Don't have Telegram yet? Try it now!
https://hgpu.org/?p=5506
Experiences with Mapping Non-linear Memory Access Patterns into GPUs