Re: [h5md-user] Strings again / parallel issue

[Peter Colberg]

On Tue, Feb 04, 2014 at 02:16:37PM -0500, Peter Colberg wrote:
> On Tue, Feb 04, 2014 at 07:42:19PM +0100, Pierre de Buyl wrote:
> > Check the other file, parallel_string_attribute.f90. The call to h5awrite_f 
> > is
> > done on every rank. And it writes the attribute correctly. The program
> > parallel_vl_string_attribute.f90 does not execute properly.
> 
> You are right; HDF5 mandates calling H5Awrite from all processes:
> 
> http://www.hdfgroup.org/HDF5/doc/RM/CollectiveCalls.html
> 
> “Because raw data for an attribute is cached locally, all processes
> must participate in order to guarantee that future H5Aread calls
> return correct results on all processes.”

Good news: collective writing and reading of attributes of
variable-length string type is working fine here.

I attached a test case. Unfortunately, this is not very helpful to
you at the moment, since MPI for Lua is not published yet :-(. The
test passes on Debian jessie using HDF5 1.8.12, and Open MPI 1.6.5
or MPICH 3.0.4, spawned with "mpirun -np 8" onto a single machine.

I will try Pierre's Fortran example next.

Peter

mpio.lua
Description: Text document