Difference between revisions of "HowTo:gamess"
(→Location of the program and setup) |
(→Submitting (parallel) Gamess jobs) |
||
(One intermediate revision by the same user not shown) | |||
Line 57: | Line 57: | ||
== Submitting (parallel) Gamess jobs == | == Submitting (parallel) Gamess jobs == | ||
− | Gamess has to be run via | + | Gamess has to be run via SLURM, which is a load-balancing program that submits batch jobs to low-load processors on the cluster cluster. [[SLURM|To learn more about this program, click here]]. A Gamess job is submitted to the Grid Engine in the form of an execution script. A reasonable execution script for Gamess looks like this: |
<pre> | <pre> | ||
− | # | + | #!/bin/bash |
− | # | + | #SBATCH --output=STD.out |
− | # | + | #SBATCH --eror=STD.err |
− | # | + | #SBATCH --nodes=1 |
− | # | + | #SBATCH --ntasks=1 |
− | # | + | #SBATCH --cpus-per-task=8 # Number of CPUs |
− | # | + | #SBATCH --mem-per-cpu=4G # memory per CPU in MB |
− | # | + | #SBATCH --time=0-00:30 # time (DD-HH:MM) |
− | + | ||
− | + | ||
− | + | export SLURM_CPUS_PER_TASK | |
+ | |||
+ | module load gamess-us/20170420-R1 | ||
+ | rungms name.inp &> name.out | ||
+ | </pre> | ||
− | + | The --output and --error lines define the standard output and standard error files, respectively. | |
− | You also need to specify the name of the input file just like in an interactive run | + | Note that all lines starting with "#SBATCH" are directives for SLURM, and will be interpreted when the script is submitted to that program. You also need to specify the name of the input file just like in an interactive run. |
− | The number of processes is specified in the " | + | The number of processes is specified in the "--cpu-per-task" line, which instructs SLURM to allocate the proper number of CPUs for you on a single node. The --nodes and --ntasks lines should be kept at 1. You do not have to specify it separately in the rungms command line, because SLURM passes the $CPUS_PER_TASK variable. |
Assuming your Grid Engine script is called "gamess.sh", it is submitted to GridEngine by typing | Assuming your Grid Engine script is called "gamess.sh", it is submitted to GridEngine by typing | ||
− | <pre> | + | <pre>sbatch gamess.sh</pre> |
No further specification of the output is necessary, since this is done inside the script and handled by GridEngine. | No further specification of the output is necessary, since this is done inside the script and handled by GridEngine. | ||
|} | |} |
Latest revision as of 17:46, 13 March 2018
Contents
Gamess (US)
This is an introduction to the usage of the electronic-structure code "Gamess" on our clusters. It is meant as an initial pointer to more detailed information, and to get started. It does not replace study of the manual.
FeaturesThe "General Atomic and Molecular Electronic Structure System" (GAMESS) is a quantum chemistry software package that was assembled from various older codes (in particular, HONDO) by M. Dupuis, D. Spangler, and J. J. Wendoloski of the National Resources for Computations in Chemistry (NRCC). The code has undergone great changes since then and is maintained now by the Gordon Research Group at Iowa State University. Gamess performs a great many of standard quantum chemical calculations. These include:
For a complete list of capabilities of GAMESS, consult this table or look at the Gamess documentation. Gamess is described in General Atomic and Molecular Electronic Structure System; M.W.Schmidt, K.K.Baldridge, J.A.Boatz, S.T.Elbert, M.S.Gordon, J.H.Jensen, S.Koseki, N.Matsunaga, K.A.Nguyen, S.Su, T.L.Windus, M.Dupuis, J.A.Montgomery; J. Comput. Chem. 14, 1347-1363 (1993) and Advances in electronic structure theory: GAMESS a decade later; M.S.Gordon, M.W.Schmidt pp. 1167-1189 in Theory and Applications of Computational Chemistry: the first forty years; C.E.Dykstra, G.Frenking, K.S.Kim, G.E.Scuseria (editors), Elsevier, Amsterdam, 2005. Location of the program and setupThe program resides on the so-called CVMFS stack (provided by Compute Canada, contains most of our application software). You also find some test examples in the program directory, which are useful to get an idea of the input format for the program. You are not allowed to copy the executable or any part of the distribution onto your local machine. However you can easily obtain the program yourself. See the GAMESS source code distribution page. Setup is done via the standard module load: module load gamess-us Scratch filesWhen you run Gamess a scratch space directory is set to /scratch/$USER and all temporary files and intermediate output will be placed in that directory. The program also requires a local temporary directory right below your home directory called $USER/scr. Make sure that you move files that you want to keep from there before running Gamess again with the same case_name. The second run requires that some temporary files be removed before re-run and will fail if they are still present in the scratch directory. For instance, if your username is "hpc9876", you will need "/scratch/hpc9876" and "/home/hpc9876/scr". Note that the former is automatically created when you obtain an account, but the latter has to be made by you explicitly. |
Running Gamess from a command lineTo run Gamess, a script file rungms is used. This file resides in the same directory as the gamess.00.x executable. Assuming that the home of the script file and executable is in your path, all you need to do is type rungms case_name 01 n_procs where case_name is the name of the input file (file extension is assumed to be .inp and must not be specified), and n_procs stands for the number of processes to be used in a parallel Gamess run. If n_procs=1 a serial run will be performed. Note: It is absolutely essential to have a good idea about the size and complexity of your calculations before you start a Gamess job. Many of the methods have terrible scaling properties, i.e. the computational cost grows very quickly with the number of electrons, degrees of freedom, or number of basis functions used. We suggest you start with a small basis set and a cheap method, and then slowly increase those parameters. Like most programs, Gamess requires an input (.inp) file that describes the system (usually a molecule) for which the calculation will be performed, specifies the level of calculation (eg, CISD), and provides other necessary information (starting orbitals, basis sets, required properties, etc). The format of the input is considerably more demanding than the one required for Gaussian (another widely used electronic-structure program), and much less information is hidden inside of defaults. This makes Gamess a very flexible program, but increases the risk of doing something wrong. Careful study of example input files and the documentation is required to run Gamess successfully. This is particularly true for CI or CAS-SCF calculations. Once an input file is prepared, you will have to make the decision if you want to run Gamess in serial or in parallel mode. Gamess supports the use of multiple processors. However, the scaling (ie, the efficiency of parallel processing) varies with the type of calculation and the systems. We suggest you perform a small test calculation of the same type as your production calculation (eg, with a minimal basis set), and rerun it several times with a varying number of processors. Compare the timings and use the maximum number of processors that yield acceptable scaling for your production calculation. Submitting (parallel) Gamess jobsGamess has to be run via SLURM, which is a load-balancing program that submits batch jobs to low-load processors on the cluster cluster. To learn more about this program, click here. A Gamess job is submitted to the Grid Engine in the form of an execution script. A reasonable execution script for Gamess looks like this: #!/bin/bash #SBATCH --output=STD.out #SBATCH --eror=STD.err #SBATCH --nodes=1 #SBATCH --ntasks=1 #SBATCH --cpus-per-task=8 # Number of CPUs #SBATCH --mem-per-cpu=4G # memory per CPU in MB #SBATCH --time=0-00:30 # time (DD-HH:MM) export SLURM_CPUS_PER_TASK module load gamess-us/20170420-R1 rungms name.inp &> name.out The --output and --error lines define the standard output and standard error files, respectively. Note that all lines starting with "#SBATCH" are directives for SLURM, and will be interpreted when the script is submitted to that program. You also need to specify the name of the input file just like in an interactive run. The number of processes is specified in the "--cpu-per-task" line, which instructs SLURM to allocate the proper number of CPUs for you on a single node. The --nodes and --ntasks lines should be kept at 1. You do not have to specify it separately in the rungms command line, because SLURM passes the $CPUS_PER_TASK variable. Assuming your Grid Engine script is called "gamess.sh", it is submitted to GridEngine by typing sbatch gamess.sh No further specification of the output is necessary, since this is done inside the script and handled by GridEngine. |
LicensingGamess is a licensed program although it is distributed freely. The license held by the Centre for Advanced Computing is limited to our computers at our main site. That means that any of our users can use the program on our machines (but nowhere else), whether they are located at Queen's or not. You are not allowed to copy the executable or any part of the distribution onto your local machine. However you can easily obtain the program yourself. See the GAMESS source code distribution page. Gamess is a very portable program, and will run on IBM PC's (Windows), on a Mac, a variety of Unix platforms (including Linux), and your cellphone (just kidding). Before you can access the Gamess executables and run the program, you have to read the license agreement and sign a statement that you have done so. Return it to us (fax to (613) 533-2015 or scan/email to cac.admin@queensu.ca) and we add you to a Unix group "gamess" which enables access to the software. Where can I get more detailed information ?Gamess is not a simple program to run. It requires careful study of the input format, and a certain degree of knowledge about the "nuts and bolts" of computational quantum chemistry.
|