MELTA: A Method Level Energy Estimation Technique for Android Development

Development of energy-aware applications requires the understanding of energy consumption at the fine-grained level of methods and functions. Several hardware and software based techniques have been proposed for estimating the energy profile of mobile applications in order to assist developers in developing energy-efficient applications. However most of these energy estimation techniques work at the application level. Developers need to be aware of the energy consumed by each module of their application in order to generate energy efficient code. In this work we propose a probabilistic technique for fine-grained energy estimation to compute the energy consumption of each method in Android applications. We extract the relevant energy consumption and execution traces using Android studio tools including dmtracedump and BatteryHistorian. We generate power profiles and execution traces of fifteen open-source applications develop an accurate energy model, MELTA, for method-level energy estimation. MELTA uses a probabilistic approach to map application power profile onto the corresponding execution traces to accurately compute method-level energy.