The ABINIT Users Mailing List ( CLOSED )

[Xenophon Krokidis <Xenophon.Krokidis@scienomics.com>]

Title: Signature text

Dear Francois,

please find my comments in context.

Best regards,

Xenophon

Xenophon Krokidis, PhD
Scienomics
T: + 33 (0)1 40 07 56 60
F: + 33 (0)1 71 19 75 83
M: + 33 (0)6 76 68 06 47

BOTTIN Francois wrote:

4A6DCF8C.7070001@cea.fr" type="cite">Dear Xenophon,

Thanks a lot for your report.

Xenophon Krokidis a écrit :

Dear François,

let me summarize  the situation. I want to perform PAW type of calculation at one k-point (0.5 0.5 0.5). Below I list the trials I did and the observations.

    * PAW
          o Parallel (wfoptalg 1; nbdblock 6): ABINIT stays idle.
          o Paralle (wfoptalg 1; npband 6): ABINIT runs fine

The npband keyword has no meaning for wfoptalg=1. I suspect you are always in sequential.

I don't think so if understand well what is said here : http://www.abinit.org/documentation/helpfiles/for-v5.8/input_variables/vardev.html#nbdblock
if wfoptalg=1, nbdblock defines the number of bands (or states) in a block, and, by the same token, the maximal number of processors that can be used for that type of parallelism.

4A6DCF8C.7070001@cea.fr" type="cite">

          o Parallel (wfoptalg 14; npband 6; paral_kgb 1, ... ):
            ABINIT runs fine
          o Sequential : ABINIT runs fine.
    * Norm-concerving pseudopotentials.
          o Parallel (wfoptalg 1; nbdblock 6): ABINIT runs fine.
          o Sequential : ABINIT runs fine

What is your keywords in these sequential cases? wfoptalg=1 or 14? And nbdblock=1 or 6?

The option wfoptalg=14 is not considered since I am not running at Gamma point.
In the sequential calculation I have nbdblock=1

That's the question of the nota bene in my previous mail.
The calculations using wfoptalg=1 and nbdblock=/1 have to work, even in sequential mode?

So it seems that there is a problem when parallelization over bands is defined via the nbdblock keyword.

4A6DCF8C.7070001@cea.fr" type="cite">Yes, there is probably a problem for the parallelization over bands using wfoptalg=1 and nbdblock=/1.

I think so but please note that this is the case only when PAW is used !

4A6DCF8C.7070001@cea.fr" type="cite">But to be sure, would you verify your calculations work in sequential using wfoptalg=1 and nbdblock=6, please?

So clearly it doesn't work. Please let me know if you need more information.

4A6DCF8C.7070001@cea.fr" type="cite">

Xenophon
Xenophon Krokidis, PhD
*Scienomics* <www.scienomics.com>
T: + 33 (0)1 40 07 56 60
F: + 33 (0)1 71 19 75 83
M: + 33 (0)6 76 68 06 47

Regards,
Francois

begin:vcard
fn:Xenophon Krokidis
n:Krokidis;Xenophon
org:Scienomics
adr:;;17 square Edouard VII;Paris;;75009;France
email;internet:Xenophon.Krokidis@Scienomics.com
title:Managing Director
tel;work:+33 (0)1 53 43 51 05
tel;fax:+ 33 (0)1 53 43 92 92
tel;cell:+33 (0)6 76 68 06 47
version:2.1
end:vcard