Don't have Telegram yet? Try it now!
https://hgpu.org/?p=6636
Optimization of mapped functions sequences using fusions on GPU