Research Of Energy Efficient Strategy On Big.LITTLE Architecture For Wearable Devices

With the development of technology and wearable smart devices,wearable smart devices are gradually widely used in military affairs,medical treatment etc.Owing to its portability and usability,wearable smart device is widely spread and applied.Different situation needs special function.Then the application requirements are ever-expanding.Consumers hope it can not only be smaller and lighter,but guarantee system performance and reduce power consumption.For example,extend the use time and standby time.While system performance and power consumption are irreconcilable contradictions.Smaller size and better system performance will cause the increasing of power consumption.In one word,power consumption problem seriously constraints the further development of wearable smart devices.ARM big.LITTLE architecture is the typical example of performance heterogeneous multi-core processors.The ARM big.LITTLE architecture consists of multiple processors with different performance and power dissipation.Processing of various applications with different performance and power processors can effectively reduce power consumption.For reasonable process scheduling and power management,ARM big.LITTLE architecture can distribute system resources according to needs,which can take both high performance and low power consumption into account.At present,the mature scheduling algorithm or dynamic power frame is designed and optimized for SMP system.It is not suitable for ARM big.LITTLE performance heterogeneous multi-core architecture,and the complicated and changing application of wearable amart device.Therefore,after analyzing the shortcomings of existing scheduling algorithms and combining with the special application scenarios of wearable smart device,the HDBDB load balancing scheduling algorithm with dynamic threshold is proposed.The algorithm adjusts the process migration threshold according to the overall load of the system,and achieves load balancing while ensuring performance and fairness.It not only can reduce the power consumption effectively,but also adapts to the extremely complicated application scene of wearable smart device.On the other hand,although the traditional scheduler and dynamic power management strategy aim at allocating resources and reducing power consumption,their own frameworks are fragmented.It is bound to interact with each other,resulting in additional performance loss and power consumption.This paper improves the scheduler and dynamic power management system,makes HMPDB scheduler as the core,to achieve a unified goal HMPDB-EAS energy-saving scheduling framework.Through the scheduler CPU and application load statistical analysis,coordinate CPUFreq FM subsystem and CPUIdle.It can reduce the power consumption and meet the load requirements of the system and the performance requirements of the application.It can increase the CPU's sleep mode time while ensuring CPU performance,then further reducing the power consumption of the wear able device.The experimental result shows,the dynamic energy saving scheduling framework HMPDB-EAS designed in this paper is more suitable for the application of the wearable smart device.Through the coordination of the scheduler and the dynamic power management framework,the system can effectively reduce the power consumption of the system while ensuring the system performance.