Why can't I use Unified Memory in two CPU thread ?

Hi,

I created two CPU thread in the main function, and new instance in every thread, in the instance , I have used Unified Memory in the instace, when I only created one cpu thread , it can run, but when I created two cpu threads, it would breakdown.
When I did not use Unified Memory, and just only use cudamalloc and cudamemcpy, it can run in two thread.

Hi,

To have a better understanding of your problem, could you share a source to reproduce this issue with us?
Thanks.

Hi,

We have updated the detail here:
https://devtalk.nvidia.com/default/topic/1030988/?offset=10#5251476

Thanks.