Scalable parallel matrix multiplication on distributed memory parallel computers

Consider any known sequential algorithm for matrix multiplication over an arbitrary ring with time complexity O(N(alpha)), where 2 < alpha less than or equal to 3. We show that such an algorithm can be parallelized on a distributed memory parallel computer (DMPC) in O(log N) time by using N(alpha)/log N processors. Such a parallel computation is cost optimal and matches the performance of PRAM. Furthermore, our parallelization on a DMPC can be made fully scalable, that is, for all 1 less than or equal to p less than or equal to N(alpha)/log N, multiplying two N x N matrices can be performed by a DMPC with p processors in O(N(alpha)/p) time, i.e., linear speedup and cost optimality can be achieved in the range [ 1..N(alpha)/log N]. This unifies all known algorithms for matrix multiplication on DMPC, standard or nonstandard, sequential or parallel. Extensions of our methods and results to other parallel systems are also presented. For instance, for all 1 less than or equal to p less than or equal to N(alpha)/log N, multiplying two N x N matrices can be performed by p processors connected by a hypercubic network in O(N(alpha)/p + (N(2)/p(2/alpha)) (log p)(2(alpha-1)/alpha)) time, which implies that if p = O(N(alpha)/(log N)(2(alpha-1)/(alpha-2))), linear speedup can be achieved. Such a parallelization is highly scalable. The above claims result in significant progress in scalable parallel matrix multiplication (as well as solving many other important problems) on distributed memory systems, both theoretically and practically. (C) 2001 Academic Press.