CUDA-accelerated radial distribution functions using NVIDIA ALCHEMI Toolkit-Ops O(N) neighbor lists and PyTorch. Compatible with ASE Atoms or MDAnalysis Universe objects.
GPU required: An NVIDIA GPU with CUDA is needed. For CUDA version selection, PyTorch wheel setup (CUDA 12 vs 13) and more information, see the Toolkit-Ops installation guide.
Benchmarking
cuRDF is benchmarked against other RDF (MDAnalysis) and neighbour list implementations on CPU (AMD Ryzen 9 9950X, 32 threads) and GPU (NVIDIA RTX 5090) for systems of varying sizes at a density of 0.05 atoms/ų over 1000 frames. Benchmarks use random positions in cubic cells sized to maintain a fixed number density (orthorhombic boxes, periodic in all directions).
Install
Latest release:
pip install cuRDF
For development:
git clone https://github.com/joehart2001/curdf.git
cd curdf
pip install -e .
Quickstart
ASE Atoms object:
from ase.io import read
from curdf import rdf
# Load trajectory or frame e.g. XYZ, extxyz, traj, LAMMPS data/dump
atoms = read("md_run.extxyz")
# Compute RDF between species C and O from 1.0 to 8.0 Å
bins, gr = rdf(
atoms,
species_a="C",
species_b="O", # species b can be the same as species a
r_min=1.0,
r_max=8.0,
nbins=200, # resolution of rdf histogram binning
method="cell_list", # neighbor list method: "cell_list" (larger systems) or "naive" (smaller systems, less overhead)
output = "results/rdf.csv" # optional output
)
# Plot RDF
plot_rdf(bins, gr, path="results/rdf.png")
MDAnalysis Universe (topology and trajectory):
import MDAnalysis as mda
from curdf import rdf
u = mda.Universe("topology.data", "traj.dcd", atom_style="id type x y z")
bins, gr = rdf(
u,
species_a="C",
species_b="O",
r_min=1.0,
r_max=8.0,
)
If the topology lacks atom names (only numeric types), supply a mapping:
bins, gr = rdf(
u,
species_a="C",
species_b="O",
atom_types_map={1: "C", 2: "H"}
)
Validation
RDFs for liquid water (64 atoms, 1 ns) computed using a trajectory from Lim et al. (2026) match reference curves for all pairs:
Citation
If you use cuRDF in your work, please cite:
@software{cuRDF,
author = {Hart, Joseph},
title = {cuRDF: CUDA-accelerated radial distribution functions using NVIDIA ALCHEMI Toolkit-Ops O(N) neighbor lists and PyTorch.},
month = dec,
year = 2025,
publisher = {Zenodo},
version = {0.5.0},
doi = {10.5281/zenodo.1085332119},
url = {https://doi.org/10.5281/zenodo.1085332119}
}
Metadata
Release files for cuRDF 0.5.4
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| curdf-0.5.4.tar.gz | 14.4 kB | Without provenance | Details for curdf-0.5.4.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| curdf-0.5.4-py3-none-any.whl | Python 3 | none | any | Details for curdf-0.5.4-py3-none-any.whl |
Total release size: 28.2 kB
Release files / curdf-0.5.4.tar.gz
| Download URL | curdf-0.5.4.tar.gz |
|---|---|
| Size | 14.4 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
24fcb0eb83de8398f8f8a5b605a860dab0fda4fd2837c6f32864dace89740505
|
|
BLAKE2b-256 checksum How to use checksums |
7c2b9df8a8ccda7ee18cf5b347018afb21c111ab2522e3cf8439fe4e7fb5919b
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.3
|
Release files / curdf-0.5.4-py3-none-any.whl
| Download URL | curdf-0.5.4-py3-none-any.whl |
|---|---|
| Size | 13.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
9cd40cd270940966aa4ea37d88828338f4ab5bb68913892b7c3a3bbd9973f300
|
|
BLAKE2b-256 checksum How to use checksums |
0fdc350600fb6e1e245ba0855a55facb05912cd49b51a477fed0f71f628828fe
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.2.0 CPython/3.12.3
|
Security for cuRDF 0.5.4
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.