Exploiting GPU Parallelism to Optimize Real-World Problems
Department of Computer Engineering & Information Technology, University of Ulsan, South Korea
1st National Conference on Intelligent Computing and Information Technology, (NCICIT’13), 2013
@article{furhad2013exploiting,
title={Exploiting GPU Parallelism to Optimize Real-World Problems},
author={Furhad, Md Hasan and Ahmed, Fahmida and Faruque, Md Faisal and Sarker, Md Iqbal Hasan},
year={2013}
}
Construction of optimal schedule for airline crew-scheduling requires high computation time. The main objective to create this optimal schedule is to assign all the crews to available flights in a minimum amount of time. This is a highly constrained optimization problem. In this paper, we implement co-evolutionary genetic algorithm in order to solve this problem. Co-evolutionary genetic algorithms are inherently parallel in nature and they require high computation time. This high computation time can be reduced by exploiting the parallel architecture of graphics processing units (GPU). In this paper, compute unified device architecture (CUDA) provided for NVIDIA GPU is used. Experimental results demonstrate that computation time can significantly be reduced and the algorithm is capable to find some good solutions in a feasible time bound.
February 11, 2014 by hgpu
