Yeah, I thought so. I believe I've seen a few of your threads on the
libMesh list.

I can tell you that I've personally run into that assertion on my own
larger problems as well which has prevented me from running my largest
simulations with DistributedMesh myself. I will say that contact definitely
does not work with DistributedMesh right now. The issue is that we haven't
designed a "physics based geometric ghosting system" yet to tell libMesh
about the extra elements that needs to keep around. libMesh obtained
support for arbitrary geometry ghosting just in this last year so it is now
possible to tell libMesh about extra elements to keep around on each
partition. In MOOSE it's currently possible to do simple things like keep
an extra halo layer of elements around or extra "point" neighbors, but it
doesn't keep an entire interface of elements around necessary for a contact
problem. Integrating that into MOOSE is also no easy feat as the order in
which MOOSE builds up the simulation is backwards from what libMesh now
requires to correctly set up extra ghosted elements for DistributedMesh.

I'm not sure what to tell you here. We are working closely with libMesh
development so that this capability will exist in the future but it may be
another year or more before it makes it's way into the framework based on
current planned development. We'd like to get everything else working first
(i.e. all of moose/test and the modules with the exception of contact and
XFEM running reliably first). After that, we plan to support pre-splitting
meshes and on-the-fly splitting of meshes. Contact and XFEM will probably
be the last things we address since they are the most difficult.

You may not be dead in the water here. I'm CC'ing Roy (our libMesh and
DistributedMesh expert) in to see if he can brainstorm any hacks that might
be able to get you up and running. I make no guarantees nor can he on these
types of problems but perhaps with a little ingenuity and some extra
perseverance we can guide you towards a solution to get the  master and
slave contact surfaces and associated elements "mirrored" everywhere as a
stop gap to your problem. I think it's doable.

Cody

On Thu, Mar 9, 2017 at 8:09 AM Maria Giuseppina Chiara Nestola <
[email protected]> wrote:

> Hi,
> actually this was just a test and i really need to use distributed mesh
> for my problem.
> Moreover  I am using the last moose version.
>
> Best
>
> Maria
>
> On 09 Mar 2017, at 16:06, Cody Permann <[email protected]> wrote:
>
> Hi Maria,
>
> We are slowly working our way towards supporting distributed mesh in the
> framework at a more "first class" level. There's a lot of active
> development in this area right now to patch up problems. Many things work,
> but there are still many more that do not work with distributed mesh at
> this time. Are you completely up to date with MOOSE or are you running a
> slightly out of date version? Are you planning to run a larger problem? 26K
> elements is really not big enough to justify the extra complication of
> distributed mesh. There are many factors that go into deciding when you
> should switch over from Replicated to Distributed mesh but unless you have
> on the order of a million elements, or hundreds of variables it probably
> won't buy you much.
>
> What kinds of problem are you running? Do you have a repository that we
> can look at?
> Cody
>
> On Thu, Mar 9, 2017 at 7:56 AM Maria Giuseppina Chiara Nestola <
> [email protected]> wrote:
>
> Dear all,
>
> i am trying to run a simulation in the MOOSE framework which is based on
> Libmesh.
> I am trying to use distribute mesh, but when i run in parallel i in debug
> mode i get the following assertion:
>
> Title:  cubit(/Users/studi/src/cgp/meshes/signorini/hemisphere_5191.e):
> 03/07/2017: 09:2
> Mesh Dimension:         3
> Number of Nodes:        5191
> Number of elements:     26888
> Number of elt blocks:   1
> Number of node sets:    0
> Number of side sets:    2
> Nodal data retrieved successfully.
> Nodal numbering map retrieved successfully.
> [0] node_num_map[i] = 56906, 57653, 57655, 57656, 56947, 57654, 57657,
> 56949, 57658, 57614, ... 62096
> All block IDs retrieved successfully.
> All block names retrieved successfully.
> Element numbering map retrieved successfully.
> [0] elem_num_map[i] = 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, ... 26888
> Reading a block of 26888 TETRA(s) having 4 nodes per element.
> Info retrieved successfully for block: 0.
> Connectivity retrieved successfully for block: 0.
> All sideset information retrieved successfully.
> All side set names retrieved successfully.
> Parameters retrieved successfully for sideset: 0.
> Data retrieved successfully for sideset: 0.
> Parameters retrieved successfully for sideset: 1.
> Data retrieved successfully for sideset: 1.
> All node set names retrieved successfully.
> Exodus file closed successfully.
> Assertion `_elements[i]' failed.
>
>
> Assertion `_elements[i]' failed.
>
>
> [3] [1] ../src/mesh/distributed_mesh.C, line 365, compiled
> ../src/mesh/distributed_mesh.C, line 365, compiled Mar  8 2017 at 11:15:31
> Mar  8 2017 at 11:15:31
> ------ BC CONDITIONS  ------
>
>
> BC_id:  1   var_ids:  0  1  2  3  application called
> MPI_Abort(MPI_COMM_WORLD, 1) - process 1
> application called MPI_Abort(MPI_COMM_WORLD, 1) - process 3
>
> Do you know what is due to?
>
> Best
>
>
> Maria
>
> ------------------------------------------------------------------------------
> Announcing the Oxford Dictionaries API! The API offers world-renowned
> dictionary content that is easy and intuitive to access. Sign up for an
> account today to start using our lexical data to power your apps and
> projects. Get started today and enter our developer competition.
> http://sdm.link/oxford
> _______________________________________________
> Libmesh-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/libmesh-users
>
>
>
------------------------------------------------------------------------------
Announcing the Oxford Dictionaries API! The API offers world-renowned
dictionary content that is easy and intuitive to access. Sign up for an
account today to start using our lexical data to power your apps and
projects. Get started today and enter our developer competition.
http://sdm.link/oxford
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to