One interesting thing that caught my eye was this:
> Hardware bugs also happened, through component or engineering failure; TX-2 was, after all, an experimental computer. The quickest way to get a bug fixed was to write a very simple program that failed in spite of obvious correctness, and the easiest way to “explain” this bug to the hardware engineers was to show them a Polaroid photograph of the panel of toggle switches that “stored” 24 words of 38 bits each and the console lights. Leaving a program in toggle switch memory that should obviously work, but failed, was a sure way to get the hardware people to fix a bug.
I checked out a memorandum on the toggle switch storage system (http://www.bitsavers.org/pdf/mit/tx-2/6M-5661_Toggle_Switch_...). From what I understand, it allowed direct manual input of data and instructions into memory through these switches. TX-2 had 24 registers, each 37 bits long. Sixteen of these could be manipulated using the toggle switches on the main control panel. So, programmers could directly alter instructions.
The system involved some kind of resistor-switch matrix (not sure exactly what it does though!). But it enabled switches to toggle between 0 (ground position) and 1 (connected to some kind of resistor driver) - this, in turn, would help represent individual bits. Neat stuff!
If anyone has fun stories related to all this, I'd love to hear them!
I’ve been working on a Sketchpad recreation recently, reading through his thesis.
One thing I loved is that the display scanned so slowly that if you did it linearly, or even interlaced, there was a visible strobing pattern across the screen and it gave people a headache. So he randomized the order in which lit pixels were drawn to give it a more distributed, “twinkle” like appearance.
If the computer has visibility of the program counter using LED's (right next to the switches) visually the operator can see what the program is doing.
Well, at least they have an idea. "It's flushing the line printer. See, how the left most PC leds are off, and the pattern being displayed. it is writing data to the printer".
Or, "something is wrong, the PC is stuck in a loop, see the pattern".
Very different way to get a feel for an operating program.
>Instead of one instruction counter, the TX-2 has 32 such counters which are assigned separately to different users of the computer, who then compete for operating time from instruction to instruction. A special part of the machine selects a particular user based partly on a predetermined priority schedule and partly on the current needs of that user. This multiple-sequence operation, in which many essentially independent instruction sequences interrupt and interleave one another, is an extension of the breakpoint operation found in DYSEAC of the National Bureau of Standards.
Is this the first implementation of a process scheduler in a multi-user system?
> A great deal had been learned about the performance of the transistors and memory, the types of logical circuits which are practical, techniques of marginal checking,
and the lesser system problems such as color scheme selection and the proper location of pencil sharpeners.
gold. is this the earliest description of bike-shedding?
One interesting thing that caught my eye was this:
> Hardware bugs also happened, through component or engineering failure; TX-2 was, after all, an experimental computer. The quickest way to get a bug fixed was to write a very simple program that failed in spite of obvious correctness, and the easiest way to “explain” this bug to the hardware engineers was to show them a Polaroid photograph of the panel of toggle switches that “stored” 24 words of 38 bits each and the console lights. Leaving a program in toggle switch memory that should obviously work, but failed, was a sure way to get the hardware people to fix a bug.
I checked out a memorandum on the toggle switch storage system (http://www.bitsavers.org/pdf/mit/tx-2/6M-5661_Toggle_Switch_...). From what I understand, it allowed direct manual input of data and instructions into memory through these switches. TX-2 had 24 registers, each 37 bits long. Sixteen of these could be manipulated using the toggle switches on the main control panel. So, programmers could directly alter instructions.
The system involved some kind of resistor-switch matrix (not sure exactly what it does though!). But it enabled switches to toggle between 0 (ground position) and 1 (connected to some kind of resistor driver) - this, in turn, would help represent individual bits. Neat stuff!
If anyone has fun stories related to all this, I'd love to hear them!