[Dune] Parallel alugrid gmsh reading

Tue Jul 7 18:03:54 CEST 2015

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 07/07/15 14:10, Jö Fahlke wrote:
> Am Tue,  7. Jul 2015, 13:42:48 +0100 schrieb Andreas Dedner:
>> Date: Tue, 7 Jul 2015 13:42:48 +0100 From: Andreas Dedner
>> <a.s.dedner at warwick.ac.uk> To: dune at dune-project.org Subject: Re:
>> [Dune] Parallel alugrid gmsh reading
>> 
>> Hi Jake. Please always remember to add which version of the
>> modules (in this case dune-grid and dune-alugrid) you are using
>> because the answer could depend on that. What seems to be working
>> fine is
>> 
>> Dune::GridFactory<GridType> factory; if( myRank == 0 ) 
>> Dune::GmshReader<GridType>::read(factory, gridfile, true, true
>> ); GridType *gridPtr = factory.createGrid(); GridType& grid =
>> *gridPtr ; grid.loadBalance();
>> 
>> A small example code demonstrating the problem would help. 
>> Andreas

Ah - thanks Andreas, that seems to work better. So the GridFactory
needs to persist outside the scope of GmshReader, rather than it just
returning an already created grid...

The grid gets read in now and the loadbalancing routines return no
errors. However, I am now beset by a bunch of these types of errors in
my later code:

PETSc Error in the PETSc function 'VecScatterCreate_PtoS' at
vpscat.c:1763: 'Petsc has generated inconsistent data'. Error message:
'ith 26 block entry 897 not owned by any process, upper bound 872'

I need to do a bit more investigating. If I can I will try and write a
simple test program that recovers the error.

The dune modules I am using are close to the most recent development
versions off git, which I have been pulling every so often to get the
cmake macros.

Thanks to both of you

- -Jake

p.s. Could I be added to the mailing list please? I did not receive
your reply Andreas. I could only read it after reading Jö's since that
message was addressed to me and cc'd to dune at dune-project.org.