Previous work on the analysis of execution time of parallel algorithms has either largely ignored communication delays or has dealt with specific interconnection structures such as the perfect shuffle and the nearest ...