dump_modify nfile problem

Dear all,

I have a small problem with
dump_modify nfile

Shortly, in the example below lammps is correctly writing a single file. However, only data of a single
processor make sense, i.e. data of roughly N/np particles. So I guess the processors are not passing
the information properly. Everything works if I remove the dump_modify command, of if I set nfile = number of processors.

> dump DUMPA_ALL all custom \{dtwrite\} dumps/dump\_{N}_\{rho\}\_{Temperature}_TW${TW}_${Rank}_P%.bin id xu yu zu
> dump_modify DUMPA_ALL nfile 1

Any suggestion?


I added these lines to bench/in.lj, for a problem size of 4000 atoms (10x10x10 lattice)

and turned things on/off.

dump 1 all custom 100 tmp. id xu yu zu #dump 1 all custom 100 tmp..bin id xu yu zu
#dump_modify 1 nfile 1

In all 4 cases (text or binary dump files, dump modify commented in or out),

I get the desired output when running on 4 procs.

E.g. 1 big file or 4 small files

So I’m either not clear on your question, or not seeing the same

Are you running the current version of LAMMPS?


Dear Steve,

thanks. The example file with the dump_modify addition works for me too.
So it must me a problem with my script or my data analysis. I will double check.
My apologize for bothering you - I should have check before.