Accurate Run-Time Prediction of Performance Degradation under Frequency Scaling
David Snowdon, Godfrey Van Der Linden, Stefan Petters and Gernot Heiser
Workshop on Operating Systems Platforms for Embedded Real-Time applications (OSPERT 2007)
Pisa, Italy, July 3, 2007
Summary
Dynamic voltage and frequency scaling is employed to minimise energy consumption in mobile devices. Typically devices are subject to timeliness or quality-of-service constraints. To meet those under frequency scaling, the performance at reduced frequency must be accurately estimated. The frequently-made assumption that performance scales linearly with core frequency has shown to be quite wrong, and better performance models are required, taking into account the effects of the memory architecture. This paper presents a methodology based on off-line hardware characterisation and run-time workload characterisation. Its evaluation shows that it provides a highly-accurate prediction of performance at arbitrary frequency settings.