Abstract

A parallel algorithm is presented for the input of data arrays (partly, of images) into processors of a distributed computer system (CS). The algorithm minimizes the input time by means of effective routing of array fragments in the CS interconnection network. The algorithm modeling shows that it is optimum for the CS with topology of the E2-graph (torus) and the optimum D2-graph (circulant). The time of the array input is determined by the number N of the graph nodes and by the degree of the CS root node which contains the input array initially, and is not dependent upon the graph type (torus or circulant) for N>20.

File
tarkov.pdf107.98 KB
Issue
Pages
79-84