Codon 0.17 Released For LLVM-Based Python Implementation For 10~100x Speedups

Codon is an open-source project that leverages the LLVM compiler infrastructure and aims for super fast Python code with as much as 10~100x speedups. Released this week was Codon 0.17 as the newest step forward for this alternative Python implementation.

Codon compiles Python code to native machine code to avoid runtime overhead. The Codon project promotes "10-100x or more" speedups over upstream Python and that the performance can be comparable to that of traditional C/C++ code. Codon also allows for native multi-threading to tap even greater performance.

With Codon 0.17 the project has re-based to upstream LLVM 17 compiler use, adds new floating point type support for float16 / bfloat16 / float128, updates to various existing functions, added input() function handling, property setters are now supported, and there have been improvements to the dynamic polymorphism handling to better match CPython.

Codon 0.17 also adds a "-disable-exceptions" flag to disable exceptions, among other fixes and refinements.

Downloads and more details on the Codon 0.17 compiler release for Python via GitHub.