Time Complexity of a Matrix Product on Message Passing Architectures

Abstract

The use of an appropriate methodology for calculating the communication cost, time complexity and performing task scheduling in massively parallel processing systems plays a significant role in improving the functionality of a message-passing architecture. This paper presents an existing message passing architecture (MS3H 4-Cube) which is based on Master-Slave Super-Hypercube (MSSHP) topology. By applying matrix multiplication, we investigate the communication cost and time complexity on MSSHP topology. However, by means of mathematical simulation and graphical presentation, it is verified that the completion time of a matrix product is shorter when is applied on MSSHP compared with Hypercube (HP) topology.