[Dune] unwanted refinement with ALUGRID_CUBE

[Matteo Semplice]

Hi,
  thanks for all your replies. For now, UGgrid with "closure none" in 
the dgf file did the trick and allowed my test to be run succesfully.

Best regards,
      Matteo

On 07/12/2012 16:03, Carsten Gräser wrote:
> Hi,
> be aware that turning of the closures in UGGrid will completely
> diable any automatic regularity control. It does not provide the
> possibility to restrict the hanging nodes to a certian level.
>
> In Subgrid you can specify the maximal level. However
> this is currently (for performance reasons) only vertex
> oriented. I.e. you control the maximal level difference
> of elements sharing a vertex.
>
> Best,
> Carsten
>
> Am 07.12.2012 15:28, schrieb Bernd Flemisch:
>    
>> Hi,
>>
>> you can switch the magic switch in the DGF file by setting the
>> corresponding grid parameter:
>> GridParameter
>> % set closure to none for UGGrid
>> closure none
>> #
>>
>> This allows for an arbitrary number of hanging nodes in UGGrid.
>>
>> Kind regards
>> Bernd
>>
>> On 12/07/2012 03:21 PM, Christian Engwer wrote:
>>      
>>> For UG there is some magic switch I've been using ages ago, which
>>> disables the closure and thus allows for hanging nodes, including
>>> arbitrary hanging nodes.
>>>
>>> An other option is to use the subgrid and mark different cells on
>>> different levels and construct you hanging nodes mesh that way. In
>>> this case you could use subgrid<yaspgrid>
>>>
>>> Christian
>>>
>>> On Fri, Dec 07, 2012 at 02:07:18PM +0000, Dedner, Andreas wrote:
>>>        
>>>> Hi.
>>>>
>>>> I believe there is only a constructor like that for the 2d setting
>>>> and there its on the base class which should
>>>> probably not be used directly - in fact non conformity with more than
>>>> one hanging node per codim 1 entity
>>>> is supported by the 2d alugrid library but not by the bindings at the
>>>> moment.
>>>> In 3d there is no such switch - I would have to look into the code
>>>> because I'm not quite sure what
>>>> the restrictions are. It's definitely one level over codim 1 entities
>>>> but I'm not quite sure what it is
>>>> over codim 2 entities (Matteo's test seem to indicate that its only
>>>> level one as well).
>>>>
>>>> Andreas
>>>> ________________________________________
>>>> From: dune-bounces+a.s.dedner=warwick.ac.uk at dune-project.org
>>>> [dune-bounces+a.s.dedner=warwick.ac.uk at dune-project.org] on behalf of
>>>> Carsten Gräser [graeser at math.fu-berlin.de]
>>>> Sent: 07 December 2012 11:55
>>>> To: dune at dune-project.org
>>>> Subject: Re: [Dune] unwanted refinement with ALUGRID_CUBE
>>>>
>>>> Hi Matteo,
>>>> this is because the level of hanging nodes is restricted and the
>>>> default is one.
>>>>
>>>> There's a constructor where you can specify the number of hanging
>>>> nodes. However, I'm not sure if this is the number per edge, per face,
>>>> or number of nested hanging nodes, i.e. the level.
>>>>
>>>> Best,
>>>> Carsten
>>>>
>>>>
>>>>
>>>> re is a default of
>>>> Am 07.12.2012 11:16, schrieb Matteo Semplice:
>>>>          
>>>>> Dear Dune,
>>>>>       I am trying to create on purpose an irregular quad-based grid
>>>>> for a
>>>>> test.
>>>>> My purpose is to have some elements whose neighbours along an edge have
>>>>> two different sizes (e.g. one of size 1/2 and two of size 1/4) like
>>>>> this
>>>>>
>>>>> |-------|  edge of big quad
>>>>> |-|-|---|  edges of neighbouring quads
>>>>>
>>>>> Is there something in ALUgrid or DUNE that prevents this?
>>>>>
>>>>> Here's my experiment.
>>>>>
>>>>> I start with a grid of a single square (level 0)
>>>>> I refine it once obtaining 4 level 1 quads.
>>>>> I mark the north-east and south-west quads and call adapt and I have a
>>>>> grid with 2 level 1 quads and 8 level 2 quads.
>>>>> I do this once more: mark the north-east and south-west level 2 quads
>>>>> and call adapt.
>>>>> This time the level2 quads that were marked get refined, but also the
>>>>> level 1 quads (not marked!) are refined.
>>>>>
>>>>> I am wondering if ALUGRID or DUNE have some mechanism that
>>>>> automatically
>>>>> causes the refinement of the unmarked level1 quads. If so, I appreciate
>>>>> that this feature is quite understandable (and desirable!) for
>>>>> production code, but is there a way to turn this feature off for a
>>>>> test?
>>>>>            
>
>
>    


-- 
Matteo Semplice			Dip. di Matematica
Phone: 011-6702831		Università degli Studi di Torino
Skype: matteosemplice		Via Carlo Alberto, 10 - Torino