Floating point maths is only very rarely required, it's slow, takes lots of space, and is inaccurate - far better to use integers instead, which is often simple with a little thought. This also applies to PC's, never mind micro-controllers!.
If there is ever a need to use them?, it's reasonably simple in assembler - you just download some floating point routines. You simply load your values into variables, call the routine you want, and read the result back. This is just how a high level language would do the same job - it just hides it from you.
And all this fuss Marks256 has been making about Z80, now he's decided he only wants to write in C! - so what's the point of building a Z80 system?, C hides that fact from you - may as well just use a single chip PIC or AVR which will out perform a Z80 at a fraction of the cost.
BTW Marks256, - if assembler scares you, go for PIC rather than AVR (or Z80), it's generally considered the simplest to learn (hence the PIC's popularity).