ICFP presentation

Matthew Fluet Matthew Fluet <fluet@CS.Cornell.EDU>
Tue, 28 Aug 2001 16:20:46 -0400 (EDT)

> The word "location" almost seems to mean the same thing as "Return" from the
> contification analysis slide.  Could you use the same word (and possibly even
> change Root to ?).

I thought about that.  Technically, this doesn't work with what's in the
paper, because you need to distinguish between Uncalled and Unknown.
(Although, one could formulate Adom by not including Uncalled functions in
the graph G, in which case you could equate Root and Unknown.)  But, for
the purpose of the presentation, it should be o.k.  And will probably
simplify things for the audience.

> Prepared question slides
> 	related work

O.k.  I guess briefly hit Reppy, Appel's loop headers, and lambda

> 	comparison to inlining

I'm going to try and address this briefly after the example.  I can make
up another slide with a couple of examples where contification gets
something that inlining doesn't.  That's sort of a weak argument, because
it really depends on how often those patterns show up in actual code.

This is one I want to include (going back to some old notation):

fun main () = ... K1 (f ()) ... K2 (f ())
fun f () = ... g () ... g () ...
fun g () = ...

Neither Acall or Acont or inlining would get g.  This was in some of my
notes on contification and it's derived from a real example for concat, so
it somewhat addresses the comment above.

Any other