Current Fastest Python Implementation?

[Carl Friedrich Bolz]

cokofreedom at gmail.com wrote:
> On Feb 18, 9:37 am, Stefan Behnel <stefan... at behnel.de> wrote:
>> samuraisam wrote:
>>> Has anyone done any recent testing as to which current python
>>> implementation is the quickest?
>> Search for a recent thread on CPython and IronPython.
>>
>>> Perhaps for Django development -
>>> though the current one certainly is fast (and I doubt micro
>>> optimizations would make much difference in overall performance).
>>> Regardless - have those pypy folks made a faster implementation, or
>>> the jython folks? Or the .NET implementation?
>> Depends on your use case. Take your application, do some benchmarking and use
>> the implementation that turns out to be a) most reliable and b) the fastest.
>>
>> In that order.

That's very good advice. Basically all four of those Python
implementations have situations where they are faster than all the
others. I guess CPython (possibly using Psyco) is still faster in many
cases, but it really depends on your application.

> PyPy [http://codespeak.net/pypy/dist/pypy/doc/home.html] is getting
> progressively faster.

This is true – PyPy is slowly getting faster. We have two students
working explicitly on speed right now: Anto Cuni is doing a phd thesis
on speeding up PyPy's Python interpreter when compiled to .NET and I am
doing a Master thesis on improving the JIT of PyPy.

> In fact for certain things it can be faster than C [http://
> morepypy.blogspot.com/2008/01/rpython-can-be-faster-than-c.html]!

This link is misleading, since it is about the speed of RPython when
translated to C, not normal Python programs. Normal Python programs tend
to be not RPython, so that figure is hardly interesting.

> However it seems it still has a way to go to be fully operational!
> Still looks like the future to me.

Cheers,

Carl Friedrich Bolz