[Q-e-developers] issue with ph.x and pseudo file location

Giovanni Pizzi giovanni.pizzi at epfl.ch
Wed Oct 10 17:11:02 CEST 2012


Dear QE Developers,
we have just noticed something odd in ph.x, i.e., that at some point the 
ph.x code (seems to) try to read the UPF files from the pseudo dir 
instead of the output 'save' dir.

Here are the steps that we have done:
1. We have done a vc-relax calculation with
   outdir = './out/',
   prefix = 'aida',
   pseudo_dir = './pseudo/',
2. at the end, we have created a new empty directory somewhere else (on 
the same machine), and within it we have copied only the ./out directory 
[actually, we also created an empty 'pseudo' dir, without any file in it].
3. We have started the phonon calculation from the new dir (that thus 
contains only an 'out' subdir, an empty 'pseudo' dir, and the aida.in 
input file).
4. The calculation starts correctly, and actually the calculation for 
the first q point (gamma) is completed correctly.
But then, at the beginning of the next q point, the code crashes because 
it can't find the pseudos in the 'pseudo' dir (see attached output file 
and stderr file).

Does anyone knows why the code tries to read from the pseudo folder? We 
believe that it should always read from the out dir (also because no 
'pseudo_dir' is specified in the ph.x input file).
[by the way, the matdyn-1 file after the crash is empty, even if the 
calculation at Gamma has finished and the code is already processing the 
new q point - maybe the file is not closed right away?]

Best,
Giovanni Pizzi and Andrea Cepellotti

-- 
Giovanni Pizzi
Post-doctoral Research Scientist
EPFL STI IMX THEOS
MXC 319 (Bâtiment MXC)
Station 12
CH-1015 Lausanne (Switzerland)
Phone: +41 21 69 31159

-------------- next part --------------
A non-text attachment was scrubbed...
Name: pseudo_phonons.zip
Type: application/zip
Size: 9401 bytes
Desc: not available
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20121010/e048695b/attachment.zip>


More information about the developers mailing list