An investigation in energy consumption analyses and application-level prediction techniques

The rapid development in the capability of hardware components of computational systems has led to a significant increase in the energy consumption of these computational systems. This has become a major issue especially if the computational environment is either resource-critical or resource-limited. Hence it is important to understand the energy consumption within these environments. This thesis describes an investigatory approach to power analysis and documents the development of an energy consumption analysis technique at the application level, and the implementation of the Power Trace Simulation and Characterisation Tools Suite (PSim). PSim uses a program characterisation technique which is inspired by the Performance Application Characterisation Environment (PACE), a performance modelling and prediction framework for parallel and distributed computing