[snappyHexMesh] SnappyHexMesh in Parallel

sjrees · April 7, 2010, 05:05

I don't recall if I got this type of error exactly but I concluded I needed to do the reconstructParMesh step and then deconstructPar again in order to get all the geometry and patch information straightened out before running in parallel (simpleFoam and pisoFoam). I did this on an 8 processor machine with 16Gb RAM and didn't hit a memory problem when dealing with a mesh of about 9m cells. I did take a quick look at memory usage and it was quite high but not a real problem so I did not figure out my mesh size limit.

bjr · April 7, 2010, 12:46

Isn't that RAM spread out over all processors and because reconstructParMesh doesn't run in parallel (to my knowledge, not in front of one of my OpenFOAM installed machines right now), doesn't that mean you actually recomposed your mesh on some small fraction of that 16GB? I was able to do 8-million on a 4-core 8GB shared memory machine using this workflow we've discussed.

bjr · April 7, 2010, 12:48

reconstructParMesh [-noZero] [-region region name] [-fullMatch] [-mergeTol relative merge distance] [-case dir] [-constant] [-latestTime] [-time time] [-help] [-doc] [-srcDoc]

The lack of -parallel is all I know to look for easily.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
[snappyHexMesh] snappyHexMesh in parallel	shailesh.nitk	OpenFOAM Meshing & Mesh Conversion	33	January 25, 2022 11:35
Error running openfoam in parallel	fede32	OpenFOAM Programming & Development	5	October 4, 2018 17:38
[snappyHexMesh] SnappyHexMesh in parallel missing 0 folder	libindaniel2000	OpenFOAM Meshing & Mesh Conversion	0	May 26, 2016 23:46
[snappyHexMesh] SnappyHexMesh in Parallel problem	swifty	OpenFOAM Meshing & Mesh Conversion	10	November 6, 2015 05:40
Explicitly filtered LES	saeedi	Main CFD Forum	16	October 14, 2015 12:58

April 7, 2010, 05:05		#21
sjrees New Member Simon Rees Join Date: Mar 2009 Posts: 12 Rep Power: 17	I don't recall if I got this type of error exactly but I concluded I needed to do the reconstructParMesh step and then deconstructPar again in order to get all the geometry and patch information straightened out before running in parallel (simpleFoam and pisoFoam). I did this on an 8 processor machine with 16Gb RAM and didn't hit a memory problem when dealing with a mesh of about 9m cells. I did take a quick look at memory usage and it was quite high but not a real problem so I did not figure out my mesh size limit.

April 7, 2010, 12:46		#22
bjr Member Ben Racine Join Date: Mar 2009 Location: Seattle, WA, USA Posts: 62 Rep Power: 17	Isn't that RAM spread out over all processors and because reconstructParMesh doesn't run in parallel (to my knowledge, not in front of one of my OpenFOAM installed machines right now), doesn't that mean you actually recomposed your mesh on some small fraction of that 16GB? I was able to do 8-million on a 4-core 8GB shared memory machine using this workflow we've discussed.

April 7, 2010, 12:48		#23
bjr Member Ben Racine Join Date: Mar 2009 Location: Seattle, WA, USA Posts: 62 Rep Power: 17	reconstructParMesh [-noZero] [-region region name] [-fullMatch] [-mergeTol relative merge distance] [-case dir] [-constant] [-latestTime] [-time time] [-help] [-doc] [-srcDoc] The lack of -parallel is all I know to look for easily.