[Q-e-developers] problem with pp

Nicola Manini Nicola.Manini at mi.infm.it
Wed Jan 11 11:48:19 CET 2012


As Paolo mentioned, this is fixed.

I did single-version tests: in both tests (4.0-working and 4.3.2-failing) I 
used the matching version of pw.x and pp.x (with the same input files; the 
total energy and other outputs of pw.x look identical for the two calculations).

In file ge100-p2x1-out.save/data-file.xml
for v.4.0 there is no trance of NON-PERIODIC_CELL_CORRECTION,
while in the same file for v. 4.3.2 the following 3 lines 26-28 occur:
     <NON-PERIODIC_CELL_CORRECTION type="character" size="1" len="4">
None
     </NON-PERIODIC_CELL_CORRECTION>
which are the ones you mention.
So it is a different problem, probably the one due to the iotk interaction 
with the gfortran compiler, as suggested by Paolo.

Ciao,
         Nick


On 01/11/2012 11:22 AM, Andrea Ferretti wrote:
>
>
> Hi Nicola,
>
> recently I've spotted that QE-4.3.1 (and newer versions) writes (and
> then wants to read) the flag "NON-PERIODIC_CELL_CORRECTION" (CELL
> section, data-file.xml).
> This makes a the code no longer back-compatible... I've found this
> working 4.3.1 on top of 4.3
>
> I'm not sure this is the only reason of your problem, a possible fix
> follows:
>
> * quick and dirty to avoid re-running the code (it worked in my case, not
> sure how general it is):
> add something like:
>
> <NON-PERIODIC_CELL_CORRECTION type="character" size="1" len="4">
> None
> </NON-PERIODIC_CELL_CORRECTION>
>
> after the <CELL> tag in data-file.xml (beware of the length of the
> string "None" that should start at the beginning of the line)
>
>
> * in order to restore the backcompatibility:
>
> - increase the version number of the QEXML fmt (because the fmt have
> been changed)
> - if the fmt is older that the one of 4.3.1, assume "None" is used
> otherwise, scan to get the proper value of the flag...
>
> (if we agree on this I can put in in)
>
> Andrea
>
>
>
>
>> Hello,
>> recent versions of pp.x seem to have turned incompatible with previous
>> input format. The following input:
>> &inputpp
>> outdir='./' ,
>> prefix='ge100-p2x1-out',
>> filplot = 'ge100.pot' ,
>> plot_num= 1 ,
>> /
>> &plot
>> iflag = 3,
>> output_format = 5,
>> fileout = 'ge100.xsf',
>> /
>>
>> was processed fine until version 4.0 of QE (of course assuming that a
>> ./ge100-p2x1-out.save/ folder containing the appropriate files was
>> there). The output log is attached (pp_v4.0.out).
>>
>> The current version 4.3.2 produces the attached log (pp_v4.3.2.out)
>> and in stopping generates the following stderr message:
>> ########################################################################################################################
>>
>> # FROM IOTK LIBRARY, VERSION 1.2.0
>> # UNRECOVERABLE ERROR (ierr=1)
>> # ERROR IN: iotk_scan_end (iotk_scan.f90:241)
>> # CVS Revision: 1.23
>> # foundl
>> # ERROR IN: iotk_close_read (iotk_files.f90:746)
>> # CVS Revision: 1.20
>> ########################################################################################################################----------------------------------------
>>
>> followed by a MPI_ABORT message.
>> Of course, no .pot & .xsf are generated.
>>
>> Is this a documented problem?
>> Maybe related to
>> http://qe-forge.org/scm/viewvc.php/trunk/espresso/PW/pw_restart.f90?root=q-e&r1=7831&r2=7830&pathrev=7831
>> ?
>>
>> For completeness, I note that also the intermediate version 4.3 fails
>> on the same input, with a similar but not quite identical error message:
>> ########################################################################################################################
>>
>> # FROM IOTK LIBRARY, VERSION 1.2.0
>> # UNRECOVERABLE ERROR (ierr=1)
>> # ERROR IN: iotk_scan_end (iotk_scan.spp:233)
>> # CVS Revision: 1.23
>> # foundl
>> # ERROR IN: iotk_close_read (iotk_files.spp:645)
>> # CVS Revision: 1.20
>> ########################################################################################################################----------------------------------------
>>
>>
>> Thank you for any help, and happy 2012!
>> Nick Manini
>>
>



More information about the developers mailing list