The problem with rand()

The old school C way of generating random numbers is to call rand(). The rand() function, which is still part of C++, takes no arguments and produces a single, uniformly distributed integer in the [0, RAND_MAX] range. The internal state can be seeded by calling the library function, srand(seed_value).

The classic code to generate a random number in the [0, x) range hasn't changed since the 1980s, shown here:

    #include <stdlib.h>

    int randint0(int x) {
      return rand() % x;
    }

However, this code has several problems. The first and most obvious problem is that it doesn't generate all x outputs with equal likelihood. Suppose, for the sake of argument, rand() returns a uniformly distributed value in the [0, 32767] range, then randint0(10) will return each value in the [0, 7] range one-3276th more often than it returns either 8 or 9.

The second problem is that rand() accesses global state; the same random number generator is shared by every thread in your...