snapshot and new backend

[Stephen Weeks]

A new snapshot is available at
http://www.star-lab.com/sweeks/mlton-2000-6-5.tgz 

This snapshot has the Cps IL without variant types and the new case
optimization.  It passes all the regression tests.  I haven't yet
tried the benchmarks or the self-compile.  I will do so this week.  I
would like Fluet to start with this code as his baseline -- not the
1999-7-12 or the Christmas version.  Hopefully, I won't have to make
too many changes to this code in order to get the self-compile working
and the merges with any new backend changes should be easy.  I think
the best thing for Fluet to get started would be to learn the Cps IL
and see how the current backend and runtime work.  One approach for
the new backend I would like to consider is using the same abstract
machine (backend/machine.{sig,fun}) as the current backend.  This
would allow almost everything to be shared, especially a lot of tricky
stuff like register allocation, limit check insertion, data
representation decisions, ...

Also, Suresh, could you make sure Fluet is on the MLton mailing list
and that he gets this mail.

Let the backend discussions begin!