DocumentCode :
1147971
Title :
Beyond execution time: expanding the use of performance models
Author :
Peterson, Gregory D. ; Chamberlain, Roger D.
Author_Institution :
Washington Univ., Seattle, WA, USA
Volume :
2
Issue :
2
fYear :
1994
Firstpage :
37
Lastpage :
49
Abstract :
Improved performance is a major motivation for using parallel computation. However, performance models are frequently used only to predict an algorithm´s execution time, not to accurately evaluate how the choices of architecture, operating system, interprocessor communication protocol, and programming language also dramatically affect parallel performance. We have developed an analytic model for synchronous iterative algorithms running on distributed-memory MIMD machines, and refined it for disrete-event simulation. The model describes the execution time of a single run in terms of application parameters such as the number of iterations and the required computation in each, and architectural parameters such as the number of processors, processor speed, and communication time. Our experience has shown us that an analytic model can not only accurately predict an algorithm´s performance but can also match the algorithm to an appropriate architecture, identify ways to improve the algorithm´s performance, quantify the performance effects of algorithmic or architectural changes, and provide a better understanding of how the algorithm works.<>
Keywords :
discrete event simulation; distributed memory systems; performance evaluation; software metrics; synchronisation; algorithm performance; analytic model; application parameters; architectural parameters; architecture; communication time; disrete-event simulation; distributed-memory MIMD machines; execution time; interprocessor communication protocol; operating system; parallel computation; parallel performance; performance models; processor speed; programming language; synchronous iterative algorithms; Algorithm design and analysis; Computational modeling; Computer architecture; Computer languages; Concurrent computing; Iterative algorithms; Operating systems; Prediction algorithms; Predictive models; Protocols;
fLanguage :
English
Journal_Title :
Parallel & Distributed Technology: Systems & Applications, IEEE
Publisher :
ieee
ISSN :
1063-6552
Type :
jour
DOI :
10.1109/88.311571
Filename :
311571
Link To Document :
بازگشت