loading...
 This Article 
   
 Share 
   
 Bibliographic References 
   
 Add to: 
 
Digg
Furl
Spurl
Blink
Simpy
Google
Del.icio.us
Y!MyWeb
 
 Search 
   
Ninth IEEE International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunications Systems (MASCOTS'01)
Low-Cost Performance Prediction of Data-Dependent Data Parallel Programs
Cincinnati, Ohio
August 15-August 18
ISBN: 0-7695-1315-8
ASCII Text
x 
 
Hasyim Gautama, Arjan J. C. Van Gemund, "Low-Cost Performance Prediction of Data-Dependent Data Parallel Programs," Modeling, Analysis, and Simulation of Computer Systems, International Symposium on, pp. 0173, Ninth IEEE International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunications Systems (MASCOTS'01), 2001.
 
 
BibTex
x
 
@article{ 10.1109/MASCOT.2001.948867,
author = {Hasyim Gautama and Arjan J. C. Van Gemund},
title = {Low-Cost Performance Prediction of Data-Dependent Data Parallel Programs},
journal ={Modeling, Analysis, and Simulation of Computer Systems, International Symposium on},
volume = {0},
year = {2001},
isbn = {0-7695-1315-8},
pages = {0173},
doi = {http://doi.ieeecomputersociety.org/10.1109/MASCOT.2001.948867},
publisher = {IEEE Computer Society},
address = {Los Alamitos, CA, USA},
}
 
 
RefWorks Procite/RefMan/Endnote
x 
 
TY - CONF
JO - Modeling, Analysis, and Simulation of Computer Systems, International Symposium on
TI - Low-Cost Performance Prediction of Data-Dependent Data Parallel Programs
SN - 0-7695-1315-8
SP
EP
A1 - Hasyim Gautama,
A1 - Arjan J. C. Van Gemund,
PY - 2001
VL - 0
JA - Modeling, Analysis, and Simulation of Computer Systems, International Symposium on
ER -
 
Hasyim Gautama, Delft University of Technology
Arjan J. C. Van Gemund, Delft University of Technology
Abstract: Current analytic solutions to the execution time distribution of an N-ary parallel composition of tasks having independent and identically distributed execution times are computationally complex, except for a limited number of distributions. In this paper we introduce an analytical solution based on approximating the execution time distributions in terms of a limited number of statistical moments. This approach allows the parallel execution time to be approximated with O(1) solution complexity for a wide range of execution time distributions, while the approximation ac-curacy outperforms comparable techniques known to date. Experiments show that the error of the predicted mean value of the parallel execution time is even less than 4% for parallel loops comprising up to 10,000 tasks whose execution times are normally distributed. Measurements on real programs (NAS-EP benchmark, PSRS sorter, and WATOR simulator) confirm these results provided the task execution distributions are independent and unimodal.
Citation:
Hasyim Gautama, Arjan J. C. Van Gemund, "Low-Cost Performance Prediction of Data-Dependent Data Parallel Programs," mascots, pp.0173, Ninth IEEE International Symposium on Modeling, Analysis, and Simulation of Computer and Telecommunications Systems (MASCOTS'01), 2001
Usage of this product signifies your acceptance of the Terms of Use.