Re: Help on code generation and register allocation

Ulrich Hobelmann <u.hobelmann@web.de> writes:




> If tail recursion were completely target independent (and an
> optimization), I don't see why not all compilers would Just Do It.
> Especially GCC in Intel doesn't do it (except maybe for
> self-tail-recursion), which means that you have to use trampolines, or
> weird source code transformations to achieve the same effects (e.g. when
> you want to transform a DFA to machine code, routines tail-calling each
> other).


Tail-recursion optimisation is primarily applied to save stack space
rather than time. Some languages (like Scheme) require tail-recursion
(direct or indirect, even if through call/cc or higher-order function
calls) to use stack space independent on the number of calls, so when
compiling these languages to machines (such as JVM) that do not
directly support tail-calls, you have to go through complex maneuvres
such as trampolines.


As you say, in some cases tail-call optimisation adds extra
operations, so you can't guarantee that you save time (though in most
cases you do). Self-tail-recursion is an obvious place where you will
almost always save time as well as space by doing the optimisation.


                Torben

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