Running Priroda on Grex

Introduction#

Priroda is a fast parallel relativistic DFT and ab-initio code for molecular modeling, developed by Dr. Dimitri N. Laikov. The code originally implemented fast resolution-of-identity GGA DFT for coulomb and exchange integrals. Later it was extended to provide RI-DFT with hybrid functional, RI-HF and RI-MP2, and parallel high-level coupled-cluster methods. All these levels of theory can be used together with an efficient all-electron scalar-relativistic method, with small-component bases supplied for all the elements of the Periodic Table. The current release of the code also includes a novel NDO-based semi-empirical method.

User Responsibilities and Access#

The code is free for academic users, but is not open source. It is distributed on request by the Author, Dr. Dimitri N. Laikov.

To access the Priroda code on Grex, the prospective users have to send us (support@tech.alliancecan.ca ) a free-form email confirming that they have read and agreed to abide by the following conditions:

Conditions for the Priroda code access on Grex:

  • I understand that the Priroda code’s ownership and copyright belongs solely to its Author, Dr. Dimitri N. Laikov. I will not incorporate any part of the Priroda code into any other program system, either for sale or for non-profit distribution, without written permission by the Author.
  • I will not copy, distribute or supply the Priroda code for any reason whatsoever to third persons or organizations. Instead, I will direct all the code requests to the Author.
  • If results obtained with the code are published, I will cite the proper Priroda code references and, when appropriate, the specific methods references, as described in the Priroda code documentation and/or Dr. Laikov’s website .
  • I understand that the Priroda code is provided “as is” and the author is not assuming any responsibilities or any liabilities that might arise from the usage of the code, whatsoever.

After receiving the email, we will add the user to the wg-prrda UNIX group that is used to control access to the Priroda program, basis sets and documentation.

Running Priroda on Grex#

The Priroda code is linked against OpenMPI built with a GCC compiler. There are several versions of them, and module spider priroda would help to locate the dependencies. As of the time of writing this documentation, the following command would load the Priroda version of 2016:

module load gcc/5.2 ompi/3.1.4 priroda/2016

The parallel Priroda executable (called p) will be in the PATH after loading of the module. Its basis sets and/or semi-empirical method parameters can be found under $PRIRODA/bin. Documentation and examples are available under $PRIRODA/doc and $PRIRODA/example, correspondingly.

The style of the Priroda input is of free format namelist groups, similar to that of GAMESS-US but more flexible (no limitations inherited from Fortran77). Examples and description of each input group are in the doc and example directories. To invoke the code interactively:

mpiexec p name.inp name.out

An archive of old Priroda documentation is here Priroda old docs from KNC

Using Priroda with SLURM#

Priroda is MPI-parallelized. The parallel efficiency varies on the method used and the kind of calculation (energies, geometry optimizations or analytical hessians) performed. Pure GGA DFT calculations are quite fast and tightly coupled, and it makes sense to use a single node with a few tasks per node, or a few nodes, as in the example below. RI-MP2 calculations would benefit from more massively parallel calculations, spanning several nodes.

It makes no sense to ask more than 4000 mb per task.

Script example for running PRIRODA on Grex
run-priroda.sh
#!/bin/bash

#SBATCH --nodes=1 
#SBATCH --ntasks-per-node=6
#SBATCH --mem-per-cpu=2000M
#SBATCH --time=0-2:00:00
#SBATCH --job-name=priroda-test-c60

SCR=$TMPDIR

echo "assuming inputs in $SLURM_SUBMIT_DIR"

module load gcc/5.2 ompi/3.1.4 priroda/2016

# copy the input file (c60.inp) locally and set the resource requests
# and temporary paths. note that the file myfile.inp
# should not have a $system .. $end group

cp c60.inp $SCR/priroda.inp

cd $SCR
echo '   ' >> priroda.inp
echo ' $system ' >> priroda.inp
echo "   memory=1000 disk=10 path=. " >> priroda.inp
echo ' $end ' >> priroda.inp

cat priroda.inp

# Copy basis sets locally:

cp $PRIRODA/bin/*.in $SCR
cp $PRIRODA/bin/*.bas $SCR

echo "Start date:`date`"

# Actually run the job

srun $PRIRODA/bin/p priroda.inp $SLURM_SUBMIT_DIR/c60.$SLURM_JOBID.log

echo "Program finished with exit code $? at: `date`"

Various scripts and utilities#

There are some simple scripts and utilities in the $PRIRODA/contrib directory. They can be used for conversions of inputs/outputs to and from Molden XYZ format, extraction of the MOs and vibrational frequencies, and restart information from the Priroda output files.