[Q-e-developers] A few questions about implementing qe post-processing

Paolo Giannozzi p.giannozzi at gmail.com
Thu Aug 25 20:02:10 CEST 2016


1) correct
2) correct
3) this doesn't sound right to me: the data-file.xml file is also
overwritten with new data, I think. We have to check.
4) different compilers may have different ideas on how to deal with "print
*", "write(*,...)", "write(0,...)". The BG/Q in particular is a rather
special machine. There is some (partly obsolete) info in
Modules/error_handler.f90

Paolo

On Thu, Aug 25, 2016 at 7:25 PM, Ye Luo <xw111luoye at gmail.com> wrote:

> Hi QE developers,
>
> I'm working on improving the interface program pw2qmcpack and I've a few
> questions.
>
> Usually we do scf->nscf->pw2qmcpack to get the final WF file. I need qe
> expert to help me understanding better the pw.x behavior.
> Assuming wf_collect=.false.
> I mean "parallel setup" by procs and k-pools. Only 1 image is used.
>
> 1, It is not required to run the nscf with the same parallel setup as the
> scf run. I tried, it seems working well. nscf doesn't need WF but density,
> right?
> 2, If I use a complete different list of k points and use different
> parallel setup for the nscf. The WF stored in "*.wfc?" in the out
> directory  will be the one correspond to nscf run because the WF of scf run
> is overwritten by nscf run. It seems so, but please confirm.
> 3, The following pw2qmcpack run needs to load the WF from the last nscf
> run. I'd like to add gatekeeping code to check whether the converter is run
> with the same parallel setup as the last nscf run. I tried checking
> nproc_pool_file as the postproc.f90. It seems that this variable keeps the
> original value from scf run rather than nscf run. So is there a way to
> access the last nscf paralle setup?
> 4, I noticed that when I tried to write(*,*) from the first MPI task
> associated to a pool. I got the contents on Intel build but not on BG/Q. It
> is not critical but introduce difficulty to print out some debug info. I
> feel strange that the same code show different behaviors. If you know the
> reason, please explain it.
>
> Thanks a lot!
>
> Ye
>
> ===================
> Ye Luo, Ph.D.
> Leadership Computing Facility
> Argonne National Laboratory
>
> _______________________________________________
> Q-e-developers mailing list
> Q-e-developers at qe-forge.org
> http://qe-forge.org/mailman/listinfo/q-e-developers
>
>


-- 
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




-- 
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/developers/attachments/20160825/cc51bf60/attachment.html>


More information about the developers mailing list