A MEAN-VALUE ANALYSIS MULTIPROCESSOR MODEL INCORPORATING SUPERSCALAR PROCESSORS AND LATENCY TOLERATING TECHNIQUES

Several approximate Mean Value Analysis (MVA) shared memory multiproce ssor models have been developed and used to evaluate a number of syste m architectures. In recent years, the use of superscalar processors, m ultilevel cal:he hierarchies, and latency tolerating techniques has si gnificantly increased the complexity of multiprocessor system modeling . We present an analytical performance model which extends previous mu ltiprocessor MVA models by incorporating these new features and in add ition, increases the level of modeling detail to improve flexibility a nd accuracy. The extensions required to analyze the impact of these ne w features are described in detail. We then use the model to demonstra te some of the tradeoffs involved in designing modern multiprocessors, including the impact of highly superscalar architectures on the scala bility of multiprocessor systems.