Hypercube matrix multiplication (Q686633)

A matrix multiplication algorithm for parallel computers (hypercubes) is given. The algorithm has running times of \(O(n)\) and \(O(\log n)\) for \(n^ 2\) and \(n^ 3/2\) processors. The sub matrices are sent in a single communication to directly connected processors in the hypercube and are also involved in sequential matrix multiplication at each processor. Experimental evidence is provided, showing that the proposed algorithm performs well on both \(iPSC/1\) and \(iPSC/2\) hypercubes.