Power contracts: A formal way towards power-closure?!

Abstract

Since energy consumption continuously becomes a limiting factor for today's microelectronics, power-aware design space exploration won significant importance in the design flows. Being strongly dependent on future design decisions and low-level parameters, the challenge results, how to derive power estimates from uncertain knowledge about later implementation details. For that purpose, high-level approaches are available, which either perform top-down synthesis and a power characterization of the concrete low-level system or re-use abstract characteristics of high-level components to derive power models and to calculate the power consumption of the composed system. Hence, these approaches suffer either performance or accuracy, due to the tradeoff between generating and considering implementation details respectively due to the inaccuracy of abstractions. Additionally, reliability of such estimations is uncertain, since system and component power models lack general validity and a traceable provability within the composed, extra-functional design space of power, function and time. To address this lack of power-closure, this paper suggests power contracts to formalize power properties and as a foundation for a more traceable, provable and thus reliable power-aware design flow. For that purpose, we introduce the formal basics of contract-based design, discuss their improvements within the design flow and propose their application within the domain of power, giving an outlook on a formal way towards power-closure.