[Dune] Intersection area and corners

Fri May 27 19:13:43 CEST 2016

Hi Martin,

I prepared the smallest case I could do.

It is a bit slow because of the computation of the element parametrizations.

It will print at stdout the corners of the element (index 43) having the sick intersection from the geometry grid and from the host one and the corners of the sick intersection from both grids.

The intersection has index 3 in this element.

As you can see in the geometry grid case corners 0 and 1 of the intersection are not at the same position of corners 6 and 7 of the element.

In the case of the host grid everything works fine.

I would no say the bug is in dune, but I cannot understand why the intersection is not coherent with its element.

It is not evident from this code but the neighbour through this intersection is not coherent too. Its index is 1025.

I remark that using the leaf mapper of the geometry grid crashes the run giving segmantation fault, see line 142 of intersectionBug.cc.
All the test I made are serial.

If you need more feel free to ask.
Thanks for your help,

Marco
________________________________
Da: Marco Cisternino <marco.cisternino at optimad.it>
Inviato: giovedì 26 maggio 2016 14.12
A: Martin Nolte; Dune
Oggetto: Re: [Dune] Intersection area and corners

Hi Martin,

I'll try to set up a small example.

The weird thing is that I have 2k elements in my coarse grid and this problem occurs for only 1 intersection. That's why I would like to see how the intersection is built, in order to understand if my code wrongly changes that intersection object.

Thanks.

Marco

________________________________
Da: Martin Nolte <nolte at mathematik.uni-freiburg.de>
Inviato: mercoledì 25 maggio 2016 17.47
A: Dune
Oggetto: Re: [Dune] Intersection area and corners

Hi Marco,

the intersection of GeometryGrid is based on the intersection of the host
grid. Then, you use the geometryInInside and the element geometry (of the
inside element) to build the global geometry of the intersection. So, in a
conforming situation, the corners of the intersection should be a subset of
the corners of the inside element.

Therefore, I consider your question a bug. Can you provide a small example?

Best,

Martin

On 05/25/2016 03:59 PM, Marco Cisternino wrote:
> Thank you Martin for your quick reply.
>
> No, for the configuration I'm debugging I'm not dealing with non-conforming
> elements. That's why I cannot really understand why intersection and element
> are not coherent.
>
> How the intersection is built? I could look in the constructor to see why its
> corners are not among the element corners.
>
> Thanks again
>
>
> Marco
>
>
>
>
> ------------------------------------------------------------------------------
> *Da:* Martin Nolte <nolte at mathematik.uni-freiburg.de>
> *Inviato:* mercoledì 25 maggio 2016 11.00
> *A:* Dune
> *Oggetto:* Re: [Dune] Intersection area and corners
>
> Hi Marco,
>
> I am not sure whether I get your question correctly. You seem to be comparing
> the geometry of the inside element to the geometry of the intersection.
>
> This is only reasonable, if the intersection is conforming. Otherwise I expect
> the corners to be different. Is it possible, that your intersection is
> non-conforming?
>
> Please note the using GeometryGrid with non-conforming grid views is a bit
> problematic. Basically you are creating a first order Lagrange function over a
> non-conforming grid. If you don't constrain hanging nodes, the position of a
> hanging node might not be on the edge of the bigger element but slightly moved.
>
> You might also want to have a look at the following merge request:
>
> https://gitlab.dune-project.org/core/dune-grid/merge_requests/60
[https://gitlab.dune-project.org/assets/gitlab_logo-cdf021b35c4e6bb149e26460f26fae81e80552bc879179dd80e9e9266b14e894.png]<https://gitlab.dune-project.org/core/dune-grid/merge_requests/60>

Bugfix/geometrygrid various (!60) · Merge Requests · Core Modules / dune-grid<https://gitlab.dune-project.org/core/dune-grid/merge_requests/60>
gitlab.dune-project.org
Grid Interface and Implementations

> <https://gitlab.dune-project.org/core/dune-grid/merge_requests/60>
>
> Bugfix/geometrygrid various (!60) · Merge Requests · Core Modules / dune-grid
> <https://gitlab.dune-project.org/core/dune-grid/merge_requests/60>
> gitlab.dune-project.org
> Grid Interface and Implementations
>
>
>
>
> It deals with normals in non-conforming grids, which might also be buggy in
> DUNE 2.3.
>
> Best,
>
> Martin
>
> On 05/25/2016 10:44 AM, Marco Cisternino wrote:
>> Good morning,
>> I have a problem with one intersection in my dune application.
>> The area I get is not what I expect.
>> So I tried to look inside the Intersection object using gdb
>> It is important to say that my grid is a GeometryGrid with a DiscreteFunction to parametrize the elements.
>>
>> If I look at these two things:
>>
>> - *(in.real.insideGeo_.mapping_)
>> - *(iGeo.realGeometry.mapping_)
>>
>> where the in variable type is
>> GeometryGrid::LeafGridView::IntersectionIterator::Intersection
>> and iGeo is
>> GeometryGrid::LeafGridView::IntersectionIterator::Intersection::Geometry
>>
>> I see different corners stored in. I mean two of the 4 corners given by *(iGeo.realGeometry.mapping_) are different from the two correspondant corners given by *(in.real.insideGeo_.mapping_). All the other corners are the same.
>>
>> If I compute the intersection area using corners given by *(in.real.insideGeo_.mapping_), I get the right area. On the other hand, if I compute the intersection area using the corners given by *(iGeo.realGeometry.mapping_), the area is wrong.
>>
>> I want to specify that the corners given by *(in.real.insideGeo_.mapping_) are the same I get using e.geometry().corners(i) where e is of type
>> GeometryGrid::LeafGridView::template Codim<0>::template Partition<Dune::InteriorBorder_Partition>::Iterator::Entity
>>
>> Could anyone help me please? Am I doing something wrong?
>> Ask for more details if you need.
>>
>> I'm using DUNE 2.3 with ALUGrid 1.52. I know it is old, but at the moment I cannot migrate to the new Dune releases.
>>
>> Thanks a lot!
>>
>> Marco
>>
>> _______________________________________________
>> Dune mailing list
>> Dune at dune-project.org
>> http://lists.dune-project.org/mailman/listinfo/dune
>>
>
> --
> Dr. Martin Nolte <nolte at mathematik.uni-freiburg.de>
>
> Universität Freiburg                                   phone: +49-761-203-5630
> Abteilung für angewandte Mathematik                    fax:   +49-761-203-5632
> Hermann-Herder-Straße 10
> 79104 Freiburg, Germany
>
>

--
Dr. Martin Nolte <nolte at mathematik.uni-freiburg.de>

Universität Freiburg                                   phone: +49-761-203-5630
Abteilung für angewandte Mathematik                    fax:   +49-761-203-5632
Hermann-Herder-Straße 10
79104 Freiburg, Germany

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dune-project.org/pipermail/dune/attachments/20160527/16a838ee/attachment.htm>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: intersectionBug.tar.gz
Type: application/gzip
Size: 1377244 bytes
Desc: intersectionBug.tar.gz
URL: <https://lists.dune-project.org/pipermail/dune/attachments/20160527/16a838ee/attachment.gz>