The linked overview on Efficiently Generating a Number in a Range by O'Neill (of PCG fame) is insightful. It gives some more background and compares several app

The linked overview on Efficiently Generating a Number in a Range by O'Neill (of PCG fame) is insightful. It gives some more background and compares several approaches (including this, Lemire's). It seems based on her testing that a very simple bit mask & rejection technique ("Apple's method", as Apple implemented it in their modification of arc4random_uniform) is the fastest among the unbiased techniques, particularly for large N, while for mixed N Lemire's technique seems competitive or better. She suggests a further optimisation, which now clearly "wins" except for large N. Cool stuff - I find random number generation fascinating. http://www.pcg-random.org/posts/bounded-rands.html http://www.pcg-random.org/posts/bounded-rands.html