forum@abinit.org

Subject: The ABINIT Users Mailing List ( CLOSED )

List archive

Re: [abinit-forum] large unit cell, few k-points and parallelisation

From: "Anglade Pierre-Matthieu" <anglade@gmail.com>
To: forum@abinit.org
Subject: Re: [abinit-forum] large unit cell, few k-points and parallelisation
Date: Tue, 25 Sep 2007 14:38:01 +0200
Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=beta; h=received:message-id:date:from:to:subject:in-reply-to:mime-version:content-type:content-transfer-encoding:content-disposition:references; b=k70+kD8xE+2Ri6saNg5778Xa/aWK7ZRX+nu63cNxgIXrtnCYRoL7IL42dgLJAeKr9VDPJ0N3E8aFh1EgydXdYex1EDgXmMzSGRhiDRuRBTzZzlLBczxC2tlF/UuvpkxOv2tyDhtoFff9MtjYd10ybQuyCjJTSZoh4Gc4U4ercnM=

Yes for sure. This is intended for that purpose. However I think this
is still, to some extent a "testing" feature of Abinit. So beware.

On 9/17/07, david.hendry@mandmd.com.au <david.hendry@mandmd.com.au> wrote:
> Hello,
>
> Low concentrations of doping requires large unit cells. Large unit cells
> require few k-points. 5.3.4 seems to have parallelisation over k-points
> only, therefore there seems to be less potential for parallelisation and
> hence memory distribution as systems grow in size. Specifically, a unit
> cell with 48 atoms and 20 k-points needs more RAM per cpu (>2GB) than can
> be provided when the system is parallelised over 20 cpus.
>
> Is there a way to get around this? I notice 5.4.3 includes beta FFT/band
> parallelisation. Would this allow more efficient distribution of memory
> across a number of cpus larger than the number of k-points?
>
> Thanks in advance,
>
> David.
>

--
Pierre-Matthieu Anglade

large unit cell, few k-points and parallelisation, david . hendry, 09/17/2007
- [abinit-forum] Reference for berryopt = -1, PGanesh, 09/19/2007
- Re: [abinit-forum] large unit cell, few k-points and parallelisation, Anglade Pierre-Matthieu, 09/25/2007