Obviously, if you have the option, you can simply increase the precision of the floating point variable, from 32-bit to 64- or even 256-bit, but this only gets you so far.