Data Availability - DDA Benchmark Paper

This repository contains the data, scripts, and figures used in the paper.

Related Repository

The benchmark software used in this work is available at:

• https://github.com/Argencle/dda-bench.git

License

• Source code is licensed under the MIT License.
• Data and figures are licensed under CC-BY-4.0.

Repository Purpose

• Provide the raw benchmark results (logs + CSV) used in the paper.
• Provide post-processing scripts to rebuild figure datasets.
• Provide plotting scripts to regenerate the final PDF figures.
• Provide patches required to reproduce the modified third-party codes:
  • ADDA
  • DDSCAT

Repository Structure

• patches/
• patches/adda_b03d648.patch: ADDA patch.
• patches/ddscat_250505.patch: DDSCAT patch.
• datas/
• logs_MPI_laptop/, logs_MPI_cluster/: CPU/MPI results.
• logs_GPU_2000Ada/, logs_GPU_6000Ada/, logs_GPU_cluster/: GPU results.
• In logs*, run logs are stored as .zip archives; only aggregated files are kept uncompressed.
• *.csv: aggregated figure datasets.
• scripts/
• run_MPI_random_and_extract.sh: runs MPI/CPU simulations and extracts metrics.
• run_GPU_random_and_extract.sh: runs GPU simulations and extracts metrics.
• sort_datafiles.py: standard sorting of result CSV files.
• fill_time_in_datafiles.py: fills elapsed times (cluster/SLURM case).
• make_figure1_data.py, make_figure2_data.py, make_figure3_data.py, make_appendix_data.py: build figure CSV datasets.
• plot_figure1.py, plot_figure2.py, plot_figure3.py, plot_appendix.py: generate PDF figures in figures/.
• figures/: output folder storing generated PDF figures.

Python Requirements

python -m venv .venv
source .venv/bin/activate
pip install --upgrade pip
pip install -r requirements.txt

Reproducing Code Modifications

1) ADDA

Expected base version: commit b03d6480b5f41b88abe0b201c847da780d1efe56.

Optional (without git clone and git checkout): download the ZIP snapshot of that commit directly from https://github.com/adda-team/adda/archive/b03d6480b5f41b88abe0b201c847da780d1efe56.zip, extract it, and apply the same patch in the extracted source tree.

git clone https://github.com/adda-team/adda.git
cd adda
git checkout b03d6480b5f41b88abe0b201c847da780d1efe56
git apply --check ../patches/adda_b03d648.patch
git apply ../patches/adda_b03d648.patch

Then compile ADDA.

2) DDSCAT

Expected base version: 7.3.4_250505.

Download the source ZIP from the official downloads page and extract it first. Then run the patch commands:

patch --dry-run -p0 < patches/ddscat_250505.patch
patch -p0 < patches/ddscat_250505.patch

Then compile DDSCAT.

Regenerate Simulation Logs

All run logs are already provided in logs* as .zip archives.

If you want to rerun simulations and re-extract metrics:

bash scripts/run_MPI_random_and_extract.sh
bash scripts/run_GPU_random_and_extract.sh

Note: both scripts contain local paths that must be adapted.

Update aggregated CSV files from logs

Aggregated CSV files are already included in logs* (uncompressed part), as *_results_sorted.csv.

Run:

python scripts/sort_datafiles.py
python scripts/fill_time_in_datafiles.py

Note:

• the ‘logs*’ folders must be unzipped in order to use these scripts.
• sort_datafiles.py will not reproduce the exact repository versions of ddscat_results_sorted.csv and ifdda_results_sorted.csv unless 4 outlier runs are removed. In the repository, these 4 lines were manually removed because they correspond to broken runs with unrealistic values.

Build final figure datasets

Final figure datasets are already provided in datas/ as CSV files.

Run:

python scripts/make_figure1_data.py
python scripts/make_figure2_data.py
python scripts/make_figure3_data.py
python scripts/make_appendix_data.py

Figure Reproduction (Plots Only)

Run:

python scripts/plot_figure1.py
python scripts/plot_figure2.py
python scripts/plot_figure3.py
python scripts/plot_appendix.py

Outputs are written to figures/.

Citation

If you use this dataset, please cite the dataset archive and the associated paper.

Dataset

Argentin, C., & Chaumet, P. C., (2026). data-dda-benchmark-paper. Zenodo.
https://doi.org/10.5281/zenodo.18847515

Paper

Argentin, C., Chaumet, P. C., Gross, M., & Yurkin, M. A. (2026).
Floating-point--consistent cross-verification methodology for reproducible and interoperable DDA solvers with fair benchmarking.
arXiv:2603.02871