An adaptive dynamic power management prediction strategy of DSP

In order to reduce CPU power consumption in portable systems, DPM (Dynamic Power Management) techniques is commonly used in the CPU low power design. This paper analyzes the time-out, prediction and random strategy of DPM, proposed an adaptive prediction-correction strategy, which makes up the deficiencies of the current prediction algorithm, that the prediction precision of algorithm in the case of local service requests non-steady state. The proposed algorithm is validated by the simulation experiment with DSP. Experimental results show that the strategy can significantly reduced power consumption of SOC (System on Chip).