Abstract
This paper addresses the problem of scheduling tasks onto Non-Uniform Memory Access (NUMA) multiprocessors with a bounded number of processors. An algorithm is proposed to schedule tasks by considering intertask communication overhead. The proposed algorithm exploits the schedule-holes in schedules; therefore, it could produce better schedules than that produced by existing algorithms. The proposed algorithm ensures performance within a factor of two times of the optimum for general directed acyclic task graphs. Experimental results demonstrate the superiority of the proposed algorithm over that presented in literature. A sharper bound on the multiprocessor scheduling problem by considering communication delay is also shown.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call