Title :
Performance of asynchronous linear iterations with random delays
Author :
Moga, Adrian C. ; Dubois, Michel
Author_Institution :
Dept. of Electr. Eng. Syst., Univ. of Southern California, Los Angeles, CA, USA
Abstract :
In this paper we investigate the speedup potential of asynchronous iterative algorithms over their synchronous counterparts for the special case of linear iterations. The space of linear iterations of size two is explored by simulation and analytical methods. We find cases and conditions for high asynchronous speedups. However, averaging asynchronous speedups over the whole set of iteration matrices reveals that the domain of matrices with high speedups is not sizable and that asynchronous iterations perform no better than synchronous ones over the entire convergence space
Keywords :
convergence of numerical methods; delays; iterative methods; mathematics computing; matrix algebra; parallel algorithms; random processes; synchronisation; analytical methods; asynchronous iterative algorithms; asynchronous linear iterations; convergence space; high asynchronous speedups; iteration matrices; parallel algorithm; performance; random delays; simulation; speedup potential; Analytical models; Convergence; Damping; Delay; Eigenvalues and eigenfunctions; Equations; Iterative algorithms; Iterative methods; Stochastic processes; Sufficient conditions;
Conference_Titel :
Parallel Processing Symposium, 1996., Proceedings of IPPS '96, The 10th International
Conference_Location :
Honolulu, HI
Print_ISBN :
0-8186-7255-2
DOI :
10.1109/IPPS.1996.508122
