Re: NAMD/MIC & CUDA on Stampede

From: Maxim Belkin (mbelkin_at_ks.uiuc.edu)
Date: Thu Jul 17 2014 - 13:32:27 CDT

Just to give an update to everyone, the problem was solved by specifying "#SBATCH -n 1"

Maxim

On Jul 16, 2014, at 1:28 PM, Jeffery Klauda <jbklauda_at_umd.edu> wrote:

> I would like to run NAMD optimized for MIC and CUDA on Stampede. I have been able to use the provided scripts in the following directory:
>
> /home1/00288/tg455591/NAMD_scripts
>
> For example, runbatch_latest_mic runs fine. The issue is I would like to use my own queue submission script and have not been able to get it to work. I have tried scripts in CSH and BASH with no luck. I am confused why runbatch_latest_mic works but when I run the script below it outputs the following error:
>
> TACC: Starting up job 3694819
> TACC: Setting up parallel environment for MVAPICH2+mpispawn.
> TACC: Starting parallel tasks...
> Charmrun: Bad initnode data length. Aborting
> Charmrun> IBVERBS version of charmrun
>
>
> **Submission Script**
> #!/bin/csh
> #SBATCH -J emre16 # Job name
> #SBATCH -o emre.o%j # output and error file name (%j expands to jobID)
> #SBATCH -N 1 # total number of nodes
> #SBATCH -n 16 # total number of mpi tasks requested
> #SBATCH -p normal-mic # queue (partition) -- normal, development, etc.
> #SBATCH -t 02:00:00 # run time (hh:mm:ss)
>
> module load fftw3
>
> # Set variables per Schulten Lab /home1/00288/tg455591/NAMD_scripts
> set BINDIR = /work/00288/tg455591/NAMD_build.latest/NAMD_2.9_Linux-x86_64-ibverbs-smp-Stampede-MIC
>
> set SCRIPTDIR = /home1/00288/tg455591/NAMD_scripts
>
> @ REALNUMPROCS = 1 * 15
> @ NUMTASKS = 1
>
> echo Running on host `hostname`
> echo Time is `date`
> echo Directory is `pwd`
>
> # Execution of the script
>
> $BINDIR/charmrun +p$REALNUMPROCS ++ppn 15 ++mpiexec ++remote-shell $SCRIPTDIR/mpiexec $BINDIR/namd2 +commap 0 +pemap 1-15 +devices 0 dyn.inp >& dyn2.out
>
> Jeff
>
> --
> *****************************************************
> Jeffery B. Klauda
> Associate Professor
> Department of Chemical and Biomolecular Engineering
> 2113 Building 90
> University of Maryland
> College Park, MD 20742
> Phone: (301) 405-1320
> Fax: (301) 314-9126
> Office: 1227A Chemical & Nuclear Engineering Bldg.
> Web: terpconnect.umd.edu/~jbklauda
> *****************************************************

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:37 CST