[Info-vax] RB730 Integrated Disk Controller (R80/RL02) usable with VAX-11/750?

[Roger Ivie]

On 2011-05-03, Bob Koehler <koehler at eisner.nospam.encompasserve.org> wrote:
> In article <ipo238$id3$1 at Iltempo.Update.UU.SE>, Johnny Billquist <bqt at softjar.se> writes:
>> 
>> What??? Are you sure? The Unibus only have 18 address lines, which means 
>> 256K addressable. And some of that is needed for I/O. I would be very 
>> surprised if memory really sat on the Unibus.
>
>    IIRC, the 730 and 725 were true UNIBUS systems, using the UNIBUS
>    as the system bus, with all its limitations, just like my PDP-11/34 
>    and 11/44.

It's been a *long* time since I've used a /730, but I'm certain it had a
UNIBUS map. I think the one I used to use had a couple megs of RAM.

AFAIK, the only VAX without a UNIBUS map (or its equivalent) was
MicroVAX I, which used QBUS memory and was therefore limited to 4MB.

My original design idea for the Firefox QBUS adapter involved having 4MB
of RAM on the adapter so that it didn't need to get on the MBUS for DMA;
the cache structure on Firefox was such that it was difficult to
guarantee that a DMA transaction could be completed within the DMA
timeout. I made the mistake of trying to sell it to VMS as "conceptually
pretty much just like MicroVAX I"; I pulled back a bloody stump because
the VMS guys hated MicroVAX I due to the lack of a scatter/gather map
and had just barely managed to clean all the MicroVAX I support out of
the code base.

Consequently, FQAM wound up using an ugly hack that was put in place to
try to make FTAM work (Firefox Tape Adapter Module, the first attempt at a
QBUS adapter for Firefox, which consisted of a CQBIC) to tame the
MBUS: while ownership of the QBUS was granted, the FQAM used the
absolute priority given to node 0 to keep the MBUS 100% busy doing short
transactions to guarantee that another node couldn't start a transaction
that would take so long to complete that it would prevent an FQAM
transaction from completing within the QBUS deadline.

Which is why Firefox can't get anything done while QBUS DMA is going on.
Ugly, but it works, for certain definitions of "work".
-- 
roger ivie
rivie at ridgenet.net