ENIAC ....

I admit that your ABC page is over my head, technical-wise. So, I asked Bill M. for his opinion. Here is what he said (with permission for me to quote it here): ---------------------------------- The [Atanasoff] machine had several operations, which were initiated by pressing the appropriate button. Two of them were decimal input and output, and just one that did arithmetic. Calling the arithmetic operation "programmable" and assigning it an adjective like "data-sensitive" is a stretch; if you use those terms so loosely, you will have to say a mechanical adding machine also has those properties. There were 30 separate adders and 30 pairs of coefs. The user had to select which pair was to be eliminated. This selection happened by the human moving a jumper cable from one set of holes to another. Selecting one of the 30 for final output happened in a similar way, I believe. The I/O for cards happened in groups of 5 and had real switches. But except for selecting an input, there are no "programmable options" on the arithmetic. There was a switch you could set, so that it would either start with a subtract or start with an add. (Once it started it would flip back and forth.) The user has to know which way to set this by examining the signs of the coefs. I wouldn't call this a programmable option as much as a hazard. If you set it the wrong way, the machine runs forever and never finishes. (Well, not forever, it will definitely finish in 2 to 50th seconds. 35M years.) As to the other part, data sensitivity, this is somewhat true. The machine is doing a "long" division by repeated adding and subtracting. In the end, the result is just a simple math operation, but the length of time it takes is dependent on the actual numbers you plug in. It is literally just testing if the sign bit flipped, and throwing a relay each time it does. I guess you could say every multiplier is data dependent.