🚀 go-pugleaf

In article <1991Jun25.151303.22228@cpu.com>, gwlester@cpu.com (Gerald Lester) writes:
> In article <1991Jun24.093746.48196@cc.usu.edu> slsw2@cc.usu.edu writes:
>>
>>The main problem with UNIBUS as far as modern systems go is that it has a
>>limited address range; 256KB as opposed to 4MB for the QBus.
>>
> Funny, I seem to remember having a PDP-11/44 (a UNIBUS machine) with 4MB of
> memory.  I think the problem is not with the UNIBUS, since it is also used
> for some of the VAX line, but with some of the older 11s.

The limited (physical) address range relates to what you can see over 
the bus.  So if (as on the 11/45, which I had experience of) the memory 
is connected to the CPU via the Unibus, then it's limited to 256KB.

However, as first the 11/70, then the VAX 11/780, and all subsequent 
Unibus VAXen and large-memory PDPs, have demonstrated, there's no reason 
you shouldn't interpose mapping registers between peripherals (talking
through the Unibus) and the memory (addressed through something 
`private').  It does impose restrictions on what you can do (limits the 
maximum physical size of transfers, for example) but on the PDP this 
isn't an issue, since the maximum segment size is 64KB, and under VMS 
(who cares about U*x?) software limits the maximum size of transfer.

In fact, the more complicated design of the Unibus is not *just* due to 
its age, but to the separation of address and data lines.  This 
separation means that Unibuses (Unibi?) intrinsically have higher 
throughput than Qbuses (Qbi).

So the only people who suffer grief are the systems programmers.

And who ever cared about us?  As my predecessor in that role here wrote in 
a comment - E Unibus plurum*
-- 
Robin Fairbairns, Senior Consultant, postmaster and general dogsbody
Laser-Scan Ltd., Science Park, Milton Rd., Cambridge CB4 4FY, UK
Email: robin@lsl.co.uk  --or--  rf@cl.cam.ac.uk

* From the Latin - plurum (n), a tear.  Boo hoo indeed.