Skip to Content.
Sympa Menu

forum - Re: [abinit-forum] MPI options

forum@abinit.org

Subject: The ABINIT Users Mailing List ( CLOSED )

List archive

Re: [abinit-forum] MPI options


Chronological Thread 
  • From: BOTTIN Francois <francois.bottin@cea.fr>
  • To: forum@abinit.org
  • Subject: Re: [abinit-forum] MPI options
  • Date: Thu, 18 Dec 2008 15:10:17 +0100
  • Organization: CEA-DAM

Hi,

If you want to use the triple
band/FFT/k-point parallelization, you have to use the following keywords:
paral_kgb=1, wfoptalg=4, nloalg=4, fftalg=401, intxc=0, and fft_opt_lob=2.
This defines the triple parallelization (paral_kgb=1) with LOBPCG (wfoptalg=4) and the FFT of Goedecker (fftalg=401)
with zero padding. The other heywords indicate some limitations which remain (intxc=0 and nloalg=4) or optimization ( fft_opt_lob=2) in LOBPCG.

For the other keywords of this parallelization (npband, npfft, npkpt ...) see
http://www.abinit.org/Infos_v5.6/input_variables/varpar.html#paral_kgb

Francois

Manuel Pérez Jigato a écrit :
20081217134209.ay3i8mzurs4kk84c@webmail4.kuleuven.be" type="cite">hi

there exist several input options for parallel running

fft_opt_lob

paral_kgb

paral_rf

except for the last one, which i guess is not needed except for perturbation theory, the first two, can they be used for standard LDA calculations?
i guess they both involve the lobpcg method. Are they compatible with Goedecker-FFT?

thanks

Manuel Perez Jigato
KU Leuven





Disclaimer: http://www.kuleuven.be/cwis/email_disclaimer.htm



-- 
##############################################################
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
##############################################################



Archive powered by MHonArc 2.6.15.

Top of Page