Is it better in C++ to pass by value or pass by constant reference?

It used to be generally recommended best practice¹ to use pass by const ref for all types, except for builtin types (char, int, double, etc.), for iterators and for function objects (lambdas, classes deriving from std::*_function).

This was especially true before the existence of move semantics. The reason is simple: if you passed by value, a copy of the object had to be made and, except for very small objects, this is always more expensive than passing a reference.

With C++11, we have gained move semantics. In a nutshell, move semantics permit that, in some cases, an object can be passed “by value” without copying it. In particular, this is the case when the object that you are passing is an rvalue.

In itself, moving an object is still at least as expensive as passing by reference. However, in many cases a function will internally copy an object anyway — i.e. it will take ownership of the argument.²

In these situations we have the following (simplified) trade-off:

1. We can pass the object by reference, then copy internally.
2. We can pass the object by value.

“Pass by value” still causes the object to be copied, unless the object is an rvalue. In the case of an rvalue, the object can be moved instead, so that the second case is suddenly no longer “copy, then move” but “move, then (potentially) move again”.

For large objects that implement proper move constructors (such as vectors, strings …), the second case is then vastly more efficient than the first. Therefore, it is recommended to use pass by value if the function takes ownership of the argument, and if the object type supports efficient moving.

---

A historical note:

In fact, any modern compiler should be able to figure out when passing by value is expensive, and implicitly convert the call to use a const ref if possible.

In theory. In practice, compilers can’t always change this without breaking the function’s binary interface. In some special cases (when the function is inlined) the copy will actually be elided if the compiler can figure out that the original object won’t be changed through the actions in the function.

But in general the compiler can’t determine this, and the advent of move semantics in C++ has made this optimisation much less relevant.

---

¹ E.g. in Scott Meyers, Effective C++.

² This is especially often true for object constructors, which may take arguments and store them internally to be part of the constructed object’s state.

Edit: New article by Dave Abrahams on cpp-next:

Want speed? Pass by value.

---

Pass by value for structs where the copying is cheap has the additional advantage that the compiler may assume that the objects don't alias (are not the same objects). Using pass-by-reference the compiler cannot assume that always. Simple example:

foo * f;

void bar(foo g) {
    g.i = 10;
    f->i = 2;
    g.i += 5;
}

the compiler can optimize it into

g.i = 15;
f->i = 2;

since it knows that f and g doesn't share the same location. if g was a reference (foo &), the compiler couldn't have assumed that. since g.i could then be aliased by f->i and have to have a value of 7. so the compiler would have to re-fetch the new value of g.i from memory.

For more pratical rules, here is a good set of rules found in Move Constructors article (highly recommended reading).

• If the function intends to change the argument as a side effect, take it by non-const reference.
• If the function doesn't modify its argument and the argument is of primitive type, take it by value.
• Otherwise take it by const reference, except in the following cases
  • If the function would then need to make a copy of the const reference anyway, take it by value.

"Primitive" above means basically small data types that are a few bytes long and aren't polymorphic (iterators, function objects, etc...) or expensive to copy. In that paper, there is one other rule. The idea is that sometimes one wants to make a copy (in case the argument can't be modified), and sometimes one doesn't want (in case one wants to use the argument itself in the function if the argument was a temporary anyway, for example). The paper explains in detail how that can be done. In C++1x that technique can be used natively with language support. Until then, i would go with the above rules.

Examples: To make a string uppercase and return the uppercase version, one should always pass by value: One has to take a copy of it anyway (one couldn't change the const reference directly) - so better make it as transparent as possible to the caller and make that copy early so that the caller can optimize as much as possible - as detailed in that paper:

my::string uppercase(my::string s) { /* change s and return it */ }

However, if you don't need to change the parameter anyway, take it by reference to const:

bool all_uppercase(my::string const& s) { 
    /* check to see whether any character is uppercase */
}

However, if you the purpose of the parameter is to write something into the argument, then pass it by non-const reference

bool try_parse(T text, my::string &out) {
    /* try to parse, write result into out */
}

Depends on the type. You are adding the small overhead of having to make a reference and dereference. For types with a size equal or smaller than pointers that are using the default copy ctor, it would probably be faster to pass by value.