(32S, 32SII) Birthday paradox, hash collision probability

[Dave Britten]

(04-27-2020 07:54 PM)Albert Chan Wrote:  

(04-26-2020 07:58 PM)Dave Britten Wrote:  Very clever! And also usable on calculators that don't have a built-in ln(1+x) function:

ln(1+x) = ln(1+x)*x/(1+x-1)

Thanks. Where do you get it ?

Note: watch out for divide-by-zero, if x is tiny.
Or, use Dieter's log1p and expm1 approximation.

Can't remember where it came from exactly, but it was related to TVM calculations. 15C Advanced Functions handbook maybe? And yes, if x+1-1 returns 0, then you just return x, since ln(1+x) approaches x as x gets smaller.