A structure-oriented power modeling technique for macrocells

摘要

To characterize the power consumption of a macrocell, a generalnmethod involves recording the power consumption of all possible inputntransition events in the look-up tables. However, though this approachnis accurate, the size of the table becomes very large. In this paper, wenpropose a new power modeling technique that takes advantage of thenstructural information of a macrocell. In this approach, a subset ofnprimary inputs and internal nodes in the macrocell are selected as thenstate variables to build a state transition graph (STG). These statenvariables can model the steady-state transitions completely. Moreover,nby selecting the characterization patterns properly, the STG can alsonmodel the glitch power in the macrocell accurately. To further simplifynthe complexity of the STG, an incomplete power modeling technique isnpresented. Without losing much accuracy, the property of compatiblenpatterns is exploited for a macrocell to further reduce the number ofnedges in the corresponding STG. Experimental results show that ournmodeling techniques can provide SPICE-like accuracy, while the size ofnthe look-up table is significantly reduced