Alternatively, look at the PROTON+ compiler.

This does true floating point calculations including trigonometry.