Energy-aware multi-threaded software systems: Measuring and modelling software energy consumption on a multi-threaded embedded processor architecture

My research seeks to further our understanding of energy consumption in embedded systems at the software level, where evidence suggests the most significant energy efficiency gains can be achieved. Effort at lower levels in the system stack, such as lower power process technologies and clock/power gating deliver limited benefits if the software running on top of them works against them.

Modern embedded processor architectures and the growth of parallel systems has created many new opportunities for advancing the state of the art in software energy modelling and optimisation. My research attempts to address three key questions:

• Can we accurately model software energy in an architecture with hardware multi-threading?
• What is the impact of communication within many core systems and what is the trade-off between communication, storage and computation?
• Can information on the protocols and timing behaviours of commonly used embedded interfaces be used within the application software to directly improve energy efficiency and expose opportunities for optimisations?