A Crazy Thought about CallContext Recycling and Caching
chromatic
chromatic at wgz.org
Tue Jan 5 01:14:44 UTC 2010
On Monday 04 January 2010 at 17:08, Austin Hastings wrote:
> I've wondered about a similar thing, mostly while listening to all the
> pcc noise a while back.
>
> The argument info, parameter info, and mro data seem very amenable to
> caching. Why not rewrite the code in-line? (Sure, there are cases where
> you have to re-check because of blah blah blah whatever, but there are
> also cases where there's only *one* sub named "say" and no reason not to
> cache it.)
Ultimately we need to do that, but changing the opcodes was enough of a hassle
that we decided to do that after the 2.0 deprecation point.
Allison and I talked recently about reviewing how we model calls in
bytecode/ops so as to allow dynamism while exploiting compile-time static
knowledge.
-- c
More information about the parrot-dev
mailing list