Date   : Wed, 10 Mar 2010 23:15:31 +0100
From   : rick@... (Rick Murray)
Subject: Tube on the Archimedes

On 10/03/2010 20:10, Rob wrote:

> I'm no expert, but the RISC machines do seem to support multiple
> processors - the 486 boards being a case in point.  I've always
> assumed that they work in a similar way to the Tube on the 8-bit
> systems, but I've no real basis for that assumption..

IIRC, there's an ASIC on the 486 card. The whole thing is mapped into 
memory, but somewhat more complicated as the second processor is also 
capable of addressing memory (it would be kinda stupid to need the ARM 
to read memory for it...). The control of the co-pro works along the 
lines of setting up the environment (memory, boot code, blah-de-blah), 
programming the ASIC, and then saying "run with this ball".

To make an assumption of my own, I would imagine the hardware part of 
the system is fairly simple (in terms of an actual PC); a CPU and the 
interface-glue chip. This support logic probably recognises certain 
standard memory locations, so that when the parallel port hardware would 
be accessed on a normal PC, it would generate some sort of interrupt to 
bring in the emulation side of things to "read" the operation wanted, do 
it, then push in the expected response.


There are numerous similarities with how the TUBE operates (async slave 
processor with I/O links to host), but it's an altogether different setup.
Aleph1's !PC/Diva code is available, for those with the patience to 
examine it. Not me, I can't off the top of my head, roll enough x86 code 
to read two values from memory and add them...


Best wishes,

Rick.


PS: Somebody else mentioned the BBC I/O podule.
     Analogue port, User port, and 1MHz bus. Plus a complicated set of
     links for interrupt support I haven't figured out how the heck it
     works. You might even have MIDI if yours isn't the "crippled 'cos
     they were too cheap to include the serial chip" version. But then
     Acorn has a history of dicking around when it comes to the serial
     port - look at the A3xx series (messed up implementation blighted
     us with special leads forevermore) and the A3000 (no 6551, lame),
     not to mention the RiscPC (combi-I/O chip can support *TWO* ports
     but only one is actually present. Pah. Perhaps it looked too hard
     to bodge the SerialDriver to support more than one port in a nice
     manner? Pah again. And Pah-more for not supporting USB within the
     machine. The fact the RiscPC came out in 1994 and USB was devised
     in 1996 is like totally irrelevant. :-) It's just a clear example
     of not thinking far enough in the future, that it was to be added
     a decade further down the line by way of the cumbersome podule...
     ...not that the main bus is that much quicker, to be honest.

PPS: Wow-OMG. Something like 15 lines all justified naturally with no
      special spacing or anything. My brain must be on a roll tonight.
      It's just a shame I can't have it to doing something useful like
      picking the winning EuroLoto numbers or devising some silly idea
      like chatroulette that I can (hopefully eventually) flog off for
      loads of dosh. Notice the justification continues? Daft brain is
      only doing this to annoy... now it knows I know. Ooooh, deeep!!!

-- 
Rick Murray, eeePC901 & ADSL WiFI'd into it, all ETLAs!
BBC B: DNFS, 2 x 5.25" floppies, EPROM prog, Acorn TTX
E01S FileStore, A3000/A5000/RiscPC/various PCs/blahblah...
 >> TO PRIVATE MAIL ME, REMOVE [BBC-Micro] FROM SUBJECT <<