A framework for design space exploration of parameterized VLSI systems

The paper presents two new approaches to multi-objective design space exploration for parametric VLSI sys-tems. Both considerably reduce the number of simulations needed to determine the Pareto-optimal set as compared with an exhaustive approach. The first uses sensitivity anal-ysis while the second uses evolutionary computing tech-niques. Application to a highly parametric system-on-a-chip for digital camera applications shows the validity of the methodologies presented in terms of both accuracy of results and efficiency, measured as the number of simula-tions needed to determine the power/execution-time trade-off front.