Function call inlining

Function call inlining

[Matwey V. Kornilov]

Hello,

Please consider the following code from the SO:
http://stackoverflow.com/a/32237033/1879547

template<class Ret, class T, class Func, int... Is>
auto apply(T&& p, int index, Func func, seq<Is...>) -> Ret
{
     using FT = Ret(T&&, Func);
     static constexpr FT* arr[] = { &apply_one<Ret, Is, T&&, Func>... };
     return arr[index](std::forward<T>(p), func);
}

I use gcc-4.8.2 and see that arr[index](std::forward<T>(p), func) call
is not inlined.
The question here is the following. All function pointers are known at
compile time. So, why don't expand this code into analogous to the
following?

switch(index) {
	case 0:
		// inline arr[0]
	break;
	case 1:
		// inline arr[1]
	// etc...
}

Re: Function call inlining

[Andrew Haley]

On 08/27/2015 04:34 PM, Matwey V. Kornilov wrote:
> I use gcc-4.8.2 and see that arr[index](std::forward<T>(p), func) call
> is not inlined.
> The question here is the following. All function pointers are known at
> compile time. So, why don't expand this code into analogous to the
> following?

I suspect that no-one has explained because no-one understands why
anyone would want to make such an "optimization".

Andrew.

Re: Function call inlining

[Matwey V. Kornilov]

2015-08-28 10:59 GMT+03:00 Andrew Haley <aph@redhat.com>:
> On 08/27/2015 04:34 PM, Matwey V. Kornilov wrote:
>> I use gcc-4.8.2 and see that arr[index](std::forward<T>(p), func) call
>> is not inlined.
>> The question here is the following. All function pointers are known at
>> compile time. So, why don't expand this code into analogous to the
>> following?
>
> I suspect that no-one has explained because no-one understands why
> anyone would want to make such an "optimization".
>

The question arised from here:

http://stackoverflow.com/questions/32235855/

Moreover, clang does this kind of optimization:

http://goo.gl/jnY5yx

-- 
With best regards,
Matwey V. Kornilov
http://blog.matwey.name
xmpp://0x2207@jabber.ru