[lammps-users] Processors command

Hi,Steve:
Recently I found that with different “Processors” command,I would get different results,Acturely when I tested with 2 processors, “Processor 1 1 2” and “Processor 1 2 1” produced different results.How does this happen and whether there is a reliable way to decide which partition manner should be adopted.
I also wonder if different processor numbers will also give different results.
thanks

Hi, Steve. A general question about this issue. In sims where random numbers are used in LAMMPS, the random number generator is parallelized. Might that not, in principle, lead to (typically slightly) different results using different processor grid and/or different numbers of processors? The same set of random numbers applied in different spatial locations might create different results, especially in simulations where “rare events” are important? Of course, these tend to occur (semi) stochastically in any case, so different seeds for the random number generators could produce such effects even in a serial code. But anyway, wondering if you (or others on the list) have thoughts on this.

Thanks,
Rob

hi rob,

a couple of comments. random number generators used in codes like
LAMMPS don't generate 'real' but 'pseudo' random numbers. if you seed
the algorithm with the same number, it should produce the same sequence
of random numbers (unless floating point math issues and optimizing
compilers get in the way, but they should not in theory).

how to make reproducible sequences of pseudo random numbers
across different processor sets in parallel is a tricky issue.

cheers,
   axel.