Software or Firmware

[Black Hole]

Hmm. Being so steeped in two's-compliment, I wasn't even aware of one's-compliment as a "thing". I can see how it works:

* The leftmost bit is a "sign" flag: 0 = positive, 1 = negative;
* Negating a number == invert all the bits;
* To add numbers: binary sum, and if one of the numbers was negative increment the result by 1.

I can't quickly think out under what circumstances a carry flag gets set.

It seems messy compared with two's-compliment, which I guess is why every microprocessor I know of uses two's-compliment. Another guess is that Unisys might have been forced into it to avoid infringing patents, or maybe they had a "not invented here" attitude.

 

[EEPhil]

I can't quickly think out under what circumstances a carry flag gets set.

Example from http://www.learnabout-electronics.org/Digital/dig15.php

Decimal          Binary
+6              00000110
-4  +           11111011 +
=========================
   (carry)  (1) 00000001

...and it gives the wrong answer of 1 (which you've already pointed out).

 

[gomezz]

It seems messy compared with two's-compliment, which I guess is why every microprocessor I know of uses two's-compliment. Another guess is that Unisys might have been forced into it to avoid infringing patents, or maybe they had a "not invented here" attitude.

They also used their own six bit character codeset instead of ASCII and 36 bit memory words which can be split into halves, quarters or sixths and memory dumps used three bit numbers which is why we used to refer to our electronic arithmetic aids as "octalators" when attacking memory dumps with highlighters and post-its trying to work out if we were looking at memory addresses, code, numbers, data tables, ASCII character strings or Fieldata character strings.