Sorry about the delay replying.
Could you provide some more information?
What specifically "doesn't work in parallel"?
Have you tried this with 0.7.0.3 or the svn head?
If it breaks with both of those, could you provide us with a test case
that replicates the problem?
On Tue, 23 Nov 2010, MARRO MASSIMO wrote:
> Dear Mr Stogner,
> I'm writing to you because I have a problem in reading/writing xda files
> Now I use the Libmesh_0.6.4. If I compile it with --disable-libHilbert, the
> code doesn't work in parallel; but if Libmesh is compiled with
> --enable-libHilbert I can't correctly read/write the solutions saved in xda
> files. I have three variables (u,v,p). I'm able to read/write the first (u)
> in a correct way, but v and p are read/written in a wrong way.
> In the following there is our last talk.
> On Wed, 2 Dec 2009 14:04:30 -0600 (CST)
> Roy Stogner <roystgnr@...> wrote:
>> On Mon, 16 Nov 2009, MARRO MASSIMO wrote:
>>> we commented out the calls to "globally_renumber_nodes_and_elements"
>>> in equation_systems_io.C, but never is changed: the first variable u was
>>> correctly written/read whereas the variable v and p were wrong.
>>> We also configured Libmesh with --disable-libHilbert and, in this case,
>>> the three variables u,v,p were written/read in wrong way.
>>> Finally, we have compiled Libmesh with --disable-libHilbert and without
>>> the calls to "globally_renumber_nodes_and_elements".
>>> The three variables u,v,p were written/read in wrong way again.
>> Sorry to vanish on you for so long; I wasn't able to find a fix two
>> weeks ago and I was off on vacation last week.
>> But I'm starting to understand what's going on, either with
>> --disable-libHilbert and/or with the equation_systems_io.C renumbering
>> commented out.
>> I'm still trying to work on a proper fix, but until then try this
>> workaround: build the libMesh SVN head with --disable-libHilbert, but
>> instead of mesh.read(inmeshfile); use mesh.read(inmeshfile, NULL,
>> true); That will force libMesh to skip the node/element renumbering
>> which seems to be the only thing breaking --disable-libHilbert
>> Let me know if that works for you or not; if I'm not yet on the right
>> track it would be good to know ASAP.