Comment by xigoi

You don’t divide a float by 256 by shifting it right eight bits; that would yield complete garbage. You subtract 8 from the exponent, then check if you got an underflow.