[yt-users] Problem with Rockstar in yt

Brian O'Shea bwoshea at gmail.com
Sun Aug 4 18:20:02 PDT 2013


Hi Daegone,

Sorry, I think my last email was imprecise: after that first email, I am
now running in parallel on 4 processors, using the command line:

mpirun -np 4 python ./test_rockstar.py --parallel

Rockstar gives a similar error both running on a single core (with mpirun)
and in parallel.  :-(

--Brian



On Sun, Aug 4, 2013 at 8:57 PM, Daegene Koh <dkoh30 at gatech.edu> wrote:

> Hi Brian,
>
> If I recall correctly, I believe you need at least 3 processors to get
> Rockstar running properly. And it must always be run in parallel.
> I'm afraid I don't know much much more beyond that.
>
> Cheers,
> DK
>
> On Aug 4, 2013, at 7:54 PM, "Brian O'Shea" <bwoshea at gmail.com> wrote:
>
> Hi folks,
>
> Just a quick followup: I've now run this on a different machine and
> verified that the reason I couldn't use Rockstar in parallel was due to a
> MPI problem.  *HOWEVER*, I still have the same Rockstar error that I was
> reporting before when I was using one core (all using the same script as in
> my previous email):
>
> http://paste.yt-project.org/show/3747/
>
> So, it does seem like there is a bug, or I am simply using Rockstar
> incorrectly...
>
> --Brian
>
>
>
> On Sun, Aug 4, 2013 at 6:26 PM, Brian O'Shea <bwoshea at gmail.com> wrote:
>
>> Dear yt-users,
>>
>> I'm trying to use the Rockstar halo finder within yt, and am encountering
>> some odd problems.  I'm using the main development tree (
>> https://bitbucket.org/yt_analysis/yt) with changeset f936432ed45d, and
>> attempting to use Rockstar to find halos on a small server running Ubuntu
>> 12.04.  When I call this script:
>>
>> ---- file test_rockstar.py ----
>>
>> from yt.mods import *
>> from yt.analysis_modules.halo_finding.rockstar.api import
>> RockstarHaloFinder
>>
>> pf = load("DD0057/data0057")
>> rh = RockstarHaloFinder(pf)
>>
>> ---
>>
>> using the command line sequence "mpirun -np 2 python ./test_rockstar.py
>> --parallel", I seem to be getting a seg fault (as can be seen at
>> http://paste.yt-project.org/show/3745/).  However, if I use a single
>> processor (with "mpirun -np 1 python ./test_rockstar.py --parallel"), I get
>> a very different error:
>>
>> Traceback (most recent call last):
>>   File "./test_rockstar.py", line 6, in <module>
>>     rh = RockstarHaloFinder(pf)
>>   File
>> "/data/bwoshea/galparttest/yt-x86_64/src/yt-hg/yt/analysis_modules/halo_finding/rockstar/rockstar.py",
>> line 230, in __init__
>>     self.pool, self.workgroup = self.runner.setup_pool()
>>   File
>> "/data/bwoshea/galparttest/yt-x86_64/src/yt-hg/yt/analysis_modules/halo_finding/rockstar/rockstar.py",
>> line 112, in setup_pool
>>     (self.num_writers, "writers") ]
>>   File
>> "/data/bwoshea/galparttest/yt-x86_64/src/yt-hg/yt/utilities/parallel_tools/parallel_analysis_interface.py",
>> line 335, in from_sizes
>>     pool.add_workgroup(size, name = name)
>>   File
>> "/data/bwoshea/galparttest/yt-x86_64/src/yt-hg/yt/utilities/parallel_tools/parallel_analysis_interface.py",
>> line 303, in add_workgroup
>>     group = self.comm.comm.Get_group().Incl(ranks)
>> AttributeError: 'NoneType' object has no attribute 'Get_group'
>>
>> I'm puzzled about the error, since I can use other parallel yt scripts
>> without any problems (a 4-processor script making projections works just
>> fine).  This machine doesn't have infiniband (as warned about at
>> http://yt-project.org/doc/analysis_modules/running_halofinder.html#rockstar-halo-finding),
>> and both FOF and Hop find several hundred halos with my dataset.  I'd just
>> use another halo finder, but I'm trying to do something that requires
>> Rockstar to be run inline with Enzo, so I'm stuck with it...
>>
>> Does anybody have any idea what might be going on?
>>
>> Thanks!
>>
>> Brian
>>
>>
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>
>
>
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.spacepope.org/pipermail/yt-users-spacepope.org/attachments/20130804/f0d3199a/attachment.html>


More information about the yt-users mailing list