higher bound for a
float point error function is
defined using ___