Task Scheduling Prediction Algorithms for Dynamic Hardware/Software Partitioning

Significant research work has been done in recent years for dynamic HW/SW partitioning and the researches are focused on the aspects of partitioning algorithm and implementation. As for the aspects of task scheduling prediction, no in-depth study has been done and no efficient prediction algorithm has been proposed. In the paper, the importance of task scheduling prediction is analyzed and the mathematical description is given. And then a period-detection-based task scheduling prediction algorithm is proposed according to the characteristics of task scheduling. After that, further two algorithms are proposed to reduce the computational complexity and to improve prediction correct rates. Finally, examples of digital audio video coding are investigated. In the experiments of task scheduling prediction of digital audio video coding, the prediction correct rates reach up to 97% for the first and second order sample individual prediction. Thus, the algorithms can predict the neighboring individuals after sample with a certain correct rate and provide reliable data reference for dynamic HW/SW partitioning.