[Pw_forum] Parallel scf calculations

[MKondrin]

Francesco Pelizza wrote:
> Hello Dear Paolo and co-workers,
>
> I am having some difficulties in obtaining output file when I run 
> parallel calculation with scf (pw.x).
>
> Using OpenMPI, I used these commands:
> mpirun -np 4 ~/Software/espresso-5.1/bin/pw.x -ni2 -nk2 -i pvdf.scf.in 
>  >pvdf.scf.out
> mpirun -np 4 ~/Software/espresso-5.1/bin/pw.x -i pvdf.scf.in >pvdf.scf.out
>
>
> and the calculation works but the problem is the output file, where 
> everything is written many times, the same number of times of the number 
> of processors i've been used, so if i use 4 processor the output file 
> data are written 4 times, if i use 3 processors the data are written in 
> the same output file 3 times and so on
> and everything in the output file became FUZZY
>
> grep or awk command give multiple data instead of only once per 
> simulation, it prints me many data as the same amount of processors i use
>
> I had the same problem with the input file with this synthax:
> mpirun -np 4 ~/Software/espresso-5.1/bin/pw.x -ni2 -nk2 < pvdf.scf.in 
>  >pvdf.scf.out
>
> I solved the problem for the input file with this synthax:
> mpirun -np 4 ~/Software/espresso-5.1/bin/pw.x -ni2 -nk2 -i pvdf.scf.in 
>  >pvdf.scf.out
>
> This synthax doesn't solve the problem:
> mpirun -np 4 ~/Software/espresso-5.1/bin/pw.x -ni2 -nk2 -i pvdf.scf.in 
> -o pvdf.scf.out
>
>
> So now the input file is written only once (avoiding problems during the 
> reading), but i do not know how to solve the problem witht the output 
> file tha tis still written many times instead of only once
>
> I am not finding anywhere a complete list of argument i can use in 
> pw.x......
>
> Can you help me?
>
> Best wishes
>
> Sincerly
>
> Francesco Pelizza
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>
>   
Did you really use parallel version of pw.x? If so you should have in 
the beginning of output file something like:

  Parallel version (MPI), running on    64 processors
     K-points division:     npool     =    8
     R & G space division:  proc/pool =    8

Your problems looks like your having serial  copies of the pw.x launched 
on different nodes  and each  of them  overwriting the same output file.

Best regards,

M. V. Kondrin (High Pressure Physics Institute RAS, Troitsk, Russia)