[MPI3 Fortran] Deprecate mpif.h?

N.M. Maclaren nmm1 at cam.ac.uk
Fri Mar 5 06:01:28 CST 2010


I don't want to follow this further, but shall just try to correct
one misapprehension.

On Mar 5 2010, Supalov, Alexander wrote:
>
> I'm afraid you miss my point. As soon as I just say "use mpi3" in an 
> application that used "include mpif.h" or "use mpi" before, I have to 
> recheck, possibly rewrite, rebuild, revalidate, and recertify my 
> application on all OS, platforms, compilers even if I don't change 
> anything else, including the compiler. This is a huge hit that seems to 
> be ignored on this thread.
>
> I understand this. However, even if I change a tiny part of a mission 
> critical application, full revalidation/recertification may be necessary, 
> which takes months of calendar time and huge resources.

Fine.  But, in that case, won't you also have to revalidate and
recertify if you change to using non-blocking collectives, anyway?

I should SERIOUSLY hope that you are, as adding more asynchronicity is
one of the most common causes of introducing errors into 'working and
tested' programs.  If any 'safety critical' project does NOT require
that when changing from synchronous to asynchronous primitives, even in
minor ways, the people who specify the revalidation and recertification
requirements should be shot.

As far as I know, there is no intention to WITHDRAW the old interfaces,
so you can carry on as at present if you don't make any major changes.

But, if you want to use major new features (and non-blocking collectives
ARE a major new feature), then you will have to bite the bullet and put
of with the cost of changing to use the new module, revalidation and
recertification.  Changing the module is simple and reliable, and the
latter are forced upon you as soon as you use the new features, anyway.

Regards,
Nick Maclaren.





More information about the mpiwg-fortran mailing list