CuNESSie.jl is an extension to the NESSie.jl package, providing CUDA-accelerated drop-in replacements for the package's numerical solvers and post-processors.

This package version requires Julia 1.8 or above. In the Julia shell, switch to the Pkg shell by pressing ] and enter the following command:

pkg> add https://github.com/tkemmer/NESSie.jl
pkg> add https://github.com/tkemmer/CuNESSie.jl

The basic usage of this package is the same as for NESSie.jl. Just replace the NESSie.BEM module by CuNESSie in your code and you're ready to go:

using NESSie
using CuNESSie  # before: NESSie.BEM
using NESSie.Format: readoff, readpqr

# I. Create model
model           = readoff("data/born/na.off")
model.charges   = readpqr("data/born/na.pqr")
model.params.εΩ = 1   # dielectric constant for vacuum model
model.params.εΣ = 78  # dielectric constant for water

# II. Apply nonlocal solver
bem = solve(NonlocalES, model)  # <-- CUDA-accelerated solver

# III. Apply postprocessor
val = rfenergy(bem)             # <-- CUDA-accelerated post-processor
println("Reaction field energy: $val kJ/mol")

CuNESSie.jl reuses the system models and solver results from NESSie.jl, so the local and nonlocal BEM solvers as well as the corresponding post-processors from both packages can be interchanged freely.

CuNESSie.jl provides tests for most of its functions. You can run the test suite with the following command in the Pkg shell:

pkg> test CuNESSie

If you use CuNESSie.jl in your research, please cite the following publications:

T. Kemmer, S. Hack, B. Schmidt, A. Hildebrandt. CUDA-accelerated protein electrostatics in linear space. Journal of Computational Science 70 (2023) 102022. https://doi.org/10.1016/j.jocs.2023.102022

T. Kemmer. Space-efficient and exact system representations for the nonlocal protein electrostatics problem. Ph. D. thesis (2021), Johannes Gutenberg University Mainz. Mainz, Germany. https://doi.org/10.25358/openscience-5689

Citation items for BibTeX can be found in CITATION.bib.