hello,

Finaly, it's work ! But I had to split the kernel in 2...
You can see a fonctional code for the new drivers here : https://github.com/ByLamacq/BitCrack

Benchmark with drivers 160.32 and Cuda 11.2 on ubuntu :
- RTX 2080 => 800 MKeys /s
- RTX 3090 => 1800 MKeys/s

Can you build and try this ?
Thx