IMPACC: A tightly integrated MPI+OpenACC framework exploiting shared memory parallelism

Jungwon Kim; Seyong Lee; Jeffrey S. Vetter

doi:10.1145/2907294.2907302

IMPACC: A tightly integrated MPI+OpenACC framework exploiting shared memory parallelism

Jungwon Kim, Seyong Lee, Jeffrey S. Vetter

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

6 Scopus citations

Abstract

We propose IMPACC, an MPI+OpenACC framework for heterogeneous accelerator clusters. IMPACC tightly integrates MPI and OpenACC, while exploiting the shared memory parallelism in the target system. IMPACC dynamically adapts the input MPI+OpenACC applications on the target heterogeneous accelerator clusters to fully exploit target system-specific features. IMPACC provides the programmers with the unified virtual address space, automatic NUMA-friendly task-device mapping, efficient integrated communication routines, seamless streamlining of asynchronous executions, and transparent memory sharing. We have implemented IMPACC and evaluated its performance using three heterogeneous accelerator systems, including Titan supercomputer. Results show that IMPACC can achieve easier programming, higher performance, and better scalability than the current MPI+OpenACC model.

Original language	English
Title of host publication	HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing
Publisher	Association for Computing Machinery, Inc
Pages	189-201
Number of pages	13
ISBN (Electronic)	9781450343145
DOIs	https://doi.org/10.1145/2907294.2907302
State	Published - May 31 2016
Event	25th ACM International Symposium on High-Performance Parallel and Distributed Computing, HPDC 2016 - Kyoto, Japan Duration: May 31 2016 → Jun 4 2016

Publication series

Name	HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing

Conference

Conference	25th ACM International Symposium on High-Performance Parallel and Distributed Computing, HPDC 2016
Country/Territory	Japan
City	Kyoto
Period	05/31/16 → 06/4/16

Funding

This research used resources of the Oak Ridge Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract DE-AC05-00OR22725. This material is based upon work supported by the National Science Foundation under Grant Number 1137097 and by the University of Tennessee through the Beacon Project. Any opinions, findings, conclusions, or recommendations expressed in this material are those of the author(s) and do not necessarily reect the views of the National Science Foundation or the University of Tennessee. The authors would like to thank NVIDIA for providing access to their PSG Cluster. This material is based upon work supported by the U.S. Department of Energy, Office of Science, Office of Advanced Scientific Computing Research. This manuscript has been authored by UT-Battelle, LLC under Contract No. DEAC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan(http://energy.gov/downloads/doe-public-access-plan).

Funders	Funder number
DOE Public Access Plan
United States Government
National Science Foundation	1137097
U.S. Department of Energy
Office of Science
Advanced Scientific Computing Research	DEAC05-00OR22725
University of Tennessee

Keywords

Clusters
Heterogeneous computing
MPI
OpenACC
Programming models

Access to Document

10.1145/2907294.2907302

Cite this

Kim, J., Lee, S., & Vetter, J. S. (2016). IMPACC: A tightly integrated MPI+OpenACC framework exploiting shared memory parallelism. In HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing (pp. 189-201). (HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing). Association for Computing Machinery, Inc. https://doi.org/10.1145/2907294.2907302

Kim, Jungwon ; Lee, Seyong ; Vetter, Jeffrey S. / IMPACC : A tightly integrated MPI+OpenACC framework exploiting shared memory parallelism. HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing. Association for Computing Machinery, Inc, 2016. pp. 189-201 (HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing).

@inproceedings{190c96f0bfe64678beabe96336dfb7f0,

title = "IMPACC: A tightly integrated MPI+OpenACC framework exploiting shared memory parallelism",

abstract = "We propose IMPACC, an MPI+OpenACC framework for heterogeneous accelerator clusters. IMPACC tightly integrates MPI and OpenACC, while exploiting the shared memory parallelism in the target system. IMPACC dynamically adapts the input MPI+OpenACC applications on the target heterogeneous accelerator clusters to fully exploit target system-specific features. IMPACC provides the programmers with the unified virtual address space, automatic NUMA-friendly task-device mapping, efficient integrated communication routines, seamless streamlining of asynchronous executions, and transparent memory sharing. We have implemented IMPACC and evaluated its performance using three heterogeneous accelerator systems, including Titan supercomputer. Results show that IMPACC can achieve easier programming, higher performance, and better scalability than the current MPI+OpenACC model.",

keywords = "Clusters, Heterogeneous computing, MPI, OpenACC, Programming models",

author = "Jungwon Kim and Seyong Lee and Vetter, {Jeffrey S.}",

note = "Publisher Copyright: Copyright {\textcopyright} 2016 by the Association for Computing Machinery, Inc. (ACM).; 25th ACM International Symposium on High-Performance Parallel and Distributed Computing, HPDC 2016 ; Conference date: 31-05-2016 Through 04-06-2016",

year = "2016",

month = may,

day = "31",

doi = "10.1145/2907294.2907302",

language = "English",

series = "HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing",

publisher = "Association for Computing Machinery, Inc",

pages = "189--201",

booktitle = "HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing",

}

Kim, J, Lee, S & Vetter, JS 2016, IMPACC: A tightly integrated MPI+OpenACC framework exploiting shared memory parallelism. in HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing. HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing, Association for Computing Machinery, Inc, pp. 189-201, 25th ACM International Symposium on High-Performance Parallel and Distributed Computing, HPDC 2016, Kyoto, Japan, 05/31/16. https://doi.org/10.1145/2907294.2907302

IMPACC: A tightly integrated MPI+OpenACC framework exploiting shared memory parallelism. / Kim, Jungwon; Lee, Seyong ; Vetter, Jeffrey S.
HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing. Association for Computing Machinery, Inc, 2016. p. 189-201 (HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing).

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

TY - GEN

T1 - IMPACC

T2 - 25th ACM International Symposium on High-Performance Parallel and Distributed Computing, HPDC 2016

AU - Kim, Jungwon

AU - Lee, Seyong

AU - Vetter, Jeffrey S.

PY - 2016/5/31

Y1 - 2016/5/31

N2 - We propose IMPACC, an MPI+OpenACC framework for heterogeneous accelerator clusters. IMPACC tightly integrates MPI and OpenACC, while exploiting the shared memory parallelism in the target system. IMPACC dynamically adapts the input MPI+OpenACC applications on the target heterogeneous accelerator clusters to fully exploit target system-specific features. IMPACC provides the programmers with the unified virtual address space, automatic NUMA-friendly task-device mapping, efficient integrated communication routines, seamless streamlining of asynchronous executions, and transparent memory sharing. We have implemented IMPACC and evaluated its performance using three heterogeneous accelerator systems, including Titan supercomputer. Results show that IMPACC can achieve easier programming, higher performance, and better scalability than the current MPI+OpenACC model.

AB - We propose IMPACC, an MPI+OpenACC framework for heterogeneous accelerator clusters. IMPACC tightly integrates MPI and OpenACC, while exploiting the shared memory parallelism in the target system. IMPACC dynamically adapts the input MPI+OpenACC applications on the target heterogeneous accelerator clusters to fully exploit target system-specific features. IMPACC provides the programmers with the unified virtual address space, automatic NUMA-friendly task-device mapping, efficient integrated communication routines, seamless streamlining of asynchronous executions, and transparent memory sharing. We have implemented IMPACC and evaluated its performance using three heterogeneous accelerator systems, including Titan supercomputer. Results show that IMPACC can achieve easier programming, higher performance, and better scalability than the current MPI+OpenACC model.

KW - Clusters

KW - Heterogeneous computing

KW - MPI

KW - OpenACC

KW - Programming models

UR - http://www.scopus.com/inward/record.url?scp=84978488301&partnerID=8YFLogxK

U2 - 10.1145/2907294.2907302

DO - 10.1145/2907294.2907302

M3 - Conference contribution

AN - SCOPUS:84978488301

T3 - HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing

SP - 189

EP - 201

BT - HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing

PB - Association for Computing Machinery, Inc

Y2 - 31 May 2016 through 4 June 2016

ER -

Kim J, Lee S , Vetter JS. IMPACC: A tightly integrated MPI+OpenACC framework exploiting shared memory parallelism. In HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing. Association for Computing Machinery, Inc. 2016. p. 189-201. (HPDC 2016 - Proceedings of the 25th ACM International Symposium on High-Performance Parallel and Distributed Computing). doi: 10.1145/2907294.2907302

IMPACC: A tightly integrated MPI+OpenACC framework exploiting shared memory parallelism

Abstract

Publication series

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this