Python,
Turbocharged

Exaloop’s Codon framework compiles Python to native machine code for 10-100x performance gains, multithreading, GPU acceleration & more. Get C performance while staying in Python.

Get started with Codon in seconds.
/bin/bash -c "$(curl -fsSL https://exaloop.io/install.sh)"
Copy to clipboard icon
Easy install:
01.
Open your terminal
02.
Run the above command
03.
Follow the prompts

Exaloop develops tools that empower anyone with basic Python programming experience to write scalable code.

codon benchmarks

Python simplicity, C performance

Benchmark (speedup over Python)

macOS
Linux
float
go
nbody
chaos
spectral_norm
set_partition
primes
binary_trees
fannkuch
word_count
taq

float — Performs a series of 3-dimensional vector operations and normalizations.

View on GitHub

go — Simulates and optimizes a game of Go.

View on GitHub

nbody — Simulates the movement of several celestial objects.

View on GitHub

chaos — Generates chaos game-like fractals.

View on GitHub

spectral_norm — Computes the largest singular value of a particular infinite matrix.

View on GitHub

set_partition — Computes all partitions of a set that fit certain conditions.

View on GitHub

primes — Counts prime numbers below a given threshold. Codon version is multi-threaded.

View on GitHub

binary_trees — Creates and traverses a number of binary trees.

View on GitHub

fannkuch — Performs, for each permutation of a list, a number of element reversals until a certain condition is met, and computes the maximum number of reversals needed across all permutations. Codon version is multi-threaded.

View on GitHub

word_count — Counts occurrences of words in a file using a dictionary.

View on GitHub

taq — Performs volume peak detection on an NYSE TAQ.

View on GitHub
float
go
nbody
chaos
spectral_norm
set_partition
primes
binary_trees
fannkuch
word_count
taq
mandelbrot

float — Performs a series of 3-dimensional vector operations and normalizations.

View on GitHub

go — Simulates and optimizes a game of Go.

View on GitHub

nbody — Simulates the movement of several celestial objects.

View on GitHub

chaos — Generates chaos game-like fractals.

View on GitHub

spectral_norm — Computes the largest singular value of a particular infinite matrix.

View on GitHub

set_partition — Computes all partitions of a set that fit certain conditions.

View on GitHub

primes — Counts prime numbers below a given threshold. Codon version is multi-threaded.

View on GitHub

binary_trees — Creates and traverses a number of binary trees.

View on GitHub

fannkuch — Performs, for each permutation of a list, a number of element reversals until a certain condition is met, and computes the maximum number of reversals needed across all permutations. Codon version is multi-threaded.

View on GitHub

word_count — Counts occurrences of words in a file using a dictionary.

View on GitHub

taq — Performs volume peak detection on an NYSE TAQ.

View on GitHub

mandelbrot — Computes the Mandelbrot set. Codon version is GPU-accelerated.

View on GitHub

Benchmarks are run on the following setups:

M1 Macbook Pro
  • 2021 16-inch MacBook Pro, Apple M1 Max, 64GB RAM, macOS Monterey 12.5.1
  • Python: 3.10.8
  • PyPy: 7.3.9 nightly build 2022-11-02, 106335-024a5669d75d
  • Clang: Apple clang version 13.1.6
  • Codon: 0.15.0, commit 5480050b0ae3fb39b30e3a71df260ce5e91e5064
Intel(R) Xeon(R) Gold
  • CentOS Linux 7 (Core), 64 cores, Intel(R) Xeon(R) Gold 5218 CPU @ 2.30GHz, 754 GB RAM
  • GPU: NVIDIA(R) Tesla V100
  • Python: 3.8.2
  • PyPy: 7.3.9
  • Clang: 13.0.1
  • Codon: 0.15.0, commit 5480050b0ae3fb39b30e3a71df260ce5e91e5064
testimonials

What our community is saying

“Thank you, for making Python fast like C/C++. It's real”

On the CPU, the Python solution runs as fast as C++. The performance on the GPU is mind boggling. It does chunking, allowing one to specify a large N (e.g. 1 billion) and not worry about depleting GPU memory.

@marioroy on GitHub

“Profound Achievement”

Your work is magnificent and inspirational. Codon is a glorious achievement and will influence the landscape of future human machine endeavors. Combining the high performance of native binaries and the intuitive natural cognitive ergonomics of Python is the best of both worlds. Big Ups to this team. Bravo.

@psytron on GitHub

“Codon support. 74 times speedup compared to Python”

This set of changes [to llama.py] adds ability to compile llama2.py with Codon. The result is a 74 X speedup!

@dmahurin on GitHub

solutions

For Enterprises

Support & Services

We offer enterprise licenses with support & services packages, custom solutions and more.

Exaloop Cloud

Run Codon in a fully managed and hosted environment. Scale on the cloud, no setup or dependency management needed.

Stay in the loop

Join our mailing list to stay updated on new features, product releases and announcements.

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.