Re: Caller/Callee saved Registers

anton@mips.complang.tuwien.ac.at (Anton Ertl) writes:
>So this is not the problem with tail call optimization in C. So what
>is? Why isn't it done?


If you allocate any local storage in the 'tail-recursive' routine and keep
pointers to it, you lose. Since some C programs do this, and testing for
it is kinda hard, I can understand why it isn't the default behavior. I
believe that this thread discussed this very issue about 2 months ago. A
reference to an article about the problems with C tail recursion was
posted, but I didn't write it down.


Since getting C to do this routinely is difficult, I have 'gone with the
flow' and suggested utilizing this behavior to incoporate 'tail recursion'
and a first-level generational GC in the same mechanism. My paper is
called 'Cheney on the MTA', and I'll be happy to send a copy to anyone who
wants it. I also have a version of the Boyer Benchmark using this style
of C coding, which I'll also be happy to send. Perhaps these should be
stored in some ftp directory? Any volunteers?


John McClain <pdp8@ai.mit.edu> wrote:
|> > The lack of tail-call optimization really hurts if you want
|> > to use a continuation passing style of programming.


The 'Cheney on the MTA' technique gives you O(1) continuation captures
'for free'.
--

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