A Routing based Genetic Algorithm for Task Mapping on MPSoC
Hiago Mayk G. de A. Rocha, Antonio Carlos Schneider Beck, Sílvia Maria Diniz Monteiro Maia, Márcio Kreutz, Monica Magalhaes Pereira · 2020
This work proposes an optimized task mapping solution called Routing Model-based Genetic Algorithm (RMGA) that combines task mapping and routing problems using an Integer Linear Programming (ILP) model as a fitness function. We compared our proposed RMGA with other Genetic Algorithms (GA) that address the mapping problem using a classical flow x distance fitness function evaluation. Experimental results evaluating communication latency demonstrate that the proposed algorithm outperforms two GAs from literature. It presents up to 30% lower delay when simulating tasks communication pattern of six different applications and varying packet injection rate and NoC dimension. Additionally, RMGA presents the lowest delay of more than 70 % in all simulated scenarios compared to the other GAs.