Re: domain decomposition error, exit codes: 137

From: Luca Tornatore <tornator_at_oats.inaf.it>
Date: Wed, 23 Jul 2008 19:53:20 +0200

Hi Sami,

you may try to lower the parameter TOPNODEFACTOR that you find at the
begin of domain.c.
Basically the particles are grouped in little chunks that actually
represents the
atomic units of the domain decomposition.
The size of those chunks is determined by TOPNODEFACTOR * NTask: i.e.,
once TOPNODEFACTOR is set, the chunk size is lower as NTask grows.
As a consequence a larger number of topnodes are created and you may run
out of memory (no messages about something similar?).

Then, you have two choices: either you decrease TOPNODEFACTOR when
NTask increase (a factor of two should work) or you increment MAXTOPNODES
in allvars.h.

All this is valid just in the case you run out of memory for the reasons
stated
above.

Just try, is takes a second

cheers,
luca

>
> Dear Gadget-2 Users,
>
> I ran into a domain decomposition error while trying to run Gadget-2
> with 1024 cores. Here's a copy what the log says (shortened): ..
> Total number of particles : 0134217728
>
> allocated 0.0762939 Mbyte for ngb search.
>
> Allocated 16.6029 MByte for BH-tree. 64
>
> domain decomposition... _pmii_daemon(SIGCHLD): PE 996 exit signal
> Killed
>
> .. (goes on and on with all PEs)
>
> _pmii_daemon(SIGCHLD): PE 888 exit signal Killed Application 57082
> exit codes: 137 Application 57082 exit signals: Killed
>
> I have run the same IC file with 512 and 64 cores without any
> problems so the problem must be related to the number of PEs.
> However, I did not change any Makefile or parameter file options when
> I moved from 512 (or 64) to 1024 cores, so everything in these files
> should be ok.
>
> The IC file I'm running contains 512**3 particles in a 20/h Mpc box
> on side, but I doubt it matters.
>
>
> Thanks in advance, Sami
>
>
>
> _______________________ Sami-Matias Niemi Student Support Astronomer
> Nordic Optical Telescope
>
> sami_at_not.iac.es +34 662 535 441 +34 922 425 424
>
> http://users.utu.fi/saniem/ http://saniem.deviantart.com
> ----------------------------------------
>
>
>
>
> ----------------------------------------------------------- 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
Received on 2008-07-23 19:53:22

This archive was generated by hypermail 2.3.0 : 2022-09-01 14:03:42 CEST