[Ecls-list] unboxed closure

[Brian Spilsbury]

On 8/30/07, Juan Jose Garcia Ripoll
<juanjose.garciaripoll at googlemail.com> wrote:
> Hi Brian,
>
> just a short email from abroad. I am currently travelling, hence the
> lack of feedback, but I am positively impressed by your progress. As
> for the closure type, that is not a bad idea as far as the data is
> aligned the garbage collector will properly interpret it.

I expected that it was probably something like that.

I think, though, that I've made some mistakes in the previous two patches.

Generating two separate lambdas at the lisp level seems to work for globals,
but it has some issue with locals.

I've added a fun-unboxed field to funs, which contains all of the
information required to call an unboxed version, from the same source
file.

I'll look at rewriting the generation code to produce the unboxed
version transparently, so that only one fun object remains in the
environment.

I don't yet understand the code mobility stuff, so I may need to ask
about that later on.

Regards,
Brian.