The best way to do it seems to be to write conversion function.
Receive the float in a u32, parse it's bits and return a s32. Can be
done without the use of floats.