An algebra for cross-experiment performance analysis

Fengguang Song; Felix Wolf; Nikhil Bhatia; Jack Dongarra; Shirley Moore

doi:10.1109/icpp.2004.1327905

An algebra for cross-experiment performance analysis

Fengguang Song, Felix Wolf, Nikhil Bhatia, Jack Dongarra, Shirley Moore

Research output: Contribution to journal › Conference article › peer-review

42 Scopus citations

Abstract

Performance tuning of parallel applications usually involves multiple experiments to compare the effects of different optimization strategies. This article describes an algebra that can be used to compare, integrate, and summarize performance data from multiple sources. The algebra consists of a data model to represent the data in a platform-independent fashion plus arithmetic operations to merge, subtract, and average the data from different experiments. A distinctive feature of this approach is its closure property, which allows processing and viewing all instances of the data model in the same way - regardless of whether they represent original or derived data - in addition to an arbitrary and easy composition of operations.

Original language	English
Pages (from-to)	63-72
Number of pages	10
Journal	Proceedings of the International Conference on Parallel Processing
DOIs	https://doi.org/10.1109/icpp.2004.1327905
State	Published - 2004
Externally published	Yes
Event	Proceedings - 2004 International Conference on Parallel Processing, ICPP 2004 - Montreal, Que, Canada Duration: Aug 15 2004 → Aug 18 2004

Keywords

Multiexperiment analysis
Performance algebra
Performance tool
Tool interoperability
Visualization

Access to Document

10.1109/icpp.2004.1327905

Cite this

@article{6c15d5bd1fa741109b7b0fd5b5f4a48a,

title = "An algebra for cross-experiment performance analysis",

abstract = "Performance tuning of parallel applications usually involves multiple experiments to compare the effects of different optimization strategies. This article describes an algebra that can be used to compare, integrate, and summarize performance data from multiple sources. The algebra consists of a data model to represent the data in a platform-independent fashion plus arithmetic operations to merge, subtract, and average the data from different experiments. A distinctive feature of this approach is its closure property, which allows processing and viewing all instances of the data model in the same way - regardless of whether they represent original or derived data - in addition to an arbitrary and easy composition of operations.",

keywords = "Multiexperiment analysis, Performance algebra, Performance tool, Tool interoperability, Visualization",

author = "Fengguang Song and Felix Wolf and Nikhil Bhatia and Jack Dongarra and Shirley Moore",

year = "2004",

doi = "10.1109/icpp.2004.1327905",

language = "English",

pages = "63--72",

journal = "Proceedings of the International Conference on Parallel Processing",

issn = "0190-3918",

note = "Proceedings - 2004 International Conference on Parallel Processing, ICPP 2004 ; Conference date: 15-08-2004 Through 18-08-2004",

}

TY - JOUR

T1 - An algebra for cross-experiment performance analysis

AU - Song, Fengguang

AU - Wolf, Felix

AU - Bhatia, Nikhil

AU - Dongarra, Jack

AU - Moore, Shirley

PY - 2004

Y1 - 2004

N2 - Performance tuning of parallel applications usually involves multiple experiments to compare the effects of different optimization strategies. This article describes an algebra that can be used to compare, integrate, and summarize performance data from multiple sources. The algebra consists of a data model to represent the data in a platform-independent fashion plus arithmetic operations to merge, subtract, and average the data from different experiments. A distinctive feature of this approach is its closure property, which allows processing and viewing all instances of the data model in the same way - regardless of whether they represent original or derived data - in addition to an arbitrary and easy composition of operations.

AB - Performance tuning of parallel applications usually involves multiple experiments to compare the effects of different optimization strategies. This article describes an algebra that can be used to compare, integrate, and summarize performance data from multiple sources. The algebra consists of a data model to represent the data in a platform-independent fashion plus arithmetic operations to merge, subtract, and average the data from different experiments. A distinctive feature of this approach is its closure property, which allows processing and viewing all instances of the data model in the same way - regardless of whether they represent original or derived data - in addition to an arbitrary and easy composition of operations.

KW - Multiexperiment analysis

KW - Performance algebra

KW - Performance tool

KW - Tool interoperability

KW - Visualization

UR - http://www.scopus.com/inward/record.url?scp=10044276950&partnerID=8YFLogxK

U2 - 10.1109/icpp.2004.1327905

DO - 10.1109/icpp.2004.1327905

M3 - Conference article

AN - SCOPUS:10044276950

SN - 0190-3918

SP - 63

EP - 72

JO - Proceedings of the International Conference on Parallel Processing

JF - Proceedings of the International Conference on Parallel Processing

T2 - Proceedings - 2004 International Conference on Parallel Processing, ICPP 2004

Y2 - 15 August 2004 through 18 August 2004

ER -

An algebra for cross-experiment performance analysis

Abstract

Keywords

Access to Document

Other files and links

Fingerprint

Cite this