# How fast can we make interpreted Python?

**Authors:** Russell Power, Alex Rubinsteyn  
**arXiv:** [1306.6047](https://arxiv.org/abs/1306.6047)  
**Date:** 2013-06-25  
**Summary page:** https://www.summarizepaper.com/en/paper/1306.6047v2/

## Key points

- Python is a widely used dynamic language with extensive libraries and extension modules
- Python is used for web development with Django and numerical analysis using NumPy
- Python's performance falls short compared to more modern languages like Lua and JavaScript
- Researchers Russell Power and Alex Rubinsteyn designed Falcon to address Python's performance issues while retaining access to existing extension libraries
- Falcon is a high-performance bytecode interpreter fully compatible with the standard CPython interpreter
- Falcon aims to accelerate the execution of Python bytecode by incorporating optimizations and introducing new techniques
- Falcon was found to be an average of 25% faster than the standard Python interpreter across most benchmarks, with some cases showing a speed improvement of about 2.5 times faster than traditional Python execution

## Summary

Python is a widely used dynamic language known for its extensive libraries and extension modules that enhance its functionality for various tasks. These include web development with Django and numerical analysis using NumPy. However, when compared to more modern languages like Lua and JavaScript, Python's performance falls short. This can be attributed to the complexity introduced by the same API and extension libraries that make Python versatile and powerful. In an effort to address this issue while retaining access to existing extension libraries, researchers Russell Power and Alex Rubinsteyn designed and implemented Falcon. It is a high-performance bytecode interpreter that remains fully compatible with the standard CPython interpreter. By incorporating well-known optimizations and introducing new techniques, Falcon aims to accelerate the execution of Python bytecode. In their evaluation, Power and Rubinsteyn found Falcon to be an average of 25% faster than the standard Python interpreter across most benchmarks. In some cases, it even demonstrated a remarkable speed improvement of about 2.5 times faster than traditional Python execution. This research sheds light on the potential for enhancing Python's performance without sacrificing access to its rich ecosystem of libraries and extensions.