[Mpi-forum] Discussion points from the MPI-<next> discussion today
nmm1 at cam.ac.uk
Fri Sep 21 13:30:53 CDT 2012
On Sep 21 2012, Anthony Skjellum wrote:
> What has been elusive, is concrete agreement beyond the 8-16 range of
> basic MPI functions, are there really subsets of functions that can be
> agreed on to form 'compliance points'... those could come with or without
> restricted semantics (e.g., no tags in point-to-point).
It's a good question, and I don't know. Only last week, someone persuaded
me that I need to add a topology lecture to my course. Tags are more
useful than data conversion, which is something that I have never seen
used, though I would be surprised if nobody used. There are quite a lot
of similar features.
> Another point was made that if you use a layered library, then you don't
> know what MPI functions it has used, and that subset could be quite
> different from what the main application uses.
That is a soluble problem, but experience is that you have to require
all components to declare their subset (and not understate it) if you
want that to work. A mandatory argument to MPI_Init would do, so a
subsetted MPI could not be fully compatible.
More information about the mpi-forum