A communication-aware and predictive list scheduling algorithm for network-on-chip based heterogeneous muti-processor system-on-chip. (March 2022)

• Record Type: Journal Article
• Authors: Yao, Yu; Song, Yukun; Ge, Hu; Huang, Ying; Zhang, Duoli
• Is Part Of: Microelectronics journal. Volume 121(2022)
• Publication Date: 2022-03
• Subjects: Heterogeneous system -- Network-on-Chip -- Parallel computing -- List scheduling; Microelectronics -- Periodicals; Microélectronique -- Périodiques; Microelectronics; Electronic journals; Journals - contents and abstracts; Periodicals; 621.3805
• DOI: 10.1016/j.mejo.2022.105367
• Languages: English
• ISSNs: 0959-8324

Abstract

Abstract: The execution efficiency of applications of network-on-chip based heterogeneous multi-processor system-on-chips highly depends on effective scheduling algorithms. In this paper, heterogeneous system list scheduling is combined with the characteristics of NoC-based system-on-chip mapping and scheduling algorithms, and a list-scheduling algorithm applicable to heterogeneous multi-processor systems-on-chips based on network-on-chip—communication-aware predictive priority task scheduling (CPPTS) is proposed. CPPTS has the features of prediction and communication-awareness. A prediction matrix in the PPTS is used for task prioritizing and processor selection. In the processor selection phase, a communication-aware method is adopted. Better scheduling results are achieved for applications on this architecture and the quadratic time complexity is maintained. The analysis of the experimental results of randomly generated DAGs and 2 real-world applications including Gaussian elimination and Cybershake shows that the performance of CPPTS is better than the other 5 algorithms when 2 different routing methods and 8 different system structures are applied.