[Python-ideas] A General Outline for Just-in-Time Acceleration of Python

[Nick Coghlan]

On 13 Jun 2014 16:22, "David Mertz" <mertz at gnosis.cx> wrote:
>
> There are many people with a better knowledge of PyPy than me; I've only
looked at it and read some general white papers from the team.  But I do
know with certainty that PyPy executes ALL Python code, not a restricted
subset[*].  You might be thinking of Cython in terms of an "annotated,
restricted, version of Python."

There's also Numba, which allows particular functions to be flagged for JIT
compilation with LLVM (and vectorisation if using NumPy). That technically
only supports a Python subset, but it's in the context of the normal
CPython interpreter, so it's generally possible to just skip accelerating
the code that Numba can't handle.

Cheers,
Nick.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.python.org/pipermail/python-ideas/attachments/20140613/60527681/attachment.html>