fftlog-lss 0.1.2

JAX-accelerated FFTLog implementation for Large Scale Structure cosmology

FFTLog

A Python implementation of FFTLog for fast logarithmic FFT transforms, developed for PyBird.

Installation

Basic installation (recommended)

pip install fftlog-lss

With JAX acceleration

pip install "fftlog-lss[jax]"

From source

git clone https://github.com/pierrexyz/fftlog.git
cd fftlog
pip install --editable .

From source with JAX

git clone https://github.com/pierrexyz/fftlog.git
cd fftlog
pip install --editable ".[jax]"

Dependencies

Required

• Python >= 3.8
• numpy >= 1.20.0
• scipy >= 1.7.0

Optional (for JAX acceleration)

• jax >= 0.4.0
• jaxlib >= 0.4.0
• interpax >= 0.3.0

Quickstart

Basic Usage (without JAX)

import numpy as np
from scipy.stats import lognorm
from fftlog import FFTLog

# Create test function (power spectrum)
k = np.logspace(-4, 0, 200)
pk = lognorm.pdf(k, 2.1)

# Initialize FFTLog
fft = FFTLog(
    Nmax=512,       # Number of points
    xmin=1e-5,      # Minimum k value
    xmax=1e3,       # Maximum k value
    bias=-0.1,      # Bias parameter
    complex=False,  # Use real FFT
    window=0.1      # Anti-aliasing window
)

# FFTLog decomposition and reconstruction
pk_reconstructed = fft.rec(k, pk, k)

# Spherical Bessel Transform
s = np.arange(1., 1e3, 5.)
xi = fft.sbt(k, pk, s)  # Fast O(N log N) transform

With JAX Acceleration

import numpy as np
from scipy.stats import lognorm
from fftlog import FFTLog
from fftlog.config import set_jax_enabled
from jax import jit
import jax.numpy as jnp

# Enable JAX mode
set_jax_enabled(True)

# Create test function
k = np.logspace(-4, 0, 200)
pk = lognorm.pdf(k, 2.1)

# Initialize FFTLog (same as above)
fft = FFTLog(
    Nmax=512, xmin=1e-5, xmax=1e3, 
    bias=-0.1, complex=False, window=0.1
)

# Convert to JAX arrays and JIT compile
k_jax, pk_jax = jnp.array(k), jnp.array(pk)
get_coef_jit = jit(fft.Coef)

# Now much faster for repeated calls
coefficients = get_coef_jit(k_jax, pk_jax)

Features

• Fast logarithmic FFT transforms
• Support for both real and complex transforms
• Spherical Bessel transforms
• Anti-aliasing windows
• Optional JAX acceleration for GPU/TPU support

Documentation

For more detailed examples and documentation, see the notebooks in the notebooks/ directory.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.