Utvärdering av programflödesanalys för WCET-beräkningar vid Volvo Construction Equipment (aw)

The purpose of Worst-Case Execution Time (WCET) analysis is to find safe upper bounds to the running time of the program. An important part of WCET analysis is to find constraints on the possible program flows, including upper bounds on the number of loop iterations, and information about infeasible paths. These constraints are not so simple to derive automatically, and current tools mostly rely the user providing the constraints by hand, as annotations to the analysed programs. To provide these constraints is time-consuming and error-prone, and it reduces the usability of the current WCET analysis tools.

The WCET project group at MdH/IDE specialises in methods for automatic program flow analysis. Our prototype tool SWEET has an advanced module for automatic flow analysis, which can find program flow constraints and pass them on to the subsequent WCET analysis.

The topic of this thesis project is to evaluate the automatic flow analysis of SWEET, in order to find out how precise it is: which program flow constraints will it detect, and which ones will it miss? The flow analysis will be run as a stand-alone tool in the experiments. It will be tested on real-time production code from Volvo Construction Equipment AB in Eskilstuna, and the thesis work will be mainly carried out on site. The production code has been previously analysed with the commercial WCET tool aiT, in an earlier M.Sc. thesis, and the automatically detected program flow constraints can be compared with the manually provided constraints which were used in that thesis project. If time permits, the automatically derived constraints will also be fed to aiT by hand, which then can be used for a direct comparison also of the computed WCET estimates.

This is a research-oriented thesis project, which can be rewarding for the right student but requires skills above average.