But thinking about this today, I have no idea what MPI_REPLACE is
supposed to do in a collective reduction. Specifically -- what value
should end up in the target buffer? It doesn't make sense.
I think that this is a grey area in the MPI standard -- MPI_REPLACE is
an MPI_Op, but it should *only* be used with MPI_ACCUMULATE, and not
with the collective reduction functions.
Any comments on this MPI interpretation? Agree / disagree?
Makes sense to limit its use.. so we should return err_arg if used in a
collective reduce op then.
G.
(It has one interesting use, everybody does an allreduce on their rank,
and it then acts as an atomic election, except its repeatable and
predictable probably)
