Performance and energy analysis of scientific workloads executing on LPSoCs

Anish Varghese; Joshua Milthorpe; Alistair P. Rendell

doi:10.1007/978-3-319-78054-2_11

Performance and energy analysis of scientific workloads executing on LPSoCs

Anish Varghese
, Joshua Milthorpe
, Alistair P. Rendell

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

1 Citation (Scopus)

Abstract

Low-power system-on-chip (LPSoC) processors provide an interesting alternative as building blocks for future HPC systems due to their high energy efficiency. However, understanding their performance-energy trade-offs and minimizing the energy-to-solution for an application running across the heterogeneous devices of an LPSoC remains a challenge. In this paper, we describe our methodology for developing an energy model which may be used to predict the energy usage of application code executing on an LPSoC system under different frequency settings. For this paper, we focus only on the CPU. Performance and energy measurements are presented for different types of workloads on the NVIDIA Tegra TK1 and Tegra TX1 systems at varying frequencies. From these results, we provide insights on how to develop a model to predict energy usage at different frequencies for general workloads.

Original language	English
Title of host publication	Parallel Processing and Applied Mathematics - 12th International Conference, PPAM 2017, Revised Selected Papers
Editors	Ewa Deelman, Roman Wyrzykowski, Konrad Karczewski, Jack Dongarra
Publisher	Springer-Verlag
Pages	113-122
Number of pages	10
Volume	10778 LNCS
ISBN (Print)	9783319780535
DOIs	https://doi.org/10.1007/978-3-319-78054-2_11
Publication status	Published - 10 Sept 2017
Externally published	Yes
Event	12th International Conference on Parallel Processing and Applied Mathematics, PPAM 2017 - Czestochowa, Poland Duration: 10 Sept 2017 → 13 Sept 2017

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	10778 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	12th International Conference on Parallel Processing and Applied Mathematics, PPAM 2017
Country/Territory	Poland
City	Czestochowa
Period	10/09/17 → 13/09/17

UN SDGs

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

SDG 7 Affordable and Clean Energy

Keywords

DVFS
Energy efficiency
Energy usage model
LPSoC
Tegra SoC

Access to Document

10.1007/978-3-319-78054-2_11Licence: In Copyright

https://rdcu.be/dA9ihLicence: In Copyright

Cite this

Varghese, A., Milthorpe, J., & Rendell, A. P. (2017). Performance and energy analysis of scientific workloads executing on LPSoCs. In E. Deelman, R. Wyrzykowski, K. Karczewski, & J. Dongarra (Eds.), Parallel Processing and Applied Mathematics - 12th International Conference, PPAM 2017, Revised Selected Papers (Vol. 10778 LNCS, pp. 113-122). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10778 LNCS). Springer-Verlag. https://doi.org/10.1007/978-3-319-78054-2_11

Varghese, Anish ; Milthorpe, Joshua ; Rendell, Alistair P. / Performance and energy analysis of scientific workloads executing on LPSoCs. Parallel Processing and Applied Mathematics - 12th International Conference, PPAM 2017, Revised Selected Papers. editor / Ewa Deelman ; Roman Wyrzykowski ; Konrad Karczewski ; Jack Dongarra. Vol. 10778 LNCS Springer-Verlag, 2017. pp. 113-122 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{83ff361f4c2b4cb980d6aa55b5be1026,

title = "Performance and energy analysis of scientific workloads executing on LPSoCs",

abstract = "Low-power system-on-chip (LPSoC) processors provide an interesting alternative as building blocks for future HPC systems due to their high energy efficiency. However, understanding their performance-energy trade-offs and minimizing the energy-to-solution for an application running across the heterogeneous devices of an LPSoC remains a challenge. In this paper, we describe our methodology for developing an energy model which may be used to predict the energy usage of application code executing on an LPSoC system under different frequency settings. For this paper, we focus only on the CPU. Performance and energy measurements are presented for different types of workloads on the NVIDIA Tegra TK1 and Tegra TX1 systems at varying frequencies. From these results, we provide insights on how to develop a model to predict energy usage at different frequencies for general workloads.",

keywords = "DVFS, Energy efficiency, Energy usage model, LPSoC, Tegra SoC",

author = "Anish Varghese and Joshua Milthorpe and Rendell, \{Alistair P.\}",

year = "2017",

month = sep,

day = "10",

doi = "10.1007/978-3-319-78054-2\_11",

language = "English",

isbn = "9783319780535",

volume = "10778 LNCS",

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

publisher = "Springer-Verlag",

pages = "113--122",

editor = "Ewa Deelman and Roman Wyrzykowski and Konrad Karczewski and Jack Dongarra",

booktitle = "Parallel Processing and Applied Mathematics - 12th International Conference, PPAM 2017, Revised Selected Papers",

note = "12th International Conference on Parallel Processing and Applied Mathematics, PPAM 2017 ; Conference date: 10-09-2017 Through 13-09-2017",

}

Varghese, A, Milthorpe, J & Rendell, AP 2017, Performance and energy analysis of scientific workloads executing on LPSoCs. in E Deelman, R Wyrzykowski, K Karczewski & J Dongarra (eds), Parallel Processing and Applied Mathematics - 12th International Conference, PPAM 2017, Revised Selected Papers. vol. 10778 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10778 LNCS, Springer-Verlag, pp. 113-122, 12th International Conference on Parallel Processing and Applied Mathematics, PPAM 2017, Czestochowa, Poland, 10/09/17. https://doi.org/10.1007/978-3-319-78054-2_11

Performance and energy analysis of scientific workloads executing on LPSoCs. / Varghese, Anish; Milthorpe, Joshua; Rendell, Alistair P.
Parallel Processing and Applied Mathematics - 12th International Conference, PPAM 2017, Revised Selected Papers. ed. / Ewa Deelman; Roman Wyrzykowski; Konrad Karczewski; Jack Dongarra. Vol. 10778 LNCS Springer-Verlag, 2017. p. 113-122 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 10778 LNCS).

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

TY - GEN

T1 - Performance and energy analysis of scientific workloads executing on LPSoCs

AU - Varghese, Anish

AU - Milthorpe, Joshua

AU - Rendell, Alistair P.

PY - 2017/9/10

Y1 - 2017/9/10

N2 - Low-power system-on-chip (LPSoC) processors provide an interesting alternative as building blocks for future HPC systems due to their high energy efficiency. However, understanding their performance-energy trade-offs and minimizing the energy-to-solution for an application running across the heterogeneous devices of an LPSoC remains a challenge. In this paper, we describe our methodology for developing an energy model which may be used to predict the energy usage of application code executing on an LPSoC system under different frequency settings. For this paper, we focus only on the CPU. Performance and energy measurements are presented for different types of workloads on the NVIDIA Tegra TK1 and Tegra TX1 systems at varying frequencies. From these results, we provide insights on how to develop a model to predict energy usage at different frequencies for general workloads.

AB - Low-power system-on-chip (LPSoC) processors provide an interesting alternative as building blocks for future HPC systems due to their high energy efficiency. However, understanding their performance-energy trade-offs and minimizing the energy-to-solution for an application running across the heterogeneous devices of an LPSoC remains a challenge. In this paper, we describe our methodology for developing an energy model which may be used to predict the energy usage of application code executing on an LPSoC system under different frequency settings. For this paper, we focus only on the CPU. Performance and energy measurements are presented for different types of workloads on the NVIDIA Tegra TK1 and Tegra TX1 systems at varying frequencies. From these results, we provide insights on how to develop a model to predict energy usage at different frequencies for general workloads.

KW - DVFS

KW - Energy efficiency

KW - Energy usage model

KW - LPSoC

KW - Tegra SoC

UR - https://www.scopus.com/pages/publications/85044780911

U2 - 10.1007/978-3-319-78054-2_11

DO - 10.1007/978-3-319-78054-2_11

M3 - Conference contribution

AN - SCOPUS:85044780911

SN - 9783319780535

VL - 10778 LNCS

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

SP - 113

EP - 122

BT - Parallel Processing and Applied Mathematics - 12th International Conference, PPAM 2017, Revised Selected Papers

A2 - Deelman, Ewa

A2 - Wyrzykowski, Roman

A2 - Karczewski, Konrad

A2 - Dongarra, Jack

PB - Springer-Verlag

T2 - 12th International Conference on Parallel Processing and Applied Mathematics, PPAM 2017

Y2 - 10 September 2017 through 13 September 2017

ER -

Varghese A, Milthorpe J, Rendell AP. Performance and energy analysis of scientific workloads executing on LPSoCs. In Deelman E, Wyrzykowski R, Karczewski K, Dongarra J, editors, Parallel Processing and Applied Mathematics - 12th International Conference, PPAM 2017, Revised Selected Papers. Vol. 10778 LNCS. Springer-Verlag. 2017. p. 113-122. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-319-78054-2_11

Performance and energy analysis of scientific workloads executing on LPSoCs

Abstract

Publication series

Conference

UN SDGs

Keywords

Access to Document

Other files and links

Fingerprint

Cite this