ORCA¶

ORCA is user software on Alps. This guide is provided based on our experiences helping users – however we can’t provide the same level of support as we do for supported software. See the main applications page for more information.

ORCA is a quantum chemistry software package implementing a wide range of methods (DFT, HF, wave-function methods, etc) with Gaussian basis sets.

It is used for CPU-only simulations on Eiger, for which this guide applies. Please do not use it on GPU platforms such as Daint.

Installation¶

You need to download the compiled executable yourself here. You first need to create a user account, and make sure you fulfill the license requirements. Select the archive for the linux x86_64 architecture. For ORCA v6.1.1, it is called orca_6_1_1_linux_x86-64_shared_openmpi418_nodmrg.tar.xz

Move the archive into your $HOME or $PROJECT, and unpack it in its own subfolder.

Running ORCA calculations¶

ORCA is shipped as a pre-compiled executable linked against OpenMPI. To be able to run, the prgenv-gnu-openmpi uenv is required. Here is a sample Slurm batch script:

#!/bin/bash -l
#SBATCH --job-name=orca_job
#SBATCH --time=00:30:00
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=64
#SBATCH --account=<ACCOUNT>
#SBATCH --hint=nomultithread
#SBATCH --constraint=mc
#SBATCH --uenv=prgenv-gnu-openmpi/25.12:v1
#SBATCH --view=default

ORCA_PATH="absolute_path_to_orca_unpacked_directory" # modify this accordingly
export PATH="$PATH:$ORCA_PATH"
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$ORCA_PATH"

ORCA_EXEC="$ORCA_PATH/orca"

$ORCA_EXEC water.inp > water.out

The number of MPI processes must be specified in the ORCA input file, and it should match the specifications of the Slurm submission script. The example input file below (non-physical stack of water molecules with RI-MP2) matches the above sample script:

!RI-MP2 DEF2-QZVPP  DEF2-QZVPP/C
%PAL NPROCS 64  END
* xyz 0 1
O   0.0000   0.0000   0.0626
H  -0.7920   0.0000  -0.4973
H   0.7920   0.0000  -0.4973
O   0.0000   2.0000   0.0626
H  -0.7920   2.0000  -0.4973
H   0.7920   2.0000  -0.4973
O   0.0000   4.0000   0.0626
H  -0.7920   4.0000  -0.4973
H   0.7920   4.0000  -0.4973
O   0.0000   6.0000   0.0626
H  -0.7920   6.0000  -0.4973
H   0.7920   6.0000  -0.4973
O   0.0000   8.0000   0.0626
H  -0.7920   8.0000  -0.4973
H   0.7920   8.0000  -0.4973
O   0.0000  10.0000   0.0626
H  -0.7920  10.0000  -0.4973
H   0.7920  10.0000  -0.4973
O   0.0000  12.0000   0.0626
H  -0.7920  12.0000  -0.4973
H   0.7920  12.0000  -0.4973
*

Warning

The ORCA executable is built with version 4 of OpenMPI, but the prgenv-gnu-openmpi uenv provides version 5. In principle, OpenMPI is backward compatible and everything should work. It is however possible that unforeseen issues arise when using methods that go beyond the simple testing we have done.

Launching multiple ORCA calculations on a single node¶

Many ORCA calculations do not scale well enough to efficiently use an entire node (128 CPU cores on Eiger). Therefore, launching a single calculation per Slurm job can be quite wasteful. If the scientific workflow allows it, it is much more efficient to launch multiple ORCA calculations in parallel from a single Slurm submission script, with each calculation using only a fraction of the node. This can be done with HyperQueue.

After downloading the hq executable, any number of ORCA jobs can be scheduled with HyperQueue, each running on a specified number of CPUs. If all jobs cannot run concurrently, some will be queued until resources become available. You can adapt the following Slurm script to your needs. Note that all input files must contain an appropriate %PAL setting, and do not forget to use the --oversubscribe option.

#!/bin/bash -l
#SBATCH --job-name=orca_job
#SBATCH --time=00:30:00
#SBATCH --nodes=1
#SBATCH --ntasks-per-node=128
#SBATCH --account=<ACCOUNT>
#SBATCH --hint=nomultithread
#SBATCH --constraint=mc
#SBATCH --uenv=prgenv-gnu-openmpi/25.12:v1
#SBATCH --view=default

ORCA_PATH="absolute_path_to_orca_unpacked_directory" # modify this accordingly
export PATH="$PATH:$ORCA_PATH"
export LD_LIBRARY_PATH="$LD_LIBRARY_PATH:$ORCA_PATH"

ORCA_EXEC="$ORCA_PATH/orca"

# Start HyperQueue server and workers
hq server start &

# Wait for the server to be ready
hq server wait

# Start HyperQueue workers
srun hq worker start --cpus 128 &

# Submit tasks. Here the 4 calculations on 32 CPUs would start first, and as soon as one is finished
# the subsequent calculations can start (using 16 or 8 CPUs, respectively).
hq submit --pin taskset --cpus 32 --stdout pal32_1.out --stderr pal32_1.err  $orca_exec  pal32_1.inp --oversubscribe
hq submit --pin taskset --cpus 32 --stdout pal32_2.out --stderr pal32_2.err  $orca_exec  pal32_2.inp --oversubscribe
hq submit --pin taskset --cpus 32 --stdout pal32_3.out --stderr pal32_3.err  $orca_exec  pal32_3.inp --oversubscribe
hq submit --pin taskset --cpus 32 --stdout pal32_4.out --stderr pal32_4.err  $orca_exec  pal32_4.inp --oversubscribe
hq submit --pin taskset --cpus 16 --stdout pal16_1.out --stderr pal16_1.err  $orca_exec  pal16_1.inp --oversubscribe
hq submit --pin taskset --cpus 16 --stdout pal16_2.out --stderr pal16_2.err  $orca_exec  pal16_2.inp --oversubscribe
hq submit --pin taskset --cpus 8 --stdout pal8_1.out --stderr pal8_1.err  $orca_exec  pal8_1.inp --oversubscribe
hq submit --pin taskset --cpus 8 --stdout pal8_2.out --stderr pal8_2.err  $orca_exec  pal8_2.inp --oversubscribe
hq submit --pin taskset --cpus 8 --stdout pal8_3.out --stderr pal8_3.err  $orca_exec  pal8_3.inp --oversubscribe
hq submit --pin taskset --cpus 8 --stdout pal8_4.out --stderr pal8_4.err  $orca_exec  pal8_4.inp --oversubscribe

# Wait for all jobs to finish
hq job wait all

# Stop HyperQueue server and workers
hq server stop

echo
echo "Everything done!"