Dear All:
I am trying to run a parallel packing as a simulation starting point just using the examples provided by LIGGGHTS 3.3.1 which lies in:
LIGGGHTS-PUBLIC/examples/LIGGGHTS/Tutorials_public/packing
As I am running on a cluster, I want to utilize all the 16 cores on a node (each node has 16 cores), thus LIGGGHTS 3.3.1 is compiled using Intel Compiler 14.0.2 using OpenMPI 1.6.3 (other versions of MPIs, e.g. mvapich2 are tested as well, with similar results), the only change I made to the in.packing is that I added one line at line 10:
processors 2 2 4
which intended to divide the entire domain into 16 sub-domains, the simulation using mpi runs with no errors however the packing result is quite different from the serial run, it seems there are big gaps at the domain boundary that divided the final particle packing and this will become a huge concern as such big gaps will affect the following simulation, does anyone know if there is any way to eliminate such gap?
The script is attached and the result screenshots are also included,
Any suggestion is very much appreciated, thank you in advance!
Roy Li
| Attachment | Size |
|---|---|
 packing_mp.tar_.gz | 3.68 KB |
 packing_single.png | 278.84 KB |
| packing_2_2_4.png | 267.36 KB |
dbxmcf | Thu, 03/17/2016 - 15:09
sorry this appears to be my
sorry this appears to be my problem as I was looking at the initial but not the final result, please disregard/delete this thread