Matt, yeah I tried pdb and no clues there. However, I believe I have determined what is going on, and I think it is not a bug. It is trying to write a huge halo mask and it is not hanging, but actually running. This is why it was so hard to determine a hang location. :) thanks for all the advice, sorry for the dumb questions. cheers, On Apr 5, 2010, at 4:50 PM, Matthew Turk wrote:
Actually, Eric, if you don't know where it's hanging, the yt docs have a section on debugging and driving that describes how to instigate a stack printout at any time as well as how to force the code to raise a runtime error, which you can catch with pdb. Try that first. :)
On Apr 5, 2010, at 12:57 PM, Eric Hallman
wrote: I'll try those. Not sure how to tell bisect which version is "good" though. How do I specify versions? How do I know which one worked?
On Apr 5, 2010, at 3:01 PM, Britton Smith wrote:
I'm a bit more confused that it's also happening in serial, but that will make debugging easier. Let us know when you track down the last working version. In fact, if you just want to insert a couple print statements into the write_filtered_list routine in HaloProfiler.py, you might be able to figure out just where it's hanging up.
Britton
On Mon, Apr 5, 2010 at 12:43 PM, Eric Hallman
wrote: Guys, thanks for the help. I was stuck on where to go next. This is breaking in the same way in serial or parallel. Don't know if that helps in diagnosing or not. It is apparently writing the virial_filter.out file.
I will go back to a working version. For sure it was working as recently as about a month ago.
On Apr 5, 2010, at 1:49 PM, Britton Smith wrote:
Hi Eric,
Matt's suggestion would be helpful. If you were running the code successfully at any time recently, it should be simple to track down the revision where it breaks. Does the halo filter list actually get written out? I assume this is only hanging when running in parallel, but correct me if I'm wrong on that. Looking at the light cone and halo profiler code, the function where the hang occurs is called from functions that have various parallel decorators around them, which are probably adding a number of barriers.
Britton
On Mon, Apr 5, 2010 at 11:35 AM, Matthew Turk
wrote: Eric, if you can run "hg bisect" (see the hg book online) to identify the revision this started at, that would be very handy. I would start b going back only a dozen or so. Does this problem occur on smaller datasets on, say your laptop or desktop?
On Apr 5, 2010, at 10:14 AM, Eric Hallman
wrote: I am getting a hang when I run the lightcone tools on ranger, and no debug info is coming out at the hang. this is in the hg version (current) of yt.
The routines I'm running is called test_vel_mask.py, which is creating a lightcone with a halo map that includes the bulk line of sight halo velocity. It includes a slight modification of HaloMask.py to include velocities. However, I removed my changes, and ran again with the same hang. The routine below is copied nearly directly from the yt docs, with minor changes.
Routine is attached as test_wo_vel.py
Hang occurs with text line: yt.lagos INFO 2010-04-05 12:06:15,657 Writing filtered halo list to /scratch/00688/tg458377/box_size_study/50Mpc_1024/ run_12_cl_5D/RD0004/virial_filter.out.
traceback is attached as test_vel_traceback.out
Thanks for any help...
cheers
Eric Hallman Google Voice: (774) 469-0278 hallman13@gmail.com
_______________________________________________ Yt-dev mailing list Yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org _______________________________________________ Yt-dev mailing list Yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
_______________________________________________ Yt-dev mailing list Yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
Eric Hallman Google Voice: (774) 469-0278 hallman13@gmail.com
_______________________________________________ Yt-dev mailing list Yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
_______________________________________________ Yt-dev mailing list Yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
Eric Hallman Google Voice: (774) 469-0278 hallman13@gmail.com
_______________________________________________ Yt-dev mailing list Yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
Yt-dev mailing list Yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org
Eric Hallman Google Voice: (774) 469-0278 hallman13@gmail.com