Diff between 11/20 and 11/45?

[Carl Lowenstein]

> From owner-pups at minnie.cs.adfa.edu.au Tue Sep  7 18:24 PDT 1999
> Date: Tue, 7 Sep 1999 17:49:36 -0700 (PDT)
> From: "Steven M. Schultz" <sms at moe.2bsd.com>
> To: pups at minnie.cs.adfa.edu.au
> Subject: Re: Diff between 11/20 and 11/45?
> 
> > From: Carl Lowenstein <cdl at mpl.ucsd.edu>
> > 
> > 	MOV SP, -(SP)
> 
> 	Similarily
> 
> 		MOV R0,(R0)+
> 
> 	won't work as expected on some 11s.  I suspect that the even less
> 	likely case of "mov pc,-(pc)" won't work either :-)
> 
> > It isn't really clear to me why one would want to use this particular
> > instruction, however it turned out to hang both BASIC and FOCAL at the
> 
> 	Fairly common when setting up call frames, etc.  You want the
> 	address of where the arguments start and since they're pushed on the
> 	stack 'sp' is the value you want.
> 
> 	There's a comment in 2BSD (I think it came from V7) where mention is 
> 	made that "we can't do sp,-(sp) because it won't work on the 11/40".
> 
> > time.  A zero-length patch wasn't too hard to figure out.
> 
> 	Hmmm, interesting.  The workaround I saw took an extra instruction.

Abbreviated due to fading memory over the years, but refreshed by some
of the current discussion.  The patch was zero-length but involved more
than the one instruction.  Something similar to:

	MOV SP, -(SP)		MOV SP, R0
	MOV (sP), R0		MOV R0, -(SP)

The net result being that the initial value of SP is now both in R0 and
on the stack.  Without doing both a SRC and DST operation on SP in the
same instruction, which is the thing that is incompatible across different
processor hardware.

    carl



Received: (from major at localhost)
	by minnie.cs.adfa.edu.au (8.9.3/8.9.3) id OAA04681
	for pups-liszt; Wed, 8 Sep 1999 14:51:11 +1000 (EST)