The Holy Grail for the BINAC is its operating manual ; I have asked about this in the few places it exists, but there doesn't appear to be a PDF copy. This was written by a fellow called Joseph Chapline, and it appears to be the first ever computer manual.
There are very occasionally some on sale, but the appear to go for $25k a time, which is a bit much really.
The most common document is a brochure http://www.studio2.org.uk/binac/brochure.pdf
This is an 9 page document with some useful stuff - it lists the whole instruction set for example - but it misses out a lot of other stuff. A BINAC has 512 words of memory, 30 bits each. Each instruction occupies 5 bits for the operation code - they were called 'orders' at this time, and 9 for the operand (an address, hence 512 words). However, each word actually contains 2 instructions - 512 instructions isn't a lot at the best of times, so by pairing instructions up, you can double the program size.
This is why the U instruction 20xxx (in octal) says 'obtain next pair of instructions from m and continue from that point' (it's an unconditional jump) and why there is an apparently pointless Skip instruction (it's for when you want an odd numbered branch)
However, digging around finds something useful. http://www.studio2.org.uk/binac/arithmetic.pdf is a mathematical paper on binary arithmetic. However, handily, the machine it uses to illustrate its sums is a BINAC. So I now know how the maths works.
This is one unusual thing about this machine. Reading another description of it, I noticed it used 2's complement arithmetic.
This was somewhat surprising. Most computers of this Era use a variety of things, but not 2's complement. One's complement is common (e.g. CDC160) which of course gives you two zeros which can cause confusion. Some machines operate in decimal units (the Harwell machine a fellow retrochallenger is doing is like this). But this is, usefully, 1's complement.
In thinking what to do with this machine, once emulated, there are two options ; the old stuff and the new stuff. There are three 'old things' which I only know a bit about.
- Roger Mills' NIM program. Possibly the first 'game' on a real computer. Might precede draughts on Turing's Pilot ACE. I think Mr Mills is still with us (one of the problem with machines of this vintage is many of the original staff are deceased)
- C-10, a sort of machine code shorthand
- Short Code, a second version which was extended to the UNIVAC.
I have a bit of a link to this era. I was a graduate of the University of Essex ; my personal tutor was a fellow called Tony Brooker. Close to retirement when I knew him, he was one of the people involved with these very early machines.