一种嵌入式软件WCET估计新方法

A new WCET estimation method of enbedded software

在实时嵌入式系统设计中,计算在最坏情况下软件执行时间的上界是很必需的,它为软硬件划分和进程调度提供了依据.由于现代微处理器使用了基于Cache存储和指令预取技术,增加了准确确定这一上界的难度,为此提出了一种基于指令Cache和指令预取联合模型的嵌入式软件性能评估新方法.该方法通过使用控制流程图和Cache冲突图,在Cache分析中联合指令预取分析,使得估计最坏情况下嵌入式软件的执行时间上界更精确;并使用了整数线性预测方法,使得求解计算复杂度降低.实验结果表明该方法估计精度可以提高近5%.

It is necessary to compute the execution time upper bound of embedded software under the worst condition in embedded system design, which decides how to partition hardware and software and how to schedule processes. Modern microprocessors which use Cache memory system and instruction pre-fetching increase the difficulty to compute the upper bound accurately. A new method for estimation of embedded software′s performance based on instruction Cache and pre-fetching model is proposed, which uses control flow graph and Cache conflict graph and combines instruction pre-fectching into Cache analysis, and it makes the estimation of execution time upper bound more accurate, the use of ILP method makes the computation complexity decrease. The experimental results show that the proposed method can improve about 5% estimation precision.