Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Posting-Version: version B 2.10.2 9/18/84; site watdcsu.UUCP Path: utzoo!watmath!watnot!watdcsu!herbie From: herbie@watdcsu.UUCP (Herb Chong [DCS]) Newsgroups: net.micro Subject: Re: demise of 8086 family? Message-ID: <1472@watdcsu.UUCP> Date: Thu, 13-Jun-85 10:55:36 EDT Article-I.D.: watdcsu.1472 Posted: Thu Jun 13 10:55:36 1985 Date-Received: Fri, 14-Jun-85 00:06:25 EDT References: <120@SCINEWS.UUCP> <656@sphinx.UChicago.UUCP> Reply-To: herbie@watdcsu.UUCP (Herb Chong [DCS]) Organization: U of Waterloo Lines: 83 Summary: In article shor@sphinx.UChicago.UUCP (Melinda Shore) writes: >> From: ned@SCINEWS.UUCP (Ned Robie) >> When IBM comes out with their 370 on a chip(s), the future of the 8086 >> family is, I believe, in serious question. > >Too late. The XT/370 has been around for at least 6 months, and clearly >hasn't done the Intel chips any damage. it's been around closer to 16 months. i was at demo in february last year. it uses the 8068 as a fileserver and first interrupt handler and a 68000 for most of the other work. the AT/370 does something similar with the 286. it was announced about 6 months ago. i also might mention that IBM has had a real 370 on a chip for at least two years. the word is that IBM doesn't think anybody will need a micro with the power of a 4341 (that's about 1.5 780's) in a package the size of a PC. with the success of the DEC microvax I and probable success of the microvax II, they may change their mind. something that powerful will be I/O bound most of the time given current small hard disk technology. >> As for 8086 segmentation... IBM needs to give its customers some reason >> to switch to the 370 when it's ready. Segmentation is simply planned >> obsolescence. > >I dunno about that -- I have nothing against segmentation per se. 64K >fixed-size segments, though ... just to clarify a little, the 370 is a segmented architecture from two different points of view. the operating system references data in segments of 1M or 64K, depending on the hardware facilties enabled. the virtual storage management instructions must deal with the appropriate segment sizes. of course, no-one writes their own operating system for 370's much anymore, so there's not much need for people to worry about it. handling data areas larger than one segment size becomes a real excercise. code must handle units of 4K in code size. the base-displacement scheme of handling instruction addressing requires that a base register be pointing to the start of a 4K page containing the code to be executed. if the code is larger than 4K, either another base register needs to be allocated or the current base register changed to point to the appropriate location. fortuantely, data can be viewed as residing in a flat address space, so that a compiler doesn't have to worry about changing registers for addressing data as much. the pdp-11 is essentially a segmented machine (when you consider split I and D). a LOT of software gets by with the limited segment size. the amount of system software available to make this transition as painless as possible allows the user to ignore this most of the time. >What, by the way, would be the point of dragging the 370 architecture over >the long haul? Is this something we really want, given newer >architectures (National Semi 32K chips, RISC, etc.)? The Intel family has >its problems, but there are alternatives other than the 370. There's been >lots of progress in machine architectures -- let's take advantage of them. there is one advantage, and IBM is using it, compatibility. IBM has already made a statement of direction implicitly by the existence of the XT/370 and AT/370. they want to have an integrated system of micros, minis, and mainframes running the same software on the same architecture. DEC appears to be thinking the same with the microvax. software development costs are reduced, training of people is reduced, and a system can grow with the user as more power is required without losing the use of software developed earlier. face it, IBM is in the market of selling to corporate customers. it's almost coincidence that the average consumer is capable of affording a PC. right now, no-one else can offer an object code compatible system of hardware spanning almost a 1000 to 1 range on CPU speed. whatever, or whoever, you are in the computer industry, you have to watch IBM. with their research and development facilities, marketing force, and sheer size, they can strongly influence the industry almost whenever they chose to. this can be a good thing, and it can be a bad thing. Herb Chong... I'm user-friendly -- I don't byte, I nybble.... UUCP: {decvax|utzoo|ihnp4|allegra|clyde}!watmath!water!watdcsu!herbie CSNET: herbie%watdcsu@waterloo.csnet ARPA: herbie%watdcsu%waterloo.csnet@csnet-relay.arpa NETNORTH, BITNET, EARN: herbie@watdcs, herbie@watdcsu