[DFTB-Plus-User] Massive parallel calculation

[Gabriele Penazzi]

Hi,

this old post can be useful:

http://www.dftb-plus.info/pipermail/dftb-plus-user/2011/000648.html

With the current version you can only run parallel tasks on shared
memory systems, so no one could try it on a massively parallel system.
I don't know how is the status with the MPI version, but I would guess
that also the parallelization scheme will be different and consequently
the scaling will change.

I would also think that the current scaling is probably limited by blas
scaling (not sure), in that case it will also depend on which blas
library you're linking to.

Gabriele


On 08/08/2012 09:24 AM, SN Ism wrote:
> Hi,
>
> Thank you for the answer.
> I do actually treat big systems with a few cores to, and I have no
> memory issue.
>
> In fact,  I was just wondering if DFTB+ could be used on a massively
> parallel system. I don't think so, because time seem not to scale with
> number of cores over 8 or 12, but maybe someone has more experience
> with this.
>
> S. Nénon
>
>
>
>
> On 07/08/2012 23:49, Amanda.Barnard at csiro.au wrote:
>> Hi,
>>
>> If you want to do hundreds of atom systems, this can be done quite
>> easily on very few cores.  I routinely do >6000 atoms on 12 cores.  
>> Is memory the issue?
>>
>> -Amanda
>>
>> _______________________________
>> Dr. Amanda S. Barnard
>> Leader, Virtual Nanoscience Laboratory
>> CSIRO Materials Science & Engineering
>> E: amanda.barnard[AT]csiro.au
>>
>>
>> -----Original Message-----
>> From: dftb-plus-user-bounces at dftb-plus.info
>> [mailto:dftb-plus-user-bounces at dftb-plus.info] On Behalf Of SN Ism
>> Sent: Tuesday, 7 August 2012 11:56 PM
>> To: dftb-plus-user at dftb-plus.info
>> Subject: [DFTB-Plus-User] Massive parallel calculation
>>
>> Dear DFTB+ users,
>>
>> Do someone have experience or comments on massively parallel DFTB+
>> calculations?
>> Is the gain interesting with hundred atoms systems (eventually with
>> some k points) on hundreds of cores?
>>
>>
>> Best regards
>>
>> S. Nénon
>>
>> _______________________________________________
>> DFTB-Plus-User mailing list
>> DFTB-Plus-User at dftb-plus.info
>> http://www.dftb-plus.info/mailman/listinfo/dftb-plus-user
>> _______________________________________________
>> DFTB-Plus-User mailing list
>> DFTB-Plus-User at dftb-plus.info
>> http://www.dftb-plus.info/mailman/listinfo/dftb-plus-user
>>
>
> _______________________________________________
> DFTB-Plus-User mailing list
> DFTB-Plus-User at dftb-plus.info
> http://www.dftb-plus.info/mailman/listinfo/dftb-plus-user


-- 
Dr. Gabriele Penazzi
BCCMS - University of Bremen

http://www.bccms.uni-bremen.de/
http://sites.google.com/site/gabrielepenazzi/

phone: +49 (0) 421 218 62337
fax: +49 (0) 421 218 62770