Comparative performance evaluation of multi-GPU MLFMA implementation for 2-D VIE problems

Carl Pearson, Mert Hidayetoglu, Wei Ren, Weng Cho Chew, Wen-Mei W Hwu

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

Abstract

We compare multi-GPU performance of the multilevel fast multipole algorithm (MLFMA) on two different systems: A shared-memory IBM S822LC workstation with four NVIDIA P100 GPUs, and 16 XK nodes (each is employed with a single NVIDIA K20X GPU) of the Blue Waters supercomputer. MLFMA is implemented for solving scattering problems involving two-dimensional inhomogeneous bodies. Results show that the multi-GPU implementation provides 794 and 969 times speedups on the IBM and Blue Waters systems over their corresponding sequential CPU executions, respectively, where the sequential execution on the IBM system is 1.17 times faster than on the Blue Waters System.

Original languageEnglish (US)
Title of host publicationCEM 2017 - 2017 Computing and Electromagnetics International Workshop
EditorsLevent Gurel
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages63-64
Number of pages2
ISBN (Electronic)9781538617328
DOIs
StatePublished - Jul 25 2017
Event2017 Computing and Electromagnetics International Workshop, CEM 2017 - Barcelona, Spain
Duration: Jun 21 2017Jun 24 2017

Publication series

NameCEM 2017 - 2017 Computing and Electromagnetics International Workshop

Other

Other2017 Computing and Electromagnetics International Workshop, CEM 2017
CountrySpain
CityBarcelona
Period6/21/176/24/17

Fingerprint

multipoles
evaluation
two dimensional bodies
water
supercomputers
workstations
Water
Computer workstations
Supercomputers
Program processors
Computer systems
Scattering
scattering
Data storage equipment
Graphics processing unit

ASJC Scopus subject areas

  • Instrumentation
  • Computer Networks and Communications
  • Electrical and Electronic Engineering

Cite this

Pearson, C., Hidayetoglu, M., Ren, W., Chew, W. C., & Hwu, W-M. W. (2017). Comparative performance evaluation of multi-GPU MLFMA implementation for 2-D VIE problems. In L. Gurel (Ed.), CEM 2017 - 2017 Computing and Electromagnetics International Workshop (pp. 63-64). [7991888] (CEM 2017 - 2017 Computing and Electromagnetics International Workshop). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/CEM.2017.7991888

Comparative performance evaluation of multi-GPU MLFMA implementation for 2-D VIE problems. / Pearson, Carl; Hidayetoglu, Mert; Ren, Wei; Chew, Weng Cho; Hwu, Wen-Mei W.

CEM 2017 - 2017 Computing and Electromagnetics International Workshop. ed. / Levent Gurel. Institute of Electrical and Electronics Engineers Inc., 2017. p. 63-64 7991888 (CEM 2017 - 2017 Computing and Electromagnetics International Workshop).

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

Pearson, C, Hidayetoglu, M, Ren, W, Chew, WC & Hwu, W-MW 2017, Comparative performance evaluation of multi-GPU MLFMA implementation for 2-D VIE problems. in L Gurel (ed.), CEM 2017 - 2017 Computing and Electromagnetics International Workshop., 7991888, CEM 2017 - 2017 Computing and Electromagnetics International Workshop, Institute of Electrical and Electronics Engineers Inc., pp. 63-64, 2017 Computing and Electromagnetics International Workshop, CEM 2017, Barcelona, Spain, 6/21/17. https://doi.org/10.1109/CEM.2017.7991888
Pearson C, Hidayetoglu M, Ren W, Chew WC, Hwu W-MW. Comparative performance evaluation of multi-GPU MLFMA implementation for 2-D VIE problems. In Gurel L, editor, CEM 2017 - 2017 Computing and Electromagnetics International Workshop. Institute of Electrical and Electronics Engineers Inc. 2017. p. 63-64. 7991888. (CEM 2017 - 2017 Computing and Electromagnetics International Workshop). https://doi.org/10.1109/CEM.2017.7991888
Pearson, Carl ; Hidayetoglu, Mert ; Ren, Wei ; Chew, Weng Cho ; Hwu, Wen-Mei W. / Comparative performance evaluation of multi-GPU MLFMA implementation for 2-D VIE problems. CEM 2017 - 2017 Computing and Electromagnetics International Workshop. editor / Levent Gurel. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 63-64 (CEM 2017 - 2017 Computing and Electromagnetics International Workshop).
@inproceedings{968996d6c97544989f531218e3fdb178,
title = "Comparative performance evaluation of multi-GPU MLFMA implementation for 2-D VIE problems",
abstract = "We compare multi-GPU performance of the multilevel fast multipole algorithm (MLFMA) on two different systems: A shared-memory IBM S822LC workstation with four NVIDIA P100 GPUs, and 16 XK nodes (each is employed with a single NVIDIA K20X GPU) of the Blue Waters supercomputer. MLFMA is implemented for solving scattering problems involving two-dimensional inhomogeneous bodies. Results show that the multi-GPU implementation provides 794 and 969 times speedups on the IBM and Blue Waters systems over their corresponding sequential CPU executions, respectively, where the sequential execution on the IBM system is 1.17 times faster than on the Blue Waters System.",
author = "Carl Pearson and Mert Hidayetoglu and Wei Ren and Chew, {Weng Cho} and Hwu, {Wen-Mei W}",
year = "2017",
month = "7",
day = "25",
doi = "10.1109/CEM.2017.7991888",
language = "English (US)",
series = "CEM 2017 - 2017 Computing and Electromagnetics International Workshop",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "63--64",
editor = "Levent Gurel",
booktitle = "CEM 2017 - 2017 Computing and Electromagnetics International Workshop",
address = "United States",

}

TY - GEN

T1 - Comparative performance evaluation of multi-GPU MLFMA implementation for 2-D VIE problems

AU - Pearson, Carl

AU - Hidayetoglu, Mert

AU - Ren, Wei

AU - Chew, Weng Cho

AU - Hwu, Wen-Mei W

PY - 2017/7/25

Y1 - 2017/7/25

N2 - We compare multi-GPU performance of the multilevel fast multipole algorithm (MLFMA) on two different systems: A shared-memory IBM S822LC workstation with four NVIDIA P100 GPUs, and 16 XK nodes (each is employed with a single NVIDIA K20X GPU) of the Blue Waters supercomputer. MLFMA is implemented for solving scattering problems involving two-dimensional inhomogeneous bodies. Results show that the multi-GPU implementation provides 794 and 969 times speedups on the IBM and Blue Waters systems over their corresponding sequential CPU executions, respectively, where the sequential execution on the IBM system is 1.17 times faster than on the Blue Waters System.

AB - We compare multi-GPU performance of the multilevel fast multipole algorithm (MLFMA) on two different systems: A shared-memory IBM S822LC workstation with four NVIDIA P100 GPUs, and 16 XK nodes (each is employed with a single NVIDIA K20X GPU) of the Blue Waters supercomputer. MLFMA is implemented for solving scattering problems involving two-dimensional inhomogeneous bodies. Results show that the multi-GPU implementation provides 794 and 969 times speedups on the IBM and Blue Waters systems over their corresponding sequential CPU executions, respectively, where the sequential execution on the IBM system is 1.17 times faster than on the Blue Waters System.

UR - http://www.scopus.com/inward/record.url?scp=85028539230&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85028539230&partnerID=8YFLogxK

U2 - 10.1109/CEM.2017.7991888

DO - 10.1109/CEM.2017.7991888

M3 - Conference contribution

T3 - CEM 2017 - 2017 Computing and Electromagnetics International Workshop

SP - 63

EP - 64

BT - CEM 2017 - 2017 Computing and Electromagnetics International Workshop

A2 - Gurel, Levent

PB - Institute of Electrical and Electronics Engineers Inc.

ER -