contification paper

[Matthew Fluet]

> I am thinking of changing the compilation statistics to the following two
> tables.
> 
> The first table gives line numbers and compile times.  I don't think it's
> necessary to report compile times for each of the four cases, since they're all
> computing all of the analyses.   So I'd just like to report the None numbers.

I guess, but that doesn't show that the actual transformation isn't that
costly.  Maybe take the max of all the times?  Or, just say that the
estimated transformation time (say, contify time of {call,cont,dom} -
contify time of none) was never more than X% of the total compile time,
where X is some small number

(One thing that is interesting is mlyacc, where None has the highest
percentage, simply because it has so many functions at the later passes.) 

> The second table would look like the current table 1, but with the lines and the 
> times removed -- i.e. just the counts.

That's fine with me.