A Flexible Operational Framework for Energy Profiling of Programs

Energy has become a serious concern in the HPC ecosystem for various reasons. The level of the issue is from crucial to critical depending on the context. Considering the variety of hardware devices and the granularity of the desired measurements, addressing the problem of energy profiling clearly requires adequate support tools for measurements and (energy aware) monitoring. There are several frameworks that aim at estimating the energy consumed by a given program on a given computing system, however they mainly suffer from some short-comings related to flexibility, portability, accuracy and programmability. Thus, we propose a tool, so-named EA2P, that is reasonably flexible from both hardware and software standpoints, supporting the major parallel paradigms including GPU acceleration. The tool is publicly available at https://github.com/HPC-CRI/EA2P.