- That is very interesting. Even if we had goos CPUs at 6.4 MHz, one clock would be about 125 nanoseconds. You can easily get SRAMS with 55 nanoseconds access time and even some with 15 to 20 nanoseconds if you can get hold of the cache RAMs that used to sit on PC mainboards.
Are do-it-yourself computers with self designed PC boards or wiring not a little to noisy for such time critical things? And how do you start each CPU at the right time after power-up or a reset?
--- In email@example.com, Lee Hart <leeahart@...> wrote:
> On 4/15/2011 10:58 AM, Richard wrote:
> > Microcontrollers are nice and well, I will even have to use a simple
> > one for the PS/2 port. The nostalgia value also is a point. When I
> > call it Elf, then there must be Elf inside :)
> > Microcontrollers would also be in the way for some ideas. Supposed we
> > had some interface with which we could upload data, commands to each
> > CPU card and download results, then I would also like to use the
> > interface to upload executable code. This way the other CPUs could
> > still be put to work, even when they are not needed for their
> > particular purpose at the moment. Then it will be practical when all
> > CPUs agree on one machine code.
> > Also, something like an operating system would be distributed among
> > the CPU cards, including the protocol for the interface. For this it
> > would again be nice to program in one machine code and not in
> > several.
> "johngodsey@" wrote:
> >> This reminds me of what the folks on the Parallax forums do with
> >> the Propeller. For extra processors attached to the 1802 why not
> >> use microcontrollers? Some of them can do video pretty well
> >> although it would take away from the nostalgia value.
> Well, I don't know about that... NASA used 1802's in spacecraft, and
> usually had a bunch of them rather than just one. The groups provided
> redundancy to improve reliability, and also allowed parallel processing
> for speed.
> The 1802 has 8 clock cycles for each bus cycle. All signals are latched
> at one of the 8 edges in each clock cycle; the data does not have be
> valid for the entire 8 cycles.
> This allows up to eight 1802's to be multiplexed to the same memory. The
> memory just needs a fast enough access time to respond in one clock
> cycle time (easy with modern memory chips).
> Simple example: Two 1802's running on the same clock frequency, but with
> their bus cycles 4 clocks out of phase. A multiplexer switches the
> memory's address and data back and forth between the two.
> Communications between the 1802's would be by flags in the shared
> memory. The 1802's could be running independent programs, or separate
> threads in one main program. FORTH would be one example of a language
> that already supports such threading.
> Lee A. Hart | Ring the bells that still can ring
> 814 8th Ave N | Forget the perfect offering
> Sartell MN 56377 | There is a crack in everything
> leeahart earthlink.net | That's how the light gets in -- Leonard Cohen
- Thanks. That was a bit hasty. It's on the list for the next changes. I will probably get home from work earlier tomorrow and then have a little more time.