Design space exploration using hierarchical composition of performance models

Bottom-up synthesis approaches based on the hierarchical composition of performance models have been proposed as a promising alternative to conventional top-down hierarchical synthesis approaches. This paper discusses problems related to the context-dependence of performance models and proposes possible solutions. Techniques for the composition of multi-dimensional performance models so that the efficiency of the design space exploration is maximized are also discussed. An active filter is used to demonstrate the accuracy and efficiency of the techniques discussed here.