Exploiting GPU Parallelism to Optimize Real-World Problems