parRSB.py

Python wrappers for parRSB - a library for parallel graph partitioning using Recursive Spectral Bisection (RSB).

Overview

parRSB.py provides Cython-based Python bindings for the parRSB C library. It exposes functionality for:

• Partitioning graphs using RSB and other algorithms.
• Reading Nek5000/NekRS meshes from .re2 files.
• Calculating mesh connectivity based on node coordinates.

The library is built on top of:

• parRSB - C library for parallel mesh partitioning.
• gslib - Highly scalable communication library.
• mpi4py - MPI for Python.
• NumPy - for array handling.

Requirements

System Dependencies

• A C compiler with MPI support (e.g., mpicc via OpenMPI)
• CMake (>= 3.21)
• A Fortran compiler (Only if you want to run the tests)

On Ubuntu/Debian, install them with:

sudo apt -y update
sudo apt install -y openmpi-bin libopenmpi-dev
sudo apt install -y cmake
sudo apt install -y gfortran

Python Requirements

• Python >= 3.10
• uv for virtual environment and package management

Install uv by following the instructions here.

Installation

1. Clone and build parRSB

First, build and install the parRSB C library into a virtual environment directory:

uv venv .venv
git clone https://github.com/thilinarmtb/parRSB.git -b general_graph
cd parRSB
cmake -B build -S . -DCMAKE_INSTALL_PREFIX=../.venv
cmake --build build --target install
cd ..

2. Build and install parRSB.py

source .venv/bin/activate
uv pip install . --target .venv -Ccmake.define.parRSB_DIR=.venv

Note: The CC=mpicc environment variable should be set so that the C extension is compiled with MPI support.