Title :
Modeling Unusual Behavior of Parallel Algorithms
Author_Institution :
Department of Computer and Information Science, Ohio State University
Abstract :
A probabilistic model of a class of parallel programs is used to investigate the counterintuitive behavior observed for some parallel algorithms. Two main points are made: 1) it may, in general, be beneficial to consider using more logical processes than physical processors in a parallel algorithm; and 2) results from order statistics are useful tools in analyzing parallel systems.
Keywords :
Integer programming; multiprocessor systems; order statistics; parallel algorithms; performance analysis; Algorithm design and analysis; Analytical models; Concurrent computing; Information science; Multiprocessing systems; Parallel algorithms; Parallel programming; Performance analysis; Random variables; Statistical analysis; Integer programming; multiprocessor systems; order statistics; parallel algorithms; performance analysis;
Journal_Title :
Computers, IEEE Transactions on
DOI :
10.1109/TC.1982.1675930
