s-l-o-w-l-y is the correct answer to this question. One can do floating-point calculations with a fixed-point CPU using special techniques.

I believe that one can emulate a full 68040 on the 68000 Macs by using available software: run 020 Simulator, then run Pseud040 and SoftFPU on top of that.

I further believe that to do so would be about as useful as Dana's stunt of booting MacOS X on the Q605.