Yes you're right - thanks for clarifying. I meant to say that not every integer in magnitude greater than 52 bit has an exact floating point representation in IEEE doubles.