Even if enum where to consume less memory than constants, they still limit the application of such a technique to normal integer values, hence the need for static constants.

He didn’t say that your code didn’t compute the answer at compile-time (which it does). He said that your solution doesn’t have a zero run-time memory requirement - what he described as “value-free”. Because you’re using static variables inside each template instantiation, you will have at least “n” static variables (admittedly, they are const, but they still consume memory space in the program’s run time data segment), where each variable is initialized by the compiler’s compile time meta-template recursion to the value of Fib(n).

If you use enums rather than static const variables inside your templates, then you wouldn’t use any memory at all.

On topic: Personally I find that template metaprogramming can be ‘handy’. But to date I’ve rarely (if ever) found a case where I’d consider using it in production code.

Great… you only give negative votes. I’m impressed

“Which is not true in your case.”

So please show me WHY it is wrong, as in this case everything is calculated at compile-time. Compile the source code and see for yourself… or explain me why the compiler does not perform the computations in this case.

“Try replacing static member variable “value” with “enum{value=…}” for a start.”

Yes, I know about enums, but the fact is that enum == int and I wanted to use long long integers rather than int. Having enum {value = …} or a static constant integer is actually the same…

Enums are one way to do it, but many times you need to manipulate other data types.

“Cleaning and simplifying the code would be next.”

???

LOL

Did you give a look at this wonderful website : http://ktd.club.fr/programmation/cpp-metaprog.php

(Ouille mes chevilles)

+
Chacha