The prime number method I am talking about DOES NOT INVOLVE FACTORIZATION -
It uses the "bit bucket" approach, or the SIEVE of ERATOSTHENES method.
Nothing you would see in grammar school - college maybe (?).
This is something the Greek mathematician came up with more than 2000 years ago - -
Also called "pebbles on a beach."
Anyway, you have to be very careful when doing the tests to end a DO LOOP,
since you can get the INTEGER OVERFLOW without knowing about it.
Actually, the formula (quote #13) I gave would be an easy fix for INTEL, since it does not involve any higher order
arithmetic. And since it only has to be computed ONCE for the entire loop, it should not significantly affect any
optimization processes. They still have to compute no. of passes, anyway, right?
The fundamental problem is that (STOP-START) in some cases does not fit within the 32 (or 64) bit word, so you can get the integer overflow before you do the divide.
Now of course, those limits have to be really large numbers for that to occur, but who is to say that a given application would NEVER run into that situation?
Especially if its determined by a formula, or by data input ?