Hi,
In order to reduce the domain box size as much as it possible, I use the command region with union style, so the box domain is an union of many smaller boxes that does a better fit of the simulation region.
Is this useful to improve the simulation time?
I used:
region reg1 block -0.4 4.35 -0.65 0.65 -0.1 0.21 units box
region reg2 block 0 0.2 -0.65 0.65 0.21 0.5 units box
region reg3 block 4.35 5.2 -0.65 0.65 -0.7 0.21 units box
region reg4 block 4.02 4.7 -0.65 0.65 -1.06 -0.7 units box
region reg5 block 4.17 4.73 -0.65 0.65 -1.31 -1.06 units box
region reg6 block 4.35 4.73 -0.65 0.65 -1.5 -1.31 units box
region reg7 block 4.68 5.01 -0.65 0.65 -1.81 -1.5 units box
region reg8 block 4.19 5.89 -0.65 0.65 -2.05 -1.81 units box
region reg union 8 reg1 reg2 reg3 reg4 reg5 reg6 reg7 reg8 units box
create_box 1 reg
when I run LIGHHHTS, it says: Created orthogonal box = (-0.4 -0.65 -2.05) to (5.89 0.65 0.5)
By the look of things, there are undesired regions that LIGGGHTS is considering, so I made a test with just one region to compare:
region reg block -0.4 5.89 -0.65 0.65 -2.05 0.5 units box
create_box 1 reg
Obviously, It also says: Created orthogonal box = (-0.4 -0.65 -2.05) to (5.89 0.65 0.5)
The first simulation (with region optimized) lasted as mush as the second one (with all the regions that I don't want to consider)
Can anybody please instruct me about this topic?
Thanks in advance
sbateman | Fri, 04/19/2013 - 23:16
The simulation domain in
The simulation domain in LIGGGHTS is always a rectangular box. You can't have a different shape for the full simulation domain, but you can have regions with different shapes to, e.g., insert particles or do some other kind of processing during the run.
The speed of the simulation shouldn't depend on the shape of the box anyway. It should only depend on the number of particles, the load balancing between processors, and whatever fixes, computes, etc you have defined.