Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.2 9/18/84 SMI; site sun.uucp
Path: utzoo!watmath!clyde!cbosgd!ihnp4!zehntel!hplabs!hpda!fortune!amdcad!amd!pesnta!qumix!qubix!sun!gnu
From: gnu@sun.uucp (John Gilmore)
Newsgroups: net.micro.mac
Subject: Re: high byte of address question
Message-ID: <1960@sun.uucp>
Date: Mon, 21-Jan-85 01:33:01 EST
Article-I.D.: sun.1960
Posted: Mon Jan 21 01:33:01 1985
Date-Received: Thu, 24-Jan-85 07:06:26 EST
References: <224@ucbopal.CC.Berkeley.ARPA> <130@redwood.UUCP>
Organization: Sun Microsystems, Inc.
Lines: 30

I'm in total agreement with Rob Warnock on the use of upper bytes of
32-bit addresses.  I saw the IBM 360 folks get burned by it and don't
intend to let any enterprise I'm associated with do it again.  This
kind of usage does NOT automatically break when you install a 68020
(the way it did in the IBM systems), since it's not part of the
instruction set.  Instead, it breaks if you ever hook up more than 24
address wires to any 68000 family member (eg 68012, 68020) that brings
these bits out to pins.  It DOES decrease cache performance in the
68020, since the same piece of code can be accessed at addresses which
the cache thinks are different.  In a data-cache based machine that
doesn't deal well with aliasing, it could also cause incorrect
results.

This whole topic has previously been discussed in net.micro.68k.

A few comments on the SummacC relocation stuff (without knowing it
personally):
	1.  It was probably expedient and it certainly worked.
	2.  It was easier than changing the C compiler.
	3.  It is not portable to larger machines.
	4.  There will be Mac-class machines with >16MB within 5 years.
	5.  There's nothing as permanent as a temporary kludge.
	6.  At least they could have cleared the high bytes after doing
	    the relocation, so the running program would not know the
	    difference.

A side comment, the 'LEA' instruction, like all address arithmetic in the
68000 family, operates on a full 32 bits.  In the 360, the equivalent "LA"
instruction cleared the high byte; this was a major problem ten years
later.