From: Chitrak Gupta (chgupta_at_mix.wvu.edu)
Date: Sun Oct 01 2017 - 19:21:52 CDT
Hi Sebastian,
I believe you would be able to specify the number of nodes and number of
procs by the "SBATCH -N" and "SBATCH -n" options respectively. So if you
want to use 8 cores and 67 procs per node, you can use the following
options in your submission script
#SBATCH -N 8
#SBATCH -n 536
And then run the script with the following line:
ibrun namd2 RUN.namd > output.log
I *think* this should work but cannot confirm because I haven't used the
default NAMD on Stampede2.
Hope this helps,
Chitrak.
On Fri, Sep 29, 2017 at 3:59 PM, The Cromicus Productions <
thecromicusproductions_at_gmail.com> wrote:
> Hi everyone,
>
> I'm trying to run some tests on Stampede2, however I'm really confused
> with using
> "ibrun" (I usually just use mpirun). Does anyone have experience on this?
> For example,
> if I want to use 8 nodes and 68 cores per node (total 544 cores), I'm
> trying
>
> ibrun -np 544 namd2 ++p 544 ++ppn 67 RUN.namd > output.log
>
> But run into errors such as
> "CmiStartThreads: Resource temporarily unavailable(11)".
>
> I tried other combinations and in the best case scenario I managed to do a
> VERY slow
> run. How can I fix this to use A nodes and B cores per node, in general?
> Found some information at https://www.nvidia.com/en-us/data-center/gpu-
> accelerated-applications/namd/
> but I don't have that much experience with parallel computing.
>
> Thank you very much,
>
> Sebastian
>
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:20:37 CST