[bitc-dev] BitC status update

[Jonathan S. Shapiro]

We now have a compiler (written in C++) that handles everything *except*
type classes. My current thinking is that we should skip implementing
type classes in this language and implement polymorphic equality and
comparison operations as a transitional mechanism. We can then implement
type classes in the BitC-hosted compiler, where GC and first class
procedures will allow us to liberate an existing implementation fairly
directly.

By "handles", I mean "up to type checking and some simplifying
transforms." I have been working for several days on the C code
generator, and I am coming to question whether going to C is a good
idea. I will send out a more complete note on this shortly.

However, ignoring the fact that the tracing logic is completely wrong,
we *have* successfully run some trivial programs that exercise
constructors, allocate storage, and so forth. As long as you don't
trigger the GC logic, very simple programs work. The code generator does
NOT yet deal with LAMBDA forms, because I needed to take time out to
review the literature on closure construction and to consider GC-related
issues.


shap