std::apply 源码分析
背景：
std::apply 适合感觉跟std::invoke 有点相似， 前边参数是可以调用的对象， 后边是一个tuple类型，在做线程的任务包装的时候，可以使用这个apply，
接下来我们就看下它的源码
std::apply 用例
| C++
#include
#include
#include <utility>
int add(int first, int second) { return first + second; }
template
T add_generic(T first, T second) { return first + second; }
auto add_lambda = [](auto first, auto second) { return first + second; };
template
std::ostream& operator<<(std::ostream& os, std::tuple const& theTuple)
{
std::apply
(
[&os](Ts const&... tupleArgs)
{
os << '[';
std::size_t n{0};
((os << tupleArgs << (++n != sizeof...(Ts) ? ", " : "")), ...);
os << ']';
}, theTuple
);
return os;
}
int main()
{
// OK
std::cout << std::apply(add, std::pair(1, 2)) << '\n';
// Error: can't deduce the function type
// std::cout << std::apply(add_generic, std::make_pair(2.0f, 3.0f)) << '\n';
// OK
std::cout << std::apply(add_lambda, std::pair(2.0f, 3.0f)) << '\n';
// advanced example
std::tuple myTuple{25, "Hello", 9.31f, 'c'};
std::cout << myTuple << '\n';
} |
https://en.cppreference.com/w/cpp/utility/apply
Output
| C++
3
5
[25, Hello, 9.31, c] |
std::apply 源码
| C++
template
inline _LIBCPP_INLINE_VISIBILITY
constexpr decltype(auto) apply(_Fn && __f, _Tuple && __t)
_LIBCPP_NOEXCEPT_RETURN(
_VSTD::__apply_tuple_impl(
_VSTD::forward<_Fn>(__f), _VSTD::forward<_Tuple>(__t),
typename __make_tuple_indices>>::type{})
) |
这个跟std::make_from_tuple基本上一样，请移步到https://blog.csdn.net/c553110519/article/details/126838742
汇编
| C++
int TestApply(int a, int b)
{
return a + b;
}
void Test()
{
int sum = std::apply(TestApply, std::tuple(1,2));
std::cout< } |
| Assembly language
0x1000036f0 <+0>: sub sp, sp, #0x30
0x1000036f4 <+4>: stp x20, x19, [sp, #0x10]
0x1000036f8 <+8>: stp x29, x30, [sp, #0x20]
0x1000036fc <+12>: add x29, sp, #0x20
0x100003700 <+16>: nop
0x100003704 <+20>: ldr x0, #0x93c ; (void *)0x00000001fd11dd98: std::__1::cout
0x100003708 <+24>: mov w1, #0x3
0x10000370c <+28>: bl 0x100003c94 ; symbol stub for: std::__1::basic_ostream >::operator<<(int)
|
可以看的出来最终转成sum = std::apply(TestApply, std::tuple(1,2));汇编后mov w1, #0x3，编译器把中间过程给做了，丝毫没有影响到性能，constexpr的特性所致