In floating-point arithmetic, what does underflow refer to?

• A. Dropped to zero.
• B. Rounding occurs to the nearest representable value.
• C. Represented as a subnormal value.
• D. Causes an exception.

Answer: (C)

Underflow happens when a result is too small in magnitude to be stored as a normal floating-point number. In that case, the value is encoded as a subnormal (denormal) value, which uses a diminished exponent and does not include the implicit leading 1. This lets numbers closer to zero be represented, providing gradual underflow instead of snapping to zero. If subnormals aren’t used on a system, the result may simply become zero. Rounding to the nearest representable value is a separate process that applies to all finite numbers, not specifically what underflow is. An exception can be raised in some environments, but underflow itself is defined by the use of a subnormal representation.