Computational chemistry and computational physics software

• c98scf
• c98scfdir
• c98integrals
• c98properties
• c98convert
• c98loptcg
• c98bulk

Example LoadLeveler job file for a serial run

#!/bin/ksh
# @ output = pHe.out
# @ error = pHe.err
# @ class = lhugex
# @ notification = never
# @ notify_user = deumens@qtp.ufl.edu
# @ checkpoint = no
# @ restart = no
# @ requirements = (Arch == "p2sc") && 
                       (OpSys == "AIX51")
# @ queue

HOST=`uname -n`
echo $HOST
MOL=pHe
SCR=/scr_1/tmp/$HOST.$MOL_ed.$$
mkdir $SCR
cp $HOME/END/$MOL.inp $SCR/
cd $SCR
echo "running version 2..."
time endyne -e 2.7.6 $MOL.inp > ${MOL}v2.log 2>&1
echo "running version 5..."
time endyne -e 5 $MOL.inp > ${MOL}v5.log 2>&1
echo "========= files ========="
ls -l
echo "========================="
cp ${MOL}*.log $HOME/END/
echo done
		

1. Solaris 9/SPARC on BUDDY,
2. Linux/IA32 (g03-C.02)
3. Linux/x86_64 (g03-D.01)
4. Linux/AMD (g03-E.01)
5. Linux/EM64T (g03-E.01) on the Linux clusters.

To run Gaussview on linx64you should set the environment variable export g03root=/share/local/lib/g03-EM64T
. $g03root/g03/bsd/g03.profile
Then the alias gv will be defined to start Gaussview. Gaussview is an X11 application. It will require an application like X-Win32 on a Windows computer to view Gaussview. MacOS and Linux desktop and laptop computers have X11 built in. The x11 also must support OpenGL for Gaussview to work. It does not work on the old QTP SUN stations.

You can start GaussView with
gv -mesagl
which will use a software OpenGL driver so that no special hardware is needed. Possible hardware conflicts with the graphics on the X-display computer can be avoided this way as well.

You should set the environment variable GAUSS_SCRDIR=/scr_1/tmp/your_directory;export GAUSS_SCRDIR before running the command g98 or g03. By default GAUSS_SCRDIR=/scr_1/tmp and then the 6 am cleaning daemon will remove all integral files. By setting the variable to the correct directory as explained in diskspace instructions, this is avoided.

You must contact Erik Deumens to sign a non-disclosure agreement before you will be allowed to access the program.

Current status of platforms and versions.

Machines	Processor	Version	Command
buddy	SPARC	g03 B.04	g03
arwen00-1d	IA32	g03 C.02	g03-32
haku00-1f surg1-7 wukong0-f	x86_64 EM64T	g03 E.01 or g03 D.01	g03-EM64T or g03-64
ra00-3i	x86_64 AMD	g03 E.01 or g03 D.01	g03-AMD or g03-64

Note that g03-64 and g03-EM64T both work on Intel EM64T nodes, aand that g03-64 and g03-AMD both work on AMD Opteron nodes; g03-32 works on all Linux nodes.

Example PBS job file for a 4-way parallel run of g03-64 on SURG

#!/bin/bash
#PBS -N g03par
#PBS -o g03par.out
#PBS -e g03par.err
#PBS -m abe
#PBS -M XXXXX@qtp.ufl.edu
#PBS -q brute
#PBS -l nodes=1:surg:ppn=4

# Use local scratch
mkdir /scr_1/tmp/`hostname`.g03.$$
GAUSS_SCRDIR=/scr_1/tmp/`hostname`.g03.$$
export GAUSS_SCRDIR
cd /scr_1/tmp/`hostname`.g03.$$
# Make sure the job.in file has the Gaussian 
# directive to use 4 CPUs
cp $HOME/job.in .
g03-64 job.in > job.out
cp job.out $HOME/job.out
cd ..
rm -rf `hostname`.g03.$$
echo done
		

1. Check out the CDs from Erik Deumens in NPB 2334. Please bring the CDs back within one day.
2. You will get two CDs. The first CD contains the standard HyperChem version 8 installation. It should be installed first. use the options standalone and software license. The CD will also install POVray, a free sofwtare package, that is needed to display molecular densities. The default installation directory is C:\Hyper80.
3. Once the installation completes succesfully, you need to remove the executable file chem.exe (the icon is a green beaker) from the folder C:\Hyper90\Program and replace it with chem.exe from the second CD. This new executable has the license to run on a computer with an IP address in the range 10.244.0.0 as used inside the New Physics Building.
4. If you install this on your laptop, HyperChem will only work when the laptop is connected to the network in the Physics building by cable, not by wireless and not in any othr location. If you need a license that works elsewhere, you must contact HyperCube and ask for an individual machine license. Then the installation of the first CD is sufficient.

An example input deck for a run on Xena II is (to be saved as nwchem.job):

#!/bin/ksh
# @ job_type=parallel
# @ class = lhugex
# @ output = nwchem.out
# @ error =  nwchem.err
# @ requirements = (Arch == "p2sc")
# @ notify_user = your_name@qtp.ufl.edu
# @ min_processors = 12
# @ max_processors = 15
# @ network.lapi = css0,not_shared,US
# @ environment = COPY_ALL; MP_PULSE=0; 
             MP_SINGLE_THREAD=yes; 
             MP_WAIT_MODE=yield; restart=no
# @ queue
export CNAME=nwchem
export CALC="/scr_2/tmp/$machine.nwchem.$$"
mkdir $CALC
export PROC=p2sc
export LOCAL="`pwd`"
date
uname -a
cd $CALC
if [ `pwd` != $CALC ]
then 
   echo "In `pwd` instead of $CALC: STOP!"
   exit
fi
echo '******* NWChem will start now *******'
nwchem $LOCAL/$CNAME".nwi" $LOCAL/$CNAME".nwo"
echo '******* NWChem exited *******'
echo "starting clean-up"
cd ..
/bin/rm -r $CALC
echo done
exit

start
title "q0l-02f, Me2N-NO2, Cs, 
         CCSD(T)-fc/cc-pVTZ//B3LYP/6-31G**"
geometry autosym
N -.567554 -.224363 .000000
N .804248 -.034262 .000000
O 1.360973 .024660 1.097057
O 1.360973 .024660 -1.097057
C -1.239866 .053279 -1.263921
C -1.239866 .053279 1.263921
H -.755338 -.502712 -2.064122
H -2.273466 -.281438 -1.163781
H -1.227028 1.121644 -1.516981
H -.755338 -.502712 2.064122
H -1.227028 1.121644 1.516981
H -2.273466 -.281438 1.163781
end
basis spherical nosegment
C library cc-pvtz
H library cc-pvtz
N library cc-pvtz
O library cc-pvtz
end
memory 512 mb
echo
ccsd; maxiter 40; freeze atomic; end
task ccsd(t)

task shell all "/bin/rm -f *"

Example LoadLeveler job file for a parallel run

#!/bin/bash
#@ shell = /bin/bash
#@ job_name = siesta_h2o
#@ output   = siesta.out
#@ error    = siesta.err
#@ job_type = parallel
#@ class    = quick
#@ node_usage = shared
#@ tasks_per_node = 1 
#@ node        = 2
#@ wall_clock_limit = 1:00:00
#@ notification     = never
#@ network.MPI      = css0,shared,us
###@ environment = MPI_SHARED_MEMORY=yes; 
            MP_SAVEHOSTFILE=hf.$(host).$(jobid); 
            COPY_ALL;
#@ queue

# Use global scratch disk
SCR=/scr_2/tmp/`hostname`.my_h2o.$$
mkdir $SCR
cd $SCR
# Copy input to work directory
cp $HOME/h2o.fdf .
# Run
/usr/bin/poe /usr/local/bin/siesta < h2o.fdf 
         > h2o.out
# Copy output to a safe place
cp h2o.out $HOME
# Remove scratch files
cd ..
rm -r $SCR

SystemName          Water molecule
SystemLabel         h2o
NumberOfAtoms       3
NumberOfSpecies     2

%block ChemicalSpeciesLabel
 1  8  O # Species index, atomic number, 
                     species label
 2  1  H
%endblock ChemicalSpeciesLabel

AtomicCoordinatesFormat  Ang
%block AtomicCoordinatesAndAtomicSpecies
 0.000  0.000  0.000  1
 0.757  0.586  0.000  2
-0.757  0.586  0.000  2
%endblock AtomicCoordinatesAndAtomicSpecies

MD.TypeOfRun             Nose # Type of dynamics:
MD.InitialTemperature    300 K
MD.TargetTemperature     300 K
MD.InitialTimeStep         1
MD.FinalTimeStep         100
MD.LengthTimeStep        1.0 fs

The codumentation explains how to create the control file and initial coordinates and other files. The command define is used to create a control file. It may refer to other files with coordinates and basis specifications.

Example LoadLeveler job file for a serial run

#!/bin/ksh 
# @ output = sertur.out 
# @ error = sertur.err 
# @ class = lhugex 
# @ notification = never 
# @ notify_user = greene@qtp.ufl.edu 
# @ checkpoint = no 
# @ restart = no 
# @ requirements = (Arch == "p2sc" && 
                     OpSys == "AIX51") 
# 
# mandatory directive to process the job. 
# @ queue 
# 
# User commands follow.... 
echo "pure serial job:" 
echo master: `uname -n` 
# ==== begin executable statement 
. /usr/local/lib/turbomole/scripts/turbopath
scr=/scr_1/tmp/`uname -n`.sg.$$
mkdir $scr
cd $scr
cp /ufl/chem/nxr/sg/RUB/OXID/SER/* .
# Geometry optimization
jobex
# single DFT SCF
#dscf > dscf.out
cp * /camp/crunch_1/chem/nxr/sg/TURBO/RUBox/SER
# ==== end executable statement 
echo "Job done." 

Example LoadLeveler job file for a parallel run

You must create a control file with define just is in the case of the serial file. Then you must prepare for a parallel run. This must be done on one of the interactive nodes of Xena II. Create a directory for the files for this particular run. Put a file host.list in it with the following content:

xena00
xena10
xena20
xena30
xena40
xena50
xena60
xena70
xena80
xena90
xenaa0
xenab0

The actual SCF or gradient or geometry calculations should be submitted to LoadLeveler with a script like the folowing. Parallel calculations are always prepared with turbo_start -batch_prepare and executed in the LoadLeveler script with turbo_start -batch. Note the following when answering the questions asked by turbo_start -batch_prepare:

• Because Turbomole uses a master-client architecture for parallel execution, it is necessary to assign one more process. Thus to run two-way parallel, you must ask LoadLeveler to assign three processes. When you create the control file as explained in the manual, you must specify that you want N clients, and you must specify N+1 processes in the LoadLeveler job file.
• The directory for scratch files should be a global directory accessible by all nodes. For Xena II this is /scr_2/tmp/.... It is recommended that you create a keep-directory, such as /scr_2/tmp/keep.sg.13-04-03, so that it will not be deleted by the scratch cleaning process and so that you can specify it in your LoadLeveler script (see example below).
• When answered whether the machine is in dedicated use, you answer yes for Xena II and no for Simu.

#!/bin/ksh 
# @ output = partur.out 
# @ error = partur.err 
# @ class = lhugex 
# @ notification = complete 
# @ notify_user = greene@qtp.ufl.edu 
# @ checkpoint = no 
# @ restart = no 
# @ requirements = (Arch == "p2sc" && 
                        OpSys == "AIX51") 
# @ wall_clock_limit = 10:00,9:00 
# 
# ---- begin specification of the kind
# ---- of parallel job 
# @ job_type = parallel 
# @ node_usage = shared 
# @ node = 3,3 
# @ tasks_per_node = 1 
# @ network.MPI = css0,not_shared,US 
# ---- end specification of the kind 
# ---- of parallel job 
# 
# mandatory directive to process the job. 
# @ queue 
# 
# User commands follow.... 
echo "pure MPI job on 2 different nodes:" 
echo master: `uname -n` 
# ==== begin executable statement 
. /usr/local/lib/turbomole/scripts/turbopath
. /usr/local/lib/turbomole/scripts/turbopathmpi
# Need global directory for parallel execution
scr=/scr_2/tmp/keep.sg.13-04-03
mkdir $scr
cd /ufl/chem/nxr/sg/RUB/OXID/PAR/
cp control control_seq basis coord alpha beta $scr/
cp xyzcase xyzproc $scr
cp DSCF-par-stat $scr
cp GRAD-par-stat $scr
cd $scr
turbo_start -batch
cp * /camp/crunch_1/chem/nxr/sg/TURBO/RUBox/PAR
# ==== end executable statement 
echo "Job done."