Improving the Performance per Watt of Asymmetric Multicores via Online Phase Classification, Adaptive Core Morphing and Dynamic Thread Scheduling
Professor of Electrical and Computer Engineering
University of Massachusetts (UMass)
DEI - Sala Seminari
6 luglio 2012
Asymmetric multicores (AMPs) have been shown to outperform symmetric ones in terms of performance and performance/watt. The improved performance and power efficiency are achieved when the program threads are matched to their most suitable cores. Since programs change their computational needs during their execution, the best thread to core assignment will likely change with time. We have therefore, developed an online program phase classification scheme that allows swapping of threads when the current needs of the threads justify a change in the assignment. The architectural differences among the cores in an AMP can never match the diversity that exists among different programs and even between different phases of the same program. Thus, we must expect to see program phases where the designed cores will be unable to support the ILP that the program can exhibit. We therefore, propose in this talk a dynamic morphing scheme. This scheme will allow a core to gain control of a functional unit that is ordinarily under control of a neighboring core, during periods of dense computation with high ILP. This way, we dynamically adjust the hardware resources to the current needs of the application. Our results show that combining online phase classification and dynamic core morphing can significantly improve the performance/watt of most multi-threaded workloads.
Affiliation and memberships