Hacker News Reader

Comment by jonjacky | Hacker News Reader

This machine made ingenious, economical use of the technology available in the mid 1950s. Only 16 instructions. The ALU used all diode logic. The CPU only used a few tubes for flipflops to store control state.

The rotating magnetic drum provided all memory -- including CPU registers. A tiny oscilloscope on the front panel showed register contents as they rolled by the read heads!

View on HN

How do you implement a whole ALU in diode logic? All i know about diode logic is what it says on the wikipedia page, but according to that, you can't implement a NOT gate, which i would have thought would make it very hard to implement anything of significance - even a half-adder requires a NOT!

Allegedly, there were just 15 flip-flops, which are reused a lot…

Probably being bit-serial and storing the CPU registers on the drum helped a lot with this. Have you looked at the magnetostrictive-delay-line calculators from the mid-1960s, which used pulses of torsion in a coil of straight-annealed wire instead of a mechanical drum?

In fact, I just have "Electronic Computers" by S.H. Hollingdale and G.C. Toothill (Pelican / Penguin Books, 1965) besides me. While meant to be popular or introductory to the EE side of computing, this is a wonderful book full of such things! Highly recommended and cheap to pick up on various market places!

Thank you for the recommendation! It doesn't seem to be on Libgen or in the Archive's texts collection, and I doubt that I can find a paper copy here in Argentina. Maybe you could scan and upload it; it is presumably in the public domain if it was originally published in the US, since the authors probably did not bother to renew the copyright.

I picked mine up at abebooks.co.uk – it seems to have been quite popular in the UK at the time.

Regarding copyright: It was published in the UK. So for the moment EU copyright regulations, 70 years from the death of the author (I guess).

Here's I found a source for this: Stanley P. Frankel on the design of MINAC and the LGP-30 in IRE Transactions on Electronic Computers. (15 flip-flops, presenting both normal and inverted output to a logical network made of resistors and crystal diodes, providing basic monotonic Boolean functions like And, Or, and Xor.)

Stanley P. Frankel, “The Logical Design of a Simple General Purpose Computer” in: IRE Transactions on Electronic Computers, March 1957, pp. 5 [1]

[1] https://www.masswerk.at/nowgobang/misc/MINAC-IRE-March-1957....

This is awesome! Thank you! This should make the Boolean equations in the manual quite a bit easier to decipher.

XOR isn't monotonic — you can't make X ⊕ Y with diode logic given just X and Y, the way you can make X ∧ Y and X ∨ Y. However, you can make it given X, Y, X̄, and Ȳ. I'm not sure if you can make it with X, Y, and just one of X̄ or Ȳ.

Regarding XOR: This is sloppy quoting of Frankel, who writes about the monotonic functions "And and (nonexclusive) Or" – so, yes, inclusive Or only. However, as described, both the normal and the inverted (negated) outputs of the flip-flops ("toggles") are fed into the network, so there's X and Y and negated X and negated Y available.

Right, and as I explained at https://news.ycombinator.com/item?id=20488446, if you have negated versions of all the literals available, you can make not just XOR but any arbitrary Boolean function as a sum of products (or a product of sums, CNF), just like a PLD does.

Notably, you need XOR (single-bit add) for an ALU anyway… (So the somewhat weak exclusion of "exclusive" in Frankel's paper is a bit luxurious.)

"Luxurious"? I'm not sure what you intend to say. Do you perhaps mean "unnecessary" or "essential" or "very strong" or "very weak" or "well-thought-out"?

As I read it, he's pointing out that since the diodes and resistors can't do things like XOR, because it's not monotonic, it's necessary for "[e]ach input variable" to be "presented in duplicate".

Another example that is in some sense more fundamental than XOR is the MUX function b if a else c. If at some point b = 0 and c = 1, the output is the negation of A, which diodes and resistors alone cannot achieve.

"Luxurious" is a term from psychology, where it's used to describe "ornamental" behavior, which may be hinting at something deeper. ;-)

In the particular case, I just found it interesting, since the duplicate outputs are already well established, before the text "withdraws" to monotonic functions (for which both inputs are used). A bit of an understatement.

Edit: I think, "luxurious" as used this way is a great term, because it doesn't attempt to apply any classification. It's just pointing out that there's a somewhat independent process, which consumes energy.

Here's a photo (by Cory Heisterkamp) showing the logic networks from the Ed Thelen's website [1]: http://ed-thelen.org/comp-hist/images/Heistercamp-lgp.jpg

(The cards with the rest of the logic are behind and below the main network plane. Mind that this a Control Data model, so it's 1965 or later.)

[1] http://ed-thelen.org/comp-hist/lgp-30.html

At the RTL or Moore-machine level, a computer consists of some register state (the flip-flops) which changes, at each clock transition, to a deterministic function of its previous state and its input lines at the time. So you need a way to compute the new state as some arbitrary Boolean function of the previous state.

The flip-flops each have a Q output and an inverted Q̄ output, which comes for free due to the symmetric construction of the flip-flop from a pair of vacuum tubes and other components. You can form any arbitrary Boolean function of the flip-flop state from a sum of products of the Q and Q̄ variables; it's a simple matter of applying De Morgan's theorem and distributivity. (Or, for that matter, you can use a product of sums by doing it in negative logic.) The flip-flops can supply not just all the memory but also all the inversion, signal level restoration, glitch elimination, and amplification, none of which can be done with diode logic.

So, indeed, diode combinational logic is not capable of computing arbitrary Boolean functions of its input lines. But it is indeed capable of computing arbitrary Boolean functions of the flip-flops' state. And that is all that is needed.

The LGP-30 manual gives a complete description of how the computation works at the Boolean equation level, a description which I have not yet managed to grok.

If you're interested in this kind of thing, you might be interested in notes/non-inverting-logic.html in Dercuano: http://canonical.org/~kragen/dercuano-20190711.tar.gz. (Like most things in Dercuano, it's unfinished.)

Reply on news.ycombinator.com