Relaxing Event Densities by Exploiting Infeasible Paths in Control Flow Graphs

Common real-time analysis techniques for embedded systems mainly concentrate on a task model where ev-ery single activation of a task leads to a single outgo-ing event that is emitted at the end of the task’s compu-tation. An extension has been introduced that allows for multiple events to occur during a single job by respect-ing the control flow inside a task in order to recalculate the worst-case density for outgoing events. This previous work does however not consider the effect that data de-pendencies have on the calculation, which leads to pes-simistic results caused by infeasible paths inherent in the control flow. In this paper we propose a method involving a flow graph transformation that reduces this pessimism when using the existing analysis and show its integration into a reasonable workflow.