Homebrewing a digital computer

From: ajp166 <ajp166_at_bellatlantic.net>
Date: Sat Nov 11 22:31:02 2000

From: Kevin Schoedel <schoedel_at_kw.igs.net>


>Sorry, I didn't express myself clearly. In Chuck's machine the
>displayed value comes from a 12 line I/O bus. He could build two
>different display devices to attach to the bus: one for everyday use,
>with LEDs and bcd-to-7-segment decoders, and one special occasions,
>with decoders and drivers and sockets for the nixie tubes that he
>borrows from whatever machine they otherwise belong to.


You can but nixies are a pita to drive. Of era lamping would be
bulb per bit. cheaper too.

>Yes, he could make 4xx an encoded operation (like pdp8 operate
>instructions) such that 400 happens to perform a clear. Or he could
>make 4xx load the accumulator with 0xx, so that 400 is again a clear;
>this may fit the spirit of the design better. In either case he need
>not disclose the extra features at first :-)


Exactly. Or the feature could be enabled/disables using a OPR like
instruction.

>Yes. But since I'm doing a pure 8-bit machine, including 8-bit memory
>address, it's *already* limited significantly. I do currently have one
>unassigned opcode, so I could add an I/O instruction if it turns out I
>have enough board space. More likely I'll just move on to design #2,
>where I can hide complexity in microcode.


You'll find it doesnt hide. The features grow to increase complexity. ;)

>
>Yes, I'm still considering something like that, though I'd probably
>keep it simpler, and just leave the PC in the accumulator: PC->AC,
>MA->PC. I know this probably seems absurdly stripped down to anyone who
>already knows what they're doing, but I don't, so I want to keep the
>control very, very basic, to reduce the chance of serious errors.


Not an unreasonable way. Worth considering as it has good
register economy.

>I think you're right. I know my current addressing scheme is too
>complicated, and I no longer know why I've been so reluctant to give up
>page 0 addressing.

With say n-bit word and n-bit indirect address page 0 is more
efficient but not required. What it does do for short words like
8bit it eats a bit that could be an address.

>Do you have any comment on PC-page (upper bits from
>PC, lower bits from instruction) versus PC-relative (PC + lower bits
>from instruction, either sign extended or not)?


Both work. If you're using an indirect address (like PDP-8) then go
with PC-page as it much simpler. PC-relative is ok but requires an
adder (or more gating to reuse the only adder).

If there is one trick to keeping things simple, the fewer paths and
gating of things will help simplify.


Allison
Received on Sat Nov 11 2000 - 22:31:02 GMT