[bgl-discuss] RM bug or user error?

[Chad Glendenin <chad@xxxxxxxxxxx>]

Is there a requirement that the number of nodes be a power of 2 or
anything like that?

I'm running a scalability study with a simple Poisson solver. It runs
either with a linear (1xP = P_total) or a square (PxP = P_total)
processor topology, so I'm trying to do each run with a perfect-square
number of CPUs. I have no problems running it on 1, 4, 9, 16, or 25 nodes,
but with 36 nodes, it gets stuck in the queue forever and never runs.

Am I doing something wrong with qsub?

> qsub -t 10 -c 36 -n 36 -C /home/chad/log /home/chad/cs340/bgl/1d/block
36000 1000 0 0 0.1 4

I'm trying to get a 1-to-1 relationship between processes and CPUs on
each test run.

Thanks,
ccg

P.S. It looks like a qsub request for 49 nodes just got stuck too.

- --------------------------------------------------------------------
To add or remove yourself from this mailing list, use the 'notifyme'
command on any BGL machine. To remove: notifyme -n, to add: notifyme -y.