Multiprocessing Performance Models for Parallel Computation: Part II. Single Processor Transfers

摘要

One of the simplest ways of coordinating a number of processors in executing a task in parallel is to leave all the data transfers to a single master processor. In this report models are formulated for this system to estimate the reduction in execution time, and efficiency of operation. Detailed expressions are derived for the performance of such system. Six examples are given to illustrate the use of the theory.