電子情報通信学会総合大会講演要旨
AS-1-1
A Fast and Exact Algorithm for Scheduling of Data-Parallel Tasks
Yang Liu・○Yining Xu・Lin Meng・Ittetsu Taniguchi・Hiroyuki Tomiyama(Ritsumeikan Univ.)
This paper studies a task scheduling problem which schedules a set of tasks on multiple cores. The proposed algorithm considers both task and data parallelisms simultaneously, and obtains the optimal scheduling length. The proposed algorithm employs depth first search to cover all possible results. Since the branch-and-bound methods are well designed, the average computation time is reduced remarkably. We also conduct several experiments to ensure that the proposed optimization algorithm runs fast and yields optimal results.