One Dimensional Optimization on Multiprocessor Systems

This paper presents a straightforward approach to determining how best to utilize an MIMD multiprocessor in the solution of one dimensional optimization problems involving continuous unimodal functions and nongradient search techniques. A methodology is presented which allows one to consider a variety of speedup functions which may occur in parallel function and systems evaluation. It is shown how the best of two parallel optimization strategies can be determined for a given accuracy, number of processors and speedup function