[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Off Topic - BBC Micro Books



On Nov 23, 4:45 am, Mark McDougall <msmcd...@no.spam.iinet> wrote:
> As an aside, I understand that there was an ARM co-processor (via the
> so-called "Tube") available for the BBC at some point?!?

If there's any errors, someone please correct me, I'm just going based
on what I've read.

Acorn's strategy for testing out new processors for the BBC's
successor was to put them in a second processor unit. That way, they
could work with their existing environment, but new hardware. Their OS
was designed very well to abstract hardware away from software, so
they could get away with this with only porting to the new CPU
architecture, and the host 6502 handled all I/O.

They also evaluated the NatSemi 32016 and the Intel 80186 in that
manner. The 32016 (and 68000, for that matter,) had some very slow
uninterruptible instructions meaning that it, at 6 MHz, couldn't do
double-density disk writing in software... when their 2 MHz 6502
could. x86 lost because Intel wouldn't license the 80286 to them,
they'd only sell Acorn chips.

I don't recall ever reading why they didn't ultimately go with the
65816 (although they did use it in one machine,) but they visited WDC,
and saw what sort of resources they were using... so Acorn decided to
just design a 32-bit RISC chip, and the rest is history. First code
ran in a behavioral model implemented in 808 lines of BBC BASIC
running on the 6502, then they got a chip under 2 years later, almost
everything worked first try, and ARM Evaluation System was made using
it. Then, they made the rest of the ARM chipset (a custom MMU that
could address 4 MiB RAM (and 4 MEMC1as could be in a system,) a custom
I/O controller, and a custom video/audio controller capable of
1152x864 monochrome, or as much as 640x480 8-bit color and eight
channels of digital audio at 41.667 kHz (or one channel at 333.333
kHz, or anything in between - completely impractical, but it can do
it,)) and built the first prototype machine, the Archimedes 500,
around it. 1987, it came out in the Archimedes 305, 310, 410 (which
may not have actually been sold,) and 440.

It's probably not healthy for me to know this much about machines that
I've only gotten interested in this year, but hey... (And, to be fair,
I did have to look up the VIDC1a's maximum audio sample frequency.)