comments

[Matthew Fluet]

> So, I just got back and pulled down the snapshot.  What's best for me to
> look at?  I'll try putting together a code size table for right now.

FYI, code sizes normalized to none range from
0.83 for cont and dom on count-graphs
to
1.03 for call on vliw

Otherwise, the range between is fairly well filled in, no clear winner. 
Is it worth devoting the space of an entire table to this, or should we
simply note the range.  Again, a lot comes back to the question mark about
how the different contification analyses interact with the other
optimizations.