How to use CUDA Memory fences using Numba python library?

Question

I want to synchronize data access for CUDA threads using Global device memory. My current implementation is not making available changes made to variable to other threads in the