except.mlton

[Matthew Fluet]

> I just looked into why MLton does so well on the except benchmark and discovered
> that it is able to turn all the raise/handle pairs into direct jumps, since it
> can figure out the control-flow at compile time.  Maybe fair, maybe not, but I
> thought you should know.  Personally, I'm pleased that the analysis does this.

This is the raise-to-jump analysis?  Very cool.  Presumably contification
and/or other analyses were able to collapse almost everything into the
same cps function.