On May 6, 2006, at 7:32 PM, Dries Kimpe wrote:
Dries Kimpe wrote:
<quote>
The problem is that the OpenMPI group takes ROMIO as-is from an
older release. Then they pass
MPI_COMBINER_SUBARRAY as-is to ROMIO, which in that release it
does not understand. Then ROMIO blows up.
</quote>
I'll try recompiling Open MPI with a newer version of ROMIO.
Found the problem;
mca/io/romio/romio/adio/common/flatten.c contains conditional code
for
MPI implementations not supporting COMBINER_SUBARRAY, COMBINER_DARRAY
and COMBINER_DUP.
OpenMPI supports these, but fails to #define the necessary symbols:
MPIIMPL_HAVE_MPI_COMBINER_DARRAY
MPIIMPL_HAVE_MPI_COMBINER_SUBARRAY
MPIIMPL_HAVE_MPI_COMBINER_DUP
Because of this, ROMIO excludes support for these datatype
constructors.
Thanks for the bug report -- I was able to duplicate your
experience. I've committed a slightly different patch to our
development trunk. It should be in tomorrow morning's nightly
tarball for any of the active release branches (the trunk, v1.0, and
v1.1):
http://www.open-mpi.org/nightly/
Thanks again,
Brian
--
Brian Barrett
Open MPI developer
http://www.open-mpi.org/
