Smartphone users rely on applications to perform various functionalities through their phones, but these function- alities may cause a significant drain on the device’s battery. To ensure that an app does not consume unnecessary energy, app developers measure and optimize the energy consumption of their apps before releasing them to the end users. However, current optimization and measurement techniques have several limitations. The energy optimization techniques only focus on refactoring energy-greedy patterns related to system events, such as garbage collection and process switching, and on providing recommendation models for API usage. Despite the fact that the energy consumption of a single API can vary depending on its configuration, and API events account for 85% of energy con- sumption in smartphone apps, existing optimization techniques do not provide guidance on how to configure APIs for energy- efficient usage. Moreover, energy measurement techniques are cumbersome because they require developers to generate test cases and execute them on expensive, sophisticated hardware. My thesis argues that we can develop a general methodology that researchers may follow to extract energy-efficient guidelines pertaining to an API, and developers may use such guidelines to develop energy-efficient apps. Additionally, it argues that we can use static analysis to estimate an app’s energy consumption. Such methodology will elevate the need for a physical smartphone and test case generation and execution. The insights and techniques that my thesis presents are particularly useful within the context of an Integrated Development Environment (IDE) or a Continu- ous Integration/Continuous Deployment (CI/CD) pipeline, where developers require results within a matter of milliseconds. Using our technique, developers would quickly receive warnings about high energy consumption caused by their code modifications, specifically those related to API usage.

Ph.D Student at University of Alberta