Aucbvax.5502 fa.works utzoo!decvax!ucbvax!works Thu Dec 17 20:54:11 1981 WorkS Digest V1 #45 >From JSol@RUTGERS Thu Dec 17 20:50:24 1981 Works Digest Friday, 18 Dec 1981 Volume 1 : Issue 45 Today's Topics: MC68000 Paging Schemes 64K And 256K Memory Chips - Is The US Falling Behind? [Sorry for the delay putting out digests, there hasn't been too much to send out -JSOL] ---------------------------------------------------------------------- Date: 16 Dec 1981 1624-PST From: Jim McGrath Subject: 64K and 256K memory chips Sender: CSD.MCGRATH at SU-SCORE Reply-To: CSD.MCGRATH at SU-SCORE I have been talking to some business folks who generally seem to have a good feel for the chip market. They were bemoaning the fact that US manufacturers have pretty much given the 64K market to the Japanese (only Motorola and TI are selling those chips in any significant quantity, and the Japanese have captured an estimated 60 to 70% of the market). I was wondering what people in the "front lines" of chip manufacture think about this. It appears that the US spent far too much time and energy on trying to come up with a denser chip to increase yeilds, resulting in practically no one coming up with a 64K chip for the general market (as opposed to in house IBM and ATT use). With this head start, the Japanese could head us off in the 256K market as well, although perhaps all the time we have spent on the 64K chip will give us a technical edge there. This is an important issue. If US industry falls behind on 64K and 256K chips for the mass market, we will lose the computer manufacturer market, which will decrease revenue and then really knock us out of the R&D race. And once the memory chips go, can logic chips be far behind? Eventually our "cutting edge" industry will subsist on DoD handouts (since they are a more reliable source of chips than the Japanese). Comments? Jim ------------------------------ Date: 14 Dec 1981 0950-PST From: Ian H. Merritt Precisely. A real implementation is onw which works. Some are better than others, but I refer to a real implementation as one which doesn't limit the instruction set at all. The 2 chip solution is a royal kluge, but it works and it doesn't have any effect on the available set of instructions; only the effective speed. IHM ------------------------------ Date: 14 December 1981 20:33-EST From: Robert A. Morris Subject: M68000 paging schemes. I have the impression that those who have adopted the two cpu schemes, eg. Apollo Computer, to overcome hardware deficiencies on pagefaults have NOT restricted the instructions useful on their machines. This would lose the benefit of the two cpu scheme. Restrictions on instructions are adopted by those who can or wish to enforce them or depend on "scouts honor" programming. I know vendors doing this who expect Motorola to produce correctly working chips before they bring any products to market and do their program development with high-level langauge compilers which circumvent the offending instructions. Such development presumes that the compilers can easily be fixed to generate more effective code when the chip handles the fault correctly and/or when the program is to run in a non-paging environement. --bob morris ------------------------------ Date: 17 December 1981 01:44-EST From: Leonard N. Foner Subject: The Real Implementation of Subject: the MC68000 Demand Paging Algorithm This was a project I was working on about a year ago. The scoop: You CAN use ANY insruction in the processor with the dual-processor system. Essentially, what happens on a page fault is that the CPU gets a LONG memory wait for the next byte. The other processor handles the stuff. Of course, you can't go off and run somebody else while you're waiting for the next page to come into memory, unlike reasonable VM machines like VAXen, but this will work. For this reason, you probably also wanna use a stacklike cache of pages that are bound for disk, to decrease "expensive" page faults which have to hit the disk. (That way, you've effectively got a larger working set, since only the page last accessed in \that/ cache gets stuck on disk, when you've gotta bring a page in from disk on a fault.) I'm not at all sure how much of an advantage this makes when you can't run somebody else during the fault, though... but good references on good paging algorithms are in the VAX Hardware Handbook. There \must/ be theoretical treatments around on how to do it efficiently, and that will really help in this situation. Obviously, how fancy you wanna get on paging is a tradeoff between price and performance, pure and simple... with the law of diminishing returns thrown in as an extra added feature. Have fun. ------------------------------ End of WorkS Digest ******************* ------- ----------------------------------------------------------------- gopher://quux.org/ conversion by John Goerzen of http://communication.ucsd.edu/A-News/ This Usenet Oldnews Archive article may be copied and distributed freely, provided: 1. There is no money collected for the text(s) of the articles. 2. The following notice remains appended to each copy: The Usenet Oldnews Archive: Compilation Copyright (C) 1981, 1996 Bruce Jones, Henry Spencer, David Wiseman.