old microcode listings
|
Prev: Software vs Hardware
Next: Searching for the PDP-3
From: "Peter "Firefly" Lund" on 5 Jan 2007 16:46 On Fri, 5 Jan 2007, ChrisQuayle wrote: >> Sure. First of all, they weren't /really/ in same time frame. >> The 8086/8088 got there years before the 68K. >> >> That tends to matter. > > I wasn't quite sure about some of the dates, so google for a timeline gets: > > 8080 and 6800 : 1974 > 6502: 1975 > > 8086: 1978 > 68k: 1979 > > So not much time slip between them. As far as I know it was a bit worse than that. The 68K was introduced in September, 1979 and the 8086 on June 8, 1978. I haven't been able to dig up any info on early availability of the 8086 but the 68K seems to have had problems. This is an early engineering sample of the 68000, serial #807, manufactured in October '79: http://www.cpu-world.com/CPUs/68000/Motorola-XC68000L%20(SN807).html That was the first, ceramic version. As far as I know, it took them ages to get to a plastic packaged version (which was cheaper) and to get a version out with an 8-bit bus (which was also cheaper). And when they did get out the 68008, they initially sold it at the same price as the 68000. -Peter
From: "Peter "Firefly" Lund" on 5 Jan 2007 17:18 On Fri, 5 Jan 2007, ChrisQuayle wrote: > I guess it depends on how much work you have done with a given cpu, as there Yes, familiarity matters, too :) > the 6502 Apple II and a 384k card for the 8086 Sirius (II ?). Still have the > Intel iAPX86 users manual somewhere, but remember it being hard work to build > any simple mental model for what was required for the design. Why? It should have been easier than the bank switching the 6502 forced you to?!? > revolutionary. I think what i'm trying to say is that the 68k was one of the > first micros to break the old mold mini model of accumulator + index register > architecture that was so pervasive in early micro designs. Yep. I like the 68000's addressing modes more than the 8086/8088's but I think the 386 (and 68020) had better addressing modes than both. The 386 could add together the value of an immediate displacement (8/16/32-bits with sign-extension), the value of a register, and then the value of another register scaled by 1/2/4/8 and use the result as an address. I know the 68020 got the same scaling capability a year earlier :) -- I'm just saying that sometimes elegance can be achieved by adding something. I don't think the pre-decrement/post-increment stuff was terribly important, though, and I've hardly missed it on other CPU's. Well, except for the 8051. It would have been nice with autoincrement and a second data pointer register pair. -Peter
From: kenney on 5 Jan 2007 17:43 In article <Pine.LNX.4.61.0701041929170.22558(a)ask.diku.dk>, firefly(a)diku.dk ( Lund) wrote: > Some say it didn't work so well but what do I know, I > never had one. I used to know people who had programmed for one, apparently it worked technically but the memory layout was real pain for programmers. Video memory was split up into 256 byte chunks scattered all over the address space IIRC > I don't know enough about TMS9900 and TMS99000 to have > an opinion other than its design led to good interrupt > response times Never used any machine based on those chips but reports are that it was slower than contemporary 8 bit machines. Later chip had 256 bytes of on chip memory for Register space. Addressing was IIRC limited to 16 bits anyway. Off course the main limit for people who were paying for home machines was price not performance. One of the main cost items in the 1980s was RAM. From memory RAM cost about twice per kilobyte then than it does a megabyte now and the difference could well have been higher. I remember paying about 380 sterling for a Video Geni, 16K of memory, cassette tape recorder for external storage and you had to use a TV for the display. Ken Young
From: Eric Smith on 6 Jan 2007 17:34 "Peter \"Firefly\" Lund" <firefly(a)diku.dk> writes: > My intention is to prove that the VAX could be pipelined and could be > made fast In what timeframe? It was pipelined and made fast in its later years. But it was getting harder to squeeze out more performance. DEC could probably have kept going down that path if they had wanted to (and been able to) spend as much money on it as Intel spent developing the Pentium and followons. Instead, they reasoned (correctly, IMHO) that they could get better raw performance by building a "speed demon" style RISC processor. And that it would be easier to take that architecture down the "braniac" path later than it would be to do so to the VAX. > My thesis is that the VAX could have been made fast relatively easily > and that the Alpha, however much we loved it, was a mistake. It might be "relatively easy" today to take a late-1980s VAX design and make it faster, but would it really have been "relatively easy" to do that in 1989? I'm skeptical. Eric
From: Eric Smith on 6 Jan 2007 17:38
I wrote: > of the MICROM chips, but there are also two PLAs in the control > chip that can't be easily extracted and affect microinstruction > sequencing, so the MICROM contents alone would be nearly worthless. "Peter \"Firefly\" Lund" wrote: > PLAs should not be too hard to extract, as long as there's no storage > involved. Oh, you mean they are embedded inside some other circuitry? Yes, there are two levels of PLAs (i.e., two AND arrays and to OR arrays) which take inputs from several internal registers, and whose outputs affect the operation of the microsequencer. > I think a cheap microscope and a camera is good enough. Tried that. Need a _good_ microscope and and a camera. > Perhaps just a camera with a "macro" setting Tried that too, with a 4 megapixel camera. Seems to be at least two orders of magnitude off. > or a camera + a magnifying glass. Haven't tried that. |