[<<][libprim][>>][..]
Sun Oct 10 09:04:29 CEST 2010

Byte code

What I conclude from [1] is that the current direct ANF-in-tree
approach can be simplified a bit.  Is it possible to distill a byte
code from simply observing the pattern of push/pop instructions?

The main pull is the conversion from expressions to flat code.  The
code pointer becomes specialized.

One thing to note though is that the compiler itself really isn't
trivial either!  I'm far from done!!

Roadmap:
  1. add reader to cellvm to make testing easier (OK)
  2. don't use c reg for storing "continuation code"
  3. copy to cellvm2.c and convert the code pointer to byte code

Would it be possible to describe the VM in scheme such that it can be
easily translated to C, so at least the VM can run also in a
scaffolded setup?

[1] entry://20101010-021347



[Reply][About]
[<<][libprim][>>][..]