Date : Fri, 19 Aug 2011 08:20:47 +1000
From : awilliams@... (Alan Williams)
Subject: Tube ROM
The point about RAM being faster than ROM is certainly an important one.
I am not sure it this would have applied to a 3MHz 6502 but the 80186 it
is almost certain to have done. Filestores copy their entire ROM to RAM
at boot before staring the OS and FS code up, so it clearly did by then.
Regarding the use of flash roms in Beebs - its not as straight forward
as you might like. Well not as straight forward as sideways RAM is,
where you bend a leg out from a RAM chip and wire that to NWDS. A
29F010 for example is not at all happy with this idea. Not withstanding
the fact that its too big for the socket it needs you to keep OE high
during writes. There are also intriguing issues with using a 6502 CPU
to generate the correct sequence of write cycles required by a flash
rom. Its seems (from some notes I have read on a apple II card using
the same chip) that the 6502 tends to do extra reads during instruction
decoding in some addressing modes, which will foul up the flash chips
write programming sequence.
I hope to have this sorted in the next short while and will publish my
findings accordingly.
29F010 should give 8 banks of sideways ROM each individually erasable.
Presently I can read the device id.
Alan
-----Original Message-----
From: bbc-micro-bounces+awilliams=linkme.com.au@...
[mailto:bbc-micro-bounces+awilliams=linkme.com.au@...] On
Behalf Of Rick Murray
Sent: Friday, 19 August 2011 2:50 AM
To: BBC MailList
Subject: Re: [BBC-Micro] Tube ROM
On 18/08/2011 18:10, Rob wrote:
> And acorn did have the foresight to allow up to 16 ROMs to be used,
I have a FlashROM that I'm going to attempt to wire in as the fifth ROM.
What with MOS, BASIC, DNFS, and GoMMC, there's not a lot of space left.
So the fifth ROM will be the Flash with, hopefully, eight images
available. The only complication is, IIRC, the Flash must be erased at
one time, however if I ensure a useful filing system (most likely
Econet) is always available, I can keep the tools and ROM images in
place to (re)program it in-situ.
> It would have complicated the co-pro startup sequence, though. If you
> have a language ROM in there too, then in what circumstances do you
> want to invoke it?
Very good point.
Another question - the service manual implies that the co-pro startup
clocks absurdly slowly (something in the order of 350kHz, IIRC?). Is
this because the ROM is way slow and it is copying the code into the
faster RAM, prior to switching to full speed for RAM-only execution?
> It would be interesting to hang a 6502 second processor off a ZX
> spectrum, say.
;-) I think the problems are twofold. Firstly, the Tube chip is a
dedicated specific purpose chip. That's not to say they could not be
obtained or cloned (a la Kortink), but it may have dissuaded others to
try. Secondly, and to be rather blunt, I think most other operating
systems of the era were way too braindead to function as Tube hosts
without a lot of reworking. Acorn took pains to devise a solid API, and
one that would work on host and co-pro alike. There's a big expansion
slot on the back of my Oric (obstensibly for a floppy drive) that might
be capable of hosting a Tube interface. I think, however, the host
mechanism would only work if the entire OS was rewritten.
We'd have more luck bunging a Tube chip onto something like a PC or a
RISC OS machine, something with enough processing grunt to fake the API
expected. But, then, why bother? BeebEm not only fakes the entire Beeb,
but the co-pro as well. ;-)
Having said that... a Spectrum with the Sprow's ARM co-pro. That would
be a mind-screw of epic proportions.
[ http://tvtropes.org/pmwiki/pmwiki.php/Main/MindScrew ]
Best wishes,
Rick.
--
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...
_______________________________________________
bbc-micro mailing list
bbc-micro@...
http://lists.cloud9.co.uk/mailman/listinfo/bbc-micro