Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Path: utzoo!mnetor!seismo!lll-lcc!pyramid!prls!mips!mash
From: mash@mips.UUCP (John Mashey)
Newsgroups: comp.arch
Subject: Re: Memory Management on microcomputers
Message-ID: <21@winchester.mips.UUCP>
Date: Sat, 14-Mar-87 17:23:27 EST
Article-I.D.: winchest.21
Posted: Sat Mar 14 17:23:27 1987
Date-Received: Mon, 16-Mar-87 04:55:35 EST
References: <4970@mit-eddie.MIT.EDU> <3810017@nucsrl.UUCP>
Reply-To: mash@winchester.UUCP (John Mashey)
Organization: MIPS Computer Systems, Sunnyvale, CA
Lines: 48

In article <3810017@nucsrl.UUCP> ram@nucsrl.UUCP (Raman Renu) writes:
>     Read the latest issue of "COMPUTER". Article by Milutinovic & ....
>     (I forget the name of the other author) have written an excellent
>     article on MMU for micros.  That article alone should be enough
>     to suggest pointers for other articles on MMUs.

The article is "A Survey of Microprocessor Architectures for Memory
Management", by Furht & Milutinovic, COMPUTER 20, 3(March 87), 48-67.

This is a clearly-written article that seems a worthwhile tutorial for
someone new to this topic.  However, be aware of a few caveats:
	a) The article appears a bit old: the latest reference cited
	is from mid-1984, hence if you're looking for a survey of the
	current state of the world, rather than a tutorial, this isn't the
	article. I'm not sure why, since COMPUTER generally has pretty short
	submission-publication timelags.
	b) Thus, there is no mention of things like:
		IBM PC/RT: ROMP + Rosetta (inverse mapping)
		Clipper: CAMMU combined cache/TLB chips
		MIPS R2000: onchip TLB with software refill & software-defined
			page-map organizations
		(Note that all of the above are actual commercial products).

	c) As a result of when the article was apparently written,
	some of what's said, although accurate (as far as I know), could
	easily mislead people as to the reality level of things being discussed.
	For example, a fair amount of the discussion revolves around the
	Intel i432 and Zilog Z80,000, in the same vein as discussions of
	286's, 68Ks, etc.  One would also expect from the article that most
	virtual 68K systems would use the corresponding MMUs, rather than
	the SRAM-based maps found in many of them [Suns, for example].

	d) In general, there is plenty of of factual information in the article,
	but little real-world evaluation, which is something one would like
	to be call something an excellent survey. There is a class of article
	that I'd call a "data-sheet" article, i.e., it passes on what is found
	in data-sheets, speaking of things that don't yet exist, may never
	exist, or may sort-of exist but don't really work very well. You
	sometime see all sorts of unevaluated comments passed along as
	gospel.  Anyway, the cited article is much better than that,
	although it does clearly fall into the "X is..." trap on occasion.

To summarize: a useful tutorial, but hardly an excellent survey of the current
state of the art.
-- 
-john mashey	DISCLAIMER: <generic disclaimer, I speak for me only, etc>
UUCP: 	{decvax,ucbvax,ihnp4}!decwrl!mips!mash, DDD:  	408-720-1700, x253
USPS: 	MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086