Abstract
In real-time systems the execution time of a program is crucial, missing a deadline can have catastrophical consequences. Today the estimation of the worst execution time is mostly done by measurements with the worst possible input to the program. These measurements are not totally reliable; there is a chance that the worst execution path of the program is not caught in the measurements. There have been a lot of research performed in the WCET (Worst Case Execution Time) analysis field in the last couple of years, and models of how to theoretically estimate the WCET have been thoroughly described. But there haven't been many attempts at applying the models to actual real-time operating system code.

The goal with this thesis project is to use today's research in the WCET analysis field, especially the work by the ASTEC WCET-group in Sweden, to develop a tool that can be used on object-code for an ARM microprocessor. The tool should be able to transform the binary executable file of a program into control flow graphs with basic blocks, so that a safe (no underestimates) and tight WCET analysis can be calculated. In this case study the WCET of the interrupt latencies in the OSE real-time operating system. A part of the work is to determine how much work that has to be done by hand, e.g. through program-specific input from the user, and how much that can be automated.

Martin has worked with the ASTEC/WCET project and ENEA OSE Systems to test WCET analysis on a real real-time operating system kernel.

Do not miss the seminar by Tomas Östlund starting at 14.15 in the same room.

ASTEC seminar
March 19, 2002

Place: Information technology, Uppsala University
Room: 6003
Time: 13.15-14.00 (+ discussions)

Room 6003 is in building 6, floor 1, room "143 Seminarierum" on construction drawing, enter via the MIC-aula entree.