Topological comparison of perfect shuffle and hypercube

Three new topological measures for static multiprocessor interconnection networks are defined and used, in conjunction with other topological measures, for a comparative analysis of such networks. The 3 new measures are: 1. disconnectivity of the network, which measures the ability to perform permutations, 2. looseness, which combines measures of different patterns of communications, and 3. cost-effectiveness of the network. Other previously known measures used in the analysis are maximal distance, fan-out time, and degree of connectivity. Applying these measures to a large set of static networks provides a comprehensive evaluation of those networks. Two of the more promising networks, the hypercube and the perfect-shuffle-nearest-neighbor networks, are given special attention. Although the hypercube promises better performance, the analysis shows that the cost-effectiveness of a large perfect-shuffle-nearest-neighbor network is preferable to that of a hypercube and is actually 2nd to none.