Parallelization of user developed solver
Hi,
I have a strange problem. I compiled a transient solver (pimpleKinematicFoam) for particle based erosion from pmpleFoam ( http://www.tfd.chalmers.se/~hani/kur...ing_report.pdf ). It runs fine on a single processor but if i try to put it in parallel it just hangs after a few iterations. Doesn't diverge/ doesn't crash just hangs at "solving 3D cloud KinematicCloud" as shown in below figure.. http://s1.postimg.org/f7x9ldlvv/hang.jpg I also get the duplicate entry error messages as shown below while running the solver. I dont know if this error is related to the problem or not. http://s19.postimg.org/dggciiaan/duplicateentry.jpg Please let me know if anybody has any clue Thank you very much With regards, Krishna |
Hello,
I am currently faced with exactly the same issue, although I can't get past that 'evolving kinematic cloud' on either my own DPMFoam solver OR the unaltered one. Have you found any solution to this? I am modelling a cyclic BC pipe and the only way I can get the code to run in parallel is to split the mesh up in the (r,theta) directions. If I split in the z direction I get this error. Have you tried decomposing the mesh in a different way? |
Hi,
no. strictly speaking i couldnt solve the problem because i never really understood the problem completely.. Need to check the iterations in the for loop step by step and i dont have any means to do it at my end. I went about it by running quite a few iterations in single processor and then decomposing the latest time. |
Hi,
Do you think this is a bug in the solver code or are we just setting up the problem wrong? I don't really have the time to dig around in the code so I think I will just use your solution and decompose the mesh at a later time step. Thanks for the quick reply! |
Now that after you have also faced the same problem in a different situation, i am inclined to think the problem lies with parallel handling of new particles being created.
|
Well I have just tried to run the solver on serial before decomposing and I am faced with the same problem .. I have even taken out the 'particle injection' from the constant/kinematicCloudProperties but it still stops at 'evolving kinematic cloud'
I have just double checked your log file and it turns out my code is freezing before solving any fluid equations where as yours manages a couple of time steps first? |
yes. mine went on for a couple of iterations before freezing.
So you are saying that even though its adding 0 parcels its not proceeding!!! I really thought the problem was with the kinematic cloud!!!! I have no idea whats happening..May be you can keep bringing down the complexity of the case to a point where it runs perfectly and then add on the requirements. |
Quote:
predict erosion rate. But when I tried the solve it does not solve for erosion rate. |
All times are GMT -4. The time now is 09:20. |