Path: utzoo!attcan!uunet!lll-winken!ames!eos!labrea!decwrl!decvax!ima!think!barmar
From: barmar@think.COM (Barry Margolin)
Newsgroups: comp.sys.mac
Subject: Re: Virtual memory init
Message-ID: <35654@think.UUCP>
Date: 25 Jan 89 05:34:59 GMT
References: <1542@csuna.UUCP> <76000334@p.cs.uiuc.edu> <13871@cup.portal.com>
Sender: news@think.UUCP
Reply-To: barmar@kulla.think.com.UUCP (Barry Margolin)
Organization: Thinking Machines Corporation, Cambridge MA, USA
Lines: 43

In article <13871@cup.portal.com> bcase@cup.portal.com (Brian bcase Case) writes:
[Someone else writes:]
>>Furthermore, nobody believed that VM could be implemented so quickly
>>on the macintosh
>Right, and it still hasn't been done.  Connectix hasn't done real
>virtual memory:  Mac applications must still be assigned and wired-down
>to places in the 8 MByte memory space.

Virtual memory is any scheme that allows applications to be fooled
into thinking that there is more memory than there actually is.
Individual applications don't have to have independent address spaces
for the memory to be virtual.  My Symbolics lisp machine provides
preemptive multitasking, but all applications (and the OS) still live
in the same 256-megabyte address space.  And on Multics, each login
session is a single process, with all applications that the user runs
during that session in that address space.

It's possible to have time sharing, protected applications,
inter-process communication, etc. without virtual memory, too.  For
example, a system that simply provides each application with the
normal physical address space and swaps all of physical memory in and
out is possible (and very similar to the first computer I used, a
PDP-8 running TSS-8).  The point is that these are independent aspects
of an OS.  There are advantages and disadvantages to each.

I think you are also misusing the term "wired-down".  In operating
systems, this term is used to mean locking a particular portion of
virtual memory to a portion of physical memory, and preventing it from
being swapped out.  It's generally used for things like I/O buffers in
DMA systems, where the device must be told a physical address in which
to store or retrieve its data, and it wouldn't be nice if the VM
mechanism were to replace it when the device isn't looking.  You're
using it to mean that an application is loaded into a particular place
in VM and it stays there, which is true on every system I'm aware of
(on systems where each application gets its own address space, every
application is normally "wired-down" to the same location (e.g. 0) in
its own space).

Barry Margolin
Thinking Machines Corp.

barmar@think.com
{uunet,harvard}!think!barmar