Re: [Psyco-devel] Generator support (was: Re: Psyco internals (fwd))

[Armin R. <ar...@tu...>]

Hello Chris,

On Sat, Jul 31, 2004 at 12:40:14AM -0400, Chris King wrote:
> > > Implementing generators in C is trivial:
> >=20
> > I don't think I agree with this sentence, either.
>=20
> Well, to try and convince you of my stance, you can download
> proof-of-concept C generators from
> http://users.wpi.edu/~squirrel/temp/cgen.c.  I did the implementation
> using the nifty ucontext mechanism, after attempting a setjmp/longjmp
> implementation.

I see.  Yes, this kind of implementation is close in spirit to how some
versions of Stackless work.  The ucontext mechanism (which I didn't know
about, thanks) appears not to be portable, though, and I don't think you =
can
do it with setjmp/longjmp.  But it does indeed work more or less like thi=
s in
Stackless nowadays.  There is even an off-product of Stackless called
Greenlets that is essentially a Python-oriented ucontext implementation. =
=20
(Both Stackless and Greenlets use a few lines of assembly code to hack th=
e
stack.)

> Of course, I'm not sure if this helps the Psyco case any...

It might do.  Christian "Stackless" Tismer offered a similar way to suppo=
rt
generators on top of Stackless, which would then more or less automatical=
ly
work with Psyco.  In the meantime, we drifted over to the present situati=
on,
which is to use bytecode rewriting of generators instead, to turn them in=
to
regular Python functions that save their state manually.  This will give =
a
different kind of generated code with Psyco: the state will be saved off =
and
restored to the C stack around a 'yield'.  In some sense it is better bec=
ause
it avoids keeping a lot of small C stacks around (and Psyco wastes quite =
a lot
of C stack space, so multiplying it by the number of active generator
instances doesn't look like a good idea).



A bient=F4t,

Armin.