Sounds interesting; I'll see how it plays with the globalstructure stuff
as soon as I can make the time. Funnily enough I was just about to send
you mail offering a suggestion for something potentially useful to spend
some time on ( and it seems you have found a little 'free' time, lucky
you).

Ages, years, ago, I started wondering if changing the codegen to be able
to just specify that inline functions were, well, inline would be any
good. Many C compilers (certainly GCC) can accept the _inline_ modifier
and I wouldn't mind betting that they might do a better job than our
textual substitution does. The only thing that stopped me was the fact
that my CC didn't support it (it does now :-) and laziness. Oh, and the
fact that last time I tried a pass at it I ran into some problem that
non-inlined CCodeGen went wrong. I can't even remember why. Or even when
it was.

It would certainly produce the code faster without the inlining phases.
It might even produce a smaller executable, depending on how well the CC
does. Has to be worth a try sometime.


tim
--
Tim Rowledge, tim at sumeru.stanford.edu, http://sumeru.stanford.edu/tim
Useful random insult:- A photographic memory, but the lens cover is glued on.