Peer to peer memory access functions

I am testing peer-to-peer memory access and samples from Nvidia that I have read suppose UVA is available. Before UVA, It was necessary to invoke cudaPeerRegister to access memory from different device that the one that allocated memory. I have been searching for that function but I have not been able to locate it although It is recognizes by nvcc.
Somewhere to look at for the behaviour of that function?