Excellent catch -- thanks!
I made a small number of mods to your fixes and committed it; it will
be included in LAM/MPI 7.1.2.
On Apr 18, 2005, at 12:08 PM, r.johnstone_at_[hidden] wrote:
>
> Hello all,
>
> I've been trying to use MPI_Alltoallw
> from a fortran program, without much
> success. I believe the reason for this
> is likely to be that (for version 7.1.1
> of LAM, at least), the relevant wrapper
> (share/mpi/f77/alltoallw_f.c) actually
> calls MPI_Alltoallv, with the same
> arguments (as Alltoallw).
>
> As a result, the displacement arguments
> are interpreted as MPI datatype extents
> instead of bytes, and data types other
> than the first in send and recvtypes
> will be ignored. I've bodged together
> the attached replacement for the
> wrapper, which seems to work for me
> (but I lack experience with this sort
> of stuff) - have I got this right?
>
> Regards,
>
> Roderick
> --
> Roderick Johnstone
> Computational Science and Engineering
> CCLRC Daresbury Laboratory
> Daresbury, Cheshire, UK
> <alltoallw_f.c.gz>_______________________________________________
> This list is archived at http://www.lam-mpi.org/MailArchives/lam/
--
{+} Jeff Squyres
{+} jsquyres_at_[hidden]
{+} http://www.lam-mpi.org/
|
