https://hgpu.org/?p=2832
Speed and Portability issues for Random Number Generation on Graphical Processing Units with CUDA and other Processing Accelerators