ILLIAC I Programming Manual (1956) [pdf] | Hacker News Reader

ILLIAC I Programming Manual (1956) [pdf]

bitsavers.trailing-edge.com

33 pointsnimz21 comments

I wish I could find the part written by J.E. Robertson. I took several graduate level courses from him as an undergraduate in the mid '60s. He became the world expert on division. He found that if you recoded to the redundant base -1, 0, 1 then number encodings could be found that had lots of zeros which he could optimize over and speed the division up. There were other sequential configurations he did special things with as well. You wouldn't believe the amount of theory he deduced from studying the redundant encoding problem and its application. The result was SRT division. He was the R and I can't remember who S and T were.

SRT division was implemented in the Illiac II.

SRT is named after Sweeney [1], Robertson [2], Tocher [3], which came up with the idea independently of each other. There is also a 1956 paper by Nadler said to have a similar algorithm, but it seems impossible to find.

[1] http://www.computerhistory.org/collections/catalog/102632302

[2] https://archive.org/details/newclassofdigita82robe

[3] http://qjmam.oxfordjournals.org/content/11/3/364.full.pdf

Thanks. [3] is essentially the outline of the first course I took from him.

It didn't occur me that SRT division might have any recent relevance, ancient history and all that, but Google tells me that the Pentium floating point bug was in fact a buggy lookup table for SRT division and that it is still apparently in use in microprocessors today. I wish he could know that.

Bitsavers.org Link (where there is even more cool stuff):

http://bitsavers.trailing-edge.com/pdf/univOfIllinoisUrbana/...

Thanks! We changed the url to that from https://dl.dropboxusercontent.com/u/12852544/serve/ILLIAC_pr....

Bitsaver is great.

Thanks to it I have collected almost all manuals from Xerox PARC systems.

The context:

http://en.wikipedia.org/wiki/ILLIAC

"ILLIAC I was built at the University of Illinois based on the same design as the ORDVAC. It was the first von Neumann architecture computer built and owned by an American university. It was put into service on September 22, 1952.

ILLIAC I was built with 2,800 vacuum tubes and weighed about 5 tons. By 1956 it had gained more computing power than all computers in Bell Labs combined. Data was represented in 40-bit words, of which 1024 could be stored in the main memory, and 12800 on drum memory."

The interesting information from the manual is that the computer had 40-bits fixied-point arithmetic, directly calculating only numbers between -1 and +1. If was certainly very demanding programming it.

Another interesting detail: the puritans still didn't influence the name of the number system, they write about "sexadecimal" coding, and interestingly use "0 1 2 3 4 5 6 7 8 9 K S N J F L." Why "K S N J F L"? I don't know.

And a really fascinating detail: it was not only punched tapes and stuff: it had a 256x256 pixels CRT as an output device too. Sample on pg. 12-9

I just had to search for it.

From: http://en.wikipedia.org/wiki/ORDVAC

"Unlike the other computers of its era, the ORDVAC and ILLIAC I were twins and could exchange programs with each other."

"Instead of the sequence A B C D E F universally used today, the digits ten to fifteen were represented by the letters K S N J F L (King Sized Numbers Just for Laughs), corresponding to the teleprinter characters on five-track paper tape."

The info you quote is a description and not the answer. The whole alphabet is on the page 9-5. I still don't know why there isn't some "order" in encoding (like it is now in ASCII, that would still result in the encoding for 10 to be A and not K. and it's not a USTTY http://en.wikipedia.org/wiki/Baudot_code there is K == 01111) but I'd expect they must have had some reason.

It looks like they took the ITA2 telegraphy code, which has the numbers in normal order but the letters in typewriter order, and re-sorted it so that the letters are alphabetical. That would explain the scrambled numbers, but not the punctuation or "K S N J F L".

The letter codes of ORDVAC and ILLIAC are clearly not ITA2 / USTTY. So we still don't know why they have been selected.

Reading this manual really brought me back to the time when being a programmer was probably closely tied to being a mechanic of the machine. We take computing for granted today but it must have been amazing to see first-hand how physical inanimate objects come together to build intelligence.

It still is, if you work with embedded systems and other lower-level stuff.

but it must have been amazing to see first-hand how physical inanimate objects come together to build intelligence.

Many CS curricula still start with building a CPU out of basic logic gates (and for those who haven't done this before, this is certainly a recommended exercise - make a simple CPU in a logic simulator that can run a short program). The first time you see how it all works seems like magic, then you remember that everything just fits together logically, and realise that it isn't magic after all.

Indeed, but when the whole room is one big logic machine that you can walk around in, it must feel somewhat different than logic simulation software.

A resource for going from NAND to a running program (Tetris in this case): http://nand2tetris.org/

A bit less flashy/more old-fashioned, Niklaus Wirth's 2013 edition of Project Oberon covers an FPGA-based CPU, operating system, compiler, tiled window manager and text and vector graphics editors. No Tetris though. http://projectoberon.com/

Mirror on archive.org: https://archive.org/details/bitsavers_univOfIlliIACILLIACpro... (as the Dropbox link is down)

Dropbox link is now up.

Dropbox is returning a 509 status code on the file, but I can't wait to take a look at this. My interest in this specific machine comes from its important place in the history of computer music with Lejaren Hiller's Illiac Suite:

https://www.youtube.com/watch?v=n0njBFLQSk8 http://ems.music.uiuc.edu/history/illiac.html https://en.wikipedia.org/wiki/Illiac_Suite

The Computer History Museum has a number of physical artifacts from the ILLIAC II, III and IV, including a number of the logic modules: metal chassis units that held from 3 to 8 vacuum tubes that performed a single function.

http://www.computerhistory.org/collections/search/?s=illiac&...

Interesting. 7.3 seems to be a very early instance of the xor swap trick, albeit with addition and subtraction as group operations.

Reply on news.ycombinator.com