I thought that too, but it's actually the difference between 1.0f and the next larger number; so it only represents the smallest difference for the non-dotzero

I thought that too, but it's actually the difference between 1.0f and the next larger number; so it only represents the smallest difference for the non-dotzero range. See here: http://www.cplusplus.com/reference/cfloat/ http://www.cplusplus.com/reference/cfloat/