The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <xavier.gonze@uclouvain.be>]

Dear Engin,

As a complement to Gilles' answer ...

On 20 Dec 2007, at 15:56, Engin Durgun wrote:

> Dear all,
>
> I am trying to perform paralel calculations which are not parallel  
> in kpoints but band/fft. When I tried to perform such kind of  
> calculations with abinit5.4.3,  I have reveived an error message  
> even for tutorial which related with Au-dimer. The error message  
> is:p4-error: interrupt SIGSEGV=11 and the input parameters (whicg  
> are exactly same with tutorial) (for 4 processors):
>
> nband 12
> iscf 5
> nstep 6
> tolvrs 1.0d-7
> wfoptalg 1
> nbdblock 2
> timopt 2

This corresponds to a band parallelisation only, that was implemented
a long time ago, and has never been very efficient.

The next that you tried is a recently implemented one (completely
different implementation), from CEA Bruyeres-le-chatel (GIlles,  
Francois Bottin, ...)

> when I tried the parameters in  /tests/paral (for 4 processors)
>
> npband 2
> npfft 2
> timopt -1
> fftalg 401 wfoptalg 4 fft_opt_lob 2
>
> The calculation is working but I noticed a warning
>
> "nproc =4  .= kpoint =1*nsppol =1
> the number of processors is larger than nkpt. This is a waste."
>
> which indicates (I quess) that the job is not working in parallel.  
> Am I doing something wrong? Does anyone has any experience about  
> such kind of paralelization?
>
> Thank you in advance...
>
> Engin
>
> ------------------------------------------------------------------
> Engin Durgun
> University of Liege
> Physics Department, Bat. B5    Phone : ++(32) (0)4-366.36.12
> B-4000 Sart Tilman                 Fax : ++(32) (0)4-366.36.29
> Belgium                                    E-mail:  
> engin.durgun@ulg.ac.be
> ------------------------------------------------------------------