# topological question

Hi everyone.

Ok so here's my situation : I have a cylinder to partition and above partitioning the 3D interior domain (tetrahedrons) which works very well, I want to do another partitioning on the cylinder's surface (triangles). This 2D mesh goes all around the cylinder so it's not just like a rectangular 2D domain.

The question is : Does partitioning the cylinder's surface imply some topological problem to METIS ?

Thanks for any answer !

Submitted by Eric26 on Thu, 2008-10-16 12:07

»

- Login to post comments

## RE: If I got you right you just

If I got you right you just wanna perform a domain decompostion for an triangular mesh, don't you? If thus, it's not a problem at all. METIS don't need any geometric informations. It's a graph based tool. Thus just define your triangle as "node" of one graph and the neighbor relations as egdes. For this problem you can abdicate ege and node weights :D

## RE: Yes that I know and on

Yes that I know and on pretty much any 2D domain, the mesh would be converted into a graph then partitionned. But here what I fear is that there would be some problem doing the partitioning because the graph is not "bounded" or "finite" in the circumferential direction. It's not like a plane or curved rectangular domain with defined edges (forgive my english I can't find better words). I'm only speculating anyway, it is intuitive so that's why I ask if there is some topological problem.

Thanks for your time !

## RE: Metis should be able to

Metis should be able to partition the graph corresponding to the cylinder without any problems.