[bitc-dev] bitc and coroutines

[Jonathan S. Shapiro]

On Sat, 2006-12-02 at 19:21 +0000, Sam Mason wrote:
> On Sat, Dec 02, 2006 at 12:34:54PM -0500, Jonathan S. Shapiro wrote:
> > Portions of the activation handler clearly need to be in assembler
> > because they deal with register-level interactions.
> 
> Indeed they will.  The aim of pointing out coroutines was that it struck
> me as a nice conceptual model that the user land scheduler could use
> internally.  I've just realised that this message should really have
> gone to the Coyotos list instead of this one!  Ah well, seemed like a
> langauge thing at the time.

No, I think this is the right list. The coroutine concept has utility
independent of the Coyotos activation issues.

> I agree that BitC's core should be as small as possible.  It's going to
> be difficult enough as it is!

Swaroop now has the core semantics documented, with one trivial
exception that I recognized today (but it's no big deal). I don't know
if he has published that as a TR yet. If not, I'll kick his butt
appropriately tomorrow. Swaroop: you've been warned! :-)

> > I'm not convinced that the language should provide
> > support for threading or co-routines. Speaking from my perspective as a
> > programmer, most language support for these features is badly flawed.
> 
> There never seems to be enough thought put into the design of these
> things...

It is always tempting to say that, but I think this is not the real
culprit. I think the real issue is that concurrency is a lot harder than
we want to believe, even in its simplest forms.

For example: co-routines generally turn into event dispatch, and there
is often some non-determinism about what co-routine runs next. This
introduces a *huge* problem for program analysis.


shap
-- 
Jonathan S. Shapiro, Ph.D.
Managing Director
The EROS Group, LLC
+1 443 927 1719 x5100