higher bound for a

float point error function less

in efficiency✕