Prof. Samuel Xavier de Souza's research targets scalability and energy efficiency of parallel systems by attempting to optimally match hardware and software. In recent years, the focus of their research has been on improving energy efficiency of computation. It proved to be viable to convert parallel software scalability into energy efficiency of parallel hardware. Simulation and results from real hardware show a factor of up to 3x energy reduction while maintaining performance. Currently, they are interested in the validation and refinement of power models for homogeneous and heterogeneous architectures in order to accurately predict energy savings.

Energy consumption, however, is generally second to performance as application design goal, mainly due to the absence of general programming tools suitable to understand and subsequently improve energy efficiency. There exist many tools to assess software performance, but there are limited tools that relate energy to the developer's code. Such information, when available, can be valuable to energy critical systems such as wireless sensors, IoT, and other systems that operate on batteries or need a lot of power for computation. Most programmers are not familiar with techniques used to measure the target hardware and therefore they must rely on an approach that makes energy transparent at the development level while being sufficiently accurate.

Together with Prof. Kerstin Eder, from the University of Bristol, Samuel has accepted the challenge to develop a framework with tools that can deliver clear and relevant information in order to empower the programmer to make energy a first class design goal. This should allow HPC systems to be more economic, battery-powered devices more autonomous, and energy harvesting more usable for the IoT.