A Bandwidth Reservation Mechanism for AXI-based Hardware Accelerators on FPGAs

Abstract

In dual priority scheduling, periodic tasks are executed in a fixed-priority manner, but each job has two phases with different priorities. The second phase is entered after a fixed amount of time has passed since the release of the job, at which point the job changes its priority. Dual priority scheduling was introduced by Burns and Wellings in 1993 and was shown to successfully schedule many task sets that are not schedulable with ordinary (single) fixed-priority scheduling. Burns and Wellings conjectured that dual priority scheduling is an optimal scheduling algorithm for synchronous periodic tasks with implicit deadlines on preemptive uniprocessors. We demonstrate the falsity of this conjecture, as well as of some related conjectures that have since been stated. This is achieved by means of computer-verified counterexamples.