中图分类号:TN47 文献标志码:A DOI: 10.16157/j.issn.0258-7998.256828 中文引用格式: 潘福跃,闫俊启,郑利华,等. 基于改进遗传算法的多芯粒NoC低功耗映射[J]. 电子技术应用,2026,52(1):33-37. 英文引用格式: Pan Fuyue,Yan Junqi,Zheng Lihua,et al. Power-efficient mapping for multi-chiplet NoC using enhanced genetic algorithm[J]. Application of Electronic Technique,2026,52(1):33-37.
Power-efficient mapping for multi-chiplet NoC using enhanced genetic algorithm
Pan Fuyue1,Yan Junqi2,Zheng Lihua1,Xu Xiaobin2
1.No.58 Research Institute of China Electronics Technology Group Corporation;2.College of Mechanical and Electrical Engineering, Hohai University
Abstract: This paper addresses the power-efficient mapping problem in multi-chiplet Networks-on-Chip (NoC) by proposing an improved Adaptive Genetic Algorithm (AGA). By introducing permutation encoding, a partial-mapping crossover operator, an adaptive swap mutation strategy, and a hybrid selection mechanism, the algorithm effectively resolves issues such as constraint conflicts, local optima, and solution space explosion that exist in traditional genetic algorithms for NoC mapping. Experiments were conducted on a 36-node 2D-Mesh topology with randomly generated communication task graphs, comparing the performance of AGA, Ant Colony Optimization (ACO), and Grey Wolf Optimizer (GWO). The results demonstrate that AGA significantly outperforms other algorithms in communication energy optimization, reducing total power consumption by 32.0% and 26.2% compared to GWO and ACO, respectively. Additionally, AGA exhibits superior global search capability and convergence stability. This research provides an efficient optimization method for power-efficient NoC design.
Key words : network-on-chip;power-efficient;genetic algorithm;optimal transmission path