[yt-users] HaloFinder- parallel is slower?
Cameron Hummels
chummels at astro.columbia.edu
Mon Dec 7 20:59:30 PST 2009
Maybe it's running low on memory in the parallel case where you use all
8 processors on a node. Why don't you try doing it with 7 or 6 ppn and
see if that changes things. Britton suggested experimenting with the
parameters in his previous email, so why not try? You can always ssh to
the node where your job is running and look at the memory consumption to
see if it's running low. Above all, experiment on your own to try to
figure these things out.
Cameron
Agarwal, Shankar wrote:
> Hi,
>
> I ran yt's HaloFinder routine on 1 processor, and then on 8 processors. Both give identical output : 40,000 halos starting from 512^3 particles.
>
> 1 processor : took 1 hour 45 minutes.
> 8 processors : took 2 hours.
>
>
> My script is hop_yt.py....
>
> from yt.mods import *
> fn = "RedshiftOutput0002"
> pf = load(fn)
> halos = HaloFinder(pf,40,padding=0.02)
> halos.write_out("HopAnalysis.out")
>
>
> commands I used....
>
> mpirun -np 1 python hop_yt.py --parallel
> mpirun -np 8 python hop_yt.py --parallel
>
>
> Any idea why 8 cpus would take longer ?
>
>
> shankar
> _______________________________________________
> yt-users mailing list
> yt-users at lists.spacepope.org
> http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org
>
>
--
Cameron Hummels
PhD Candidate, Astronomy Department of Columbia University
Public Outreach Director, Astronomy Department of Columbia University
NASA IYA New York State Student Ambassador
http://outreach.astro.columbia.edu
More information about the yt-users
mailing list