From: Jesper Sørensen (jesorensen_at_ucsd.edu)
Date: Tue Jan 24 2012 - 14:35:39 CST
Thanks George. That works.
I'm still curious about all the specialized flags set on the other version on the cluster, you'd think if someone went to all that trouble, i'd be worth it...
Best,
Jesper
On Jan 24, 2012, at 10:22 AM, George M Giambasu wrote:
> Here is what I use to run namd2.7 on ranger. I have not observed any
> performance issues with this script.
>
> George
>
> #$ -S /bin/bash
> #$ -A XXXXXXXXX # replace with your TG account #
> #$ -V # pass current environment variables
> #$ -cwd # script will run from the directory that the job was
> submitted from
> #$ -N NAMD
> #$ -pe 16way 512 # 16 cores per node, 4 nodes, 64 cores total
> #$ -q normal
> #$ -l h_rt=24:00:00 # 1 hour walltime
>
> module unload mvapich
> module swap pgi intel/10.1
> module load mvapich-old/1.0.1
> module load fftw2
>
> export VIADEV_SMP_EAGERSIZE=64
> export VIADEV_SMPI_LENGTH_QUEUE=256
> export VIADEV_RENDEZVOUS_THRESHOLD=50000
> export VIADEV_ADAPTIVE_RDMA_LIMIT=0
>
> ibrun tacc_affinity /path/to/namd2 md2_prod_2.inp > eq_9.inp.log
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2012 - 23:21:09 CST