POET: A tool for the analysis of the performance of parallel algorithms
Publication
, Journal Article
Lastra, AA; Starmer, CF
Published in: Proceedings of the International Conference on Parallel Processing
December 1, 1988
The tool consists of a simple language for describing algorithms and an interpreter that determines the execution time on a given number of processors. The key concept is the separate specification of local computation and remote memory access in the algorithm description. This allows the interpreter to simulate the communications of the target parallel machine. An accurate simulation of the memory access delays, coupled with the specified amount of local computation, results in the predicted parallel execution time. Operation of the system has been validated by comparing predicted versus observed execution times for numerical algorithms on the Butterfly and Butterfly Plus parallel processors.
Duke Scholars
Published In
Proceedings of the International Conference on Parallel Processing
ISSN
0190-3918
Publication Date
December 1, 1988
Volume
1
Start / End Page
126 / 129
Citation
APA
Chicago
ICMJE
MLA
NLM
Lastra, A. A., & Starmer, C. F. (1988). POET: A tool for the analysis of the performance of parallel algorithms. Proceedings of the International Conference on Parallel Processing, 1, 126–129.
Lastra, A. A., and C. F. Starmer. “POET: A tool for the analysis of the performance of parallel algorithms.” Proceedings of the International Conference on Parallel Processing 1 (December 1, 1988): 126–29.
Lastra AA, Starmer CF. POET: A tool for the analysis of the performance of parallel algorithms. Proceedings of the International Conference on Parallel Processing. 1988 Dec 1;1:126–9.
Lastra, A. A., and C. F. Starmer. “POET: A tool for the analysis of the performance of parallel algorithms.” Proceedings of the International Conference on Parallel Processing, vol. 1, Dec. 1988, pp. 126–29.
Lastra AA, Starmer CF. POET: A tool for the analysis of the performance of parallel algorithms. Proceedings of the International Conference on Parallel Processing. 1988 Dec 1;1:126–129.
Published In
Proceedings of the International Conference on Parallel Processing
ISSN
0190-3918
Publication Date
December 1, 1988
Volume
1
Start / End Page
126 / 129