Automated estimation of power consumption for rapid system level design

Abstract

This paper describes an early power estimation method for Electronic System Level(ESL) design, which provides a scalable API to support automated power profiling and analysis at the early stages of the design process. The proposed framework utilizes a high-level power modeling mechanism along with an automated profiler to extract energy activity from the simulated system model. These two features are integrated into PowerMeter, a framework that automatically annotates power meters as well as energy and performance functions into the executable model. This integrated profiling helps the designer to rapidly explore the design space, trading off performance against power cost in order to make best design decisions. Our approach also provides the designer with the ability to quantify the effect of revisions in the ESL design models, in terms of both power and performance. Despite the high abstraction level, our results show that the PowerMeter delivers rapid estimates with high fidelity and at minimal cost.