Don't have Telegram yet? Try it now!
https://hgpu.org/?p=15040
Optimizing CUDA Shared Memory Usage