Problem when simulating larger number of particles

Submitted by knoe_ph on Thu, 10/04/2012 - 16:19

Hallo everbody,

I've got a Problem when simulating larger number of Particles.

For instance, i've created a simple simulation for testing with 10000 Particles over 50 timesteps. The simulation starts and finishes quite fine in a few seconds.

Then I take the exact same simulation (particles are insertet via: fix ins nve_group pour/dev nparticles ${n_Particle} 1 distributiontemplate pdd1 region bc) but with n_Particle equal 1000000.

Now simulation starts and on my screen it is displayed:

...
Setting up run ...
Memory usage per processor...
step atoms ....

then it should begin to run, but nothing happens anymore (tried for over 20 hours)

Does anyone know the reason for this or how it can be fixed?

Kind regards
Philipp

Forums:

ckloss's picture

ckloss | Thu, 10/04/2012 - 16:27

Hi Philipp,

the fact that you use pour/dev means that you are using an old version of LIGGGHTS. Please update to 2.1.2

Cheers, Christoph

knoe_ph | Fri, 10/05/2012 - 10:36

Hi Christoph,

I updated to the latest version of Liggghts and adapted my input script to the new commands and syntax (particle insertion now via : fix ins nve_group insert/pack seed 1 distributiontemplate pdd1 insert_every once overlapcheck yes particles_in_region ${n_Partikel} region bc).

Unfortunately the problem remains.

Kind regards,
Philipp

ckloss's picture

ckloss | Fri, 10/05/2012 - 11:23

Have you checked your time-step? Without further info, it's not possible to tell what is causing the trouble. I suggest you step-wise debug your simulation to find out the reason

CHristoph

knoe_ph | Mon, 10/08/2012 - 14:46

I tried again with the latest version of LIGGGHTS, this time it worked somehow.

It took almost 2 hours to complete the entire Simulation though, containing 1*10^6 Particles and 50 timesteps (with 1.8*10^(-5) seconds per timestep). This seems a really long time for me to finish the simulation, or is it a reasonable amount of time?

ckloss's picture

ckloss | Mon, 10/08/2012 - 15:14

Hi,

that's definitely too much - it shouldn't take more than a few sec to half a minute under normal circumstances. As written above, I suggest you step-wise debug your simulation to find out the reason

Cheers, Christoph

knoe_ph | Wed, 10/10/2012 - 10:09

Hi Christoph,

I did as you told me and simplified my simulation even more. Now the simulation takes only a couple of seconds like it should be. I'm not sure which option caused the problem, though. Anyway thanks for your help!

Cheers, Philipp