Unladen-Swallow: A faster python

[Luis M. González]

This is a new project started by two Google engineers to speed up
python:
http://code.google.com/p/unladen-swallow/

[Tim Roberts]

I read this with a skeptical eye, but they have some very interesting ideas
here. IronPython has certainly shown that Python can be successfully
implemented in a JIT compiled VM in a performant way, but it has issues
running C extension modules.

I'll be curious to see where this project goes.
--
Tim Roberts, ti...@probo.com
Providenza & Boekelheide, Inc.

[Dotan Cohen]

> This is a new project started by two Google engineers to speed up
> python:
> http://code.google.com/p/unladen-swallow/
>

There is a decent discussion going on at slashdot:
http://tech.slashdot.org/article.pl?sid=09/03/27/1934256

--
Dotan Cohen

http://what-is-what.com
http://gibberish.co.il

[Irmen de Jong]

Tim Roberts wrote:
> Luis M. González <lui...@gmail.com> wrote:
>> This is a new project started by two Google engineers to speed up
>> python:
>> http://code.google.com/p/unladen-swallow/
>
> I read this with a skeptical eye, but they have some very interesting ideas
> here. IronPython has certainly shown that Python can be successfully
> implemented in a JIT compiled VM in a performant way, but it has issues
> running C extension modules.

My little experience with IronPython is that its raw compute speed is actually
faster on my machine than CPython (test.pystone is 50% faster in ironpython), but:
- it takes a very long time to start up (almost 5 seconds)
- I/O and networking (sockets) is slower

To compare, Jython starts in about half a second on my machine, which makes it
usable to run short scripts for instance. (Unfortunately its raw compute speed
is slower than CPython: test.pystone runs at half the speed).

I hope 'unladen swallow' won't suffer from a long startup time and has decent I/O speed!

> I'll be curious to see where this project goes.

Me too.
Although CPython has always been plenty fast for me :)

--irmen

[andrew cooke]

Tim Roberts wrote:
> [...] IronPython has certainly shown that Python can be successfully

> implemented in a JIT compiled VM in a performant way, but it has issues
> running C extension modules.
>
> I'll be curious to see where this project goes.

given the comments on python-dev i wonder if this is the first indication
that python is going to split into separate implementations for windows
and unix (via .net and llvm, respectively)?

andrew

[Mark Hammond]

On 28/03/2009 9:50 PM, andrew cooke wrote:
> Tim Roberts wrote:
>> [...] IronPython has certainly shown that Python can be successfully

>> implemented in a JIT compiled VM in a performant way, but it has issues
>> running C extension modules.
>>
>> I'll be curious to see where this project goes.
>

> given the comments on python-dev i wonder if this is the first indication
> that python is going to split into separate implementations for windows
> and unix (via .net and llvm, respectively)?

What comments are they? There is no indication that unladen-swallow is
fundamentally broken for Windows, just temporarily broken due to the
lack of windows developers/contributors...

Saying-no-to-fud ly,

Mark

[sk...@pobox.com]

Dotan> There is a decent discussion going on at slashdot:
Dotan> http://tech.slashdot.org/article.pl?sid=09/03/27/1934256

>From that thread:

person1> I'm not quite sure what benefits this gives that Psyco doesn't
person1> already.

person2> It doesn't get as stabby.

:-)

--
Skip Montanaro - sk...@pobox.com - http://www.smontanaro.net/

[sk...@pobox.com]

Andrew> given the comments on python-dev i wonder if this is the first
Andrew> indication that python is going to split into separate
Andrew> implementations for windows and unix (via .net and llvm,
Andrew> respectively)?

The Windows limitations for Unladen Swallow are due primarily to the poor
support for Windows in LLVM. As that improves, Windows support in Unladen
Swallow will improve.