exaloop/codon

Codon is a Python compiler that translates Python code into native machine code, giving it the performance of a compiled language like C or C++ while keeping Python's familiar syntax. Standard Python runs code through an interpreter, adding overhead that makes it much slower for computation-heavy tasks. Codon removes that overhead by compiling ahead of time. The speed difference can be dramatic. The README shows a benchmark where the same recursive Fibonacci calculation takes about 18 seconds in standard Python and 0.28 seconds in Codon, roughly a 65x speedup. NumPy-based code also runs significantly faster. Beyond raw speed, Codon removes Python's GIL, the Global Interpreter Lock, which normally prevents Python code from using multiple CPU cores simultaneously. Codon supports true multi-threading using a parallel annotation you add to loops, specifying how many threads to use and how to divide up the work. It also supports writing GPU kernels, letting you run computations on a graphics card directly from Python-like code. Codon is not a complete drop-in replacement for Python: some dynamic features of standard Python do not work in a statically compiled system. However, you can call standard Python libraries from Codon when needed, and use Codon selectively inside larger Python projects via a decorator that compiles specific functions. Codon includes a fully compiled implementation of NumPy, the popular numerical computing library, and is built on LLVM, the same compiler infrastructure used by C and C++ compilers. You would use Codon when you have Python code doing heavy number crunching and want much faster execution without rewriting in another language.