Jihad needs scientists
in [Design]
|
From: krw on 11 Feb 2007 11:49 In article <87ire947jd.fsf(a)nonospaz.fatphil.org>, thefatphil_demunged(a)yahoo.co.uk says... > Eeyore <rabbitsfriendsandrelations(a)hotmail.com> writes: > > Ken Smith wrote: > > > > > IMO The best processor line Intel introduced was the 8051. It is too bad > > > that they didn't think to extend it in the obvious ways. Others have now > > > taken up the lead on that. > > > > The 8051's been going for 25 years now. > > > > Is there any other processor whose original core has lasted as long as is still > > in widespread volume use ? > > Freescale is still shipping bucket-loads of 6800 family and 68000 > family devices. By that argument, Intel and AMD are still making a few x86 family devices too. That's what, 29 years now? -- Keith
From: MassiveProng on 11 Feb 2007 11:54 On 11 Feb 2007 18:26:05 +0200, Phil Carmody <thefatphil_demunged(a)yahoo.co.uk> Gave us: >jmfbahciv(a)aol.com writes: >> > You can't even keep a memory long enough to know what a simple >> >discussion is about. >> >> You were the one who has a misconception that NT was a rewrite. > >You do realise that all the evidence, such as the fact that non-NT >windows at that time was increasingly written in x86 assembly, and >that NT was not just portable across many architectures, but was >actually first developed for non-x86 archs, and that the development >team under ex-DEC man Cutler was an entirely independent team from >the win3.x team, actually points towards "rewrite" being the most >appropriate word to use. > >Phil Yes, and WHEN it WAS ported to the x86 arch, it was MP capable, pre-emptively multi-taslking, and win 3.11 was not in any way capable of such things at all. Win 3.11 WAS a GUI over a largely re-used DOS underpinning. NT was most certainly not so. Thus re-write.
From: Phil Carmody on 11 Feb 2007 11:59 krw <krw(a)att.bizzzz> writes: > In article <87ire947jd.fsf(a)nonospaz.fatphil.org>, > thefatphil_demunged(a)yahoo.co.uk says... > > Eeyore <rabbitsfriendsandrelations(a)hotmail.com> writes: > > > Ken Smith wrote: > > > > > > > IMO The best processor line Intel introduced was the 8051. It is too bad > > > > that they didn't think to extend it in the obvious ways. Others have now > > > > taken up the lead on that. > > > > > > The 8051's been going for 25 years now. > > > > > > Is there any other processor whose original core has lasted as long as is still > > > in widespread volume use ? > > > > Freescale is still shipping bucket-loads of 6800 family and 68000 > > family devices. > > By that argument, Intel and AMD are still making a few x86 family > devices too. That's what, 29 years now? Care to highlight the major architectural differences between the current FSL offerings in the 6800 family to the original processor? And on-die serial controllers et al. do not count as part of the processor architecture. Phil -- "Home taping is killing big business profits. We left this side blank so you can help." -- Dead Kennedys, written upon the B-side of tapes of /In God We Trust, Inc./.
From: Ken Smith on 11 Feb 2007 12:22 In article <MPG.203912b7a00e4217989f97(a)news.individual.net>, krw <krw(a)att.bizzzz> wrote: >In article <eqlntm$421$3(a)blue.rahul.net>, kensmith(a)green.rahul.net >says... >> In article <MPG.2037d8dcaeddd806989f8e(a)news.individual.net>, >> krw <krw(a)att.bizzzz> wrote: >> >In article <eqkv5i$2v2$5(a)blue.rahul.net>, kensmith(a)green.rahul.net >> >says... >> > >> ><snip> >> > >> >> Intel tried to sell the 432 at just about the same time as they tried >> >> to sell the 286. It didn't take a genius to notice that N 286s with some >> >> overhead would out perform N 432s. >> > >> >The iAPX432 came out in '81, a tad before the 80286. I don't think >> >Intel actually tried to sell it. It was a real turkey, performance >> >wise. Too much hardware falling all over itself. >> >> They were already trying to sell at the time they were trying to sell the >> 432. You couldn't get either chip(s). The 432 required two chips and >> they ran too hot to be shipped. > >Nope. The '432 was announced well before the '286. I was at the >Intel unveiling (one of IBM's representatives) at the Grand Hyatt >in NYC. BTW, the 432 was three chips, not two. Did you actually see any chips that worked. At the time I was visited by the Intel folks, the 432 was not yet available and the 286 was already planned. This was late 1981 or early 1982. > >> >> The 432 was supposed to do >> >> multiprocessor systems efficiently but it failed badly. >> > >> >The 432 had nothing to do with multi-processor. It was actually >> >quite like IBM's FS in the early '70s (which was killed before >> >implementation) and the AS400. >> >> The advertising they did to me touted the multiprocessor application of >> the device. > >That wasn't by any means the supposed advantage of the '432. MP >had been done before. It was supposed the be their "mainframe" >entry. Everyone laughed. The sales guy acted dead serious when he said it. It may have been a last gasp effort before they killed it. Since the 286 was going to come out, they may have seen this as the only remaining parket. >It's most "interesting" aspect was the single-level store. Quite >like IBM's ill-fated FS and later AS400 (as I mentioned). I thought the most interesting was the protection against attempts to do inappropriate operations of variables. It doubled the number of memory operations needed so it cost hugely on perfromance. >> >Nope. 432 <> Itanic. ;-) The 432 was supposed to be a "micro- >> >mainframe". It had nothing to do with x86, architecturally or >> >market-wise. >> >> I suggest that it did. Mot was working on a 16 bit machine. Intel only >> had a 8 bit machine. If the market went for 16 bit and swung over to the >> 68K, it would cut into the cash flow into Intel until the 432 got to >> market. > >I suggest that you have no clue. 80868 is a 16bit machine. You >might want to try to count again. The PC market wasn't going to >"swing" anywhere, particularly to the 68K. Your argument is simply >silly. I said "the market" not the "PC market". I suggest you go back and reread what I wrote. It was correct. You simply missed the point because you were thinking PC. >> >> The X86 ended >> >> up with so many bad ideas in it because the real brains were being wasted >> >> on the 432 project. >> > >> >Wrong. x86 was extended because they could. Backwards >> >compatibility is key, much like the history behind the IBM 360->z9. >> >Break backwards compatibility and the competition is on an even >> >playing field. >> >> I disagree. > >Get real! > >> The 8086 processor had a huge collection of bad ideas in it. > >...and you think the 8051 was the greatest processor of all time?! ><boggle> No, I said it was the best on Intel introduced. I suggest again that you read more carefully. >> Needed 32 bits worth of register to access 20 bits worth of address space >> was only one of many. > >So what? One didn't change segments on every access. Indeed one >could say the opposite, one only had to supply 16bits of a 20bit >address 99% of the time. ...more efficient. No, it used up a 16 bit register to supply only 4 more bits of addressing. It also required a trip through the ALU to do a memory access. This is part of why the 8086 was so darn slow. >> The fact that a trip through the ALU was needed for >> a jump instruction was another. > >Microarchitecture nit to save gates. Another adder could have been >used but the gates weren't available. Later x86 processors didn't >do this (though the P4 used the FPU for FX multiply; just as >stupid). No, an implementation of a stupid idea. It means that you either had to run through the existing ALU or add one just to get to memory. This means that instead of using the ALU to do usefule stuff, you had to wait while it fiddled about. >> The x86 had to keep the 8086 instruction >> set because Intel knew that they couldn't strand the code and expect to >> sell many processors. > >Wrong! They had the PC market and wanted to keep it. A radical >change would have put them in the dumper. See: Itanic. You just rephrased what I said after calling me wrong. > >> >> IMO The best processor line Intel introduced was the 8051. It is too bad >> >> that they didn't think to extend it in the obvious ways. Others have now >> >> taken up the lead on that. >> > >> >Best processor? You think the x86 is kludgey and like the '51? >> >I've used the 8051 several times and wouldn't shy away again, but >> >the ISA is a mess. >> >> I think the x86 is kludgy in a more extreme way than the 51. The silly >> fools put in stuff like "ascii adjust for divide" but no 32 bit add >> instruction. The REP prefix and the LOOP instruction both imply the CX >> register. > >Ascii adjust was useful and simple enough to implement (no large >use of valuable transistors). How many programs have you ever used it in? Its a stupid instruction to bother to implement. > >Why would there be a 32b add in a 16b processor? That would >require a 32b ALU or microcode specifically for that one >instruction. Sheesh, what a silly argument! The 8080 could add 16 bit values. >Since SC is the "count" register, it sorta makes sense that it >would be used for loop counts and repeats, eh? Assuming you mean "CX" not "SC" It makes many routines take extra instructions because they didn't have the ability to redirect the register used. > You could have >complained that x86 wasn't as orthogonal as you like, but the 8051 >is even less so. Compared to even x86, the 8051 is a mess. I still disagree. The 8086 was much worse and for no good reason. The 8051 was a very limited machine to keep it simple. From the point of view of actually creating programs with it, the instruction set is not too bad. There is only one opcode not used. If you wanted to add more instructions, you'd have to give up something. > >-- > Keith -- -- kensmith(a)rahul.net forging knowledge
From: krw on 11 Feb 2007 13:11
In article <87zm7k3a0l.fsf(a)nonospaz.fatphil.org>, thefatphil_demunged(a)yahoo.co.uk says... > krw <krw(a)att.bizzzz> writes: > > > In article <87ire947jd.fsf(a)nonospaz.fatphil.org>, > > thefatphil_demunged(a)yahoo.co.uk says... > > > Eeyore <rabbitsfriendsandrelations(a)hotmail.com> writes: > > > > Ken Smith wrote: > > > > > > > > > IMO The best processor line Intel introduced was the 8051. It is too bad > > > > > that they didn't think to extend it in the obvious ways. Others have now > > > > > taken up the lead on that. > > > > > > > > The 8051's been going for 25 years now. > > > > > > > > Is there any other processor whose original core has lasted as long as is still > > > > in widespread volume use ? > > > > > > Freescale is still shipping bucket-loads of 6800 family and 68000 > > > family devices. > > > > By that argument, Intel and AMD are still making a few x86 family > > devices too. That's what, 29 years now? > > Care to highlight the major architectural differences between the > current FSL offerings in the 6800 family to the original processor? > And on-die serial controllers et al. do not count as part of the > processor architecture. A Core-2-duo will still execute 8088 binaries. -- Keith |