The CUDA documentation for scheduling the launching a host function (cuLaunchHostFunc) says:
Completion of the function does not cause a stream to become active except as described above.
I couldn't quite figure out what's "described above". As far as I understand how streams work - the next consecutive piece of work scheduled on the stream after the host function should begin right after the host function execution concludes, i.e. the stream should "become active". Am I missing something? Perhaps I'm misunderstanding what being "active" means?