FPowerTool: A Function-Level Power Profiling Tool

Guang Wei,Lin Wang,Zhongzhi Luan,Hailong Yang,Depei Qian

doi:10.1109/access.2019.2961507

Guang Wei, Lin Wang + Show 3 more

Open Access

https://doi.org/10.1109/access.2019.2961507

Copy DOI

Abstract

This paper presents the design and implementation of FPowerTool, a lightweight function-level profiling tool for measuring the energy consumption of program code. Based on the trace of energy usage and the timestamps of programs execution events, FPowerTool monitors and analyzes the power consumption of a program at runtime. Two issues have been addressed: associating the power consumption data with the individual function code segments and reducing the overhead of power profiling. FPowerTool adopts an approach which takes advantage of both sampling and dynamic instrumentation to address those issues. The energy consumed by code at the granularity of function is measured by sampling the energy usage hardware counters built in the CPU and the sampled value is associated to the corresponding code segments by offline analysis. Experiment results show that this approach significantly reduces the overhead of the program power profiling without significant interference to the timing of the original program execution.

Highlights

Modern computing platforms such as datacenter servers or battery-powered mobile devices have played an indispensable role in people’s daily life
This paper introduces the design and implementation of a function-level power profiling tool FPowerTool
DESIGN OF FPOWERTOOL FPowerTool was designed to be a lightweight function-level power profiling tool which can be used for measuring the energy consumption of program code, even for the non-debugging binary codes

Summary

Introduction

Modern computing platforms such as datacenter servers or battery-powered mobile devices have played an indispensable role in people’s daily life. The energy consumption issue related to those equipment has been an increasingly significant concern. If we could improve the energy efficiency of the datacenters, there will be a big saving in energy. With the widespread usage of mobile computing devices, energy efficiency has become a big concern in hand-held devices such as mobile phones. The endurance (the time for the device to work continuously without recharging) of this kind of devices is limited by the capacity of battery. Higher energy efficiency means better endurance performance. Great efforts have been put on reducing the power consumption of the above computing platforms. Among them low-power hardware design and hardware/software coordinated energy-saving mechanism are two common approaches and have been studied

Methods

Results

Discussion

Conclusion