Quantum computing with real numbers

I would appreciate hearing from anybody who has any information on how quantum computers might handle real numbers. Is classic floating point (k-bit binary exponent and n-bit binary mantissa) really appropriate and how good might its performance be?

As an example, if a quantum computer was instructed to divide 1.0 by 3.0, how quickly would it get a result and to what precision? Or is it a matter of waiting until you get a precision that you find satisfactory? Or, is the operation itself virtually instantaneous as long as you keep it in a quantum state and use it in further quantum calculations, and only when you wish to extract the value into a non-quantum format do you need to do all the traditional heavy-lifting?