Python execution speed

Martin von Loewis loewis at informatik.hu-berlin.de
Tue Nov 20 12:57:57 CET 2001


"Morten W. Petersen" <morten at thingamy.net> writes:

>   1. What is the plan for psyco (inclusion in the standard
>   distribution, etc) ?

Here is what I think should/will happen:

* Psyco needs more help. Poor Armin is trying to get people
  interested, but nothing happens. I guess this is *also* an aspect of
  free software: Even if you follow the rules (publish early, publish
  often), success is not guaranteed.

  However, it may be that he'll give up one day if there isn't more
  interest in his work beyond people posting "looks cool" in a
  newsgroup...

* Psyco, as is, is unstable and incomplete. So the natural next step
  would be completion of the work done.

* Psyco currently supports x86 only, so volunteers might be interested
  in adding support for other hardware.

* I don't think Psyco will be integrated into the standard
  distribution in any foreseeable feature, especially if nobody is
  using it. Putting something into the standard distribution is no
  guarantee that it will be stable, supported etc. Instead, being
  stable and supported is a prerequisite for being included in the
  standard distribution.

>   2. Will there be a JIT compiler for Python ?

Well, Psyco is one. I doubt anybody would start another one.

>   3. Are there any similar efforts for Jython [5] ?

Sun is making big efforts in producing code just in time for Java on
different machines, and Jython uses the Java JIT compiler today.

>   4. Are efforts made to optimize Python's execution speed in
>   general ?

Yes, constantly. Most Python contributors believe that this is where
the realistic improvements "for the masses" can be made in the near
future, and concentrate on improving the performance of the
interpreter.

Regards,
Martin



More information about the Python-list mailing list