Power modeling of mobile applications using Hardware Performance Counters
No Thumbnail Available
URL
Journal Title
Journal ISSN
Volume Title
School of Science |
Master's thesis
Checking the digitized thesis and permission for publishing
Instructions for the author
Instructions for the author
Authors
Date
2011
Department
Tietotekniikan laitos
Major/Subject
Tietokoneverkot
Mcode
T-110
Degree programme
Language
en
Pages
[9] + 51
Series
Abstract
Power consumption issues are one of the most significant factors affecting the future of performance improvements in modern microprocessors. This is all the more important for mobile and handheld devices running on battery. This work demonstrates a first order, linear power model using Hardware Performance Counters (HPC) to estimate the CPU and memory power consumptions. The experiments are done on a Nokia N810 device based on a TI OMAP 2420 processor. We derive and validate a set of power weights using regression modelling for performance and power and present a general power model. We also present two application specific power models, one for audio player application and one for a video player application. The use of HPC for performance-power prediction seems to be more popular than simulator based approaches which are time consuming and error prone. On most processors, the number of HPCs which can be monitored simultaneously is very limited. To the best of our knowledge, this is the first model based on just 3 HPC. We present our preliminary results from our experiments and discuss how in future this model can be combined with power modelling for data transmissions to create a full system energy profiler of mobile devices.Description
Supervisor
Ylä-Jääski, AnttiThesis advisor
Xiao, YuKeywords
power modeling, performance counter, mobile devices, regression modeling