Lets Talk About Something Diabolical: A PyPy Intro (David Beazley PyCon keynote) • Hacker News Board

Re: Lets Talk About Something Diabolical: A PyPy Intro (David Beazley PyCon keynote)

#11

Post by Lozzer » Sun, Mar 18, 2012, 2:52 PM UTC

ctkrohn wrote: post #9

Ingaz wrote: post #8

Can somebody explain the end of lecture? Why Ruby and PyPy looks similar in this benchmark?

Yeah, he skipped over what might have been the most interesting part of the talk! I have a very cursory understanding of how the GIL works, so it would be nice to understand what's going on here.

He referenced a previous talk about the GIL:

http://blip.tv/rupy-strongly-dynamic-conference/david-beazly...

Re: Lets Talk About Something Diabolical: A PyPy Intro (David Beazley PyCon keynote)

#12

Post by stefantalpalaru » Sun, Mar 18, 2012, 5:37 PM UTC

The RPython translation to C sounds a lot like what cython is already doing with a subset of Python. Are the PyPy guys reinventing the wheel here?

Re: Lets Talk About Something Diabolical: A PyPy Intro (David Beazley PyCon keynote)

#13

Post by hogu » Sun, Mar 18, 2012, 6:18 PM UTC

stefantalpalaru wrote: post #12

The RPython translation to C sounds a lot like what cython is already doing with a subset of Python. Are the PyPy guys reinventing the wheel here?

Cython generates c based on the cpython api, you can only use it to build cpython extensions.

Re: Lets Talk About Something Diabolical: A PyPy Intro (David Beazley PyCon keynote)

#14

Post by kingkilr » Sun, Mar 18, 2012, 6:50 PM UTC

stefantalpalaru wrote: post #12

The RPython translation to C sounds a lot like what cython is already doing with a subset of Python. Are the PyPy guys reinventing the wheel here?

I seriously doubt it, considering a) RPython is a strict subset of Python, operating on the in memory representation of Python objects and does aggressive type inference, whereas Cython adds additional features (and syntax) to the language and has it's own parser and no type inference (AFAIK), and b) RPython predates Cython, AFAIK.

Re: Lets Talk About Something Diabolical: A PyPy Intro (David Beazley PyCon keynote)

#15

Post by fijal » Sun, Mar 18, 2012, 7:24 PM UTC

stefantalpalaru wrote: post #12

The RPython translation to C sounds a lot like what cython is already doing with a subset of Python. Are the PyPy guys reinventing the wheel here?

Cython is both differently defined and serves a different purpose. First of all, RPython is runnable, Cython is not. RPython has type inference, but more importantly, because it works from live objects, it has Python as a metaprogramming language. It's also much easier to extend than Cython.

Purpose wise cython is for optimizing pieces of code, rpython is for writing interpreters. RPython is quite faster than Cython, but it's also much more restricted, making it a bit unusable for the purpose of "just" speeding up pieces of code (you can't have PyObject equivalents).