Installation¶

Basic Install¶

Clone the repository and install with pip:

git clone https://github.com/taladjidi/NLSE.git
cd NLSE
pip install .

For development (includes pytest, ruff, mypy, etc.):

pip install -e ".[dev]"

Requirements¶

Python: 3.10 or later
Platforms: Linux, macOS, Windows

Core dependencies (numpy, scipy, matplotlib, numba, tqdm) are installed automatically.

GPU Backends¶

By default, NLSE uses the CPU backend (NumPy + Numba, with scipy.fft for the transform). For GPU acceleration, install one or more of the following:

CUPY (NVIDIA CUDA)¶

For NVIDIA GPUs with CUDA support:

pip install cupy-cuda12x   # for CUDA 12.x
# or
pip install cupy-cuda11x   # for CUDA 11.x

See the CuPy installation guide for details.

OpenCL (Cross-platform GPU)¶

For AMD, Intel, or NVIDIA GPUs via OpenCL:

pip install pyopencl

You also need an OpenCL runtime installed for your GPU. On macOS, OpenCL is available by default.

MLX (Apple Silicon)¶

For Apple Silicon Macs (M1/M2/M3/M4):

pip install mlx

MLX provides native GPU acceleration on Apple hardware.

Backend Selection¶

NLSE auto-detects the best available backend. The priority order is:

CUPY (if CuPy is installed and a CUDA GPU is available)
MLX (if MLX is installed on Apple Silicon)
CPU (always available)

You can override this with the NLSE_BACKEND environment variable:

export NLSE_BACKEND=CPU   # force CPU backend

See the Backends page for more details.

The CPU transform¶

The CPU backend transforms with scipy.fft (pocketfft), which needs no planning, no wisdom file and no configuration. It uses every core by default. To limit that, cap the thread pool scipy draws on:

import scipy.fft

with scipy.fft.set_workers(4):
    E_out = simu.out_field(E_in, L)

Running Tests¶

pytest tests/               # run all tests
pytest tests/ -v --tb=short  # verbose with short tracebacks