View Issue Details
|ID||Project||Category||View Status||Date Submitted||Last Update|
|0001547||OpenFOAM||[All Projects] Bug||public||2015-02-25 22:46||2015-02-25 23:33|
|Priority||low||Severity||text||Reproducibility||have not tried|
|Fixed in Version|
|Summary||0001547: In regards to Issue #1450: "decomposePar with scotch breaks for cyclicACMI patch..."|
|Description||Since the issue was closed before I managed to answer, this bug report acts a bit as an addendum:|
1. There was a typo in the patch that I had attached on that report, in the line of "WarningIn", where it states "cyclicAMIPolyPatch", it should instead be "cyclicACMIPolyPatch" (i.e. missing the "C" in "ACMI"). This was unfortunately also inherited in the commit that was pushed to 2.3.x.
2. Using the same tutorial case "incompressible/pimpleDyMFoam/oscillatingInletACMI2D", with 96 subdomains decomposed with either scotch or hierarchical "(2 48 1)", there were no occurrences of a partial mismatch of sizes. In other words, "tgtMask_" was always either 0 or of equal size.
3. Nonetheless, since this case has a very simple mesh, it will require some deeper investigation into the source code to ascertain whether such a situation of non-zero lists not having the same size is even possible. Either way, the implemented hack does seem to cover most of the potential problems.
|Tags||No tags attached.|