Implementation Of The Genetic Algorithm By Means Of CUDA Technology Involved In Travelling Salesman Problem

Abstract

The research was intended to solve the travelling salesman problem by means of genetic algorithms. The implementation of the algorithm was by virtue of CUDA technology. The research was focused on checking how much the system can improve if instead of classical CPU processors one uses GPU graphical processors enabled to perform the operations parallel. The algorithm was implemented in the high level CUDA C language. Thus, measuring the pure time of performance of the algorithm could be the single but reliable point of comparison between two above mentioned types of processors. Making some operations mutually independent and using CUDA technology makes the task much faster to execute. Due to it complex issues can be solved in a shorter time.

Parallel Hierarchical Genetic Algorithm for Scattered Data Fitting through B-Splines

“…In this regard, the implementation of metaheuristics for several problems has been addressed using GPUs and CPUs mainly using CUDA, OpenCL, OpenMP and MPI. Table 1 shows some works which have addressed classic problems such as the traveling salesman [21][22][23] and others more specialized like path planning [24], etc. ; being OpenCL and CUDA the technologies more used.…”