Much research focuses on the development of specialized architectures. To increase the utility of task-oriented or specialized parallel networks, one can simulate one parallel network on another. Of primary interest are lower bounds on the cost of such simulation. This project investigates the actual performance of proposed theoretical embeddings strategies by simulating them on a hypercube machine. The contribution of this practical investigation will be to establish how quickly the increase in embedding costs prohibit an actual simulation. By simulating different architectures on the hypercube, one may compare how similar measures of embedding costs, and their growth, are affected by architectural design. For example, an edge congestion cost greater than one may be more deleterious for an asymmetric network. Furthermore, the effects of even a slight increase in such a cost may have more dramatic effects in asymmetric architectures. However, such differences may be minimized if the cost can be localized or redistributed. Pure embeddings do not consider these possible revisions but a simulation may suggest an efficient means for obtaining improvements.
