From: "Rolf Rabenseifner" <rabenseifner@XXXXXXX>
Date: January 22, 2008 9:56:11 AM CST
To: mpi-21@XXXXXXXXXXXXX
Subject: Re: [mpi-21] Re: MPI-2 MPI_Comm_spawn inconsistency
Reply-To: mpi-21@XXXXXXXXXXXXX

If the MPI-Forum would decide to not return the error codes
on the non-root processes, then this would be a interface change
that may break existing application code (although I expect that most  
are using default abort on errors).
Therefore, it cannot be part of MPI 2.1.
MPI 2.1 makes mainly clarifications and error corrections (typos,
inconsistencies).
MPI-2 page 84 line 45 defines array_of_errcodes without saying
"significant only at root".

Your question may be still senseful!

For me, it implies another question:
Is it allowed that some processes are passing an array_of_errcodes
while some other processes are passing MPI_ERRCODES_IGNORE?

Allowing this, would not break user codes. Therefore, the proposed
advice should be more clear:

Proposal for MPI 2.1, Ballot 4 (2nd version):
---------------------------------------------
Add in MPI-2.0, page 88, after line 24:

Advice to users. If the non-root processes do not use  
MPI_ERRCODES_IGNORE,
then they have to allocate the appropriate number of entries (see  
maxproc
at root) in the array_of_errcodes although the maxproc argument is
unused in non-root processes. It is allowed to use an array_of_errcodes
at some of the calling processes and MPI_ERRCODES_IGNORE at some others.
(End of advice to users.)
-------------------------------

But we should also vote on your "proposal":

Alternative-proposal for MPI 2.1, Ballot 4:
-------------------------------------------
MPI 2.0 page 84 line 45 reads (in MPI_COMM_SPAWN):
     OUT   array_of_errcodes   one code per process (array of integer)
but should read:
     OUT   array_of_errcodes   one code per process (array of integer,
                               significant only at root)

MPI 2.0 page 89 line 42 reads (in MPI_COMM_SPAWN_MULTIPLE):
     OUT   array_of_errcodes   one code per process (array of integer)
but should read:
     OUT   array_of_errcodes   one code per process (array of integer,
                               significant only at root)

Another reason, not to do this modification, is that the non-root
processes have no chance (in error-return-mode) to detect an error.
And after an error, MPI does not guarantee that MPI communication still
works. Only MPI_Abort should be guaranteed to work.

Best regards
Rolf



On Fri, 18 Jan 2008 17:43:47 -0300
  "Lisandro Dalcin" <dalcinl@XXXXXXXXX> wrote:> On 1/18/08, Rolf Rabenseifner <rabenseifner@XXXXXXX> wrote:>> This is a proposal for MPI 2.1, Ballot 4.
>>   Significance of non-root arguments to MPI_COMM_SPAWN
>> Proposal for MPI 2.1, Ballot 4:
>> If the non-root processes do not use MPI_ERRCODES_IGNORE,
>> then they have to allocate the appropriate number of entries
>
> Are you propossing that the root process have to implicitely inform
> all other processes in comm group if an error occured?
>
> Is'nt it better to return error codes only at root, and let the
> programmer take the decision if that information should be broadcasted
> to other procs in group?
>
> Perhaps I missing something, but I would prefer to make
> array_of_errcodes only relevant at root, and filled-in only at root,
> and others procs in comm group just ignore this entry.
>
>
> -- 
> Lisandro Dalc’n
> ---------------
> Centro Internacional de MŽtodos Computacionales en Ingenier’a  
> (CIMEC)
> Instituto de Desarrollo Tecnol—gico para la Industria Qu’mica  
> (INTEC)
> Consejo Nacional de Investigaciones Cient’ficas y TŽcnicas (CONICET)
> PTLC - GŸemes 3450, (3000) Santa Fe, Argentina
> Tel/Fax: +54-(0)342-451.1594
>



Dr. Rolf Rabenseifner . . . . . . . . . .. email rabenseifner@XXXXXXX
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)