Python JIT

Python source → CPython bytecode → interpreter starts running it → JIT notices “this code runs a lot” → compiles that hot path to machine code → later calls jump to faster machine code

Python JIT / Numba mental model

Normal CPython:

Python source → bytecode → CPython interpreter executes bytecode`

With @numba.njit: Python function -> Numba wrapper -> compile hot function to machine code -> run machine code

A decorator like:

@njit
def f(x):
    return x * x

CPython still runs the overall program. But inside the @njit function body, Numba can skip CPython bytecode interpretation and execute native machine code directly.