Why are floating point types invalid template parameter types for template functions?

Question

I read on a website that the declaration

template <int x>
int func() {
  return x;
} 

is valid while the following is not

template <double x>
double func() {
  return x;
}

Why is the first a legal declaration for a template function while the second is not?

Answer 1

It is not valid because it is not an integral type. There are certain restrictions on nontype template parameters and this is one of them, which says ...

Floating-point numbers and class-type objects are not allowed as nontype template parameters.

template <double VAT>       // ERROR: floating-point values are not
double process(double v) { // allowed as template parameters
    return v * VAT;
}

template <std::string name> // ERROR: class-type objects are not
class MyClass {             // allowed as template parameters
  /* ... */
};

The above is quoted from C++ Templates. I take no credits for it.

The reason why they are not valid for template initialization, as per my understanding, is because types like float and double don't have a defined implementation in C++. So when a template like

template <double VAT> double process(double v);

is initialized with two different double values as

template <(double) 2/3> double process(2.3)

template <(double) 1/3> double process(2.4);

they might not have same bit representation because of double nontype, which confuses the compiler.