Research On Task Scheduling Algorithm Of Single Core Real-time Embedded System Based On Cache Locking

Real-time embedded systems usually have strict real-time constraints.However,the application of cache makes it impossible to accurately predict the worst execution time of a task,which usually results in an overestimation of the worst execution time of a task,reducing the effectiveness of the scheduling algorithm and negatively affecting the real-time analysis of the system.Cache locking mechanism is introduced to accurately estimate the worst execution time of a task.The core idea is to lock certain instructions and data in the cache,so that these locked contents will not be replaced during task execution.Based on the accurately calculated worst execution time of a task,an efficient task scheduling algorithm can be developed to ensure that the time constraints of the real-time embedded system are met and its execution efficiency is improved.Therefore,the research on cache locking algorithm and task scheduling method is of great significance and industrial application value.This dissertation presents a detailed study of the cache locking problem for single tasks and the task scheduling problem for single-core real-time embedded systems.First,this dissertation investigates the I-Cache cache locking problem,proposes a dynamic I-Cache locking method based on a genetic algorithm to reduce the worst execution time of a single task,and compares them with two I-Cache locking methods by using a set of tasks from the MRTC benchmark.The experimental results show that the approach proposed in this dissertation achieves significant improvements compared to the other two methods.Secondly,this dissertation proposes a task scheduling method that integrates cache locking and task scheduling to minimize the worst-case execution time of real-time embedded systems.The approach in this dissertation can allocate appropriate cache space for each task and achieve feasible scheduling.We implemented this dissertation's approach using the SNU real-time benchmarking test suite and compared it with a non-preemptive random cache locking approach and a static analysis approach(STA).The experimental results show that the approach in this dissertation yields better performance.