A few days ago Calvin_Hsia posted Floating Point calculations: comparing with zero in his blog at http://blogs.msdn.com/calvin_hsia/archive/2005/4/13.aspx


>Please go back and reread any of the prior threads we've had discussing floating point calculations on binary computers. And then in the future remember to use an epsilon difference calculation to determine if two floating point numbers are close enough to each other to be considered equivalent.