forum@abinit.org

Subject: The ABINIT Users Mailing List ( CLOSED )

List archive

Re: [abinit-forum] Memory distribution in Band Structure calculation

From: "Anglade Pierre-Matthieu" <anglade@gmail.com>
To: forum@abinit.org
Subject: Re: [abinit-forum] Memory distribution in Band Structure calculation
Date: Thu, 6 Dec 2007 08:31:14 +0100
Domainkey-signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=message-id:date:from:to:subject:in-reply-to:mime-version:content-type:references; b=utM0Zs6fH62QWcewkqtO/mgUkk1CGOHIbOd7cmKC8evEXTm9fykVHkDvnNd0O5vFsf6nQIonzUB/Hn2JoHhI+dNY8EPAigDZevdv6a0axYnkMugmpyY/RGlcSHtn9jzlHDBPr7kz+hmMQlVvTs7VWqnagd6IT4bOQkm97ajT294=

Hi,

Reducing the number of bands and ecut at the expense of convergency is likely not the thing to do because your results will become meaning less. The best way to reduce memory consumption are:
Go to parallel execution (very nice)
Use mffmem and mkmem (nice)
I think you will get your best results out of a combination of those two.

REgards

PMA

On Dec 4, 2007 5:01 AM, Hua Bao < hbao@purdue.edu> wrote:

Dear all,

I am calculating thei band structure for a large system with 120 atom in
the unit cell. The estimate memory need is 20 GB. I am wondering how
will the memory need distribute to each node. Currently I am using a
dual processor dual core system with the memory of 8GB. Is it possible
to do such a calculation using parallel abinit with three such computers?

The other way is to limit the accuracy of this calculation. I have
already limit the ecut and nband, and also using a fairly small grid for
k point. Is there any other parameter affect the memory need?

Thanks,
Hua

--
Pierre-Matthieu Anglade

Memory distribution in Band Structure calculation, Hua Bao, 12/04/2007
- Re: [abinit-forum] Memory distribution in Band Structure calculation, Anglade Pierre-Matthieu, 12/06/2007