CSE

[Matthew Fluet]

> > I think that what is currently missing is a way of recognizing when
> > the success branch of a Transfer.Prim has the Transfer.Prim as it's
> > unique predecessor.
> 
> Easy enough.  Just count the in-degree of each label.

Yes; the bigger question was whether or not enough other passes might want
this information that it should be computed once, say as part of the
dominator tree, rather than in each pass. 

> We could even record the other direction as well -- down the overflow
> branch we know that a particular expression overflows.  But probably
> not worth it.

It was easy enough to do after getting the success branch to work
correctly.  I checked in the CSE mods.