Energy-driven Statistical Profiling: Detecting Software Hotspots

Abstract

Energy is a critical resource in many computing systems, motivating the need for energy-efficient software design. This work proposes a new approach, energy-driven statistical profiling, to help software developers reason about the energy impact of software design decisions. We describe a prototype implementation of this approach for the Itsy pocket computing platform. Our experimental results using the prototype with 13 benchmark programs show that (i) energy measurement tools that ignore system/kernel effects can give erroneous results about energy hotspots, and (ii) using execution time profiles to develop intuition about energy usage can often be misleading.

8 Figures and Tables

Cite this paper

@inproceedings{Chang2002EnergydrivenSP, title={Energy-driven Statistical Profiling: Detecting Software Hotspots}, author={Fay Chang and Keith I. Farkas}, year={2002} }