jcallen@Encore.COM (Jerry Callen) (11/17/90)
Speaking of stack-oriented architectures, has anyone else noticed that Harris has stopped work on a 32 bit version of the RTX-2000 stack-based processor? There was a blurb to this effect in the November 12th issue of Electronic Buyer's News, of all places. I think this is the processor Harris touted in ads with a big fish hook, saying something like "Don't buy that line about what a RISC can do for your embedded system." Note that Harris is using the Moto 88K in its line of real-time computers. I don't know much about the RTX-2000 other than it is stack-based and its preferred language is Forth. Does anyone have more details about either the chip or the cancellation? -- Jerry Callen jcallen@encore.com
koopman@a.gp.cs.cmu.edu (Philip Koopman) (11/17/90)
In article <13287@encore.Encore.COM>, jcallen@Encore.COM (Jerry Callen) writes: > Speaking of stack-oriented architectures, has anyone else noticed > that Harris has stopped work on a 32 bit version of the RTX-2000 > stack-based processor? There was a blurb to this effect in the > November 12th issue of Electronic Buyer's News, of all places. This is an understatement. To quote an official Harris policy statement: "Harris has decided to discontinue future investments in new RTX standard products and development tools...". A more comprehensive version of this statement was posted on comp.lang.forth. The reasoning was that they are de-emphasizing digital in order to concentrate on analog, power, and mixed signal products. So, they are severely curtailing work (one is tempted to say, somewhat incorrectly stopped work), on all 16-bit RTX systems, and have definitely stopped all 32-bit RTX work. RTX is by no means the only project to suffer this fate at Harris, but certainly the highest profile project to be cut back. > I think this is the processor Harris touted in ads with a big fish > hook, saying something like "Don't buy that line about what a RISC > can do for your embedded system." Note that Harris is using the > Moto 88K in its line of real-time computers. Yes, this is the processor (and they took some heat for those ads too...). 1) The RTX was, IMO, an excellent solution for moderate- to low- cost embedded controllers. We saw several instances where RISC system solutions were just too expensive or not capable of doing the job (because of predictability problems or high interrupt overhead, usually) in things such as disk drive controllers and LAN controllers. Harris made its decision based on cash flow and high-level management woes, not on the quality of RTX technology 2) The division of Harris that makes the Nighthawk is not the same division as Harris Semiconductor. They don't talk. They don't collaborate. It seems that they'd rather avoid getting involved with each other. Also, the Nighthawk is a "big" real time system, clearly out of the class of system the RTX 2000 was ever intended for, so the technology would have been an inappropriate match. The 32-bit RTX maybe, but that was never really far enough along to be considered for the Nighthawk. > I don't know much about the RTX-2000 other than it is stack-based > and its preferred language is Forth. Does anyone have more details > about either the chip or the cancellation? The RTX 2000 was a stack-based "Forth" machine. But, it also had a C compiler. Its strength was low system cost and high real time responsiveness. Its weakness was being different than all the rest of the world. It appears that since Harris won't give it a real chance, we may never find out whether it was really a better solution. If you want to know more about stack oriented architectures, I recommend my book "Stack Computers", P. Koopman, Ellis Horwood Ltd, 1989. You can get it through inter-library loan, so this is not a pitch to actually buy it. Also, there is a summary of stack-based machines in Jack Woehr's article "Forth Machines" in the November 1990 issue of Embedded Systems Programming (although this article is perhaps as much or more about the players than the actual technology). Phil Koopman koopman@greyhound.ece.cmu.edu Arpanet 2525A Wexford Run Rd. Wexford, PA 15090 *** this space for rent *** I _used_ to be the senior scientist in charge of architecture for the RTX group at Harris. But, the above is only me talking since I don't work for Harris anymore.
my@dtg.nsc.com (Michael Yip) (11/17/90)
I know that one of our customer uses the Harris Forth processor on a networking product and he was quite please with the processor's performance. -- Mike
shri@ncst.ernet.in (H.Shrikumar) (11/18/90)
In article <13287@encore.Encore.COM> jcallen@encore.Com (Jerry Callen) writes: >Speaking of stack-oriented architectures, has anyone else noticed that Harris >has stopped work on a 32 bit version of the RTX-2000 stack-based processor? It happened in comp.lang.forth a while ago, it did not occur to me that comp.arch might be interested. Quoted below is a recent comp.lang.forth article from Phil Koopman (who was in Harris till this event, and is a regular in the forth newsgroup). which fills in the story. >I think this is the processor Harris touted in ads with a big fish hook, saying >something like "Don't buy that line about what a RISC can do for your embedded >system." Note that Harris is using the Moto 88K in its line of real-time >computers. This is news to c.l.f ... maybe I ought to quote this there :-) >I don't know much about the RTX-2000 other than it is stack-based and its >preferred language is Forth. Does anyone have more details about either the >chip or the cancellation? (1) Quite a few external memory busses (heap, stack, dictionary), (2) absolutely no pipelining (predictable instruction timing, said to be good for tight "enbedded controller" loops). (3) Quite fast clock speeds, (10 Mhz and above ?) By 2 and 3 above memory had to be 25 ns static ram, cant do much with traditonal caches. And by 1 above, would need three banks of such RAM. so bound to be expensive board, OK for a small memory requirement, fast predictable response controller application. But nevertheless, (IMHO, plus some hand-waving) you could build a cache system if you were smart enough to exploit some a-priori knowledge of the Forth language. (what, how can the h/w engg talk to a s/w engg ! refer comp.arch about a year ago :-) Also Harris had not-long-ago organised a big, and fairly well recieved RTX design contest (winner gets a RTX board, and $10K; and a RTX board for a 1000 good design submissions) and had announced intensions of C for the RTX family ... and then this bolt! >-- Jerry Callen > jcallen@encore.com ----------------------------------------------------------------------------- Path: ncst.ernet.in!uunet!snorkelwacker.mit.edu!usc!wuarchive!ukma!uflorida!mlb.semi.harris.com!jujeh.mlb.semi.harris.com!krl From: krl@jujeh.mlb.semi.harris.com (Ken Lyons) Newsgroups: comp.lang.forth Subject: Re: Forth Engines / Harris Message-ID: <1990Nov14.150715.28620@mlb.semi.harris.com> Date: 14 Nov 90 15:07:15 GMT References: <1982.UUL1.3#5129@willett.pgh.pa.us> Sender: news@mlb.semi.harris.com Organization: Harris Semiconductor, Melbourne FL Lines: 34 Nntp-Posting-Host: jujeh.mlb.semi.harris.com For starters, everything to follow is my opinion and not the official position of Harris. Besides, Friday is my last day there. I think the primary reason for cutting funds to the RTX line (which is not yet self-supporting) is that the market is down and sales are lousy. That doesn't necesssarily answer the question, "Why RTX and not something else?" It seems that presently the payback is better for other lines, so the decision appears to be made for short-term benifits, which is all too usual. In terms of long-term benifits, there are serious obsticles to overcome when introducing a new microprocessor architecture, market inertia for one. It appears a lot easier for a customer to stick with a given vendor than to relearn a new architecture. It is the path of least resistance to say, "We're an Intel house," or, "We're a Motorola house." It takes time to investigate all the possibilities so the easy or exciting ones are examined while the rest accumulate in a file cabinet somewhere. This is not to say that the big companies don't offer advantages: they have the critical mass of sales to provide 99% of the expensive support their customers want and still turn a profit. In addition to this, for a new processor to be exciting it must provide at least 2X obvious bang for the buck. When introducing a new architecture, the improvement factor has a large, application dependent range, so the advantage is not so obvious. If after considering all this, you still want to do something new, I would bet that the intellectual property rights to RTX could be picked ----(*) up cheap. There were many ideas we had to improve the architecture that may be lost all because of a stubborn refusal to do things right the first time. All the same it's been fun; no regrets. Regards, Ken Lyons Hardware Designer, looking. ----------------------------------------------------------------------------- (Used to say "Harris" above here -- shri) (*) Comment: Due to a technicality, ownership of the intellectual property rights could (or has ?) revert back to Novix, the original co. wanting to make the FORTH chip. -- shri ....and Phil Koopman's comment to that.... ----------------------------------------------------------------------------- Path: ncst.ernet.in!uunet!crdgw1!rpi!zaphod.mps.ohio-state.edu!ub!uhura.cc.rochester.edu!rochester!pt.cs.cmu.edu!a.gp.cs.cmu.edu!koopman From: koopman@a.gp.cs.cmu.edu (Philip Koopman) Newsgroups: comp.lang.forth Subject: Re: Forth Engines / Harris Summary: a supplemental view on what went wrong Message-ID: <11111@pt.cs.cmu.edu> Date: 15 Nov 90 11:42:55 GMT References: <1982.UUL1.3#5129@willett.pgh.pa.us> <1990Nov14.150715.28620@mlb.semi.harris.com> Organization: Carnegie-Mellon University, CS/RI Lines: 56 In article <1990Nov14.150715.28620@mlb.semi.harris.com>, krl@jujeh.mlb.semi.harris.com (Ken Lyons) writes: > For starters, everything to follow is my opinion and not the official > position of Harris. Besides, Friday is my last day there. I don't work for Harris now, so my opinion is strictly my own as well. Ken Lyons writes: > I think the primary reason for cutting funds to the RTX line (which is > not yet self-supporting) is that the market is down and sales are > lousy. That doesn't necessarily answer the question, "Why RTX and > not something else?" It seems that presently the payback is better > for other lines, so the decision appears to be made for short-term > benefits, which is all too usual. I think that the "market down/sales lousy" reason (which is superficially true) is more a convenient excuse for upper-level Harris management than a reality. From the market research I saw while at Harris, it was evident that it _always_ takes at least 3 years for a new proprietary processor to become self-sufficient (this means one with a new instruction set, not a next-generation compatible chip). You can't make money by selling one- and ten-piece orders to people building prototypes -- but you make lots of money a year or two later when they ramp up production. RTX has always been given insufficient resources (and now, insufficient time) to achieve success within Harris. Look at the Transputer -- about 10 years and perhaps just barely there (but, still funded and growing every year). It is true that Harris Semiconductor has a tremendous cash flow crunch. This, combined with a sector-wide deemphasis of digital technology spelled doom for RTX despite the fact that it was in reasonably good shape for long-term customer growth. So, the demise of RTX had almost nothing to do with technology problems. It had something to do with marketing (but, they were getting smarter and could probably have recovered if given the chance). It had something to do with Harris' problems producing low-cost RTX silicon on their old 2.0 micron fab line (RTX's can be made for much less than it cost Harris to make them, thus lowering price, but you really want to be on 1.0 micron technology to do it). It had a lot to do with cash flow (which is a stupid reason to cancel a new processor line when you have N large customers poised on placing large orders). It had even more to do with what Harris would like to call a "deemphasis" on digital technology, meaning that *any* new proprietary processor (stack, register, whatever) wouldn't fit into their plan of concentrating on analog, power, and mixed signal. I would say that RTX technology still makes sense for systems that must have high real-time performance at moderate cost. The question is, who is going to do it? Phil Koopman koopman@greyhound.ece.cmu.edu Arpanet 2525A Wexford Run Rd. Wexford, PA 15090 *** this space for rent *** ----------------------------------------------------------------------------- ( Used to say "Harris" here too -- shri ) -- shrikumar ( shri@ncst.in )