[MPI3 Fortran] Deprecate mpif.h?

Supalov, Alexander alexander.supalov at intel.com
Tue Mar 9 04:26:00 CST 2010

Dear Jeff,

Thanks. I'm very pleased to hear that no-one is going to remove the mpif.h and the mpi module. I knew this upfront. I argue on top that we should not deprecate them, and should instead keep them as current as possible in order to facilitate the transition to the MPI-3. This point of mine seems to escape your attention. If you wish, we can take the rest of this exchange offline as suggested on Friday.

As to the numbers, I'm looking forward to the official analysis of the December data. In the meantime, I feel entitled to interpreting what I have access to, expressing enough reservations on the way to keep this discussion constructive. I would probably do this better if we had the official data published and analyzed by now. Alas, this hasn't happened yet. This is really frustrating.

In any case, I'll be back with "my" ISV feedback numbers as soon as I feel that they are complete. It will be great to have the December data analyzed and published by May, so as to allow an objective comparison of both the data and the conclusions drawn therefrom. FYI, here are the questions I'm asking now c/o our support team to learn what I want from the commercial ISVs we work with:

1.	What Fortran language version is used by your ISV now? What will they use in 5 years from now? Possible answers: Fortran IV, Fortran 66, Fortran 77, Fortran 90, Fortran 95, Fortran 2000, Fortran 2003, Fortran 2008.
2.	What MPI version is used by your ISV now? What will they use in 5 years from now? Possible answers: MPI-1, MPI-1.1, MPI-1.2, MPI-2, MPI-2.1, MPI-2.2, MPI-3.
3.	How does your ISV include the MPI into their code? Possible answers: “include mpif.h”, “use mpi”, they use C for the MPI part.
4.	Will your ISV be ready to recompile/revalidate/recertify the application for “use mpi3” in order to profit from the MPI-3 extensions? Yes/No
5.	[Optional] If you answered “No” to 4, will your ISV want to keep their Fortran usage model but use some MPI-3 extensions? Yes/No
6.	[Optional] If you answered “Yes” to 5., what is the MPI-3 extension are they most interested in: Fault Tolerance, Non-blocking collectives, Fast RMA, Threads as processes, other?

When you answer, please base your reply on the actual language and MPI standard features used by the application, rather than nominal indications (i.e., think whether they use process spawning or one-sided communication or MPI I/O, rather than look into what is said in the used mpif.h about the MPI version).

Best regards.


-----Original Message-----
From: mpi3-fortran-bounces at lists.mpi-forum.org [mailto:mpi3-fortran-bounces at lists.mpi-forum.org] On Behalf Of Jeff Squyres
Sent: Monday, March 08, 2010 6:01 PM
To: MPI-3 Fortran working group
Subject: Re: [MPI3 Fortran] Deprecate mpif.h?

On Mar 5, 2010, at 10:26 PM, Supalov, Alexander wrote:

> In the meantime, the feedback I'm getting across a wide commercial  
> ISV range is that a backward incompatible MPI-3 will be a major  
> issue for them.

I'm curious to know *EXACTLY* what you asked them, and how they  
understood the question.

Analogous example: if you ask someone, "Should I hit John?", they'll  
undoubtedly answer "NO!"  But if you leave out the context that you're  
the dealer in a blackjack game and John only has 2 cards that total a  
face value of 5, then they're not getting all the relevant information  
to answer the question properly.

> Moreover, those who include mpif.h now won't move until they see  
> MPI-3 offering compelling value.

And that's a perfectly reasonable approach.  What's the problem?

> "Survey data points (with the usual disclaimers)
>  Only 16% think recompiling to MPI-3 is bad: Good!
>  Only 11% willing to change source code to MPI-3: Boo!"
> I.e., with all the necessary reservations, we may have about 11% of  
> users who would be willing to add the MPI-3 features to their code.

*** NO!!! ***

This is NOT what the survey says.  That is NOT what the question asked.

The question asked if they wanted to move their application to an  
MPI-3 implementation -- it did NOT ask anything about using new MPI-3  
functionality.  Implying that this question has anything to do with  
people's intentions to using new MPI-3 functionality is just plain  

> Further in the same slide we see (again in Jeff's words):
> "Do we care:
>  Running MPI-3 codes in MPI-1/2 implementations
>        I don’t think so
>  Integrating MPI-3 features in existing MPI-1/2 codes
>        I DO think so
>  Run MPI-1/2 codes without source code changes in MPI-3
>        YES (per survey data)"
> This, again with all the necessary reservations, indicates that  
> users directly ask us not to break the backward compatibility.

Q: Who has said anything about breaking backwards compatibility?
A: Only Alexander Supalov.

No one else intends to break backwards compatibility.  I don't know  
why you keep harping on a point that simply is not true.

This is what is so frustrating in this email thread and really brings  
my blood to a boil.  You are totally taking things out of context and  
leaving out details.

> From this, I derive that the desire to introduce a new, backward  
> incompatible MPI-3 Fortran interface may go against the overwhelming  
> user will for backward compatibility (89% above), as may the desire  
> to deprecate or freeze the conventional mpif.h and the mpi module.

Both of those conclusions are totally unsubstantiated by the survey  

As has been said many times on this thread, your premise of removing  
mpif.h and use mpi is just plain wrong.  They're not going away.   
Period.  Stop saying that they are.

Jeff Squyres
jsquyres at cisco.com
For corporate legal information go to:

mpi3-fortran mailing list
mpi3-fortran at lists.mpi-forum.org
Intel GmbH
Dornacher Strasse 1
85622 Feldkirchen/Muenchen Germany
Sitz der Gesellschaft: Feldkirchen bei Muenchen
Geschaeftsfuehrer: Douglas Lusk, Peter Gleissner, Hannes Schwaderer
Registergericht: Muenchen HRB 47456 Ust.-IdNr.
VAT Registration No.: DE129385895
Citibank Frankfurt (BLZ 502 109 00) 600119052

This e-mail and any attachments may contain confidential material for
the sole use of the intended recipient(s). Any review or distribution
by others is strictly prohibited. If you are not the intended
recipient, please contact the sender and delete all copies.

More information about the mpiwg-fortran mailing list