But it has been explained why using a floating-point value is not a good idea:

"not only is the use of floating-point unnecessary - it could actually give you the wrong result!"
http://www.8052.com/forum/read.phtml?id=96034

The reason it's not a good idea due to the fundamental nature of the implementation of floating-point arithmetic - thus it is not compiler-specific, nor even processor-specific.

All you need to do is to change your XTAL definition from a floating-point number of MHz to an integer number of Hz (or even kHz), and scale accordingly.