The erand48 function generates pseudorandom numbers using the
    linear congruential algorithm and 48-bit integer arithmetic.

    It returns nonnegative, double-precision, floating-point values
    uniformly distributed over the range of y values, such that 0.0
    <= y < 1.0.

    The erand48 function works by generating a sequence of 48-bit
    integer values, Xi, according to the linear congruential formula:

           Xn+1 = (aXn+c)mod m        n >= 0

    The argument m equals 248, so 48-bit integer arithmetic is

    performed. Unless you invoke the lcong48 function, the multiplier
    value a and the addend value c are:

          a = 5DEECE66D16 = 2736731631558

          c = B16 = 138

    The erand48 function requires that the calling program pass an
    array as the xsubi argument. For the first call, the array must
    be initialized to the value of the pseudorandom-number sequence.
    Unlike the drand48 function, it is not necessary to call an
    initialization function prior to the first call.

    By using different arguments, the erand48 function allows
    separate modules of a large program to generate several
    independent sequences of pseudorandom numbers; for example, the
    sequence of numbers that one module generates does not depend
    upon how many times the function is called by other modules.