Re: Stackless compilers?

neptundancer <Karol.Skocik@gmail.com> writes:




> Could you recommend me some papers on constructing a stackless
> compilers? I can only find lots of links about Stackless Python, but I
> am more interested in combination of PI calculus and lambda calculus.
> I know about CubeVM, but that's just pure PI calculus, and it directly
> evaluates AST. I would like to read something about compiler design
> for stackless language, but could not find anything.


It is not clear whether you want the compiler itself to not use a
stack or the generated code to not use a stack? In the latter case,
are you limited to static memory allocation or can you allocate on a
heap?


SML of New Jersey doesn't use a stack but places everything on the
heap -- closures, activation records, data structures etc. The
motivation was that by using a single shared area, you are more
flexible and that by using a good GC, heap management isn't
(considerably) more expensive than stack management. Since the
compiler is bootstrapped, both compiler and target code are stackless.


Many Scheme compilers use a similar strategy, as they have to support
call/cc, which makes simple stacks unsuitable.


Torben

Return to the comp.compilers page.
Search the comp.compilers archives again.