cuRDF documentation

cuRDF computes GPU-accelerated radial distribution functions from ASE structures and MDAnalysis trajectories using NVIDIA ALCHEMI Toolkit-Ops and PyTorch.

Start with the quickstart to install the package and compute a curve. The methods page explains normalization and periodic boundaries. The API reference lists parameters and output formats, and troubleshooting covers setup and input errors.

Contents:

• Quickstart
  • Installation
  • A single ASE structure
  • An ASE trajectory
  • An MDAnalysis trajectory
  • Saving and reusing results
• Methods
  • What the RDF measures
  • Radial bins
  • Same-species normalization
  • Cross-species normalization
  • Trajectory averaging
  • Cells and periodic boundaries
  • Numerical and performance controls
• API
  • rdf: structures and trajectories
  • compute_rdf: one frame of coordinates
  • accumulate_rdf: a frame iterator
  • plot_rdf: plot a saved or computed curve
• Troubleshooting
  • CUDA is not available
  • A species selection is empty
  • A cell is missing or invalid
  • No frames or no possible pairs
  • Neighbour capacity or GPU memory is insufficient
  • The RDF differs from another package
  • A saved file or plot is missing
  • What the CI tests cover

Citation

• DOI: https://doi.org/10.5281/zenodo.20613283

• See CITATION.cff in the repo for citation metadata.