Dual-space multilevel kernel-splitting method for Monte Carlo simulation

This is a C++ implementation of the dual-space multilevel kernel-splitting method for Monte Carlo simulation long-range electrostatics systems.

How to use

1. Clone the repository
2. Run git submodule update --init --recursive to clone the submodules
3. Run cmake . -B build to create the build directory
4. Run cmake --build build -j 8 to build the project with 8 threads
5. Run ctest --test-dir build to run the tests

Tips: for usage on rusty, need to run module load gcc fftw openmpi

Julia interface

A lightweight Julia wrapper is provided in the julia/ directory. After building the shared library (libhpdmk), activate the Julia project and construct trees directly from Julia:

using Pkg

Pkg.activate("julia")
using PDMK4MC

params = PDMK4MC.HPDMKParams(L = 20.0)
coords = rand(3, 100)
charges = randn(100)
tree = PDMK4MC.create_tree(coords, charges; params=params)
energy = PDMK4MC.eval_energy(tree)

The wrapper expects MPI.jl to load an Open MPI runtime so that its communicators remain compatible with libhpdmk. Configure MPI.jl once via MPIPreferences.jl before importing PDMK4MC:

julia> using MPIPreferences

julia> MPIPreferences.use_jll_binary("OpenMPI_jll")

By default the wrapper looks for libhpdmk using the standard library search path. Set the environment variable HPDMK_LIBRARY to point to the shared library if it lives in a non-standard location. The bindings ask libhpdmk itself to initialise MPI so that Julia and the native library always share the same MPI runtime; the optional comm keyword can be left as nothing to use the library's MPI_COMM_WORLD, or set to an existing communicator (for example MPI.COMM_WORLD) so long as it comes from the same Open MPI installation. The precision keyword selects either Float64 (default) or Float32 computations.