Some parallel sorts on a mesh-connected processor array and their time efficiency

Journal Article · Mon Sep 01 04:00:00 EDT 1986 · J. Parallel Distrib. Comput.; (United States)

DOI:https://doi.org/10.1016/0743-7315(86)90023-7· OSTI ID:6849179

Sado, K; Igarashi, Y

In this paper the authors study properties of the parallel bubble sort and show some efficient parallel sorting algorithms with very simple control hardware. They introduce an interesting function COUNT (upsilon, s) on the set of pairs of values and stages, and prove that the function value of COUNT (upsilon, s) is the exact number of stages necessary to route value upsilon at stage s to its final position. This function is a powerful tool for designing efficient parallel sorting algorithms and for analyzing their time efficiency. The authors design parallel sorting algorithms that can be implemented on a mesh-connected parallel processor array. It is proved that the time complexities of some of our algorithms are O (n log n). The algorithms are faster than the parallel bubble sort, but they are asymptotically slower than the implementations of Batcher's algorithms on the mesh-connected model. However, in the average case the algorithms are as fast as those implementations of Batcher's algorithms for practical values of n, 1 delta n delta 128.

• Research Organization: Dept. of Computer Science, Gunma Univ., Kiryu 376
• OSTI ID: 6849179
• Journal Information: J. Parallel Distrib. Comput.; (United States), Journal Name: J. Parallel Distrib. Comput.; (United States) Vol. 3:3; ISSN JPDCE
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990210* -- Supercomputers-- (1987-1989)
ALGORITHMS
ARRAY PROCESSORS
COMPUTER ARCHITECTURE
COMPUTERS
DIGITAL COMPUTERS
MATHEMATICAL LOGIC
PARALLEL PROCESSING
PERFORMANCE
PROGRAMMING
SORTING
SUPERCOMPUTERS
TIME DEPENDENCE