[Pw_forum] Corrupting "input_tmp.in" when starting jobs at once
lmartinsamos at gmail.com
Mon Feb 6 14:04:46 CET 2012
yes allways!!! even with -in -inp .... open_input_file first dump to
2012/2/6 Silvia Bahmann <Silvia.Bahmann at physik.tu-freiberg.de>
> When using the following line in the jobfile "input_tmp.in" also is
> and possibly read?!
> mpirun -np 8 pw.x -in test.in > test.out
> On Monday 06 February 2012 13:47:46 Paolo Giannozzi wrote:
> > On Feb 6, 2012, at 13:26 , Layla Martin-Samos wrote:
> > > Dear Silvia, we moved to this intermediate input_tmp.in for
> > > preparing pw
> > > to the automatic detection of xml input format. There is no simple way
> > > to overcome this step.
> > there are a few possibilities to prevent overwriting by other processes:
> > - add a unique identifier, e.g. the process ID, to the file name;
> > - add a random number to the file name;
> > - rename the file to input_tmp_N.in, with N= smallest number such that
> > no input_tmp_N.in is detected;
> > but none is trivial to implement in a portable and safe way.
> > I don't think however that the input file is copied to input_tmp.in,
> > when
> > it is read using the -in option (as opposed to " < file"). So everything
> > should work if the -in (or -inp, or -input) option is used.
> > Paolo
> > ---
> > Paolo Giannozzi, Dept of Chemistry&Physics&Environment,
> > Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
> > Phone +39-0432-558216, fax +39-0432-558222
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the users