HPC clusters

With pip, the released version into a plain virtual environment:

module purge            # optional; stops other modules leaking packages via PYTHONPATH
module load python      # any Python >= 3.11, e.g. Python/3.12.3-GCCcore-13.3.0
python -m venv ~/envs/qorona && source ~/envs/qorona/bin/activate
pip install qorona

With conda, the development version from source:

git clone https://github.com/RayanDhib/Qorona.git
cd Qorona
conda env create -f environment.yml
conda activate qorona

Two flags are made for batch jobs: --workers pins the kernel threads to your allocation (numba otherwise takes every core it sees), and --quiet keeps job logs readable.

One node, one task, many CPUs

Qorona is shared-memory threaded (numba), not MPI: request a full node's cores with --cpus-per-task and do not use mpirun.

On GPU nodes, load the site's CUDA module (e.g. module load CUDA/12.4.0) so numba finds the toolkit; --device gpu errors loudly if the GPU is unusable rather than silently falling back. See GPU acceleration for the backend details and timings.

SLURM templates

Both templates live in hpc/ in the repository; replace the <placeholders> with your site's values. The --cluster line is only needed where one scheduler serves several clusters (e.g. VSC); drop it otherwise.

CPU node

#!/bin/bash -l
#SBATCH --job-name=qorona-cpu
#SBATCH --cluster=<cluster>
#SBATCH --partition=<partition>
#SBATCH --account=<account>
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=36
#SBATCH --mem-per-cpu=4G
#SBATCH --time=00:30:00
#SBATCH -o qorona-cpu-%j.out
#SBATCH -e qorona-cpu-%j.err

# Qorona is threaded (numba), not MPI: one node, one task, many cores.
module purge
module load Python/3.12.3-GCCcore-13.3.0    # adjust to your site: `module spider Python`
source <path-to-venv>/bin/activate          # the venv Qorona is in (or: conda activate qorona)

cd <case-directory>

# Build the volume once, then render any number of viewpoints off it.
# --workers ties the threads to the cores SLURM gave this job.
qorona build solution.CFmesh -o solution.qor --outer-radius 8 \
    --device cpu --workers "$SLURM_CPUS_PER_TASK" --quiet

qorona render solution.qor -o eclipse.png --fov 8 --longitude 317 --latitude 6.2 \
    --workers "$SLURM_CPUS_PER_TASK" --quiet

GPU node

#!/bin/bash -l
#SBATCH --job-name=qorona-gpu
#SBATCH --cluster=<cluster>
#SBATCH --partition=<gpu-partition>
#SBATCH --account=<account>
#SBATCH --nodes=1
#SBATCH --ntasks=1
#SBATCH --cpus-per-task=8
#SBATCH --gpus-per-node=1
#SBATCH --mem-per-cpu=4G
#SBATCH --time=00:15:00
#SBATCH -o qorona-gpu-%j.out
#SBATCH -e qorona-gpu-%j.err

# One GPU builds the volume; a few CPU cores feed it. One node, one task.
module purge
module load Python/3.12.3-GCCcore-13.3.0 CUDA/12.4.0    # adjust versions to your site
source <path-to-venv>/bin/activate                     # the venv Qorona is in (or: conda activate qorona)

cd <case-directory>

# --device gpu builds on the GPU (it errors loudly if the GPU is unusable).
qorona build solution.CFmesh -o solution.qor --outer-radius 8 \
    --device gpu --workers "$SLURM_CPUS_PER_TASK" --quiet

qorona render solution.qor -o eclipse.png --fov 8 --longitude 317 --latitude 6.2 \
    --workers "$SLURM_CPUS_PER_TASK" --quiet