Safety-Critical Deadline Compliant Static Task Scheduling for Heterogeneous Architectures

Abstract

Currently, the prevalence of heterogeneous multicore architectures even in real-time safety-critical embedded systems is rising, as an increasing number of applicationspecific accelerators are introduced into the core architecture to tackle more and more complex problems. This does create a challenge for the modern developer, as the hardwaresoftware mapping is consequently becoming more complex as well, increasing development times. The developer can be assisted by leveraging automated scheduling algorithms, which calculate optimal hardware-software mappings and execution orders. As safety-critical timing constraints are a critical part of realtime embedded systems, being able to quantify the deterministic safety-compliance of the system at compile-time is a considerable upside of static scheduling approaches. However, to be able to fully incorporate static scheduling in the development cycle of safety-critical systems, solutions that allow for automated safetycompliance are needed. In this paper, we show how existing static scheduling methods can be extended to calculate schedules with minimal overall latency while respecting one or more intermediate deadlines. We utilize this to achieve full deadline compliance for up to three simultaneous intermediate deadlines with an average accuracy of 99% for easy problems, and up to 69.3% for very hard problems.