Research On Non-CPU Energy Saving Scheduling Technology In Embedded System

With the rapid development of integrated circuit technology and computer technology,performances of embedded systems have improved greatly.But at the same time,power issues have become an important constraint on system design and development.CPUs are traditionally considered as the most energy-intensive component in embedded systems,and the energy-saving studies in CPUs have become very mature.With the increasing of memory requirements and extensive application of real-time systems,reducing the power consumption of non-CPU components—memory and devices,has become an important part of reducing system power consumption.The current memory's energy-saving technologies usually take advantage of the low-power memory default state to achieve energy conservation by combining hardware and software.However,it does not take into account the relationship between tasks and memory access behaviors,and its energy-saving effect is limited by memory hardware.Currently,the energy-saving effect of devices' energy-saving technologies is limited by the manufacturing processes,supply voltages,frequencies and others factors.In addition,with the development of real-time applications and the research of hard-soft real-time task models,it needs to consider the energy-saving issues when the hard-soft real-time tasks execute in the devices.This paper puts forward the energy-saving scheduling technology for memory and devices,which meets the requirements of real-time of tasks and reduces the energy consumption of components.The study of this paper includes the following two points:(1)This paper presents a memory energy savings scheduling technology based on task scheduling for memory.It is based on the memory access behavior of tasks,and it establishes the corresponding relationship model between tasks and memory blocks.The task scheduling module uses this model to schedule tasks with the similar memory access behavior,which reduces the state switching energy consumption of memory blocks in order to achieve the purpose of memory energy savings.(2)This paper presents an energy-savings scheduling technology based on taskscheduling for devices.This technology is based on(m,k)-firm to classify the jobs of the periodic tasks and relax the response time of partial jobs.The task scheduling module uses the relaxed response time to schedule jobs as centralized as possible and reduces the energy switch consumption of device states,so as to achieve the effect of memory energy savings.In this paper,we get the data of component's state switches and energy consumption by software simulation,and analyze the experimental results.The experimental results show the methods proposed in this paper can obtain better performance in collecting the valid sleep time of components and the energy savings than the existing algorithms.