Multi-core Chip Dynamic Power Management Framework Based on Reinforcement Learning br

Multi-core chips can provide mighty computing capability for mobile intelligent terminals, but their performance is constraint by thermal and power issues. For this problem, this paper proposes a multi-core chipdynamic power management framework based on reinforcement learning. First, based on GEM5, a dynamicvoltage and frequency scaling simulation system of the multi-core chips is established. Second, a chip powermodel characterization method is adopted, which takes CMOS physical characteristics into consideration torealize online real-time power monitoring. Finally, a gradient reward method for the multi-core chips isdesigned, and a Deep Q Network (DQN) algorithm is used to learn the power management strategy for themulti-core chips. Compared with conventional Ondemand and MaxBIPS schemes, the simulation results showthat the proposed framework achieves 2.12% and 4.03% improvement in computational performance of the multi-core chips respectively