cps & contification

Stephen Weeks MLton@sourcelight.com
Wed, 24 Jan 2001 15:43:38 -0800 (PST)


> It  seems  to me (from a position of almost complete ignorance) that it would
> be rather easy to change things now so that you can have  mutually  recursive
> continuations.

It's probably one afternoon's worth of work, provided no bugs are introduced.

> Is  their lack the reason for the dominator calculation? 

No. 

> If
> so then this would seem to be a really strong argument for going for mutually
> recursive  continuations.  If not then what is the dominator calculation for?

The dominator calculation is so that we can find labelings for functions
that meet the safety conditions, in particular, the tail call condition
        *** For all tail calls (f, g) in T,
                A(f) = Uncalled
                or A(g) in {f, A(f), Unknown}

Matthew, could you please send out the latest summary of the entire analysis so
that Henry could read through that instead of reading through all of our
mistakes?  Thanks.