Trace-based performance analysis framework for heterogeneous multicore systems

Shih Hao Hung, ChiaHeng Tu, Thean Siew Soon

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

5 Citations (Scopus)

Abstract

Performance evaluation is key to the optimization of computer applications on multicore systems. While many techniques and profiling tools are available for measuring performance on homogeneous multicore platforms, most of them depend on the hardware support from the vendors. For developing applications on heterogeneous multicore systems, very few analysis tools exist to help the developers. This paper describes a software-based trace collection and performance analysis framework that can be ported to a variety of platforms via code instrumentation at the source level. A pure software profiling toolkit, called ParallelTracer, were implemented based on ANTLR, an open source parser generator, to support this framework. In this paper, we present our framework and toolkit. We use the IBM Cell processor as a case study to demonstrate the capability of ParallelTrace. Our results show that ParallelTracer provided useful information for programmers to understand program behaviors and identify potential performance bottlenecks via graphical visualization. We also discuss the runtime overhead of ParallelTracer. With proper usage, the performance and code size overhead introduced by our toolkit are limited around 19% to 5% and 9%, respectively, for the benchmark program in the case study.

Original languageEnglish
Title of host publication2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010
Pages19-24
Number of pages6
DOIs
Publication statusPublished - 2010 Apr 28
Event2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010 - Taipei, Taiwan
Duration: 2010 Jan 182010 Jan 21

Publication series

NameProceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

Other

Other2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010
CountryTaiwan
CityTaipei
Period10-01-1810-01-21

Fingerprint

Computer applications
Visualization
Hardware

All Science Journal Classification (ASJC) codes

  • Computer Science Applications
  • Computer Graphics and Computer-Aided Design
  • Electrical and Electronic Engineering

Cite this

Hung, S. H., Tu, C., & Soon, T. S. (2010). Trace-based performance analysis framework for heterogeneous multicore systems. In 2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010 (pp. 19-24). [5419926] (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). https://doi.org/10.1109/ASPDAC.2010.5419926
Hung, Shih Hao ; Tu, ChiaHeng ; Soon, Thean Siew. / Trace-based performance analysis framework for heterogeneous multicore systems. 2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010. 2010. pp. 19-24 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).
@inproceedings{6f8b15df84584024889f29895931ba2c,
title = "Trace-based performance analysis framework for heterogeneous multicore systems",
abstract = "Performance evaluation is key to the optimization of computer applications on multicore systems. While many techniques and profiling tools are available for measuring performance on homogeneous multicore platforms, most of them depend on the hardware support from the vendors. For developing applications on heterogeneous multicore systems, very few analysis tools exist to help the developers. This paper describes a software-based trace collection and performance analysis framework that can be ported to a variety of platforms via code instrumentation at the source level. A pure software profiling toolkit, called ParallelTracer, were implemented based on ANTLR, an open source parser generator, to support this framework. In this paper, we present our framework and toolkit. We use the IBM Cell processor as a case study to demonstrate the capability of ParallelTrace. Our results show that ParallelTracer provided useful information for programmers to understand program behaviors and identify potential performance bottlenecks via graphical visualization. We also discuss the runtime overhead of ParallelTracer. With proper usage, the performance and code size overhead introduced by our toolkit are limited around 19{\%} to 5{\%} and 9{\%}, respectively, for the benchmark program in the case study.",
author = "Hung, {Shih Hao} and ChiaHeng Tu and Soon, {Thean Siew}",
year = "2010",
month = "4",
day = "28",
doi = "10.1109/ASPDAC.2010.5419926",
language = "English",
isbn = "9781424457656",
series = "Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC",
pages = "19--24",
booktitle = "2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010",

}

Hung, SH, Tu, C & Soon, TS 2010, Trace-based performance analysis framework for heterogeneous multicore systems. in 2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010., 5419926, Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC, pp. 19-24, 2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010, Taipei, Taiwan, 10-01-18. https://doi.org/10.1109/ASPDAC.2010.5419926

Trace-based performance analysis framework for heterogeneous multicore systems. / Hung, Shih Hao; Tu, ChiaHeng; Soon, Thean Siew.

2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010. 2010. p. 19-24 5419926 (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC).

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

TY - GEN

T1 - Trace-based performance analysis framework for heterogeneous multicore systems

AU - Hung, Shih Hao

AU - Tu, ChiaHeng

AU - Soon, Thean Siew

PY - 2010/4/28

Y1 - 2010/4/28

N2 - Performance evaluation is key to the optimization of computer applications on multicore systems. While many techniques and profiling tools are available for measuring performance on homogeneous multicore platforms, most of them depend on the hardware support from the vendors. For developing applications on heterogeneous multicore systems, very few analysis tools exist to help the developers. This paper describes a software-based trace collection and performance analysis framework that can be ported to a variety of platforms via code instrumentation at the source level. A pure software profiling toolkit, called ParallelTracer, were implemented based on ANTLR, an open source parser generator, to support this framework. In this paper, we present our framework and toolkit. We use the IBM Cell processor as a case study to demonstrate the capability of ParallelTrace. Our results show that ParallelTracer provided useful information for programmers to understand program behaviors and identify potential performance bottlenecks via graphical visualization. We also discuss the runtime overhead of ParallelTracer. With proper usage, the performance and code size overhead introduced by our toolkit are limited around 19% to 5% and 9%, respectively, for the benchmark program in the case study.

AB - Performance evaluation is key to the optimization of computer applications on multicore systems. While many techniques and profiling tools are available for measuring performance on homogeneous multicore platforms, most of them depend on the hardware support from the vendors. For developing applications on heterogeneous multicore systems, very few analysis tools exist to help the developers. This paper describes a software-based trace collection and performance analysis framework that can be ported to a variety of platforms via code instrumentation at the source level. A pure software profiling toolkit, called ParallelTracer, were implemented based on ANTLR, an open source parser generator, to support this framework. In this paper, we present our framework and toolkit. We use the IBM Cell processor as a case study to demonstrate the capability of ParallelTrace. Our results show that ParallelTracer provided useful information for programmers to understand program behaviors and identify potential performance bottlenecks via graphical visualization. We also discuss the runtime overhead of ParallelTracer. With proper usage, the performance and code size overhead introduced by our toolkit are limited around 19% to 5% and 9%, respectively, for the benchmark program in the case study.

UR - http://www.scopus.com/inward/record.url?scp=77951243904&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=77951243904&partnerID=8YFLogxK

U2 - 10.1109/ASPDAC.2010.5419926

DO - 10.1109/ASPDAC.2010.5419926

M3 - Conference contribution

SN - 9781424457656

T3 - Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC

SP - 19

EP - 24

BT - 2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010

ER -

Hung SH, Tu C, Soon TS. Trace-based performance analysis framework for heterogeneous multicore systems. In 2010 15th Asia and South Pacific Design Automation Conference, ASP-DAC 2010. 2010. p. 19-24. 5419926. (Proceedings of the Asia and South Pacific Design Automation Conference, ASP-DAC). https://doi.org/10.1109/ASPDAC.2010.5419926