After allowing the performance counter, ncu correctly profiles my program.

If you have same problem, follow this page

1. Open the NVIDIA Control Panel in your window 11
2. Check "Enable Developer Settings"
3. Under 'Developer' > 'Manage GPU Performance Counters', select 'Allow access to the GPU performance counter to all users'

Why do I have to set these settings in "Window", even if I profile CUDA programs in Ubuntu-18.04, WSL2?

Following this page, this page says:

Once a Windows NVIDIA GPU driver is installed on the system, CUDA becomes available within WSL 2. The CUDA driver installed on Windows host will be stubbed inside the WSL 2 as libcuda.so, therefore users must not install any NVIDIA GPU Linux driver within WSL 2.

I think this is the reason why I need to check the driver in Window. The point is I was not in the native linux, I was in the linux with the WSL