Energy Consumption Analysis of Software for Smartphones

Occurrences : 


Nombre d'étudiants minimum: 


Nombre d'étudiants maximum: 


Nombre d'instances : 


The service uptime of battery-powered devices, e.g., smartphones or tablets, is a sensitive issue for nearly any user. Optimizing the battery-lifetime of smartphones is therefore a hot topic in the research and the industry arenas. Minimizing energy consumption of electronic devices can be achieved at any stage in the device's conception process. On the other hand, energy can also be controlled during runtime, for example by clock gating active components, introducing sleeping states for I/O devices, or also via dynamic voltage frequency scaling (DVFS).

Dynamic voltage and frequency scaling (DVFS) allows the CPU to operate at a regime that is less energy demanding, while however trading of some performance. It is shown that by scaling the CPU frequency the energy consumption is convex for simple algorithms. This means there exist an optimal CPU frequency that is not the maximum nor the minimum CPU frequency, which minimizes the energy consumption.

Yet little is known about the generalizability of the energy-frequency graph's convexity. Therefore extended accurate measurements are to be conducted based on a variety of software benchmarks to verify the convexity of the energy-frequency graph. During the PAF software benchmarks will be run on a smartphone testbed while conducting power measurements. The impact of stressing different hardware parts on the energy-frequency graph's convexity will be analyzed. Eventually this should also allow us to define an energy profile for software programs (running on smartphones), similar to the energy profiles used for hardware circuits.

This project will be held in english.