Abstract
Fast and accurate power analysis tools are crucial for effective system design evaluations. However, due to the tight coupling between power models and simulators, existing techniques are either unsatisfactorily slow or inaccurate. In this thesis, we observe that a high-level instruction execution always triggers a same set of resources and leads to same activity effects. Hence, we propose a new idea that maps instructions to microarchitecture components for efficient resource-oriented power evaluations. By pre-characterizing the instruction-resource effects in details, we can efficiently compute accurate power values using high-level instruction-set simulators. Furthermore, based on the concept of resource we can accurately and effortlessly capture the power waveform at any time point for power profile, peak power and dynamic thermal distribution analysis. The experimental results show that the proposed approach is nearly as accurate as gate-level simulators, with less than 1.2% error rate while achieving a simulation speed up to 20 MIPS, five orders faster than a commercial gate-level simulator.