Skip to main content

Posts

Showing posts from April, 2010

Good performance of SSB-D for NMR

We have been testing our newly developed SSB-D functional on NMR shieldings of simple and more complex molecules, and it has withstood the scrutiny.

Similar to other properties, the new functional functions better than either PBE or OPBE. The latter was recently (J. Comp. Chem. 2007, 28, 2431-2442; Chem. Phys. Lett.2006< 421, 383-388; J. Comput. Chem.2008, 29, 1798-1807; J. Phys. Chem. A2008, 112, 6794–6799) reported to already be one of the best DFT functionals.

In a recently published paper, we show that SSB-D improves upon OPBE.

CFOUR in parallel

Lately I have managed to get CFOUR working in parallel with the following options for the configuration script:

FC=ifort CC=gcc MPIFC=mpif90 ./configure --with-blas=blashc \
--enable-mpi=openmpi --with-mpirun="mpirun -np \$CFOUR_NUM_CORES" \
--with-exenodes="mpirun -np \$CFOUR_NUM_CORES"

The only thing that remains to be done is add the $CFOUR/bin directory to your PATH:

export PATH=$PATH:$CFOUR/bin

and set the CFOUR_NUM_CORES environment variable:

export CFOUR_NUM_CORES=`cat $PBS_NODEFILE|wc -l`

And this works, especially for larger systems.

NB. In the output it should say:

PARALLEL RUN STARTED ON 8 NODES.

and

parallel ececution on node 0 of 8

On 1 node, you can see the scale-up:

1 cpu
@CHECKOUT-I, Total execution time : 920.9800 seconds.
2 cpus
@CHECKOUT-I, Total execution time : 465.4900 seconds.
4 cpus
@CHECKOUT-I, Total execution time : 241.4300 seconds.
8 cpus
@CHECKOUT-I, Total execution time : 138.0800 seconds.

Mikael Johansson starts as Juan de la Cierva

Today dr. Mikael Johansson has started working in my group within the IQC. He will be working on spin-state reactivity and transition-metal complexes.


He had obtained a prestigious Juan de la Cierva post-doc position (3 years), which is being financed by the Spanish Ministry of Science and Innovation (MICINN).