[MPI3 Fortran] MPI Fortran bindings
donev1 at llnl.gov
Fri Jun 5 13:34:57 CDT 2009
On Friday 05 June 2009 11:25, Jeff Squyres wrote:
> I guess I'm a little confused. The intent here is to provide a
> parameter that any valid buffer can be passed through (probably not
> expressions) -- something like C's (void*).
If that is your intent, that is what you should use, not some cryptic
notation <>. An interface is either a valid Fortran interface, meaning
it parses through a compiler and the standard explains what it means
(like a valid C prototype), or it is something else, which to me makes
no sense unless you do what you suggested: You give a valid interface
and then you say "an implementation may use other mechanisms but the
intent/semantics must be obeyed as if this interface were used"...in
which case you better be clear about what the intent/semantics is.
> It is my understanding
> that there is a standard way to do this, but not all compilers
> support it yet. Is that incorrect?
No, there is no officially published way to do this yet. This is why
some of us suggested it is premature to publish the revised interface.
More information about the mpiwg-fortran