Also I'm not sure how the 16 bit operations are handled on the 8 bit AVR. So there is a good change for whatever reason to have a division by zero.
it is not 16 bit problem, it is a float (32bit IEEE754) problem.
The only thing I can remember right now was that direct multiplication between calRAdevice and byteCode will always result in an overflow.
what are typical values for bytecode?