Solution Level Parallelization of Local Search Metaheuristic Algorithm on GPU?

Local search metaheuristic algorithms are proven & powerful combinatorial optimization strategies to tackle hard problems like traveling salesman problem. These algorithms explore & evaluate neighbors of a single solution. Time Consuming LSM algorithms can be improved by parallelizing exploration & evaluation of neighbors of a solution. GPU architecture is suitable for algorithms of single program multiple data parallelism. Implemented algorithm reduces time consuming memory transfers and improves computational time by efficient use of memory hierarchy.