static uint32_t _rand32();
static uint64_t _rand64();

// Borrowed from https://www.pcg-random.org/posts/bounded-rands.html
static uint32_t _bounded_rand(uint32_t range) {
	uint32_t x = _rand32();
	uint64_t m = (uint64_t)x * (uint64_t)range;
	uint32_t l = (uint32_t)m;

	if (l < range) {
		uint32_t t = -range;
		if (t >= range) {
			t -= range;
			if (t >= range)
				t %= range;
		}
		while (l < t) {
			x = _rand32();
			m = (uint64_t)x * (uint64_t)range;
			l = (uint32_t)m;
		}
	}

	return m >> 32;
}

// https://prng.di.unimi.it/#remarks
static double _uint64_to_double(uint64_t num, bool inclusive) {
	// A standard 64bit double floating-point number in IEEE floating point
	// format has 52 bits of significand. Thus, the representation can actually
	// store numbers with 53 significant binary digits.

	double scale;
	if (inclusive) {
		scale = 1.0 / ((1ULL << 53) - 1); // [0, 1]
    } else {
        scale = 1.0 / (1ULL << 53);       // [0, 1)
    }

	double ret = (num >> 11) * scale; // Top 53 bits divided by 1/2^53

	//printf("Double: %0.15f\n", ret);

	return ret;
}

static float _uint32_to_float(uint32_t num, bool inclusive) {
    float scale;
    if (inclusive) {
        scale = 1.0f / ((1U << 24) - 1);  // [0, 1]
    } else {
        scale = 1.0f / (1U << 24);        // [0, 1)
    }

    float ret = (num >> 8) * scale;

	return ret;
}

// Why This Works
//   (x + 0.5) offsets each uint32_t value into the center of its floating-point "bin," reducing bias.
//   Multiplying by 1.0 / 4294967296.0 scales it into the range [0,1).
static double _uint32_to_double_old(uint32_t x) {
    return (x + 0.5) * (1.0 / 4294967296.0);  // 1/2^32
}

// MurmurHash3 Finalizer (Passes SmallCrush)
static uint64_t _hash_mur3(uint64_t x) {
    x ^= x >> 33;
    x *= 0xff51afd7ed558ccd;
    x ^= x >> 33;
    x *= 0xc4ceb9fe1a85ec53;
    x ^= x >> 33;
    return x;
}