I tried running the simulation at 64, 128 and 256 cores, with the same in file and the executable. The simulation works fine for 64 cores but when we run it on 128 and 256, the particles seem to fly out or seem to be get deleted somehow.
I hope someone could give me an insight about this.
I have attached the in file, the only difference between the 3 being the number of processors.
64 cores -processors 16 2 2
128 cores -processors 32 2 2
256 cores -processors 64 2 2
Also attached is the final image of the simulation.
| Attachment | Size |
|---|---|
 infinalcopy.txt | 36.6 KB |
 64_cores_output.png | 48.82 KB |
| 128_cores_output.png | 34.35 KB |
| 256_cores_ouput.png | 34.11 KB |
j-kerbl | Tue, 06/06/2017 - 14:29
Hi,
Hi,
this seems odd, but maybe the decomposition is at its limit in x direction. I do have some questions: Have you also tried different decompositions? Did you get any warnings in the log-file?
You can have a look at the decomposition with the dump decomposition/vtk command ( http://www.cfdem.com/media/DEM/docu/dump.html )
How many particles are in the simulation?
Another thing: It is possible to disturb insertions with distributing the insertion zone too much in parallel. Is the insertion region slightly smaller than the mesh? Maybe the "all_in yes" keyword does the trick for you.
Cheers,
Josef
cs222 | Thu, 06/08/2017 - 22:30
Thank you Josef
Hey Josef,
Thank you for your reply. I carried out a few decomposition studies and the simulations worked.!
I was also looking into the all_in yes command but rather than having an insertion face I have a region from where the particles are continuously added thus, according to the documentation it may not be valid.
Regards,
Chai