processorAgglomeration broken for processor boundaries
Summary
GAMG processorAgglomeration behaves differently w.r.t v2306. Convergence when embedded in GAMG seems to be worse.
Might originate from
Steps to reproduce
Get e.g. potentialFoam/pitzDaily. Set up processorAgglomerator masterCoarsest and use random decomposition method. Number of GAMG iterations is very different compared to previous versions.
Example case
See above
What is the current bug behaviour?
See above
What is the expected correct behavior?
Repeat of v2306 behaviour
Relevant logs and/or images
Environment information
- OpenFOAM version : v2312 (v2306 and before is ok)
- Operating system :
- Hardware info :
- Compiler :
Possible fixes
Processor 'patches' should be merged into internal faces - so the boundary coefficients become internal coefficients. This seems not to happen.
Edited by Mattijs Janssens