charles.unix.pro at gmail.com
Tue Nov 3 01:04:35 CST 2015
On Mon, Nov 2, 2015 at 1:28 PM, rod <rodsmallwood52 at btinternet.com> wrote:
> Well I've had a good look at the section marked 'Display'
> It looks like the same row of 37 lamps is repeated on a number of sub
> panels with up to four rows on a panel.
> The lamps look to be pushed through from behind and held every so often
> with a chrome screw from the front.
> The photo of the lower half shows four sub panels
> CO 4 ROWS
I think that should be 'CU'; Control Unit (CPU). First row is instruction
decoder; second row is the instruction buffer; third row is the PC and
instruction sequencer state; fourth row is a bunch of internal state bits.
Fifth row is the computed address (the current state of the incredibly
complex 6180 indirect addressing model).
OU 1 ROW
Operation unit; fixed point and boolean math.
> DU 2 ROWS
Decimal Unit. Does decimal math.
APU 1 ROW
Append Unit. Does virtual address translation.
Towards the top are the "DATA" and "ADDRESS" panels which allow displaying
and toggling in data. In addition the DATA switches are readable from by
the software; Multics requires certain magic settings on the data switches,
probably indicated by the bits of white tape in the picture.
Also, the INITIALIZE button that does just that.
The upper half of the display panel has the "Execute" button which causes a
CPU fault based on the ADDRESS switches; Multics uses this to break out of
wedged operation and reenter Bootload Command Environment (GRUB), where the
operator can dump memory to a disk partition for analysis and reboot.
Below the EXECUTE is the indicator register display; things like
"negative", "carry", "zero" and the like.
To the right of that is the history register display; the last 16 or 64
instructions and their operand values are kept in a circular queue; very
handy for low-level debugging.
Below that is ICT and FUNCTION REGISTER tracks communication between the
I don't see the A and Q register display; they must be on the other panel.
They have a distinctive 'crawl pattern' when Multics is idling. (You can
see it in the youtube clip.)
You may notice that there is no RUN/STOP switch. The 6180 runs all of the
time. When it is first turned on, memory is zero and the PC is zero.
Instruction code 0 is illegal, which causes a fault. It then executes the
illegal operation fault handling word pair, which are zeros, and cause a
fault while processing a fault: a Double Fault. It toddles off and
executes the double fault handling word pair, causing a double fault.
Somewhere on that display panel is the fault code display, which would show
a steady double fault code.
The BOOT button (sorry, not IPL here) is on the Input/Output Manager
cabinet, not the CPU. It reads a tape record into low memory overwriting
the interrupt handler word pair, and interrupts the CPU, which executes the
interrupt word pair which now contains instructions to jump into the code
read in from tape. There are some interesting gyrations to make the first
tape record both a boot loader and a legitimate tape label.
Anyway, I know what most of those lights mean.
Initially, I am just going to build a plexiglass board and use LEDs, laid
out in a similar order to the display; I'm mostly interested in CPU state.
That will give me a better idea of the scope of the project, and allow me
to make more educated decisions about the Mk. II display.
A lot of helpful ideas from cctalk; thank you all very much.
More information about the cctech