NULAPACK: NUmerical Linear Algebra PACKage

NULAPACK is a lightweight, high-performance numerical linear algebra library. It provides a set of core subroutines implemented in Fortran for efficiency, with convenient C++ and Python interfaces.

Installation

nulapack can be installed from PyPI or built from source. Using uv is recommended for fast and reproducible environments.

---

Install from PyPI

Using pip

System-wide (user install):

pip install --upgrade --user nulapack

You may need to use pip3 instead of pip depending on your Python installation.

Inside a virtual environment:

python -m venv .venv
source .venv/bin/activate
pip install --upgrade nulapack

---

Using uv (recommended)

Install into a new environment:

uv venv
source .venv/bin/activate
uv pip install nulapack

Add to an existing uv project:

uv add nulapack
uv sync

---

Using pipenv

pipenv install nulapack

---

Using poetry

poetry add nulapack

---

Using hatch

hatch add nulapack

---

Declaring as a Dependency

pyproject.toml:

[project.dependencies]
nulapack = ">=0.1.0"

requirements.txt:

nulapack>=0.1.0

---

Building from Source

Building from source is useful if you want the latest features or need to modify the Fortran core.

Prerequisites

• CMake ≥ 3.10
• Fortran compiler (e.g. gfortran)
• C/C++ compiler (e.g. gcc, clang)
• Python ≥ 3.9
• uv

Build and install

git clone https://github.com/eggzec/NULAPACK.git
cd NULAPACK

uv venv
source .venv/bin/activate

uv run bin/build.py develop --with-py

---

Usage

Python

import numpy as np
from nulapack import doolittle

A = np.array([[4, 3], [6, 3]], dtype=np.float64)
L, U, info = doolittle(A)

if info == 0:
    print("L:\n", L)
    print("U:\n", U)

C++

#include <iostream>
#include <vector>
#include "Doolittle.h"

int main() {
    int n = 2;
    std::vector<double> a = {4, 3, 6, 3};
    std::vector<double> l(n * n, 0.0);
    std::vector<double> u(n * n, 0.0);
    int info;

    doolittle(&n, a.data(), l.data(), u.data(), &info);

    if (info == 0) {
        // Use l and u
    }
    return 0;
}