There are functions like _isnan, _isnanf, _finite, and _fpclass for detecting that a floating point value is one of the special values like NaN, but how do you actually generate one of these values?

You can access these values from the std::numeric_limits template.

std::numeric_limits<float>::infinity(); // positive infinity
std::numeric_limits<float>::quiet_NaN(); // non-signalling NaN

Wait, where’s negative infinity? The compiler folks provided these handy little definitions for when you need to generate a special value (as opposed to merely detecting one), and for which the numeric_limits template comes up short.

DECLSPEC_SELECTANY extern const float FLOAT_POSITIVE_INFINITY = ((float)(1e308 * 10));
DECLSPEC_SELECTANY extern const float FLOAT_NEGATIVE_INFINITY = ((float)(-1e308 * 10));
DECLSPEC_SELECTANY extern const float FLOAT_NaN = ((float)((1e308 * 10)*0.));

Disclaimer: Applies to Microsoft Visual Studio. Your mileage may vary. Use the template when available.

Bonus chatter: Note that you must use functions like _isnan to detect special values, because floating point special values behave very strangely in comparisons. (For example, NaN does not compare equal to itself!)