The Power of the NORC 3763

On Sun, 18 Jun 2006 23:10:26 GMT, Brian Inglis

Yes; I've seen information about that. I could be wrong, but I thought there was a standard now for floating-point in bases other than binary.

Despite everything, though, I feel few manufacturers will see the point in the extra circuit complexity just to avoid conversions - even though techniques like carry-save adders, Wallace tree multiplication, and, of course, Goldschmidt division, can all be applied to decimal arithmetic... and Chen-Ho encoding can also be used to keep any waste of storage space to a minimum.

One problem *I* see is that if you use one bit for the sign, you can't chop up a decimal number neatly into two halves that are decimal numbers with signs without waste. Not even if you use IBM's Densely Packed Decimal, that lets you use the last four bits of the 10-bit code to stand for one digit, and the last seven bits to stand for two digits!

I've actually come up with a *solution* to that problem. I said to myself, whatever IBM did for base-10 can also be done, by taking one bit away from each digit, for base-5. So, make the code *symmetrical*, and you can chop off the leading three or six bits (but not one bit) from a decimal number *in nine's complement form*.

And it will work just as well if the underlying decimal number is in ten's complement form. The property is that not only the last seven bits of the code for 011, say, code for 11, but also the last seven bits of the code for 988 code for 88.

But even this isn't going to make decimal popular.

John Savard Usenet Zone Free Binaries Usenet Server More than 140,000 groups Unlimited download