comments

Suresh Jagannathan suresh@emphora.net
Wed, 14 Mar 2001 16:26:44 -0500


You're right that Appel's inliner would not handle this case; 
flow-directed inlining does which is what I was thinking about.

You're also correct that no beta reductions have been performed
yet.

I concede your points although I must admit that when I first
read the example, my initial reaction was to think of other
standard simplification strategies that would lead to the same
result.

As I said, it's not a big deal.

-- Suresh

> -----Original Message-----
> From: Stephen Weeks [mailto:sweeks@intertrust.com]
> Sent: Wednesday, March 14, 2001 4:12 PM
> To: MLton@sourcelight.com
> Subject: RE: comments
> 
> 
> 
> > A syntactic inliner would have inlined loop at the call
> > to sum even though its body has a recursive call;
> 
> I disagree.  For example, the inliner described in Appel's book 
> would not do
> anything here.  Would your inliner in "Flow directed inlining" do 
> anything with
> loop? 
> 
> I don't see any connection between the contification 
> transformation and inlining
> on this example.  No beta-reduction has been done -- no actual 
> arguments have
> been substituted for formal parameters.  There are still the same 
> number of
> calls to loop.  In particular, the outer call loop (v, 0, 0) is 
> still there.
> 
> > The meta-level comment here is that anyone who
> > isn't very familiar with contification might be confused 
> > (legitimately, I think) into questioning whether the end
> > effect of contification is any different from what a better
> > known aggressive code-motion optimization like inlining
> > might achieve.
> > 
> > It's not a big deal, but I think it would be nice to add
> > a bit of clarification about what the differences are.
> 
> OK.  Assuming you agree with my point above, I'll add something 
> similar to the
> paper.
>