Re: Caller allocates space for callee-save registers

pardo@cs.washington.edu writes:
> ... also wonders aloud if anybody has done recent studies on stack usage.
> The studies used in designing the SPARC showed very modest call nesting.
> I've heard it said that some LISP compilers avoid register windows because
> the overhead of using them is too large. The same compilers might also wind
> up wasting a lot of stack space using the standard convention.


I'm also a bit curious about this. How well does the stack window
paradigm stand up against the stack frame paradigm? A stack window is
essentially a cache on the topmost portion of a stack allocated in
register space. It is written to memory upon overflow and read from
memory upon underflow. Some small overlap (the term is hysteresis?) is
retained to avoid degenerate behavior at boundaries. I can envision
definable underflow and overflow traps for more ambitious memory
management. Would it be feasable to make such an object addressable?


What the heck, let's make it a survey for analyses. I'll summarize all
mailed responses to this topic (unless specifically requested otherwise)
periodically. I'll even cough one up, although I suspect mine'll look
like a mule amidst thoroughbreds. "Enter as often as you like, no
purchase required."


Thanks in advance, [Ag] gaynor@paul.rutgers.edu
[It is my impression that with a reasonably smart compiler stack windows
are a big lose because the compiler can usually cut the register save
traffic to memory far below what windows do. The IBM RISCs were all
designed along with industrial-strength optimizing compilers and the
closest thing they have to windows is load and store multiple. I'll try
to dig up some references. -John]
--

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