A distributed scheduling framework for asymmetric multi-core systems based on gridware

Abstract

Asymmetric multi-core systems represent an increasingly popular class of computer architecture due to their potential performance and power efficiency. Gridware is an innovative and effective computing model used to design asymmetric multi-core systems. However, scheduling of fusions in Asymmetric multi-core systems is still a challenging problem.This work proposes a distributed scheduling framework to solve this problem. It maps the atomic fusion to nodes of corresponding type, and interleaves the execution of fusions to improve the utilization of atomic fusion nodes. The scheduling decision is made asynchronously and independently on each atomic fusion node. Moreover, the state of fusions is subdivided. The experimental result shows that the distributed scheduling framework achieves high utilization of the asymmetric multi-core platform.