Thanks to visit codestin.com
Credit goes to Github.com

Skip to content
/ cyfaust Public

A cross-platform cython wrapper of the faust interpreter with rtaudio.

License

Notifications You must be signed in to change notification settings

shakfu/cyfaust

Repository files navigation

cyfaust

The cyfaust project provides a minimal, modular, self-contained, cross-platform cython wrapper of the Faust DSP language and the RtAudio cross-platform audio driver.

  • Dual Backends: Interpreter (default, ~8MB) or LLVM JIT (~71MB, faster execution)
  • Box API: Fully wrapped with both functional and object-oriented interfaces
  • Signal API: Fully wrapped with both functional and object-oriented interfaces
  • Platform Support: macOS, Linux, and Windows
  • Build Variants: Dynamic (libfaust.so|dylib) and static (libfaust.a|lib)
  • Faust Version: 2.83.1

Installation

To install cyfaust with the faust interpreter backend (default):

pip install cyfaust

To install with the LLVM JIT backend:

pip install cyfaust-llvm

Features

  • Python-oriented cross-platform implementation of the Faust DSP language

  • Two backend options:

    • Interpreter (default): Compact ~8MB binary, good for development and most use cases
    • LLVM JIT: ~71MB binary with native code compilation for maximum performance
  • Provides the following submodules (in the default build):

    • cyfaust.interp: wraps the faust interpreter and the rtaudio audio driver

    • cyfaust.box: wraps the faust box api

    • cyfaust.signal: wraps the faust signal api

    • cyfaust.common: common utilities and classes

  • Self-contained, minimal, and modular design

  • Default build uses interpreter backend to keep binary size low (~8MB vs ~71MB with LLVM)

  • Can generate code using the following backends:

    • c++
    • c
    • rust
    • codebox
  • Can generate auxiliary files such as svg block diagrams of dsp

  • Dual functional/oo design for box and signal apis with minimal code duplication.

  • Implements Memorviews for read/write buffer using numpy or array.array.

  • Both dynamic and static build variants can be packaged as a self-contained python wheel.

  • Includes several github workflows to automate the testing and building of cyfaust wheels for a number of supported platforms

  • Command-line interface for common operations

Command-Line Interface

cyfaust provides a CLI accessible via the cyfaust command or python -m cyfaust:

cyfaust <command> [options]
# or: python -m cyfaust <command> [options]

Commands

version - Show cyfaust and libfaust version information:

cyfaust version

info - Display DSP metadata, inputs, outputs, and dependencies:

cyfaust info synth.dsp

compile - Compile Faust DSP to target backend (cpp, c, rust, codebox):

cyfaust compile synth.dsp -b cpp -o synth.cpp
cyfaust compile synth.dsp -b rust -o synth.rs

expand - Expand Faust DSP to self-contained code with all imports resolved:

cyfaust expand filter.dsp -o filter_expanded.dsp
cyfaust expand filter.dsp --sha-only  # output only SHA1 key

diagram - Generate SVG block diagrams:

cyfaust diagram synth.dsp -o output_dir

play - Play a DSP file with RtAudio:

cyfaust play osc.dsp              # play until Ctrl+C
cyfaust play osc.dsp -d 5         # play for 5 seconds
cyfaust play osc.dsp -r 48000     # use 48kHz sample rate

params - List all DSP parameters (sliders, buttons, etc.):

cyfaust params synth.dsp

validate - Check a DSP file for errors:

cyfaust validate filter.dsp
cyfaust validate filter.dsp --strict  # treat warnings as errors

bitcode - Save/load compiled DSP as bitcode for faster loading:

cyfaust bitcode save synth.dsp -o synth.fbc
cyfaust bitcode load synth.fbc

json - Export DSP metadata as JSON:

cyfaust json instrument.dsp --pretty
cyfaust json instrument.dsp -o metadata.json

For detailed help on any command:

cyfaust <command> --help

Building

It has two build variants:

  1. The default build is dynamically linked to libfaust.dylib, libfaust.so or faust.dll depending on your platform and consists of a python package with multiple compiled submodules and embedded resources (faust libraries and architecture files):

    % tree -L 3
    .
    └── cyfaust
        ├── __init__.py
        ├── box.cpython-311-darwin.so
        ├── common.cpython-311-darwin.so
        ├── interp.cpython-311-darwin.so
        ├── signal.cpython-311-darwin.so
        └── resources
            ├── architecture
            └── libraries
  2. The static build is statically linked (with libfaust.a or libfaust.lib) and consists of a python package with a single compiled submodule and embedded resources (faust libraries and architecture files):

    % tree -L 3
    .
    └── cyfaust
        ├── __init__.py
        ├── cyfaust.cpython-311-darwin.so
        └── resources
            ├── architecture
            └── libraries

While this project was initially developed and tested primarily on macOS (x86_64, arm64), Linux (amd64, aarch64) and Windows (amd64) are now supported in recent releases.

See the project's TODO for remaining enhancements.

Setup and Requirements

cyfaust has a build management script, scripts/manage.py, which simplifies cross-platform project setup and builds and also build automation in the case of github workflows.

usage: manage.py [-h] [-v]  ...

cyfaust build manager

options:
  -h, --help     show this help message and exit
  -v, --version  show program's version number and exit

subcommands:
  valid subcommands

                 additional help
    build        build packages
    clean        clean detritus
    setup        setup prerequisites
    test         test modules
    wheel        build wheels

A brief guide to its use is provided in the following table:

# platform step command
1* common install prerequisites python3 scripts/manage.py setup --deps
2 common build/install faustlib python3 scripts/manage.py setup --faust
3 common build cyfaust (dynamic) python3 scripts/manage.py build
4 common build cyfaust (static) python3 scripts/manage.py build --static
5 common test cyfaust python3 scripts/manage.py test or pytest
6 common build cyfaust wheels python3 scripts/manage.py wheel --release
7 common test cyfaust wheels python3 scripts/manage.py wheel --test

[*] Prerequisites consists of general and platform-specific requirements.

The general requirements are:

  1. libfaust configured to be built with the c, c++, codebox, interp_comp, and rust backends and the executable, static and dynamic targets.

  2. libsndfile and libsamplerate for faust soundfile primitive support

Platform specific requirements are covered in the next sections.

Windows

To build cyfaust you will need python to be installed (3.10+), a c++ compiler such as visual studio community edition and make sure to install c++ and Windows SDK development support, as well as git, and cmake.

Then do something like the following in a terminal:

git clone https://github.com/shakfu/cyfaust.git
cd cyfaust
pip install -r requirements.txt # or python3 scripts/manage.py setup --deps
python scripts/manage.py setup --faust
# then pick from 3, 4 or 6. For example
python scripts/manage.py build
pytest
# etc..

macOS & Linux

On macOS and Linux, a Makefile is available as a frontend to the above manage.py script to make it a little quicker to use.

The following setup sequence is illustrative of this and is also useful if you want to setup cyfaust manually.

For macOS, having Xcode or the CommandLine tools (xcode-select --install) installed is required, and then you will need to have python and cmake installed. If you use Homebrew, this is simply:

brew install python cmake

For Linux, if you are on a Debian-derived system, you will need something like the following:

sudo apt update
sudo apt install build-essential git cmake python3-dev # you probably have these already
sudo apt install libasound2-dev patchelf # for alsa sound driver and fixing dynamic libs
sudo apt install libsndfile1-dev libsamplerate0-dev # specifically for faust

Then

git clone https://github.com/shakfu/cyfaust.git
cd cyfaust
make # or python3 scripts/manage.py setup --faust
  • This will download faust into the build directory, then configure (and patch) it for an interpreter build, build it, and install it into the following (.gitignored) folders in the project directory:

    • bin, containing the faust executables,
    • lib, the dynamic and static versions of libfaust and
    • share, the faust standard libs and examples.
  • Faust version 2.83.1 will be used as it is a stable reference to work with and is used by the setup scripts.

  • The script can be run again and will create (and overwrite) the corresponding files in the bin, include, lib and share folders.

To build the default dynamically-linked package and/or wheel:

make

or using uv (the project now uses scikit-build-core):

uv build --wheel

For a wheel:

make wheel

For the static variant, set the STATIC=1 environment variable with make, or use CMAKE_ARGS for direct builds:

make STATIC=1

or

CMAKE_ARGS="-DSTATIC=ON" uv build --wheel

LLVM Backend (cyfaust-llvm)

The LLVM backend compiles Faust DSP code to native machine code via LLVM JIT, offering significantly faster execution compared to the interpreter backend. It is distributed as a separate package: cyfaust-llvm.

Interpreter vs LLVM

Feature cyfaust (interpreter) cyfaust-llvm
Binary size ~8MB ~33MB
DSP execution Bytecode interpreter Native machine code
Compilation speed Fast Slower (JIT compilation)
Runtime performance Good Best
Use case Development, lightweight deployment Production, performance-critical

Installation

Install from PyPI:

pip install cyfaust-llvm

Both packages provide the same API via from cyfaust import cyfaust - only the backend differs.

Building from Source

Prerequisites

macOS:

# Install Xcode CommandLine tools if not already installed
xcode-select --install

# Install dependencies via Homebrew
brew install python cmake

Linux (Debian/Ubuntu):

sudo apt update
sudo apt install build-essential git cmake python3-dev
sudo apt install libasound2-dev patchelf
sudo apt install libsndfile1-dev libsamplerate0-dev
sudo apt install libtinfo-dev  # Required for LLVM
Build Steps (macOS & Linux)
  1. Clone the repository:

    git clone https://github.com/shakfu/cyfaust.git
    cd cyfaust
  2. Download the prebuilt LLVM-enabled libfaust:

    # Using make
    make faustwithllvm
    
    # Or using manage.py directly
    python3 scripts/manage.py setup --llvm

    This downloads libfaustwithllvm.a (~300MB) from the official Faust releases:

    • macOS: Downloads DMG for your architecture (arm64 or x86_64)
    • Linux: Downloads zip for your architecture (x86_64 or aarch64)
  3. Build for development:

    make build-llvm
  4. Run tests:

    make test-llvm
Building Wheels

To build a distributable wheel:

make wheel-llvm

This creates dist/cyfaust_llvm-<version>-<platform>.whl.

Or build directly with uv/cmake:

# First ensure libfaustwithllvm.a is downloaded
make faustwithllvm

# Generate static sources
python3 scripts/generate_static.py

# Build wheel
CMAKE_ARGS="-DSTATIC=ON -DLLVM=ON" uv build --wheel

# Rename to cyfaust-llvm (required for PyPI)
python3 scripts/rename_wheel.py dist/cyfaust-*.whl cyfaust-llvm --delete

Usage

The API is identical regardless of backend:

from cyfaust import cyfaust

# Create a DSP factory from Faust code
factory = cyfaust.create_dsp_factory_from_string(
    "volume",
    "process = *(hslider(\"gain\", 0.5, 0, 1, 0.01));"
)

# Create DSP instance
dsp = factory.create_dsp_instance()
print(f"Inputs: {dsp.get_num_inputs()}, Outputs: {dsp.get_num_outputs()}")

Detecting Backend at Runtime

import cyfaust

# Check which backend is available
print(f"LLVM Backend: {cyfaust.LLVM_BACKEND}")

if cyfaust.LLVM_BACKEND:
    # LLVM-specific: get target machine info
    print(cyfaust.get_dsp_machine_target())
    # e.g., 'x86_64-pc-linux-gnu' or 'arm64-apple-darwin24.6.0'

Platform Support

Platform Architecture Status
macOS arm64 (Apple Silicon) Supported
macOS x86_64 (Intel) Supported
Linux x86_64 Supported
Linux aarch64 Supported
Windows x86_64 Planned

To run the tests

make test

or

make pytest

Prior Art of Faust + Python

  • DawDreamer by David Braun: Digital Audio Workstation with Python; VST instruments/effects, parameter automation, FAUST, JAX, Warp Markers, and JUCE processors. Full-featured and well-maintained. Use this for actual work! (pybind11)

  • faust_python by Marc Joliet: A Python FAUST wrapper implemented using the CFFI (updated 2015). There's a more recent (updated in 2019) fork by Patrik Lechner. (cffi)

  • pyfaust by Alexandru Stan: Embed Faust DSP Programs in Python. (cffi)

  • faust-ctypes: a port of Marc Joliet's FaustPy from CFFI to Ctypes. (ctypes)

  • faustpp: A post-processor for faust, which enables more flexible code generation.