[MPIWG Fortran] Deprecate MPI_SIZEOF in MPI-4.0

[Rolf Rabenseifner]

Jeff and Craig,

I'm also good with this, but technically, in my opinion,
the following changes are needed:

- Page 604, lines 19-20 of
  https://github.com/mpi-forum/mpi-issues/files/221301/mpi-report-deprecate-mpi-sizeof.pdf
  read

    The following Fortran subroutines are deprecated; 
    the Fortran language storage_size() and c_sizeof() 
    intrinsic functions provide this functionality.

  but should read (proposal)

    The following Fortran subroutines are deprecated; 
    the Fortran language storage_size() and c_sizeof() 
    intrinsic functions provide similar functionality.
    Whereas MPI_SIZEOF() and c_sizeof() provide the size
    in bytes, the intrinsic storage_size() provides the
    size in bits.  

  Otherwise it is a direct path to the hell when people
  updating there deprecated source code.


- Page 631, lines 5-8 read

    MPI_TYPE_MATCH_SIZE can be used to obtain a size-specific
    type that matches a Fortran numeric intrinsic type by 
    first calling storage_size()in order to compute the
    variable size, and then calling MPI_TYPE_MATCH_SIZE to find
    a suitable datatype. In C, one can use the C function 
    sizeof(), instead of storage_size().

  It is necessary to mention that storage_size() returns 
  bits and not bytes. It should be written in a way that a 
  Fortran reader Need not to read the C text and vice versa.
  Proposal:

    MPI_TYPE_MATCH_SIZE can be used to obtain a size-specific
    type that matches a Fortran numeric intrinsic type by 
    first calling storage_size() in order to compute the
    variable size in bits, deviding it by 8, 
    and then calling MPI_TYPE_MATCH_SIZE to find
    a suitable datatype. In C, one can use the C function
    sizeof(), which returns the size in bytes, 
    instead of storage_size() (which returns the size in bits).


- On page 18 in Table 2.1 the following mail should be added:

   MPI_SIZEOF  MPI-4.0   ____  storage_size() 5)

  With footnote 
   5) Fortran intrinsic. It returs the size in bits instead of bytes. 
 

- In the General index on page 721, I would recommend to add 
  the keyword "MPI_SIZEOF and storage_size()"
  and all Locations of the both routines.

Best regards
Rolf


----- Original Message -----
> From: "Craig Rasmussen" <rasmus at cas.uoregon.edu>
> To: "MPI-WG Fortran working group" <mpiwg-fortran at lists.mpi-forum.org>
> Sent: Wednesday, April 20, 2016 1:52:18 AM
> Subject: Re: [MPIWG Fortran] Deprecate MPI_SIZEOF in MPI-4.0

> I'm good with this too.
> 
> -craig
> 
> 
> On Fri, Apr 15, 2016 at 11:03 AM, Bill Long < longb at cray.com > wrote:
> 
> 
> 
> On Apr 15, 2016, at 12:41 PM, Jeff Squyres (jsquyres) < jsquyres at cisco.com >
> wrote:
> 
>> Re-sending because the reply-to on the original mail was incorrect (i.e., if you
>> replied and didn't hand-edit the reply address, it likely bounced).
>> 
>> 
>>> On Apr 15, 2016, at 11:57 AM, Jeff Squyres (jsquyres) < jsquyres at cisco.com >
>>> wrote:
>>> 
>>> We brought this up forever ago (December 2014), and never progressed the issue
>>> (IIRC).
>>> 
>>> It came up again today in a different context (in the Open MPI community). So I
>>> took the liberty of re-creating the issue in the Github MPI Forum system:
>>> 
>>> https://github.com/mpi-forum/mpi-issues/issues/51
>>> 
>>> We still want to move forward with this, right?
> 
> I’m in favor.
> 
> Cheers,
> Bill
> 
>>> 
>>> --
>>> Jeff Squyres
>>> jsquyres at cisco.com
>>> For corporate legal information go to:
>>> http://www.cisco.com/web/about/doing_business/legal/cri/
>>> 
>>> _______________________________________________
>>> mpiwg-fortran mailing list
>>> mpiwg-fortran at mpi-forum.org
>>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran
>> 
>> 
>> --
>> Jeff Squyres
>> jsquyres at cisco.com
>> For corporate legal information go to:
>> http://www.cisco.com/web/about/doing_business/legal/cri/
>> 
>> _______________________________________________
>> mpiwg-fortran mailing list
>> mpiwg-fortran at lists.mpi-forum.org
>> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran
> 
> Bill Long longb at cray.com
> Fortran Technical Support & voice: 651-605-9024
> Bioinformatics Software Development fax: 651-605-9142
> Cray Inc./ Cray Plaza, Suite 210/ 380 Jackson St./ St. Paul, MN 55101
> 
> 
> _______________________________________________
> mpiwg-fortran mailing list
> mpiwg-fortran at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran
> 
> 
> _______________________________________________
> mpiwg-fortran mailing list
> mpiwg-fortran at lists.mpi-forum.org
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpiwg-fortran

-- 
Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner at hlrs.de
High Performance Computing Center (HLRS) . phone ++49(0)711/685-65530
University of Stuttgart . . . . . . . . .. fax ++49(0)711 / 685-65832
Head of Dpmt Parallel Computing . . . www.hlrs.de/people/rabenseifner
Nobelstr. 19, D-70550 Stuttgart, Germany . . . . (Office: Room 1.307)