Static WCET Analysis of Task-Oriented Code for Construction Vehicles
Daniel Sehlberg · 2005
When creating real time systems for embedded vehicle computers it is important to know how they will behave during the worst possible scenario. To know each task’s WCET (Worst Case Execution Time) is a prerequisite for being able to know if their deadlines always will be met. A missed deadline in a real time system can lead to severe consequences. Volvo CE is using the Rubus real time operating system for their vehicles. Before Rubus is creating the task schedule, the user has to specify the period time, offset time and WCET for each task. The estimated WCET for each task is today based on measures from automatic tests or by experiences. However, it can not be guaranteed that these automatic tests have found the longest possible path through each task. This can result in a too low estimate of the WCET, which means that the behaviour of the system is unreliable. This thesis investigates if WCET analysis tool aiT can be used to find safe WCET estimates for the tasks in Volvo CE’s applications, and how much user input is needed to get the result. Analysis results from aiT are compared to measured results and the WCET estimates used today. We have found that the times set in Rubus are in many cases greatly overestimated, and that Volvo CE would have good use of analyzing the tasks with aiT in order to get tighter WCETs. By giving annotations for the tool, the WCET estimates can become even tighter, but this also requires some manual work.