[blockMesh] Parallel meshing strategy of a (very) simple case

ThibaultGET · September 7, 2021, 12:54

Hi all !

It's the first time for me on the forum.

I have a quite simple problem to solve.

I would like to set up a simple OpenFOAM case described by a cube. But I need to :

- 1* have a linear mesh refinement in one direction (as the simpleGrading option of blockMesh could do)
- 2* create a huge case with a billion of element, supposed to be ran on supercomputer.

My problem : the usual workflow (blockMesh, decomposePar) could work by using fat node on supercomputer, but can take hours and hours to perform, and sometimes hit the allowed job-time limit of supercomputer.

What would be your strategy if you wanted to perform this "simple" case set-up in parallel ? I though about creating coarse mesh, decompose it and then refine it using refineMesh, but it seems i can't control the mesh grading the way i would like (condition 1*).
(Extra-bonus question : is it mandatory to have a repertory for each MPI-process ?)

Thanks for your great help !

Thibault

cyss38 · September 9, 2021, 06:56

Hi Thibault,

Have you tried the following command?

mpirun -np XXX redistributePar -decompose -parallel

doc : https://www.openfoam.com/documentati...-parallel.html

September 7, 2021, 12:54	Parallel meshing strategy of a (very) simple case	#1
ThibaultGET New Member Thibault XAVIER Join Date: Sep 2021 Posts: 2 Rep Power: 0	Hi all ! It's the first time for me on the forum. I have a quite simple problem to solve. I would like to set up a simple OpenFOAM case described by a cube. But I need to : - 1* have a linear mesh refinement in one direction (as the simpleGrading option of blockMesh could do) - 2* create a huge case with a billion of element, supposed to be ran on supercomputer. My problem : the usual workflow (blockMesh, decomposePar) could work by using fat node on supercomputer, but can take hours and hours to perform, and sometimes hit the allowed job-time limit of supercomputer. What would be your strategy if you wanted to perform this "simple" case set-up in parallel ? I though about creating coarse mesh, decompose it and then refine it using refineMesh, but it seems i can't control the mesh grading the way i would like (condition 1*). (Extra-bonus question : is it mandatory to have a repertory for each MPI-process ?) Thanks for your great help ! Thibault

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[DesignModeler] DesignModeler Scripting: How to get Full Command Access	ANT	ANSYS Meshing & Geometry	53	February 16, 2020 16:13
Problem running movingCylinders case in parallel with foam-extend-3.1	mhkenergy	OpenFOAM Running, Solving & CFD	5	March 3, 2017 06:20
implementation of mapFields into parallel transient case	simpomann	OpenFOAM Pre-Processing	4	August 2, 2016 05:41
simpleFoam parallel	AndrewMortimer	OpenFOAM Running, Solving & CFD	12	August 7, 2015 19:45
Free surface boudary conditions with SOLA-VOF	Fan	Main CFD Forum	10	September 9, 2006 13:24

September 9, 2021, 06:56		#2
cyss38 Member Cyrille Bonamy Join Date: Mar 2015 Location: Grenoble, France Posts: 86 Rep Power: 11	Hi Thibault, Have you tried the following command? mpirun -np XXX redistributePar -decompose -parallel doc : https://www.openfoam.com/documentati...-parallel.html