An energy efficiency task scheduling algorithm for streaming applications on multiprocessor SoC

In this paper, we propose a two-step V/F assignment and task scheduling algorithm (VATSA) using DVS and DPM techniques for energy efficiency of streaming applications on multiprocessor System-on-Chip (MPSoC). In this algorithm, the dependencies of tasks are eliminated by software pipelining. The optimal voltage/frequency (V/F) assignment is achieved from the solution of the corresponding linear programming (LP) problem. Then, sub-tasks are scheduled to processors with identified V/F distributions. The V/F assignment and task scheduling are executed separately, which simplifies the problem and reduces computations. Experiments show that our algorithm can achieve a 15%~30% energy reduction compared with the approach in [1].