[Q-e-developers] Undefined variables in espresso

[Filippo Spiga]

You are all right, the third dimension of becsum can have size 4 even if nspin_mag is equal 1. Thanks Vittorio to spot this problem!

I do not think the GPU version of addusdens is affected by that because both nspin and nspin_mag are used apparently in the proper way. I will double-check.

F


On Jun 10, 2013, at 4:01 PM, Andrea Dal Corso <dalcorso at sissa.it> wrote:
> 
> On Mon, 2013-06-10 at 15:38 +0800, Filippo Spiga wrote:
>> Dear Vittorio,
>> 
>> On Jun 10, 2013, at 2:47 PM, Vittorio Zecca <zeccav at gmail.com> wrote:
>>> In addusdens.f90 line 104 there is a shape mismatch because tbecsum
>>> has one element,
>>> while becsum has four.
>> 
>> 
>> 'tbecsum' is a one-dimentional array created to 'store' values of
>> 'becsum' varying 'nspin_mag' (latest dimension of a three-dimensional
>> array). It can have only one or two or four elements (it depends by
>> the input case, see modules/noncol.f90). I said that 'becsum' is
>> a three-dimensional array because inside the routine 'addusdens_g' the
>> array 'becsum' comes from USPP (Modules/uspp.f90). If you look at the
>> file you will see at line 136-137 that 'becsum' is defined as …
>> 
>> 
>>  REAL(DP), ALLOCATABLE :: &
>>       becsum(:,:,:)           ! \sum_i f(i) <psi(i)|beta_l><beta_m|
>> psi(i)>
>> 
>> 
>> Dimensions across these arrays seems consistent to me. Searching
>> across the source code I can find some four-dimensional "becsum"-like
>> arrays but they also have different names…
>> 
>> 
>> Carlo? Paolo? Anything to add?
> 
> It seems to me that Vittorio is right, tbecsum should be declared nspin,
> otherwise in the case in which nspin_mag=1 and nspin=4 you are copying 4
> elements of becsum in an array with one element. 
> In allocate_nlpot becsum is declared nspin.
> Maybe some compilers do not note the problem (so far I had not noticed
> it) because you use only one element of tbecsum, but I think it is not
> correct fortran.
> 
> Andrea
> 
>> 
>> 
>> Regards,
>> Filippo
>> 
>> --
>> Mr. Filippo SPIGA, M.Sc.
>> http://www.linkedin.com/in/filippospiga ~ skype: filippo.spiga
>> 
>> «Nobody will drive us out of Cantor's paradise.» ~ David Hilbert
>> 
>> 
>> _______________________________________________
>> Q-e-developers mailing list
>> Q-e-developers at qe-forge.org
>> http://qe-forge.org/mailman/listinfo/q-e-developers
> -- 
> Andrea Dal Corso                    Tel. 0039-040-3787428
> SISSA, Via Bonomea 265              Fax. 0039-040-3787249
> I-34136 Trieste (Italy)             e-mail: dalcorso at sissa.it
> 
> 
> _______________________________________________
> Q-e-developers mailing list
> Q-e-developers at qe-forge.org
> http://qe-forge.org/mailman/listinfo/q-e-developers

--
Mr. Filippo SPIGA, M.Sc.
http://www.linkedin.com/in/filippospiga ~ skype: filippo.spiga

«Nobody will drive us out of Cantor's paradise.» ~ David Hilbert

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20130610/68abbf55/attachment.html>