Underflow does occur, about talking, whenever the response to an arithmetic procedure is indeed tiny so it can’t be trapped in their desired spot formatting without enduring a rounding problem this is certainly larger than normal.
TABLE 2-11 displays the underflow thresholds for individual, double, and double-extended accuracy.
The positive subnormal rates are the ones amounts amongst the least standard amount and zero. Subtracting two (beneficial) tiny amounts being close to the least typical numbers might make a subnormal quantity. Or, dividing the littlest favorable typical amounts by two generates a subnormal outcome.
The current presence of subnormal amounts supplies greater detail to floating-point data which entail tiny numbers, although the subnormal quantities on their own have fewer pieces of detail than normal quantities.