Pypy and JIT

anirudh215 · 2011-06-18 15:44:06

I just read about this project called Pypy and Just In Time compiler it uses a.k.a JIT. http://pypy.org/ .Since it is so much faster than the original python, I was wondering why the people writing the official python code did not adopt this. This would also save time as the people heading the Pypy project wouldn't have to re-write their code every time. Would it detrimental to either project to merge their efforts?

tavianator · 2011-06-18 17:14:22

CPython ought to implement a JIT compiler rather than a bytecode interpreter, and I have no idea why they haven't. However, you can generate very fast CPython-compatible code with Cython, and in my personal opinion, if you're writing a CPU-bound application, you shouldn't be using a scripting language anyway.

anirudh215 · 2011-06-19 05:26:51

I too agree that heavy duty programs ought not to use scripting languages. I myself use C for such programs. However, the majority of my programs are written for physics based projects and so I find myself using the numpy/scipy/matplotlib modules extensively. I find that these modules are excellent and have not found anything quite like it. In light of this, I was hoping that these projects would come together to help those of us in science function better.

Last edited by anirudh215 (2011-06-19 05:56:20)

Mr.Elendig · 2011-06-19 12:03:08

PyPy is faster than cPython at some times, but slower than others. There are also many libs that currently doesn't work with it. My suggestion is to for the time just use cPython, and write your performance critical loops etc in cython.

anirudh215 · 2011-06-24 15:55:53

Yes, maybe that's what I should do. The Pypy website now says they are working on including the numpy and scipy libraries, which is great news.

Arch Linux

#1 2011-06-18 15:44:06

Pypy and JIT

#2 2011-06-18 17:14:22

Re: Pypy and JIT

#3 2011-06-19 05:26:51

Re: Pypy and JIT

#4 2011-06-19 12:03:08

Re: Pypy and JIT

#5 2011-06-24 15:55:53

Re: Pypy and JIT

Board footer