Scheme: Industry Fellowship
Organisation: University of Bristol
Dates: Sep 2010-Sep 2011
Summary: Energy management on mobile electronics is paramount since the amount of energy available is limited by battery size or the effectiveness of energy scavenging methods (e.g solar). The demands made by sophisticated computing devices such as smartphones that need to operate unpluged for as long as possible can quickly deplete these energy sources. This problem has become more accute recently as the demand for advanced mobile electronic products (eg tablets, dual-core smartphones) continues to increase.
To be able to manage the available energy, it is necessary to be able to understand which are the energy sinks in the system and how the interactions among these sinks affects overall energy consumption. This should be done before the system is built since at that point wrong decisions cannot be correctly easily and could lead to product failures.
Previous research has focused on obtaining data regarding the energy requirements of the core processor as one of the main system components but much less attention has been paid to other parts of the system (display, memory, etc). More importantly and overall vision of the system from a power point of view is missing. This lack of knowledge means that the effects in energy consumption following changes in one of the system components are not well understood and can lead to wrong decisions taken during the system definition and development.
My current research has focused on investigating a new methodology for early stage power modelling and analysis. The approach is based on the exploration of the design space using statistical regression models and extends previous work that focuses on the microprocessor to include non-core components. The results highlight how the effects of power interdependency among the hardware components and the hardware and software affect the overall power consumption in the system.