An Optimized Nature-Inspired Metaheuristic Algorithm for Application Mapping in 2D-NoC.

Area of Science:

• Computer Engineering
• Artificial Intelligence
• Hardware Architectures

Background:

• Mapping application task graphs onto intellectual property (IP) cores in Network-on-Chip (NoC) architectures is a computationally challenging problem.
• Network performance is significantly influenced by the efficiency of mapping techniques and the optimization of critical metrics like power, reliability, area, thermal distribution, and delay.

Purpose of the Study:

• To introduce a novel and efficient mapping technique for NoC architectures.
• To minimize power dissipation and improve mapping speed using a new optimization algorithm.

Main Methods:

• A state-of-the-art mapping technique named the Sailfish Optimization Algorithm (SFOA) is proposed.
• The algorithm employs a shared k-nearest neighbor clustering approach to empirically minimize power dissipation.
• The SFOA is evaluated on six standard benchmarks for performance comparison.

Main Results:

• The SFOA demonstrates quicker mapping times compared to existing methods across the evaluated benchmarks.
• The proposed technique effectively minimizes power dissipation in NoC architectures.
• Experimental results show superior performance of SFOA over other nature-inspired metaheuristic approaches, particularly for large application task graphs.

Conclusions:

• The Sailfish Optimization Algorithm (SFOA) presents a highly effective and efficient solution for mapping application task graphs onto NoC.
• SFOA offers significant advantages in terms of reduced power consumption and faster mapping times, outperforming existing metaheuristic methods.