[BangPypers] unladen-swallow

Anand Balachandran Pillai abpillai at gmail.com
Wed Apr 1 08:18:47 CEST 2009


On Wed, Apr 1, 2009 at 12:42 AM, Shekhar <pythonic at gmail.com> wrote:
> Already slashdotted and many of us should already be knowing but could
> not resist posting.
> http://arstechnica.com/open-source/news/2009/03/google-launches-project-to-boost-python-performance-by-5x.ars
> http://code.google.com/p/unladen-swallow/wiki/ProjectPlan
> Apart from performance boost there is also mention of removal of GIL.
> Very exciting.
>
> Would love to hear opinions of bangpypers.

Talk about GIL is fine, but more importantly  if they can reduce the
memory footprint of Python by 5x just that would be a great
achievement. I also
liked their focus on improving cPickle performance, which is an area
that badly needs improvement.

 Basing it on LLVM is a good idea, and if they can remove opcodes
 nothing like it, since a great lot of time in Python's eval loop is spent
 in dispatching them. There have been bug fixes and improvements
 in this, but a lot of them are dependent on new features in GCC.
 I think they should also take a look at the work already done in
 Parrot and Cython which can be reused.

 This line caught my eye: "Apart from these, our benchmark suite
includes several crap benchmarks like Richards, PyStone and PyBench",
followed by

"For charting the long-term performance trend of the project, Unladen
Swallow makes use of Google's standard internal performance
measurement framework" - I am not sure if the Python community will
take kindly to
the remark that PyStone is "crappy" ;)

Regards


>
> Shekhar
> _______________________________________________
> BangPypers mailing list
> BangPypers at python.org
> http://mail.python.org/mailman/listinfo/bangpypers
>



-- 
-Anand


More information about the BangPypers mailing list