Application of an Adaptive Genetic Algorithm for Task Mapping Optimisation on a Wormhole-based Real-time Network-on-Chip
Jessé Barreto de Barros, Maurício Ayala-Rincón, Carlos Humberto Llanos · 2019
The task mapping problem of real-time applications onto a homogeneous multiple processors system-on-a-chip (MPSoC) is an NP-complete problem that can be addressed using search-based meta-heuristic methods. A particular case is the priority preemptive wormhole-based Network-on-a-Chip (NoC) communication architecture that takes into consideration the end-to-end schedulability analysis. In this paper, we present a Adaptive Genetic Algorithm (AGA) with adaptive parameter control strategies that is capable of efficiently achieving schedulable task placements. In our experiments, we perform analytic methods in different scenarios with multiple synthetically generated real-time applications, as well as various platforms with a ranging number of processors to provide a statistical study comparing our algorithm against a standard genetic algorithm approach.