Every single processor made it to at least 184? You have this *for sure*?
What happens if you barrier after every sphere instantiation, to
ensure they are sync'd up? Print at the barrier too so we know that
they *are* synced up?
On Wed, Mar 4, 2009 at 12:22 PM, Stephen Skory
My resident size is 1.5gb per process, and it's not increasing at all. I made some modifications to your script, but none that were a big deal:
http://paste.enzotools.org/show/73
It's just going and going over here. Up to sample ~1500, on two procs on the interactive node. How many did it crash after on yours?
On all processors it did at least 184 samples, then it crashed. Some did many more.
Maybe the Ranger MPI stack is crap. I was going to test my stuff on Kraken XT5 today, but it's down, as you probably know.
I just tried with 2 procs, it went 197.
_______________________________________________________ sskory@physics.ucsd.edu o__ Stephen Skory http://physics.ucsd.edu/~sskory/ _.>/ _Graduate Student ________________________________(_)_\(_)_______________ _______________________________________________ Yt-dev mailing list Yt-dev@lists.spacepope.org http://lists.spacepope.org/listinfo.cgi/yt-dev-spacepope.org