Date   : Sun, 14 Feb 2010 11:18:24 -0600
From   : jules.richardson99@... (Jules Richardson)
Subject: Subject: Re: 'BASIC 512K'

jgh@... wrote:
> Rick Murray wrote:
>> I wonder if the co-pro was a side effect of Acorn looking for a
>> worthwhile RISC chip for their new range of processors?
> 
> It was. Acorn were looking for something to step from 8-bit 6502s
> to a suitable 32-bit CPU. They approached Intel to see if they
> could license the 80x86 core, but Intel were only interested in
> selling physical chips.

What puzzles me is why they didn't go m68k - everyone else was around that 
time. Did they go out of their way to try and be different, or was there some 
other reason for rejecting it? (I'd have to check to see if Torch's Neptune 
copro was around by then - maybe it was, and so they considered the m68k 
market to be already cornered)

> The 32016 gave Acorn the experience that showed that short, fast
> instructions and fast interupt response was what was important.
> Some 32016 instructions are 12 bytes long, BBC BASIC on an 8MHz
> 32016 barely manages to perform as well as BBC BASIC on a 2MHz
> 6502.

Yes. Some of the info I have on Xenix port to the 32016 suggests that they 
were having *major* problems with interrupt response (and in part due to the 
slow transfers across the TUBE) - reading between the lines it was clear by 
then that a BBC micro / copro setup just wouldn't cut it any more, no matter 
how good the copro was.

Funny thing is, I did spot some later documentation which suggests that they 
(Acorn) were thinking about ditching ARM when they were thinking more 
seriously about getting into the UNIX market, even though ARM had proved 
itself a capable-enough CPU by that point. (I also unearthed some stuff which 
hints at a Svr4-based UNIX release on Acorn hardware, circa 1993 or so - that 
would have been into A4000/A5000 territory, I think)

> The beauty of the Tube system was that it takes a very small bit of
> hardware and firmware to throw together a system that lets you
> experiment with a new CPU without having to build an entire system
> - the system is already there in the host processor. Typically, you
> only need about 1K of client code, a Tube ULA and a couple of glue
> chips.

It was certainly nice that Acorn had the foresight to make it a built-in part 
of the firmware. I don't think the hardware side of it was particularly 
unique, but the software bit was (to a great extent).

cheers

Jules