[Wien] case.scf Concatenation in wien2K [Parallel execution]

[Javad hashemifar]

> Dear Javad
> Thank you for your response,  please note the following scenario with
> which i am working.
> 1. I am NOT using the parallel Version binaries or scripts provided with
> the application.
> 2. I dont have a NFS system where we can share the files from one
> machine to another.
> 3. We are sending the klist and lapw1_*.def and lapw2_*.def files onto
> remote machines.
> 4. In other Words my way of execution is different from that of the
> Parallel Execution carried out on a cluster or Share Memory Systems.
>
> Question:
> Is it possible to split the case.klist file so that we can send each
> kpoint to a remote machine and the processing for that kpoint is done
> independently over there without using data from other kpoints?

Dear Malik

Please have a look at the figures 5.1 and 5.2 of the user manual.
These figures are useful to find out how the program splits the job
during kpoint parallel calculation. Based on these figures it seems
that it is not possible to do both lapw1 and lapw2 programs on
different machines independently. Because after finishing lapw1 on all
machines and for calculation of the Fermi energy, all case.energy_X
files should be accessible for lapw2. When the Fermi energy is
calculated then again job (that is lapw2) will be split and solved on
different machines independently. After finishing lapw2 on all
machines the case.clmval_X files stored on different machines should
be combined together to make case.clmval file that is the input file
of mixer.
In the automatic parallel calculation it is sumpara program that sum
up all case.clmval_X files and I am not sure if it is an
straightforward task to be done manually.

Javad Hashemifar



> Javad hashemifar wrote:
> > Dear Malik
> >
> > Please note that case.scfX files are not the input files but they are
> > only output summary of the corresponding programs (lapwX, lcore, ...).
> > For example case.scf1 is the output summary of lapw1 NOT the input
> > file of lapw2.
> > But if you are still interested to combine the scf files of all group
> > of kpoints in a parallel run, please look at the lapw1para and
> > lap2para scripts. In these scripts (close to their ends) there are
> > commands that creates case.scf1 and case.scf2 out of case.scf1_X and
> > case.scf2_X files.
> > for example in the case of lapw1para the following commands does this job:
> >
> > sed "/NUMBER OF K-POINTS:/s/[0-9].*/$klist/" < $case.scf1${updn}_1
> >
> >> $case.scf1$updn
> >>
> >
> > It means that to make case.scf1 it is enough to write the correct
> > number of total kpoints at the end of case.scf1_1 and other
> > case.scf1_X files are not required.
> >
> > Bests
> > Javad Hashemifar
> >
> >
> >> I have a question regarding concatenation of multiple case.scf files.
> >> I am running wien2k on multiple sites, by splitting the case.klist file
> >> into multiple case.klist files and creating an appropriate lapw1_*.def
> >> and lapw2_*.def.
> >> i split the case.klist file in such a way that each case.klist_* file
> >> contains one kpoint data.
> >> after completing one iteration of the application i end up with multiple
> >> case.scf1_* and case.scf2_* files corresponding to each case.klist_* file.
> >> How do i combine these files to get a proper case.scf file which i can
> >> use for input to the second iteration.
> >>
> >>
> >> Many thanks for your help.
> >>
> >> best regards,
> >>
> >> Malik
> >>
> >> PS: Please note that neither i am using the -parallel switch nor i use
> >> the mpi version of  the application for parallelization.
> >> _______________________________________________
> >> Wien mailing list
> >> Wien at zeus.theochem.tuwien.ac.at
> >> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
> >>
> >>
> >
> >
> >
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien
>


-- 
======================================
Seyed Javad  Hashemifar,    Ph.D.
[current:]
Tel:+49-203-3794743        Fax:+49-203-3794742
Fachbereich Physik (AG Kratzer), Universitat Duisburg-Essen
47048 Dusiburg, Germany
[permanent:]
Tel:  +98-311-3912375       Fax: +98-311-3912376
Physics Department, Isfahan University of Technology
84156 Isfahan, Iran
---------------------------------------------------------------------------