[Dune] globalRefine and ALUGRID_SIMPLEX

[Carsten Gräser]

Am 08.03.2010 15:29, schrieb Martin Nolte:
> There are 3 different ways of subdividing a tetrahedron into 8 smaller
> tetrahedra. Depending on the chosen variant, this leads to different
> centers. As far as I know ALUGrid, chosen variant depends on the order
> of the vertices.
> It may be a clever idea to sort them such that version produce the same
> output. This might need some complex change in the DGFGridFactory of
> ALUGrid, though.
IMHO the grid factory should only sort such that it is guaranteed
that the reference simplex itself (with correct vertex order) will
be devided into rectangular simplices.

I never used the interval feature of dgf. However the described behaviour
is what I always assumed: If the dgf reader itself creates a rectangular
tet it should create it with the 'correct' vertex ordering. However
this requires that the gridfactory provides the above guarantee.

Regards
Carsten

> 
> By the way: The number of tetrahedra is the same for both DGF files,
> since 6*8 = 8*6 (i.e. make6 + red refinement = red refinement + make6).
> 
> Martin
> 
> Andreas Dedner wrote:
>> Hi.
>> This would really be a surprising error in the refinement strategy of
>> ALU - which I have difficulties to believe.
>> Again your tow test cases cannot lead to the same grid -
>> are we agreed on that?
>>
>> Perhaps you could reduce the test case by starting with one tetrahedron
>> and calling globalRefine - since you have analyzed the problem so far
>> it should be easy enough to write a dgf file which only contains
>> one of the six tetrahedron ALU starts with and then we have only
>> 8 tetrahedrons after refinement. This grid should be easy to
>> visualize.
>>
>> Andreas
>>
>> Dragan Vidovic wrote:>
>>> This is difficult to see because it
>>> looks all right from the outside, but four of the 48 tetrahedra in my
>>> final grid have cell centres at z=0.5, which means that they are not
>>> similar to the big tetrahedra that they were made from.
>> Could you explain this argument at bit more?
>>
>>
>> _______________________________________________
>> Dune mailing list
>> Dune at dune-project.org
>> http://lists.dune-project.org/mailman/listinfo/dune
> 


-- 
----------------------------------------------------------------------
Carsten Gräser           | phone: +49-30 / 838-75349
Freie Universität Berlin | fax  : +49-30 / 838-54977
Institut für Mathematik  | email: graeser at math.fu-berlin.de
Arnimallee 6             |
14195 Berlin, Germany    | URL  : http://page.mi.fu-berlin.de/graeser
----------------------------------------------------------------------