JetsonLeap: A Framework to Measure Energy-Aware Code Optimizations in Embedded and Heterogeneous Systems

Energy-aware techniques are becoming a staple feature among compiler analyses and optimizations. However, the programming languages community still does not have access to cheap and precise technology to measure the power dissipated by a given program. This paper describes a solution to this problem. To this end, we introduce Jetson-Leap, a framework that enables the design and test of energy-aware code transformations. JetsonLeap consists of an embedded hardware, in our case, the Nvidia Tegra TK1 System on a Chip Device, a circuit to control the flow of energy, of our own design, plus a library to instrument program parts. We can measure reliably the energy spent by 400.000 instructions, about half a millisecond of program execution. Our entire infra-structure - board, power meter and circuit - can be reproduced with about $500.00. To demonstrate the efficacy of our framework, we have used it to measure energy consumption of programs running on ARM cores, on the GPU, and on a remote server. Furthermore, we have studied the impact of OpenACC directives on the energy efficiency of high-performance applications.