Re: bytecode unification for scripting (and other!) languages

[Shriram Krishnamurthi <sk@cs.brown.edu>]

Nicolas Neuss wrote:

> You certainly know the type inference which CMUCL is using.  

No, I don't.  That is, I know it's there, and I've seen examples of
it, but I don't understand how it works.  (I've read the manual.  This
gives me several examples, but no spec.  Maybe the manual has changed
since I last read it?)  

I don't know what mechanism they use to implement type inference, and
how it stays tractable in the presence of union types and range types
and so forth.  I don't fully understand "dynamic type inference"; it
looks more like "static flow analysis" to me.  Certainly what they
accomplish looks impressive -- eg, if-splitting followed by
optimizations -- but how?

Surely someone on this list can point me to a technical paper that
explains exactly what they do?  I'd be much obliged.

Shriram