On Mon, 2001-12-24 at 19:18, Armin Rigo wrote:
> Petru Paler a =E9crit :
> > What would be the implications of trying to combine Psyco with Stackles=
> > Python? I don't understand enough about either of them to see what woul=
> > happen, but the result would be quite interesting...
> I don't know enough about Stackless to give a definite answer, but I
> believe that it won't easily work. Stackless means that the interpreter
> but clearly separates the C stack with Python's. But the code emitted by
> Psyco extensively uses the processor stack (i.e. the C stack). So I
> think that implementing all the interesting features of Stackless would
> be either hard or slow. Implementing even only Python 2.2's simple
> generators ('yield') is too tricky right now.
Judging from the comp.lang.python discussions about Stackless, the
current implementation is quite intrusive and some of the developers
don't like it. Too bad, tail recursion and continuations would be two
nice things to have in Python.
Which reminds me: could Psyco eventually detect and optimize tail
recursive functions to simple loops instead of calls?
Petru Paler, http://www.ppetru.net