odometer:
In other words, the "double" keyword is a lie, as it is not really double-precision floating point.
It's not a lie. Nothing is being concealed. It is explained in the reference material.
You shouldn't be using either double OR float unless you REALLY have to.
Without a floating point unit, all floating point maths will be very slow.
If at all possible use integer maths (multiply up to use bigger values) or fixed point arithmetic (like Q15 etc).