[pypy-dev] Re: [ann] Minimal Python project

[Samuele Pedroni]

From: "Christian Tismer" <tismer at tismer.com>
> > Will you give some thought to going to a tagged representation of
> > object references, and maybe a more traditional GC?  That way you
> > avoid a lot of memory traffic, and also don't have to adjust reference
> > counts every time you touch anything.
> >
> > I think these would give a big performance boost, and also would also
> > simplify the C API.  It might be possible to supply a compatibility
> > layer to make the old C API still useable.
>
> Can you give me a hint how this would be done?
> I have no experience with tagged representations,
> but if this can save so much, I should learn more
> about it.

this is a relevant survey paper:

http://citeseer.nj.nec.com/gudeman93representing.html

Representing Type Information in Dynamically Typed Languages

e.g.

integer in the range [-2**30,2**30-1] would be represented by their bit
patterns shifted left by 1 bit

all other objects would be boxed, and the address left shifted by 1 and ORed
with 1.

The fact that Python is slowly removing the long/integer separation can make
less disruptive such an approach.

OTOH I think such an approach would make the life a bit more complicated for
psyco when detecting that a datum has machine-size integer type, because such a
datum would be either in the first form _or_ a boxed (long) integer in the
larger range -2**31 2**31-1.

Another approach is to carry data around as a struct where both the tag and the
datum are machine words, this approach is AFAIK currently used in the Gwydion
Dylan compiler (www.gwydiondylan.org) originally developed at CMU.

regards.