Tony Firshman wrote:
>>> The history is documented here
>>> http://www.cs.berkeley.edu/~wkahan/ieee754status/754story.html
> Can someone please define "Gradual Underflow".

I might be mistaken (as usual ;), but as I understand it that's the
denormalised numbers we have discussed here, i.e. when the exponent
underflows the value is not set to 0 but it is tried to represent the
number using denormalisation as close as possible. Only when the
mantissa runs out of bits to hold at least the highest one bit (all
zero), then it's a "complete" underflow.

Marcel

Reply via email to