The probable cause of the error is the use of a floating point variable for computing of value a. Any floating point types have limited precision, and looks like some times it is not enough. Consider to use some integer types instead.