View Issue Details
ID | Project | Category | View Status | Date Submitted | Last Update |
---|---|---|---|---|---|
0001494 | OpenFOAM | [All Projects] Bug | public | 2015-01-20 10:29 | 2016-02-03 21:03 |
Reporter | matteoL | Assigned To | henry | ||
Priority | normal | Severity | major | Reproducibility | always |
Status | resolved | Resolution | fixed | ||
Platform | centos | OS | Other | OS Version | (please specify) |
Product Version | |||||
Fixed in Version | |||||
Summary | 0001494: mirrorMesh does not work in parallel | ||||
Description | Hello, it seems mirrorMesh does not work in parallel even if it has the -parallel flag. When I run it in parallel I get: Create time --> FOAM Warning : From function Time::operator++() in file db/Time/Time.C at line 1055 Increased the timePrecision from 7 to 100 to distinguish between timeNames at time 0 --> FOAM Warning : From function Time::operator++() in file db/Time/Time.C at line 1064 Current time name 0 is the old as the previous one 0 This might result in overwriting old results. [node291.cluster.racing.lc:77034] 7 more processes have sent help message help-mpi-btl-sm.txt / knem fail open [node291.cluster.racing.lc:77034] Set MCA parameter "orte_base_help_aggregate" to 0 to see all help / error messages Mirroring points. Old points: 190025 New points: 380050 Mirroring faces. Old faces: 998867--> FOAM Warning : From function mirrorFvMesh::mirrorFvMesh(const IOobject&) in file mirrorFvMesh.C at line 158 Found coupled patch procBoundary0to1 Mirroring faces on coupled patches destroys the ordering. This might be fixed by running a dummy createPatch afterwards. --> FOAM Warning : From function mirrorFvMesh::mirrorFvMesh(const IOobject&) in file mirrorFvMesh.C at line 158 Found coupled patch procBoundary0to8 Mirroring faces on coupled patches destroys the ordering. This might be fixed by running a dummy createPatch afterwards. --> FOAM Warning : From function mirrorFvMesh::mirrorFvMesh(const IOobject&) in file mirrorFvMesh.C at line 158 Found coupled patch procBoundary0to32 Mirroring faces on coupled patches destroys the ordering. This might be fixed by running a dummy createPatch afterwards. --> FOAM Warning : From function mirrorFvMesh::mirrorFvMesh(const IOobject&) in file mirrorFvMesh.C at line 158 Found coupled patch procBoundary0to33 Mirroring faces on coupled patches destroys the ordering. This might be fixed by running a dummy createPatch afterwards. --> FOAM Warning : From function mirrorFvMesh::mirrorFvMesh(const IOobject&) in file mirrorFvMesh.C at line 158 Found coupled patch procBoundary0to40 Mirroring faces on coupled patches destroys the ordering. This might be fixed by running a dummy createPatch afterwards. New faces: 1997734 Mirroring patches. Old patches: 11 New patches: 8 Mirroring cells. Old cells: 407326 New cells: 814652 Mirroring cell shapes. Creating new mesh [3] [5] [5] [5] --> FOAM FATAL ERROR: [5] Unknown polyPatch type processor for patch procBoundary5to4 Valid polyPatch types are : 18 ( cyclic cyclicACMI cyclicAMI cyclicSlip empty genericPatch mappedPatch mappedWall mappedWallVariableThickness nonuniformTransformCyclic oldCyclic patch regionCoupled regionCoupledWall symmetry symmetryPlane wall wedge ) [5] [5] [5] From function polyPatch::New(const word&, const word&, const label, const label, const label, const polyBoundaryMesh&) [5] in file meshes/polyMesh/polyPatches/polyPatch/polyPatchNew.C at line 58. [5] FOAM parallel run exiting [5] [4] [4] [4] --> FOAM FATAL ERROR: [4] Unknown polyPatch type processor for patch procBoundary4to3 | ||||
Steps To Reproduce | Take a decomposed mesh and run mirrormesh in parallel. | ||||
Tags | No tags attached. | ||||
|
This is a known limitation and seems easy enough to fix. The problem is that the patches for the mirrored mesh get created from components instead of being cloned from the old mesh. We will look into this. |
|
mirrorFvMesh.C (12,238 bytes) |
|
elipsekkLOmega.tar.gz (4,698 bytes) |
|
Attached are the following files: - elipsekkLOmega.tar.gz - It's a modified case of the tutorial "incompressible/pimpleFoam/elipsekkLOmega", adapted for testing the mesh generation with mirrorMesh in serial and in parallel. It uses it twice, so it's both the only and the almost perfect test case. - mirrorFvMesh.C - updated file for replacing "applications/utilities/mesh/manipulation/mirrorMesh/mirrorFvMesh.C". This is directly based on Mattijs fix that is present in commit 769abd353ca2e324b513b7c51524d12c481bc1d3 in the OpenFOAM-history repository: https://github.com/OpenCFD/OpenFOAM-history/commit/769abd353ca2e324b513b7c51524d12c481bc1d3 The attached file "mirrorFvMesh.C" is already adapted to OpenFOAM-dev, and possibly can also be used in 3.0.x. The test was done in serial with the attached case in OpenFOAM 3.0.x without the test fix, and then in OpenFOAM-dev with the fix in serial and parallel. The resulting "log.checkMesh" files are nearly identical of 3.0.x serial vs dev parallel. Therefore it does seem that this fixes the reported issue and has been tested with at least this test case. @matteoL: Any chance you still have your case that lead to the initial report, so that you can test the attached fix as well? |
|
Thanks Bruno, I have applied the change to OpenFOAM-dev for evaluation: commit a24eec402c39fbc69dd546fffa4c0d4ea6aff182 |
|
No problems with the patch have been reported so I assume this is now resolved. Thanks for the fix Bruno. |
Date Modified | Username | Field | Change |
---|---|---|---|
2015-01-20 10:29 | matteoL | New Issue | |
2015-03-11 12:13 |
|
Note Added: 0004081 | |
2015-03-24 00:17 | liuhuafei | Issue cloned: 0001622 | |
2015-11-29 21:00 | wyldckat | File Added: mirrorFvMesh.C | |
2015-11-29 21:02 | wyldckat | File Added: elipsekkLOmega.tar.gz | |
2015-11-29 21:08 | wyldckat | Note Added: 0005688 | |
2015-11-29 21:43 | henry | Note Added: 0005690 | |
2016-02-03 21:03 | henry | Note Added: 0005904 | |
2016-02-03 21:03 | henry | Status | new => resolved |
2016-02-03 21:03 | henry | Resolution | open => fixed |
2016-02-03 21:03 | henry | Assigned To | => henry |