Energy-aware scheduling of malleable fork-join tasks under a deadline constraint on heterogeneous multicores

Abstract

This paper proposes an energy-aware scheduling of malleable fork-join (MFJ) tasks on heterogeneous multicores. This work allows a task to be split into multiple sub-tasks for fork-join parallel execution. The number of the sub-tasks is determined simultaneously with scheduling. Our scheduling technique aims at the minimization of energy consumption under a deadline constraint. In addition, this paper proposes a technique for simultaneous scheduling and core-type optimization. The technique optimally decides types of cores (to be either "big" or "little") at the same time as MFJ task scheduling in order to further reduce energy consumption.