Enes -- I'm pretty much the worst person to ask about CUDA stuff, but
wanted to point to this issue:
https://github.com/dealii/dealii/issues/18210
I assume this is not the same as what you have, given that your program
crashes at the *end*, not the beginning of the program?
If not, do you think you could obtain a backtrace for the segfault?
Best
W.