Can ParMETIS use the geometric information of a mesh.

dear all,

I'm using ParMETIS to partition FE meshes and I use "ParMETIS_V3_PartMeshKway" routine.
I get very pleasing results with this routine, but not always. So I was wondering if it is possible to use the geometric information (co-ordinates) of the mesh to improve the results.
If someone has an idea on this, please share it.
Also please share, if you know of any other way to improve the mesh partition.

RE: You need to generate the

You need to generate the graph from the mesh, and then use the corresponding graph partitioning routine that uses coordinates to compute the partitioning.

RE: Co-ordinates of Vertices

dear sir,
thanks for your reply. I have one more question.
I have the dual graph of my mesh, but I'm not sure about how to set the argument 'xyz' for 'ParMETIS_V3_PartGeomKway'.
Because I have the co-ordinates of the nodes in my mesh, but 'ParMETIS_V3_PartGeomKway' requires the co-ordinates of the vertices(elements).
So is it safe to use the centroid of the elements?

thanks,
Prashanth Sheshappa

RE: You can just create a

You can just create a coordinate for the vertex (i.e., element) by taking the average of the coordinates of the nodes that make up the element.

RE: Quality of the partion

dear sir,

'ParMETIS V3 PartGeomKway' is giving me a partition which is bad in comparison to 'ParMETIS V3 PartKway'.
I took the average of the co-ordinates of the nodes of an element and populated it into the 'xyz' local array and then called 'ParMETIS V3 PartGeomKway'.
So I would like to know if I'm missing anything?

RE: The use of coordinates will

The use of coordinates will not necessary lead to a better quality partitioning solution. Where the coordinates help, is in reducing the runtime of ParMetis. Having said that, the average quality of PartKway and PartGeomKway should not be all that different. When you said it was "bad", how bad is it?

RE: thanks

thank you very much.

regards,
Prashanth S