Parts of this answer are necessarily speculation, especially since the way the machine worked (and its bizarre programming model) are buried deep in time, or at least deep in various museums(1) :-)
First off, I have to say this behemoth was truly a thing of beauty. The way it was programmed was with a paper tape with each instruction holding parts (practically RISC/VLIW in the general approach):
- which devices would output to the "bus";
- which devices would input from the "bus"; and
- an operation which would be executed or started.
All uses of the calculator boiled down to a sequence of simple operations like adding, multiplying, interpolation, and function calculations. These were used in conjunction with "constant" registers that would use these operations for specific tasks.
The whole thing was driven by a (very mechanical) sequencing unit, using motors and gears to advance the paper tape and use the current tape "instruction" to control every other part of the machine. Hence it was, for all intents and purposes, the clock signal in modern computers:
It appears it was even multi-core since you could dedicate the entire calculator to one task or use the duplication on right and left to do two calculations concurrently.
The following layout shows the various parts of the system.
On the other side of the main "winged" control panel stood the six rather large relay cubicles, which each contained various panels, including panel
F in the third relay cubicle, where the moth in question met its glorious and famous end. A non-layout view of the left three relay cubicles (and the left wing) is shown here:
F in the layout did not have a more mathematically obvious labels (such as
INTERPOLATOR). Instead, panels
G were both labelled
SEQUENCE, and panel
TRANSFERS AND JUMPS.
So they were all probably responsible for program or data flow, such as controlling flow of information between various other panels, or somehow affecting the "clock generator" shown above.
The image of a
JUMP instruction causing a motor/gear combo to back the paper tape up a few feet, rather then just changing the instruction pointer register, seems particularly funny to me (I have no idea how loops were actually done, that thought just popped into my head as one possibility).
(1) Note that all images and most research in this answer has come from either:
- the Grace Hopper online exhibitions at the Smithsonian; or
- the Computer History Museum in California.
I'm not affiliated with them in any way but I urge you to visit (or otherwise support) them if you're interested in retro-computing, especially if you're interested in times before the unholy trinity. I visited the CHM about a decade ago and still plan to visit the Smithsonian at some point in the future, assuming the chances of a treasonous insurrection during my visit can be minimised :-)