How dare they say ++it is faster than it++

This is NOT true for the iterators of 100% of the STL containers - including the commonly used ones - std::vector, std::unordered_map, std::array, std::list, std::unordered_set, std::map, std::set, std::queue and ALL others.

This is obviously NOT true for primitive types. No explanation required.

This is NOT true for the iterators of 99.99999999% of the non-STL templated containers (custom implemented) ever written in the entire C++ world except those which were written for extremely specialized use case in which they couldn’t ensure TriviallyCopyable requirnment for iterator… but hey, for those super specialized 0.00000001% use case, they are not likely to support it++ anyway, so nothing to worry.

Why this discussion on the first place

Because of misleading information:

Misleading Posts

Why `++it` performs identically to `it++`

because cost of copying a TriviallyCopyable object is ZERO if it’s unused later.

It’s NOT almost-zero. It’s absolute ZERO.

Note:

1). Compilers can figure out if it’s unused or not. It works in O1, O2 and O3 mode.

2). operator++ for templated iterator is inlined.

In this code:

struct R {
  ...;
};

struct S {
  R* x;
  int y;
};

S Func1(S& s) {
  S s1_copy = s;
  s.x++;
  s.y++;
  return s1_copy;
}

S& Func2(S& s) {
  s.x++;
  s.y++;
  return s;
}

Func1(s);
Func2(s);

Func2(s) is NOT faster than Func1(s).

Let’s look at the machine code:

Let’s take an example of std::list:

int F3(const std::list<int>& v) {
    int output = 0;
    for (auto it = v.begin(); it != v.end(); it++) {
        output += *it;
    }
    return output;
}

int G3(const std::list<int>& v) {
    int output = 0;
    for (auto it = v.begin(); it != v.end(); ++it) {
        output += *it;
    }
    return output;
}