1. Building from Source

1.1 Prerequisites

Package	Min. version
clang/clang++	19
cmake	4.3.0
ninja	1.11
python	3.10
pybind11	2.12
BLAS + LAPACK	64‑bit ints
HDF5	1.12
OpenMP · OpenCL	latest

1.2 Build with Conda (recommended)

git clone https://github.com/raspa3/raspa3.git
cd raspa3
conda env create -f env.yml
conda activate raspa3
cmake --preset=linux_conda  # or mac_conda / windows_conda_raspa3
ninja -C build
ninja -C build install      # optional

Note: remove the package ocl-icd from env.yml for Mac and Windows installations, as OpenCL is only available for Linux. For MacOS using the system OpenCL is recommended.

Note: it is often necessary to add the conda installed libraries raspa3 links against to your dynamic library path using export LD_LIBRARY_PATH=${CONDA_PREFIX}/lib:$LD_LIBRARY_PATH.

1.3 Linux examples

Ubuntu 24.04

sudo apt install -y git ca-certificates cmake ninja-build \
  llvm lld clang clang-tidy libc++-dev libc++abi-dev \
  libomp-dev libclang-rt-dev python3 python3-dev pybind11-dev \
  liblapack64-dev libblas64-dev
 
git clone https://github.com/raspa3/raspa3.git
cd raspa3
cmake -B build --preset=linux-x86_64-core-avx2-ubuntu-24
ninja -C build
ninja -C build install

Fedora 40

sudo dnf install -y git llvm lld cmake clang clang-tools-extra ninja-build \
  libomp-devel libcxx libcxx-devel libcxxabi libcxxabi-devel \
  lapack64 blas64 python3 python3-devel pybind11-devel
 
git clone https://github.com/raspa3/raspa3.git
cd raspa3
cmake -B build --preset=linux-x86_64-core-avx2-fedora-40
ninja -C build
ninja -C build install

1.4 macOS (Homebrew)

brew install llvm lld libomp hdf5 libaec ninja cmake doxygen graphviz lapack pybind11
export PATH="$(brew --prefix llvm)/bin:$PATH"  # clang‑18
export HDF5_ROOT="$(brew --prefix hdf5)"
git clone https://github.com/raspa3/raspa3.git
cd raspa3
cmake -B build --preset=macos-apple-silicon  # or macos-x64-core-avx2
ninja -C build
ninja -C build install

2. CMake Presets

Run cmake --list-presets to see every preset. Common shortcuts:

Preset	Target
linux_conda	Generic Linux + Conda
linux-x86_64-core-avx2-ubuntu-24	Ubuntu 24, AVX2
linux-x86_64-skylake-avx512-fedora-40	Fedora 40, AVX‑512
mac_conda	macOS + Conda
macos-apple-silicon	macOS Apple Silicon
windows_conda_raspa3	Windows + Conda