PyPy 2.4 "Snow White" Released

[phoronix]

Phoronix: PyPy 2.4 "Snow White" Released

Version 2.4 of the PyPy Python interpreter and JIT compiler has been released...

http://www.phoronix.com/vr.php?view=MTc5NDQ

[rgvivA]

Pyston

Drobbox is working on an alternative to PyPy, a jit based python implementation based on LLVM.

https://tech.dropbox.com/2014/04/introducing-pyston-an-upcoming-jit-based-python-implementation/

Really curious to see what comes out of this.

[schmidtbag]

I keep meaning to try this. I'd be very interested to see benchmarks of this on ARM. Actually, I'm interested how python in general performs on ARM clock for clock vs x86. Since python is an interpreted language, it isn't exactly efficient or (to my knowledge) capable of taking full advantage of CISC, so there's a good chance that the language is relatively very efficient on python.

[TheBlackCat]

Since python is an interpreted language, it isn't exactly efficient or (to my knowledge) capable of taking full advantage of CISC, so there's a good chance that the language is relatively very efficient on python.

In this context, it is more useful to say python is a glue language. Python code itself won't benefit from processor features, but a lot of the slow code paths inside python are coded in C, and those will benefit.

[schmidtbag]

In this context, it is more useful to say python is a glue language. Python code itself won't benefit from processor features, but a lot of the slow code paths inside python are coded in C, and those will benefit.

Whoops I meant to say "...so there's a good chance that the language is relatively very efficient on ARM." Stupid brain fart.

Sure, the slow code paths inside python may benefit, but wouldn't that only be the case if they're properly compiled to do so?


@PMrgvivA
I've heard of that but last time I checked it was windows-only and somewhat limited on libraries. Personally, I'd much rather see the dropbox devs help out pypy. Then again, dropbox is creating their own interpreter for their own needs first, so they can really optimize it for their workload better than pypy. But then in another perspective, why not just code straight from C++?

[majorkush]

There's a lot of talk about Pyston (DropBox's Python JIT), but I wonder where DropBox thought PyPy came up short? For most needs I've faced, standard Python has sufficed and the few times it was "too slow" it was because I was using a slow algorithm. But I'm no where near as talented as the guys on the Pyston project, I'm sure their rebuilding for reasons us simpletons will never understand lol.

Never the less, I'm excited for this new PyPy! At this rate, it should be the new official fork of Py2.7