clipka wrote:

> * Did you think about how to implement a stack (and subroutine calls)?

Yes: With great difficulty. ;-)

There are instructions to load and save the PC. Your "stack pointer" is 
going to be a RAM register rather than a CPU register. You'll have to 
manually incriment/decriment it, which will be absurdly tedious. But it 
can be done.

> * The address space is not only piffling, but outright insufficient, 
> especially given the limited instruction set;

Well, there is that.

I'm not seriously expecting to make a usable computer out of it. I'll 
just be amazed if it actually ****ing works at all! Currently it works 
in the simulator, but I don't think the simulator is taking gate delays 
into account, and I'm really not sure how that affects all the 
edge-triggered circuits I've used. (!!) There's also the minor detail 
that even the trivial little design I have now is going to take an 
absurd number of gates to realise in hardware...

> some things you could do about it:
> 
> - As your commands are all 2 bytes wide, you can require that they be 
> 16-bit aligned, so you can increase address space available for programs 
> to 512 bytes (=256 instructions).

In other words, make memory 16-bit addressible rather than 8-bit 
addressible.

> - Harvard architecture.

That could work...

The other possibility, of course, is to just connect up more address 
lines. Makes little or no difference in software, but in hardware I 
suddenly have to put a hell of a lot more wires together. ;-)

-- 
http://blog.orphi.me.uk/
http://www.zazzle.com/MathematicalOrchid*

Post a reply to this message