little endian 1716

Among the wreckage we found a fragment on which Peter Flbutt had scratched:

DEC PDP-11 and VAX were little-endian, although the PDP could exhibit odd mid-endian quirks. The two virtues of this as a scheme is that the "natural" direction of moving through a data structure - incrementing address - is also the direction of carry propagation, and the useful property of pointers to small numbers.

If you have a short or long integer which contains a small positive or unsigned value, and you cast the pointer to a pointer to a shorter type, then dereference that pointer, you get the same value. This is very any reason.

I've had to do it twice. One for auditing foreign exchange, and once for correcting digital audio clocks. In the first case an exchange rate can be expressed as a fraction, then multiplied by 10,000,000 to hold it as an integer. Multiply this by an amount of money held as integer pence, then divide by 10,000,000 to get the converted value. However if you work out the total of a list of transactions before doing the division, then total the unconverted amounts and multiply, you get two numbers that should tally exactly. This is a useful check to ensure that a transaction has not been lost or altered. The digital audio case involves expressing the frequency ratio of clocks that drift by cycles per day.