[Pw_forum] What are wfc files outside prefix.save folder?

[Paolo Giannozzi]

There are two ways to write wavefunction files at the end of a pw.x or cp.x
run: "collected", that is, a single file per k-point whose content is
independent upon the number of processors used in the run; and
"distributed", that is, each processors writing its own slice of
wavefunctions. The former can be read on any number of processors; the
latter, only on the same number of processors that was used to write it.

The original format, and the internal format used in QE, is the distributed
one. To ensure back-compatibility, distributed files are still written one
level up the *.save directory, which contains collected files. In order to
manage both types of files in a transparent way, all codes that use data
from pw.x or cp.x read collected files if present, re-writing them as
distributed files; if not present, they read the distributed format. It is
a quick-and-dirty hack, but it is also dumb and wasteful. Distributed files
will disappear (hopefully) soon and this hack will disappear with them.
Note however that the phonon code may write other files, such as \Delta\psi
and \Delta V\psi, in the distributed format, and will likely continue to do
that, until some serious cleanup is done.

Paolo


On Sat, Dec 2, 2017 at 2:50 PM, balabi <balabi at qq.com> wrote:

> Dear developers,
>      I am using QE 6.2.
>      Normally, all the binary wfc files for each k points are all located
> inside prefix.save folder.
>      But I found that after running bands.x, there will be several wfc
> files outside prefix.save folder generated, and they are significantly
> larger than those wfc inside prefix.save folder. What is the purpose of
> generating these wfc files? Are they different from those wfc files inside
> prefix.save? Judging from the file size, they seems to be some kind of
> merged version of those wfc files inside prefix.save?
>      What is more, I found same behaviour when doing phonon GRID
> calculation( GRID_example/run_example_3).
>      I tested two kinds of running
>      1. The default "run_example_3" runs  alas.ph.wfc.in first, then
> followed by running alas.ph.wfc.in.
>           I then check the outdir, there are wfc files located inside each
> _ph0/prefix.q_{i}/prefix.save. Which is the normal behaviour.
>      2.  I found that if I skip the running of alas.ph.wfc.in, and
> directly run alas.ph.in0. The whole process proceed correctly ( I mean
> bands can be found when running each $q.$irr ) . But I found there are no
> wfc files inside each _ph0/prefix.q_{i}/prefix.save, however there are
> larger wfc files located in  each _ph0/prefix.q_{i}, that is outside
> prefix.save folder.
>
> So I got confused. It seems that no matter wfc is located inside .save
> folder or outside .save folder, they can both detected correctly by QE. Am
> I right? Then why there are two kinds of output? Is this some kind of
> inconsistency?
>
> I packed test files into a tar can be downloaded here
> https://drive.google.com/open?id=1rOm8V_jpAYBcCpwtgf9Czup-Cg0iSfm_
>
> best regards
>
>
>
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://pwscf.org/mailman/listinfo/pw_forum
>



-- 
Paolo Giannozzi, Dip. Scienze Matematiche Informatiche e Fisiche,
Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/users/attachments/20171204/cc459bb3/attachment.html>