This directive tells the compiler to use maximum precision when evaluating floating point expression, i.e. use the maximum possible precision, even if the members in the calculation have less than maximum precision. It accepts a boolean value (ON/OFF or +/-).