Re: work-load imbalance (again, but for SPH calculation)

From: Yves Revaz <yves.revaz_at_epfl.ch>
Date: Wed, 30 Jun 2010 10:39:01 +0200

Hi Yanbin,

Did you first try to change

 if(TopNodes[sub].Count > All.TotNumPart / (TOPNODEFACTOR * NTask * NTask))
to
 if(TopNodes[sub].Count > All.TotNumPart / (TOPNODEFACTOR * NTask))

as proposed by Volker ?

http://www.mpa-garching.mpg.de/gadget/gadget-list/

see : Domain decomposition takes too much time


Cheers,

yves

> I feel the imbalance is linked to the domain decomposition. Because
> during the merger, the gas component may be decoupled from baryon
> strongly, while the domain decomposition is a global process ignoring
> the type of particles (I don't know if I'm correct), Thus when N_CPU
> increases, more regions are created. Some CPUs may have very large
> number of gas particles, some may have few gas particles.
> One thing I should tell you is that at the early stage of the simulation,
> the work-load looks nice, the imbalance become significant after
> the interaction begins.
>
>
> Am I right? If I'm, is that possible to improve that? Any comments are
> welcome.
>
> Thanks in advance.
> Yanbin.
>
>
>
>
> -----------------------------------------------------------
>
> If you wish to unsubscribe from this mailing, send mail to
> minimalist_at_MPA-Garching.MPG.de with a subject of: unsubscribe gadget-list
> A web-archive of this mailing list is available here:
> http://www.mpa-garching.mpg.de/gadget/gadget-list
>


-- 
                                                 (o o)
--------------------------------------------oOO--(_)--OOo-------
  Dr. Yves Revaz
  Laboratory of Astrophysics EPFL
  Observatoire de Sauverny     Tel : ++ 41 22 379 24 28
  51. Ch. des Maillettes       Fax : ++ 41 22 379 22 05
  1290 Sauverny             e-mail : Yves.Revaz_at_epfl.ch
  SWITZERLAND                  Web : http://www.lunix.ch/revaz/
----------------------------------------------------------------
Received on 2010-06-30 10:39:01

This archive was generated by hypermail 2.3.0 : 2023-01-10 10:01:31 CET