[Saleh Abuhanieh]

Hi Foamers,


I am working on a simple two solid regions heat conduction case. I got the correct result for single processor (serial).


When I try to run the same case in parallel, I got a segmentation fault error. The error is indicating the library (.so) file for the BC which I have created to coupled the two region.


I did the following steps:
- Started from the single processor case
- Added the decomposeParDict for the main system folder and for the system folder of each region
- I selected the simple method for the main decomposeParDict file
- I ran my solver in parallel


Some threads recommend to use the preserveFaceZones, it did not work since the problem not inside the solid1 or solid2 region, but between them, and this patch/bounday between the two regions cannot be included under one processor (as far I know) since the decomposition works for each region independently.


I am sure many Foamers have faced such problem, I use OF5.0
Do I need to consider any additional functionally in my BC for parallel running?


Thanks to share your valuable suggestions

[pete20r2]

Hi Saleh,
Try the scotch method only at the top level and no region specific decomposition.
Also try preservePatches instead of faceZones and be sure to add both sides of the boundary.

[Saleh Abuhanieh]

Hi Peter,


Thanks for your reply!


I've a concern here, in both cases (using preservePatches or faceZones) the decomposition on the main case will result no mesh per processor per region (ex processor0/constant/solid1/polymesh). That means the parallel run can't be executed:
mpirun -np xx solverName -parallel


I tried that, the solver asks for the polymesh files.



Regards,
Saleh

[Saleh Abuhanieh]

any other ideas?

[ViktorKL]

Have a look on this post to apply manual decomposition.
The method presented here is already more advanced and works with a python script.
Regards,
Viktor