Don't have Telegram yet? Try it now!
https://hgpu.org/?p=8053
Designing OP2 for GPU architectures