> disables IEEE gradual underflow, producing zero instead of a
> denormalised number.
>
> Can someone explain what is a "denormalised number"? I've never seen
> that terminology before.
Many floating-point patterns can represent the same number. A
normalised number is one with the most significant bits in the mantissa.
What happens when a number becomes very small? When the exponent is as
small as possible, one could make it even smaller---but less
accurate---by adding "leading zeros" to the mantissa. This is a
denormalised number.
|