higher bound for a

float point error function is

in efficiency✕