[Mpi-21] A couple of bugs in 2.1

[Rolf Rabenseifner]

Rajeev,

1) your first item (MPI_REPLACE in MPI_ACCUMULATE)
should you handle purely as an MPI-2.2 item,
because this text is voted twice as Ballot 4, Item 13.
The MPI-2.1 edited text is therefore correct.

It should be a new MPI-2.2 discussion, because
the two atomic operations (replace by B) and (replace by C)
applied to an existing window item, are not commutative.
All other operations are commutative
(with exception with MAXLOC and MINLOC on the "location").

Therefore, the MPI-2.2 Forum must decide whether
current sentence
"MPI_REPLACE, like the other predefined operations, is
defined only for the predefined MPI datatypes."
should read
a.  "MPI_REPLACE is
      defined only for the predefined MPI datatypes."
or
b.  "MPI_REPLACE, like the other predefined operations, is
      defined only for the MPI datatypes defined above."

2) Your errata to MPI_MAX_OBJECT_NAME is correct.
I have set-up the following wiki MPI-2.1 errata page:

https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/mpi22/MPI_MAX_OBJECT_NAME

I linked this page in both MPI-2.1 errate summary lists,
see end of
https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/MpiTwoTwoWikiPage

Best regards
Rolf

On Mon, 21 Jul 2008 21:18:13 -0500
  "Rajeev Thakur" <thakur_at_[hidden]> wrote:
> Sent it to the wrong list earlier... 
> 
> -----Original Message-----
>From: owner-mpi-21_at_[hidden] [mailto:owner-mpi-21_at_[hidden]] 
>On
> Behalf Of Rajeev Thakur
> Sent: Monday, July 21, 2008 9:12 PM
> To: mpi-21_at_[hidden]
> Subject: [mpi-21] A couple of bugs in 2.1
> 
> I found a couple of bugs in MPI 2.1:
> 
> * pg 332, ln 16-21: "MPI_REPLACE, like the other predefined 
>operations, is
> defined only for the predefined MPI datatypes. Rationale: The 
>rationale for
> this is that, for consistency, MPI_REPLACE should have the same 
>limitations
> as the other operations. Extending it to all datatypes doesn't 
>provide any
> real benefit."
> 
> 
> That is not true. Derived datatypes are allowed in MPI_Accumulate 
>for the
> predefined reduction operations as explained in the paragraph just 
>above it.
> 
> 
> * pg 237, line 1 and line 3: MPI_MAX_OBJECT should be 
>MPI_MAX_OBJECT_NAME.
> 
> Accordingly, the Change Log needs to be updated for both these 
>items.
> 
> The question is can these be fixed before the 2nd vote in Sept.? My
> understanding of the voting process is that the reason for having 
>two
> separate votes in two separate meetings is to have a chance to catch 
>such
> bugs. So it should be fixable. (Otherwise, what is the use of the 
>2nd vote?)
> 
> Rajeev
> 
> 
> _______________________________________________
> mpi-21 mailing list
> mpi-21_at_[hidden]
> http://lists.mpi-forum.org/mailman/listinfo.cgi/mpi-21

Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner_at_[hidden]
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: Allmandring 30)