Hi everyone,
When I run the attached case (5200 CFD cells and 90000 DEM particles) in 12 processors for a day (around 1 second in simulation), it always exceeds the memory limitation, which is 45 G and should be sufficiently large.
I tried to change the neighbor and neigh_modify settings but it didn't work. I cannot figure out why the case consumes such a large memory.
Please could anyone help to identify the reason or let me know if I made any mistakes.
Liggghts files are attached.
Thanks!
Jason
| Attachment | Size |
|---|---|
 log.txt | 20.62 KB |
| in_liggghts_initial.txt | 3.46 KB |
| in_liggghts_run.txt | 2.19 KB |
alice | Wed, 12/07/2016 - 13:00
Hi Jason,
Hi Jason,
it is a little hard to tell, but there are some things that come to my mind: first of all, the init script seems to use 24 processors and 2 periodic boundaries, while the second script uses 12 processors and only one periodic boundary. Imho the number of particles involved is not ideal for 12, let alone 24, processors. 4 or maybe 8 should be sufficient. While the different amount of processors should not be an issue, the periodic/non-periodic boundaries might be. Other factors that might have a bad influence on the calculation could be possibly large stl-files. Furthermore it seems that there is quite a high number of iterations necessary for solving the p-equation, which could point to numerical problems or problematic settings in fvSchemes or fvSolution...
Hope any of this helps!
Best regards,
Alice
llib11ik | Mon, 12/12/2016 - 05:15
Hi Alice,
Hi Alice,
Thanks for your reply and helping me to find so many possible problems!
When I tried to decrease the number of processors to 4, it seemed getting better. But when the simulation ran around 6 seconds (5 times longer than before), same error happened again.
I can restart it and continue running. But I'd like to find out the reason of error, if possible.
For periodic boundary, how does periodic or non-periodic affect the memory usage?
For stl-files, does the "large" mean large area of the wall, or should I improve the stl settings?
For p-equation, I will the try to improve the grids and other related settings.
Thanks again!
Jason