Binary Multi-Objective Salp Swarm Optimization for Task Mapping in NoC-based Heterogeneous MPSoCs

Abstract

This work presents an innovative approach for optimizing static task mapping of computation-intensive applications onto Network-on-Chip (NoC)-based heterogeneous Multi-Processor Systems-on-Chip (MPSoCs). The objective is to minimize both makespan (time) and total energy consumption, critical metrics for embedded systems performance. Task mapping, an NP-hard problem, assigns application tasksto processing elements in a 2D mesh NoC, significantly influencing system efficiency. To address this, we propose a Binary Multi-Objective Salp Swarm Algorithm (BMSSA), leveraging Pareto dominance and an external archive with crowding distance to maintain diverse, non-dominated solutions. Evaluated on benchmark applications with 10 to 50 tasks mapped to 4 to 20 processors, BMSSA achieves a reduction of up to 50% in makespan and 9% in energy compared to state-of-the-art methods, including NSGA-II, MOPSO, and MOACO. Statistical analysis via Wilcoxon signed-rank tests confirms BMSSA’s significant improvements, particularly over MOPSO. These results underscore BMSSA’s scalability and effectiveness for energy-aware, performance-optimized task mapping in next-generation MPSoCs.