Quote:Original post by Rockoon1
Is an estimate good enough?
For instance the practical difference between 10^1000 frames and 10^1000 + 100,0000,000 frames is moot, right? The extra 100 million is a very minor component in such a case.
Doing estimation, just two 64 bit integers should be more than enough for you. One holds the 'mantissa' of a value and the other holds the 'exponent' (preferably base-2)
Working out the precise details involved should be quite easy. Multiplication involves multiplying mantissas and adding exponents, taking care to compensate for overflows by shifting the mantissa and biasing the exponent.
Yeah, the main reason for which I need that number is for remaining time calculations. A result of 100 years with a +-1 year error margin won't trouble me much (as far as I can give a rough error esteem).
Note that I don't need to keep track of the current frame's number, unless I decide to show a progress bar and a dynamic time countdown, which sounds pretty useless. Maybe I might do the opposite, like telling the algorithm "start from 35%", alias "skip the first n frames", but I can work around this so that I can keep every big number outside the main loop.
Can you give me a better explanation of what you said? Are you telling me to "simulate" a larger IEEE float functioning?