Re: LZ: Re: A few questions, a few comments, blahb
[Prev][Next][Index][Thread]
> On Tue, 9 Oct 1996 22:11:20 +0000, you wrote:
>
> > (1) It is too slow. Instructions are executed much faster (I
> >think a few hundred instructions per second, but correct me if I'm
> >wrong.) than the link port can handle (a few thousand bytes per
> >second).
> >
>
> Actually, the processor executes at about a few hundred *thousand*
> instructions per second. At a 6 Mhz clock, this makes each T-State
Sorry, thats what I meant to say. :)
> about 0.0000001 seconds, and with about 7 to 12 T-States making up the
> average instruction, were looking at almost a million instructions per
> second (1 MIP)! Of course, this pales in comparison to modern
> processors, but it's surely fast enough for our purposes. Your right,
> though, because every time you would want to fetch the next
> instruction from the expander, you would have to wait almost 10,000 T
> states to execute it! Talk about a slow down...
>
> > (2) It is not possible to snag the attention of the processor after
> >each instruction, except to add a wire inside. Even if this was
> >done, the processor would be forced to run at the speed of the link
> >port (see above).
> >
>
> I don't even think it would be possible by adding traces. In a Z80,
> there are two request lines, the memory request line and the IO
> request line. The IO request line goes low every time you access a
> port, which doesn't just happen when we want to access the link port.
> That line is also called when you access the timers, the screen, the
> keypad, etc. It would be an impossible mess to distinguish where the
> interrupt came from if we directly tied IOREQ to the IRQ line! Oh
> well, it doesn't really matter anyways.
>
What I meant was to tie the memory request/memory enable line (or
whatever it is called on the Z80) to the IRQ line. But this probably
still wouldn't work, and it wouldn't be worth it anyway.
Ben
shakal@ns.net
References: