The ABINIT Users Mailing List ( CLOSED )

[Gilles Zerah <gilles.zerah@cea.fr>]

Hi Nichols,

It is not for sure in 4.6.4 and will be included in a merge soon to come, as soon as possible
in a "for developers" version.
This is a MPI //lelisation (in which bands are distributed among processors)

Gilles

Hi Gilles,

One clarification, when you say parallelized do you mean with MPI or OpenMP..

Thanks,
On 10/20/05, Nichols A. Romero <naromero@gmail.com> wrote:
  

Hi Gilles,

Thanks for writing back. So, wfoptalg=4 is NOT parallelized in ABINIT 4..6.4?

Do I have to wait for version 4.6.5 or 5.x.x?

I ask because I am somewhat pressed to make progress on a project.

Bests,
On 10/20/05, Gilles Zerah <gilles.zerah@cea.fr> wrote:
    

Hi,
If you can wait the coming release, the (blocked)lobpcg algorithm
(wfoptalg=4) has been parallelized
in the norm conserving case (the paw version should come next).
It can be used for the number of procs you mention.
Gilles Zerah

      

Hi,

I have a question about the blocked conjugate gradient algorithms
(wfoptalg=1,11) in ABINIT 4.6.4.

Was this intended to be used with a large number (16-32) of
processors? Will this scale well on large systems?

Bests,
--
Nichols A. Romero, PhD.
1613 Denise Dr. Apt. D
Forest Hill, MD 21050
443-567-8328 (C)
410-306-0709 (O)




        

--
Nichols A. Romero, PhD.
1613 Denise Dr. Apt. D
Forest Hill, MD 21050
443-567-8328 (C)
410-306-0709 (O)

    

--
Nichols A. Romero, PhD.
1613 Denise Dr. Apt. D
Forest Hill, MD 21050
443-567-8328 (C)
410-306-0709 (O)