Parallel software performance metrics as continuous functions of event times

S. D. Prestwich

doi:10.1016/S1571-0661(05)80057-X

Parallel software performance metrics as continuous functions of event times

S. D. Prestwich

Research output: Contribution to journal › Article › peer-review

Abstract

An important role played by metrics in tuning properties such as task granularity and load balancing while developing software for parallel architecture was discussed. It is found that the designer of parallel performance metrics must take great care to ensure that it is meaningful. It was argued that small inaccuracies in event times should not produce large effects on metrics and hence on numbers, graphs, pictures or animations produced by analysis or visualization. The results show that continuous metrics are insensitive to time slot length and hence are found to be well-behaved under changes in time scale.

Original language	English
Pages (from-to)	302
Number of pages	1
Journal	Electronic Notes in Theoretical Computer Science
Volume	40
DOIs	https://doi.org/10.1016/S1571-0661(05)80057-X
Publication status	Published - Mar 2001
Event	MFCSIT2000, The First Irish Conference on the Mathematical Foundations of Computer Science and Information Technology - Cork, Ireland Duration: 20 Jul 2000 → 21 Jul 2000

Access to Document

10.1016/S1571-0661(05)80057-X

Cite this

@article{bea32b81db074a60a2b2cd3a50eec13c,

title = "Parallel software performance metrics as continuous functions of event times",

abstract = "An important role played by metrics in tuning properties such as task granularity and load balancing while developing software for parallel architecture was discussed. It is found that the designer of parallel performance metrics must take great care to ensure that it is meaningful. It was argued that small inaccuracies in event times should not produce large effects on metrics and hence on numbers, graphs, pictures or animations produced by analysis or visualization. The results show that continuous metrics are insensitive to time slot length and hence are found to be well-behaved under changes in time scale.",

author = "Prestwich, \{S. D.\}",

year = "2001",

month = mar,

doi = "10.1016/S1571-0661(05)80057-X",

language = "English",

volume = "40",

pages = "302",

journal = "Electronic Notes in Theoretical Computer Science",

issn = "1571-0661",

publisher = "Elsevier B.V.",

note = "MFCSIT2000, The First Irish Conference on the Mathematical Foundations of Computer Science and Information Technology ; Conference date: 20-07-2000 Through 21-07-2000",

}

TY - JOUR

T1 - Parallel software performance metrics as continuous functions of event times

AU - Prestwich, S. D.

PY - 2001/3

Y1 - 2001/3

N2 - An important role played by metrics in tuning properties such as task granularity and load balancing while developing software for parallel architecture was discussed. It is found that the designer of parallel performance metrics must take great care to ensure that it is meaningful. It was argued that small inaccuracies in event times should not produce large effects on metrics and hence on numbers, graphs, pictures or animations produced by analysis or visualization. The results show that continuous metrics are insensitive to time slot length and hence are found to be well-behaved under changes in time scale.

AB - An important role played by metrics in tuning properties such as task granularity and load balancing while developing software for parallel architecture was discussed. It is found that the designer of parallel performance metrics must take great care to ensure that it is meaningful. It was argued that small inaccuracies in event times should not produce large effects on metrics and hence on numbers, graphs, pictures or animations produced by analysis or visualization. The results show that continuous metrics are insensitive to time slot length and hence are found to be well-behaved under changes in time scale.

UR - https://www.scopus.com/pages/publications/18944374117

U2 - 10.1016/S1571-0661(05)80057-X

DO - 10.1016/S1571-0661(05)80057-X

M3 - Article

AN - SCOPUS:18944374117

SN - 1571-0661

VL - 40

SP - 302

JO - Electronic Notes in Theoretical Computer Science

JF - Electronic Notes in Theoretical Computer Science

T2 - MFCSIT2000, The First Irish Conference on the Mathematical Foundations of Computer Science and Information Technology

Y2 - 20 July 2000 through 21 July 2000

ER -

Parallel software performance metrics as continuous functions of event times

Abstract

Access to Document

Other files and links

Fingerprint

Cite this