A comparison between evolutionary and local search techniques applied to NoC design space exploration. (30th October 2020)

• Record Type: Journal Article
• Authors: Silva, Jefferson; Maia, Sílvia M.D.M.; Magalhães, Monica; Kreutz, Márcio E.
• Is Part Of: International journal of innovative computing and applications. Volume 11:Number 4(2020)
• Page Start: 193
• Page End: 203
• Publication Date: 2020-10-30
• Subjects: evolutionary computing -- networks on chip -- NoC -- genetic algorithm -- iterated local search -- ILS -- memetic algorithm -- design space exploration; Evolutionary computation -- Periodicals; Neural networks (Computer science) -- Periodicals; Genetic programming (Computer science) -- Periodicals; Biologically-inspired computing -- Periodicals; Swarm intelligence -- Periodicals; Quantum computers -- Periodicals; 006.3
• Languages: English
• ISSNs: 1751-648X

Abstract

Networks on chip emerged as a communication architecture that overcame the limitations of bus architecture. As more cores were being incorporated in a single die, shared communication architectures reached out limitations in terms of scalability and performance. However, the NoC-based communication architecture has many configuration parameters leading to a huge design space to be covered during the design phase. Aiming to find optimised configurations, some methods should be employed to explore the design space and speed up this process. This paper investigates three techniques, genetic algorithm, memetic algorithm, and iterated local search. Results have shown the lowest execution time for ILS when compared with the other two. In relation to the quality of the solution, the MME overcame both in almost all scenarios, even considering execution time or number of evaluation as stop criteria.