Preemption-Aware Allocation and Deadline Assignment for Conditional DAGs on Partitioned EDF

International audienceComplex heterogeneous hardware platforms are increasingly used for implementing critical real-time applications like ADAS and autonomous driving. To better support real-time workloads, GPUs have evolved to allow preemption for computationally intensive tasks and for graphical tasks. However in some cases the cost of preemption can be very high, and must be accounted for in the design and in the scheduling analysis.In this paper, we address the problem of allocating a set of real-time tasks, modeled by conditional directed acyclic graphs, onto multiprocessor platforms under partitioned preemptive Earliest Deadline First scheduling, assuming a non-negligible cost of preemption. We propose methods for assigning intermediate deadlines and offsets to real-time C-DAGs, so to remove unnecessary preemptions and reduce the total preemption overhead. The effectiveness of the proposed techniques is evaluated using a large set of synthetic tasks sets