Analisis Kinerja Genetic Algorithm yang diakselerasi untuk Travelling Salesman Problem pada Platform Multicore CPU dan CUDA

Alexander Thomas Kurniawan Pratomo (Program Studi Informatika)
Henry Novianus Palit (Program Studi Informatika)
Darian Gunamardi (Program Studi Informatika)

Jurnal Infra
Vol 8, No 1 (2020)

Advancement in technology brings about both new challenges and opportunities. One of which is the opportunity to accelerate an algorithm which usually took a very long time to finish. Genetic algorithm is one such algorithm that can be accelerated using these advancements. Certain ways to accelerate this algorithm is done by tuning the parameters, but these methods are usually unable to retain the quality that is obtained from previous, non-accelerated method. This research applies parallel computing using the multicore technology of CPU and GPU to accelerate genetic algorithm without any changes to the parameters. The purpose of the research is to analyze the differences in performance of the algorithm upon being accelerated with the technologies. The technologies used are CUDA platform and OpenMP API for GPU and CPU respectively. Aside from the technology itself, choosing the algorithm segments on which the implementation is done will also greatly affect the performance of the algorithm. According the testing results, Genetic Algorithm can be accelerated with parallelization using either OpenMP with CPU or CUDA with GPU.