Approximation algorithm for scheduling applications on hybrid multi-core machines with communications delays

Massinissa Ait Aba (1) , Lilia Zaourar (1) , Alix Munier-Kordon (2)

1 DACLE-LIST - Département d'Architectures, Conception et Logiciels Embarqués-LIST
2 ALSOC - Architecture et Logiciels pour Systèmes Embarqués sur Puce

This paper presents an efficient algorithm with performance guarantee (approximation algorithm) to solve task scheduling problem on hybrid platform. The underlying platform architecture in this work is composed by two types of resources CPU and GPU, often called hybrid parallel multi-core platforms. We consider here for each type of resource identical nodes with communications delays. We focus in finding a generic approach to schedule applications presented by DAG (Directed Acyclic Graph) that minimizes makespan by considering communication delay between processors and tasks. A 6-approximation scheduling algorithm is proposed and evaluated in comparison to exact solutions and to another method. We demonstrate that the proposed algorithm achieves a close-to-optimal performance. Finally, our algorithm has been experimented on a large number of instances. These tests assess the good practical behavior of the algorithms with respect to the state-of-the-art solutions whenever these exist.

Mots clés: hybrid CPU GPU; approximation algorithm; makespan; DAG applications; scheduling

Domaines: Informatique [cs]

Massinissa Ait Aba, Lilia Zaourar, Alix Munier-Kordon. Approximation algorithm for scheduling applications on hybrid multi-core machines with communications delays. IPDPS Workshops 2018, 2018 IEEE International Parallel and Distributed Processing Symposium Workshops, May 2018, Vancouver, Canada. pp.36-45, ⟨10.1109/IPDPSW.2018.00016⟩. ⟨hal-02073600⟩