摘要
利用代码插装完成实时任务对自身剩余最长执行时间的运行时自我预估,以满足实时动态调度领域的相关需要.该方法以全面的、规范的相关性分析为基础,构建形式化模型,简洁一致地刻画程序的控制流和相关性;提出了一种高效的预估点代码生成算法;并给出了权衡预估精度和运行时开销的若干优化实现技术.
In order to meet relevant requirements in the area of real-time dynamic scheduling, a novel approach was proposed to enable real-time programs to estimate their own remaining worst-case execution times (RWET) by means of code instrumentations. Based on fully and specifically analyzing all the correlations in a program, a formal model was constructed to unify both control-flows and correlations. Then an efficient algorithm was presented to insert estimation codes at reasonable places. Besides, some optimizations for implementation were also introduced to trade off the estimation precisions against runtime overheads.
基金
国家自然科学基金(60273042)
安徽省自然科学基金(03042101)资助
关键词
实时动态调度
剩余最长执行时间
代码插装
编译器
real-time dynamic scheduling
remaining worst-case execution time
code instrumentation
compiler