ABSTRACT
In this chapter, a number of parameters are investigated which differ from those discussed thus far. The purpose here is to produce a comprehensive study of the shared memory referencing strategies and further evaluate the performance of the various algorithms under a variety of conditions. Several different shared memory storage and referencing methods are analyzed in the first section. The implementations of the Uniformly Distributed and Locally Cached schemes are described in detail in this section. A framework is presented which allows a straightforward comparison of these schemes using the task partitioning implementations discussed in the previous chapter. In the second section, the effect of machine param eters such as the operating system and architectural characteristics are evaluated in regard to algorithm performance. In the third section, a number of additional characteristics such as image and object complexity are varied to see how overall algorithmic performance is affected. The comparisons in this chapter are
intended to provide a broader base for determining the relative merits of each of the parallel approaches which have been implemented.
