|March 22, 2012, 11:41||
conjugateHeat multi-processors: "preserve shadowPatches"
Join Date: Apr 2010
Posts: 1Rep Power: 0
I am working on problems where fluid and solid domains are coupled. The numerical approach that I use is analogous to the one in conjugateHeat as implemented in OF1.5-dev (I didn’t try with OF-1.6-ext yet). I want to run parallel simulations but I noticed that, depending on the way in which I decompose the domains, the solution becomes wrong. Observing various cases and trying different decomposition methods, I concluded that the problem could be related to the fact that the corresponding shadowPatches at the fluid-solid interface are not always located in the same processor.
In the most recent versions of OF I found in the decomposeParDict some additional options like preservePatches (for cyclic BCs) and preserveFaceZones (interface1_faces interface2_faces) ”to keep owner and neighbour on same processor for faces in zones” when using the MultiRegion solver. Maybe for conjugateHeat-like problems it is required to use a similar idea and “preserveShadowPatches”.
I am still investigating the problem, but it would be interesting to know if the problem has been already observed by other users (so far I could not find any useful hint) or if it has a different origin. Thanks.
|conjugateheat, multi-processors, shadowpatches|
|Thread||Thread Starter||Forum||Replies||Last Post|
|64-bit processors for home computing||Ananda Himansu||Main CFD Forum||2||March 16, 2004 13:48|