Abstract de la publi numéro 4834

In this paper, we present a framework for computing the Worst-Case Execution Time of programs. Because the WCET analysis proves to be complex and time-costly, we suggest that most of th eeffort should be put into the critical parts of the program under analysis. They should be analyzed with a high level of accuracy while other parts are evaluated more roughly. We are currently implementing a generic and free framework satisfying openness and re-usability constraints. In its architecture, the program code is viewed as a spine where WCET information is attached by successive analysis passes. This approach should maximize the possibilities of combining different analyses and should facilitate the integration and implementation of WCET computation.