https://hgpu.org/?p=12930
Random Address Permute-Shift Technique for the Shared Memory on GPUs