A couple of observations.
Taking the PDP-11 as a fairly typical example, the switches are "data" and
"address". While running, the data switches were visible to the software, and
could do something if you wanted to (typically this wasn't done). When
stopped, you could set an address in the address switches, and examine memory
at that address (showing it in the lights) or change it (with the data supplied
by the data switches). You could also start at the address in the address
switches.
Also, while running, the lights would display some bit of system state. On a
fair number of PDP-11s, you could select what to display. When running RT-11
you would typically ask for the "display register", a register written by
software that appears in the lights and contains the idle pattern. The pattern
would slow down as the system got busy.
On other operating systems, say RSTS/E, you'd ask for "data paths" which was
some internal state except during a WAIT instruction, when it would show the
contents of R0. That was where RSTS would put its idle lights pattern. The
result is that you'd see the clear pattern on a nearly idle machine, a blur if
the machine is busy, or a frozen pattern if the machine was stuck in an
infinite loop. With experience, you could gauge what sort of thing a system
was doing by its lights. This is why in RSTS/E development we had Field
Service remove the "remote diagnostics" (no lights) consoles and put the lights
ones back in.
Very early PDP-11 systems needed to have some bootstrap toggled in, but by the
time I saw my first one (1973) diode based boot cards (16 words, just enough
for early disk drives) had appeared.
Other machines might be different. Some had vast quantities of lights to show
a lot of internal state at the same time; I remember Burroughs mainframes and
some IBM/360 models (360/65?) in particular. A few had none at all: the CDC
6000 mainframes had a CRT display under program control, but no lights
whatsoever. It did have a boot loader, in the form of a 12 by 12 array of
toggle switches -- a 12 word changeable ROM. :-)
The first machine I got my hands on, a Philips PR8000 -- 24 bit minicomputer --
had 8 digit light displays -- 8 octal digits for the 24 bits of data. And the
data entry was with 8 sets of buttons, one set for each octal digit. The
buttons were marked 4/2/1 and arranged so you could enter a digit with one
finger press:
| 2 |
| 4 | 1 |
so pressing in the middle would get you 7, bottom center a 5, etc.
The "boot ROM" was mentioned, which applies to various DEC machines. On some
other machines, the boot operation might not be a small program but a
specialized hardware action instead. That applies to the CDC mainframes, for
example, as well as machines like the IBM 1620. There a boot operation
("deadstart", "initial program load") might put the CPU and I/O devices in some
specific state and then start a data transfer from some selected device to some
fixed memory location.
paul
