In MPI 3.0, is MPI_Get_address subject to the same threading rules as
other MPI calls? In particular, in MPI_THREAD_SERIALIZED mode, am I
required to ensure that MPI_Get_address is not called at the same time as
itself or other MPI code? If it does require synchronization currently,
would it make sense to change that requirement to be looser?
-- Jeremiah Willcock
