[lammps-users] Minimization question

Dear Lammps users,

I have noticed that I obtain different values of the potential energy after the minimization of the same data file (just by changing the number of parallel processors). In all case, as listed below, the energy is minimized but the number of steps is different, as well as, the final energy value. Perhaps this is correct since the stopping criteria might be local, but it called my attention. In my case, I am using the class2 force file (bonds, angles, dihedrals) and non-bonded interactions, no shake, no coulombic forces; any comments?

Procs, Steps, Min
28 128 4099.24
26 111 4248.35
24 138 4016.21
22 115 4053.56
20 111 4149.20
10 101 4220.48
8 122 4115.64
4 108 4213.59

BTW, the class2 bond page is broken, the link should be:


Regards, Javier

Where is the link to the class2 bond page coming from?

Re: min in parallel - once you are near the minimum, I'm not
suprised that the convergence could be slightly different on
differernt numbers of procs, due to round-off and precision issues.