PowerDrive: a fast, canonical POWER estimator for DRIVing synthEsis

Abstract

The computational complexity of a probability based combinational power metric lies in the creation of a BDD for each node in the circuit. We formalize the problem of finding an intermediate support set which controls the size of BDD. We propose an exact algorithm to solve it. We also propose an heuristic solution, PowerDrive, for estimating the power of large circuits. Apart from being more accurate and several times faster than methods by H. Choi and S.H. Hwang (1997) and B. Kapoor (1994), PowerDrive possesses the unique quality of being canonical and of constant complexity, a very desirable quality for a power metric guiding a synthesis tool. Finally, the proposed power metric was able to guide the synthesis tool (S. Roy et al., 1998) to optimize large circuits which could not be synthesized by POSE (S. Imam and M. Pedram, 1995), thus proving the effectiveness of our power metric.