fast-math-0.1: Non IEEE-754 compliant compile-time floating-point optimisations