2020/5/29

N2 - Energy is one of the most important objectives for optimization on modern heterogeneous high performance computing (HPC) platforms. The tight integration of multicore CPUs with accelerators in these platforms present several challenges to optimization of multithreaded data-parallel applications for dynamic energy.In this work, we formulate the optimization problem of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution. We propose a solution method to solve the problem. It consists of a data-partitioning algorithm that employs load imbalancing technique to determine the workload distribution minimizing the dynamic energy consumption of the parallel execution of an application. The inputs to the algorithm are discrete dynamic energy profiles of individual computing devices.We experimentally analyse the proposed algorithm using two multithreaded data-parallel applications, matrix multiplication and 2D fast Fourier transform. The load-imbalanced solutions provided by the algorithm achieve significant dynamic energy reductions (on the average 130% and 44%) compared to the load-balanced ones for the applications.

AB - Energy is one of the most important objectives for optimization on modern heterogeneous high performance computing (HPC) platforms. The tight integration of multicore CPUs with accelerators in these platforms present several challenges to optimization of multithreaded data-parallel applications for dynamic energy.In this work, we formulate the optimization problem of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution. We propose a solution method to solve the problem. It consists of a data-partitioning algorithm that employs load imbalancing technique to determine the workload distribution minimizing the dynamic energy consumption of the parallel execution of an application. The inputs to the algorithm are discrete dynamic energy profiles of individual computing devices.We experimentally analyse the proposed algorithm using two multithreaded data-parallel applications, matrix multiplication and 2D fast Fourier transform. The load-imbalanced solutions provided by the algorithm achieve significant dynamic energy reductions (on the average 130% and 44%) compared to the load-balanced ones for the applications.

KW - High performance computing

KW - Heterogeneous platforms

KW - Energy of computation

KW - Multicore CPU

KW - GPU

KW - Xeon Phi

UR - https://2019.euro-par.org/contributors/conference-poster/

U2 - 10.1007/978-3-030-48340-1_25

DO - 10.1007/978-3-030-48340-1_25

M3 - Conference contribution

SN - 9783030483395

T3 - Lecture Notes in Computer Science

SP - 320

EP - 332

BT - Euro-Par 2019: Parallel Processing Workshops

A2 - Schwardmann, Ulrich

A2 - Boehme, Christian

A2 - Heras, Dora B.

A2 - Cardellini, Valeria

A2 - Jeannot, Emmanuel

A2 - Salis, Antonio

A2 - Schifanella, Claudio

A2 - Manumachu, Ravi Reddy

A2 - Schwamborn, Dieter

A2 - Ricci, Laura

A2 - Sangyoon, Oh

A2 - Gruber, Thomas

A2 - Antonelli, Laura

A2 - Scott, Stephen L.

PB - Springer

T2 - Euro-Par 2019 International Workshops

Y2 - 26 August 2019 through 30 August 2019

ER -

Khaleghzadeh H, Fahad M, Manumachu RR, Lastovetsky A. Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution. In Schwardmann U, Boehme C, Heras DB, Cardellini V, Jeannot E, Salis A, Schifanella C, Manumachu RR, Schwamborn D, Ricci L, Sangyoon O, Gruber T, Antonelli L, Scott SL, editors, Euro-Par 2019: Parallel Processing Workshops: Euro-Par 2019 International Workshops, Göttingen, Germany, August 26–30, 2019, Revised Selected Papers. Springer. 2020. p. 320-332. (Lecture Notes in Computer Science). doi: 10.1007/978-3-030-48340-1_25

Optimization of data-parallel applications on heterogeneous HPC platforms for dynamic energy through workload distribution

Abstract

Publication series

Workshop

Keywords

UN SDGs

Access to Document

Fingerprint

Cite this