A Module-Interlock Technique for Task Scheduling in a Multi-core Processor System

Abstract

Data dependence and mutual exclusion are problems must be overcome in the task management of multi-core systems for parallel computing. In this paper, we propose a directed acyclic graph (DAG) based task scheduling technique, module-interlock technique (MIT), to keep the processor utilization as high as possible, so as to obtain a shortest execution time. According to the critical path scheme on the DAG, the result of the task scheduling can be presented in a Gantt chart that is then translated into bricks for further scheduling. Bricks in the proposed system are categorized in four types, rectangular, protruding, concave, and uneven, according to the usage of processors. Applying the proposed technique by interlocking proper type of bricks to perform the task re-scheduling, a higher processor utilization with shorter execution time can be achieved.