From mboxrd@z Thu Jan  1 00:00:00 1970
From: jmk@plan9.bell-labs.com
To: 9fans@cse.psu.edu
Subject: Re: [9fans] how people learn things (was architectures)
MIME-Version: 1.0
Content-Type: text/plain; charset="US-ASCII"
Content-Transfer-Encoding: 7bit
Message-Id: <20010713163252.E5AAD199C0@mail.cse.psu.edu>
Date: Fri, 13 Jul 2001 12:32:50 -0400
Topicbox-Message-UUID: cccd69e8-eac9-11e9-9e20-41e7f4b1d025

Please, lighten up. There's no point in raking over this stuff again
and again. All architectures have had their missteps, even the venerable
PDP-11. As Phil Winterbottom said when Ken Thompson was bemoaning
the demise of Mips Technologies and the coming age of all the world's a
386 - they won, we lost, get over it.

IA32 'superservers' can address >4GB of PHYSICAL memory due to page-table
extensions, it's nothing to do with segmentation and doesn't need to be
written in assembler.

On Fri Jul 13 12:21:25 EDT 2001, lucio@proxima.alt.za wrote:
> On Fri, Jul 13, 2001 at 03:26:08PM +0000, Douglas A. Gwyn wrote:
> >
> > Lucio De Re wrote:
> > > No more than from Intel (they are to be blamed for CP/M, too),
> > > overlooking the fact that their address bus was wider than the
> > > register size, or scrapping the i860/i960 developments.
> > > They are rushed decisions that can't be reversed.  ...
> >
> > Even the original PDP-11 had a wider address bus than its
> > register (word) size.  It wasn't a "rushed decision", although
> > in the long run it caused enough trouble that a whole new
> > architecture was designed to replace it.
>
> Do you mean to tell me that DEC had a segmented architecture, with
> haphazard default "base" registers, a LOCK instruction to lock the
> bus for the following fetch cycle (whatever for? even the Univac
> 1106 had the more sane test-and-set) and a faulty MOV SS,XX that
> did _not_ lock the bus for the following MOV SP,YY as was the
> intention?
>
> I bet the 20-bit address was an oversight, one that is still being
> dragged along in this era of 64-bit registers.  For heaven's sake,
> Intel's marketing hype was full of praise for the ability to have
> 2^12 (overlapping) segments, and to this day the Pentium has segment
> registers.  Worse, one reads of superservers addressing up to
> 64Gigabytes of main memory - must be programmed in assembler 'cause
> it's a long time since I've seen a compiler capable of producing
> segmented architecture code.
>
> > architecture was designed to replace it.
>
> No, not to "replace it", but to propagate it.
>
> ++L