[Pw_forum] How to parallelize phonons

[William Parker]

I want to follow up on my last mailing.  I have since found, thanks to some discussion with Alejandro Rébola, that, if I copy only the data-file.*.*.xml files from the _ph* directories into _ph0, the recover step works correctly.

So my script goes something like this (using one processor per k-point pool):

mpirun -n $npool         pw.x                 -npool $npool -inp $pw_input         >& $pw_output
mpirun -n $npool*$nimage ph.x -nimage $nimage -npool $npool -inp $ph_input         >& $ph_output
i=1
while [ "$i" -lt "$nimage" ]
do
        cp _ph$i/$prefix.phsave/data-file.*.*.xml _ph0/$prefix.phsave/
        i=`expr $i + 1`
done
mpirun -n $npool         ph.x                 -npool $npool -inp $ph_recover_input >& $ph_recover_output

$ph_recover_input and $ph_input differ only in that the former contains the additional line:
recover=.true.

This procedure produces $prefix.dyn* files containing all of the expected eigenvalues and eigenvectors. 
I imagine a more fluent person could write the copy stage better.  May I update the documentation for ph.x to indicate this vital copy step?  

--William


On Mar 28, 2012, at 10:16 PM, William Parker wrote:

> Thanks for the followup, Paolo.
> 
> The recover step happens, producing the appropriate "Restart after Phonon calculation" message, but the output insists the representations calculated on other images are still "To be done."  ph.x produces a new .dyn file for the first q-point but no others.  So, the remaining .dyn files from the other images are incorrect because they are incomplete, even though all the data are correctly stored in the data-file.*.xml files in the _ph*/(prefix).phsave/ directories.   Any ideas?  Need any more information?
> 
> --William
> 
> On Mar 28, 2012, at 4:08 PM, Paolo Giannozzi wrote:
> 
>> 
>> On Mar 28, 2012, at 21:47 , William Parker wrote:
>> 
>>> The second step does generate all of the right dynamical matrix entries,
>>> but I have been unable to get the recover step to complete successfully
>>> yet, though I am currently working on this (if anyone else on the list has
>>> any comments on this, please speak up).
>> 
>> why have you been unable to perform the last step? what happens?
>> 
>> P.
>> ---
>> Paolo Giannozzi, Dept of Chemistry&Physics&Environment,
>> Univ. Udine, via delle Scienze 208, 33100 Udine, Italy
>> Phone +39-0432-558216, fax +39-0432-558222
>> 
>> 
>> 
>> 
> 
> *********************************************************
>  William D. Parker                 phone: (630) 252-4834
>  Computational Postdoctoral Fellow   fax: (630) 252-4798
>  MSD-212, Rm. C-215
>  Argonne National Laboratory
>  9700 S. Cass Ave.
>  Argonne, IL 60439
> *********************************************************
> 
> 
> 
> 
> 

*********************************************************
  William D. Parker                 phone: (630) 252-4834
  Computational Postdoctoral Fellow   fax: (630) 252-4798
  MSD-212, Rm. C-215
  Argonne National Laboratory
  9700 S. Cass Ave.
  Argonne, IL 60439
*********************************************************