|
|
NWChem on Redwood at MCSR
NWChem 4.5 and 4.6 are installed on redwood. Each version was built atop MPT (SGI's MPI implementation.)
Steps for running an NWChem job on redwood:
- If you don't have a redwood account, apply for one via the MCSR Online Account Manager.
- Ensure that you have a file called .nwchemrc in your $HOME directory on redwood. The contents of this file will depend on whether you are running NWChem 4.5, or NWChem 4.6. You may copy or link to our tested versions of these files:
cp /usr/local/appl/nwchemrc.4.5 $HOME/.nwchemrc
OR
ln /usr/local/appl/nwchemrc.4.5 $HOME/.nwchemrc
cp /usr/local/appl/nwchemrc.4.6 $HOME/.nwchemrc
OR
ln /usr/local/appl/nwchemrc.4.6 $HOME/.nwchemrc
- Create a PBS script to submit your job. See PBS at MCSR. Your script must invoke the respective nwchem executable using mpirun:
mpirun -np NPROCS /usr/local/appl/nwchem-4.5/bin/LINUX64/nwchem YOURINPUTFILE.nw
OR
mpirun -np NPROCS /usr/local/appl/nwchem-4.6/bin/LINUX64/nwchem YOURINPUTFILE.nw
The following example PBS file requests 2 processors, 1GB memory, and 100 hours of total CPU time. Notice that it changes directory to a subdirectory called "nwchem" under the user's /ptmp account, where there exists a NWCHEM input file called myfile.nw, then invokes the NWChem 4.5 executable as 2 processes via mpirun. Note that you must specify the number of processors in two places.
#PBS -S /bin/csh
#PBS -l ncpus=2
#PBS -l mem=1GB
#PBS -l cput=100:00:00
cd /ptmp/$USER/nwchem/
mpirun -np 2 /usr/local/appl/nwchem-4.5/bin/LINUX64/nwchem myfile.nw
Please only use redwood for those high performance jobs that will not run on sweetgum or mimosa, or that will not run efficiently on those systems. If you would like access to larger queues, we will ask for a demonstration that your job can achieve good speedups on the numbers of processors you are currently using. If you have an NWChem job that you've run on redwood and also either sweetgum or mimosa, we'd be interested in knowing the relative computation times of those runs, including the number of processors, the memory requested/used, and the version of NWChem being used.
To report performance information, or if you experience difficulty running your job, please email assist@mcsr.olemiss.edu, and let us know
the version of NWChem you are using, the location of your PBS script, and the error or problem you
are having.
Memory Issues Running NWChem on Redwood
If your job requires much shared memory space, and not enough is available, you may see an error message like the following in your output file. If so, this may be because other NWChem jobs have already been allocated most or all of the shared memory that is available for the entire Altix system. In this case, you can try your job later. As of November 22, 2004, the Maximum Shared Memory available system wide is 32GB.
The application attempted to allocate a shared memory segment of XXXXXXXXX bytes in size. This might be in addition to segments that were allocated succesfully previously. The current system configuration does not allow enough shared memory to be allocated to the application.
This is most often caused by:
1) system parameter SHMMAX (largest shared memory segment) being too small or
2) insufficient swap space.
Please ask your system administrator to verify if SHMMAX matches the amount of memory needed by your application and the system has sufficient amount of swap space. Most UNIX systems can be easily reconfigured to allow larger shared memory segments,
see http://www.emsl.pnl.gov:2080/docs/global/support.html
In some cases, the problem might be caused by insufficient swap space.
*******************************************************
0:allocate: failed to create shared region : -1
|
