Returns
a vector containing the largest representable floating-point integral
values less than or equal to the values of the corresponding elements
of the given vector.
Note:vec_floor provides
the same functionality as vec_roundm, except that vec_floor could
trigger the IEEE-inexact exception.