forum@abinit.org

Subject: The ABINIT Users Mailing List ( CLOSED )

List archive

Re: [abinit-forum] benchmark calculations for parallel ABINIT

From: "Anglade Pierre-Matthieu" <anglade@gmail.com>
To: forum@abinit.org
Subject: Re: [abinit-forum] benchmark calculations for parallel ABINIT
Date: Sat, 29 Sep 2007 10:01:27 +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=CzvDSclLlnTpfDae1XHF9bHaNS63yxGqx0jOM8cSMJvpMl59a7af1MjJqXj9u/B3aiZ3mUswKaD+ou8V2MBwB5PA2BQCuDU5tHcd4gBiWi8WqjDH2EPRzEIJ83euNXhXr5dZRlRjdH2qi15DWN3maCQ/ruS8GUTwOEWHRZGfuGI=

Hi,

If you use k-point parallelism, the scaling is almost perfect. Even if
your network is slow because there is very few communications.
When using band/FFT parallelism, the scaling can be quite good, yet it
is very much machine dependent. So benchmark won't transfer well to
your own machine. If you still want some, you can have a look at the
presentation by François Bottin on this page:
http://www.abinit.org/workshop_07/program.html

Best,

PMA

On 9/28/07, brsahu@physics.utexas.edu <brsahu@physics.utexas.edu> wrote:
> Dear Abinit users,
>
> Does any of the users have information about the CPU time versus number of
> processors (ie scaling in an approximate way)for a given system using
> parallel ABINIT code. The ABINIT webpage lists benchmarks for single
> processor runs.
>
> Do I have to take it as a parallel code running on one processor or a
> serial code running on one processor.
>
> Bhagawan
>

--
Pierre-Matthieu Anglade

benchmark calculations for parallel ABINIT, brsahu, 09/28/2007
- Re: [abinit-forum] benchmark calculations for parallel ABINIT, Anglade Pierre-Matthieu, 09/29/2007