c++11 - Performance difference in pointer checking - smart pointers [closed]

Question

Is there any difference between testing a smart pointer, e.g. shared_ptr, using operator bool

if (!smart_ptr)
{
    // ...
}

and using operator == ?

if (smart_ptr == nullptr)
{
    // ...
}

I know the difference will be small (if any), but it could also help deciding the coding style at the same time.

Answer 1

In gccv4.8:

#include <memory>

bool is_empty(const std::unique_ptr<int>& p) {
  return p == nullptr;
}

produces assembly code:

is_empty(std::unique_ptr<int, std::default_delete<int> > const&):
    cmpq    $0, (%rdi)
    sete    %al
    ret

---

#include <memory>

bool is_empty2(const std::unique_ptr<int>& p) {
  return !p;
}

produces assembly code:

is_empty2(std::unique_ptr<int, std::default_delete<int> > const&):
    cmpq    $0, (%rdi)
    sete    %al
    ret

Consequently, it makes no difference in a decent modern compiler.

Live Demo by Jonathan Wakely