https://hgpu.org/?p=6625
Leveraging Parallelism with CUDA and OpenCL