News for hackers. Stuff that matters.
Re:Fast Approximate Logarithms, Part III: The Formulas by jason_s on Saturday June 27, 2015 @11:46AM (#9790485)
in [part I](http://www.ebaytechblog.com/2015/05/01/fast-approximate-loga...) the author argues against using [1,2) because it causes large relative error near x=1 (since log x = 0 there). I would disagree (why would you ever care about relative error in a logarithm -- IMHO absolute error makes much more sense... unless for some crazy reason you're dividing by log x near x=1) but that's his argument.
Re:Fast Approximate Logarithms, Part III: The Formulas by pkhuong on Saturday June 27, 2015 @01:52PM (#9790931)
It's an interesting trade off; you get a nice bump in precision for a small increase in code complexity.