atomic_fetch_inc_modulo_wrap

/* Fetch and increment with modulo wrap
 * Although there's theoretically a race condition,
 * it's practically impossible to increment 32^2 times
 * while the number is wrapped */
static inline uint64_t atomic_fetch_inc_modulo_wrap(atomic_uint_fast64_t *val,
                                                    uint32_t wrap) {
  uint64_t fetched = (*val)++;
  uint32_t hi = (fetched & 0xFFFFFFFF00000000) >> 32;
  if (hi == wrap) {
    atomic_fetch_and(val, 0xFFFFFFFF);
  }
  return fetched % wrap;
}