Floating point precisions and std::numeric_limits digits, exponents, min/lowest members.

[abhi-k9]

Channel

"C++Weekly" request.

Topics

1. Floating point precision
2. std::numeric_limits<T>::digits, std::numeric_limits<T>::digits10, std::numeric_limits<T>::max_digits10 differences.
3. std::numeric_limits<T>::min, std::numeric_limits<T>::lowest differences.
4. std::numeric_limits<T>::min_exponent, std::numeric_limits<T>::min_exponent10, std::numeric_limits<T>::max_exponent, std::numeric_limits<T>::max_exponent10 differences if time permits.

Length

Long form if need be, but could possibly be a short one.

[Dharmesh946]

NB floating point accuracy is not limited to these types but also to the kind of operations performed, floating point promotion and the underlying HW implementation of the types (typically, it can have 80 bits for intermediate computation, which can lead to unexpected behaviors).