[QE-users] wavefunctions writing

Ivo Batistic ivo at phy.hr
Wed Dec 4 14:45:25 CET 2019


Hi Paolo and Pietro,

Just for you to know how things got resolved.
After I have made the changes that Pietro had suggested,
the calculations crashed with no message.

OK, I cannot restart the job, but two weeks are not lost.
The already calculated energies were saved in
the file data-file-schema.xml.

I extracted k-points where energies were zero (not set),
and in order to create a list of k-points for 'bands' type of calculation.

I erased the existing wavefunction files, making more disk space
for new wavefunctions, and I started a new job, (with disk_io='medium').

And I hope I will be able to merge the new data with old one.
(in data-file-schema.xml?)

Thank you all very much.
And my best regards

Ivo Batistic
Department of Physics
Faculty of Science, University of Zagreb

03. 12. 2019. u 19:38, Paolo Giannozzi je napisao/la:
> I am not sure I understand the situation: are only eigenvalues required? 
> if so, there should be a *.xml file that contains already computed 
> eigenvalues. You may just compute the remaining k-points and make a merge.
> 
> For future reference: disk_io='none' does not write the xml file, but it 
> is sufficient to move the first instruction in "punch.f90":
>      IF (io_level < 0 ) RETURN
> after
>     CALL pw_write_schema( only_init, wf_collect )
> Maybe this should actually be the default behavior for disk_io='none'. 
> In case you are young and wondering where this funny name "punch" comes 
> from: it is a relict of the era of "punched cards"
> 
> Paolo
> 
> On Tue, Dec 3, 2019 at 12:09 PM Pietro Delugas <pdelugas at sissa.it 
> <mailto:pdelugas at sissa.it>> wrote:
> 
>     in qe-6.5Beta it is at line 56 but yes that's the line.
> 
>     you can restart the job with a different io just specifying it in
>     the namelist.
> 
>     Be aware that you could encounter memory issues.
> 
>     On 03/12/19 11:52, Ivo Batistic wrote:
>>
>>     Hi Pietro,
>>
>>     Let me just explain the situation:
>>
>>     * I have started nscf run two weeks ago with disk_io='medium',
>>     * Few day ago I have noticed that disk is almost full,
>>       so, I have suspended calculations, (prefix.EXIT)
>>     * and I want to restart (and not to lose two weeks)
>>       but without wavefunctions writing.
>>
>>     I do not know can I restart job with a different disk_io.
>>     Replace value in *.xml files?
>>
>>     So I stick with disk_io  = 'medium'.
>>
>>     OK, regarding your suggestion,
>>
>>     I cannot find io_level at line 57 (QE-6.3?).
>>
>>     The closest not commented line is 61:
>>
>>     CALL open_buffer( iunwfc, 'wfc', nwordwfc, io_level, exst_mem,
>>     exst_file )
>>
>>     where I can replace io_level with zero.
>>     Is this what you meant?
>>
>>     Ivo
>>
>>     PS: similar line at QE-6.4.1 code base is at line 54!
>>
>>     03. 12. 2019. u 11:18, Pietro Delugas je napisao/la:
>>>     Hello Ivo
>>>
>>>     if you use disk_io = 'low' and verbosity = 'high' the program
>>>     should print the band structure in output and avoid the usage of
>>>     scratch files, the return in io_base preventing the printout of
>>>     collected wavefunctions.
>>>
>>>     If instead you want to use disk_io  = 'medium' you can try to
>>>     replace io_levelat  line 57 ofPW/src/wfcinit.f90with a 0.
>>>
>>>     hope it helps
>>>
>>>     greetings - Pietro
>>>
>>>
>>>     On 03/12/19 10:49, Ivo Batistic wrote:
>>>>
>>>>     Thanks very much Pietro,
>>>>
>>>>     I have checked and
>>>>     this indeed prevents wavefunctions writing in
>>>>     the final stage, in prefix.save directory.
>>>>
>>>>     However, if one use disk_io='medium',
>>>>     there are temporary files for wavefunction storage
>>>>     in working directory,
>>>>     prefix.wfcN
>>>>     which at each step (k-point) grow bigger.
>>>>
>>>>     Is there any magic, like RETURN, to prevent
>>>>     wavefuctions writing in these temporary files.
>>>>
>>>>     I would be very  grateful for this info.
>>>>
>>>>     Best regards
>>>>
>>>>     Ivo B.
>>>>
>>>>     Department of Physics
>>>>     Faculty of Science, University of Zagreb
>>>>
>>>>
>>>>     02. 12. 2019. u 11:18, Pietro Delugas je napisao/la:
>>>>>     Hi Ivo
>>>>>
>>>>>     quick and dirty: just put a RETURN statement in
>>>>>     Modules/io_base.f90  at line 74 or there around and then
>>>>>     recompile pw.x. Obviously once done remember to remove it ;)
>>>>>
>>>>>
>>>>>     for a more durable implementation one could add more modulation
>>>>>     in disk_io strings ...
>>>>>
>>>>>     hope it helps
>>>>>
>>>>>     regards - Pietro
>>>>>
>>>>>     On 02/12/19 11:03, Ivo Batistic wrote:
>>>>>>     Hi to QE developers,
>>>>>>
>>>>>>     Is is possible to disable wavefunctions writing
>>>>>>     in nscf calculations.
>>>>>>     for Fermi-surface calc, afterwards.
>>>>>>
>>>>>>     (in QE-6.3 if it is version dependent)
>>>>>>
>>>>>>     disk_io='none' is not an option, since
>>>>>>     it does suppress also writing of the energies.
>>>>>>
>>>>>>     I. Batistić
>>>>>>
>>>>>>     Department of Physics
>>>>>>     Faculty of Science, University of Zagreb
>>>>>>     _______________________________________________
>>>>>>     Quantum ESPRESSO is supported by MaX
>>>>>>     (www.max-centre.eu/quantum-espresso
>>>>>>     <http://www.max-centre.eu/quantum-espresso>)
>>>>>>     users mailing list users at lists.quantum-espresso.org
>>>>>>     <mailto:users at lists.quantum-espresso.org>
>>>>>>     https://lists.quantum-espresso.org/mailman/listinfo/users
>>>>>
>>>>>     _______________________________________________
>>>>>     Quantum ESPRESSO is supported by MaX
>>>>>     (www.max-centre.eu/quantum-espresso
>>>>>     <http://www.max-centre.eu/quantum-espresso>)
>>>>>     users mailing list users at lists.quantum-espresso.org
>>>>>     <mailto:users at lists.quantum-espresso.org>
>>>>>     https://lists.quantum-espresso.org/mailman/listinfo/users
>>>>>
>>>>
>>>>     _______________________________________________
>>>>     Quantum ESPRESSO is supported by MaX
>>>>     (www.max-centre.eu/quantum-espresso
>>>>     <http://www.max-centre.eu/quantum-espresso>)
>>>>     users mailing list users at lists.quantum-espresso.org
>>>>     <mailto:users at lists.quantum-espresso.org>
>>>>     https://lists.quantum-espresso.org/mailman/listinfo/users
>>>
>>>     _______________________________________________
>>>     Quantum ESPRESSO is supported by MaX
>>>     (www.max-centre.eu/quantum-espresso
>>>     <http://www.max-centre.eu/quantum-espresso>)
>>>     users mailing list users at lists.quantum-espresso.org
>>>     <mailto:users at lists.quantum-espresso.org>
>>>     https://lists.quantum-espresso.org/mailman/listinfo/users
>>>
>>
>>     _______________________________________________
>>     Quantum ESPRESSO is supported by MaX
>>     (www.max-centre.eu/quantum-espresso
>>     <http://www.max-centre.eu/quantum-espresso>)
>>     users mailing list users at lists.quantum-espresso.org
>>     <mailto:users at lists.quantum-espresso.org>
>>     https://lists.quantum-espresso.org/mailman/listinfo/users
>     _______________________________________________
>     Quantum ESPRESSO is supported by MaX
>     (www.max-centre.eu/quantum-espresso
>     <http://www.max-centre.eu/quantum-espresso>)
>     users mailing list users at lists.quantum-espresso.org
>     <mailto:users at lists.quantum-espresso.org>
>     https://lists.quantum-espresso.org/mailman/listinfo/users
> 
> 
> 
> -- 
> 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
> 
> 
> _______________________________________________
> Quantum ESPRESSO is supported by MaX (www.max-centre.eu/quantum-espresso)
> users mailing list users at lists.quantum-espresso.org
> https://lists.quantum-espresso.org/mailman/listinfo/users
> 



More information about the users mailing list