Software energy profiling

A software energy estimation methodology is presented that avoids explicit instruction level analysis and predicts energy consumption to an accuracy within 3% for a set of benchmark programs evaluated on the StrongARM SA-1100 and Hitachi SH-4 microprocessors. A technique has also been proposed to separate the switching and leakage energy components of software. A web-based tool, JouleTrack, has been developed that incorporates the proposed techniques.