Re: Prothon on CPython intrepreter? [PROTHON]

Mark Hahn wrote:

> Christian Tismer wrote:

[big snip]

>>This is why I started PyPy, in January 2003.
>>I knew that it is impossible, even (or maybe just) for me,
>>to get around this in any other way than to have 10 people
>>on a radically different approach:
> 
> I didn't know you started PyPy.  How did you get 10
> people to work on it?

Well, it started on the python-de mailing list. I was
musing with Holger Krekel about a small, nice,
configurable and pluggable Python.
After short time, we realized that we must write Python
in Python.
We got Armin Rigo interested, who made the ideas wider
by a power of ten. Michael Hudson, Samuele Pedroni
(the Jython maintainer) and some more. Later, Laura
Creighton and Jacob Hallén got interested, and they pushed
towards EU funding. Alex Martelli followed.
Later this year, the funding will start, and I will work
full-time on the project for two years.

[another big snip]

>>If you want to do a better CPython, ignore CPython and
>>come to PyPy. There you can find a group of people of
>>at least your craziness, and you'll have great support.
> 
> Will it run all the CPython libraries?  How fast is it?  How close is it to
> beta?

We are planning to interface to CPython libraries.
But mainly, we are going to rewrite CPython libraries
with pure Python.
The goal is to get rid of as much of C as possible.
Nobody needs that. We will have our own code generators,
and C is too static, the real showstopper.

How fast? At the moment about 10000 times slower than
CPython. It is an interpreter on top of it, not optimized
at all for speed, but for abstraction.

One first speed-up will be a simple code generator that
emits C code for the interpreter. That should run
very much faster. Still, there is no hand-written C code,
it will be generated.
We have different and also more dynamic paths which we
are trying in parallel. I just gave a simple example.

But we are confident that we can make it considerably faster
than CPython, by the dynamic features of specialization.
Compare Psyco. PyPy is the future of Psyco and Stackless,
both projects will become useless when PyPy rules.

beta? I think we are far away from even alpha. :-)

...

> I will take a more in-depth look at the Cpython code over the next couple of
> days.  If I see a way to do what I want to do I will ask you for
> verification that my ideas will work.  If I can't figure out a way to do it
> then I will agree that it is a bad idea and move on to other ideas, probably
> .Net.

Yes, please do that, read ceval.c, typeobject.c and a few
object implementations, for instance. The code is very readable,
but it is quite a lot.
Feel free to ask me anything about the code.

all the best -- chris
-- 
Christian Tismer             :^)   <mailto:[email protected]>
Mission Impossible 5oftware  :     Have a break! Take a ride on Python's
Johannes-Niemeyer-Weg 9a     :    *Starship* http://starship.python.net/
14109 Berlin                 :     PGP key -> http://wwwkeys.pgp.net/
work +49 30 89 09 53 34  home +49 30 802 86 56  mobile +49 173 24 18 776
PGP 0x57F3BF04       9064 F4E1 D754 C2FF 1619  305B C09C 5A3B 57F3 BF04
      whom do you want to sponsor today?   http://www.stackless.com/