| As an important branch of embedded systems,Mixed-Criticality Systems are widely welcomed because they can integrate multiple subsystems with different criticality levels on a shared hardware platform.Integration brings great challenges to the design and development of Mixed-Criticality Systems,among which task scheduling optimization is the most prominent.Designers should not only ensure that high criticality tasks can pass the certification,but also consider improving the resource efficiency of Mixed-Crticality Systems.At present,most researches on Mixed-Criticality Systems are based on Dual-Criticality Systems,that is,there are only two criticality levels.Therefore,based on the Dual-Criticality Systems and preemption threshold,we research the related issues of resource scheduling optimization.This thesis includes the following work:(1)Preemption threshold scheduling introduces a non-preemptive mechanism in full preemption scheduling,including the advantages of preemptive scheduling and non-preemptive scheduling,which not only effectively improves the schedulability,but also reduces the system stack usage.To improve the schedulability of task sets in Mixed-Criticality Systems under preemption threshold scheduling,we propose a heuristic priority assignment algorithm named MRT and use the maximal preemption threshold assignment algorithm to set preemption threshold for task sets.MRT algorithm,using optimal low-priority task retrieval and redundant time as heuristic factors,can solve the problem of priority and preemption threshold assignment with large task sets in the limit time.Experimental results show that the performance of MRT is very close to that of the optimal priority assignment algorithm,and MRT can effectively improve the schedulability of task sets in MCS by up to 88.9%,compared with other heuristic priority assignment algorithms.(2)To deploy Mixed-Criticality Systems on a multiprocessor,based on preemption threshold adaptive mixed-criticality(PT-AMC)scheduling,we study the assignment of tasks to processors on a multiprocessor,which optimizes the system stack usage while ensuring systems schedulability.We propose four algorithms: a branch and bound algorithm,which combines several effective pruning rules to find the task assignment with the smallest system stack usage;a simulated annealing algorithm,which can find a locally optimal task assignment in a limited time as a benchmark for comparison;a heuristic assignment algorithm called MUDP,which can provide a high-quality solution in polynomial time and greatly improves the schedulability of the systems;a heuristic assignment algorithm called MUDP-BM,which based on MUDP and backtrack and greatly reduces the system stack usage.(3)Semi-clairvoyant,as an emerging technology,can determine the execution mode of a task when it is released,which can effectively improve the scheduling of the Mixed-Criticality Systems.Preemption threshold scheduling is also an effective method to improve system schedulability and reduce system stack usage.Integrating preemptive threshold scheduling and semi-clairvoyant technology into the Mixed-Criticality Systems,we propose a model called PT-AMC-sem,and give the corresponding schedulability test.Compared with the existing methods,it can effectively improve the schedulability of the systems and reduce the system stack usage. |
