facebookincubator/cinder

★ 3,785PythonAudience · researcherComplexity · 5/5LicenseSetup · hard

Mindmap

mindmap
  root((Cinder))
    What it is
      CPython fork
      Meta-maintained
      Python 3.14
    Purpose
      Performance research
      Scale optimization
      Upstream proposals
    Building
      configure script
      make
      PGO and LTO
    Related
      cinderx JIT
      Official CPython

mindmap root((Cinder)) What it is CPython fork Meta-maintained Python 3.14 Purpose Performance research Scale optimization Upstream proposals Building configure script make PGO and LTO Related cinderx JIT Official CPython

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Study how Meta modifies the Python interpreter to gain performance at scale by comparing this fork to upstream CPython.

USE CASE 2

Build and test a customized Python runtime fork using the same configure-and-make workflow as standard CPython.

USE CASE 3

Research Python runtime internals and performance optimization techniques used in large production deployments.

Tech stack

PythonCCPython

Getting it running

Difficulty · hard Time to first run · 1day+

Building from source requires a full C toolchain and familiarity with CPython internals, this is a research reference, not a drop-in Python install.

Free to use, modify, and distribute under Python's standard open-source license with no commercial restrictions.

In plain English

Cinder is Meta's internal fork of CPython, which is the standard Python interpreter that most people and servers use when they run Python code. Meta (formerly Facebook) maintains this separate copy so they can apply performance improvements and experimental changes without waiting for those changes to be accepted into the official Python project. The name "cinder" is historical. The actual JIT compiler and performance extensions that made this fork notable have since been moved to a separate companion repository called cinderx. What remains here is the modified Python runtime itself, currently tracking Python version 3.14. From a technical standpoint, this repository behaves like the standard CPython source. You can build it the same way: running a configure script and then make on Unix-like systems, or following Windows-specific instructions for that platform. It supports the same optimization flags, such as Profile Guided Optimization and Link Time Optimization, that the main CPython project offers. For most developers, this repository is more useful as a reference than as something to run directly. If you want to understand how a large technology company modifies and extends an open-source language runtime to squeeze out more performance at scale, studying the differences between this fork and the upstream CPython project is informative. If you are looking for the JIT compiler work specifically, the cinderx repository is the right place to look instead. Meta runs Python at an enormous scale internally, which means even small interpreter-level gains translate into significant infrastructure savings. This fork represents the company's ongoing investment in Python performance research, and over time, some of the ideas developed here have fed back into proposals and patches for the official CPython project that everyone uses.

Copy-paste prompts

Prompt 1

Show me how to build the Cinder CPython fork from source on a Linux machine, including the configure and make steps with PGO enabled.

Prompt 2

What are the main differences between Meta's Cinder fork and standard CPython, and which changes have been proposed upstream?

Prompt 3

How does Cinder's JIT compiler work, and where should I look in the cinderx companion repo to find that code?

Prompt 4

I want to study Python interpreter performance optimization, what should I read first in the Cinder source to understand Meta's approach?

Open on GitHub → Explain another repo

← facebookincubator on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.