Reducing Energy per Instruction via Dynamic Resource Allocation and Voltage and Frequency Adaptation in Asymmetric Multicores

With the advent of multicore processors the emphasis incomputation is moving from sequential to parallel processing. Still, applications that require strong sequential performance do not achieve their highest performance/power when executing on current multicoresystems. As the computational needs vary significantly across different applications and with time, there is a need to dynamically allocate appropriate computational resources on demand to suit the applications' current needs, in order to minimize the energy consumption. The Energy per Instruction (EPI) could be further decreased by dynamically adapting the voltage and frequency to better fit the changing characteristics of the workload. Not only can a core be forced to a low power mode when its activity level is low, but the power saved by doing so could be opportunistically re-budgeted to other cores to boost the overall system throughput. To this end, we propose a holistic solution to energy efficiency improvement by seamlessly combining heterogeneity, Dynamic ResourceAllocation (DRA) and Dynamic Voltage and Frequency Adaptation (DVFA) capabilities to adapt the core resources to the changing demands of applications. Our results show that the proposed scheme provides anEPI reduction of about 17.9% when compared to the baseline heterogeneous multicore, 14% when compared to the baseline heterogeneous multicore with DVFA only and about 16.5% when compared to the baseline heterogeneous multicore with DRA only.