> >Perl 6 will have optional type declarations
> >and I imagine the statically typed registers
> >in Parrot are there also to allow the
> >compiler to exploit them.
>
> Nope, that's not why they're there. The typed registers may be used for
> them as a sort of cache, the way hardware registers are used now, but typed
> variables will still be PMCs.
So they are there for very specific potential optimisations, e.g.
for 'for' over a known range, and YMMV.
Some questions:
there is some idea about using at least behind the scene, tagged
arithmetic in Parrot at some point.
[Simon Cozens on perl.com]
> We've already designed the vtables and the structure of PMCs
There are some textual traces of that to read? I'm curious,
I mean, it's a central point.
I have read the thread about %MY. My impression is that
there is a strong believe there will be enough information
around to avoid big penalties (I mean in general
also beyond the %MY issue): properties of subs will be known.
But is that true for OO code?
The problem of getting the call graph of a OO program is not that easy
and without call a graph I wonder whether much information is available.
regards.
