🚀 go-pugleaf

For languages without "upward funargs" (e.g. Pascal), a global display
is sufficient, and furthermore requires only one extra push/pop per
procedure call. 

Let D[0..n] be the display registers.  Then the code for entering a procedure 
with lexical level n is:

	push D[n];
	D[n] := local_frame_pointer;

And the code for procedure exit is:

	pop D[n];
	return

One need not copy the entire display for each procedure call.
This was the Burroughs 6700's way of handling the display. I also notice 
that it was rediscovered in SIPLAN notices last year.  I suppose the
``expensive display'' myth is related to the ``expensive procedure call'' 
myth.  

I've heard that the "static link" approach is not inefficient with a
global optimizer.  The "chase" becomes a common subexpression and is
computed just once.

Arch D. Robison
University of Illinois at Urbana-Champaign
	
CSNET: robison@UIUC.CSNET
UUCP: {ihnp4,pur-ee,convex}!uiucdcs!robison
ARPA: robison@B.CS.UIUC.EDU (robison@UIUC.ARPA)