How to Run Gaussian94 (g94) on the
IBM-SP

IBM-SP:

Introduction

Gaussian can be run in both serial and parallel, via shared memory, on the IBM SP.

The following sections demonstrate how to run Gaussian94 in parallel and serial in both interactive and batch modes.

The scripts rg94 and qg94 described below were created by the user support staff to make running Gaussian easier. These scripts are not part of the Gaussian distribution.

Available Version

The D.2 revsion of Gaussian94 is the only version of Gaussian94 available on the IBM-SP. It is available by loading the g94 module.

module loadd g94

Setting the Scratch Directory

The scratch directory where Gaussian stores temporary files can be specified in two different ways.

Set the environment variable GAUSS_SCRDIR in the .cshrc file or in the LoadLeveler script. This is done by adding the following line to either file:
where scr_directory is the directory you want Gaussian to use for the temporary files.
Use the rg94 and qg94 scripts (described below) to set the scratch directory. If the GAUSS_SCRDIR environment variable has not been set, then the script will create a temporary directory in the scratch space, scratch1 by default, for Gaussian to use for the temporary files. The temporary directory will be removed when the calculation has finished.
If the GAUSS_SCRDIR environment variable has been set, then the scripts will not create a temporary file in the scratch space.

Note: Setting the GAUSS_SCRDIR environment variable is particularly useful when one or both of the scratch directories are unavailable.

Interactive running of g94

4.A. Running g94 in serial:

load the g94 module
launch the executable
where input_file is the Gaussian input file. If the GAUSS_SCRDIR environment variable has been set, Gaussian will use the designated directory for the temporary files. If it has not been set the rg94 script creates a temporary directory in /scratch1 that is used by Gaussian to store the temporary files. This temporary directory is removed when the calculation finishes. The Gaussian output file, input_file.out, will be placed in the same directory as the input_file.
Example:

4.B. Running g94 in parallel:

load the g94 module
add the line
to the Gaussian input file. This line tells Gaussian that it can use up to 2 processors. Currently, 4 processors is the limit on the SP when running interactively. This is because the winterhawk+ nodes (the interactive nodes) have 4 processors per node, therefore the maximum number of processors that can be used with shared memory is 4.
launch the executable
where input_file is the Gaussian input file. If the GAUSS_SCRDIR environment variable has been set, Gaussian will use the designated directory for the temporary files. If it has not been set the rg94 script creates a temporary directory in /scratch1 that is used by Gaussian to store the temporary files. This temporary directory is removed when the calculation finishes. The Gaussian output file, input_file.out, will be placed in the same directory as the input_file.
Please see the webpage, http://www.msi.umn.edu/sp/info/policies/#bgjobs for details about the policies and time limits for interactive calculations on the IBM SP.
All interactive jobs are run on the winterhawk nodes+.
Example:

Batch running of g94:

5.A. Running g94 in serial

load the g94 module
launch the executeable
where input_file is the Gaussian input file and the options for the qg94 script are as follows: (default values in brackets)
```
  -h prints the help menu
  -e [never] email notification (options: end or never)
  -n [winter] type of node (options: winter or night)
  -p [1] total number of processors (max of 4 proc)
  -r [yes] submit the job? (options: yes or no)
  -s [/scratch1/user] scratch space (options: scratch1 or scratch2)
  -t [5:00:00] the amount of wall clock time (format: hh:mm:ss)
```
The qg94 script creates the LoadLeveler file input_file.ll, creates a temporary directory for the Gaussian temporary files (if GAUSS_SCRDIR has not been set), and submits the job to the queue. If the -r flag has been set to "no" then the job will not be submitted; however the LoadLeveler script and the temporary directory (if GAUSS_SCRDIR has not been set) will still be created. The temporary directory is removed when the calculation has finished.
Examples:
1. Print out the options menu for qg94
```
qg94 -h
```
Run the Gaussian job, h2o using the defaults; run serially on the winterhawk+ nodes with a maximum wall clock time of 5 hours and place the temporary files in /scratch1.
The output file, h2o.out is created in the current working directory with the input file. The LoadLeveler script, h2o.ll is also placed in the same directory as the input file.
Run the Gaussian calculation, h2o, serially on the nighthawk nodes with maximum wall clock time of 12 hours and place the temporary files in /scratch2.
Create the LoadLeveler script, but do not submit the job to the queue.
The job, h2o.ll, can be run by typing
This command is echoed in the h2o.ll file.

5.B. Running g94 in parallel

load the g94 module
launch the executeable
where input_file is the Gaussian input file and the options for the qg94 scripts are given above. When running in parallel the -p option MUST be set to n > 1, where n is the number of processors.
Notes:
1. Don't forget to add the
  line to the Gaussian input file, where n is the number of processors to be used by Gaussian.
2. The number of processors specified in the Gaussian input file must be the same as the number of processors specified with the -p flag.
3. The maximum number of parallel processors available at this time (for Gaussian calculations) on the winterhawk+ and on the nighthawk nodes is 4. This is because bothe the winterhawk+ and nighthawk nodes have 4 processors per node.
Examples:
1. Run the Gaussian job, h2o-p, in parallel on the winterhawk+ nodes with the default wall clock time and default scratch space.
  or
  where h2o-p is a parallel Gaussian input file.
2. Run the Gaussian job, h2o-p, in parallel on the nighthawk nodes with 2 processors, 12 hours of wall clock time, and placing the temporary files in /scratch2.
3. Run the Gaussian job, h2o-p, in parallel with 4 processors, 24 hours of wall clock time, and placing the temporary files in /scratch1.
  Remember the maximum number of parallel processors available at this time (for Gaussian calculations) on the winterhawk+ nodes and on the nighthawk nodes is 4.

Please direct questions or problems to help@msi.umn.edu
The University of Minnesota is an equal opportunity educator and employer.

This information is available in alternative formats upon request by individuals with disabilities. Please send email to alt-format@msi.umn.edu or call 612-624-0528.

	URL: http://www.msi.umn.edu /software/gaussian94/content.html
This page last modified on Thursday, 22-May-2008 16:00:15 CDT
Please direct questions or problems to help@msi.umn.edu
Website related questions or problems should be directed to webmaster@msi.umn.edu The Supercomputing Institute does not collect personal information on visitors to our website. For the University of Minnesota policy, see www.privacy.umn.edu.

How to Run Gaussian94 (g94) on the IBM-SP