Pyston 2.3.5 Boosts Python Performance with Major CPython Compatibility

Pyston 2.3.5 has been released, delivering significant performance improvements and enhanced compatibility with upstream CPython.

Pyston was originally developed by Dropbox as a fast open‑source Python implementation using the LLVM JIT framework. Although Dropbox stopped supporting it years ago, the team continued maintenance and joined Anaconda to keep advancing high‑performance Python.

The new version, forked from CPython 3.8, aims to stay highly compatible while providing double‑digit percentage performance optimizations over the upstream.

The release can be downloaded from GitHub: https://github.com/pyston/pyston/releases/tag/pyston_2.3.5 Pyston v2 focuses on optimizing network‑service workloads, offering notable acceleration across many benchmarks. A new public Python macro‑benchmark suite was created to evaluate common Python projects. In micro‑benchmarks such as chaos.py and nbody.py, Pyston v2 runs about twice as fast as standard Python.

Test results on a target benchmark (djangocms + flaskblogging) show an average latency improvement of 1.22×, a p99 latency improvement of 1.18×, and only a few megabytes of additional memory per process.

The development team highlighted several technologies used in Pyston v2:

Low‑overhead JIT using DynASM

Quickening

Standard CPython optimizations

Improved build process

CPython 3 is typically much slower than compiled languages like C, Go, or JIT‑compiled languages such as Java and JavaScript.

Because Pyston is a CPython fork, it is one of the most compatible alternative Python implementations available today, supporting all CPython features and the C API.

References:

https://www.infoworld.com/article/3618169/pyston-project-open-sources-its-faster-python.html

https://www.theregister.com/2021/05/06/the_quest_for_faster_python/

https://mp.weixin.qq.com/s/QRX7NSQrNHgwC0A_EQji7w

https://www.bullfrag.com/pyston-the-version-of-python-that-promises-to-be-30-faster-is-going-to-gain-traction-with-the-support-of-anaconda-who-has-hired-his-team/