Scalable low power computing via scheduling on subsets of multicore processors / by Faisal Yousef Hamady.

Abstract

Given the accelerated growth in tablet devices, smartphones, and netbooks, designers are faced with serious challenges to meet the needs of mobility in terms of battery life and form factor. It is vital to exploit the architectural features of todayΓÇÖs hardware platforms to provide innovative solutions that would deliver the best mobile experience to users while ensuring adequate levels of performance. In this thesis, we present a novel approach to power management for multicore processor systems by exploiting the operating system scheduler. Using the various power levels and the utilization statistics of the cores, we limit the execution of software threads to a subset of the available cores while leaving the others idle to allow them to enter into deeper power-saving states. To show the potential gains from a system power management perspective, we have implemented our scheme on a mobile platform featuring the Second Generation Intel Core i5 processor, and tested it on a wide selection of workloads and benchmarks. Our experimental results show significant thermal power reduction (up to 61percent) in a variety of scenarios, while system performance was sustained in most cases but sacrificed in a few other uncommon situations.