What are some recommended numerical analysis books on floating point numbers? I'd like the book to have the following
In depth coverage on the representation of floating point numbers on modern hardware (the IEEE standard).
How to do arbitrary precision floating point calculations with a reasonably fast modern algorithm.
How to compute the closest 32-bit floating point representation of a dot product and cross product. And do this fast, so no relying on generic arbitrary precision calculations to get the bits of the 32-bit floating point number right.
From what I can infer from doing some searches most books tend to focus on stuff like the runge kutta and not put much emphasis on how to make floating point calculations that are ultra precise.