Failure of computer software in a hard real-time system leads to severe consequences and must be avoided by proving the correctness of the systems software. A prerequisite for this is the determination of an upper bound for the worst-case execution times (WCET) of the tasks in the system. We show that for modern CPUs, WCETs can be obtained by static program analysis methods even for CPUs with execution history sensitives components like caches and pipelines. This is the first time that complex CPU features (out-of-order execution, speculation, etc) have been included in a comprehensive and safe analysis. The approach presented in this thesis is able to handle the analysis of very complex architectures (PowerPC 755) by first modeling the CPU and peripherals of the system and then using abstractions on some components of the system to obtain an analysis. The analysis computes WCET for the basic blocks of the program by simulating the abstract system model. The correctness of the approach is shown. A tool has been built based on this approach, which was evaluated under reallife industry conditions by Airbus France in the course of the DAEDALUS project, showing the practical applicability of the methodology.
展开▼
