Re: define and set! with multiple-values

[Bob Kerns <rwk@world.std.com>]

    Date: Sun, 03 Jan 93 16:49:43 -0500
    From: Bob Kerns <rwk@world.std.com>
    
    Not in my experience.  I don't mean it's *no* work for the
    implementor, or that it doesn't require a little bit of thinking,

By the way, as long as we're into C comparisonsm, what I've never
figured out how to squeeze out (except in certain unusual
architectures) is the cycle or so that goes to return NIL
in the case where in C the function would be declared VOID.

So if you meant to complain of the difficulty of optimizing
both the 1 AND ZERO value cases to the level of C, then yes,
I'll agree it's hard.  In fact, I suspect it's impossible in
many architectures.

If Dylan allowed you to declare zero return values in such
a way that the compiler could tell that it will effect *all*
callees as well as the caller, it could match this.

Of course, I think it's silly to spend too much time optimizing
out a single instruction.  Usually you could use the same time
to optimize out a dozen easier ones, weighted by frequency of
usage.