[Openmcl-devel] Grand Central Dispatch and CCL

[Brian Mastenbrook]

On Sep 14, 2009, at 8:08 PM, Gary Byers wrote:

> I think that you're right in thinking that most of what's involved
> in being able to use GCD from CCL has to do with supporting blocks,
> and other stuff in Cocoa is also likely to start depending on that
> support as well.  It's obviously desirable for CCL to provide that
> support, but I don't yet understand the issues well enough to know
> whether that's trivial or what might make it non-trivial.

I don't think any block support is necessary: every GCD function has a  
corresponding _f variant that takes a regular function pointer.

> We have had a customer express interest in this and as we pursue that
> it'll probably be necessary to think about it enough to have a less
> handwavy notion of the work involved.  I suspect that there needs to
> be at least some additional runtime support in the lisp (mostly having
> to do with support for anonymous callbacks that can be
> established/disestablished on the fly) and I vacillate between
> thinking that there needs to be some compile-time support as well
> and thinking that my reasons for believing that are invalid.

About the only things I can think of for compiler support would be  
some way of inlining the trampolining back into Lisp-land for  
efficiency, and maybe something having to do with specials. (What are  
the rules for special bindings with GCD?) What would you think of as  
being the reasons for adding some kind of compile-time support?

--
Brian Mastenbrook
brian at mastenbrook.net
http://brian.mastenbrook.net/