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/

 

[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

 

[Irmen de Jong]

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

 

[Mark Hammond]

[...] 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

 

[skip]

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]

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.