> At the moment, the JIT is only used if the function contains the JUMP_BACKWARD opcode which is used in the while statement but that will change in the future.
It's a bit less underwhelming if you consider that only function objects with loops are being JITed. nb: for loops in Python also use the JUMP_BACKWARD op.
PyPy was never able to get fast enough to replace CPython in spite of its lack of compatible C API. CPython is trying to move fast without breaking C API, and 2--9% improvement is in fact very encouraging for that and other reasons (see my other comment).
> The initial benchmarks show something of a 2-9% performance improvement.
Which is underwhelming (as mentioned in the article), especially if we look at PyPy[0]. But it's a step forward nonetheless.
[0] https://speed.pypy.org/