Installation¶

This guide covers installation for all supported languages.

Prerequisites¶

Core (CPU)¶

Requirement	Version
C compiler	GCC or Clang with C11 support
CMake	≥ 3.14
OpenMP	Optional, auto-detected by CMake

A C++17 compiler is additionally required for the libdress++ (C++) wrapper.

CUDA (GPU acceleration)¶

Requirement	Details
NVIDIA GPU	Compute-capable device
CUDA Toolkit	Provides `nvcc` and `libcudart`

CUDA is built separately from the core library, producing libdress_cuda.so. Build it with:

./build.sh cuda
# or manually:
make -C libdress/src/cuda

MPI (distributed computing)¶

Requirement	Details
MPI implementation	OpenMPI or MPICH
`mpicc`	Must be on `PATH`

For C / C++ via CMake, pass -DDRESS_MPI=ON:

cmake -S . -B build -DDRESS_MPI=ON
cmake --build build

The CUDA Makefile auto-detects mpicc and includes MPI+CUDA support when available.

Per-language extras¶

Language	CUDA requirement	MPI requirement
C / C++	Link against `libdress_cuda.so` + `-lcudart`	CMake flag `-DDRESS_MPI=ON`
Python	`libdress_cuda.so` on `LD_LIBRARY_PATH`	`pip install mpi4py`; run with `mpirun`
Rust	`libdress_cuda.so` on `LD_LIBRARY_PATH`	Cargo feature `mpi`; `mpicc` on PATH
Go	CGO: `libdress_cuda.so` + `-lcudart`	CGO: `-lmpi`; MPI headers on include path
Julia	`libdress_cuda.so` in `julia/` dir (or on path)	`MPI.jl` package; `libdress.so` with MPI symbols
R	`configure` auto-detects `libdress_cuda.so`	`configure` auto-detects `mpicc`
Octave	`libdress_cuda.so` on `LD_LIBRARY_PATH`; `nvcc` at build time	Not supported (CPU only)
WASM	Not supported (CPU only)	Not supported (CPU only)

Python¶

From PyPI (Remote)¶

pip install dress-graph

From conda-forge¶

conda install -c conda-forge dress-graph

From Source¶

cd python
pip install .

To install in editable mode for development:

pip install -e .

Rust¶

From Crates.io (Remote)¶

cargo add dress-graph

From Source¶

Add the local path to your Cargo.toml:

[dependencies]
dress-graph = { path = "path/to/dress-graph/rust" }

JavaScript / WASM¶

From npm (Remote)¶

npm install dress-graph

From Source¶

Build the WASM package first:

./build.sh wasm

Then install from the local directory:

cd wasm
npm install .

Or link it:

cd wasm
npm link
# In your project:
npm link dress-graph

Julia¶

From GitHub (Remote)¶

using Pkg
Pkg.add(url="https://github.com/velicast/dress-graph", subdir="julia")

From Source¶

using Pkg
Pkg.develop(path="julia")

R¶

From CRAN (Remote)¶

CRAN hosts a stable release. For the latest version, install from GitHub.

install.packages("dress.graph")

From GitHub (Remote)¶

# install.packages("remotes")
remotes::install_github("velicast/dress-graph", subdir="r")

From Source¶

R CMD INSTALL r

Or from within R:

install.packages("./r", repos = NULL, type = "source")

Go¶

From GitHub (Remote)¶

go get github.com/velicast/dress-graph/go

From Source¶

If you have the repository cloned locally, you can use a replace directive in your go.mod:

go mod edit -replace github.com/velicast/dress-graph/go=../path/to/dress-graph/go

C / C++¶

Homebrew (macOS / Linux)¶

brew tap velicast/dress-graph
brew install dress-graph

vcpkg (Overlay Port)¶

If you use vcpkg, you can use the vcpkg/ directory in this repository as an overlay port.

vcpkg install dress-graph --overlay-ports=/path/to/dress-graph/vcpkg

From Source (CMake)¶

You can build the library using CMake directly:

cmake -S . -B build -DCMAKE_BUILD_TYPE=Release
cmake --build build --target install

From Source (Script)¶

The provided build script handles libdress (C), libdress++ (C++), and libdress-igraph.

./build.sh c cpp igraph

MATLAB / Octave¶

From Remote (Octave)¶

pkg install "https://github.com/velicast/dress-graph/releases/download/v0.8.1/dress-graph-0.8.1.tar.gz"

From Source (Octave)¶

Build the tarball and install locally:

./build.sh octave

pkg install dress-graph-0.8.1.tar.gz
pkg load dress-graph

From Source (MATLAB)¶

Add the matlab directory to your MATLAB path:

addpath('path/to/dress-graph/matlab');
savepath;