Re: Tail recursion and hobbit (was: finite state machines in guile)

[Jim Blandy <jimb@red-bean.com>]

Bernard, tell me if I've got this right:

Hobbit does turn some tail calls into jumps, if it is able to figure
out how.  However, tail calls to other top-level functions, and some
tail calls within a single top-level function, are not optimized, and
allocate additional stack space.

I don't see how Hobbit could do better, given that it has decided to
match the SCM calling conventions.  It has the same problems GCC
does: if it could know the nature of the callee ahead of time, it
could use a better calling convention, but failing that, it can't do
tail calls.