[Wowbagger]

Hi to all,

I am trying to run a case in parallel. My problem is that I want to refine the mesh with SHM. Depending that I read that the scotch method should be avoided because of possible trouble in that combination.

But the scotch method is that one which is obviously the fastest to get the calculation results.

Is there a way to decompose first for SHM and after that to decompose again with the scotch method?

I tried various attempts like this in my Allrun-script but without success:

#!/bin/sh
cd ${0%/*} || exit 1    # Run from this directory

# Source tutorial run functions
. $WM_PROJECT_DIR/bin/tools/RunFunctions

rm log.*
rm -rf ./processor*

application="$(getApplication)"

runApplication blockMesh

rm ./system/decomposeParDict

cp ./system/_decomposeParDict_SHM ./system/decomposeParDict
runApplication decomposePar -copyZero -force
runParallel snappyHexMesh -overwrite
runApplication foamListTimes -rm
rm ./system/decomposeParDict
mv log.decomposePar log.decomposePar_SHM

runApplication reconstructParMesh -constant

runApplication extrudeMesh
runApplication transformPoints "translate=(0 -0.02 0)"

runApplication renumberMesh

cp ./system/_decomposeParDict_Pimple ./system/decomposeParDict
runApplication decomposePar -force 
rm ./system/decomposeParDict
mv log.decomposePar log.decomposePar_Pimple

runParallel $application

runApplication reconstructPar

Does anybody have experience with that?


Best regards,
Michael

[Wowbagger]

I was able to get results by having a look into the "turbineSiting" tutorial.
Maybe this helps some other users.


Best
Michael