The ABINIT Users Mailing List ( CLOSED )

[Deyu Lu <deyulu@yahoo.com>]

For people who want to take the advantage of the
band/fft parallelization in non-self-consistent
calculations on, e.g., isolated molecules, my colleage
and I figured out a way within the current
implementation. It can be done in two datasets. In the
first one, converge the GS density as usual with e.g.
"toldfe". Then run the subsequent dataset with
"tolwfr" to converge the orbitals. One also needs to
set "nnsclo" to a large enough value in order to
converge "tolwfr" within the 1st SCF circle. A example
input is given below:

ndtset      2
istwfk      1

npband 4 npfft 1
timopt -1
fftalg 401 wfoptalg 4 fft_opt_lob 2
iprcch 0 intxc 0

# DATASET 1
toldfe1  1.0d-8    # Will stop when this tolerance is
achieved on total energy
nband1      40         # Number of (occ and empty)
bands to be computed
prtden1     1

# DATASET 2
getden2  1             # Read previous density file
tolwfr2  1.0d-12        # Still get it converged
nband2   100
nnsclo2  100
nbdbuf2  2

..............


Best
Deyu Lu


--- BOTTIN Francois <francois.bottin@cea.fr> wrote:

> Dear Deyu Lu,
> You're right. The Band/FFT parallelization is not
> yet available in the 
> framework of non-self-consistent calculations.
> Sorry, this point is not documented. Moreover, there
> is no check of 
> input variables.
> This feature is under development in our lab and
> will be available as 
> soon as possible.
> Regards
> Francois
> 
> deyulu@yahoo.com a écrit :
> > Dear Abinit users:
> >      This is partially mentioned in my previous
> post, but I feel that it is better to put the
> discussion under a proper subject. I was testing the
> band/fft parallelisation in version 5.4.3 on a
> powerpc-ibm-aix5.2.0.0 machine (the hostname.ac and
> input file are listed below). I was able to run the
> test case 
> > t_bandfft.in and also cases within SCF mode. Then
> the job crashed when I tried to run
> non-self-consistent calculations with iscf=-2. In
> the log files, I found that the convergence was
> reached/nearly reached, and then the job crashed.
> Depending on the input, I got two types of error
> messages. 
> >
> > 1. npband 4 npfft 1:  Segmentation fault
> > 2. npband m npfft n(n>1) : 
> > -P-0000  prep_fourwf: BUG -
> > -P-0000   ndatarecv******** /= dimtabs ********
> > -P-0000
> > -P-0000  leave_new : decision taken to exit ...
> >
> >      Does someone know how to fix it?
> >      Thanks
> >      Deyu Lu
> >   
> 
> 
> -- 
>
##############################################################
> 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
>
##############################################################
> 
> 



      
____________________________________________________________________________________
Don't let your dream ride pass you by. Make it a reality with Yahoo! Autos.
http://autos.yahoo.com/index.html