Dear All,
My setup is gas particle confined in solid nanogap, and I am simulating it using LJ potential.
I would like to set up neighbor list at the very beginning of the simulation at once, and would like to update only gas-gas and gas-solid interaction during the simulation (basically I would like to turn off the interaction between solid-solid since it does not change the neighbor list at all) to save the neighbor list build up time.
It looks like “exclude” command excludes the interaction (they do not include the interactions at all during the entire simulation).
Also, it seems like “include” check the interaction among the group only, but when I use this option by grouping both gas and solid together, it check solid-solid interaction as well.
Do you have any comments/suggestions?
Thanks, Gisuk.
Dear All,
My setup is gas particle confined in solid nanogap, and I am simulating it
using LJ potential.
I would like to set up neighbor list at the very beginning of the simulation
at once, and would like to update only gas-gas and gas-solid interaction
during the simulation (basically I would like to turn off the interaction
between solid-solid since it does not change the neighbor list at all) to
save the neighbor list build up time.
do you have some dependable data, as to how much of your simulation
time this would actually save you.
many times, when people come with questions like yours, they are
asking about issues, that are a minor fraction of the total time and
thus the savings are often less than the time spent on implementing
them. therefore it is always a good idea to quantify the potential of
savings.
It looks like “exclude” command excludes the interaction (they do not
include the interactions at all during the entire simulation).
yes, and the major savings is less so the time spent on building the
neighbor list, but from the fact that the number of pair interaction
to be computed drops. if the particles in your solid structure are not
time integrated, this is the way to go. if those particles *are* time
integrated, there is not much to be saved from tweaking the neighbor
lists, as you'll have to do the binning and looping over particles in
bins anyway. as i mentioned the major savings do not come from the
neighbor list construction.
with setups like these, you should also look into optimizing the
processor load distribution, e.g. via the processors or the balance
command.
Also, it seems like “include” check the interaction among the group only,
but when I use this option by grouping both gas and solid together, it check
solid-solid interaction as well.
yes, of course.
Do you have any comments/suggestions?
what else are you looking for?
axel.