I didn't say you need to call a bus a given thing. What I was suggesting is
using and staying close to the ISA hardware because ISA hardware is so
cheap. You can go to almost any thrift store and get a video board or a
serial board, or a disk controller for $3. U.S. They throw them away by
the 55-gal drumful.
Once you've got the hardware, YOU decide how to make the interrupts work,
and YOU decide how to use the DMA control lines. You can call if FRED for
all I care.
Have a look below for additional comments embedded in your reply.
Dick
-----Original Message-----
From: Tony Duell <ard_at_p850ug1.demon.co.uk>
To: Discussion re-collecting of classic computers
<classiccmp_at_u.washington.edu>
Date: Sunday, July 04, 1999 2:00 PM
Subject: Re: OT: A call to arms (sort of)
>> >
>> >Well, IMHO there is _if you want to call the bus ISA_.
>> >
>> >There's no reason not to use this type of DMA on a homebrew system.
>> >There's no reason not to use cards that have the same form factor and
>> >same connectors and ISA cards.
>> >
>> I've seen little reason to use DMA at all when processors generally have
the
>> capacity to move data at the bus bandwidth with block transfer
instructions.
>
>If the system is properly designed (adequate cache, etc) then the CPU can
>run a program from the cache _while_ a DMA transfer is going on. That's
>one advantage of DMA.
Well, I don't know what you want to do, so I can't size a cache for you.
Moreover, most DMA depends on external events to schedule its access to the
bus. If the external event is concurrent with the time the processor is
running from CACHE, it works out well, otherwise, it doesn't. I don't know
why you'd want DMA in a typical system. The CPU is quick enough to
accommplish the transfers as fast as they can go, and it just sits waiting,
generally, when it's allowing DMA, since it can't use the bus. Floppies,
for some stupid reason, use DMA in a PC, though it's not warranted, yet hard
disks, yes, including CDROMs generally don't. There are some PCI
controllers which use DMA, but we're talking ISA for now.
>> It's not a religious issue for me to call the bus whatever seems
>
>Hmmm... One thing I _really_ object to is calling a bus (or whatever) a
>standard name when it doesn't meet the standard. At one time Apple were
>particularly bad about this - they had 'RS422 ports' that were nothing
>like what they should have been, etc.
>
I agree that it's inappropriate to call an interface by a name unwarranted
by its characteristics.
>
>As I said, design your own bus, sure. If I was doing it, I'd use those
>DIN41612 connectors because you don't need to etch (and preferably
>gold-plate) PCBs. But don't call it ISA.
>
If I were using the DIN connectors, I'd get a standard out, read and
understand it, and then use it, calling it VME, its rightful name.
Unfortunately, I'd not be able to get a two-port serial board or a LAN board
for VME from the local thrift store.
>
>[As an aside, my MG1 workstation has the ISA adapter board in it. This is
>a 32016-based workstation, and the ISA adapter goes to great trouble to
>exactly emulate the ISA bus, even down to 'useless' things like a refresh
>address appearing on the address lines, all the clocks, etc. The result
>is that all ISA cards will work in this machine]
>
But to what end??? If you have a "decent" system already, why would you
want to use ISA. For that matter, why the crusade to invent a new bus?
>
>> with video boards. There aren't really any terribly interesting parallel
>> I/O boards, and if you want to use IEEE488 stuff you'll play hell beating
>> the WIndows software to help you do it.
>
>Eh? Maybe the Windows software is OK if all you want are 'virtual
>instruments' (but I really can't see the point of that - physical
>controls are a lot easier) but for real automatic test/measurement
>systems give me a decent GPIB driver, a good compiler and a real
>operating system...
>
A friend of mine uses it at the Cape and on the west coast for launching
rockets. I think it probably works O.K. I guess it's a matter of
preference.
>
>> >Suppose you don't have a central DMA controller. What do you propose
>> >doing with the DRQ/DACK signals? Sure you can make them effectively
>> >bus request signals. But now the peripheral card has to know to generate
>> >the address. And no standard ISA card would do that.
>> >
>> Just exactly which boards do you wish to use that do that? If they don't
>
>Err, any ISA card that uses DMA, like CD-ROM controllers, FDCs,
>soundcards, tape controllers, etc, etc,etc.
I know of no ISA I/O devices which require DMA. I know of only a few which
CAN use DMA, and of those, only the high-end SCSI adapters e.g. ADAPTEC 154x
series actually use it, albeit poorly. I have fairly fast equipment, in
general, and find that programmed I/O performs as well as DMA in the cases
I've tested, because the software I use does little else while the DMA is
progressing. Now I've seen little FDC based tape controllers, the kind you
use to interface one of those QIC toys to a PC which use DMA, but only
because they're grafting themselves onto the FDC handler. They don't NEED
DMA, they just inherited it.
>You may say you don't want the above. Fine. Design your own bus. But
>don't call it ISA, because it isn't..
>
Remember, I just suggested making the choices compatible with ISA cards to
save guys like yourself money. You're always complaining about the cost of
computer hardware.
>
>> >> too soon? I made no such assertion! There are lots of processors
which
>> >> have block transfer instructions which operate at the bus bandwidth.
>> Even
>> >> the Z80 did that.
>> >
>> >And IIRC the Z80 block moves were ridiculously slow...
>> >
>> Yes, perhaps it was ridiculously slow, but it was the bus bandwidth at
the
>> time.
>
>Not so. A Z80 LDIR doesn't hit the bus bandwidth _at all_. Heck, it
>fetches the instruction each time for one thing. IIRC at least one of the
>Z80 block instructions is slower than doing it in hand-optimised machine
>code.
>
Well, it was fast enough to transfer a whole track of data from an ST-506 in
one revolution of the disk . . . that's 10416 byte-times, nominally, in 16.6
ms . . .
>
>> >> which used it, I'd say that's a non-issue. You don't need the
schematic,
>> >> though,since the board you'll be using will be an IDE interface with
>> onboard
>> >
>> >For the <n>th time, the aim is to get 'open hardware'. That means (at
>> >least to me) available schematics. Not schematics of things that _might_
>> >work the same (e.g. WD1003 .vs. IDE). It means scheamtics and
>> >documentation for the hardware that's actually in the machine.
>> >
>> So, you want schematics of the disk drives as well, and the keyboard, and
>> the floppy drive? . . . and when you have them, how are you going to
stick
>
>YES!! That's exactly what I want.
>
>> your 'scope probe into that IC, and how are you going to fix it when it's
>> broken. It's a custom IC, after all, and they will cost 10x what a new
disk
>
>Why do you insist on wanting custom ICs? You can do an awful lot with
>standard chips, you know.
Except buy them, since they don't make them anymore.
>> drive costs if you try to buy just one. If "OPEN" means to you that you
>
>At the moment maybe. But in 20-30 years time it will be nice to be able
>to fix the machine when new modules simply aren't available. Now that you
>can't get new 5.25" drives, I'm darn glad I've got a pile of service
>manuals for them.
>
First of all you CAN get new ones, and secondly, death is inevitable. Once
they stop making the stuff you want to use, it's only a matter of time
before something you can't fix will fail.
>
>> have access and rights to all the intellectual property contained in your
>> computer, you might as well give up right now. If what you want is
enough
>
>Why? OK, I am not going get the rights to copy standard chips, but I
>don't see why the PCB schematics, etc can't be open.
The reason is because YOU don't make the rules regarding other people's
intellectual property. I agree it would be nice, but if a board costs $15
new, and $2 used, why would anyone worry about fixing them? You just buy a
couple of identical spares while the opportunity is in front of you. Most
commercial boards, nowadays, have custom firmware of some sort. There's no
reason at all why anyone should give you that. I doubt you'll have a
problem getting schematics for "standard" chips. It's the custom logic,
e.g. FPGA's, CPLD's, PAL's, and PROM's that will be the problem.
>As I have said many times before, I am using a '100% documented' PC. It's
>not open, in that schematics, etc are copyright IBM etc, but I didn't sign
>any NDAs to get them. The _only_ thing I don't have the schematics for is
>the hard disk. I have them for the motherboard, PSU, keyboard, monitor,
>floppy drives, CD-ROM, expansion cards, etc.
>
Yes, but as you said, you don't have the firmware listing for the keyboard
interface.
>
>> >I've found it _very_ hard to get data on the typical ASICs that you find
>> >on modern PC motherboards and I/O cards. In fact I've not managed to do
>> >it in a lot of cases.
>> >
>> That's the reason, precisely, why you don't use them.
>
>You're the one who wants to use standard PC I/O cards. I don't...
>
Well, what would you use instead? Would that save money? Would it save
time?
>
>-tony
>
Received on Sun Jul 04 1999 - 16:14:28 BST