View Issue Details
ID | Project | Category | View Status | Date Submitted | Last Update |
---|---|---|---|---|---|
0002875 | OpenFOAM | Bug | public | 2018-03-12 13:54 | 2018-03-12 14:01 |
Reporter | Jess | Assigned To | henry | ||
Priority | normal | Severity | major | Reproducibility | have not tried |
Status | closed | Resolution | no change required | ||
Platform | Linux | OS | Ubuntu | OS Version | 16.04 |
Summary | 0002875: Difference parallel and serial running | ||||
Description | Case description: Air blowing into an open space with only a small wall section. Turbulence model used: RNG kEpsilon Problem : When running the exact same case using various methods of decomposition for parallel run on one to 8 processors, the number of iterations before convergence (steady state case) varies greatly (from 245 to 1252), as does one value extracted from each end result (TKE varies from 0.255 to 1.198). Similar results are obtained for OpenFoam 2.4. Laminar case was tried for same geometry and initial and boundary conditions with less extreme differences between various decompositions. (though the differences exist) Issue #0000747 was consulted at https://bugs.openfoam.org/view.php?id=747 describing a similar problem, where it is said to have been resolved, which doesn't seem to be the case? | ||||
Steps To Reproduce | Use uploaded compressed folder containing DecomposePar dictionary for Scotch decomposition into 4 processors. decomposePar mpirun -np 4 simpleFoam -parallel > log & reconstructPar postProcess -func sampleDict Look at value k at position (-0.3 6 0) in the last iteration folder in postprocessing folder. Repeat same actions varying each time the processor decomposition. Results obtained in table below. | ||||
Additional Information | Decompositions tried out Number of iterations before convergence Key value (k at position (-0.3 6 0) Simple: 8 (2 2 2) 785 0.711 Scotch: 8 291 0.354 Scotch: 2 375 0.556 Simple: 8 (1 8 1) 348 0.564 Simple: 2 (1 2 1) 513 1.17 Simple:4 (2 1 2) 266 1.15 Scotch: 4 245 0.255 Simple: 8 (4 1 2) 1252 1.198 No Split up (1 processor) 380 0.57 | ||||
Tags | No tags attached. | ||||
Date Modified | Username | Field | Change |
---|---|---|---|
2018-03-12 13:54 | Jess | New Issue | |
2018-03-12 13:54 | Jess | File Added: simpleductred.zip | |
2018-03-12 14:01 | henry | Assigned To | => henry |
2018-03-12 14:01 | henry | Status | new => closed |
2018-03-12 14:01 | henry | Resolution | open => no change required |
2018-03-12 14:01 | henry | Note Added: 0009414 |