Re: More IO issues
Steve Huss-Lederman (lederman@cs.wisc.edu)
Thu, 17 Apr 1997 12:38:00 -0500 (CDT)
> > But I say this tounge in cheek because John pointed out that MPI IO is
> > supposed to be layered on top of MPI-2, and I still don't see a way to
> > put a function in MPI::Comm if this is supposed to be layered code.
> > :-( So my above point is probably moot.
>
> Careful, here! The MPI Forum has been exceedingly fickle on this point. Most
> recently, for example, we did decide that the MPI-IO datatypes should be full
> citizens i.e. *not* layered. So my guess is that in the true spirit of
> compromise, we'll probably end up with the worst of both worlds; the coupling
> will be just enough to prevent layerable implementations but not strong enough
> to be really useful.
Actually, the datatypes were made first class to make I/O layerable in
a first class way. If the datatype constructors were done in I/O then
the decoding functions that are part of MPI-2 but not in I/O would not
be able to return the calling sequence used - just an equivalent one.
Moving them out of I/O means that they are no different - which they
would be if layered.
Steve