How to disable memory overcommitment when cuda is enabled?

we usually disable overcommitmend by using prlimit64 and limiting the memory to the available physical memory. This does not work when we compile our program with cuda support enabled. Right after startup it already uses 16GB virt RAM on our Tx2 (with only 8GB RAM in total).
Where is this huge usage coming from? Any ideas how to disable over-commitment?

Sorry for the late response, have you managed to get issue resolved or still need the support? Thanks


Could you share more details about your use case?
A sample to reproduce the issue will help a lot.