XXDP on PDP-11/03
jnc at mercury.lcs.mit.edu
Wed Aug 14 17:24:01 CDT 2019
> From: Allison Parent
> ! Seriously? ... Memory of some form there is a must.
I don't know about you, but my approach in looking into hardware issues is
often to start by reducing things to the simplest possible configuration that
exhibits the failure.
(I asssume the various reasons for that approach are obvious.)
So, the OP couldn't get ODT to work. Well, what's the simplest config one
needs for ODT? Well, a CPU (but it won't be executing any instructions, so one
could leave HALT on), the console serial card (with a working terminal
attached), a bus/backplane to plug them into, and a power supply.
But no, the LSI-11 machines also want memory - although it's unused by ODT
after a single read cycle at power-on.
It's probably worth pointing out that this is _not_ true of the F-11 machines;
those do ODT just fine without memory. Perhaps DEC got some complaints about
the behaviour of the LSI-11, and made a change?
I don't know if the front console on the early UNIBUS machines works without
any memory on the UNIBUS - I'm too lazy to check. I have this vague memory that
they do, though.
> The architecture of pdp-11 has the first 256 words as interrupt vectors
> and software locations.
Some 'internal' interrupts from the CPU (e.g. NXM) are at fixed, low,
locations (in Kernel D space on some of the models with MMU, to be technical -
I don't know about the /40 and /34, etc), but there's nothing that restricts
_device_ interrupts to be in low memory (either physical, or virtual on those
machines which get vectors from Kernel virtual).
E.g. in the "pdp-11 bus handbook" (EB 17525 20), pg. 119, it says "Place
Vector on BDAL <15:00> L" - so one could use 0140000 if one wanted.
Most DEC devices that do the vector with jumpers don't have posts for all 15
bits, it is true, but AFAIK no CPU looks at only the low bits on the bus.
> How else would the console vectors at 60 work.
ODT doesn't use interrupts.
More information about the cctalk