Fast and Scalable Run-time Scheduling

As explained in previous chapters, a run-time scheduler is indispensable to efficiently explore the design space and make system level trade-off according to the dynamic context. For that sake, a fast and effective heuristic is needed. In this chapter, we first review again why we need a two-phase approach for task scheduling and how it is applied. The problem is then defined in a more formalized way and a greedy heuristic is described. After that, experimental results on both randomly generated and real-life applications are explained. In this chapter, we will illustrate our method on 2-dimensional Pareto trade-offs with execution time vs energy as axes. But the underlying techniques can also be applied to other axes and more dimensional trade-offs, which will be demonstrated in the next chapter.