[Pw_forum] ph.x - from broyden : error ...

Paolo Giannozzi giannozz at nest.sns.it
Wed Nov 14 16:56:24 CET 2007


On Nov 14, 2007, at 13:25 , Katalin Gaal-Nagy wrote:

> once again a question about a crashed phonon job...
> It happened to me as it happened to Marcel Mohr (30/10/2007), that a
> phonon calculation crashed and the recover did not work ... well,  
> that's
> the question.
>
> In my case the original calculation ended because of
>       from davcio : error #        99
>       i/o error in davcio

it looks like the disk is full or, worse, faulty...

> Do I have to restart the calculation from scratch or is there some  
> way to
> recover it avoiding this error? Maybe it would be enough to restart  
> the
> calculation of Representation 62 from scratch by deleting/editing some
> files in the tmp_dir?

the way restart is currently implemented makes it very hard to deal
with a case like yours. First of all, files with information on the  
status
of the calculation are unformatted; second, the overall logic is  
complicated
and one is never sure what has been already done and what hasn't.
Until the day somebody with enough good will and knowledge will
need to restart long phonon runs in a reliable way, the phonon restart
procedure will stay the way it is, i.e., adventurous.

In my opinion, your best chance is to modify the routine
PH/phq_recover.f90 so that the code restarts from the
beginning of the current mode (the one where the code
crashed). I vaguely remember that there was a potential
problem with some pieces of the dynamical matrix (dyn0
and the like) but I don't remember exactly what it was.

Paolo
---
Paolo Giannozzi, Dept of Physics, University of Udine
via delle Scienze 208, 33100 Udine, Italy
Phone +39-0432-558216, fax +39-0432-558222






More information about the users mailing list