DESCRIPTION

To satisfy portable code, srand48(), seed48(), or lcong48() should be called to initialize the subsystem. In OpenBSD the seeding parameters are ignored, and strong random number results will be provided from arc4random(3). In other systems, the parameters prime a simplistic deterministic algorithm.

If the standardized behavior is required then srand48_deterministic(), seed48_deterministic(), and lcong48_deterministic() can be substituted for srand48(), seed48(), and lcong48(). That will cause subsequent calls to drand48(), lrand48(), and jrand48() to return results using the deterministic algorithm.

drand48() and erand48() return values of type double. The full 48 bits of r(n+1) are loaded into the mantissa of the returned value, with the exponent set such that the values produced lie in the interval [0.0, 1.0].

lrand48() and nrand48() return values of type long in the range [0, 2**31-1]. The high-order (31) bits of r(n+1) are loaded into the lower bits of the returned value, with the topmost (sign) bit set to zero.

mrand48() and jrand48() return values of type long in the range [-2**31, 2**31-1]. The high-order (32) bits of r(n+1) are loaded into the returned value.

In the deterministic mode, the rand48() family of functions generates numbers using a linear congruential algorithm working on integers 48 bits in size. The particular formula employed is r(n+1) = (a * r(n) + c) mod m where the default values are for the multiplicand a = 0xfdeece66d = 25214903917 and the addend c = 0xb = 11. The modulus is always fixed at m = 2 ** 48. r(n) is called the seed of the random number generator.

For all the six generator routines described next, the first computational step is to perform a single iteration of the algorithm.

drand48(), lrand48(), and mrand48() use an internal buffer to store r(n). For these functions the initial value of r(0) = 0x1234abcd330e = 20017429951246.

On the other hand, erand48(), nrand48(), and jrand48() use a user-supplied buffer to store the seed r(n), which consists of an array of 3 shorts, where the zeroth member holds the least significant bits.

All functions share the same multiplicand and addend.

srand48_deterministic() is used to initialize the internal buffer r(n) of drand48(), lrand48(), and mrand48() such that the 32 bits of the seed value are copied into the upper 32 bits of r(n), with the lower 16 bits of r(n) arbitrarily being set to 0x330e. Additionally, the constant multiplicand and addend of the algorithm are reset to the default values given above.

seed48_deterministic() also initializes the internal buffer r(n) of drand48(), lrand48(), and mrand48(), but here all 48 bits of the seed can be specified in an array of 3 shorts, where the zeroth member specifies the lowest bits. Again, the constant multiplicand and addend of the algorithm are reset to the default values given above. seed48_deterministic() returns a pointer to an array of 3 shorts which contains the old seed. This array is statically allocated, so its contents are lost after each new call to seed48_deterministic().

Finally, lcong48_deterministic() allows full control over the multiplicand and addend used in drand48(), erand48(), lrand48(), nrand48(), mrand48(), and jrand48(), and the seed used in drand48(), lrand48(), and mrand48(). An array of 7 shorts is passed as parameter; the first three shorts are used to initialize the seed; the second three are used to initialize the multiplicand; and the last short is used to initialize the addend. It is thus not possible to use values greater than 0xffff as the addend.

Note that all three methods of seeding the random number generator always also set the multiplicand and addend for any of the six generator calls.