The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <gonze@pcpm.ucl.ac.be>]

Dear Francois Lallet and Francois Bottin,

This problem is fixed in v5.4.3 : now, the getng routine takes into  
account
the constraint due to nproc_fft to generate ngfft values.
Would be nice if you could check this when v5.4.3 is available (soon) !

Best regards,
Xavier

On 13 Jul 2007, at 21:28, Lallet François wrote:

> Dear Francois,
>
> I followed the modifications you suggested and the problem is fixed  
> now. Thank you very much for help.
> Regards
>
> Francois
>
>
>
> -----Original Message-----
> From: BOTTIN Francois [mailto:francois.bottin@cea.fr]
> Sent: Fri 7/13/2007 3:13 AM
> To: forum@abinit.org
> Subject: Re: [abinit-forum] Value of ngfftdg(3) uncorrect
>
> Dear Francois,
>
> francois.lallet@umontreal.ca a écrit :
> > Dear all,
> >
> > I tried to launch a band-fft job with npband=8 and npfft=4. I  
> > found the following error in the log file:
> >
> >  npfft and npband           4           8
> >  in initmpi_grid: me_fft and me_band are           0           0
> >  in initmpi:me_fft, me_band are           0           0
> >
> >  getng is called for the coarse grid:
> >  For input ecut=  1.500000E+01 best grid ngfft=      96       
> > 90      60
> >        max ecut=  1.506396E+01
> >  input values of ngfft(1) = 96 ngfft(2) = 96 ngfft(3) = 64 are  
> > alright and will be used
> >  getng: value of mgfft=      96 and nfft=      147456
> >  getng: values of ngfft(4),ngfft(5),ngfft(6)      97      97      64
> >  getmpw: optimal value of mpw=     959
> >
> >  getng is called for the fine grid:
> >  For input ecut=  3.500000E+01 best grid ngfft=     150      
> > 128      90
> >        max ecut=  3.530641E+01
> >
> >  getng : BUG -
> >   The third dimension of the FFT grid, ngfft(3), should be  a  
> > multiple of the number of processors for the FFT, nproc_fft.   
> > However, ngfft(3)=   90 and nproc_fft=    4
> >   Action : contact ABINIT group.
> >
> > .Delivered    1 WARNINGs and   1 COMMENTs to log file.
> > -P-0000
> > -P-0000  leave_new : decision taken to exit ...
> > MPI: MPI_COMM_WORLD rank 0 has terminated without calling  
> > MPI_Finalize()
> > MPI: aborting job
> I think your last dimension of ngfftdg  is not take into account  
> because
> it is not a multiple of 2,3 and 5: 92=2^2*23.
> Try to put ngfftdg(3)=96.
> > but in my input file I wrote:
> >
> > ecut 15
> > nband 584
> > ngfft 96 96 64
> >
> > ngfftdg 152 128 92
> >
> > npband 8
> > npfft 4
> > pawecutdg 35
> >
> > in order to prevent such a mistake... But obviously the input for  
> > ngfftdg was not read...
> > Note that I had a similar mistake but for ngfft with another job  
> > (npband=8 and npfft=4)
> > The log file:
> >
> >
> >  npfft and npband           4           8
> >  in initmpi_grid: me_fft and me_band are           0           0
> >  in initmpi:me_fft, me_band are           0           0
> >
> >  getng is called for the coarse grid:
> >  For input ecut=  1.500000E+01 best grid ngfft=      72       
> > 72     135
> >        max ecut=  1.535428E+01
> >
> >  getng : BUG -
> >   The third dimension of the FFT grid, ngfft(3), should be  a  
> > multiple of the number of processors for the FFT, nproc_fft.   
> > However, ngfft(3)=  135 and nproc_fft=    4
> >   Action : contact ABINIT group.
> >
> > .Delivered    1 WARNINGs and   1 COMMENTs to log file.
> > -P-0000
> > -P-0000  leave_new : decision taken to exit ...
> > MPI: MPI_COMM_WORLD rank 7 has terminated without calling  
> > MPI_Finalize()
> > MPI: aborting job
> >
> >
> > and the input file:
> >
> > ecut 15
> > nband 384
> >
> > ngfft 72 72 136
> It may be the same error: 136=2^3*17
> > ngfftdg 108 108 216
> > npband 8
> > npfft 4
> > pawecutdg 35
> >
> >
> > Any idea?
> > Thanks to all
> >
> > Francois
> Please note the following point. If you want to DECREASE the ngfft or
> ngfftdg with respect to the computed ones,
> you have also to decrease boxcutmin and bxctmindg.
>
> Regards
> Francois
>
> --
> ##############################################################
> Francois Bottin                    tel: 01 69 26 41 73
> CEA/DIF                            fax: 01 69 26 70 77
> BP 12 Bruyeres-le-Chatel         email: Francois.Bottin@cea.fr
> ##############################################################
>
>
> <winmail.dat>