Combining phase identification and statistic modeling for automated parallel benchmark generation

Ye Jin, Mingliang Liu, Xiaosong Ma, Qing Liu, Jeremy Logan, Norbert Podhorszki, Jong Youl Choi, Scott Klasky

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

1 Citation (Scopus)

Abstract

Parallel application benchmarks are indispensable for evaluating/optimizing HPC software and hardware. However, it is very challenging and costly to obtain high-fidelity benchmarks reflecting the scale and complexity of state-of-the-art parallel applications. Hand-extracted synthetic benchmarks are time- and labor-intensive to create. Real applications themselves, while offering most accurate performance evaluation, are expensive to compile, port, reconfigure, and often plainly inaccessible due to security or ownership concerns. This work contributes APPRIME, a novel tool for trace-based automatic parallel benchmark generation. Taking as input standard communication-I/O traces of an application's execution, it couples accurate automatic phase identification with statistical regeneration of event parameters to create compact, portable, and to some degree reconfigurable parallel application benchmarks. Experiments with four NAS Parallel Benchmarks (NPB) and three real scientific simulation codes confirm the fidelity of APPRIME benchmarks. They retain the original applications' performance characteristics, in particular the relative performance across platforms.

Original languageEnglish
Title of host publicationProceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP
PublisherAssociation for Computing Machinery
Pages269-270
Number of pages2
Volume2015-January
ISBN (Print)9781450332057
DOIs
Publication statusPublished - 24 Jan 2015
Event20th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPoPP 2015 - San Francisco, United States
Duration: 7 Feb 201511 Feb 2015

Other

Other20th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPoPP 2015
CountryUnited States
CitySan Francisco
Period7/2/1511/2/15

Fingerprint

Statistics
Personnel
Hardware
Communication
Experiments

Keywords

  • Automatic benchmark generation
  • HPC applications
  • Phase identification
  • Statistical profiling
  • Trace

ASJC Scopus subject areas

  • Software

Cite this

Jin, Y., Liu, M., Ma, X., Liu, Q., Logan, J., Podhorszki, N., ... Klasky, S. (2015). Combining phase identification and statistic modeling for automated parallel benchmark generation. In Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP (Vol. 2015-January, pp. 269-270). Association for Computing Machinery. https://doi.org/10.1145/2688500.2688541

Combining phase identification and statistic modeling for automated parallel benchmark generation. / Jin, Ye; Liu, Mingliang; Ma, Xiaosong; Liu, Qing; Logan, Jeremy; Podhorszki, Norbert; Choi, Jong Youl; Klasky, Scott.

Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP. Vol. 2015-January Association for Computing Machinery, 2015. p. 269-270.

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

Jin, Y, Liu, M, Ma, X, Liu, Q, Logan, J, Podhorszki, N, Choi, JY & Klasky, S 2015, Combining phase identification and statistic modeling for automated parallel benchmark generation. in Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP. vol. 2015-January, Association for Computing Machinery, pp. 269-270, 20th ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPoPP 2015, San Francisco, United States, 7/2/15. https://doi.org/10.1145/2688500.2688541
Jin Y, Liu M, Ma X, Liu Q, Logan J, Podhorszki N et al. Combining phase identification and statistic modeling for automated parallel benchmark generation. In Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP. Vol. 2015-January. Association for Computing Machinery. 2015. p. 269-270 https://doi.org/10.1145/2688500.2688541
Jin, Ye ; Liu, Mingliang ; Ma, Xiaosong ; Liu, Qing ; Logan, Jeremy ; Podhorszki, Norbert ; Choi, Jong Youl ; Klasky, Scott. / Combining phase identification and statistic modeling for automated parallel benchmark generation. Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP. Vol. 2015-January Association for Computing Machinery, 2015. pp. 269-270
@inproceedings{09b066fa9fb54f2c9892b0e4e85fc23e,
title = "Combining phase identification and statistic modeling for automated parallel benchmark generation",
abstract = "Parallel application benchmarks are indispensable for evaluating/optimizing HPC software and hardware. However, it is very challenging and costly to obtain high-fidelity benchmarks reflecting the scale and complexity of state-of-the-art parallel applications. Hand-extracted synthetic benchmarks are time- and labor-intensive to create. Real applications themselves, while offering most accurate performance evaluation, are expensive to compile, port, reconfigure, and often plainly inaccessible due to security or ownership concerns. This work contributes APPRIME, a novel tool for trace-based automatic parallel benchmark generation. Taking as input standard communication-I/O traces of an application's execution, it couples accurate automatic phase identification with statistical regeneration of event parameters to create compact, portable, and to some degree reconfigurable parallel application benchmarks. Experiments with four NAS Parallel Benchmarks (NPB) and three real scientific simulation codes confirm the fidelity of APPRIME benchmarks. They retain the original applications' performance characteristics, in particular the relative performance across platforms.",
keywords = "Automatic benchmark generation, HPC applications, Phase identification, Statistical profiling, Trace",
author = "Ye Jin and Mingliang Liu and Xiaosong Ma and Qing Liu and Jeremy Logan and Norbert Podhorszki and Choi, {Jong Youl} and Scott Klasky",
year = "2015",
month = "1",
day = "24",
doi = "10.1145/2688500.2688541",
language = "English",
isbn = "9781450332057",
volume = "2015-January",
pages = "269--270",
booktitle = "Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP",
publisher = "Association for Computing Machinery",

}

TY - GEN

T1 - Combining phase identification and statistic modeling for automated parallel benchmark generation

AU - Jin, Ye

AU - Liu, Mingliang

AU - Ma, Xiaosong

AU - Liu, Qing

AU - Logan, Jeremy

AU - Podhorszki, Norbert

AU - Choi, Jong Youl

AU - Klasky, Scott

PY - 2015/1/24

Y1 - 2015/1/24

N2 - Parallel application benchmarks are indispensable for evaluating/optimizing HPC software and hardware. However, it is very challenging and costly to obtain high-fidelity benchmarks reflecting the scale and complexity of state-of-the-art parallel applications. Hand-extracted synthetic benchmarks are time- and labor-intensive to create. Real applications themselves, while offering most accurate performance evaluation, are expensive to compile, port, reconfigure, and often plainly inaccessible due to security or ownership concerns. This work contributes APPRIME, a novel tool for trace-based automatic parallel benchmark generation. Taking as input standard communication-I/O traces of an application's execution, it couples accurate automatic phase identification with statistical regeneration of event parameters to create compact, portable, and to some degree reconfigurable parallel application benchmarks. Experiments with four NAS Parallel Benchmarks (NPB) and three real scientific simulation codes confirm the fidelity of APPRIME benchmarks. They retain the original applications' performance characteristics, in particular the relative performance across platforms.

AB - Parallel application benchmarks are indispensable for evaluating/optimizing HPC software and hardware. However, it is very challenging and costly to obtain high-fidelity benchmarks reflecting the scale and complexity of state-of-the-art parallel applications. Hand-extracted synthetic benchmarks are time- and labor-intensive to create. Real applications themselves, while offering most accurate performance evaluation, are expensive to compile, port, reconfigure, and often plainly inaccessible due to security or ownership concerns. This work contributes APPRIME, a novel tool for trace-based automatic parallel benchmark generation. Taking as input standard communication-I/O traces of an application's execution, it couples accurate automatic phase identification with statistical regeneration of event parameters to create compact, portable, and to some degree reconfigurable parallel application benchmarks. Experiments with four NAS Parallel Benchmarks (NPB) and three real scientific simulation codes confirm the fidelity of APPRIME benchmarks. They retain the original applications' performance characteristics, in particular the relative performance across platforms.

KW - Automatic benchmark generation

KW - HPC applications

KW - Phase identification

KW - Statistical profiling

KW - Trace

UR - http://www.scopus.com/inward/record.url?scp=84939146769&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84939146769&partnerID=8YFLogxK

U2 - 10.1145/2688500.2688541

DO - 10.1145/2688500.2688541

M3 - Conference contribution

AN - SCOPUS:84939146769

SN - 9781450332057

VL - 2015-January

SP - 269

EP - 270

BT - Proceedings of the ACM SIGPLAN Symposium on Principles and Practice of Parallel Programming, PPOPP

PB - Association for Computing Machinery

ER -