System Level Power Optimization And Verification For Loongson 2K1000B Processor

System power optimization has always been an important research content of mobile device product design.The thesis is based on 2K1000 B mobile intelligent terminal,designed and implemented a system-level power consumption management method for its video decoding scenario and standby scenario.First,a power consumption evaluation model based on PMC(Performance Monitoring Counters)is proposed,which aims to evaluate the operating power consumption of the processor without adding external devices.For video decoding scenarios,V-DFS(Video Dynamic Frequency Scaling)strategy was proposed,and the CPU idle mechanism was applied.According to the decoding performance requirements of different videos,through the Qo S(Quality of Service)mechanism,the sampling period and frequency range are dynamically set to reduce the system decoding energy consumption.CPU idle mechanism is optimized by executing WAIT instruction in idle process.During standby,DCS(Dynamic Cores Scaling)strategy is applied to dynamically adjust the number of processor cores in the system according to the system load.Further,the STR(Suspend to RAM)strategy is applied to save the running state of the processor and suspend most of the hardware peripherals to reduce the system power consumption.Based on the power consumption model of PMC,the average relative error at Unix Bench is 2.14%.The VDFS strategy can reduce the processor power consumption by 3.05% compared to the default on-demand strategy in multiple resolution H.264 and MPEG2 format video decoding scenarios.Taken together,when the video decoding task is running,V-DFS and IDLE strategies can reduce the average processor power consumption by 7.62% and board-level power consumption by 8.22%.During standby,DCS and STR strategies can further reduce processor power consumption by 25.8% and board-level power consumption by 31.4%.