How expensive is cudaIpcOpenMemHandle?

Is cudaIpcOpenMemHandle expensive to call?
Does it copy the data on GPU so that it runs slower on large matrices/data?

Thanks

Just to follow up this question.