A Parallel Iterated Local Search Algorithm on GPUs for Quadratic Assignment Problem
MetadataShow full item record
In this study, quadratic assignment problem, which is a hard combinatorial optimization problem, is examined to solve by a new approach. To reach the optimal results by using mathematical programming approaches cannot be possible even for some sorts of small and middle scaled problems in a reasonable time interval. Huge amounts of data are being progressed simultaneously by graphics processing units located on computers’ graphics card. Therefore, a parallel iterated local search algorithm has been proposed to solve the quadratic assignment problem by using graphics processing units’ simultaneously progressing property. This parallel algorithm and the sequential one on central processing units are tested and compared for test problems in literature. Indeed, it is observed that the parallel algorithm works averagely 6.31 times faster for Skorin problems and 11.93 times faster for Taillard problems faster than sequentially one.