what pattern does decomposePar follow to divide surface points among processors

Mohsin1 · March 14, 2024, 02:02

When decomposePar is run, it divides the original mesh into multiple folders named processorN (N=0,1,2,3..)

In original mesh, which is at constant/polyMesh
We can find out the points/nodes ids related to a particular bounary patch by using boundary file and faces.gz file.
But when its divided in many folders, the the node ids in paces.gz start from zero again and numbering is no longer same as the original faces.gz.

So, its not possible to use files processorN/constant/polyMesh/boundary and processorN/constant/polyMesh/faces.gz to find out the absolute node ids related to a particular boundary patch.

So, does any one know of a pattern in which decompose points among multiple processors?

Or is there any other easy way to find out absolute node ids in each processor?

olesen · March 16, 2024, 07:43

The various cell/face/point ProcAddressing files give the correspondence between the serial view of the mesh and the rank-local mesh. So this could serve as one view of the global numbering. Another view could just be to use the rank-local offsetted numbering such as used in globalIndex and polyMesh globalMeshData. It is frequently this view of global numbering that is the more useful version. In fact, you will find a whole series of gather/scatter routines attached to globalIndex for thus reason.

March 14, 2024, 02:02	what pattern does decomposePar follow to divide surface points among processors	#1
Mohsin1 New Member Mohsin Join Date: Jul 2023 Posts: 13 Rep Power: 2	When decomposePar is run, it divides the original mesh into multiple folders named processorN (N=0,1,2,3..) In original mesh, which is at constant/polyMesh We can find out the points/nodes ids related to a particular bounary patch by using boundary file and faces.gz file. But when its divided in many folders, the the node ids in paces.gz start from zero again and numbering is no longer same as the original faces.gz. So, its not possible to use files processorN/constant/polyMesh/boundary and processorN/constant/polyMesh/faces.gz to find out the absolute node ids related to a particular boundary patch. So, does any one know of a pattern in which decompose points among multiple processors? Or is there any other easy way to find out absolute node ids in each processor?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Error in gmshToFoam on ESI v2106	siefer92	OpenFOAM Programming & Development	4	July 17, 2021 05:11
[snappyHexMesh] snappyHexMesh generates not planar surface	krzychu111	OpenFOAM Meshing & Mesh Conversion	2	April 23, 2020 16:38
OpenFOAM error	Vinay Kumar V	Main CFD Forum	0	February 20, 2020 09:17
[ICEM] Surface mesh does not follow node spacing	findtheinvisiblecow	ANSYS Meshing & Geometry	7	August 26, 2014 11:17
CFX4.3 -build analysis form	Chie Min	CFX	5	July 12, 2001 23:19

March 16, 2024, 07:43		#2
olesen Senior Member Mark Olesen Join Date: Mar 2009 Location: https://olesenm.github.io/ Posts: 1,686 Rep Power: 40	The various cell/face/point ProcAddressing files give the correspondence between the serial view of the mesh and the rank-local mesh. So this could serve as one view of the global numbering. Another view could just be to use the rank-local offsetted numbering such as used in globalIndex and polyMesh globalMeshData. It is frequently this view of global numbering that is the more useful version. In fact, you will find a whole series of gather/scatter routines attached to globalIndex for thus reason.