Tpetra::Map::locallySameAs slow when running computations on GPU
Created by: ambrad
Tpetra::DistObject::doImport calls Map::operator!= even when ! defined HAVE_TPETRA_DEBUG. That means isSameAs and locallySameAs must be fast.
If pointer equality, contiguity, uniformity, etc. checks fail, then locallySameAs iterates over GIDs on the host. This is very expensive relative to GPU computations.
I don't require an urgent fix for this problem, as I can work around this, but I'm opening this issue because I think the problem needs to be fixed at some point.