Hi,
the memory usage of LIGGGHTS 3.0.1 seems to be decreased a lot in comparison to 2.3.8. I did run the same script on both versions starting with a cloud of levitating particles which are compacted by stepwise particle growth algorithm, so the number of particle contacts is constantly increasing. At the start the two branches of LIGGGHTS seems to use nearly the same amount of memory. The more the particles are getting in contact, the more memory is used. At the end LIGGGHTS 3 uses 10 to 20 (!) times less memory then LIGGGTS 2 (20.000 – 120.000 particles). For comparison the numbers given by the output of the run command were applied. Since htop shows not such a big difference but includes cache etc, can I rely on the output of run command? In the release notes I couldn't find any hint to this remarkable improvement.
I also noticed a decrease in calculation time, thanks for this.
Regards,
Paul
PaulWinkler | Tue, 05/20/2014 - 10:57
Hi,
Hi,
since the memory limits loosen, I explore new ones ;-) So I found one.
In a simulation nearly 20 Million (20E6) particles should be inserted by insert pack, but these fails immediately with the warning that not all particles could be inserted. Actually not a single particle is inserted, so I guess in preparation for the insertion procedure there is something like an out of range exception. What are the limit of the fix insert pack?
Regards,
Paul
ckloss | Fri, 06/06/2014 - 16:39
Hi Paul,
Hi Paul,
yes, in 3.X there is less memory allocation for polydisperse systems
I guess there is an overflow somewhere, you should normally get a warning about this?
If not, can you upload the case?
Thanks
Christoph