This is a small header-only library implementing a handful of routines for uniform sampling in floating-point. See the blog posts Sampling in Floating Point (1/3): The Unit Interval and Sampling in Floating Point (2/3): 1D Intervals for further context and documentation.
The main entrypoints are:
Sample01(): returns uniform floating-point values in [0,1).SampleToPowerOfTwo(x): uniformly samples in [0,2^x).SampleRange(a, b): uniformly samples in [a,b).
It is expected that the user will provide implementations of two functions to generate random values:
uint64_t Random64Bits();
uint32_t Random32Bits();For previous implementations of these ideas, see:
- Christoph Conrads's Rademacher Floating Point Library, in particular the make_uniform_random_value() function.
- Olaf Bernstein's dist_uniformf_dense().
- Marc Reynolds's Higher density uniform floats blog post.