PEP 659 – Specializing Adaptive Interpreter

This PEP seems to be proposing introducing a sort of JIT-lite interpreter to CPython. It would make optimisations based on a much smaller number of examples than a traditional JIT compiler and would also back out much more cheaply. It's too early to estimate performance improvements but they are hoping for something on the order of 25-50% (very preliminary).

The main compromise here seems to be to make a small, simple change while keeping the maintenance and understandability of CPython.

The python-dev thread introducing it is at https://mail.python.org/archives/list/python-dev@python.org/...