[Gmsh] a question on partitioning using gmsh

[Umut Tabak]

Dear all,

I am trying to use gmsh for partitioning operations on simple geometries 
for what if tries for my research.

The input model is 'model_for_parts.mesh' and the model that is 
partitioned into 2 subdomains is given in 'simple_domain_part.msh' 
however there are some points that I do not understand on the output 
file, 'simple_domain_part.msh'

Well the elements section of the partitioned file is given below

43
1 1 2 0 -1 36 7
2 1 2 0 -1 18 35
3 1 2 0 -1 35 36
4 1 4 0 0 1 2 1 3
5 1 4 0 0 1 2 3 4
6 1 4 0 0 1 2 4 5
7 1 4 0 0 1 2 5 6
8 1 4 0 0 1 2 6 7
9 1 4 0 0 1 1 7 8
10 1 4 0 0 1 1 8 9
11 1 4 0 0 1 1 9 10
12 1 4 0 0 1 1 10 11
13 1 4 0 0 1 1 11 2
14 3 4 0 0 1 1 12 14 27 26
15 3 4 0 0 1 1 14 15 29 27
16 3 4 0 0 1 1 15 16 31 29
17 3 4 0 0 1 1 16 17 33 31
18 3 5 0 0 2 1 -2 17 18 35 33
19 3 5 0 0 2 2 -1 18 19 37 35
20 3 4 0 0 1 2 19 20 39 37
21 3 4 0 0 1 2 20 21 41 39
22 3 4 0 0 1 2 21 22 43 41
23 3 4 0 0 1 2 22 13 23 43
24 3 4 0 0 1 1 26 27 28 25
25 3 4 0 0 1 1 27 29 30 28
26 3 4 0 0 1 1 29 31 32 30
27 3 4 0 0 1 1 31 33 34 32
28 3 5 0 0 2 1 -2 33 35 36 34
29 3 5 0 0 2 2 -1 35 37 38 36
30 3 4 0 0 1 2 37 39 40 38
31 3 4 0 0 1 2 39 41 42 40
32 3 4 0 0 1 2 41 43 44 42
33 3 4 0 0 1 2 43 23 24 44
34 3 4 0 0 1 1 25 28 11 2
35 3 4 0 0 1 1 28 30 10 11
36 3 4 0 0 1 1 30 32 9 10
37 3 4 0 0 1 1 32 34 8 9
38 3 5 0 0 2 1 -2 34 36 7 8
39 3 5 0 0 2 2 -1 36 38 6 7
40 3 4 0 0 1 2 38 40 5 6
41 3 4 0 0 1 2 40 42 4 5
42 3 4 0 0 1 2 42 44 3 4
43 3 4 0 0 1 2 44 24 1 3

so partitioning operation creates some ghost cells which are not a part 
of the original mesh structure, for instance the first 3 line 
elements(which are really on the interface.) are not a part of the mesh.

However, gmsh also marks some of the quad elements as ghost cells while 
mentioning that they have 5 tags, however the manual only mentions 4 tags.

Is there a systematic way to get rid of the real ghost cells on the 
partitioned mesh structure. As far as I could understand from the 
manual, elements 28-29 (which are on the interface) are also mentioned 
as ghost cells however they are a part of the mesh and belong to 
different partitions, so they should not be marked as ghost, at least 
this is what I think.

I was wondering if I should always check the output mesh data after 
doing the partitioning and reorder that by hand extracting data that I 
want?

I wanted to automate this process for simple geometries however that did 
not seem possible for the moment, could there be a workaround on this 
problem, at least for simple 2d/3d mesh structures?

Best regards,
Umut
-------------- next part --------------
A non-text attachment was scrubbed...
Name: simple_domain_part.msh
Type: model/mesh
Size: 1656 bytes
Desc: not available
URL: <http://www.geuz.org/pipermail/gmsh/attachments/20111121/49f39e1d/attachment.msh>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: model_for_parts.msh
Type: model/mesh
Size: 2331 bytes
Desc: not available
URL: <http://www.geuz.org/pipermail/gmsh/attachments/20111121/49f39e1d/attachment-0001.msh>