Improving Cache Locality for Ray Casting with CUDA