[Yt-svn] New issue 320 in yt: MPI error in parallel halo profiling

[Bitbucket]

--- you can reply above this line ---

New issue 320: MPI error in parallel halo profiling
https://bitbucket.org/yt_analysis/yt/issue/320/mpi-error-in-parallel-halo-profiling

mqk on Tue, 8 Nov 2011 21:31:32 +0100:

Description:
  There's a bug in parallel halo profiling, specifically in the make_profiles() part of HaloProfiler.

When I run make_profiles() like so

{{{
hp = HaloProfiler(dir + fn, halo_list_file=hop_fn)

hp.add_halo_filter(VirialFilter,
   must_be_virialized=True,
   overdensity_field='ActualOverdensity', 
   virial_overdensity=200,
   virial_filters=[['TotalMassMsun','>=','1e8']],
   virial_quantities=['TotalMassMsun','RadiusMpc'])

hp.make_profiles(filename="VirialHaloes.out",
                 format="%12.7e")
}}}

in parallel (-np 8 in this instance), I get an MPI error. This is on yt-2.3dev tip (f376abaf6942). 

Here's the traceback: http://paste.yt-project.org/show/1927/

Use of 'hg bisect' (what an excellent tool btw!) reveals:

{{{
The first bad revision is:
changeset:   4723:b98516052e9a
branch:      yt
user:        Matthew Turk <matthewturk at gmail.com>
date:        Tue Oct 18 10:55:34 2011 -0400
summary:     Consolidating "_mpi_concatenate_array_*" operations into the _mpi_catarray
}}}

No idea how to proceed.

Mike



--

This is an issue notification from bitbucket.org. You are receiving
this either because you are the owner of the issue, or you are
following the issue.