TURBOMOLE
Overview
TURBOMOLE is an ab initio quantum chemistry simulation package for ground and excited state electronic structure calculations. In addition to Hartree Fock and DFT, a number of second order approximations are available, including MP2 and coupled cluster.
The developers claim TURBOMOLE to be one of the fastest and most stable codes available for standard quantum chemistry applications. To this end, development is primarily focused on the provision of a fast and stable code to treat molecules of industrial relevance within reasonable time and memory constraints.
Version 7.6, 7.5 & 7.3 are currently installed on the CSF.
Restrictions on use
The license agreement, a copy of which is available in
/opt/apps/apps/binapps/turbomole/TURBOMOLE_license_2015.pdf
must be strictly adhered to by users. Important points to note are:
Who may use TURBOMOLE?
The software is only available to users from EPS and they must be a member of Staff or a Student of the University of Manchester or visiting scientists who are collaborating with a member of Staff and who have been given an official University of Manchester IT username.
Users are reminded that they must not share their password with anyone, or allow anyone else to utilise their account.
Access to the software cannot be granted to users whose primary location is not the main University campus.
What may TURBOMOLE be used for?
The software may be used for academic research and educational purposes only.
No commercial work must be undertaken using the software.
Further details are available in clause 3 of the license.
What may TURBOMOLE not be used for?
TURBOMOLE may not be used for any commercial work or calculations for third parties. See clause 3 of the license agreement for further detail of how this is defined.
You must not:
- Copy or distribute the software,
- decompile, disassemble, or reverse engineer the software,
- share the data generated with the software with third parties for purposes other than academic research or publication in a scientific journal,
- use the data generated in a patent application,
Further details are available in clause 3 of the license.
How should the program be cited?
Data generated may be published in scientific journals and presentations. You must correctly cite the software using this format:
“TURBOMOLE , TURBOMOLE GmbH, Karlsruhe, , http://www.turbomole.com”
All COSMOxxx products should be cited clearly by name and version number.
See clause 5 of the license agreement and the TURBOMOLE manual/website for further guidance.
Getting access to the software
Anyone wishing to use TURBOMOLE must request access to the turbomole unix group and confirm that they have read the above T&Cs via its-ri-team@manchester.ac.uk.
Set up procedure
One of the following modulefiles must be loaded, depending on whether you wish to run a single core (serial), shared memory (smp), or multi-node (mpi) job:
module load apps/binapps/turbomole/7.6/serial module load apps/binapps/turbomole/7.6/smp module load apps/binapps/turbomole/7.6/mpi module load apps/binapps/turbomole/7.5/serial module load apps/binapps/turbomole/7.5/smp module load apps/binapps/turbomole/7.5/mpi module load apps/binapps/turbomole/7.3/serial module load apps/binapps/turbomole/7.3/smp module load apps/binapps/turbomole/7.3/mpi
Note that the GlobalArrays library was removed as of version 7.2. Users should load the mpi module instead.
TURBOMOLE 7.1, 7.1.1 and 7.2.1 have also been made available for users who require it for continuity of ongoing projects and where results cannot be reproduced exactly with 7.3 or 7.5. Please refer to the section Earlier Releases below for some additional setup instructions required to use these on CSF3. Users are strongly advised to use the current release unless it is essential for ongoing work to continue with an older version.
Running the application
Please do not run turbomole on the login nodes: All work must be submitted to the batch system.
Many of the environment settings needed for the different job types are set by default in the relevant modulefile, so you do not need to set everything specified in the turbomole manual. The examples below contain all the key information for running a job.
All of the input files required for your job must be located in the directory from which you submit your job. Turbomole input should be in a file named control.
Serial batch job submission
The following batch script will setup a serial job. Simply replace TURBOMOLE_COMMAND with the turbomole code that you wish to run, e.g. dscf.
#!/bin/bash --login #$ -cwd module load apps/binapps/turbomole/7.6/serial TURBOMOLE_COMMAND
To submit the job, enter qsub jobscript, where jobscript is the name of your batch script.
Single node jobs, 2-32 cores
The following script sets up an 8 core smp job. Again, simply replace TURBOMOLE_COMMAND with the turbmole code that you wish to run, e.g. ridft. Please note that not all turbomole codes offer smp execution, so do consult the manual first. For single node parallel jobs running on a small number of cores, the smp version will typically be more efficient than the equivalent mpi version, where available. Set PARNODES=$NSLOTS to tell turbomole to spawn the appropriate number of threads.
#!/bin/bash --login #$ -cwd #$ -pe smp.pe 8 module load apps/binapps/turbomole/7.6/smp # Inform turbomole how many cores to use export PARNODES=$NSLOTS TURBOMOLE_COMMAND
For larger single node jobs, the mpi version might be more efficient:
#!/bin/bash --login #$ -cwd #$ -pe smp.pe 24 module load apps/binapps/turbomole/7.6/mpi # Inform turbomole how many cores to use export PARNODES=$NSLOTS TURBOMOLE_COMMAND
To submit the job, enter qsub jobscript, where jobscript is the name of your batch script.
MPI jobs (multiples of 24 cores, minimum 48)
The following script sets up a 48 core job. Again, simply replace TURBOMOLE_COMMAND with the turbmole code that you wish to run, e.g. ricc2. As for smp, not all turbomole codes have mpi versions available, so please consult the manual to find the appropriate version for your job. Set PARNODES=$NSLOTS to tell turbomole to launch the appropriate number of processes. Note that turbomole will launch one extra process, the first being a server that spawns and monitors the worker processes.
#!/bin/bash --login #$ -cwd #$ -pe mpi-24-ib.pe 48 module load apps/binapps/turbomole/7.6/mpi # Inform turbomole how many cores to use export PARNODES=$NSLOTS TURBOMOLE_COMMAND # # For example: dscf # Note that you should NOT use mpirun to start the MPI versions of turbomole apps.
To submit the job, enter qsub jobscript, where jobscript is the name of your batch script.
Earlier Releases
The following release are available for use where this is essential to ongoing work:
apps/binapps/turbomole/7.1/mpi apps/binapps/turbomole/7.1/serial apps/binapps/turbomole/7.1/smp apps/binapps/turbomole/7.1.1/mpi apps/binapps/turbomole/7.1.1/serial apps/binapps/turbomole/7.1.1/smp apps/binapps/turbomole/7.2.1/mpi apps/binapps/turbomole/7.2.1/serial apps/binapps/turbomole/7.2.1/smp
Please note that these releases require an additional step in the jobscript for multi-node MPI jobs, otherwise all processes will run on a single node. The following sample jobscript demonstrates how to do this:
#!/bin/bash --login #$ -cwd #$ -pe mpi-24-ib.pe 48 module load apps/binapps/turbomole/7.1/mpi # Additional step needed for MPI in older versions of turbomole source setup_turbompi TURBOMOLE_COMMAND
Sourcing the “setup_turbompi” script creates an MPI hostfile and sets the relevant environment variable required for turbompi to reference it, then sets PARNODES for you. This ensures that processes are distributed amongst all of the nodes assigned to your job. Please note that the script must be sourced rather than executed, such that the required environment variables are available to turbomole.
Further info
Further information can be found on the TURBOMOLE or COSMOlogic websites including Turbomole news, FAQs, the latest basis sets, documentation, and tutorials. There is also a user mailing list to announce news, bugs and fixes, and a TURBOMOLE User Forum.
Updates
None at present.