[mpich-discuss] MPI_Pack max buffer size

[Jeff Hammond]

No, your count has to be representable as a C int per
http://www.mcs.anl.gov/research/projects/mpi/www/www3/MPI_Pack.html
and the MPI standard.

The issue with 32-bit integers and counts >2^31 is well known to the
MPI Forum but there is not a consensus on how to resolve it in all
cases.

The good news is that splitting your >2 GiB messages will have no
measurable overhead.

Unlike other MPI functions, MPI_PACK/UNPACK doesn't have a datatype
workaround.  I will bring it up with the MPI Forum to see if there's
not something to be done abou this, or at least make sure that people
know that it is an issue for >0 users.

Jeff

On Wed, Oct 3, 2012 at 3:10 AM, Christoph Sprenger
<csprenger at wetafx.co.nz> wrote:
> Hi,
>
> just wanted to double check if there is a way to Pack/Unpack bigger char
> buffers than numeric_limits<int>::max() without tokenizing the message into
> multiple chunks ?
> seems like one of my messages is hitting these bounds unfortunately.
>
> Cheers,
> Christoph
>
>
>
> _______________________________________________
> mpich-discuss mailing list     mpich-discuss at mcs.anl.gov
> To manage subscription options or unsubscribe:
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss



-- 
Jeff Hammond
Argonne Leadership Computing Facility
University of Chicago Computation Institute
jhammond at alcf.anl.gov / (630) 252-5381
http://www.linkedin.com/in/jeffhammond
https://wiki.alcf.anl.gov/parts/index.php/User:Jhammond